UNITED STATES OF AMERICA
+ + + + +
FOOD AND DRUG ADMINISTRATION
+ + + + +
CENTER FOR DEVICES AND RADIOLOGICAL HEALTH
+ + + + +
RADIOLOGICAL DEVICES PANEL
+ + + + +
+ + + + +
TUESDAY, DECEMBER 10, 2001
The Panel met at 8:30 a.m. in the Walker/Whetstone Rooms of the Gaithersburg Holiday Inn, Two Montgomery Village Avenue, Gaithersburg, Maryland, Dr. Minesh P. Mehta, Chairman, presiding.
MINESH P. MEHTA, M.D. Chairman
HARRY K. GENANT, M.D. Member
GEOFFREY S. IBBOTT, Ph.D. Member
ALICIA Y. TOLEDANO, Sc.D. Member
PRABHAKAR TRIPURANENI, M.D. Member
EMILY F. CONANT, M.D. Temporary Voting Member
REGINA J. HOOLEY, M.D. Temporary Voting Member
MARILYN R. PETERS, M.N., M.P.H. Non-Voting Consumer Representative
ERNEST L. STERN Non-Voting Industry Representative
ROBERT J. DOYLE Executive Secretary
PRESENT FROM FDA:
HARRY F. BUSHAR, Ph.D.
JOHN C. MONAHAN
ROBERT A. PHILLIPS, Ph.D.
WILLIAM SACKS, Ph.D., M.D.
STANLEY STERN, Ph.D.
KARLEEN CALLAHAN, M.D.
KEVIN HUGHES, M.D.
YURI PARISKY, M.D.
STEVE RUST, Ph.D.
C O N T E N T S
Call to Order and the Chair's Introduction 5
Minesh P. Mehta, M.D.
Update on FDA Radiology Activities 8
Robert A. Phillips, Ph.D.
FDA Introductory Remarks 10
Robert J. Doyle, Executive Secretary
Development of Amendments to the U.S. 16
Radiation Safety Standard for Diagnostic
X‑Ray Computed Tomography
Stanley Stern, Ph.D.
Open Public Hearing - (No Respondents) 37
Open Committee Discussion 39
Charge to the Panel 39
Minesh P. Mehta, M.D.
Presentations on P010035 by Sponsor 39
Introduction - John Brenna 39
Clinical Use - Yuri Parisky, M.D. 43
Clinical Use - Kevin Hughes, M.D. 47
Engineering and Technology - 52
Clinical Trial and Results - 63
Karleen Callahan, M.D.
Statistics - Steve Rust, Ph.D. 74
Efficacy Results - Yuri Parisky, M.D. 77
C O N T E N T S (CONTINUED)
Presentations on P010035 by Sponsor
Performance Results and Their Effect 84
on the Health Care System - Steve Rust, Ph.D.
Summary of Presentation - 87
Questions and Answers 90
Presentations on P010035 by FDA
PMA Overview - John Monahan 123
Clinical Review - William Sacks, Ph.D., M.D. 126
Statistical Analysis - Harry Bushar, Ph.D. 137
Clinical Review Continued - 152
William Sacks, Ph.D., M.D.
Recognition of Outgoing Panel Members - 165
Nancy Brogdon, Director, Division of
Reproductive, Abdominal, and Radiological
Devices, Office of Device Evaluation
Panel Discussion of PMA 167
FDA Questions - Robert A. Phillips, Ph.D. 230
Open Public Hearing 264
Panel Recommendations and Vote 268
CHAIRMAN MEHTA: I think if everybody will take a seat, we would like to go ahead and get started on time. Several people have come from long distances, and we would like to stay on time today, if possible.
So I would like to call this meeting of the Radiological Devices Panel to order. I also want to request everyone in attendance at the meeting to sign in on the attendance sheet that's available at the door.
For the record, I note that the voting members present constitute a quorum, as required by 21 CFR Part 14, and at this time I would like to have each of the Panel members at the table to introduce themselves, state their specialty, position title, institution, and status on the Panel.
I'll begin with myself. My name is Minesh Mehta. I'm a radiation oncologist at the University of Wisconsin, and I'm currently the Chair of the Radiological Devices Panel.
We'll move to the right and we'll have individuals introduce themselves, and we'll go around the table and come back, so that the introductions are complete.
DR. CONANT: Good morning. I'm Emily Conant, and I'm Chief of Breast Imaging at the University of Pennsylvania, Associate Professor, and I'm here as, I think, a clinical breast imager.
DR. HOOLEY: Hello. My name is Regina Hooley. I'm an Assistant Professor of Diagnostic Radiology at Yale University with a specialty in mammography. I'm here as a consultant as a breast imager.
DR. IBBOTT: I'm Geoff Ibbott. I'm a medical physicist, and I'm an Associate Professor at MD Anderson Cancer Center in the Department of Radiation Physics there. I'm also Director of the Radiological Physics Center at MD Anderson, and I'm a member of the Panel.
MR. STERN: My name is Ernie Stern. I'm the industry representative. I'm the Chairman of Thales Components Corporation, the U.S. subsidiary of Thales in France. We manufacture defense electronics and medical electronics products.
DR. TRIPURANENI: I'm Prabhakar Tripuraneni, radiation oncologist by training and trade. I'm the head of the Radiation Oncology at Scripps Clinic in La Jolla, California, and I'm a Panel member.
MS. BROGDON: Good morning. I'm not a member of the Panel. I'm Nancy Brogdon. I'm the Director of FDA's Division of Reproductive, Abdominal, and Radiological Devices.
MS. PETERS: I'm Marilyn Peters. I'm a consumer representative, and I'm a health education consultant in Los Angeles.
DR. GENANT: Good morning. I'm Harry Genant. I'm a radiologist. I'm Professor of Radiology Medicine and Orthopedic Surgery at the University of California, San Francisco, and the Executive Director of the Osteoporosis and Arthritis Research Group at UCSF. I'm a Panel member.
DR. TOLEDANO: Good morning. My name is Alicia Toledano. I'm Assistant Professor at the Center for Statistical Sciences at Brown University. My specialty is with diagnostics, diagnostic radiology, and I am a Panel member.
MR. DOYLE: And my name is Bob Doyle. I'm the Executive Secretary of this Panel.
CHAIRMAN MEHTA: Thank you, everybody.
At this time Dr. Robert Phillips, the Chief of the Radiology Branch of the Office of Device Evaluation, would like to give us a brief update on the FDA radiological activities over the past several months.
DR. PHILLIPS: Good morning. It's a little chilly out, I think, isn't it?
I want to give you an update of what has happened as far as our major approvals since the last time we met. If you recall, the last meeting of the Panel was in March 2001.
Since then, we have approved for marketing the following devices:
Sirtex, which was Sirspheres, which is a brachytherapy product for treatment of liver cancers. This is an injectable, and I believe this was brought to the Panel some time ago.
Deus Technologies, which makes a rapid-screen CAD which is used for detecting solitary nodules in the lung from a flat-film lung x‑ray.
Diagnostic Medical Systems, the UBIS Bone Sonometer. This is another bone sonometer. You know, we have seen several of these, and the Panel looked at one of these quite some time ago.
We've approved the Fisher Imaging Corporation's SenoScan, which is a full-field digital mammography system.
We've also approved the Lorad Digital Breast Imager, which is also a full-field digital mammography system.
And then the CADx Medical Systems, which is a second look. It's used as a second look for breast mammography.
And the same thing for Intelligent Systems Software, which is a MammoReader CAD. Again, it is a second look or back-up CAD system for digital mammography.
Along with that, we have approved supplements for Soft Copy Imaging for the Digital Mammography Systems.
All of the PMAs that we have have summaries of safety and effectiveness. If anybody on the Panel is interested in any of these, if you would just drop me a line or leave me a note, I will be glad to send them to you, so you can see what our basis for approval was.
Thank you. Any questions?
CHAIRMAN MEHTA: Any questions for Dr. Phillips?
No? Thank you, Dr. Phillips.
At this time Mr. Doyle would like to make some introductory remarks.
MR. DOYLE: Thank you, Dr. Mehta.
Pursuant to the authority granted under the Medical Devices Advisory Committee charter, dated October 27th, 1990, and as amended August 18th, 1999, I appoint the following individuals as voting members of the Radiological Devices Panel for the meeting of December 10th, 2002. These individuals are Emily F. Conant, M.D., and Regina J. Hooley, M.D.
For the record, these individuals are special government employees and consultants to this Panel under the Medical Devices Advisory Committee. They have undergone the customary conflict-of-interest review and have reviewed the material to be considered at this meeting. This authorization is signed by David W. Feigal, Jr., Director, Center of Devices and Radiological Health.
Now for the conflict-of-interest, the following announcement addresses conflict-of-interest issues associated with the meeting and is made part of the record to preclude even the appearance of an impropriety.
To determine if any conflict existed, the agency reviewed the submitted agenda and all financial interests reported by the Committee participants. The conflict-of-interest statute prohibits special government employees from participating in matters that could affect their employer's financial interests. However, the agency has determined that participation of certain members and consultants, the need for whose services outweighs the potential conflict-of-interest involved, is in the best interest of the government.
Therefore, waivers have been granted to Drs. Regina Hooley, Geoffrey Ibbott, and Prabhakar Tripuraneni for their interest in firms that could potentially be affected by the Panel's recommendations. The waivers allow them to participate fully in today's deliberations.
Dr. Hooley's waiver involves stockholdings valued between $25,001 to $50,000 in the parent of a competing technology manufacturer.
Dr. Ibbott's waiver involves a consulting arrangement with a competing technology firm. For this unrelated consulting services, he receives less than $10,000 a year.
Dr. Tripuraneni's waiver involves an unrelated consulting agreement with a firm that has a financial interest in a competing technology manufacturer. He receives less than $10,000 a year for this service.
Copies of these waivers may be obtained from the agency's Freedom of Information Office, Room 12A‑15 of the Parklawn Building.
We would like to note for the record that the agency took into consideration other matters regarding Drs. Ibbott, Mehta, and Tripuraneni. They reported interest in firms at issue, but in matters not related to today's agenda. The agency has determined, therefore, that they may participate fully in all discussions.
In the event that the discussion involves any other products or firms not already on the agenda for which an FDA participant has a financial interest, the participant should excuse him or herself from such involvement, and the exclusion will be noted for the record.
With respect to all other participants, we ask, in the interest of fairness, that all persons making statements or presentations disclose any current or previous financial involvement with any firm whose products they wish to comment upon.
If anyone has anything to discuss concerning these matters, please advise me now, and we will leave the room to discuss them.
Seeing none, I will proceed.
The FDA seeks communication with industry and the clinical community in a number of different ways. First, FDA welcomes and encourages pre-meetings with sponsors prior to all IDE and PMA submissions. This affords the sponsor an opportunity to discuss issues that could impact the review process.
Second, the FDA communicates through the use of guidance documents. Towards this end, FDA develops two kinds of guidance documents for manufacturers to follow when submitting a pre-market application.
One type is simply a summary of information that has historically been requested on all devices that are well-understood in order to determine substantial equivalence.
The second type of guidance document is one that develops as we learn about new technology. The FDA welcomes and encourages the Panel and industry to provide comments concerning our guidance documents.
I would also like to remind you that the next two meetings of the Radiological Devices Panel are tentatively scheduled for February 4th and March 20th next year. You may wish to pencil these dates on your calendar, but please recognize that these dates are tentative at this time.
MS. BROGDON: Excuse me, Mr. Doyle. You said March 20th for the next meeting?
MR. DOYLE: May. Did I say March? Oh, excuse me. Thank you. May 20th.
CHAIRMAN MEHTA: Thank you for the correction, Nancy.
Thank you, Mr. Doyle.
The first item on our agenda today is a presentation by Dr. Stanley Stern from the Office of Surveillance and Biometrics. Dr. Stern will discuss the development of amendments to the U.S. Radiation Safety Standards for diagnostic x‑ray computed tomography.
DR. STERN: Thank you very much. Just a minor correction: I'm from the Office of Health and Industry Programs, and the Office of Surveillance and Biometrics is kindly sponsoring this presentation.
The presentation is provided solely for your information. It grows out of the collaborative efforts of an FDA group of science, regulation, and economics staff. We're working to facilitate radiation dose reduction through consideration of amendments to the existing CT Equipment Radiation Safety Performance Standard.
I just might mention, by the way, today was published in The Federal Register proposed amendments to the X‑Ray Performance Standard.
The Work Group's current thinking and my own personal views and analysis here, presented here, don't necessarily reflect any official position of the FDA or any of its components. Many items in the slides are annotated with superscripted numbers that refer to citations and notes listed at the end of the presentation. Reference to any products, manufacturers, models of CT systems, or external websites does not imply FDA endorsement.
Computer tomography is a vitally-important, beneficial modality whose radiation doses are relatively higher than those of other x‑ray exams. The scope of CT applications is broad, and CT is used in many different ways, from diagnosis to staging, to treatment planning, and, more recently, for real-time visualization during interventional operations.
Our motivation is the proposition that the current federal regulations covering CT, in place since the mid-1980s, have not kept pace with technological developments and with the need to assure the lowest dose for the best image quality practically achievable.
What is prompting us to consider updated standards? The items on the lefthand of this slide underscore some post-market public health concerns ensuing from the growth and use of CT. The righthand side lists the preliminary responses of CDRH in addressing these concerns.
First, we are faced with the problem of determining the scope of radiological exposure from CT. How many CT examinations are going on annually? And just how large are the doses from what particular exams?
CDRH provided the principal technical direction for a survey conducted through the nationwide evaluation of X‑Ray Trends Program, administered by the Conference of Radiation Control Program Directors.
Between April 2000 and July 2001, state inspectors surveyed examination doses and workloads in 263 CT facilities randomly selected in 39 states to provide the first national understanding of the magnitude of collective dose from CT since the first CT survey in 1990.
A related project is the ongoing development of a handbook of patient doses associated with approximately 50 of the most common CT examinations. Such a handbook would force the risk communication between medical staff and patients, and it would enable medical physicists and radiologists to evaluate patient tissue doses and effective dose for their facility's CT systems and adjust their protocols that CT techs follow in order to reduce doses.
In February 2001, The American Journal of Roentgenology published a series of papers describing the potential risk associated with inappropriate equipment settings and scanning techniques in CT examinations of children. A great deal of publicity resulted from these studies and our concerns were voiced at the 2001 meeting of the Technical Electronic Product Radiation Safety Standards Committee. That's the advisory committee to the FDA.
Following that Committee's recommendation, in November 2001, CDRH issued a public health notification to radiologists, radiation health professionals, risk managers, and hospital administrators alerting facilities to the problem and providing practical advice on how to reduce risks associated with CT dose in pediatric and small adult patients.
Following up, FDA was active in planning and participating in an NCI-sponsored Symposium on Patient Dose held just five weeks ago by the National Council on Radiation Protection and Measurements.
There's been burgeoning popularization of a group of applications commonly referred to as CT screening of self-referred individuals who are asymptomatic of any particular disease. Among these applications include whole body examinations, examinations of the lungs for cancer, and calcium scoring of the heart as a purported indicator of potential heart disease.
Right now CT screening makes up only a tiny fraction of the number of CT procedures performed annually in the U.S. Our main concerns are the risks associated with false positive results and with radiation dose.
False positive results could needlessly lead to follow-up tests or procedures that might be invasive associated with surgical risks of anesthesia, bleeding, infection, scarring, or entail additional radiological exams.
Radiation doses and diagnostics CT are among the highest of those of all x‑ray modalities, and screening CT doses are significantly large, even when low-dose protocols might be applied.
There are no scientific studies demonstrating that whole body CT screening of asymptomatic people is efficacious. Were it a useful screening test, it would be able to detect particular diseases early enough to be managed, treated, or cured and advantageously spare a person at least some of the detriments associated with serious illness or premature death.
At this time any such presumed benefit of whole body CT screening is, in fact, uncertain, and the benefit may not be great enough to offset the potential harm such screening could cause.
Last April FDA posted a web page about CT screening. The page provides information about our concerns. It contains brief explanations of computer tomography, radiation risks, radiation quantities and units, the regulatory status of CT, and includes links to related resources.
It's hoped that an objective presentation from a government institution whose fundamental mission is to protect public health will clarify the natures of the risks and presumed benefits in a way that persuades people to carefully consider these aspects of CT screening before deciding whether or not to have such exams.
Finally, we are aware of the small, but growing, use of what's called CT fluoroscopy or dynamic CT to visually guide interventional procedures and those involving biopsy, drainage, device placement. CT fluoroscopy refers to the capability of a CT system to update images in nearly real time, as the x‑ray field and detectors rotate multiple times around the patient at a fixed z-position; that is, without table movement.
In general, interventional fluoroscopic procedures are a concern with respect to large radiation dose, all the more so, as often as not, they may be performed by physicians in a broad range of specialties outside of radiology, physicians who may not have had particular training in radiation safety. Hence, equipment features that tend to reduce dose systematically automatically, irrespective of professional background, might be desirable.
Recent reports cite mean values of entrance skin dose of approximately 100 to 400 milliGray, below the threshold for skin injury. Several years ago a small CDRH group did guidance for reviewers and manufacturers of CT systems capable of CT fluoroscopy, but to move to formal adoption, a final guidance has been on hold in view of the relatively small probability for skin injury in the most common procedures, and also since preliminary findings of the 2000 CT survey indicated that only 5 percent of the most frequently-used CT units and facilities have the capability of doing CT fluoroscopy in the first place.
The baseline of radiation with respect to CT equipment is prescribed by the federal government through performance standards established under the Radiation Control for Health and Safety Act. The regulations in place now date back approximately 20 years. These rules apply to manufacturers of CT equipment, not to the facilities that use the equipment.
The basic mandate is documentary. Manufacturers must provide users with specified documentation of dose values for CT systems under typical operating conditions. Because this mandate predates special or new modalities, such as electron beam, multi-slice, spiral, fluoroscopic, or cone beam CT, the doses manufacturers report don't necessarily pertain to those modes of operation. There is no regulatory ceiling on patient dose, and there are few major equipment requirements particular to CT per se.
Possible amendments to the current Radiation Safety Performance Standard would require particular technical features for CT equipment. The initial focus of the Work Group effort is on three possible features:
One, display and recording of standardized dose indices. Two, automatic control of x‑ray exposure according to individual patient thickness, and, three, x‑ray field-size limitation for multi-slice systems.
This amendment would require each new CT system to provide users with options to display and record one or more dose indices for every patient's examination. The dose indices and related terminology would be standardized through formal definition and regulations.
An example is the volume Computed Tomography Dose Index, which is the CTDI, corrected for scanning with gaps between slices or scanning with overlapping slices. Another example is the dose-length product which is proportional to the length of the patient volume scanned in an exam.
Such an amendment would enable an aspect of facility quality assurance that today is feasible only with extra effort or through features available on just some newer scanner models. The basis of this quality assurance is the use of what are called reference dose values as norms to which individual examination doses could be compared.
If reference values are exceeded, facilities could follow up anomalies by looking at possible problems to see if exposures could be reduced without compromising image quality. A reference dose value corresponds to the 75th percentile of the distribution of measured dose values for particular radiological procedures. Reference values may be generated based on a facility's own records of dose distributions for various CT exams or based on regional or national dose distributions.
Recent experience in the United Kingdom leads us to assume that the systematic use of dose index display or recording in a facility audit program could reduce patient CT dose on average on the order of 15 percent.
Of the three technical areas that we are considering, probably the largest dose reduction, at least for thinner patients, would be brought about by requiring every newly-manufactured CT system to provide the capability of automatically adjusting the amounts of x‑ray emissions into those needed to image particular patient anatomy.
In other words, as the x‑ray beam probes a thinner portion of the anatomy, which would not require as much radiation as a thicker portion would in order to reach the detectors, the CT system would automatically reduce the average tube current or voltage or some combination of radiological variables to spare that thinner part unnecessary dose. Conversely, when the beam encounters thicker anatomy, the CT system would automatically increase the tube output to levels needed for adequate visualization.
An automatic exposure control system offers a technical answer to facilities where, for practical or clinical reasons, it is not the practice to change manual techniques on a patient-by-patient basis, let alone readjust techniques within a single patient exam.
With an AEC system in place, the presumption is that pediatric and thinner adult patients would receive lower doses than thicker patients. Calculations and measurements suggest that use of a sophisticated automatic exposure control system could reduce patient dose by approximately 30 when compared to systems where the techniques are set manually.
We are concerned that a number of different multi-slice CT models produce images with a technologically-inefficient application of radiation. This inefficient technology has been dubbed "overbeaming."
The two figures represent a comparison of the spatial distributions of radiation incident on a patient. The figure on the left depicts the distribution for a single-slice CT scanner, whereas the one on the right corresponds to that of a multi-slice scanner.
Here's the important point in this comparison: Although the amount of radiation applied to construct one image with the single-slice scanner or to construct a set of images with the multi-slice system is the same for each configuration, for the multi-slice CT system the radiation distribution is much wider than that of the single-slide system.
Why? Multi-slice CT imaging requires that radiation incident on the patient be consistently distributed across each of the separate areas subtended by the detectors. Such consistency can be achieved by opening up the z-collimation of the source radiation so that only the most spatially-uniform region of the x‑ray field, the umbra, is subtended by the detectors.
Furthermore, since the x‑ray focal spot tends to wander around spatially, multi-slice models broaden the umbra by opening the collimation even more to compensate for x‑ray source excursions. All of the radiation that falls beyond the spatial extent of the detectors is not used by the detectors for image construction, but it is, nevertheless, incident on the patient and contributes to the dose.
To mitigate the inefficient use of radiation in multi-slice computed tomography, we suggest consideration of an x‑ray field-size limitation. Such an amendment would require that all new CT systems be capable of automatically limiting field sizes to those no larger than needed to construct multi-slice images. Several technical approaches to enable such limitation have been patented, and one, in fact, has been implemented.
The approach implemented uses some of the x‑ray detectors lying behind those capturing the clinically-useful signal to track the wandering of the penumbral regions of the x‑ray field and feed back instructions to motor-driven collimator cams to readjust their positions. The result is that the umbra remains subtended by the clinical signal detectors.
The chart on the right depicts two multi-slice dose profiles measured in a head phantom on the same CT system. For the same 5-millimeter-wide imaging sensitivity profile, the dose profile in black is obtained when there is no tracking and collimation update system, whereas the dose profile in fuchsia is obtained when the tracking update system is activated.
It is evident that the non-tracking dose profile is approximately 50 percent wider than the tracking profile. All of the radiation represented by the difference between the two profiles would correspond to radiation which is absorbed by a patient, but not used to construct imagines. Data suggests that the kind of x‑ray field-size limitation enabled by tracking collimation adjustment could reduce dose in multi-slice CT systems on the order of 30 percent.
If all CT equipment were to include the technical features just proposed for consideration as mandatory standards, then, based on the relative dose reductions and the collective dose attributable to CT, one can estimate an annual collective dose savings of 193,000 person-Sieverts per year. For an annual collective dose savings of 193,000 person-Sieverts, on the order of 870,000 radiation-induced cancer mortalities are projected to be avoided per year, beginning 20 years after each annual collective exposure.
The yellow shading is intended to highlight the uncertainty in this projection, which is based on an extrapolation to the CT dose region of a mortality risk estimate derived from larger-dose epidemiological data. Other methods of extrapolation could yield higher or lower estimates of the number of radiation-induced cancer deaths, and it is conceivable that the estimated dose savings would not result in any significant avoidance of cancer death at all.
In the United States in the year 2000 the annual number of deaths linked to cancer from all causes not specifically associated with radiation is approximately 550,000. There would also be a significant benefit and pecuniary savings associated with society's willingness to pay to avoid mortality risk, and economists have estimated that society pays on the order of $5 million per premature mortality that it perceives might be avoided.
We have come up with a framework for analysis that will lead to what is called the Concept Paper for possible development of amendments. That is an internal document which will be the basis for CDRH decisions on how to proceed.
In the block on the right, the green shading indicates the technical areas summarized in this presentation, and the red shading contains areas where we have an interest that is deferred for the time being. The yellow-shaded block on the left lists some general categories of issues: technical feasibility, impact on clinical aspects such as efficacy, frequency of utilization, harmonization with international consensus standards, CDRH resources required to develop test methods and to incorporate the administration of new rules in a compliance program. The arrows indicate that in principle each of these issues can be applied as a basis of assessment to each technical area under consideration.
Although the equipment features that I have discussed today may all be technically-feasible, there remain a number of particular problems outstanding. Here are a few examples:
First, for the purpose of display or recording in a quality assurance program, not only would we have to select a representative index of patient dose, we would need to specify whether the dose index could be based on average values, determined by manufacturers for all models of scanners, or whether it must be specific to the particular unit actually being used in a facility.
Perhaps the dose index displayed or recorded could be based on real-time measurements made during actual patients' examinations. It is not clear how the index would represent values in an automatic exposure control mode. Parameters based on CTDI may not be good candidates to represent skin dose, particularly for CT fluoroscopy. A good skin dose index may need to be developed. A recording capability for a dose index may affect practice and use, and we ought to consider such impact.
Second, with respect to automatic exposure control, in addition to specifying what kind of technological approach is best, perhaps the key issue is how to define the optimal amounts of radiation needed by the detectors for particular imaging tasks. These amounts would effectively set the points of detection equilibrium, driving the modulation of emissions from the x‑ray source according to patient anatomy thickness.
Perhaps we need to set standards to optimize detection. It's not clear who ‑‑ manufacturers, radiologists, FDA ‑‑ should set the equilibrium points and how that would be done.
In a related issue, Philip Judy, a prominent medical physicist, cautioned that while automatic exposure control may reduce dose to thinner patients, it also might on average increase dose to thicker patients.
Third, a primary challenge in developing an amendment for x‑ray field-size limitation or for automatic exposure control, and most likely other areas as well, would be how to prescribe performance standards, not design standards, forward-looking enough to transcend limitations that might be present in current technological approaches.
In conclusion, an FDA Work Group has identified several areas of possible development of mandatory CT equipment radiation safety performance standards. The initial focus is on technically-feasible features that would reduce patient dose; dose index standardization; display and recording; automatic exposure control, and x‑ray field-size limitation.
Were these features implemented on all CT systems, the projected collective dose savings in the United States would be approximately 193,000 person-Sieverts yearly. The Work Group has established a framework of issues for analysis that would be detailed in a regulatory concept paper for internal decisions on how to proceed. We expect industry, professional groups, and states to contribute to our development process.
Our timeline for the initial stage of this process is to begin a Concept Paper by the end of this year, and next year brief the FDA Advisory Committee, the Technical Electronic Product Radiation Safety Standards Committee, from whom we would seek further recommendations.
I thank you for your attention.
CHAIRMAN MEHTA: Thank you, Dr. Stern.
Any questions for Dr. Stern?
I think we are about five to ten minutes early, but we will get started early. That will give us an opportunity to devote enough time to the primary agenda of today's meeting, which I did not state at the beginning. So I'll restate that for the record.
So the purpose of the meeting is to discuss, make recommendations, and vote on a pre-market approval application, PMA P010035, for a device that produces a computerized thermal image of the breast of women recommended for biopsy.
We'll begin the public hearing session of the meeting. This will be the first of two half-hour open public hearing sessions for this meeting. The second half of our open public hearing session will follow the Panel discussion later this afternoon. At these times, public attendees are given an opportunity to address the Panel to present data or views relevant to the Panel's activities.
It is my understanding that no individual has given advance notice of wishing to address the Panel. If there is anyone now wishing to address the Panel, please identify yourself at this time.
Seeing none, I would like to remind the public observers at this meeting that, while this portion of the meeting is open to public observation, public attendees may not participate except at the specific request of the Chair.
I would like at this time that persons addressing the Panel now or later to come forward to the microphone and speak clearly, as the transcriptionist is dependent on this for providing an accurate transcription of the proceedings of the meetings.
If you have a hard copy of your talk available, please provide it to the Executive Secretary, so that they can use this and the transcriptionist can help provide an accurate record of the proceedings.
We are also requesting that all persons making statements either during the open public hearings or the open Committee discussion portions of the meeting to disclose if they have financial interests in any medical device company. Before making your presentation to the Panel, in addition to stating your name and affiliation, please state the nature of your financial interest and the organization you represent. Of course, no statement is necessary from employees of the sponsoring organization.
Definition of financial interests in the sponsor company may include compensation for time and services of clinical investigators, their assistants and staff, in conducting the study and in appearing at the Panel meeting on behalf of the applicant; a direct stake in the product under review ‑‑ for example, inventor of the product, patentholder, owner of shares of stocks, et cetera, or owner or part-owner of a company.
We can now begin the first open public portion of this meeting. We will begin with presentations on P010035 by the sponsor.
So this will conclude the open public portion of the meeting. We will now begin the open Committee discussion. Again, this is for PMA 0100035, for a device that produces a computerized thermal image of the breast of women recommended for biopsy.
The sponsor, Computerized Thermal Imaging, Inc., CTI, will state its case for the PMA and be followed by the FDA presentations.
The first speaker will be John Brenna, the President and CEO of CTI.
MR. BRENNA: Good morning. My name is John Brenna. I would like to thank the members of the Panel Review Board, the Food and Drug Administration, ladies and gentlemen from the press, and the general public for the opportunity today to introduce CTI's breast imaging system known as the BCS 2100.
This is a non-invasive, adjunctive device that's designed to work with mammography to obviate biopsy of benign masses. It's a system that provides physiological information based on infrared technology that supplements the anatomical information provided by mammographic x‑ray.
Before I introduce the other members of the team, I would like to take a few moments to highlight some of the key points about this device that you will be hearing about in more detail in the following presentations.
This is a diagnostic breast imaging system that provides a painless, non-invasive patient procedure, a procedure that takes less than ten minutes to complete, that captures over 100 dynamic images and collects over 8.3 million temperature values per imaged breast. It is adjunctive to mammography x‑ray and it provides physiological information.
A clinical manuscript describing the clinical trials of the BCS 2100 has been reviewed and accepted for peer review and will be published by The American Journal of Roentgenology, which is scheduled later this month.
Our product focus is to market the BCS 2100 exclusively to MQSA-certified facilities and under control of Board-certified radiologists.
Now I would like to take the time to introduce the members of the CTI presentation team, and from the CTI organization I would like to introduce Lynn Satterthwaite, our Vice President of Engineering. Lynn, would you stand up?
Dr. Karleen Callahan, the Director of Clinical Research.
I also would like to introduce our principal investigators, Dr. Yuri Parisky, Associate Professor of Radiology from the University of Southern California School of Medicine, also a Director of Breast Imaging Services at USC/Norris Comprehensive Cancer Center in Los Angeles, and was a principal investigator during the clinical trials. Dr. Parisky.
I also would like to introduce Kevin Hughes, Assistant Professor of Surgery, the Harvard Medical School; Surgical Director of Breast Screening at the Massachusetts General Hospital, and also a principal investigator at the Lahey Clinic. Dr. Hughes.
Advisors to the CT organization include Dr. Kathy Plesser, former Chief of Breast Imaging at St. Vincent's Comprehensive Cancer Center in New York. Not joining us today, but an advisor to the organization is Dr. Pat Romilly, Assistant Professor of Radiology at H. Lee Moffitt Cancer Center.
Other advisors: Dr. Steven Rust, Senior Research Leader from the Battelle Institute. Steve.
Dr. Loraine Sinnot, Research Scientist at the Battelle Institute, and Elizabeth Nelson, a Senior Regulatory Consultant from the Catalyst Group.
Our agenda today will cover the following topics: Dr. Parisky will lead off with a presentation regarding the clinical environment, followed by a presentation by Dr. Hughes on patient management and case presentations, a device description and an operation ‑‑ and you may note off to your right we do have a system with us to demonstrate during the break periods ‑‑ followed by a clinical trial process and statistical procedures presentation by Dr. Callahan, and a wrapup on the efficacy of results and indications for use by Dr. Parisky.
DR. PARISKY: Thank you very much, John. Thank you. It's good to meet distinguished members of the Panel as well as members of the audience.
I'll give you a brief perspective on mammography and diagnostic breast imaging; I'm sure most of you are familiar with it. Mammography made its introduction in the sixties with subsequent development of film screen mammography for the purpose of detecting occult or clinically-occult breast cancer. The majority of abnormalities at that time went to biopsy, most likely surgical biopsy.
The introduction and proliferation of diagnostic mammography with specialized views, unfortunately, did not add significant sensitivity or specificity and did not detract significantly from the biopsy rates.
The introduction of ultrasound within the past decade introduced an early drop in the number of breast biopsies performed because of the recognition of cysts, but now the proliferation of ultrasound-guided biopsies have led to an increase in biopsies of abnormalities detected.
Other modalities have been introduced and are discussed in literature and at scientific conferences, notably Sestamibi, PET scan, MRI. All of these modalities advocate or publish sensitivities in the high eighties to low nineties with specificities ranging anywhere from the twenties to the eighties.
The ethical imperative I have as a full-time clinical diagnostic radiologist who practices in breast imaging is to do no harm to the patient. My job is to detect breast cancer with at the same time the imperative to do no harm by unnecessarily biopsying the patient.
Every year at least 1.3 million to maybe upwards of 1.5 million women undergo breast biopsies to determine if they have cancer of the breast. At least 80 percent of these biopsies are benign. That means over a million women undergo a procedure for benign process.
There are a number of inconveniences and discomforts and traumas associated with the breast biopsy procedure: the initial discomfort, inconvenience, the anxiety that is associated with waiting for the biopsy to occur, and statistics or publications have commented upon the trauma, even though the result is a benign breast biopsy, may linger for many years with that woman. Fortunately, it doesn't preclude them from seeking further screening, but it is a traumatic impact. There's psychological trauma and physical trauma and potential for complication.
The proliferation of literature amongst radiologists and clinicians in this field discussing the need to reduce breast biopsies over the last decade, unfortunately, has not resulted in the reduction of breast biopsies. Why? Because the sensitivities remain as they are in the high eighties to low nineties, and the specificities range, as I stated.
Next slide. A patient who undergoes a screening mammogram today in the patient who is asymptomatic has approximately a 10 percent chance of being recalled for further diagnostic views. That patient who is recalled, the patient will undergo approximately 1.5 to 2 diagnostic examinations, at which point the clinician or the radiologist is left with the observation of whether to proceed to biopsy or not.
We shall present to you a technology which is non-invasive that in the prospective study that was performed in the selected, prospective subset of masses which account for nearly 50 percent of all lesions that go to biopsy will show you a negative predictive value approaching 100 percent, a sensitivity that approaches 100 percent, and a specificity that is 20 percent of women already selected by all the other examinations to undergo biopsy.
We hope by introducing a technology which offers the physician and the patient a physiological perspective to complement the anatomic imaging that is performed by mammography, diagnostic workup, and ultrasound, a chance to counsel that patient that the fact that a negative predictive value in her mass may obviate the need for a biopsy.
Physiological imaging is likely based on the exclusion or recognition of proliferative changes within the breast likely due in the case of malignancy for angiogenesis or the absence thereof.
I would like now to introduce a colleague, a co-author on the paper, a principal investigator, Dr. Kevin Hughes from Harvard Medical School.
DR. HUGHES: Good morning. As a breast surgeon, I wanted to give my perspective on where this machine will fit into the work of patients with a breast abnormality.
As Dr. Parisky has pointed out, mammography or other screening modalities identify patients who are at risk of possibly having a cancer and require additional workup. Those for whom no further workup is needed go on to routine screening. Those who need additional workup normally undergo an ultrasound or physical exam to determine whether or not this lesion is suspicious enough to require a biopsy.
At this point in time, if the lesion is suspicious enough to require a biopsy by mammography, by physical exam, or by ultrasound, our only option is to go on to biopsy, either a core biopsy or open biopsy, to determine whether or not this is cancer.
Next slide, please. How we are hoping that our imaging will fit into this routine is that the same patients now who have been identified as being suspicious by mammography, by ultrasound, or physical exam requiring biopsy, rather than going directly to biopsy, will go to IR imaging as their next step. If the IR imaging is negative, that patient potentially could go on to followup instead of a breast biopsy.
I think it's important to point out that the IR imaging is like any other imaging modality and has to be looked at in conjunction with mammography, ultrasound, and physical exam. Even if IR imaging comes out negative, if this is highly suspicious, we'll still go on to biopsy. However, if we believe that this patient is likely benign, this may help us to confirm that impression and avoid a biopsy.
For those patients where the IR imaging is positive, these patients go on to biopsy, as suggested. We are hoping that this will decrease the rate of biopsy for these individuals.
Next slide, please. This is a case just to show, again, how this will fit into clinical practice. This is a patient found on screening mammography to have a mass lesion in her right breast.
Next slide, please. Additional views on ultrasound confirm this to be a solid mass lesion in the breast which was not palpable.
Next slide, please. At this point in time the patient would have been or was scheduled for a breast biopsy. If the IR imaging device is approved, what we then do is take the patient for IR imaging, the table as you see over here and will demonstrate. The patient lies flat on the table. The breast being imaged fits into the hole here. The non-imaging breast goes into one of these holes for patient comfort. Each breast is imaged individually, and then those images are combined.
Next slide, please. This is a look from inside the table. The hole where the breast comes in is here. That hole is surrounded by six first-surface mirrors which collect the heat image of the breast. Those images are reflected off the reflector here into the collecting camera.
The image that's taken of the breast are these six individual mirror images made by these mirrors, plus an en face image of the breast. The outline of the breast itself is made in red by the technician. The nipple is also marked in red by the technician.
Next slide, please. At this point in time we look at the two images of the right and left breast. As you know, in this patient the lesion was in the upper outer quadrant of the right breast.
It is important to realize that this is not a visual image that can be easily interpreted by a radiologist or a surgeon. Essentially, looking at this area, you cannot glean any information just visually. What you need is the algorithm to determine whether this is, indeed, a suspicious lesion.
Next slide, please. At this point we bring a region of interest over the area that we consider to be involved by the mammographic lesion. Trying to get the lesion placed as perfectly as possible is important, but not critical. Within a short distance of the lesion appears to be adequate.
Once we identify the region of interest, the machine runs through its algorithm and gives us a report, which comes up as either negative or positive, shown here.
Next slide, please. Here, again, shown as a negative report. So, basically, we have a lesion on a mammography we believe is benign, but we require a biopsy under current clinical practice, even though we all suspect this to be a benign lesion.
We undertake this test, which shows that it does not light up in the way that we would consider malignancy from the thermal imaging. This should give us enough information to help us avoid a biopsy in this individual.
Next slide, please. In this particular case, under this study, we did proceed with biopsy of all these patients, and this patient, indeed, had a benign lesion showing periductal and stromal fibrosis.
At this point I'll turn the podium over to Lynn Satterthwaite. Thank you.
MR. SATTERTHWAITE: Good morning.
The first slide here shows the electromagnetic spectrum. Because sensing infrared energy is so central to our device success, I show the infrared light as part of that spectrum. You can see there that it is shorter than the radio waves we listen to and longer than the x‑rays that are part of mammography and imaging sessions.
Next slide. Infrared energy is emitted by the body as a result of the physiologic processes that take place in the body. Our camera passively senses that infrared energy and records that information as temperatures. Therefore, we do no harm to the patient. This is a non-invasive device. There is no risk involved in sensing infrared energy given off by the body.
What we're going to do for a few minutes is to discuss the device. There are two primary functions, the acquisition of the data and then the evaluation of the data. We'll first talk about the device itself ‑‑ let's back up, please ‑‑ the device itself, which we have a model, actually a working model, over here, a working device. Then the evaluation system I don't have here, but is very similar to a simple desktop computer.
Next slide. Here I show the functions that are performed during the imaging session in sequence. We'll first talk about data acquisition. Those that are outlined in green are those functions, and we'll discuss those first today.
Next slide. Let me just direct your attention for a minute over here to the bed. You've seen a picture of the bed. The bed is composed of a support unit. There's a table-top unit here. Inside the bed is an optical system. Dr. Hughes gave a fairly good explanation of that optical system. Also inside the bed is a camera and then a simple cooling system.
This bed functionally is equivalent to what was used in the clinical trials. It is exactly functionally equivalent, except that some of the functions that were operated manually by the technologist are now done automated by the computer.
Let me just take a minute and take you through the sequence here. Once the patient has been enrolled or it is decided that this image is going to be taken, the patient will disrobe and equilibrate.
We then bring the patient here. The technologist will lay the patient prone on the bed. You noted on the pictures on the screen ‑‑ you won't see it very well here, but we'll invite you to step up here and take a closer look anytime in the breaks ‑‑ there are three holes. The center hole is the imaging hole. The two holes on the side will accommodate the other breast while imaging takes place.
So the patient lays prone on the bed. We suspend the breast pendulously in the center. This center hole is where those six mirrors surround the breast tissue for that imaging session.
The computer controls the beginning of imaging. That's initiated by the technologist once they note that the patient is properly positioned. The imaging begins, and then 30 seconds into the process the computer controls the turn-on of the cooling process, and the cooling continues as the camera continues to record through the remainder of our taking over a hundred images of the patient.
So there's a lot of simplicity here. We have an optical system and a camera, a cooling system. It's important to note that they are controlled by a computer that takes the human element out of that.
I also have here a model that I'll invite you to come over and take a look at. This is the old "a picture is worth a thousand words" approach. But this model has a miniature set of six mirrors at the top.
Let's go to the next slide. Those six mirrors that you see on the left of your screen up there are first-surface gold-coat mirrors. They're arranged so as to optimally image the suspended breast. They're at the top of this device.
That information is reflected off a main bounce mirror that is down here at the bottom. That bounce mirror then is viewed by the camera, and all the information of the bounce mirror is what's recorded.
On the device here is a plastic model of a breast, and when you take a look here, you will be able to see how we are able to see all sides of the breast through the optical subsystem that is available here.
When the imaging is complete, the technologist will see two composite images similar to what you see here.
Next slide. The camera that's central to this process is a scanning camera. It uses a mercury-cadmium-telluride sensor that's sensitive in the 8- to 12-micron band.
Next slide. What I show here is the electromagnetic spectrum in the IR band, noted here on the bottom in microns from 3 to 13 microns. The arrows represent the emission of the human body of IR energy. Note that the optimum emission is at about 10 microns, though the body emits at several wavelengths here, as you can see.
Our camera is sensitive in the 8- to 12-micron band, so you can see what we have done is optimize the recording of the optimum signal given off by the body. Other sensors that have been used in the past will pick up infrared energy, obviously, but it won't be optimum.
Next slide. The camera then is important because it will detect differences in temperature as low as a tenth of a degree between two pixels. Translated into the human breast, between two points on the breast, we can see differences that are very small in temperature. Our camera resolves the area of the breast down to less than 2 millimeters, and it takes just about a millimeter-and-a-half as the smallest area that's resolvable, where you can see a different temperature on the skin.
So with the technology we've talked about, the sensitivity and the IR band, the resolution of the camera, and the temperature sensitivity of the camera, we're able to capture the infrared energy that's emitted by the body in the form of temperatures, and that provides physiological information that will supplement the anatomical view provided by x‑rays.
Because the cooling challenge is important, so important, to the information that we gather with the system, I would like to just talk you through that cooling challenge.
Next slide. First, the technologist will position the patient. I show a timescale here on the bottom. Once the patient is positioned, the camera begins the imaging session, which continues for over three minutes.
About 30 seconds into that imaging session, the computer turns on the cooling, and that cooling continues through the remainder of the little over three-minute imaging session, where we take 103 images.
It is important here to note that it's the computer that's doing the control there. It is also important to note that, because we have a computer in the system, we're accurate and precise. The cooling challenge elicits that physiological challenge, and then we record that view in the camera and provide that to supplement the anatomical view that's provided by mammographic x‑ray.
This is an image. We're going to attempt here to show you what happens during that cooling challenge. So we're going to compress about three-and-a-half minutes into a few seconds here.
Notice the central image there, the en face image, it's here in the center. The six mirrors that Dr. Hughes discussed earlier present those parts of the breast tissue. Then, on command, you will see that light areas are warmer; dark areas are cooler. During the cooling challenge you see what happens there. Let's try it one more time.
Notice how it gets cooler as time goes on. Again, we have compressed three-and-a-half, a little over three minutes into a few seconds here, but you get the idea of what is happening.
It is important to point out, again, that there's no diagnostic information when the physician views one of these images. This is an analysis-intensive modality.
Next. Once captured, the infrared data is processed and we develop a single, composite image.
Now I would like to take you through the second part of this. The process for our system is to evaluate the data. It is outlined in the functions on the bottom here in green letters.
Initially, the technologist places outlines on the breast. You can see here that the outlines have been placed. Basically, the technologist has a number of ellipses there that they can move around and just capture breast tissue. The physician then confirms and changes those outlines as necessary.
Once those outlines are placed, then the physician, utilizing mammographic x‑ray views, will localize using those x‑ray views and transfer that localization onto the composite image here. You can see here a region-of-interest marker is placed on the breast in the appropriate location.
Because we recognize that between modalities placing the ROI may have some variance, our system does a search to improve that localization. What I show here is this breast tissue area here has been blown up here or magnified, and that region-of-interest marker is the center of this circle here. What we do is calculate the pixels inside this outline and take about one-twelfth of that area and surround it around this region-of-interest marker and do the calculations then to potentially improve that localization. We look for a stronger IR signal.
So we gather over 8 million temperature datapoints, and then analyze select datapoints in order to come up with our diagnostic information. Again, it is important to point out that the infrared images are not visually interpreted. This is an analysis-intensive approach.
We then go on and do the index-of-suspicion calculation. This is what is presented to the physician.
Next. Once the region of interest has been placed, it takes less than three seconds for the calculation to take place, and then we present the results shown here, a negative result. We actually do have an index of suspicion. Here this is the version of the system at this time, and we anticipate then taking out this index of suspicion and just presenting a test result that is either negative or positive.
Next. CTI provides training to both the technologist and the physician for those things that are important to the success of the use of our device. The technologist is trained to position the patient to do the proper procedure. The technologist is prompted by the computer to do certain things as far as positioning and looking for artifacts that might render the image not useful, and then would prompt them to go back and re-image, if necessary. Then we teach them about outlining the breast and then some about upkeep and maintenance that is done by the technologist.
Next. The physician, we train the physician in lesion localization, particularly how to translate from the x‑ray view that they already are familiar with onto our composite image the placement of that ROI marker, and then the utilization of the IR test result.
In summary, our system performs a computerized analysis to differentiate between malignant and benign tissue. It is comprised of non-invasive, safe components. There's no exposure of risk to the patient. It is designed to be a non-invasive, adjunctive medical device for use by the radiologist.
I now would like to introduce Dr. Karleen Callahan, who is Director of Clinical Studies.
DR. CALLAHAN: Good morning. I'll go through the clinical study protocol, talking about the study objectives, hypothesis, study design and procedures, as well as the study flow, our subject enrollment demographics, safety results, and then, finally, our efficacy groups and statistical analyses.
The study objective, as stated in the protocol, was then to determine whether the CTI BS 2100, when used in conjunction with mammography, increases the ability of physicians to differentiate benign from malignant breast abnormalities.
The hypothesis, therefore, was that this differentiation between benign breast lesions from malignant breast lesions was based on the relatively lower strength of the infrared signal in benign tissue. The goal, then, therefore, is to reduce the number of benign biopsies.
The study was designed as a blinded investigation. The initial protocol involved one site and 600 subjects. It was eventually expanded to five sites in order to obtain a sufficient number of malignancy for analysis purposes.
Our effectiveness evaluation criteria, as outlined in the original study protocol, were area under the curve, sensitivity, specificity, as well as subpopulation analyses.
Our clinical sites are listed here and were throughout the country: Dr. Parisky, who you have already heard from, at USC, and Dr. Hughes, who is in Lahey Clinic outside of Boston. We also had investigators, Dr. Robert Hamm here in Washington, D.C., at Providence Hospital; Dr. Esserman at Mt. Sinai Medical Center in Miami, and Dr. Sardi at St. Agnes Health Care in Baltimore. So we had wide geographical representation.
The study flow was such that subjects were identified and enrolled. They underwent infrared imaging and then proceeded to biopsy. The biopsy results, a pathology was sent to an independent research organization, Quintiles, and kept vaulted from CTI.
The IR imaging data as well as patient information obtained in a case report form and mammography were sent to CTI. A trial read was done by independent evaluators, as I'll discuss subsequently.
After the trial read was completed and the database controlled and locked, at that time the pathology results then were unblinded and sent to CTI and Battelle for analysis purposes.
Our evaluator panel for the infrared procedure were seven independent mammographers who are currently practicing. Three of these mammographers were director of their breast imaging centers; one had participated in MQSA standards development. They were all currently practicing and read several thousand mammograms each year. Again, they were independent from the sites and blinded to lesion pathology results.
The clinical research organization that was involved in this study is Quintiles. They monitored the investigative sites, did 100 percent source documentation. They held the blinded pathology results and, again, after the IR evaluation phase, a locked database was supplied to them prior to unblinding.
I'll next go over the study inclusion/exclusion criteria. The study inclusion in the original protocol was subjects that were recommended for biopsy based on mammography and/or clinical findings. These subjects did receive and sign IRB-approved informed consents.
Exclusion criteria were subjects that had previous breast surgery. I will give a little bit more details in a moment. Other exclusion is if they had had radiation in the breast that was to undergo biopsy, if they had had either breast implants or breast reduction surgery. There was a weight limitation of 300 pounds, based on the table weight limit. Other exclusions were patients that were pregnant or had previous diagnosis of breast cancer.
For the study protocol, there were two amendments during the study trial period. The first amendment was in November of 1998, and this amendment changed the method of evaluating the mammogram and IR imaging.
The original protocol had intended that the site investigator that enrolled the patient would, in fact, assign mammographic LOS, level of suspicion, similar to BIRADS categorization, as well as do the IR imaging evaluation. It was determined in this amendment that, in fact, these reference independent radiologists would do that procedure. I will talk a little bit more about some of the developmental problems with this independent evaluation, trying to determine level of suspicion from mammograms.
The second protocol amendment was initiated in June of 1999, and it reduced the prior breast exclusion criteria. There initially had been no prior breast surgery for three years; that was reduced to a one-year time period.
So to reiterate, the investigators would screen, the principal investigators would screen and enroll subjects. The IR data acquisition would begin at those sites, and the original investigator would record a variety of subject data, including lesion descriptor; that is, was this a mass, microcalcification, architectural distortion, and so forth? They would describe the lesion size and they would complete the pathology outcome.
The reference or the evaluating radiologist would do a level-of-suspicious determination, and when they was done, this was based, then, on radiology reports and overread of the mammography films that had been provided by the original investigative site.
It became apparent that there were some issues related to this because it was not being done in real time. These reference radiologists did not always have all the information that would have been available to the original investigator; that is, they wouldn't have had prior films, and oftentimes a radiology report might contain additional information, for example, ultrasound information.
So a comparison based directly on mammography alone was difficult to establish. Because of this, we eventually decided that comparison for efficacy to the biopsy decision ‑‑ that is, these patients went to biopsy; we had the pathology result ground truth that this comparison was more rigorous.
Our study procedures for the physician evaluators then were that they reviewed the mammogram. They checked the breast outline, as you have heard. They placed the ROI marker on the IR image, and then an assignment of either a negative or positive IR result was obtained.
I will discuss the subject enrollment briefly. This shows it across our five sites. Our total patient enrollment for the study was 2,407 patients, and our enrolling centers ranged from 170 up to over 800 patients.
You will hear some of the additional demographics later on from the FDA. So I will just talk about lesion size.
The majority of lesions were either between ‑‑ and this is in our final efficacy group that I'm referring to here of 490 masses ‑‑ our lesion size ranged between .5 and 1 centimeter or greater than 1 sonometer. There were a small number of masses less than .5 sonometers.
For safety issues, of the 2,400 subjects enrolled, all were followed for safety. There were four adverse events noted. Two of these were mild and felt related to the device and related to discomfort lying on the imaging bed or just exacerbation of a previous condition. Two other adverse events not related to the device were also reported.
I am going to spend a little bit of time about the withdrawn cases for study flow. So I will go through a series of charts here.
We started out with, as I just mentioned, 2,407 subjects. Seven hundred of those were in an algorithm development group and 275 were not part of the original unvaulting. So when we began the analysis phase, we had 1,432 subjects that had some with multiple lesions. So we had 1,660 lesions.
There were three categories of withdrawn cases that were determined prior to the evaluation phase. The first group were those that we're calling non-conformatory to study protocol. Basically, these were patients that did not undergo biopsy, and the flow of the study was such that patients were ‑‑ these were often referral centers ‑‑ patients had been recommended for biopsy. They came in, and perhaps they were scheduled for an ultrasound, not a biopsy, and a cyst, fluid-filled cyst, was identified and they did not proceed to biopsy. So there were a number of cases that did not undergo biopsy. Of course, they were dropped.
There were also cases where we had missing or incomplete mammography films. That is, if there weren't sufficient views for our independent evaluators to localize, then they could not be entered into the evaluation phase.
Finally, there were some cases with unusable infrared images. A lot of those difficulties were related to cooling, the start of cooling, which wasn't automated at the time of our clinical trial, as well as positioning issues. With technologist training, we think that we have mitigated, attenuated that problem. So at this point we had withdrawn cases. This was determined prospectively prior to entering the evaluation phase.
Those subjects that did go on to the evaluation phase, there were also a few more withdrawn cases, primarily because the physician evaluator could not evaluate the case. This could be due to, for example, if they did not see a lesion in the area that had been described on the case report form or the lesion didn't correspond with the descriptor.
There were a few also that were location discrepancies; that is, a priori we made a determination model for, if the ROI marker was significantly outside of the location of the biopsy lesion as described in the case report form, it was removed from the analysis phase. Again, all these withdrawn cases were determined prospectively prior to unblinding of our results.
Finally, then, for our original study, we described 875 lesions in 769 subjects. We had, as you will hear from Dr. Parisky subsequently, we had good results with all the lesions. However, we also looked at a subset we had described that we would do prospectively, and that is the masses.
So the points I wanted to make were that the lesions withdrawn were done prior to IR evaluation or after IR evaluation prior to unblinding, and we did do statistical analysis comparing those withdrawn lesions to those that were analyzed. This says no statistically-significant differences were found. One difference was that more of the spiculated masses were evaluated, and probably Dr. Rust will address that.
So our mass, when we did all the cases, we also then decided to look at the masses and focus on masses, because to minimize the risk to patients, maximize benefit for labeling and patient benefit purposes, it was felt that our target population would focus on masses.
The FDA did request a confirmatory study for masses, and we did another small confirmatory study with those 275 cases that remain vaulted looking at the masses. So our efficacy groups that Dr. Parisky will describe include our original study with all lesion types, those of masses from the original group, a confirmatory study, and then our combined results.
So our confirmatory study involved those 275 patients. They underwent the identical evaluation process, and the result was 78 masses.
Dr. Rust, our statistical expert from Battelle, then will talk about the subsetting. Thank you.
DR. RUST: My name is Steve Rust, and I am employed by Battelle Memorial Institute, and I would like to disclose that Battelle does have a fee-for-service contract with CTI to provide statistical services and that neither Battelle nor I have any equity position with CTI.
Because the target population of masses was selected after the pathology for the original study subjects was unblinded, it presents a bit of a challenge in how to interpret the original study data. I believe it is important to point out that this subsetting, two masses, was based on results from a prospectively-planned subsetting analysis by lesion type. What I mean by prospectively-planned is that the study protocol did indicate that analyses would be conducted by lesion type for subsetting purposes.
Now that subsetting analysis could have resulted in seven different possible target populations if you take all combinations of the lesion types that were prospectively stated; that is, masses, calcifications, and distortions. So we have listed on this slide the seven possible target populations that could have resulted from the analysis.
Next slide, please. Now given that we could be here before you today reporting on one of seven possible target populations, and we are, in fact, proposing that the device be targeted for the subset for which the best performance was obtained, it is necessary to carefully handle the statistical analysis.
There are two options for handling this situation. One would be to perform an entirely new study and draw your statistical conclusions only from the new study data.
Another option is to apply a correction that validates the statistical conclusions drawn from the original study data. Either approach is scientifically correct, and I should also point out that it is possible to apply a combination of the two approaches, which is in fact what CTI did.
As Dr. Callahan pointed out, a confirmatory study was carried out using a subset of 78 masses from the 275 patients that remained unvaulted at the time that the target population was focused on masses, and that data from the new confirmatory study was added to the clinical trial dataset.
A Bonferroni correction then was applied to the results for the combined data from the original study and the confirmatory study to correct the statistical inferences for the fact that seven potential target populations were considered prospectively. Now when I say ‑‑ could you back up, please? ‑‑ when I say, "conservatively applied," I want to point out that no credit was taken for the fact that the data from the confirmatory study is actually new and independent data. So the Bonferroni correction was simply applied to all of the data, taking no credit for new data.
The result is that statistical conclusions that we would report for any of the seven possible target populations are valid, and, therefore, the conclusions we're reporting for the target population, the masses, are, therefore, valid.
Okay, next slide. The result is that, if you simply apply direct statistical procedures to the data and calculate confidence intervals for sensitivity and specificity with no correction, you get a confidence interval for sensitivity going from 95.6 to 100 percent and for specificity from 16 percent to 22.8 percent.
Applying the correction that corrects for the fact that we're here reporting on the best of seven target populations, the correction essentially widens the confidence intervals to make them valid, and the confidence interval for sensitivity now becomes 93.5 to 100 and for specificity 14.5 percent to 24.6 percent.
Thank you. I would like to now turn it back over to Dr. Parisky.
DR. PARISKY: I am remiss in the fact that I forgot to mention my financial disclosures, of which I have none. I am principal investigator. I do serve as a consultant to the company. I have no equity in CTI. I apologize for that.
I will now share with you the efficacy results from the original trial and then the subset and confirmatory trials.
These are the numbers which I will speak about. First, the original trial, looking at all subgroups.
Next. Of the 875 lesions studied, and these are again, I remind you, prospectively-chosen patients who by mammographic and clinical criteria were destined for biopsy and then studied, there was a 14 percent specificity, 96 of the 688 benign lesions. Lesions including descriptors such as mass, microcalcifications, architectural distortion were assigned a negative IR result.
Next. Interestingly, though, there was a 97 percent sensitivity. Of the 187 malignant lesions that were in this original population, including masses and calcifications, 180 were correctly assigned a positive IR result; seven received a falsely-assigned negative IR result.
I would like to draw the Panel's attention, and especially the mammographers and those who clinically practice in the field of breast treatment or breast surgery or breast diagnosis. So let's look at the false negatives.
The population were that of microcalcifications. That was the descriptor. The pathology was four DCISes, two DCISes with focal microinvasion and only one intraductal and infiltrating ductal carcinoma that was described as calcifications rather than mass. I was impressed by this, that there were no invasive malignancies described as mass in the false negative.
When we looked at masses alone within the original study group, this target population which was prospectively selected or targeted, and looking specifically at masses, we increased our specificity to 18 percent.
Next. We rendered 100 percent sensitivity. Of the 90 malignant masses within this population group, there was no false negative assigned.
The confirmatory study, as Dr. Callahan explained, the 78 patients or 78 masses in the 200-and-some-odd patient groups in this confirmatory set of 78 masses, specificity increased was 25 percent. Sensitivity was reduced. Of the 15 malignancies, 14 were correctly assigned a positive IR result and one was incorrectly assigned a negative result. Let's take a second to pause and take a look at what that cancer was.
That cancer, as pathologically evaluated, including basement membrane standing, was DCIS, a non-invasive malignancy presenting as a mass, which those of us who practice radiology know that occurs approximately 3 to 7 percent of the time, presentation of DCIS.
If we combine these two groups, the combination yielded a specificity of 19 percent and a sensitivity of 99 percent. Again, the one incorrectly or falsely-negative cancer was a non-invasive malignancy.
Next. This is a distribution of the lesions that were identified as malignant masses. We had both invasive and some non-invasive masses, reasonable distribution.
Go to the next one. Benign masses accounted, fibrocystic disease but not cysts. These are fibrocystic disease, primarily fibrosis and microcysts, fibroadenomas and a host of others, including fibrous mastopathy and the spectrum of benign etiologies that pathologists are confronted with.
Next. So we step back and take a look and see what is the intended population of this device. Again, I remind you that well over 1.3 million women are biopsied annually. A fair percentage of those are for the subgroup masses.
When a radiologist today is confronted with a mass observed on the mammogram, they are offered several tools, diagnostic mammography and ultrasound being primary. Much more expensive modalities such as Sestamibi, MRI, and PET scan have been proffered. There's published sensitivities and predictive values I don't believe approach what we have presented here.
This is a tool that would be used to provide information, as Dr. Hughes stated, that the radiologist looks at a mass. He or she performs an ultrasound, and they get a feeling and there's a perspective on whether or not to send this patient to biopsy. Those are based on anatomic criteria. So far, physiologic imaging, including Doppler ultrasound, have not provided a very good indicator or a good indicator for whether to proceed with biopsy or not.
This is a tool that appears to measure physiological changes like these related to blood flow in the region. I think the data supports use of this in masses. I hope the Panel and members of the audience are somewhat excited about the fact that we ‑‑ with the DCIS. I think that eventually we'll stratify some of that DCIS and look at the low grade and high grade, because we're now looking at the physiology. We're looking at the innerworkings of the breast in a rather inexpensive and non-invasive way.
We will use this to counsel our patients to say: You have a mass. Ultrasound tells me it's solid. Another examination tells me that, with a very high, near absolute negative predictive value, we could obviate the need for biopsy or we need to proceed to biopsy. It is a piece of reassurance to both the doctor and to the patient.
Mind you that the numbers I showed you, these specificity, that was superimposed on patients who were already mammographically-determined to proceed to biopsy.
Next. I think I have discussed intended population as masses. Continue.
In the schematic I'll reintroduce: The patient is seen. The mammogram says that it requires a further workup. The lesion is characterized as a mass. At that point in time, or in conjunction, in parallel to ultrasound, as part of the diagnostic workup, a new tool is now available which will allow the physician to determine, based on the results of positive or negative, whether or not to consider biopsy or to consider short-term followup, as is pretty standard in clinical practice today, without the additional physiological test.
Next. We avoided biopsy in 74 benign masses in a little over 380 patients. At risk was delaying one biopsy of a non-invasive malignant mass in over 100 malignancies.
I would like to thank the Panel for its consideration. Thank you.
I would like to now reintroduce Dr. Rust.
DR. RUST: What I would like to do is end the presentation of technical material as part of the sponsor presentation by putting the performance results that Dr. Parisky presented in perspective in terms of their effect on the health care system.
This 2x2 table simply takes the performance data that Dr. Parisky presented and puts it in the form of a 2x2 table where a true pathology is indicated in one dimension and the results of the IR test are indicated in the other direction.
Of course, in the trial lesions with a negative IR result did receive a biopsy because that is what is required by current practice. However, if you interpret these results, what this implies is that these 74 lesions for which there was a negative IR result would not have gone to biopsy and, therefore, 74 benign biopsies would have been prevented.
Again, this one lesion did receive a biopsy in the trial, but the implication is that this one lesion would have been a cancer for which diagnosis would have been delayed.
Now to take these performance results and put them in perspective in terms of effects on the health care system, what I'm going to do is basically extrapolate them up to the annual population to which this device could be applied. The way I am going to do that is to start with the 1.3 million biopsy figure that Dr. Parisky mentioned earlier and apply a 45.5 percent factor to determine the number of masses biopsied annually. That is where the 591,500 total mass biopsy figure comes from.
The 45.5 percent figure that I apply is simply what we observed in the clinical trial. All of the other numbers in the table simply follow by extrapolating up from the table on the previous slide to this 591,500 number.
The impact is that in practice we would expect to prevent approximately 90,000 benign mass biopsies annually in the U.S. if the device was applied to the entire population to which it is intended, and 1,207 malignant masses would have a delayed diagnosis.
Next slide, please. Now if you incorporate cost information into this picture to do a cost/benefit analysis, in the way of costs you first have to consider the cost of the IR procedure. We are, in fact, adding cost into the health care system. At a median level of the procedural cost of $225, we are, in fact, adding 591,500 new procedures into the system at a cost of $133 million. Another cost is that 1,207 cancers would have a delayed diagnosis.
Now the benefits of the device would be that approximately 90,000 benign biopsies would be prevented at an average cost of $3,000 per biopsy, resulting in $268 million of cost removed from the health care system, and the negative effects of 90,000 benign biopsies would also be mitigated.
So in that, 1,207 cancers would receive delayed diagnosis; the net cost savings to the health care system would be $135 million annually, and the negative effects of approximately 90,000 benign biopsies would be mitigated.
I should point out that the $225 figure that I used for the IR procedure is the midpoint of an anticipated range for the procedure of $150 to $300. If you apply that entire range, this net health care cost savings actually ranges between $90 million and $179 million, depending on the cost of the IR procedure.
So now I would like to turn this over to Lynn to basically summarize our presentation.
MR. SATTERTHWAITE: I'll apologize; hopefully, John Brenna is just not feeling well temporarily here. I'll try to finish up in his place.
Our proposed indication is that we are intended for use as an adjunct to mammography, to safely avoid biopsy of benign breast masses that would otherwise have gone to biopsy.
Next. We're recommended for all patients receiving a negative IR test result be similar to the recommendation for care of mass that is assigned a mammographic BIRADS category 3.
Next. We want to take a minute and just briefly go through the history of our interaction with the FDA. We have enjoyed a great relationship with the FDA people in counseling us, coaching us, reviewing, and so on.
Our original submission was done in June of 2001. We have talked about our movement to a subset of masses which we presented to the FDA in the form of an amendment in February of 2002, where they indicated to us there were things that we needed to do to confirm that data. We worked with the FDA personnel to come up with a plan to utilize the 275 patients for which the pathology was still vaulted to do a confirmatory study using that set of patients, which had the 78 masses we have talked about.
The FDA folks, reviewers, indicated that they were fine with our plan. We moved ahead to evaluate and analyze those patients and provided the results of that analysis in Amendment 5. That is what we call "Confirmatory Study Results" there.
We have had site audits by the Office of Compliance. We've had a sponsor audit by the Office of Compliance, and by all measures we think we successfully completed those audits.
We have been invited to a panel in July of this year, and then subsequent to that invitation, in working with the FDA, we mutually agreed to postpone the date for this meeting for administrative and logistical reasons.
Next. So somewhat in a conclusion manner here, the original study protocol was developed with the FDA. A confirmatory study plan to deal with the confirmation of the subset of masses was reviewed and approved by the FDA. That did target, formalized our targeting of masses or lesions with mass as a descriptor. Those results were combined with original study results. On review of those results, the FDA scheduled this Panel meeting.
In summary, let me just recap here. We have a device that is non-invasive. It's safe, painless, to be adjunctive to mammography x‑ray. It complements the anatomical view with the physiological view.
Our clinical study performance is an improvement over the biopsy decision. Nineteen percent specificity we think is significant. We believe that it has the potential to reduce health care costs. We believe that we have demonstrated that we are safe and effective medical device for the proposed indication for use.
I ask that you recommend approval of our device as you complete your review. Thank you very much.
CHAIRMAN MEHTA: We would like to thank the sponsor for their presentation. At this point, if there are questions from the Panel specifically in terms of clarifications only ‑‑ we'll have discussion questions later on in the afternoon ‑‑ but if there are clarification questions from any of the Panel members, this would be a good time to ask the sponsor for that.
DR. TOLEDANO: The first one is, looking at your table, I am reminded that in many mammographic procedures women with large breasts require multiple images. How well do you accommodate women with large breasts on your table?
DR. CALLAHAN: I'm not aware in our clinical trial that we had any sort of dropout because of non-accommodation of size. Perhaps Dr. Hughes or Dr. Parisky could comment. So I am not aware that that has been a problem.
DR. TOLEDANO: Okay.
CHAIRMAN MEHTA: I do have a question. I was somewhat confused about the specific indication that is being sought in terms of the BIRADS category. In the earlier presentation by Dr. Hughes, he indicated that, using this nice flow diagram, what one would do is perform a physical exam, mammography, ultrasound if necessary. If there's a highly-suspicious lesion ‑‑ i.e., a BIRADS category, say for example, of 5 ‑‑ one would still go ahead and do the IR imaging. Then if the IR imaging is negative, you would still go and do the biopsy, because, obviously, the clinical information suggested that this was a highly-suspicious lesion.
But I think one of the final cites suggested that the BIRADS category would be restricted to 3. Can you clarify for us whether there is a specific BIRADS category you are asking for in this or not?
DR. HUGHES: I believe it's for BIRADS 4 and 5. I believe what the slide was trying to say was that we might be able to take a 4 or 5, which is where the test would be done, and then downgrade it to a 3; whereas, rather than or instead of doing a biopsy on a 4 or 5, we would call it a 3 and do a followup. So we are not looking for BIRADS 3 at all. BIRADS 4 or 5. Is that accurate, Lynn?
DR. CALLAHAN: Well, I would say that that's fairly accurate, but we're not restricting it to a BIRADS categorization. The decision is, if this is a patient that the physician feels a biopsy might be warranted, then the IR procedure would be an alternative or an adjunctive test.
For example, we know that there are BIRADS 3s that the ACR recommendation is six-month followup, but for reasons, either personal reasons of the patient or the physician, oftentimes these patients do not desire for that six-month followup. So in that sort of situation this procedure would be appropriate, we believe.
CHAIRMAN MEHTA: I'm sorry for my confusion, but three or four slides into the last presentation there was a statement about a BIRADS category. Can you put that slide back up again? There you go. Can you clarify this slide for us?
DR. CALLAHAN: This statement refers to what would be the labeling indication for a negative IR result; that is, a negative IR result, the recommendation would be that the followup be six-month followup, similar to a BIRADS 3. So I am sorry for that confusion.
DR. CONANT: I have a quick question, I hope quick.
I'm not sure I completely understand the flow, the clinical flow, accrual. The level of suspicion that then would prompt one to biopsy or not is based purely on the mammography or the combination of the mammography and ultrasounds?
Because, for example, one case that was shown, I think the clinical case of two views of the breast, looked like an asymmetric density, perhaps not a mass. Then if the ultrasound was negative ‑‑ I am wondering how the ultrasound plays into all of that and whether the determination of mass is made by ultrasound or mammography.
DR. CALLAHAN: For this study the determination of the descriptor was supposed to be based solely on mammography. It was started, was initiated in 1997. Ultrasound was not part of the study protocol. However, the fact is that we know that standard clinical practice was utilized.
So that if women had something that appeared to be a mass by mammography and had been recommended for biopsy, they would have been enrolled in our study, but then ultrasound may have been performed; it was found to be a fluid-filled cyst; the biopsy was cancelled. So that accounted for some of our dropout.
So the study protocol did not specifically collect the contribution or the impact of ultrasound.
DR. CONANT: So there would be cases where a mass was considered on mammography, but the ultrasound could have been negative?
DR. PARISKY: Correct.
DR. CONANT: Okay, and I'm just curious.
DR. CALLAHAN: I should have asked our clinicians.
DR. CONANT: No, I was asking about the clinical flow of things. It was confusing to me where the level of suspicion came from and how ultrasound contributed to that, because there are frequently times ‑‑ the example that was shown, for example, that looked like not truly a mass by BIRADS, but an asymmetric density, and if followed by ultrasound that was negative, the management might be quite different if you separate ultrasound from mammography.
So I am wondering if the trial is driven purely by the mammography ‑‑
DR. PARISKY: The trial was driven ‑‑ you know, it so happened that the real explosion in ultrasound was in the last five years. When the trial protocol was written and presented to the FDA, it was based on mammographic findings.
So if a lesion was suspect mammographically, the patient, at least at my center, and I think most of the centers, was then enrolled for consideration into this trial. In some centers, in talking to my colleagues, any ultrasoundographically-solid mass is biopsied where in other centers ultrasound is used to try to attempt to characterize based on work that you know, Stavros and such, to try to obviate the need for biopsy.
Ultrasound I believe was at least, we know from reviewing the medical charts, was employed in a fair percentage of these patients, but was not the determinant to proceed to biopsy. At that time clinical judgment by the physician in attendance determined.
It should be noted that a number of these patients who were enrolled initially was with a suspect mass required for the workup, and part of the large withdrawal pool was because they were enrolled and, subsequently, by ultrasound to be shown to be cysts.
I think the FDA Panel is familiar with a process or subselection out like that, considering that that was presented to you during one of the ultrasound panels in which enrollment was made, and then a great majority of the cases were dropped from consideration because cysts were discovered. So the patients were enrolled based on mammographic findings.
DR. CONANT: But, clinically, at that point one wouldn't be recommending a biopsy for the inclusion criteria.
DR. PARISKY: A patient would be considered for biopsy if they had an abnormality on mammography or referred, in my instance which is a referral center, would be referred for consideration for a biopsy.
DR. CONANT: Okay, that to me would be like a category zero, that ultrasound was needed. That's okay; we can talk later.
DR. PARISKY: Yes, that's arguable as to how one addresses that.
DR. CONANT: And were there asymmetric densities? That's quite a common category in BIRADS that I didn't see included.
DR. PARISKY: Asymmetric density, do we have the numbers on asymmetric?
DR. CONANT: Not quite a mass, for those of you who don't use that term, it's ‑‑
DR. PARISKY: You know, again, it's open to personal conjecture that an asymmetric density seen in one view ‑‑ we did have cases that, if it could be seen, if the density could be seen in two views, some physicians might categorize that as a mass, having been able to describe it in two views.
We were not particularly rigid in terms of specific criteria for determination of mass and left it to the individual investigators.
DR. CONANT: For example, two-thirds of the margin's convex ‑‑
DR. PARISKY: No, that was ‑‑
DR. CONANT: -- on two views equals a mass versus ‑‑
DR. PARISKY: That was not, those rigorous criteria were not applied.
DR. CONANT: And the BIRADS characterization was done by just one radiologist from the mammogram?
DR. PARISKY: Again, the introduction and mandate of utilization of BIRADS occurred in the midst of the trial.
DR. CONANT: I think 1997, is that right, or 1996?
DR. PARISKY: No.
DR. CONANT: No?
DR. PARISKY: The rule, I think, came in 1999?
DR. CONANT: I'm not sure.
DR. PARISKY: April 1999 was the final rule. So by mandate, that was not to be included in each of the reports until 1999, so through half of the reports.
So the difficulty with BIRADS, some reports reviewed all of them; some reports included BIRADS; some negligently didn't include BIRADS even after the mandate date. So we attempted to have the physicians in the separate pool try to develop a level of suspicion, but, again, as Dr. Callahan pointed out, they didn't have access to prior films; they didn't have access to some of the additional imaging or the films that were provided were just that of the lesion and the breast in question.
So that proved to be very cumbersome, which is why, again, we went back to what we thought was the gold standard, acknowledging the fact that there is a 3 to 5 percent false negative even with the gold standard.
DR. CONANT: The gold standard of?
DR. PARISKY: Being biopsy.
DR. CONANT: Oh, okay. Not variability within readers ‑‑
DR. PARISKY: No.
DR. CONANT: -- which is quite large?
DR. PARISKY: Well, more so in ‑‑
DR. CONANT: Thirty percent.
DR. PARISKY: More so in DCIS than invasive, and I think that that's a consideration, something, hopefully, you dwell upon, given the fact that we looked, you know, at the low-grade DCISes that were false negatives. But in terms of invasive cancer, I think the variability is much less than it is in DCIS.
I apologize, I was acting in professional capacity just now.
CHAIRMAN MEHTA: Dr. Hooley, any questions?
DR. HOOLEY: I have a question in line with Dr. Conant's. The presentation emphasized mammographically-detected masses, but I'm unclear if the study included masses that were detected only on clinical breast exam.
DR. PARISKY: Yes, it did.
DR. CALLAHAN: The original study protocol was intended to analyze both patients that were enrolled based on clinical exam alone, say, you know, palpable lesions that weren't identified, that weren't visible mammographically. That was the original study protocol.
But when it became apparent that really the utilization of this device requires the evaluator, the physician, to localize on the IR image, and because we had this independent panel of reviewers that did not have access to the patient, if it was a palpable lesion that wasn't visible mammographically, then they couldn't perform a localization and assessment.
So, in the end, our efficacy claim would be that it's a mammographically-apparent mass that can be localized. Dr. Hughes may comment on, and maybe perhaps you want to defer this discussion until later, but Dr. Hughes can comment on the protocol at Mass. General, where I believe they are enrolling patients with palpable lesions, because the enrolling physician can go ahead and localize based on the palpable lesion location that they are aware of without having mammographic evidence. But our clinical trial could not assure that, because our evaluators were blind and independent.
DR. HOOLEY: I have one more question.
CHAIRMAN MEHTA: Go ahead.
DR. HOOLEY: The cost that you cited for the biopsies of $3,000 per biopsy, I thought that was in a very high range. Yet, you used, your range for your procedure, you used like $220 or something like that, giving a medium range of between $150 and $300, but $3,000 for a breast biopsy seems excessive.
DR. PARISKY: In defense of that, that number is taken from a compilation of both core biopsy and surgical biopsy. Since the national average is approximately 50/50, while it may not be in academic centers and certain centers of excellence, of a number of mammographically-apparent lesions still undergo needle localization, and we'll probably continue to do so, which has a much higher cost. That's how that figure was arrived at.
And it's not particularly different, accounting for inflation and rise in health care costs, from the numbers quoted by Jim Brenner and Ed Sickles from their article about seven years ago, which I believe was $1,900 or $1,700. So taking that into account, I think $3,000 is a fair number.
DR. CONANT: How about the cost of the followup for those IR-negative lesions, the six-month followups and the cost of those women who were imaged and uninterpretable?
DR. PARISKY: Well, the uninterpretable I will address that first. There was a learning curve for both the machine and I think for the technologists in terms of uninterpretable. Steps have been taken; again, as a practicing clinician, I want to minimize the number of repeats, and so forth. So I believe that that has been mitigated.
In terms of followup, patients who you would, in the absence of this technology, would follow up as well, based on giving them a BIRADS 3, we didn't calculate that score, and that would be based on physician preference.
So you are talking about additional diagnostic mammogram or ‑‑
DR. CONANT: Would you recommend this as a followup as well?
DR. PARISKY: This would have a followup as well and the appropriate clinical judgment of the clinician. So there may be an increased cost, a slight increase in cost, in followup, but we didn't subtract out or we didn't calculate, if you conventionally sent patients to BIRADS 3, what the costs are with that.
DR. CONANT: Sure.
CHAIRMAN MEHTA: Go ahead, Geoff.
DR. IBBOTT: I have a couple of technical questions.
DR. PARISKY: Oh, good. I'll sit down.
DR. IBBOTT: The description of the use of the images focused on the central en face image, outlining the breast and the region of interest. How are the peripheral six images used or are they used?
MR. SATTERTHWAITE: You'll note from those images that there are anatomical features that are apparent. In the training that CTI does of both the technologist and the physician, we train them on how to localize or identify the breast by looking at outlines of the breast in the side mirrors and drawing lines across the en face image in order to clearly identify exactly where breast tissue is. So those are used primarily to identify nipple location and to outline the breast.
DR. IBBOTT: I see. What sort of calibration and quality assurance procedures are required? If this device were to be approved, how would those procedures be implemented in a routine setting?
MR. SATTERTHWAITE: The device that will be marketed has a calibration that takes place before every imaging session. There's a black body device in there that allows us to calibrate each time.
There is regular maintenance on the camera that is associated just with the camera. Every year there's things that happen with the camera. There are other things that we do to assure quality, like check the mirrors and a number of things that are in that manual that has been provided the FDA.
DR. IBBOTT: And the region of interest was expanded by the software to something approximately one-twelfth of the area, I guess, of the breast tissue. How was that one-twelfth value selected? Because it seems like choosing a smaller area that's limited to the mass would make the device more sensitive.
MR. SATTERTHWAITE: I'll ask Dr. Rust to come and back me up here on that one. I mean, obviously, there was some work that we did, but we recognize the fact that to localize between modalities there's some variance just because of the presentation of breast tissue in those various modalities. So we saw the need to improve that localization by looking for a stronger IR signal in the vicinity.
So the actual selection of the one-twelfth, I will let Dr. Rust address. I believe that was your primary question.
DR. RUST: Maybe I will start by clarifying how that procedure works. The area that is one-twelfth of the breast region is the size of the search region for looking for a point of both higher IR signal and higher contrast with surrounding pixels.
Once that location is determined, what is actually used for determination of a negative or a positive outcome is actually a circle of radius five pixels, much smaller than that one-twelfth region. So the one-twelfth area is a search region. The decision is made on a much smaller focused region of interest.
DR. IBBOTT: Thank you.
CHAIRMAN MEHTA: Dr. Rust, while you're up there, I did have one question for you.
DR. RUST: Sure.
CHAIRMAN MEHTA: You were very clear in outlining to us that the mass subset was prospectively defined and the planned assessment was included in the protocol.
DR. RUST: Yes.
CHAIRMAN MEHTA: I do have a question that goes to the general concept of prospectively-planned statistical analysis and assessment in the protocol and have several subsections to the question.
Did the protocol state that this assessment for masses would be done after the biopsy results were available or before? Because it appears to me that this assessment was done after the biopsy results became available.
Secondly, did the protocol prospectively allow for an expansion in patient pool from 600 to 2,400, a fourfold increase? Was that prospectively defined in the protocol? Did the protocol prospectively allow the inclusion of 275 patients that would not be evaluated and would be unblinded at a later date to be evaluated?
These are sort of confusing issues and aspects in the conduct of the protocol, unclear whether this was prospectively planned for and, if so, how did you convince IRBs about the justification of such a prospective plan?
DR. RUST: To your first point, the protocol stated that performance analysis would be done by lesion type, which implicitly requires that it be done after pathology is unblinded. Otherwise, it is impossible to look at lesion type ‑‑
CHAIRMAN MEHTA: I don't think a lesion typed labeled "mass" is a pathologic diagnosis.
DR. RUST: No, it's not, but what are you analyzing by lesion type? You are analyzing performance results, and there are no performance results to analyze prior to unblinding of the pathology. So implicitly I believe that those analyses were prospectively planned to be done after unblinding of pathology. Otherwise, there is nothing to analyze.
Your second question was the expansion of the ‑‑
CHAIRMAN MEHTA: Six hundred to 2,400.
DR. RUST: I do not believe that the expansion was prospectively planned for in the protocol.
DR. CALLAHAN: The study protocols were submitted independently to each IRB, and each one stated the 600 enrollment definition. So each protocol was independently under the each institution IRB with a sample size of 600.
It wasn't clearly stated, a total enrollment goal, in the protocol, I'll say that.
CHAIRMAN MEHTA: I'm sorry, does that mean 600 patients from each site?
DR. CALLAHAN: Each site was independently, yes, provided the study protocol with an enrollment goal of 600 patients.
CHAIRMAN MEHTA: And you had six sites. That would require ‑‑
DR. CALLAHAN: Well, we had actually five, six enrolling centers. Dr. Parisky's site at USC involved LA County initially and then Norris Cancer Center. So it would have been 600 at that one site, with two enrolling centers, and then our four other sites.
CHAIRMAN MEHTA: So that would have required 3,000 patients, 600 per five, per site, times five. Was the protocol prospectively designed to stop at 2,400?
DR. CALLAHAN: I cannot answer that, as I was not there at the initiation of these study protocols that began 1998. I don't have that information. I don't know that any member of our team has that information.
There was no, as far as I know, there was no prospectively-defined written document saying that the total enrollment goal would be a specific number.
CHAIRMAN MEHTA: Will you, at a later point in the meeting, be able to provide us site-specific enrollment data?
DR. CALLAHAN: We have that on one of our slides ‑‑
CHAIRMAN MEHTA: Okay.
DR. CALLAHAN: -- if you want to go back. I can state that they range from the lowest enrolling site was 170, and our highest enrolling site was a little over 800.
DR. GENANT: I have a question.
CHAIRMAN MEHTA: Go ahead.
DR. RUST: There was a third part to your question.
CHAIRMAN MEHTA: Yes.
DR. RUST: Would you like me to address that?
CHAIRMAN MEHTA: If you could.
DR. RUST: Okay. I can simply tell you what happened in terms of how the 275 patients were actually identified. There was a schedule put in place to complete analyses, to go into Module 5 for submission to the FDA. In order to support that schedule, a freeze was placed on the clinical database in terms of patients enrolled before a particular date. Then that frozen database was taken forward through all the analysis procedures, results analyzed, put into the original Module 5 submission, and submitted.
Centers continued, three of the centers continued to enroll patients after that database freeze, and that was the source of the additional 275 patients that ultimately were analyzed in the confirmatory study.
CHAIRMAN MEHTA: Okay. Go ahead.
DR. GENANT: Yes, I have questions with regard to the reproducibility, No. 1, of the machine itself. Do you have data that would indicate that, if a woman was measured on two occasions, that there would be comparability of results?
And, secondly, you have pooled the readers' results, and I wonder, do you have information on the individual readers and how they related to each other, what the reproducibility was, and how would that extrapolate to clinical practice, where one would have a single reader?
MR. SATTERTHWAITE: I'll address your first question, which is the reproducibility. We have not imaged a single patient in two different situations. We have not addressed that.
DR. RUST: In terms of pooling the information from multiple evaluators, let me describe exactly how we do that. Our analyses were all lesion-based, and we took the three evaluations of a particular lesion and included all three evaluations in the analysis, giving each one a weight of one-third in the analysis, taking credit for only one lesion with that analysis.
So we did not pool the three evaluators into a single positive/negative score which was then analyzed. We did include each individual evaluation, giving it a weight which caused each lesion to get a total weight of one in the analysis. So I wanted to clarify that.
So, to answer your question, do we have data on the individual evaluators' scores, yes, we do, and it was incorporated into the analysis in the fashion that I just described.
DR. GENANT: Can you share with us what that inter-reader variability was?
DR. RUST: In fact, we did an analysis to answer a question that the FDA posed to us on inter-reader variability in the IOS score that leads to a positive/negative test result. I'm going to have to back up here and describe what the IOS score is.
It is an index on a scale of zero to a hundred, and small values are associated with less-suspicious lesion, large values with more-suspicious lesions. We compare that to a threshold to get the positive and negative test result.
The result of our analysis on inter-reader variability was that the inter-reader standard deviation was approximately four units, which in this context appeared to be rather small, both to us and the FDA, I believe.
Does that address your question?
DR. CONANT: May I ask a question along that line? That's for the IR part. I am very interested, I think I mentioned before, about the inter-reader variability, about what brought the patient to the category of biopsy necessary. Do you have that information?
DR. RUST: Okay, I don't believe I can address that question.
DR. CONANT: Because there's at least a 30 percent variability, unfortunately, I think, in most practices.
Another question similar to that is this threshold idea. I read about the threshold and how it was chosen. I am very curious about the distribution of the IR readings about threshold and how they correlated with the LOS, the level of suspicion, in terms of the radiologist.
It sounds like a great threshold. I mean, it only missed one. But I am wondering how close things really were in there and whether they were all very borderline, and there's a whole group of gray zone cases ‑‑
DR. RUST: Right, we ‑‑
DR. CONANT: -- because those are very difficult clinically.
DR. RUST: We did, in fact, do an analysis of sensitive ‑‑ "sensitive" may be a bad word ‑‑ sensitivity of the performance parameters to the threshold selection and basically summarized that in the form of a plot showing the changes in sensitivity and specificity near that threshold.
Does that address your question at all?
DR. CONANT: Well, actually, I would like to see the raw data and how it fell, not just the sensitivity. I have questions with the sensitivity because there are so many excluded patients. So I would rather look at the raw data.
I mean, it's an interesting ‑‑ I was just trying to crunch the numbers here, but Dr. Parisky said at the beginning about 50 percent of the cases that go to biopsy are masses. Your population, I think, if I did the numbers right, was around 20 percent. So sensitivity I am not sure reflects the population, but actually raw data around that threshold level ‑‑
DR. RUST: I just want to get a clarification on your question. You think that only 20 percent of our study population presented with masses?
DR. CONANT: No, but it looked like, when you came down to that ‑‑ maybe I did my numbers wrong. Maybe we can talk about that later, but it's not of the overall, but when you take out all those exclusions, when you crunch down to the numbers that then actually are eligible and then get a ‑‑ it's on page ‑‑
DR. RUST: Actually, I believe that -‑
DR. CONANT: -- page 12.
DR. RUST: I believe that that ratio in our study, and correct me if I'm wrong, was 412 out of 875.
DR. TOLEDANO: What would that have looked like if you excluded the masses up at the top? So you excluded for those; you excluded for that. And I know that this is the way that this study actually happened, but I think what Dr. Conant is interested in, and what I'm interested in, is, what if you excluded, what if you subdivided masses from non-masses up at the top?
DR. RUST: We do have the data to do such an analysis. I can't guess at what the answer would be.
DR. CONANT: But then to look at the raw numbers of the IR without choosing ‑‑ I mean, I'm sure you have a very good reason for your threshold, but it's just, you know, I would love to know.
DR. PARISKY: You're interested in the IOS related to malignancy, malignancy distribution or ‑‑
DR. CONANT: No, no. I'm really interested in ‑‑ that's one question, yes, but ‑‑
DR. TOLEDANO: Aren't you asking for the distribution of IOS in the malignants ‑‑
DR. CONANT: Yes.
DR. TOLEDANO: -- and the distribution of IOS in the benign?
DR. CONANT: Yes, yes.
DR. RUST: And that data could certainly be made available, yes.
DR. PARISKY: And we have looked at it.
DR. CONANT: Great.
CHAIRMAN MEHTA: I think Prabhakar had a question.
DR. TRIPURANENI: In the confirmatory study you had 275 patients, out of which you had 78 masses. The 275 patients, is it before excluding certain groups of patients or after?
DR. RUST: It is before. That is the total number of patients before any cases were withdrawn for the various reasons that Dr. Callahan indicated would cause patients to be withdrawn.
DR. TRIPURANENI: So if you take the 78 masses out of that 275 patients, the percent of masses in those patients comes to approximately 30 percent or so. The same number into the original study comes to approximately 432 patients. Is it the equivalent number within the confirmatory study with the original study?
DR. RUST: I'm not clear on your question. I'm sorry, I didn't follow all ‑‑
DR. TRIPURANENI: Two hundred and seventy-five patients now.
DR. RUST: Yes.
DR. TRIPURANENI: What is the counterpart number in the original study? Is it 432 or is it 840?
DR. RUST: Eight hundred and seventy-five is after exclusion of stuff.
DR. TOLEDANO: I could do it.
DR. RUST: It's on Dr. Callahan's slide. I believe it is where we say 1,660 masses and ‑‑
DR. TOLEDANO: Okay, do you want me to do it? Okay, 2,406 patients goes down to 1,432 in the FDA dataset, when you exclude all the unvaulted. Of the 1,432, you have approximately 800 who were in the original FDA evaluable dataset, of which 432 had masses. In the 275 you had 171 who were evaluable, of which 69 had masses. Did I get that right?
DR. RUST: I believe the comparable number is 1,432.
DR. TRIPURANENI: Then in that case, the percent of masses from the original study to the confirmatory study actually is almost double the number. Fourteen percent of the original values actually had the masses whereas in the confirmatory study it is almost 29 percent. Were there significant enrollment differences between the original study to the confirmatory study?
DR. RUST: Okay, I'm thinking the comparable ratio here is 412 to 1,432 versus 78 to 275, and are those really that different, I guess it the question I'm asking.
DR. CONANT: I think they both are in the 20 percent range, low twenties.
DR. RUST: I'm calculating on my feet here, but they don't seem all that different.
DR. CONANT: I think they're similar ‑‑
DR. RUST: Uh-huh.
DR. CONANT: -- but the question is, how does that deviate from what goes to biopsy in reality, which is about 50 percent masses? I'm not sure it's that high, but I don't have that number in front of me. I think it's lower.
DR. CALLAHAN: I think what we need to do to do this calculation is look at, like Dr. Toledano suggested, look at the number of masses in each of these groups prior to any exclusion to see what the percentages were and to see if there's differences, and that's something we can certainly do over the break.
CHAIRMAN MEHTA: I think we'll go to it at this time in order to stay on time. Let's go ahead and take a five- to ten-minute bathroom break. We'll reconvene at 11:00, so that the FDA can do its presentations. We will have another opportunity for questions later on in the afternoon.
(Whereupon, the foregoing matter went off the record at 10:54 a.m. and went back on the record at 11:05 a.m.)
CHAIRMAN MEHTA: If could start having everyone in their seats, the first speaker will be Jack Monahan, the leader for the PMA.
MR. MONAHAN: Good morning. I would like to take this opportunity to thank the Chair and the other members of the Panel for taking time from their busy schedules to help us on the deliberations associated with this PMA.
As you know, we have had the sponsor present, and I would like to just give a few introductory remarks prior to moving into the clinical and statistical discussion.
This particular PMA is what we refer to as a modular submission, and we received the first module back in 1999. I was the primary reviewer for that module, which contained the preliminary information related to the device.
Module 2 was subsequently submitted, and that contained primarily the software material, and that was reviewed by Joseph Jorgens in our Office of Science and Technology.
Module 3 consisted of all of the manufacturing information, and that was reviewed in the Office of Compliance by Xuan Vo.
The fourth module was the engineering material associated with the product, and that was reviewed by Jim Seiler, who is in the Division.
The PMA was submitted and contained principally the clinical data, the protocol, and the labeling material for this product. This is what we are dealing with today.
I am the lead reviewer on the PMA. The clinical reviewer is Dr. William Sacks, who you will be hearing from in a moment. The statistical reviewer is Harry Bushar, and we also had a bioresearch monitoring review of this submission, and that was done by Kevin Hopson.
What I would like to do is very briefly go over the proposed indication for use. I have abbreviated a little bit on the screen, but you will note that I have highlighted some of the words in there because I believe that these are important aspects of this proposed indication for use.
The CTI BCS 2100 is a dynamic, computerized, infrared-based imaging, image acquisition and analysis system. It is intended for use as an adjunct to mammography, to safely avoid biopsies of benign breast masses that would otherwise have gone to biopsy.
Physicians should not base a decision for patient care solely on the results of testing with this device, but rather on results of this test in combination with all other findings and risk factors associated with a specific patient.
The CTI BCS 2100 provides additional information to guide a breast biopsy recommendation. Because demonstration of device effectiveness was limited to breast lesions that included mass as a lesion descriptor, use of the CTI BCS 2100 should be limited to the evaluation of breast lesions that include mass as a lesion descriptor. The presence of another lesion descriptor does not contraindicate use of the CTI BCS 2100 if the lesion is also described as a mass.
It is recommended that the appropriate recommendation for care of all patients receiving a negative IR test result be similar to the recommendation for care of a mass that is assigned a mammographic category of 3 or a BIRADS 3. That is short-interval followup is recommended in order to establish the stability of the findings.
This is the proposed indication for use, and I would now like to turn this over to Dr. William Sacks to discuss the clinical study.
DR. SACKS: Just for those of you who don't know me, I'm a radiologist and used to be a physicist. So I have some familiarity with numbers as well.
I want to stress a number of aspects of the device that I will enlarge on as I go on. First of all, this is a new type of thermographic device. Secondly, it's an adjunct to mammography. Thirdly, it renders a positive or negative result, as you have seen, and that is based, as the company has explained, on an index-of-suspicion score.
It is intended for women on their way to biopsy only. It is, furthermore, intended for women on their way to biopsy who have mammographic masses only, and the intended use is to save biopsies of lesions that turn out to be benign.
The points I am going to cover are what the BCS is and is not intended to do; how the device does it. Then I'm going to have Dr. Bushar give the clinical trial results, and I will come back and make some assessment of those results and, finally, a few labeling issues that we would like the Panel to consider.
Before I start on what the BCS is and is not intended to do, I want to make a clear distinction in the minds of the Panel between a device and its intended use. It is very important to keep that in mind as we go on. One and the same device can have a number of different intended uses, and, indeed, for any given intended use, there may be one or more devices that will satisfy that use.
We will be talking predominantly about the clinical trial. A clinical trial is always designed based to demonstrate that the particular chosen intended use of the device is safe and effective. So it will always be an underlying issue here that we are talking about the company's intended use for this device.
Now in case any of you come here with any baggage or prejudice from the past about breast thermography, I want to make a clean break with that. Historically, it has not had the sensitivity and specificity to either replace screening mammography or to be a complementary screening test; that is, it hasn't had the sensitivity or specificity to be a screening test.
However, the BCS is a new type, as I said, of thermographic device, and it is new in two ways. One, it uses a new application of technology which lies predominantly in the cooling of the breast with the fan, and that enlarges the temperature contrast between malignant tissue and benign tissue, it is thought. The reason for that is that the benign tissue will cool, whereas the malignant is fed by angiogenesis and has a higher metabolic rate, will not cool as fast.
So that if you were to track the time course, as this device does, over the cooling, the contrast between malignant and benign tissue would be enhanced. So that is one aspect that is new.
The second one is that it targets a different group of women from that which conventional thermography in the eighties tried to target. In the eighties the attempt was to make this a screening device that, hopefully, would replace mammography or at least work alongside it for all women screened. This device, however, targets, as you have heard, and as I myself have mentioned, a subgroup of screened women.
I've already mentioned that, that the cooling is the issue here, and the different group of women is the ones whose screening tests, mammography and/or palpation, along with other factors, indicate a need for biopsy.
So it is not intended as a screening device ‑‑ that's very important ‑‑ and it's neither, therefore, a replacement nor a complement to screening mammography, but rather as an adjunct, and, in particular, an adjunct to mammography, not to clinical palpation, as was hoped at the time of the original protocol, but to mammography, and, indeed, only for women on their way to biopsy and only for those with mammographic masses.
Let me say a few words about the difference between a complementary test and an adjunctive test. These are somewhat confusing concepts, and adjunctive is itself probably the most confusing.
Let me say something about complementary tests to begin with. A test that's complementary to a screening test is used on all persons screened; that is, it is itself a screening test, and, therefore, its results may by themselves determine the next step in clinical management.
Complementary screening tests, therefore, are on equal footing with each other. One easy example is screening mammography and clinical examination. Women over 40 get annually, should get annually, a clinical palpation as well as screening mammography. If either one of these shows need for a biopsy, such as a palpable mass, even if it's invisible on the mammogram, then the clinical examination will be the thing that will decide the woman's clinical management.
If, on the other hand, there is nothing to palpate, but the mammogram shows a suspicious finding, the woman will still go on and get further workup. So these two exams are complementary to each other because either one by itself can determine the next step.
Adjuncts, on the other hand, are subordinate to the index screening test; that is, the screening test to which they are adjuncts. They can be subordinate in one of two ways or both.
They are either not used on all the persons screened, and I will come back to the examples in a second, or if they are used on all the persons screened, their results do not by themselves determine the next step in clinical management. Let me give you examples of each of these.
The BCS itself is the first type. It is not used on all persons screened. However, on those on whom it is used its results do generally by themselves, or will, or it is intended that, its results will determine the next step in clinical management; namely, whether or not the woman goes on to biopsy or not.
In a sense, it is inherent in the nature of the device, which is a black box that pops out a number, you can't make a judgment, the company has stressed. You have no ‑‑ it is not a visual issue of the image itself; the device gives you a number. So insofar as it does, you are forced to listen to the device.
Now you are not forced to do what the device tells you, but if you do, it is not a ‑‑ let me put it this way: If you have a woman that you really think needs a biopsy and you subject her to this test, and this test gives you a negative result and you decide to send her to biopsy anyway, my suggestion is, don't do the test. There's no point in having done it. You could have foreseen that ahead of time.
Another example of an adjunct of this type is the ultrasound of solid breast masses, as it is being done by a growing number of people, Stavros and others, and so on. It is used only on a subgroup of those who go through mammography or palpation, and it may by itself determine whether the woman needs a biopsy or not, if you happen to use ultrasound in that fashion for solid masses. Extra-mammographic views are another example, and so on, and even biopsy itself.
Now the other type, the results by themselves don't determine the next step in clinical management, even if they are used on all the women screened ‑‑ and a perfect example of that is a mammography computer-assisted diagnostic system. It is used on everybody, but it is the radiologist who decides, after it points out places, "Have you looked here, here, and here," whether or not to do something about that.
So those two types of subordination, either one of those or both will throw a device into an adjunctive status.
The intended use ‑‑ and I stress it again ‑‑ the intended use, as currently intended, of the BCS is to confirm the need for biopsy or change a woman's clinical management. If it is changed, it is to be changed from biopsy only to short-term followup, not to come back in a year for the next screen.
Thereby, it can only decrease the number of biopsies, and in statistical language that means it can only increase the specificity. It cannot increase the detection of cancers as it is currently intended to be used and as the trial was conducted. That is, it cannot increase sensitivity.
An advantage of the particular selection of target population ‑‑ that is, only women on their way to biopsy ‑‑ is that device false positives, and I define a device false positive as a woman who has a mass that is benign but gets a BCS-positive result. It's that simple. Such device false positives have no impact on clinical management. After all, these are women who were on their way to biopsy anyway, and if you get this positive result, you will simply go on and do what was recommended in the first place. Therefore, there's no impact upon clinical management.
We would like the Panel, during the discussion this afternoon, to consider an issue as to whether they have any concern over the potential psychological impact of a positive mammogram followed by a false positive BCS result, that is, on a woman who does not, in fact, have cancer. Let me just say a word about that.
If I were to get a mammogram based on which a recommendation that I get a biopsy was made, my main fear would not be of a biopsy procedure; it would be that I had cancer. Now I'm offered a test that says we can do one other test that may obviate the need for a biopsy, and if I get a positive result from that test, now I'm a little more convinced that I must have cancer, even though ‑‑ and it can be explained to women, and this is what we want you to discuss, whether labeling or anything like that needs to be addressed ‑‑ is your chance of having cancer zoomed from about 20 to 23 percent. I mean, you're still overwhelmingly not likely to have cancer, even though both tests were positive, but this is a subject that one of our questions will be designed to ask you to discuss.
Is the BCS an alternative to biopsy? In 80 to 85 percent of women who obtain the test, it will end up being in addition to biopsy. Only 15 to 20 percent of such women will end up not getting a biopsy in addition, at least not immediately.
How does the BCS do this? This is somewhat repetitious, but it is good to hear a little redundancy when so much information is being thrown at you.
It calculates an index-of-suspicion score for the region of interest selected by the radiologist, and the radiologist bases that selection on the mammographic location of the mass. The device then compares whatever that number is, which ranges from zero to a hundred, with the determined threshold that was determined by the company during their training set of the first 700 ‑‑ a slight detail there, but Dr. Bushar will talk about that ‑‑ but, roughly, the first 700 out of the 2,407 patients, did some training and picked the threshold of 20.59 so as to keep a very high sensitivity.
If the IOS score for this woman falls below that threshold, the device will read negative results, and that means change her path to biopsy to short-term followup. That is very similar to the BIRADS 3 category. If the IOS score is at or above the threshold, the device output will read positive, and that means just continue with the plan to biopsy.
There is a side effect, and that is that some cancer diagnoses may be delayed, and we can talk about that.
Now I would like to have Dr. Bushar give you some of the statistics here.
DR. BUSHAR: Thank you very much, Bill.
Good morning. My name is Harry Bushar. I'm the statistician who reviewed this PMA on the computerized thermal imaging Breast Cancer System 2100. I will be doing the statistical presentation. An outline of what I will be presenting is I want to discuss the clinical study protocol, including objective design, population, demographics, and evaluation, both effectiveness and safety, and then get into the actual PMA clinical study and what was done there in terms of effectiveness and safety.
And, finally, continue to move on to Amendment 4, which again brings out the effectiveness in the clinical study, and Amendment 5. Both of these amendments were caused by letters that the FDA sent, deficiency letters, to the company. These are their responses to those. And, finally, Amendment 7, which gives an adjustment to the effectiveness, and I want to make some statistical conclusions.
I would like to make it clear that what I did is I reviewed the sponsor's analysis. I did not actually analyze the sponsor's data. So what I will be showing you are the actual results of the sponsor.
In the clinical study protocol, the study objective was to determine if the CTI system, when used in conjunction with clinical examination and/or diagnostic mammography, increases the ability of physicians to differentiate benign from malignant or suspicious breast abnormalities.
Now what had to be done at a later date was to drop the clinical examination alone because they found they could not focus the BCS unless they had the actual mammography results.
Continuing with the clinical study protocol, the study design, this is a prospective study. There's a blinding to histology, which meant that the actual histology results were not made available until the data was actually analyzed.
It is a multi-center study. There were actually six physical sites. The design of the study was intended to compare the level of suspicion ‑‑ that's LOS ‑‑ which is a number, 0, 1, 2, 3, 4, 5, of malignancy of suspicious breast lesions for clinical examination or diagnostic mammography before the BCS used, through the combination score LOS plus the BCS index of suspicion, IOS, which is a score ranging from zero to one hundred and measured in hundredths. So it is almost a continuous score, in contrast to the very discrete score of the LOS. This is, again, of malignancy of suspicious breast lesions.
Biopsy was used as the gold standard for pathology. One thing that had to be changed there was, again, clinical examination had to be dropped because the diagnostic mammography was required for the BCS to be used.
In the study population, the original study population was 600 patients with biopsy. The actual study population was 2,407 patients with biopsy.
The demographics: gender, there were actually 15 males; ethnicity, mostly Caucasian, a good deal of African-Americans, some Latinos, and others; age, most people were 40 to 60 with a lot being greater than 60.
The primary effectiveness was on the overall population. The evaluation was to be the area under the ROC curve, the AUC, to compare results of diagnostic LOS score without BCS and the diagnostic mammography with BCS. That is a combined score, LOS plus IOS.
They also mentioned that they would look at sensitivity and specificity, and the CTI system will be considered effective if its performance in conjunction with diagnostic mammography and/or clinical examination is clinically better than mammography and/or clinical examination alone.
ROC is the receiver operating characteristic.
There were secondary effectiveness mentioned on subpopulations. They didn't say what they were going to do with this, but they did say they would look at the mammography lesion type by three categories: calcifications, masses, and distortions. These lesion types are not mutually-exclusive categories since a lesion can be more than just one.
The mammographic lesion size was measured in three categories broken into a half centimeter and one centimeter, and the mammographic lesion depth was simply stated "as available in the protocol." It wasn't specified.
Safety evaluation was by occurrence of adverse events. In the PMA clinical study population, the sponsor acquired BCS images from 2,407 patients at six physical U.S. clinical sites from December of 1996 through April of 2001. The sponsor actually analyzed only those patients with both mammography, not those with just clinical examination, and biopsy within 60 days.
Bob, could you turn on the slide? This will show a flowchart of the sponsor's clinical study, starting at 2,407 and then fanning off in various directions, depending on what part of the study I'm talking about. Hopefully, that will help. I don't know whether that helps or not.
Look at your hard copy. It's a lot easier to read than what's on the screen, but if you want to glance up there, it's there.
The PMA training clinical study consisted of 700 patients. This consisted of the first 220 patients plus an additional 480 patients which were randomly selected from among the next 1,912 patients. These were used by the sponsor to set the following BCS IOS, index of suspicion, cutoff, and that is 20.59, which implies a recommendation for biopsy of a given lesion, or less than 20.59, which implies a recommendation for short-interval followup of a given lesion.
There were 1,432 patients enrolled from December 1996 through October of 2000, and these were initially available to test the effectiveness of the BCS, both in the original PMA and in Amendment 4. Now what's missing here are the 275 patients which were in the pipeline at the time the data was frozen and analyzed.
Out of these 1,432 patients, there were 769 patients with 187 malignant and 688 benign lesions that were actually included in the effectiveness evaluation. This is true both for the original PMA and for Amendment 4. The population didn't change from one submission to the other.
Note that each patient had from one to four lesions, and the sponsor assumes that lesions within patients are independent in their analyses.
The PMA clinical study results: The primary effectiveness was by ROC, area under the curve. I'm just going to mention that the last two analyses the sponsor did, because I think they make a point, the sponsor found, after excluding calcifications alone, a statistically-significant greater area for the combined score IOS plus LOS1. By "LOS1," I mean that this is an LOS score where the unknowns and zeroes were eliminated.
Then for mammography, LOS score, which score is 1, 2, 3, 4, alone had a significance level of .05. Now what they then found after, again, excluding the calcifications, but now expanding the mammography LOS1 score to add an additional two intermediate categories, 3.5 and 3.75, which was obtained by rereading the 4s, no statistically-significant difference in the area. This is referred to as LOS2 because now it includes the numbers 1, 2, 3, 3.5, 3.75, and 4.
Now I have done some plots here. What I did is I just crudely reproduced what the sponsor has in the PMA submission. You can see there that the purple line indicates the combined score, the IOS plus LOS1, which is essentially the IOS score just bumped up or down, depending on what the LOS was. You can see that's almost a continuous curve, and you can see what happens when you use the ‑‑ you can see the purple curve is almost continuous. This represents the IOS score bumped up or down by the LOS.
Now when we go to LOS, we only have a few numbers. We have 1, 2, 3, 4. You can see there's a point here, and there's, of course, an obvious point down there. What has happened is that those two points are just connected with a straight line, which makes the area under the LOS curve smaller, statistically-significantly smaller, than the area under the combined curve.
Now I want to show that that is an artifact because, if we go to the next curve, now I have added in two more points. I have added in just two, 3.5 and 3.75, and you can see now the curve, the LOS curve, moves up. In fact, it even intertwines with the combined score.
Here, although physically the area under the IOS-plus-LOS curve is greater, the statistical significance is lost. I think this shows the problem of trying to compare a continuous curve with a discrete curve. There's a definite bias against the discrete curve.
Now continuing with the PMA clinical study results for safety, the following four adverse events occurred out of 2,407 subjects. This is from December 1996 all the way to the end, April 2001.
There were two mild, possibly-related, resolved adverse events with both associated with patient discomfort during positioning, and then there was one serious and one mild, not-likely-related, resolved adverse event for hospitalization for treatment of a preexisting metabolic disorder and dizziness when sitting up after thermal imaging, respectively.
Now with Amendment 4, the sponsor essentially dropped their ROC analysis and just focused on sensitivity and specificity. This was mentioned in the PMA, but no detail was given. Here we have a little bit more detail on what the sensitivity and specificity look like using the 20.59 cutoff.
You see overall, with 187 malignant, to determine the sensitivity, and 688 benign, to determine the specificity, we get a sensitivity around 97 percent with a 95 percent confidence interval from 94 to 99 percent and a specificity of 14 percent with a confidence interval from 12 to 17 percent.
Now this overall result was rejected in terms of looking at the various lesion categories. The numbers given here are a little bit strange because they don't add up, and the reason they don't add up is some calcifications also were masses, and whatnot. So what's being done here is any calcification is shown at the top, any mass in the middle, and any distortion at the bottom.
You can see from this that they got perfect results for both masses and distortions in terms of sensitivity. That is 100 percent. The sponsor interpreted these effectiveness clinical results by lesion type to specifically exclude that didn't come out to be 100 percent; namely, the calcifications, where it was only 95 percent sensitivity, which they claimed was not acceptable. And you can see the confidence intervals associated with those estimates.
The Amendment 4 interpretation: The sponsor's initial rejection of the overall effectiveness, followed by the sponsor's differential findings among the three lesion-type subpopulation, clearly indicates exploration, which does require confirmation and must be based on new data.
That came in Amendment 5, in which they did the so-called post-PMA or PPMA population. These were the 275 additional patients that had been in the pipeline at the time of the original submission and now were analyzed for the first time.
The gender is almost the same as before, which is one male; ethnicity, mostly Caucasian, some African-Americans, and a few others; age, again, mostly 40 to 60 with some large number greater than 60. So very similar to the original population.
These 275 additional patients were enrolled from November now of 2000 through April of 2001 at just three of the six original U.S. clinical sites which were initially available for confirmation of the effectiveness of the BCS in Amendment 5. Out of these 275 additional patients, there were 173 patients after exclusions with 43 malignant plus 159 benign lesions that were actually included in the Amendment 5 effectiveness evaluation.
Similar to the original, these patients had from one to three lesions. Again, the sponsor assumes that lesions within patients are independent.
The overall results were, based on 43 malignant, sensitivity 94 percent, and 159 benign, specificity 20 percent, and the confidence intervals shown there, which are fairly wide because of the small numbers here.
This is what the PPMA Amendment 5 results look like when broken down, again, by lesion type. Again, they don't add up because some categories overlap.
We see here that the sponsor interpreted these effectiveness results by lesion type to specifically include only masses. Masses gave the highest sensitivity and the highest specificity.
Of course, we see what happened to distortions. They were 100 percent on sensitivity before; now they're down to 75 percent.
Of course, calcifications were 95 percent before; now the sensitivity is 93 percent. So they've definitely shown that calcifications need to be excluded.
Now in Amendment 7 the sponsor makes an adjustment. The sponsor attempted to use Bonferroni in response to FDA Deficiency 1(a). In other words, we specifically asked them what they were going to do about the fact that they are combining the data from the PMA Amendment 4 and the PPMA Amendment 5, and looking at all the data, how are they going to handle that?
What they said: We'll widen the sensitivity/specificity confidence interval estimates which are based on a simple, direct combination. They just added the data together, both the exploratory and the confirmatory clinical data, to test a theoretical possible set of hypotheses.
They first did the seven lesions types, which were explained by Dr. Rust: the mass, distortions, and the calcifications, and all combinations thereof, or possibly 63 lesion types, sizes and depths. Here they're multiplying seven by three sizes, and then they are breaking the depths into three sizes also.
Now these hypotheses are not explicitly included in the protocol. In the protocol they simply say: We're going to look at these things. They don't say what they are going to do with them.
This is not statistically-acceptable because the sponsor simply estimates sensitivity, specificity, and confidence intervals for various subpopulations without actually statistically testing any hypotheses. At this point in their analysis, what they are doing is they are just telling us what the sensitivity and specificity is. I don't see any hypotheses there. Therefore, I don't see any need for any Bonferroni adjustment. Therefore, as far as I am concerned, there are no multiple comparisons requiring adjustment.
In conclusion, one thing I want to make clear is that diagnostic mammography, not just clinical examination, is required for use of BCS. The sponsor's primary effectiveness demonstration using ROC, area under the curve, loses statistical significance when mammography LOS1 through 4 is expanded by just two additional intermediate categories after excluding calcifications alone.
The sponsor's initial rejection of overall sensitivity, followed by rejection of calcification alone sensitivity, indicates exploration which requires confirmation, which requires new data. The sponsor's attempt at Bonferroni adjustment to make sense out of putting all of the data together by widening the confidence interval estimates is not statistically-acceptable.
Thank you very much for your attention. I am going to turn the podium back over to Bill Sacks to continue with the clinical.
DR. SACKS: Before I do that, I just want to make a point about safety. We have heard that there were four adverse events out of 2,400, which is a very important aspect of safety, but there are two aspects of safety for any diagnostic device. It's not peculiar to this device.
That is the accuracy of the diagnostic output of the device also involves a question of safety. So as far as the adverse events were concerned, they were very few and minor, but from the point of view of the BCS output, we should focus on safety is more closely related to the question of sensitivity; that is, on cancers or the false negative rate. In other words, how many cancers have their diagnoses delayed? Also, in the context of the psychological impact, the false positive rate also can be regarded as a question of safety.
Effectiveness is more closely related to the specificity; that is, its performance on the benign masses, because the intent, the intended use of the device is to save biopsies of benign masses.
Now let's look at what the clinical trial demonstrated. I am going to just summarize this briefly for the history of these again.
There were four relevant clinical submissions here: the PMA, Amendments 4, 5, and 7. After reviewing the PMA, the FDA sent a letter to the company listing a number of deficiencies, and the company's response was Amendment 4.
In Amendment 4, for their conclusions concerning the effectiveness, the company retrospectively selected from the PMA data one of two analytical indices, namely, sensitivity and specificity, as opposed to ROC curve comparison, and two of three lesion types at first, masses and architectural distortion, and not microcalcifications.
In that same amendment, however, the revised labeling further deleted architectural distortion and referred to masses alone. So that was sort of both steps were involved in Amendment 4.
The FDA sent another deficiency letter, and the response was Amendment 5. Amendment 5 was offered as a test of the device in additional subjects. That's those 275, although not all of them were evaluable, additional subjects who had not previously been analyzed. That was because Amendment 4 had contained retrospective selections.
The company refers to this additional dataset as the "post-PMA." That is PPMA for short.
This amendment confined its analysis of the PPMA data, that is, the newly-analyzed data, just to the newly-chosen analytical index, namely, sensitivity/specificity, in the newly-chosen subgroup, masses. That was done before unvaulting that data. So as far as this data is concerned, that was prospectively done.
In addition to presenting data on a new set of subjects, the amendment also contained an analysis, as you have seen, of the combined datasets from Amendment 4 and the PPMA.
Because of the retrospective selections in Amendment 4, the FDA asked the company to justify combining that data with the PPMA data, and the response was Amendment 7.
In Amendment 7 the company applied the Bonferroni correction, as you've heard, in an attempt to compensate for retrospective selection and the smallness of the additional PPMA sample.
Now, as we go through this, there are two overriding issues. One is the adequacy of the data, and the second is the interpretation of the data. That is, do they demonstrate safety and effectiveness of the device, assuming that we accept that the data is adequate?
On the question of the adequacy, a question that we will have the Panel consider this afternoon, and before that we will give you the questions as they are phrased more precisely. This is paraphrasing. Can the data from Amendment 4 contribute to the judgment of safety and effectiveness when it consists of retrospective selections? Is a Bonferroni correction applicable in this context? Are the data from the PPMA alone adequate for the judgment of safety and effectiveness?
In looking at the interpretation of the data, it is noteworthy for the following discussion that no formal hypotheses were explicitly put forward for testing either in the PMA or in the subsequent amendments, and let me hasten to add here that, to qualify as a testable hypothesis, there must be a quantitative criterion whereby either a point estimate may imply rejection or a confidence interval may entail exclusion.
There were two implicit hypotheses. One was that the ROC area for the device and mammography combined would exceed that of mammography alone with statistical significance.
The second was ‑‑ and this was derived from the training set of 700 subjects, of whom 150 were cancers, and the sensitivity of the device ‑‑ that is, the threshold for the device was set such that 149 of those 150 cancers were positive, with one being negative, which is a sensitivity setting of 99.3 percent. There was an implicit hypothesis that the point estimate for sensitivity would be at least 99.3 percent in at least 75 percent of simulations with the data.
The protocol otherwise contained only non-quantitative statements of what the company hoped to achieve in the clinical trial. One example, quote, "The objective of the study is to determine if the BCS, when used in conjunction with clinical examination and/or diagnostic mammography, increases the ability of physicians to differentiate benign from malignant or suspicious breast abnormalities." But there is no quantitative criterion by which we can judge success or failure on this, except through ROC area comparisons, but those were dropped.
In the original PMA submission, the comparison of ROC areas failed to achieve statistical significance except, as Harry has shown you, as an artifact of too few points in the mammography alone curve. It was, therefore, not pursued in any of the amendments.
In addition, the sensitivities failed to achieve a level of 99.3 with 75 percent confidence in any of the datasets. Here is a diagram that shows them. I finally get to use this pointer.
This is the upper righthand corner of an ROC plot. It is about a quarter in both dimensions. Mammography alone, because of this being the universe here is women on their way to biopsy based on mammography, was 100 percent sensitive. That is just an artifact of the choice of the universe here.
It was, similarly, zero percent specific. That is, there were no non-biopsied people here.
Now the PPMA ‑‑ I'm sorry, the original PMA point estimate with 187 cancers ‑‑ this "N" is just the number of cancers ‑‑ turned out to be 97.1 percent. For reference, this line here is the 99.3 percent level that was involved in that implicit hypothesis of trying to keep it above 149 out of 150. Its confidence limits are, as you see here, 94.1 to 98.8.
The next data was Amendment 4, where out of this set of 187 were culled the 90 masses. It is the same as ‑‑ these 90 are part of this on the 97. In those 90 we got point estimate of 100 percent sensitivity. The higher confidence bound, of course, is also 100 percent because you can't go over that, and the lower one is about 96.7.
Because of the retrospective selection of these out of this group, the next set of data in Amendment 5 was the PPMA, of which there were 15 cancers. The point estimate there was 93.3 percent because one of those turned out to be negative, so 14 out of 15.
The confidence interval on this, because the number is so small, 15, is rather wide. Interestingly, the lower confidence bound is actually below the chance line.
When the two sets of data are combined, if you think it is valid to combine these two, you get a point estimate of 99.0, which is still below the 99.3, and its lower confidence limit is about 95.6. So that sort of displays all of the data in reference to that 99.3 implicit hypothesis.
The potential safety and effectiveness in the U.S. population as a whole ‑‑ this is a bit of a busy slide, but I'll walk you through it. The percent of U.S. biopsies that are potentially obviated by the BCS, if used on all eligible women, and we've seen these figures, and mine are very close to the company's, 1.3 million U.S. women biopsied each year, of which I use the number 45 percent; the company used 45.5.
Forty-five is not only a typical figure for the country at large, but happened to be exactly the percentage in the used data combining the PMA and the PPMA; 45 percent of them were cancer. So I used that figure. That's about 585, which is very close to ‑‑ Steve Rust gave you a figure that was 591,000, very close ‑‑ of which 80 percent, roughly, are benign. That's about 468,000, of which 15 to 20 percent, using the various ranges of specificity that we got for the device, which would be 70,000 to 94,000, would be BCS-negative and, therefore, save the biopsy.
So 70,000 to 94,000 out of 1.3 million is roughly 5 to 7 percent of the 1.3 million U.S. biopsies would be obviated. That is if the BCS were used on all 585,000 women who are eligible; namely, mammographic masses on their way to biopsy.
In addition to saving biopsies on these benign masses, approximately 1 to 6 percent of the malignant masses ‑‑ that's, again, the range from the data ‑‑ and a half to 3 percent of all breast cancers, that is, not just of masses, might be delayed in diagnosis.
A couple of labeling issues involved the size of the mass and the depth of the mass. We are going to ask you for some discussion on this this afternoon.
The size of the mass: The effect of small lesion size on device sensitivity was difficult to evaluate since only 2 out of the 105 cancers in the two combined sets were smaller than 5 millimeters. Here are the figures for how they fell. This is different from the figures that Karleen Callahan gave you, but she was including the ones that we didn't have the data for. This is just the two combined, Amendment 4 and PPMA data. Only two of the malignant masses were less than a half a centimeter, 5 millimeters. So it is hard to make any statement about it.
With the chosen threshold, there was no definite effect of lesion depth on BCS result, but, as the mammographers here know, the effect of lesion depth is difficult to evaluate because depth is not easily gauged on the mammogram. Worse yet, we are imaging women in a position in which the breast is pendulent. It is a fairly mobile structure. Depth is variable. A given lesion has different depths in the breast, depending on the position.
Therefore, we really have difficulty from this data making any judgments or conclusions about depth. However, one should realize that, just from the physics of the situation and the physiology, that the deeper the lesion, the less effect a cancer will have on contrast of temperature on the overlying skin. So that might affect device sensitivity, but we can't make any statement about it.
Conclusions then: In summary, only 4 out of 2,407 subjects had an adverse event, all minor. In that regard, the device seems safe.
There were no explicit, quantitative hypotheses. There were two implicit, quantitative hypotheses. Neither hypothesis was fulfilled. Most of the data was selected retrospectively. Bonferroni correction we feel is not applicable in this context, in part because there were no hypotheses; therefore, no alpha levels to protect, and so on. But if you did use the correction to widen the confidence limits with the point estimates already below the implicit hypothesis of 99.3, that doesn't help keep them above it.
Finally, using the trial results, if the BCS were in general use in the U.S., it would obviate 5 to 7 percent of the 1.3 million biopsies a year, and approximately 1 to 6 percent of these obviated biopsies would turn out to be malignant and their diagnoses would, thus, be delayed.
The people who are still awake will notice this 1 to 6 percent is not the same figure as 1 to 6 percent I gave before because one was looking at the percent of malignants that would be negative and this is looking at the percent of negatives that would be malignant. It turns out that the diagonal members of the 2x2 table are about the same, so these figures come out to be the same, or perhaps it's not so coincidental.
CHAIRMAN MEHTA: I think, before we leave for lunch, I would like to remind you that the open Committee deliberations will resume at 1:00 p.m., but the Panel members are requested to be back here at 12:30 for a Panel-members-only Closed Session from 12:30 to 1:00 p.m.
(Whereupon, the foregoing matter went off the record for lunch at 12:00 noon and went back on the record in Closed Session at 12:36 p.m.)
CHAIRMAN MEHTA: Before I call the meeting back to order, Nancy Brogdon, Director of the Division of Reproductive, Abdominal, and Radiological Devices of the Office of Device Evaluation, has a few words that she would like to say.
MS. BROGDON: Thank you, Dr. Mehta.
I wanted to let you know that two members of this Panel are completing their terms before the next scheduled meeting.
First is Ms. Marilyn Peters, the Panel's consumer representative. Ms. Peters was a Patient Health Education Coordinator for the Department of Veterans' Affairs at the West Los Angeles Health Care Center. She recently retired from this position.
Serving as the consumer representative is a difficult task, as you know, due to the wide spectrum of devices that come under the radiology umbrella. Although meetings of the Panel required Ms. Peters to travel across the country, her attendance has been perfect during her tenure.
Thank you, Ms. Peters, for your service on the Panel.
Next is Dr. Alicia Toledano, the Panel's biostatistician. Dr. Toledano is Assistant Professor in the Center of Statistical Sciences at Brown University.
Dr. Toledano has served the Panel well, providing insightful input on all the various devices that have come before the Panel during her tenure. In addition, she's been called by at least one other panel for her statistical expertise and, when requested, she has provided written reviews that are very comprehensive for statistical analyses of various PMAs. We'll miss her contributions as a Panel member, but, if she agrees, we hope that she will stay on as a consultant, so that we can call on her for other difficult statistical issues in the future.
And thank you, Dr. Toledano, for your service on the Panel.
We'll be sending both of you recognition plaques as soon as they're signed by our new Commissioner, Dr. Mark McClellan. Thank you very much, and best wishes to both of you.
CHAIRMAN MEHTA: I would now like to call the meeting back to order.
I would like to remind the public observers at the meeting that, while this portion of the meeting is open to public observation, public attendees may not participate unless specifically requested to do so by the Chair.
We will now continue with the Panel's discussion of the PMA. What we will do here is I would like to ask the Panel to pull out the Panel discussion questions, which were actually placed in your blue folders, and maybe use these as a starting for discussion points for the Panel discussion.
Who would like to go first?
DR. TOLEDANO: Are you sure you want me to go first?
CHAIRMAN MEHTA: Yes, if you would like to, go ahead.
DR. TOLEDANO: Okay. My first question for today has to do with the feasibility of using this device in clinical practice. We have seen numerous exclusions, many of them for technical factors. We have reference to the fact that the device has changed, more things are automated, there's better training, the company thinks that the performance in terms of being able to obtain results on women has improved. So I guess my concern is, have you proved to me that you would be able to obtain results in the women who undergo this procedure?
Now there's a second thing that's sort of buried in some of the appendices, and maybe it comes to my mind easily because I'm a statistician. The blinded investigators were assigned three to each subject. If all three ‑‑ when it happened in the first study, the subject was reviewed by two investigators, and if there was a disagreement, the third investigator was brought in.
You can explain if I've got this in any way incorrect. For each woman there were up to three scores weighted one-third each, but there were some women who only had two evaluations and there were some women who you could only get one evaluation and there was some women who you got no evaluations.
So even beyond everybody who was excluded because you got no evaluations, there were instances where you tried to get evaluations but you couldn't. That, to me, is a big concern. So could you address that a little bit?
CHAIRMAN MEHTA: Does the sponsor want to suggest someone who might be able to address that?
DR. RUST: Let me start just by saying that I think you do have it correct in terms of how we analyzed the data, but, just for everyone's benefit, we attempted to get three evaluations for each lesion. If we got three, all three were used in the analysis. We weighted each evaluation one-third. In some cases there was a failure to evaluate on a single, leaving us with two evaluations. Those were both included, weighted one-half. In some cases we failed to get two but got one, and that was given a weight of one in the analysis, and, yes, there were cases where we failed in all three cases to get an evaluation. So I believe that that is the correct interpretation of how we handled the data.
Lynn Satterthwaite will address the feasibility of using the device.
MR. SATTERTHWAITE: We'll address it in two parts, the device itself and then the other part of it Dr. Callahan will address.
In the clinical trial we utilized a device that, as we have characterized earlier, was mostly a manual device. Manual had much to do with the cooling challenge itself, but there's a little more to it than that in that the images were checked well after the fact, well after the imaging session.
To address several of the problems, problems included movement of the patient, poor positioning of the patient, and the possibility that the cooling challenge wasn't adequate. Much of that had to do with the reliance on the technologist to manually throw a switch. The protocol required that they watch the second hand, and 30 seconds into the imaging procedure they were required to turn that switch. In many cases that switch was not thrown, it was thrown right at the beginning, or thrown too late, and those kinds of problems.
So these issues have been addressed in the following way: The automated system, the functions performed by the computer now include doing this temperature challenge check right at the imaging session time. That will prevent us from accepting that image if it was not adequate.
That takes into consideration all factors. Was it adequate? Was its timing done in the right place? It also prompts the technologist in the session to review the image at that point in time. So there's some discipline here that we enforce because the technologist must look at the screen and go through it step by step.
They look at a cinema view, much like you saw today, and that would allow them to make the decision as to whether they truly were positioned correctly and whether there were no artifacts. At times we would see a gown strap that's down in the image and those sorts of things, and those cause problems.
So we believe that the design of the system, which functionally does exactly the same thing, the same camera, the same cooling ‑‑ it is just the protocol is enforced by the computer ‑‑ will solve many of these problems. We do believe that in clinical practice we can enroll patients, at least from an imaging standpoint, and get good images or know that we don't have good images and we'll know to do it again, if we have to.
DR. CALLAHAN: And I'll just briefly address technologist training. Another thing that we have instituted since identifying that there were a number of images that were not usable is we have developed a set of training cases to use for technologist training. We've implemented that in our study with Dr. Hughes at Mass. General. That study just began a few weeks ago, but I understand the first eight or ten patients that have been enrolled, the images have been taken; there's been no quality problems. They are evaluable images, albeit a small number.
MR. SATTERTHWAITE: And let me follow that up. When we enroll patients in the clinical trial, if any of the factors that made their image inevaluable were there, then we would lose that patient. With the new, automated system and the prompting by the computer, we would know to go back and do the imaging then. So there is a number that we wouldn't lose because of that factor.
DR. TOLEDANO: Okay. So can I follow up?
CHAIRMAN MEHTA: Yes, go ahead, Alicia.
DR. TOLEDANO: Okay. So that addresses the issue of being able to obtain usable images from the system. Now the physician has to look at them and obtain a score, a test result.
We are still left with the issue that in the trial images that were potentially evaluable that were presented to physicians, they couldn't obtain a test result. How often? What's been done to address that?
DR. HUGHES: What we have been dealing with is that the central reviewers received a set of mammograms and then had to figure out where the lesion was relative to the mammogram. Any lesion that didn't show up on a mammogram they could not use.
Within the study we're currently doing is, how do we incorporate it into clinical practice? The way we're doing this is I'll examine a patient. I'll review the mammogram. I'll schedule that patient for a biopsy. We'll then do the imaging. I know exactly where the lesion is. I know where I felt it. I know where it is on the mammogram, and I can pick where to go.
Is that your question?
DR. TOLEDANO: That's not the question.
DR. HUGHES: Then what is your question?
DR. TOLEDANO: Okay. The question is, three physicians were presented with images that they were supposed to be able to work with. Sometimes not all three got it. How often? How do you address that in clinical practice?
DR. HUGHES: Within clinical practice I can hit the area of interest for the patient that I'm dealing with because I know where the region of interest is. Other than that, I'm not sure what the three disagreements are.
DR. TOLEDANO: I'm not stating it clearly?
DR. CALLAHAN: The question, if I understand it correctly, is we had three evaluators. They each had the same films. They each had the same thermal image. They each were given the same opportunity to evaluate, and in some cases not all three chose to evaluate the case or were not able to evaluate the case.
The issues are partly as Dr. Hughes described, in that they may not have felt ‑‑ there's differences in mammographers' interpretation ‑‑ they may not have felt, one of the three may have said ‑‑ well, if you've seen the evaluation sheet that the evaluators got, it described the case to the extent that was in the patient database as to where the lesion was located, as well as the lesion descriptor. So it says there was a mass in the upper outer quadrant of one centimeter size, for example.
If the evaluator felt like he or she did not see that lesion with that descriptor, they would choose not to evaluate it because there could be other things on that mammogram perhaps that they thought was suspicious, but they were instructed ‑‑ the instruction on the evaluation sheet was to attempt to evaluate the lesion that was actually biopsied.
So that accounted for some of the cases where not all three evaluators interpreted the information in the same way. I think that was the predominant situation.
CHAIRMAN MEHTA: Let me follow up on that question. The use of the expert panel for evaluation was carried out through an amendment. It was not the original intent of the protocol to have such a panel.
DR. CALLAHAN: That's correct.
CHAIRMAN MEHTA: Why was it done? When was it done, and how many patients had already been enrolled and evaluated prior to the panel coming into place?
DR. CALLAHAN: Okay, that amendment was done in September of 1998. My understanding is that that was done for a couple of reasons.
One is it was felt that the original investigator might, by knowing the patient outcome, biopsy outcome, or something, their evaluation might potentially be biased at that time, if their evaluation of the IR image could have been, you know, biased by the pathology outcome.
CHAIRMAN MEHTA: But the original investigator did not have the pathology at the time of the IR image.
DR. CALLAHAN: There was no requirement ‑‑ and Dr. Parisky can perhaps address this or Dr. Hughes ‑‑ but there was no requirement. They did the IR imaging prior to biopsy, but they were not required to do the evaluation in real time at the clinical sites.
CHAIRMAN MEHTA: But they did the ROI in real time?
DR. CALLAHAN: No. No, they did not.
DR. PARISKY: I'll demonstrate a patient flow. I would enroll a patient, get informed consent for the IR imaging and also informed consent for biopsy, and would take the patient because I have a mass. Rather than do an ultrasound, see the mass, take the patient off the table, take them to another room, do a test, and then return for the biopsy, at my center specifically Dr. Silverstein prides himself that we will give you a result within that day.
So I enrolled the patient, would go to ultrasound, would do a biopsy, and while I can't prove to you scientifically, those of us who do that, the way we watch a needle go into a lesion, we can tell you if it's likely cancer or not. We felt, in discussion amongst ourselves, or we had information such as an MRI or we had information such as obtaining prior films, that showed a rapid growth, things that would be biasing the investigators at that site that perhaps would have a bias.
It was felt by the company and in counsel with the principal investigators at the various sites that, to try to remove this non-quantifiable bias, that it would be evaluated best by a separate panel.
CHAIRMAN MEHTA: Let me follow up on that. Given the fact that the assessment is actually done by a computer using a number, it's either a yes or no, all it would have taken would have been for the diagnostic radiologist to place the ROI, and the computer could have done it at any point because the computer didn't know what the pathology was.
DR. PARISKY: If this was all in a work area that was immediate within my clinic or within some of the clinics, I think, and if patient flow allowed for that. While hindsight may have suggested that, the course of events were that we would in a busy clinical practice move a patient from the IR table, you know, pull together the images necessary for the biopsy ‑‑ this is my situation ‑‑ and proceed forth.
I appreciate your comments and criticism.
CHAIRMAN MEHTA: So in that case, in your busy clinical practice, how do you see incorporating this and giving the patient the results on the same day, if you can't do it in real time?
DR. PARISKY: The same way I incorporate giving them my results to each of my patients following diagnostic workup. I will incorporate that as ‑‑ will then take the time to do so, once all the information has been compiled.
Conducting research is one thing. Providing patient care adequately, we'll make time for it, sir.
DR. CONANT: How long does it take to get the IR reading?
DR. PARISKY: It takes minutes, but in terms of locality as to where the patient was taken, in some instances it was not on the same floor.
DR. CONANT: No, the numerical IR output?
DR. PARISKY: Three seconds, yes.
CHAIRMAN MEHTA: Let me just follow up, and anyone can answer this question. The reason I'm sort of going on this direction of questions is, to some extent, the use of a panel is useful. To another extent, it detracts from reality in terms of how this machine is intended to be used. It is intended to be used by a practice in reality, not necessarily at the Norris Cancer Center, where there are hundreds, if not thousands, of women that are undergoing mammography on a daily basis, but much smaller numbers of women that are having mammography.
The ability in a clinical trial to demonstrate the effectiveness of this machine in that context is negated by using a panel that replaces an individual radiologist's decisionmaking. That is one of my concerns.
DR. HUGHES: Just from a logistical point of view, I have the machine in my clinic. I'll see a patient. I'll say, "You're going to need a biopsy." I'll have her consented for the biopsy. She will go and have her thermography done. It takes about 25-30 minutes to do.
The image is immediately available for me to look at, and I evaluate the image immediately at that time. I'm not using these images to change my care, because that's not part of the protocol, but getting the image done, seeing it, putting the ROI in place, getting a reading out is done within the context of my normal clinic day.
CHAIRMAN MEHTA: And I'm sorry, do you do this in the context of a clinical trial right now or what?
DR. HUGHES: This is a clinical trial we're running currently, which is ‑‑
CHAIRMAN MEHTA: It's a separate clinical trial from ‑‑
DR. HUGHES: That's correct, which is to look at ‑‑ I guess, to answer your question, and we want to answer the same question: When we put this into clinical context, how much does it help us with our patients? How logistically is it possible? How often are we not able to find an ROI? That's not happened as of yet. How often are the images unevaluable? That's not happened as of yet.
And this is eight or ten patients, but we're not having the problems that you have when you're taking these images, shipping them across the country, and asking somebody else to evaluate them.
DR. CONANT: So these are prospectively patients that are coming to your office after having their mammogram and ultrasound?
DR. HUGHES: These are patients who are in my office who require a biopsy. They may or may not have had a mammogram or ultrasound, depending on their age. But at the end of the visit, when I say, "You need to have a biopsy done," based on mammography, ultrasound, whatever is available, my exam, we then do the imaging and use that to predict what we would do, were we to use this as an FDA-approved device, which it currently is not.
DR. CONANT: So they are having mammography and ultrasound before you recommend the biopsy?
DR. HUGHES: Well, if the patient is 20 years old, I don't do mammography.
DR. CONANT: No, but ultrasounds.
DR. HUGHES: So in the majority of cases they would have a mammogram, ultrasound, and physical exam, that's correct.
DR. CONANT: And how many of them go to biopsy that have a negative ultrasound and a negative mammogram?
DR. HUGHES: Out of ten patients, I can't answer that quite yet. But within my practice I have a fair number of women who have come in with breast lumps that turn out to be fibrocystic changes that the patient is concerned about. That may be 10-20 percent of the biopsies I see.
DR. CONANT: With negative ultrasounds?
DR. HUGHES: Negative ultrasounds, patients with calcifications, patients with mammography, and this is a different trial.
But I can't give you the percentages of those things currently.
DR. CONANT: Okay. Thank you.
CHAIRMAN MEHTA: Alicia, go ahead.
DR. TOLEDANO: So going back to the patient flow, they all are recommended for biopsy. So you've done all your preliminary workup. They're all recommended for biopsy. You do the infrared imaging. They proceed to biopsy.
Now in the clinical trial results that you present to us in support of your PMA you have excluded the women who did not have biopsies, even, for example, if they went for the biopsy for like an ultrasound-guided biopsy, and the ultrasound determined that it was a fluid-filled cyst.
Now you're looking at two different populations, because I cannot tell, when I'm doing the IR imaging in clinical flow, I can't tell whether they're going to end up having a fluid-filled cyst. I don't know if it is unguided biopsy.
Doesn't the exclusion of, I think it was 20 percent of women, 10 or 20 percent of women, because they didn't have biopsies, doesn't that exclusion open up the results of your clinical trial to some significant bias?
DR. HUGHES: I'll let the statistician handle that one, I think.
DR. RUST: I think asking the question whether or not those exclusions create bias is a very legitimate question. Now one of the significant classes of patients that were excluded that way are patients that, upon ultrasound, would not have been scheduled for biopsy.
So I think what we have here is a sequencing-of-events type of thing. In Dr. Hughes' clinic that ultrasound occurs before he makes the decision on IR imaging. In our case, because of the enrollment criterion for the trial, we enrolled based on mammographic findings and clinical findings only. So a patient got enrolled and had the ultrasound post.
But I believe that what you end up with are the same populations in both cases, that the population we end up studying, analyzing, is the same population that Dr. Hughes ends up sending to the IR procedure. So, in fact, those exclusions I believe are a good exclusion as opposed to a possibly negative exclusion.
DR. CONANT: I think Dr. Callahan this morning mentioned, in going over the exclusion groups, that some of them were complex cysts or questionable cysts that on aspiration, or were aspiratable but a needle was placed in them. I think they probably had already a recommendation for intervention, though, which would be a biopsy recommendation, and then they were aspirated.
I think that's a subgroup that Alicia may be referring to that actually is suspicious on mammography, whether they have the ultrasound or not, where complex or maybe even appearing solid ‑‑ this happens quite a lot clinically ‑‑ then had an aspiration, then were excluded. That's a bias, I believe, on those recommended for biopsy, excluded before they got to the analysis.
DR. PARISKY: The protocol that was reviewed with the FDA required histological proof of disease. While the scenario that you address, you're absolutely right, that needle was stuck in there, but any specimen that you might get from an intervention like that would be cytological and also fraught with the known cytological problems. Strictly following the approved protocol, we were required to disenroll patients who did not have histology.
DR. CONANT: Did you then have patients that might have had an FNA in someone's office for a suspicious area on exams that were positive and malignant and, therefore, known to the enrolling radiologist, an FNA based on palpation that wasn't histologic but ‑‑
DR. PARISKY: In my own center there was one case that was excluded because we knew the answer. We voluntarily excluded it on an ethical basis, but it didn't fit the criteria.
DR. CONANT: Okay, but these women still would have had IR imaging in the flow ‑‑
DR. PARISKY: I don't remember if they went to completion. You know, we enrolled patients, and then some decision points were made actually at times before IR imaging, you know, because we would get a result back. Sometimes patients were ‑‑ yes?
DR. CONANT: I'm just trying to think how it fits into the clinical flow of things.
DR. PARISKY: Yes. I think the way Dr. Hughes ‑‑ in terms of a standard clinical flow, I would give my own experience, on mammographically-apparent lesions we would run parallel or complementary to ultrasound. In a negative ultrasound we would then perform the IR. If it was a solid mass, we would perform the IR. If it was a cyst, we wouldn't perform the IR in my clinical practice, and it would be done within the setting of the two hours that the patient's in my hands.
DR. CONANT: There's been some good data published recently about palpable areas, which I suspect you're including in this flow, palpable areas, negative mammogram, negative ultrasound, and negative predictive value. It's almost 100 percent. Dan Kopans, Mary Scott Soo, Sue Weinstein, they've all ‑‑ three different publications.
So I'm just wondering how this would fit in and influenced ‑‑
DR. PARISKY: I would welcome the opportunity to evaluate patients with palpable abnormalities in a setting of a negative or a dense mammogram usually setting, and it would be complementary to ultrasound.
The problem we had in generating numbers was that patients who actually fit that scenario, not all centers placed markers, metallic markers, on the mammograms to allow the tertiary parties to ‑‑
DR. CONANT: Okay. So those would have been excluded? Those were part of that exclusion? Okay. I see.
DR. CALLAHAN: I think that both Dr. Toledano's and your questions are insightful. I think we just don't have all the data that you would like to know about how, you know, the impact or the potential impact of these other diagnostic procedures.
That is one of the points of the protocol at Mass. General, is to look at how this will integrate into actual clinical practice. There the enrollment criteria is women, mammography and/or ultrasound, whatever diagnostic tests lead to the decision for biopsy.
DR. TOLEDANO: Could you take your existing data and re-analyze it in an exploratory manner as if you had separated out masses in the beginning, had the ultrasounds before the IR? Could you do this? And if you could do it, have you done it?
DR. CALLAHAN: We could take our existing data and we certainly have a case report form for all those that have mammograms. I mean that was part of the information that was collected, as well as those that were enrolled only on clinical examination; that is, there wasn't mammographic data available. We could do an analysis like that.
Does that answer your question?
DR. TOLEDANO: Uh-hum.
DR. RUST: I think we can do the separation based on mass/non-mass in the beginning, as you're indicating. We do not have data on why the biopsy was not performed. So that's where our inability to do that analysis ‑‑
DR. CALLAHAN: And let me make it clear, we obviously couldn't do ‑‑ I was talking about potential bias in patient sets between those that weren't evaluated.
DR. TOLEDANO: That's right.
DR. CALLAHAN: Obviously, if they weren't evaluated, we don't have the IR data ‑‑
DR. TOLEDANO: Right.
DR. CONANT: -- so we could do bias, you know, potential bias.
DR. TOLEDANO: Dr. Brenna and I had the same reaction to those answers.
DR. GENANT: Well, overall, I have concerns about the primary efficacy and just how robust the examination is. I think that we do see that the primary efficacy endpoints of the area under the curve did not reach levels of significance and hardly showed even a trend.
Then it was in the post hoc analyses where you did reach levels of significance of sensitivity/specificity, but I do believe that the PPMA data cannot stand alone, and the Amendment 4 data are basically post hoc. So I think that you have problems with regard to statistically-significant endpoints in this study.
I think on top of that, we have been discussing the relevance of the manner in which the clinical trial was conducted to the way in which you anticipate using the system clinically. Now this is often a problem when you're doing clinical trials and then you're trying to extrapolate to the clinical practice, and I recognize that.
But I think it is also clear that there are many issues that at this stage could be better addressed in a prospective study than, in fact, they were, as we look back on the way that this study was conducted.
I remain particularly concerned on the evaluation side of the images. I don't really have a sense of how robust that is. You've told me that, for example, among readers, between readers, that one would see perhaps plus or minus four points; that on, say, a 100-point scale, that doesn't sound like it's terribly critical.
On the other hand, since you are using a threshold-based yes/no, it would be particularly important to know in the cases that are perhaps plus or minus ten around the threshold what kind of reproducibility there was, in fact, from one reader to the next.
And do you, in fact, have to have a trained panel of three readers reaching a consensus for this type of an approach to work or can a single reader, relatively reliably, give consistent information? I don't know that from the data that you have shown us to date. I think that is extremely critical when it comes down to using this in practice.
DR. RUST: Would you like me to respond?
CHAIRMAN MEHTA: If you would like to.
DR. RUST: I guess the last one first, and then I might have to ask you to remind me of the earlier points.
In terms of the way we analyzed the data, we allowed ourselves to be penalized for reader-to-reader variability within a lesion. In other words, we did not take the three evaluators, reach a consensus evaluation, and then evaluate that against the gold standard.
We did, in fact, if a lesion went two-thirds one way and one-third the other in terms of its evaluation, then that two-thirds error versus one-third correct is included in our performance statistics. So we did penalize ourselves for reader-to-reader variability within lesion in terms of positives and negatives. Okay?
DR. GENANT: Well, by virtue of the way that you did it, I'm not certain that actually three readers, three trained readers working kind of in concert ‑‑ in a sense, you're using their data combined ‑‑ may be more reliable than the single reader would be.
We simply don't have data, you haven't shown us data, on the impact that that might have. Yet, it has a lot of relevance about how you could, in fact, use this in practice.
DR. RUST: Yes, I guess our rationale ‑‑ and it's easier to think, if we have two evaluators ‑‑ our rationale was, if you have two evaluators, and let's say it's a malignant patient. One calls it positive; the other calls it negative. We would consider that a lesion for which it's more difficult to determine whether it's positive or negative, and so we put part of that lesion in the correctly-classified category and part of that lesion in the incorrectly-classified category.
So we are, as I think you can see by that explanation, penalizing ourselves for having multiple readers, and them not necessarily agreeing. So there was no averaging through consensus.
DR. GENANT: Well, I think that one could look at this and simply determine what the threshold was for each of the three readers and how that was in terms of positive/negative, and see what that level of agreement is. We don't know, but I presume that you probably have done that analysis at some point, I would think.
DR. RUST: Looking at the threshold for each reader?
DR. GENANT: Yes, and then also a reader doing the same examination twice in terms of, do they place the region sufficiently nearby that it doesn't impact your result? I don't know that.
DR. RUST: We did not have the same reader place the same ROI twice because we felt we couldn't get an independent placement the second time. Our reasoning was that, if you have three readers doing it, that includes both within-reader variability and reader-to-reader variability in your observed variability. That was our rationale.
But you had other points earlier. I don't know if you want me to address any of those.
DR. GENANT: Well, I was basically commenting on, I think, some of the deficiencies with regard to how robust the statistical analyses were and what we can draw from that. Because I simply feel that, based upon the information that we have, that I don't have a high level of confidence that the device is effective.
DR. RUST: Okay. You specifically indicated that our analysis of the original study data, you considered that to be post hoc. I guess what I would like to ask the Panel to consider is whether or not that analysis is truly and entirely a post hoc analysis.
What I have to draw your attention to is that in the protocol there was a prospectively-placed plan to do analyses by lesion type. Given that that was prospectively planned for, that bounded any of the analyses that would fit within that plan.
We specifically have corrected our analyses to be correct for any of those analyses in that bounded plan. Therefore, in my judgment as a statistician, it's not subject to being a post hoc analysis because it was done according to a prospectively-planned set of analyses.
There is a necessity to correct significance levels associated with the inferences, and we feel that we have done that. So I would just ask you to reconsider your thoughts on whether it is post hoc or not.
CHAIRMAN MEHTA: Dr. Rust, as you point us to the direction of looking at the protocol, to look at the plan, could you also point out to us where in the protocol the quantifiable variables for specificity and sensitivity, for analysis by mass, are included in the statistical evaluation plan in the protocol?
DR. RUST: I'm going to, as far as hypotheses and objectives, I'm going to turn to Dr. Callahan.
DR. CALLAHAN: You're correct, there is no quantitative definition in the protocol.
CHAIRMAN MEHTA: Thank you.
DR. TOLEDANO: There's a lovely article in AJR which you have an article coming out in January. It is in 1996 by a woman named Nancy Obuchowski. Dr. Obuchowski is one of our foremost people in statistics and diagnostic radiology.
In that article she discusses conduct of multi-reader studies and proper analysis of multi-reader studies. The convention in this field, in the field of multi-reader studies, is to present the measures of diagnostic accuracy per reader and then overall.
I do not see anywhere, and I have looked through everything, sensitivity for masses for reader one in your post-PMA or even within your pre-PMA dataset. Reader one, sensitivity, specificity, AUC; reader two, sensitivity, specificity, average. I don't see that. I need that. Do you have that?
DR. RUST: In fact, we did do such an analysis in the original submission, and there is a Table 5.9-M ‑‑
DR. TOLEDANO: Tell me which one is that. Module 5?
DR. RUST: Module 5, page 505.
DR. TOLEDANO: I still don't think that's what I need.
DR. RUST: Well, I don't think it's what you need, either, because it goes all the way back to the original submission, but I did, in answering your question, I did want to point out that we, in fact, did that analysis in the original submission, and you are correct, we have not repeated that analysis in any of the subsequent submissions.
DR. TOLEDANO: Okay.
CHAIRMAN MEHTA: Dr. Hooley, do you have any questions?
DR. HOOLEY: Yes. I'm concerned about the broad definition that you use for masses. In clinical practice a mass can have various definitions, meaning that one can be a mass as only seen clinically. Another one is the mass is only detected on mammography. The third definition is a mass that is detected on mammography and ultrasound as well.
I feel that ultrasound has been neglected in your review of the decisionmaking process. I am wondering if you have any data on the ultrasound characteristics of the masses and if that is available.
DR. CALLAHAN: The short answer to that question is we did not collect ‑‑ it was not part of the protocol, and we did not collect ultrasound data. So it's not available.
DR. HOOLEY: But, yet, in clinical practice almost all of these masses will undergo ultrasound, and ultrasound can be used to determine what the clinical outcome or whether or not a biopsy is going to be used.
I think that it would be important to know whether this would be useful in masses that were given sort of a BIRADS 3 or 4 on ultrasound as opposed to a BIRADS 5.
DR. CALLAHAN: Right, and, parenthetically, I can say that we do know that where ultrasound would normally have clinically been employed, it appears that it was because a lot of the radiology reports refer to ultrasound results. But as a fact of the protocol, it wasn't required that this be collected or documented or verified. We just do not have that information.
So I think the fact is that our enrolling physicians were conducting clinical practice as you described, that those masses that required ultrasound ‑‑ and I don't want to speak for our investigators; certainly two of them are here and can add to what I might have to say.
DR. CONANT: I agree very much with Dr. Hooley. It seems from looking at your data that there's a bias from the beginning because some of these patients definitely had ultrasounds, and cysts were seen, and in some cases possible solid masses were aspirated and they were found to be complex cysts, when they were thought to be masses, solid masses, but they still got the IR imaging and they're still included in the data.
Then there is a group that didn't get ultrasound factored into the level of suspicion, and biopsy was recommended, and they got IR imaging. So I think even within this data there may be a bias when ultrasound was used and when it wasn't. There wasn't a cut point as ultrasound definitely needs to be used to evaluate or should not be included.
I know from just our clinical practice we often, I believe, base our outcome, our category 0 through 5, on the ultrasound. We incorporate that into our 0-through-5 bottom line and recommendation. So I am concerned about the same thing.
The other thing that is sort of on that line is there were quite a few 0, 1, 2, and 3s included in the data, I believe. Those aren't cases that usually go to biopsy recommended by the radiologist. Now they may have palpable areas, but they don't seem to have been excluded. I may just be missing something here, but they did have IR imaging and they were included in the mass group, I believe. Is that true?
DR. PARISKY: Correct.
DR. CONANT: Why would a 0, 1, or 2 be included in biopsy? I know sometimes the patient drives at the doctor ‑‑
DR. PARISKY: One is that the radiologist felt that this was negative, and the surgeon or the patient pursued that something was there. More commonly, that happens with a 2. A 2 means that there's a benign finding, benign finding mammographically. It is still up to the referring surgeon and patient whether or not to remove something. I think you probably have that conversation with women on a daily or weekly basis where they say, "I want it out."
DR. CONANT: But if I am looking at it correctly, and please correct me if I'm wrong, the numbers of 0s, 1s, 2s, and 3s were really quite large. I'm looking at ‑‑
DR. HOOLEY: I see it's close to being 30 percent.
DR. CONANT: Yes. Page 29, in answer, response, to Question 9 from the FDA. It's in this white thing. It's about 30 percent.
DR. CALLAHAN: Okay, again, I would remind the reviewers that these BIRADS assignments were done by our independent evaluators based on the radiology reports. A lot of times there were zeroes because the radiology reports would say, "Need additional information." That would account for a lot of the zeroes.
We do know that there were palpable because the original protocol called for not just mammography, but was "and/or" clinical examination. We had a lot of palpable ‑‑ or we had palpable cases enrolled, where in fact they could have a 1 that is a negative mammogram, but a palpable finding that would lead them to go on to biopsy.
I think we tried to address the fact that the BIRADS recommendation and the ACR recommendations for treatment, and what's actually been described in the literature, is in fact that BIRADS in all categories do proceed to biopsy.
DR. CONANT: But not on a 30 percent rate. If they were palpable only and not mammographically-visible, they should have been excluded from the final mass category and then not reflected in this group.
DR. CALLAHAN: And they would have been excluded from the final category that could have been evaluated, because if they were palpable and there wasn't a mammographically-apparent lesion ‑‑
DR. CONANT: Even if on the ultrasound or in retrospect, sometimes you look back at the mammogram. I mean, it just seemed to be included in the end.
DR. CALLAHAN: Well, yes, our independent reviewers had to use the mammograms to assign and use an ROI. So I think that is the explanation.
CHAIRMAN MEHTA: Dr. Tripuraneni, do you have any questions?
DR. TRIPURANENI: Region of interest, when you had three independent reviewers pick the region of interest, was there significant variability?
DR. RUST: One analysis that we did perform, at the request of the FDA, was to determine if there was ever more than a 10-pixel difference between the ROI marker location and the center of the region of interest. We found that that never occurred. So there was never more than the 10-pixel difference.
DR. TRIPURANENI: Does the size of the breast have any impact on how it is cooled, on ultimately the results of positivity on the IR?
DR. RUST: We did not perform any analysis along those lines.
DR. TRIPURANENI: Would it?
DR. RUST: That's not a question for a statistician to answer.
CHAIRMAN MEHTA: Do you want someone else from the sponsor to answer the question?
DR. TRIPURANENI: Just curious.
DR. CALLAHAN: I mean that's a good question. We don't have any data on that, though, that we can report at this time, whether breast size makes a difference. We did not collect that information.
DR. TRIPURANENI: You don't have any thermometry about the temperature of the skin on the breast? If you look at the subset analysis with the size of the breast, would there be any difference?
DR. CALLAHAN: Again, we didn't collect information about breast size. We only collected information about lesion size. So we don't have the data to address that.
CHAIRMAN MEHTA: Dr. Ibbott?
DR. IBBOTT: Well, that leads into a question that I have been thinking about for a few minutes. That is, you decided on an equilibration time of, I think, ten minutes. I wonder if that number was selected arbitrarily or if you have some data looking at the effects of things like room temperature or whether the patient ran up the stairs to the clinic before coming in for the exam, things like that.
It seems like equilibration time would be important, and I wonder how much effect it has on the exam.
MR. SATTERTHWAITE: Let me ask you to restate your question, so we've got it clear here.
DR. IBBOTT: Well, the short version is, did you determine the effects of environmental factors such as room temperature on the equilibration process before the IR exam was performed and, if so, what effect that might have on the quality of the exam?
MR. SATTERTHWAITE: Okay. Before the trial started, we recognized, in reviewing other work that was done, the problems associated with sensitivity in a device that was an IR imager, and recognized the fact there was a lot of variability for that kind of thing. That was one of the reasons why we both asked for an equilibration and introduced the coolant challenge.
The coolant challenge is, we believe, what helped us differentiate ourselves, our diagnostic test from those of the past that have recognized the very issue you are addressing, which is that there would be some variability to that. But doing the coolant challenge allowed us to confirm the kind of characteristics that we looked for.
DR. IBBOTT: So, in effect, that excludes, doing the cooling challenge then excludes any impact of differences in the equilibration time? I'm just thinking of different cities. I work in Houston where it's obviously frequently very warm and humid, but at the same time we know how to air condition in Houston. People often complain.
So if somebody is coming in out of ‑‑ I don't need to go through that. You know what I'm trying to get to.
My point is, does the cooling challenge then negate all of those other potential influences?
MR. SATTERTHWAITE: We don't have the data to really address your question so specifically; only to tell you that we recognized the fact that that would be a real issue, and that we needed to do some things that would allow us to try to offset those kinds of problems and think that we probably still could do better, but we think that we have done an adequate job now to be able to identify those things that differentiate.
DR. IBBOTT: Thank you.
CHAIRMAN MEHTA: Dr. Hooley?
DR. HOOLEY: Yes. I would also like to clarify how you chose the threshold of the index of suspicion. My understanding is that you used a number of subjects and you included all lesions: masses, architectural distortions, and calcifications. You came up with a number of 20.5-something and then applied that to determine mass suspicion. That seems to be incongruent to me.
DR. RUST: Okay. The threshold was determined prior to unblinding of the pathology for the original study. The analysis by lesion type that led to a focus on masses was done after unblinding of the original study.
We felt that we had no leeway at all to change that threshold after unblinding. So we stayed with that threshold. Even though it was established using masses and non-masses in the dataset, we didn't feel that there were any degrees of freedom in terms of modifying threshold, that we were basically obligated to stick with the algorithm and the threshold that was determined prior to unblinding the original study data.
I would like to, if I could, take this opportunity to say that our objective, when setting that threshold, was, as Dr. Sacks stated in his talk earlier, to try to achieve an approximately 99.3 sensitivity. When we did that with the training dataset, we attempted to do that with 75 percent confidence by using a simulation procedure, and that is how we came up with that threshold.
That was the criterion for establishing the threshold. That criterion was never intended to be a hypothesis to be used for performance evaluation, as I think Dr. Sacks indicated he thought it was. It was simply to set the threshold, and that was it. It was not intended to be used as a performance criterion.
DR. GENANT: Just a little bit of a followup on that issue, the threshold, for example: You indicated that you have not performed duplicate measurements or triplicate measures in an individual patient. Yet, when you move from some of the qualitative aspects of imaging to a purely quantitative aspect, where you have a threshold base, it becomes important to know that you have stability of your equipment and that there is reliable reproducibility in some fashion for the equipment itself, whether that be on phantoms or whether it be on ‑‑ ultimately, one would like to see some data on patients indicating that it is reproducible.
How have you established that the instrument with regard to a threshold base has reliability?
MR. SATTERTHWAITE: So I understand the question to be, how do we assure you that this device will operate reliably each time?
DR. GENANT: And consistently.
MR. SATTERTHWAITE: Consistently. Again, I'll reiterate that we have not tested a patient in more than one imaging session for that kind of reproducibility. But we initiated early on an early-morning system test that the technologist was required to run. They would actually fax the results of those tests into us. But it required temperature sensors to be placed and the cooling system to be turned on in a very specific manner. Then we would measure various factors that assured us that we would have success there.
A lot of that, not a lot of that, all of that now is automated and done without technologist intervention. All they need to do is invoke the software that controls cooling and computer imaging, and that takes place.
So we believe that we have implemented the tests that will assure us that we have images consistent enough that our algorithm and the application of the coolant challenge will take care of those differences that might exist there.
DR. GENANT: But you haven't applied this in any objective fashion to determine whether it actually works? I mean, for example, with a phantom of some sort or ‑‑
MR. SATTERTHWAITE: Well, that's true, not with a phantom, but a reproducible test that has ‑‑ this system here has a number of temperature sensors that are tied to the computer, and the computer monitors all those temperature sensors while this beginning-of-the-day test takes place. Those sensors are located so as to sense the breast chamber and some other areas that we know, if we measure them, we know what they should be consistently, will give us a confidence that that device is going to work like you're suggesting it should.
CHAIRMAN MEHTA: Mr. Stern?
MR. STERN: Yes, I'm just curious to know if there is a next-generation device on the horizon and if you can say anything about it.
MR. SATTERTHWAITE: I'm sorry, I'm multiplexing here.
Your question specifically again? I'm sorry.
MR. STERN: Yes, I was just curious to know if there is a next-generation device on the horizon and if you can say anything about it.
MR. SATTERTHWAITE: I guess we're allowed, since we have been asked? Is that right?
There is a next-generation device. This actually represents some part of the characteristics of the next-generation device in that it is lower than the device that was used in clinical practice but functionally the same.
We, indeed, are ‑‑ we do have phantom work that is going on right now. We do tests in our own research facility that we think will improve the kind of tests I have already suggested or told you that we are running with the software.
MR. STERN: Thank you.
CHAIRMAN MEHTA: Ms. Peters, do you have any questions?
MS. PETERS: Pretty much just some comments, and I don't know if I'm really addressing it to the Panel or to the presenters here. But in a medical setting you go from your primary care physician who has ordered your mammogram, I mean who has asked you to do your mammogram. If the results come back that there's something suspicious, then your primary care physician says, "Well, I'm going to refer you to a surgeon" ‑‑ this is how a lot of your health groups work ‑‑ "refer you to a surgeon to talk about whether you're going to have biopsy or what are your options."
So then is this in the realm ‑‑ what I am hearing is that this machine is in radiology. So does the radiologist reading a film see something there and suggest to your primary care physician that this can be done to determine whether you should have a biopsy or not or to help in the decision of whether to proceed to biopsy? I'm not sure where this falls in the patient going through the system, the health care system.
DR. PARISKY: As I understand, your question is, using a model where a patient is referred for a mammogram and then the primary care physician is the quarterback in directing the patient's care ‑‑
MS. PETERS: Right.
DR. PARISKY: -- and how does that fall in? I think that model is becoming less and less frequent in terms of breast imaging.
Your experience in San Francisco, our experience, I'm sure your experience is the same, where a lot of these decisions are being relegated to the radiologist.
I think Dr. Sacks talked about the trauma. I think it is unfair to do a test and send a patient back to her primary care physician, wait for that letter to get to the primary care physician, send the patient back for another test.
We try to encompass as many tests as we possibly can. The patient comes to radiology. Now insurance reimbursement, for instance, screening, it's tough to do tests the same day. But if the patient is referred back for diagnostics, I think the prevailing philosophy of most radiologists is to do as many tests as appropriate to come to the right answer.
MS. PETERS: So then the radiologist would ask for the patient to come back ‑‑
DR. PARISKY: Yes, Ma'am.
MS. PETERS: -- to have the additional work done?
DR. PARISKY: Yes, yes, yes, and actually, by policy and by law, we are the ones that are instituting the recall. We do the obligatory letter to your clinician, but we have to notify you that there is a problem.
MS. PETERS: Okay, it's coming.
Then my other concern was, when we were talking about the heat sensitivity of it, in the age range, the majority of your patients were between 40 and 60 years old, and that's time where some people are having hot flashes during that time. What impact might that have on the cooling and overall body temperature?
DR. CALLAHAN: I think that's an excellent question. It's something that we haven't formally addressed in this investigation, but it is certainly, you know, a question that bears looking at.
We do have, I mean as part of the information that's collected, we have the patients, we have all this information about the beginning skin temperature and then throughout time. So it's possible that sort of analysis that you're describing, I mean we could look and see if, let's say, those patients start at a higher temperature perhaps.
One thing I would add to what our engineer said is, for the cooling challenge, the ten-minute equilibration period, part of the protocol is that the room climate control be at a fairly tightly-controlled room temperature, from around 67 to 73 degrees. So we're not saying, have a patient come in and equilibrate at 90 degrees. I mean that's part of the equilibration.
CHAIRMAN MEHTA: I'm going to go ahead and ask a question regarding patient consent on this particular study. Obviously, patients were not deriving any direct benefit by participating in this study. So in the consent form and in the consent process, what were the patients told as to the rationale for participating in this study?
Were they, for example, contributing to the development of science? Were the data going to be used for the benefit of women in the future? Could you clarify that for me?
DR. HUGHES: Essentially, in the study we made it very clear to the patients that we were not going to use these results for their care. What I would tell a patient specifically is that we have a new device that we're testing, that it may help us tell what lesions are without having to have surgery done, that we will not have results for them because this was a blinded study, and that they would have this test done and would have no benefit whatsoever, but it may help them or somebody else in the future.
Women uniformly in my practice were happy to do this.
CHAIRMAN MEHTA: So would it be fair, then, to assume that every patient who signed on the study assumed that their data were going to be used for the benefit of science in the future?
DR. HUGHES: I would think so. That's what research is for.
CHAIRMAN MEHTA: Okay. Did the 275 women whose data were not initially used, were they informed that the data were going to be cut off because of a timetable deadline, and that only at the behest of the FDA for request of further data would the data be included in the analysis?
DR. HUGHES: I'm actually not sure when the deadline occurred for that. I can't answer that question. I would expect that any data we collected would have been used eventually, whether it was used within this study or not.
CHAIRMAN MEHTA: In the initial submission they were not included. It was only when the request was made to get more data were those data dredged out.
DR. CALLAHAN: These patients were all enrolled under the same protocol. We continued collecting data for those additional 275 patients, as we had ongoing agreements and enrollment ongoing at those sites, for the additional 275, with the intention, the intention was always to utilize that patient data for analysis purposes.
So there was no difference in the consent amongst those. They had IRB approval under the same protocol that was ongoing, and there was never any intention not to use their data or to use it for, you know ‑‑
CHAIRMAN MEHTA: If the intention was to always to use the data from those 275 patients, why didn't you simply wait until you had those data and then included them, rather than going back into a cutoff date?
DR. CALLAHAN: There were time responsibilities. There were obligations and reasons that we felt like we had collected enough data for an initial submission.
We also can, frankly, say that there was always a possibility that additional data would be requested. So by remaining blinded to this, we felt like we had valid cases, the same protocol, the same evaluation procedure, that could have been used to answer questions that arose after review of the first submission.
CHAIRMAN MEHTA: Well, my concern is this: This morning you told me that you were planning to enroll 3,000 patients, 600 in each institution times five, and that's not what happened. There were a few hundred at some; there were more at others. It wasn't a 600 times five. Three thousand patients were not enrolled; 2,400 were. Two hundred and seventy-five were cut off from the data and were subsequently used.
So I'm just concerned about the entire consent process of how this all happened.
DR. CALLAHAN: We note your concern.
CHAIRMAN MEHTA: Any other questions from the Panel? Go ahead.
DR. CONANT: I have a few questions about the false negatives and in terms of efficacy. I gather there was one in the summed group of masses. I am wondering if, No. 1, that false negative was read by one, two, or three readers, how many of them felt it was a false negative, and how many other possible false negatives there might have been read by one, two, or, you know, a non-minority number, or that the averages then pulled up above to a positive result?
DR. CALLAHAN: Okay, for that particular case, it was read by all three readers. I can tell you that all three got an IOS score close to our threshold, below the threshold but close to the threshold for that case.
DR. CONANT: Were there other lesions that were read by at least one, and how many other lesions were read by at least one reader as a negative?
DR. CALLAHAN: For our masses, if any of our total efficacy group of 490 and 105 malignants, any of the other 104 that were read, all the assigned evaluations would have been above the threshold. If, for example, when we looked at our microcalcification cases, when we were looking at all the cases, or in general, if there was one read below the threshold, that was scored as a miss, as a false. If it was, in fact, malignant, it would have been considered a false negative.
DR. CONANT: I guess it comes back to the pooling of the data, because some readers, if there's really one that says ‑‑ I mean it's all in the way you run the readers today, but I understand now.
DR. RUST: Well, let me clarify how that would show up in the summary data, if it occurred. If one reader had missed, had had a negative for a malignant mass, and there were two other readers that have a positive, then what you would have seen is 1.33 missed cancers. So it would show up in our statistics.
DR. CONANT: Did that happen often?
DR. RUST: It did not happen at all in the case of masses. All readers on all 104 other masses obtained a positive test result.
DR. CONANT: Okay. And then back to that little false negative one, it was a small lesion. I believe ‑‑ I have been searching for where I read about that case, but I think it was a 5-millimeter or smaller; I thought it was a 4-millimeter mass. Correct me if I'm wrong, but I thought it was a tiny one.
No. 1, I would like to know more about that case.
DR. CALLAHAN: It was actually a 1-centimeter mass ‑‑
DR. CONANT: It was a 1-centimeter mass?
DR. CALLAHAN: -- but I'll let Dr. Parisky talk about that.
DR. PARISKY: I do have the pathology.
DR. CONANT: That's a question: Was the size determined pathologically or mammographically of the lesions?
DR. PARISKY: Both. Mammographically, but in the pathology report there's an acknowledgment of size.
DR. CONANT: So for the comparison of grades ‑‑
DR. PARISKY: Yes.
DR. CONANT: -- I mean of sizes here in the data, is that pathologic or mammographic sizes?
DR. PARISKY: No, those are mammographic determinations, but in this particular case, which I had a tremendous interest in because you can imagine why, we had a pathological size as well.
DR. CONANT: Okay. Then I'm just wondering, your comments, clinically, why there were so few small lesions, two out of all of those cancers. Why do you think that is? I mean, we tend to find small cancers, at least histologically small. Sometimes they look bigger mammographically, but ‑‑
DR. PARISKY: You're talking about only two lesions below 5 millimeters in size?
DR. CONANT: Yes.
DR. PARISKY: Yes. I think the majority of mammographic lesions are greater than 5 millimeters, the great majority. So we're looking at people reacting to a 3- or 4-millimeter may very well BIRADS code 3.
DR. CONANT: Yes. Well, it might be ‑‑ I think if you threw the ultrasound side in there, you would start seeing some more.
DR. PARISKY: We would love to throw the ultra ‑‑ no, I understand that, but do you react to every 3-millimeter lesion you see if it's, let's say, smooth?
DR. CONANT: If it's a change, definitely.
DR. PARISKY: No, de novo. You may very well ‑‑
DR. CONANT: No, I ultrasound them.
DR. PARISKY: You ultrasound them. You don't see it ultrasoundographically. You have a decision tree which you're going to follow in six months or needle localize it or maybe do another adjunctive test.
DR. CONANT: I'm just asking.
DR. PARISKY: And I'm getting back to you ‑‑
DR. CONANT: So then I've got another question.
DR. PARISKY: Go ahead.
DR. CONANT: I'm interested, and I may have missed it in the protocol, how you dealt with that contralateral breast. Because we know there are things lurking in that contralateral breast, and there must have been lots of positives or things that lit up, and it was used as a control ‑‑
DR. PARISKY: You only light up if you look, and this is not a screening tool. This is a targeted ‑‑
DR. CONANT: But aren't you using it as a control to the involved breast, and, therefore, can't it cause a problem in calculating? If one's hot, but it's not the one that I'm recommending a biopsy on, my comparison, therefore, may be askew, and ‑‑
DR. RUST: The theory here is that we are using the entire contralateral breast area, the largest circle you can fit ‑‑
DR. CONANT: An average.
DR. RUST: -- inside it, and average across that as the ‑‑
DR. CONANT: Not the quadrant, matching or ‑‑
DR. RUST: That's right, as the control parameter, and that the effect, if there were a lesion there that would have an effect, there's a small local effect that does not have enough of an effect on the entire breast reading to negate its value as a control. That's the theory.
DR. CONANT: Okay, that's an interesting theory to pursue. I mean just as the hormonal, the cyclical changes, we know from MRIs that there's significant vascular changes throughout the cycle, and I assume that they are symmetric, but I'm not sure we know that.
CHAIRMAN MEHTA: I think we have about 45 minutes left in the discussion time. So what I would like to do at this point is to ask Dr. Phillips to sequentially project the discussion points that the FDA would like addressed by the Panel.
Copies of these, again, are in the folder, so we can look at these as Dr. Phillips is pointing the questions to us. We can continue the discussion once the questions have been pointed out.
DR. PHILLIPS: Do you want all of them or do you want ‑‑
CHAIRMAN MEHTA: Well, we could do them one at a time. If you want to project them, maybe we could just use these as discussion points.
I don't think that it is necessary to read this out. Everyone can read this, and you have it in your handout. So if you would just take five ‑‑
DR. SACKS: The record needs it.
CHAIRMAN MEHTA: Oh, okay, I guess we need it for the record.
DR. SACKS: Yes, the record needs it. All right, go ahead.
DR. PHILLIPS: I'll do it.
"No. 1, clinical data: (a) The data in Amendment 4 were selected retrospectively from the original PMA dataset, albeit based on lesion-type analyses that were prospectively planned for in the clinical trial protocol. Aren't the data from Amendment 4 applicable for the assessment and determination of the effectiveness of the BCS 2100?"
CHAIRMAN MEHTA: Who on the Panel would like to start either asking questions about this or providing opinions or discussion points about this?
DR. GENANT: Well, I've already commented on this, but I could just reiterate it. I mean, I think that these are not valid in and of themselves to document effectiveness. I think they are encouraging, and they're in the right direction, but I think that they do, in fact, represent quantitatively a post hoc analysis and, as such, need to be applied to an independent set of data and show that it's statistically-significant.
CHAIRMAN MEHTA: Why don't we ask our statistician, Dr. Toledano, about this as well?
DR. TOLEDANO: I was really hoping you wouldn't make me answer this one.
Okay. Prospectively, there was a plan to analyze performance by lesion type. Prospectively, there was no plan to exclude certain lesion types from the indications for use.
The current indications for use are based on masses. You need data and analysis for masses, and that can't come from the data that told you to limit to masses. Sorry.
MR. SATTERTHWAITE: Can I ask a clarifying question?
CHAIRMAN MEHTA: I think if the Panel members want a question answered, they will specifically request you to do so.
Any other comments from Panel members?
DR. CONANT: A quick comment: I would be interested in this prospective data collection, looking at the group of masses already analyzed, sliding the threshold to look at different sensitivities and specificities, and then assigning a threshold to the prospective data collection. Because I think that the threshold was established after the fact and after the analysis, the threshold for the IR numbers.
Well, it was from multiple lesion types, and it wasn't specifically for masses. But I think that going back and looking at data you've already accrued and the hard work you've done, and look at where that threshold may be ‑‑ it may be different for masses, as Dr. Hooley mentioned, and then determine that, and make a plan moving forward to gather more data to test that hypothesis.
CHAIRMAN MEHTA: Yes, go ahead, Alicia.
DR. TOLEDANO: If that recommendation to develop a new threshold based on the indication solely for masses, and using the data available to you, is taken, one suggestion would be to do this with higher power. Your current power for establishing the minimum sensitivity with that threshold is power of 75 percent. So you might want to be more certain that you would achieve the necessary sensitivity, perhaps a power of 80 percent, perhaps a power of 90 percent.
CHAIRMAN MEHTA: Any further discussion on this point? Any comments from the Panel members?
You can go to 1(b) then.
DR. PHILLIPS: "1(b) The additional data in Amendment 5 consists of 78 masses. Are these additional data by themselves sufficient for the assessment of determination of effectiveness of the BCS 2100?"
CHAIRMAN MEHTA: Should we start from the other end of the table this time? Any views from anyone? Prabhakar?
DR. TRIPURANENI: I think a simple answer is no. We just don't have enough number of patients to actually confirm the data. I think that Dr. Sacks has quite nicely shown that the variability is quite large. I don't think that by just looking at this small number of patients, compared to the bigger picture, that one can draw a conclusion on the effectiveness.
CHAIRMAN MEHTA: Alicia?
DR. TOLEDANO: Given that it is too small and that you would need more patients, we should also consider one of the points raised by Dr. Sacks, that the key measure of diagnostic accuracy here is specificity, establishing specificity.
So we don't need to worry about area under the ROC curve. That blends sensitivity and specificity. We're not concerned about sensitivity too much. We're not concerned about AUC too much. We're concerned about, can you be specific?
So when you consider how many patients you need data on, look at adequate power to establish specificity. That's just a comment.
CHAIRMAN MEHTA: Any other comments on 1(b)?
Hearing none, we can move on to 1(c).
DR. PHILLIPS: "1(c) When combined, Amendment 4 provides 84 percent ‑‑ in other words, 412 ‑‑ of the masses, and Amendment 5 provides 16 percent, or 78 of the masses. What is the validity of combining these data to assess and determine the effectiveness of the BCS 2100?"
CHAIRMAN MEHTA: Anyone? I'll put in my two cents, and then we'll have Alicia give the statistical version of this.
I don't think you can make a yes by combining two noes.
DR. TOLEDANO: Not valid.
CHAIRMAN MEHTA: Okay, we can move to Question 2 then.
DR. PHILLIPS: Question 2 is the same series of questions but related to safety. "2(a) The data in Amendment 4 were selected retrospectively from the original PMA dataset, albeit based on lesion-type analyses that were prospectively planned for it in the clinical trial protocol. Are the data from Amendment 4 applicable for the assessment and determination of safety of the BCS 2100?"
CHAIRMAN MEHTA: Perhaps we could have our radiology experts comment on the safety aspect of this device.
DR. CONANT: Well, I'm still concerned about the false negatives certainly in terms of safety and the false sense of security potentially for lesions read as negative. I don't think I have the data to really base that on right now from the limited numbers. So I'm not sure that I can really address safety at this point.
I mean, no adverse effects, that's really great, but I'm more concerned about prolonged followup of malignant lesions, delaying diagnosis. I think with larger numbers, there is potential for even ‑‑ well, I'm not sure what would happen with larger numbers.
CHAIRMAN MEHTA: Well, I guess the safety question has three components to it: the false negative component, the false positive component, and the actual physical safety of the device itself.
I suspect in this particular component, if we start by limiting to physical safety, we might get to a quick answer, and then address the other aspects of it.
Any concerns about physical safety or performance safety, mechanical safety of the device? (No response.)
So the false negatives appear to be a concern. There is a question about false positives later on. We can defer to later.
Do other people on the Panel have concerns about the false negative rate and its implications for patients? Please go ahead and state your concerns, if you have them.
DR. GENANT: Well, to the extent that we don't feel that we have compelling data for the overall effectiveness of the device, and that false negatives are a concern, and this carries a safety implication, I have concerns.
CHAIRMAN MEHTA: Alicia, you were indicating some concerns?
DR. TOLEDANO: Yes, I'm concerned. I think we need sufficient data to establish high sensitivity, so that we have a very little, tiny false negative rate, and we need to establish, also, that the recommendation for short-term interval followup is an appropriate recommendation to preserve patient safety.
CHAIRMAN MEHTA: Let's move to Question 3.
DR. PHILLIPS: Question 3 or do you want to do (b) and (c)? It's repetitive of the earlier one.
"(b) the additional data in Amendment 5 consists of 78 masses. Are these additional data by themselves sufficient for the assessment and determination of safety of the BCS 2100?"
CHAIRMAN MEHTA: I think I can probably speak for the Panel ‑‑ you can correct me if I'm wrong ‑‑ that we have probably addressed (a), (b), and (c) collectively in the discussion for Question 2. Unless the Panel feels differently, we can skip to Question 3.
DR. PHILLIPS: Fine, we'll go to three. No. 3: "Please discuss whether safety and effectiveness has been established. As part of this, please discuss the risk/benefit tradeoff whereby a false negative results in a six-month delay of cancer diagnosis and a true negative obviates a biopsy that would otherwise turn out to be benign."
DR. HOOLEY: I think there's a high risk/benefit tradeoff with the amount of false positives. I think that the patients, about 80 percent of the patients with a false positive will be subjected to another test, a little more emotional turmoil during the time of their diagnosis. I think that that is a concern. I think that that is fairly high, and it is also going to add an expense.
CHAIRMAN MEHTA: One of the implicit suggestions in this question is that a false negative results in only a six-month delay in cancer diagnosis. Do we have data to support this? And is the Panel concerned about the fact that we haven't seen any data to suggest that it's only six months and not longer? Because we don't really know in clinical practice what that is going to be. It may be six months for most patients. We don't really know that.
MS. PETERS: Well, patients can tend to fall between cracks in terms of doing their own outreach for their health care. They can fall between the cracks. So it could be longer, I would think.
CHAIRMAN MEHTA: Go ahead, Alicia.
DR. TOLEDANO: I also don't know the natural history of DCIS to establish if a six-month interval is appropriate. Not being a radiologist, I asked some of my respected radiology colleagues, and I was told, no, we don't really know the natural history of DCIS. So I'm concerned about that.
I'm concerned about the recommendation for six months in the first place, and then I'm concerned about what happens with that recommendation in practice.
DR. HOOLEY: By convention, we usually use a two-year followup in the mammography to determine benignity of a suspected lesion. There's really no clear data on ultrasound followup, but generally we use about two, possibly three, years.
CHAIRMAN MEHTA: Prabhakar, as a physician treating breast cancer patients, would you be concerned about a six-month delay in diagnosis if, indeed, that was the case for every patient?
DR. TRIPURANENI: I think if you look at the bigger picture of all the breast cancer patients that are diagnosed, I think there is quite a variation. But in DCIS patients waiting for six months is probably not deleterious, as opposed to somebody who has a small infiltrating cancer that actually could blossom out in the next six months.
I think that's a rather difficult question, and I suspect if the device were to be approved and used in the clinic routinely, you do probably see a spectrum of the patients from one end to the other end.
CHAIRMAN MEHTA: So perhaps we can summarize the answer to that question by stating that, if, indeed, it were a DCIS, and if, indeed, the delay were only six months, which we don't know a priori, then we wouldn't be concerned. But if it were an infiltrating ductal carcinoma or the delay were significantly greater, which are real possibilities in the clinical world, then we would be concerned.
Go to Question 4. I'm sorry, one more point, Alicia.
DR. TOLEDANO: I did want to say that I think that the data does suggest efficacy, and very strongly suggest safety and suggest efficacy. So it is not, in my mind, a complete wash. There is a suggestion that the device is effective.
CHAIRMAN MEHTA: We can go to Question 4.
DR. PHILLIPS: "Is the proposed labeling adequate to ensure safe and effective use of the BCS 2100? Please include in your discussion the following specific items:
"(a) Given that only two of the 105 cancers were smaller than 5 millimeters, should the labeling specify a lower size limit for an eligible mammographic mass? If so, what size limit?
"(b) Should the labeling address lesion depth? If so, in what way?"
CHAIRMAN MEHTA: Let's take Question 4(a) first. Who on the Panel would like to try and deal with this question? Geoff, go ahead.
DR. IBBOTT: I will comment, but I'm sure Alicia should comment as well here.
And it's the same answer as before. There isn't enough data here to say anything about such small lesions. So, no, I don't think -‑ or, rather, I suppose the answer is, yes, it should specify a lower limit that is somewhere above the point where there are significant numbers of examinations.
CHAIRMAN MEHTA: Other opinions on this?
DR. CONANT: Not getting into size and depth so much, but more sort of general use of it, I hate to keep harping on ultrasound, but I would specify that this would be to be considered after completion of conventional imaging, which in this day and age means mammography and ultrasound.
There were too many cases excluded and too many cases imaged that didn't seem to get ultrasound until the fact, and I think that's a very confusing part of this data. So that I think that needs to be incorporated, that a mass is a mass, and it's determined solid by ultrasound. It's not an asymmetric density. It's not seen by ultrasound, but it, indeed, is there.
CHAIRMAN MEHTA: Can we get a statistical answer on two out of 105, Alicia?
DR. TOLEDANO: I think Geoff already gave us the statistical answer on two out of 105.
CHAIRMAN MEHTA: Okay. Let's talk about lesion depth then. Are there concerns? If so, in what way?
DR. HOOLEY: I have concerns about lesion depth. First of all, I think in a very large woman with pendulous breasts, I think imaging the patient in the prone positioning and imaging on the CTI table and imaging her upright with mammography would produce a large degree of discrepancy and introduce a lot of variability. So I think that needs to be addressed.
I am also surprised that depth was not thought to be a considerable factor, based on the design of the table. I know with my experience with stereotactic biopsy that lesions that are very far posterior or in the axillary tail cannot be imaged, and I would imagine that these would be excluded by the thermal imaging.
Also, women with very, very small breasts, there's just not enough tissue going through the aperture of the table to be imaged.
DR. CONANT: Have you found that an issue, the small breasts away from the detector, or does that make a difference at all? Because the distance of the pad is fairly substantial. I think distance to ‑‑
DR. CALLAHAN: Is this a question?
CHAIRMAN MEHTA: Yes, it's a question.
DR. CONANT: Yes.
CHAIRMAN MEHTA: Please, if you could answer it?
DR. CALLAHAN: We did look originally at depth, and that was assessed mammographically in our first, original study. We had the mammographers quantify depth as superficial, intermediate, or deep, and didn't see a degradation in performance.
We have subsequently looked at ‑‑ we've done a study of normals, not patients with going to biopsy, but looking at issues of physiological changes based on breast size and age, and we have not found that small-breasted or large-breasted women, that those pose particular difficulties with just the basic imaging process.
DR. PARISKY: Do you want me to comment on depth?
CHAIRMAN MEHTA: Yes, please, go ahead.
DR. PARISKY: Because of the references you made to stereotactic and lesion location, let me remind you we're not measuring the lesion directly. Let me remind you that the anatomy of the breast is likely looking at venous drainage, which in the breast goes to the surface.
So you are looking at lesions that, if malignant, their environment is affected by, in all likelihood, increased capillary flow, and that relatively coming to the surface. So we're not in terms of depth, while we did look at depth and found no degradation based on depth, what we really are looking at is warmer blood or greater volume of warmer blood in a cancer milieu coming to the surface.
I apologize for not saying that upfront during the comments about angiogenesis and such. I thought that that was fairly common knowledge.
DR. CONANT: But I would think that a very large, pendulous breast with a small posterior central chest-wall lesion would not bring blood flow so much to the surface?
DR. PARISKY: That would be speculative because we did not encounter a false negative invasive cancer in the hundred-and-some-odd that we looked at, of all the distributions, and I apologize.
DR. CONANT: No, that's the way they came.
DR. PARISKY: That's the way they came, and we got them all.
DR. CONANT: We have a lot of big-breasted women in our practice actually. Do you find that where you all are at the sites?
I mean, but it's a big issue. So I'm just wondering, are there women who can't fit in there?
DR. PARISKY: We had no exclusion, at least in my center we had no exclusion, and I had both the County Hospital with a large Latino population ‑‑ I'm not making any references, but I would think that in some instances ‑‑ we had a large Latino; we had a relatively large African-American and Caucasian population distributed along Los Angeles County, and we did not have any of those patients rejected because of breast size.
DR. CONANT: Great.
CHAIRMAN MEHTA: So I think the consensus Panel answer for 4(a) would be that we feel there aren't enough data and, therefore, we are concerned about making categorical recommendations for all sizes, and that we simply don't know enough about lesion depth to give an appropriate or an erudite answer to that question.
So we can move on to Question 5.
DR. PHILLIPS: No. 5: "Should the labeling be revised to address any potential psychological impact of a positive mammogram followed by a positive BCS 2100 result on a woman who does not, in fact, have cancer; i.e., false positive?"
CHAIRMAN MEHTA: Maybe we can have Ms. Peters get us started on this question first.
MS. PETERS: Well, this is always a scary thing for women. Having something that's going to or having a person having the time to reassure the patient that this does not necessarily mean that they do have cancer, but a lot of times time is precious and that isn't taken, so that patients can feel a lot of anxiety in the waiting and the not knowing, and their own assumptions, especially if they aren't into asking questions of health care providers about what is going on and they just take the word of the provider at face value and go home with their fears and anxiety.
I don't know how labeling that could help, except that the person who is using the instrument or making the decisions needs to have the time to give adequate information to their client patient.
CHAIRMAN MEHTA: Further comments from the Panel? Prabhakar?
DR. TRIPURANENI: No test is perfect. Every test has some false positives and false negatives. I think in this particular group of patients that a positive mammogram, presumably ultrasound is such that today they probably would have gone for a biopsy anyway, but for this test, and presumably the test will save some patients going to biopsy.
So I'm not too concerned about this question. I think that is the fact of life and medicine is such that there are going to be patients that will require biopsy, as statistics have shown over and over again. We do lots of biopsies to absolutely exclude the cancer.
It's the cancer that you miss, the importance that could have on the life is more worrisome than this.
CHAIRMAN MEHTA: Does anyone else have further comments on this?
I do have a point I would like the Panel to think about a little bit. One of the things that we have heard is that in this study and in real-life practice there are many women that have abnormalities that are identified and are thought not to be highly suspicious, and a recommendation is actually made based on available published information not to biopsy these.
Yet, a significant proportion of these women demand biopsies. As we heard in the context of this study, these biopsies get done primarily to alleviate anxiety and potentially because of the litigious medical climate that some of us live in.
As a consequence, this device, too, may face that same scenario, that this device might identify women who could potentially be saved a biopsy; yet, because of a considerable level of anxiety amongst these women, they may decide not to follow that advice and go on to biopsy and, therefore, reduce potentially the proportion of women who did not actually go on to biopsy.
So there is some potential psychological impact. Is it enough to require labeling information? Is it enough to actually do a study on what are the psychological consequences of this?
Go ahead, Alicia.
DR. TOLEDANO: No.
CHAIRMAN MEHTA: Does anyone else feel different?
Okay, then we have the answer to that question.
We'll move on to No. 6.
DR. PHILLIPS: No. 6: "Do the above or any other issues (a) require resolution before approval of the PMA; (b) suggest the need for a post-marketing study?"
CHAIRMAN MEHTA: Well, this is a good opportunity for the Panel to think about other issues that need resolution. So, as we have been thinking through this and discussing this, we have brought up many, many issues. Are there other issues? For example, there were some questions about reproducibility. There were issues about phantom studies. Do you wish to bring these up as issues that should be resolved prior to the approval of the PMA?
DR. GENANT: Well, I think that we certainly raised a number of really critical issues that need to be addressed and resolved. It is not clear that they can be resolved within the context of the clinical data that currently exists, but I think we need some answers to those issues before recommendation for approval.
CHAIRMAN MEHTA: Well, if you could at least even highlight some of those areas, that would be useful for the FDA, I think. So maybe we can give them a list of the concerns and areas that we think should probably be considered for resolution.
DR. GENANT: Well, two of the areas that I had raised before deal with reproducibility. I think we need information about the readers and their capability to reproduce the results, and we need information about, and some feeling of security about, the stability of the instrument and ways to assess its stability. I think we also need information along this line with regard to repeat measurements on the same individual.
CHAIRMAN MEHTA: Geoff, any QA issues?
DR. IBBOTT: Yes, I would repeat my request that there be some sort of quantitative QA procedures to ensure that, as has just been mentioned, the instrument does work reproducibly one day to the next, one patient to the next.
Oh, and I'm still concerned about the equilibration. Not every clinic is going to maintain the room temperature between 67 and 73. I think there should be some data to see what effect there might be of equilibrating the patient at 75 or 77 or some other temperature.
And my third comment would be the establishment of the threshold for the IOS. I think that is a serious concern.
CHAIRMAN MEHTA: Dr. Hooley, you had IOS special concerns as well. Do you want to expand on that a little?
DR. HOOLEY: Well, I think that we would benefit by having a direct comparison of a control of masses and determining the IOS threshold based on just masses, since that's what we want to target the use of the CTI for, just masses. So that the threshold should be focused on just masses, not a wide variety of lesions.
CHAIRMAN MEHTA: Go ahead, Alicia.
DR. TOLEDANO: I would like to see effectiveness established in a target population that is directly relevant to the proposed indications for use, taking into account clinical flow of the patients.
DR. CONANT: That's my biggest concern, is to really be very strict about the inclusion criteria and where imaging fits in with ultrasound, and, also, that for inclusion the BIRADS categories should really be limited to those recommending biopsy. Now I know that others go, but I don't think those are the ones that should be included in this prospective trial. They should really be 4s and 5s.
DR. TOLEDANO: I'm going to argue a little bit against Dr. Conant. Sorry, Emily.
Because we know that recommendation for biopsy does occur in women who are not BIRADS 4 and 5, I think we need to say, if the indication for use is recommended for biopsy, then you include recommended for biopsy. If the indications for use were BIRADS 4 or 5, then that's what you would include. But since the indication for use is recommended for biopsy, I'm fine with any BIRADS category as long as they have been recommended for biopsy.
DR. CONANT: So then you will be including patients with palpable lesions that have areas possibly not seen on mammography, that it has to be clarified that the lesion has to be localized by ultrasound and then labeled by whomever, so that the readers can go forward with that.
DR. TOLEDANO: Right. The current indication is mammographically-something mass. It's seen mammographically, or whatever it is. So it is in the indication for use.
DR. CONANT: Well, then that has to be clarified, whether it is just mammographic masses or things going to biopsy, which include palpable lesions as well. So just a clarification of that.
DR. TOLEDANO: I agree.
CHAIRMAN MEHTA: Any other issues that the Panel feels require resolution?
Six (b). Sorry, go ahead. I didn't mean to cut you off.
DR. TOLEDANO: So now that we have discussed what issues might require resolution, could I note that it might be possible to resolve some of those issues without additional formal clinical trial data? So we should be looking very carefully at which issues can be resolved without additional clinical trial data and which issues require additional clinical trial data. That is just a recommendation.
CHAIRMAN MEHTA: Thank you.
Six (b) "Do the above or any other issues suggest the need for a post-market study?"
DR. TOLEDANO: Yes.
CHAIRMAN MEHTA: Any other thoughts?
MS. BROGDON: Could I ask Dr. Toledano to clarify her answer?
DR. TOLEDANO: I guess I should clarify that, if it turns out that this company has to go back and do a whole other clinical trial for their pre-market approval, then no. But if it turns out that we are able to use what we know from the existing clinical data and from physics and from experience to posit answers that are good enough to allow approval, then we do need to see the actual empirical data, the clinical data, that bears that out in a post-market sense.
So does that help?
MS. BROGDON: Thank you.
CHAIRMAN MEHTA: Alicia, I'm going to ask you to keep on talking a little bit. So you've obviously made two suggestions. One is, assuming that there is a PMA approval, that you would suggest a post-market study, and that if there isn't a PMA approval, then obviously the post-market study question is a moot question because the idea would be that the study would be repeated.
Assuming that there is the need for a post-market study, do you have a suggestion what that should be?
DR. TOLEDANO: This is like "Study Design in 30 Seconds," right? Great.
Well, you start out looking at your indications for use, and perhaps before you design your post-market study or before you ‑‑ actually, it's post-market would come after approval, right?
CHAIRMAN MEHTA: Correct.
DR. TOLEDANO: Okay, so that's it. You're stuck with the indications for use. You can't change them. You have to use the ones that exist.
You look at the indications for use. You define a target population. You enroll that population.
You determine, what are you trying to establish for the effectiveness and safety? We have discussed safety means an incredibly high sensitivity. We've discussed effectiveness means a non-zero specificity, because all these women have already been recommended for biopsy.
You design up a study that is either based on the local reads, the real-time reads, like what Dr. Hughes is doing in practice, possibly supplemented by a secondary reader study that would allow you to look at inter-reader variability, possibly supplemented by studies that would allow you to look at intra-reader variability.
Use acceptable methods to design that study. Give it adequate power. Analyze them with methods for multiple-reader, multiple-modality, diagnostic accuracy data, and prove the indications, so much easier said than done.
I'll entertain a question.
DR. PARISKY: Would you be interested in us providing invasive and non-invasive malignancy because the results ‑‑
CHAIRMAN MEHTA: Please come up to the microphone.
DR. PARISKY: I appreciate all your comments.
How about also dividing, subsetting the masses we're going to look at into pathologically invasive and non-invasive? Because I think it has clinical implications.
DR. TOLEDANO: I think it depends on what you would like to say in the labeling. If you would like ‑‑
DR. PARISKY: Yes, because you can't predict ‑‑
DR. TOLEDANO: Okay, so you know?
DR. PARISKY: Yes.
DR. TOLEDANO: Right, you can't predict it.
DR. CONANT: You don't know that upfront.
DR. PARISKY: Huh?
DR. CONANT: You don't know that upfront ‑‑
DR. PARISKY: You don't know that upfront, but ‑‑
DR. CONANT: -- when you're scanning the patient.
DR. PARISKY: You don't know that upfront, but, I mean, I think the numbers here would have shown if we had divided them upfront between invasive and non-invasive. There would have been a much stronger argument.
DR. TOLEDANO: Right. So you can't say that in terms of designing up a trial, but what you can say is that you would be going for the possibility of putting something in the labeling that says: Having a false negative result is associated, was associated with non-invasive cancer in our study.
DR. PARISKY: That's what I'm getting at. Thank you.
DR. TOLEDANO: Okay.
CHAIRMAN MEHTA: So you could have a false negative for invasive malignancy. You could have a rate for false malignancy when there's a malignancy which would be a special rate that you don't want to exceed.
DR. GENANT: I think that we have raised enough serious issues with regard to the validity of the data in the various cuts that have been taken at the data, such that I don't think that it is appropriate to recommend a post-approval study at this point, because I believe that it will require actually new data to be obtained.
Basically, we saw that the equipment was changed to some extent during the study. We also had various amendments that impacted the validity of the analyses. We have considerable concerns about the manner in which the data were analyzed relative to clinical practice and whether it is really relevant, and problems with the matter of the imaging interpretation.
I think that, in sum, that we simply don't enough have a sufficiently high level of comfort with the documentation of the efficacy and safety to be making recommendations with regard to a post-approval study.
CHAIRMAN MEHTA: Okay. Alicia?
DR. TOLEDANO: I think that whether the study should be pre-market approval or post-market approval depends in large part on the extent to which we expect results from such a study to differ from the current results.
CHAIRMAN MEHTA: Okay. Seeing no further comments, I think we are coming to the end of the Panel discussion questions.
We have time for a short break to let people use the bathrooms and things of that sort. Please know that we are still in open Committee, so don't discuss things outside this room.
Let's plan on meeting back, say, in ten minutes, and then we will begin the open public hearing at that time.
(Whereupon, the foregoing matter went off the record at 2:56 p.m. and went back on the record at 3:08 p.m.)
CHAIRMAN MEHTA: Are there any open public members, speakers, who want to address the Panel?
Seeing none, before we move to the Panel recommendations and vote, is there anything additional the FDA would like to address?
DR. PHILLIPS: Yes. Somebody asked me on the way out how long it would take to get a cab here when you're over with.
I spoke to the desk. They said it could take as long as 30 minutes. If you talk to Tom, who is the gentleman in the white shirt behind the desk, he will help you get something expeditiously.
CHAIRMAN MEHTA: A very important piece of information. Thank you.
Is there anything else the sponsor would like to address to the Panel at this point?
DR. PARISKY: As an academic and practicing radiologist who is quite familiar with adjunctive testing in breast cancer issues, I would like to remind you that the thresholds that you set today for safety in terms of false negative cannot be met by any single existing modality, nor do I believe it can ever be met by any future modality, because the standard you have set is so high.
For example, PET scan, meta-analysis shows 5 to 6 percent false negative rate. MRI, if it's DCIS, it's upwards of 30-40 percent false negative rate. If it's invasive cancer, it's 10-15 percent false negative rate.
Let's take another. Let's take ultrasound, which we have talked about. You know the false negative rate for that. It is well below the so-called safety standard which you have set today.
So in your consideration, another aside, in terms of exclusion of data, the ATL presentation of their ultrasound PMA started with 1,200 subjects, then excluded cysts. They went to 400. Mr. Monahan will verify that. I believe those were the numbers. There did not seem to be an issue about exclusion.
You've taken it upon yourselves to set a standard today that will only allow the existing technologies, because they've already met approval, to persist and may very well have pulled the rug out for any other innovative technologies by setting that high a standard for false negative.
Nevertheless, I thank you for your time. I thank you for the exchanges and the ideas and the comments you have made. For myself, for a learning experience, and for my fellow investigators, thank you.
MR. BRENNA: Just a couple of comments myself, as the president of this company. We have worked very closely over the past couple of years with FDA. There was quite a bit of discussion and agreement along the way regarding all the amendments, the confirmatory studies, the confirmatory plan.
I am a little confused today, and you need to know that. You are looking at a product that is non-invasive, painless, safe. It is adjunctive. It is categorized as a product that will be under control of the radiologist as an adjunctive device.
It shows extreme promise with the sensitivity results. This gentleman is concerned about missing cancers. We can prove that our databases show at about 99.3 percent sensitivity and about a 19.2 percent increase in specificity over current biopsy methods.
Now I agree with Dr. Parisky that you need to consider the decision you are making today. This is new, innovative technology that has only one place to go over the next five-to-ten-year period.
If I go back 20 years ago and I look at a CT scanner from 1982, and I look at a CT scanner today in 2002, I say it has changed quite a bit, and there was dissension back then.
But that type of technology was allowed to enter into the radiology community. Through the product maturation process controlled by radiology, you brought it to where it is today.
I am just very concerned, and I share the same sentiment that Dr. Parisky did, and I thank you.
CHAIRMAN MEHTA: Any further comments from the sponsor? It is an opportunity to tell us anything additional that you might want to at this time.
Mr. Doyle will address the Panel.
DR. PHILLIPS: We will now move to the Panel's recommendations concerning PMA P010035.
The medical device amendments to the Federal Food, Drug, and Cosmetic Act, referred to as "the Act," as amended by the Safe Medical Devices Act of 1990, allows the Food and Drug Administration to obtain a recommendation from an expert advisory panel on designated medical device pre-market approval applications, PMAs as they're often called, that are filed with the agency.
The PMA must stand on its own merits, and your recommendation must be supported by safety and effectiveness data in the application or by applicable publicly-available information.
Safety is defined in this case, in the Act, as reasonable assurance based on valid scientific evidence that the probable benefits to health under conditions of intended use outweigh any probable risks.
And effectiveness is defined as reasonable assurance that in a significant portion of the population the use of the device for its intended uses and conditions of use, when labeled, will provide clinically-significant results.
Now your recommendation options for the vote are as follows, and there are three of these:
Approvable. That's straight approval if there are no conditions attached.
Then approvable with conditions. This Panel may recommend that the PMA may be found approvable subject to specified conditions, and these include physician or patient education, labeling changes, or further analysis of existing data. Prior to voting, all of the conditions should be discussed by the Panel.
And the third, not approvable. The Panel may recommend that the PMA is not approvable if the data do not provide a reasonable assurance that the device is safe or if a reasonable assurance has not been given that the device is effective under the conditions of use prescribed, recommended, or suggested in the proposed labeling.
If you should vote for non-approvable, the Panel will have to indicate what steps the sponsor may take to make the device approvable.
CHAIRMAN MEHTA: Would anyone on the Panel like to make a motion?
DR. TOLEDANO: I'll do it. I move for approvable subject to specified conditions. Do you want the conditions?
CHAIRMAN MEHTA: Is there a second?
CHAIRMAN MEHTA: Is there discussion of the main motion?
DR. CONANT: I'm still concerned that, not that there hasn't been some wonderful data and suggestions that this, indeed, will be helpful to some women, but I'm not sure that I'm at the point with the data presented that I can establish that.
I think there have been too many exclusions, too narrow a group, too many cases that I feel didn't follow a regular clinical flow. So that the amount of data I have right now, small number of cancers and very limited patients, I'm not yet comfortable with in terms of the effectiveness.
DR. GENANT: I would like to echo that. I also share some of the concerns with regard to the data that we have to analyze and feel that there is insufficient information and support for effectiveness to approve, to vote for approval.
CHAIRMAN MEHTA: Alicia?
DR. TOLEDANO: So I would subset that out because it is not straight approval, and we have been told that the specified conditions include further analysis of existing data. We have been told that our recommendation can take into account applicable publicly-available information in order to show effectiveness.
My opinion is, my very thoughtfully-considered opinion is that, with reanalysis of the current data in an exploratory manner, and with knowledge that already exists, we can establish a high probability of effectiveness for this device.
So that's why I make the recommendation of approval subject to conditions.
CHAIRMAN MEHTA: Is there any other discussion on the main motion?
If not, let's proceed to vote on the main motion, which was approvable with conditions.
All those members in favor of the motion for approval with conditions raise your hands, please.
(Show of hands.)
MR. DOYLE: Three.
CHAIRMAN MEHTA: For the record, we count three.
DR. CONANT: Can we try that again after we talk about the conditions?
MR. DOYLE: No.
DR. CONANT: Okay. I just thought I would ask.
DR. TOLEDANO: May I ask you a question of clarification?
CHAIRMAN MEHTA: Yes.
DR. TOLEDANO: Okay. In previous meetings when we have recommended approval subject to specified conditions, those specified conditions have reflected only changes in the labeling.
My understanding from what Mr. Doyle has just stated is that the conditions can include further analysis of existing data, and I just wanted to confirm that.
MR. DOYLE: That's correct.
CHAIRMAN MEHTA: I think that finishes the vote.
So is there anybody else who wants to make a different motion?
I was assuming that those who didn't raise their hands were against, but we need to clarify that that's really the case.
So all those members who are not in favor of the motion please raise your hands.
(Show of hands.)
MR. DOYLE: It's three to three. So you can see there's a tie vote, so the Chair can now vote.
(The Chair votes no.)
MR. DOYLE: Now I guess we need another motion.
CHAIRMAN MEHTA: We need a second motion at this point. Does anybody want to make a second motion?
DR. HOOLEY: I motion that the PMA is not approved because there are significant questions on the efficacy of the study and how it was performed and omissions in the clinical reality of how we work up breast masses.
CHAIRMAN MEHTA: Is there a second?
DR. CONANT: I second.
CHAIRMAN MEHTA: Is there discussion on this main motion? Go ahead, Alicia.
DR. TOLEDANO: May I discuss? One of the most difficult things that I have learned over three years being on this Panel, and previous experience with the Panel, is the difficulty of separating out the effectiveness of a device from what happens once the device is released into market.
I think that, as we make our motions and as we make our votes, we need to be considering the device itself.
CHAIRMAN MEHTA: Any further discussion of the main motion?
If not, let's proceed to vote on the main motion.
All those members in favor of the motion, which is for disapproval raise your hands.
(Show of hands.)
All those members against the motion for disapproval please raise your hands.
(Show of hands.)
I guess I get to cast the vote for disapproval.
What we are going to do at this point is we're going to poll all the voting members for the reasons for their recommendations, and we will also ask the industry and consumer representatives for their comments on the recommendations.
As part of your comments, specific statements regarding what it would take to obtain approval would be very useful at this point, specifically for the FDA and the company.
So perhaps we can just start at one end of the table and go around. Prabhakar, we can start at your end.
DR. TRIPURANENI: The reason I was in favor of approving with conditions is it is a relatively non-invasive machine, hardly any invasiveness. The patient comes in for ten minutes and then gone.
I think it does have its utility, and I think we have probably seen the first wave of these things. I think as the data gets finetuned a little more, as the clinicians get more experienced on the machine, the company gains more experience, I think they can finetune the data a little bit more. Perhaps presumably in the real clinic at this point somebody will get a mammogram followed by ultrasound, and probably this IR imaging at this point in time.
So, for all those reasons, even though the data is not as clean as I would like to see, but, once again, with all the vagaries of doing a clinical trial, being a clinician, I was in favor of doing this.
Now that officially the Panel has recommended, once again I abide by the Panel's majority disposition at this point in time that this is not approvable, but I think my own bias is that the second clinical trial, if there is going to be one ‑‑ I presume there will be one ‑‑ will be take the real-life situation such as a mass, perhaps followed by ultrasound and followed by this machine in some shape or form, basically, directly going right to what to do to get the final approval to be done.
CHAIRMAN MEHTA: Mr. Stern, do you have comments?
MR. STERN: Had I been able to vote, I would have voted with the doctor (referring to Dr. Tripuraneni). I believe that if some tens of thousands of women in America can be spared the psychological trauma, thanks to a negative reading with the BCS 2100, I'm for the PMA.
CHAIRMAN MEHTA: Geoff?
DR. IBBOTT: Well, I voted in favor of the first motion, but, like Dr. Tripuraneni, I'm quite comfortable with the approval of the second motion.
I have concerns, as you know, about this device. I am not concerned that any significant number of women will be injured or hurt by this device, but I do have the concerns that have been mentioned about the psychological effects of the results, but I also have the concerns that not very many women will be positively impacted by the use of the device. It is only a small number of women whose course of therapy appears likely to be changed, based on the data that have been presented.
As you know, I have concerns about the physics of the device, the reproducibility from a physics and engineering point of view, and quality assurance issues, and the concern about the procedure that was used to set the threshold for the IOS. I think that can be done better, perhaps with the existing data, perhaps not, but it does need to be done.
DR. HOOLEY: I think that a future study should be a prospective study which better characterizes the definition of mass, whether the mass is just clinically detected by only mammography or seen with ultrasound. I think the omission of ultrasound in the characterization of masses is significant, and I think that that should be addressed in the future.
DR. CONANT: I certainly look forward to this because I think it is a promising device, and I certainly do want to cut down unnecessary biopsies. However, I am concerned that a real-life population is women with masses felt on exam, not seen mammographically. I think in the case of the data presented this was an artificial exclusion.
I think that the power with a prospective study will be very convincing, and I look forward to that data. So I am very optimistic, but I am concerned about implementation in real life, and if this was to go out now, what it could be used for, masses, and that that is not as defined as it could be and I think should be, what a mass is.
I am also very interested in refining threshold for the IOS based on masses alone for this indication, rather than on all different lesion types, and moving forward to test that as a hypothesis.
The talk about sensitivity again confuses me because the sensitivity is 100 because our population is incoming with mammographic masses. Again, that is not reality. It goes back to the definition of a mass, and a mass on an exam, a physical exam, and I think that is a very important population to address. So the sensitivity part, I'm really looking forward to the improvement in your specificity that I think you may show us in the future.
Oh, the reproducibility of the exam which would be shown, hopefully, by inter- and intra-reader studies, I look forward to that data.
CHAIRMAN MEHTA: I think the sponsor has done an excellent job of building what looks like an exciting device. I think they mounted a clinical trial that was very broad, which in its original design had very limited quantitative analysis built in, which I suspect in hindsight was a statistical error.
It had a variety of concerns in terms of trial design, such as, for example, the inclusion of 600 patients, and then somehow 2,400 showed up, and then we were told maybe there was supposed to be 3,000, but, sorry, there's only 2,400, but, oh, by the way, it's not 600 per institution.
Then not all data were analyzed. There were many, many exclusion criteria, and eventually a subgroup was identified where it appears that there may, in fact, be up to a 15 to 20 percent benefit in terms of potentially delaying or avoiding biopsies, or at least allowing these patients to be screened closely.
This appears to be a finding in a subset of patients. In fact, if, indeed, this device is going to benefit this group of women, then a properly-designed clinical trial in this subset of women should be conducted to verify that this was not artificial finding as a consequence of this broad study, but a real benefit in these women.
After all, we are talking about a target population of half a million women. Before we allow a half million women to be subjected to this, let's be absolutely sure that this benefit is real. And that's the reason I voted for disapproval.
DR. TOLEDANO: So it's my turn now. I made a motion to approve subject to conditions, and those conditions, as I brought up in my request for clarification, I greatly wish I could have stated those conditions before we took the vote. Unfortunately, that's not allowed.
I did bring up something in the clarification, that the conditions could be different from the usual things that we see approval with conditions, change the labeling.
And I appreciate all the issues that have been brought up about clinical practice and I value everybody's opinions and their experience. But when I look at this, what I see is somebody who has a product that could work, and I see data that shows that it probably does work, and the data may not have been gathered in the optimal way, and it may not have been analyzed in the optimal way, but I, honest to goodness, believe that if you go out and collect new data in the optimal way and analyze it in the optimal way, you're going to come up with the same answer.
To me, that doesn't mean sending somebody back to the drawing board. To me, that means saying I think we're going to come up with the same answer, approvable subject to conditions.
Look at your data the right way. Go through the literature. Go through your physics. Tell me what you think is going to happen. And if you can prove to me that your device is going to be effective, get your approval and do a post-market study.
So that was the reason for my motion. Unlike my esteemed colleagues, who I have really come to enjoy, I am not comfortable with the recommendation to disapprove.
DR. GENANT: I basically agree with the comments made by our Chairman. I thought that he captured my own feelings about this issue very, very well.
Perhaps, in addition, maybe I would just comment. To the sponsors, I think that you have a very exciting technology, and there are definitely great possibilities.
I think it represents a refinement and a substantial advance over earlier work that was done with thermography and that has kind of lingered under a cloud for some many years. I think you have the opportunity now in designing prospectively a study that will address the various issues that we have raised and will bring to the larger community somewhere down the road, hopefully, not too far down the road, a technology that will, in fact, bring benefit to women in this particular setting.
MS. PETERS: If I was able to vote, I would have voted for the PMA with conditions, approval of the PMA with conditions.
It is an adjunct therapy. It is not used as screening, which would make it what I think some of that would really be necessary for. But I think with its being non-invasive, that with education and some of the additional changes in looking at the data, it would make it very good.
MR. DOYLE: Before we adjourn for the day, I would like to remind the Panel members that they are required to return all materials they were sent pertaining to the PMA itself. Of course, the list of Panel members and agendas, and so forth, you're welcome to keep.
Any materials you have with you may be left at your table. Any others that you may not have brought, you can send back to me at the FDA as soon as possible. Thank you.
CHAIRMAN MEHTA: At this point I would like to thank the speakers and the members of the Panel for their preparation and participation in this meeting.
I would also like to thank the sponsors for being here to present the data to us, and to all the members for attending.
Since there is no further business, I would like to adjourn this meeting of the Radiological Devices Panel. Thank you.
(Whereupon, the Committee was adjourned at 3:35 p.m.)