May 17, 2005
Gaithersburg Holiday Inn
Obstetrics and Gynecology Devices Advisory Panel Meeting
May 17, 2005
Kenneth L. Noller, M.D.
Tufts University Medical School
Evelyn R. Hayes, Ph.D., R.N., CS-FNP
College of Health and Nursing Sciences
University of Delaware
Paula J. A. Hillard, M.D.
University of Cincinnati College of Medicine
Hugh S. Miller, M.D.
Arizona Health Science Center
Jonathon W. Weeks, M.D.
Maternal-Fetal Medicine Center
Marcelle I. Cedars, M.D.
University of California
San Francisco, CA
Ralph B. D’Agostino, Ph.D.
Amir H. Gandjbakhche, Ph.D.
National Institutes of Child Health and Human Development
National Institutes of Health
Melvin V. Gerbie, M.D.
Northwestern University Medical School
Yulei Jiang, Ph.D.
University of Chicago
Thomas M. Julian, M.D.
University of Wisconsin
Joseph S. Sanfilippo, M.D.
Magee Women’s Hospital
The University of Pittsburgh School of Medicine
Russell R. Snyder, M.D.
University of Texas Medical Branch at Galveston
Phillips Medical Systems
Michael Bailey, Ph.D.
CALL TO ORDER
Panel Chair Kenneth L. Noller, M.D., called the meeting to order at 8:34 a.m., to review a PMA for the MediSpectra LUMA™ Cervical Imaging System (P040028). Panel members introduced themselves. Panel Executive Secretary Michael Bailey, Ph.D., announced that the next Panel meeting would be June 23, 2005; meetings for the remainder of the year are tentatively scheduled for August 15-16 and November 14-15.
FDA PRESENTATION: Condition of Approval Studies: Recent CDRH Changes
Susan Gardner, Ph.D., Director, Office of Surveillance and Biometrics (OSB), reported on a major programmatic change in CDRH condition of approval (CoA) studies, including the move of the Condition of Approval Studies Program to her office. Her office plays a role in premarket reviews, conducts signal detection to monitor the postmarket arena, assumes responsibility for characterizing device risk by reviewing postmarket data, and interprets medical device reporting regulations. The legislative authority for their work comes from 21 CFR 814.82(a) (2).
This shift began a few years ago when CDRH decided to evaluate its CoA program, initially to look at the quality of the studies required for condition of approval. They noted that 45 of the 127 PMA studies approved between 1998 and 2000 had CoA studies. They also realized that they had no standardized tracking process for this area, and high turnover among reviewers. There were no results for 22 percent of studies.
The official move to OSB was January 1, 2005, preceded by a 3-year pilot program. One of their first steps was to develop an automated tracking system to help them acknowledge to industry that they have received the report and conduct any necessary follow-up. The second change added an epidemiologist to the PMA team to develop the postmarket plan during the premarket review.
Dr. Gardner said that the changes should provide industry with increased motivation to carry out the CoA studies. The office will post the status of these studies on the CDRH website. As well, under Section 522, they have the ability to mandate postmarket studies; this will allow them to apply civil penalties to incomplete studies. The Panel will be asked to provide feedback and advice on better approaches to finish the studies. Her office will also provide the Panel with an update on the progress of the CoA study.
Colin Pollard, Chief, Obstetrics and Gynecology Devices Branch, made a few introductory remarks to the Panel. He noted that the efforts for this technology dates back several years; the Panel met several years ago to draft a guidance document on this technology. The sponsor has been collaborating with the FDA for a number of years with respect to this device.
Panel Secretary Michael Bailey, Ph.D., read into the record the conflict of interest statement. A waiver has been granted to Dr. Sanfilippo for his consulting interests with a subsidiary of a competing firm. He was allowed to participate but could not vote in the day’s deliberations. The Agency took into consideration certain matters regarding Dr. D’Agostino’s past personal interests and his current employer’s interest with firms at issue, but in matters not related to the day’s agenda. He was allowed to participate fully in the day’s deliberations. Temporary voting status for the extent of the meeting was granted to Drs. Cedars, D’Agostino, Gandjbakhche, Gerbie, Jiang, Julian, and Snyder.
OPEN PUBLIC HEARING
Dr. Noller read the open public hearing speaker statement. Diane Solomon M.D., National Cancer Institute, NIH, DHHS, began her presentation on data from the ASCUS/LSIL Triage Study (ALTS) study. She noted that the ALTS clinical trial received support in the form of equipment and supplies at below cost from Digene, Cytek, Tripath, and National Testing Laboratories.
The study enrolled just over 5,000 women with a community cytology diagnosis of either atypical squamous cells of undetermined significance (ASCUS) or low-grade squamous intraepithelial lesions (LSIL). The trial was a randomized with a 2-year follow-up. The women were divided into three management arms: immediate colposcopy; human papillomavirus (HPV) triage; and conservative management.
The ALTS trial results were as follows: no test was completely able to detect cumulative cervical intraepithelial neoplasma 3 (CIN3); colposcopically-dervived biopsies were less sensitive that previously assumed—a finding she said was a surprise; and undetected CIN3 did not regress, while CIN2 showed significant regression. The ALTS results demonstrated that colposcopy is a weak link in the screening chain of events. CIN3 detection was essentially the same in the immediate colposcopy arm versus the conservative management arm. She contrasted this with detection levels of CIN2. Undetected CIN2 actually regressed, and CIN2 detection was 9.9 percent in the immediate colposcopy arm versus 5.8 percent in the conservative management arm. This demonstrated that CIN2, as compared to CIN3, is not a precancerous condition but more likely a mix of high-grade and low-grade disease that looks severe, but is destined to regress.
Mark Schiffman, M.D., M.P.H., National Cancer Institute, NIH, DHHS, compared data for first and second biopsies taken at colposcopy in the ALTS trial. The study looked at what a second biopsy added when done with colposcopic direction and no adjunctive device. They discovered that the first biopsy did not yield any more significant information than the second biopsy. However, the severity of the two biopsies was positively associated (weighted kappa=0.37; 95 percent confidence interval=0.29-0.45). Therefore, the investigators focused on looking at the second biopsy and its effect on sensitivity.
Their results show that there is a clear need to improve colposcopic performance and localization of biopsy sites—this was a major concern. They realize that any additional biopsies will increase sensitivity to some degree, but they question what the proper comparison is for any device. They believe that a proper comparison would not be random punches of normal epithelium.
The Panel asked for clarification on whether there was any effort to improve the performance of the pathologies at the initial reading. Dr. Solomon said they that they did have a pathology quality control group. This group tended to call more CIN3s than CIN2s, but there was no significant increase in sensitivity with this group. In answering a question about procedure, Dr. Schiffman said that colposcopists were instructed to do the worst site first. A Panel member asked about the qualifications of the ALTS colposcopists; Dr. Schiffman said that they were screened by a quality control group and underwent extensive training.
Theresa Wingrove, Ph.D., Senior Director, Regulatory and Clinical, MediSpectra, offered a regulatory and clinical history of the LUMA™ Cervical Imaging System (P040028). Their collaboration with the FDA on this device started in 1998. The FDA issued a nonsignificant risk determination letter in 2001, and the Pilot Study with 753 patients ran from May 2001 through February 2002. The first Pivotal Study was held from August 2002 through September 2003. This study enrolled about 2,300 subjects and was designed as a randomized control trial that compared LUMA and colposcopy versus colposcopy alone.
The second Pivotal Study ran from July 2003 through October 2003. It was designed as a multi-center, single-arm study, with more than 700 expected subjects; however, it was terminated early with 227 patients by the sponsor’s Board of Directors for solely financial reasons. Most of the data was still in the progress of being collected, so those making the termination decision did not have data from the second study; they did have the results from the first study. In June 2004, the sponsor filed the PMA and it was assigned expedited review status.
The current indications state that “LUMA is indicated for use as an adjunct to colposcopy in identifying high-grade disease (CIN2, CIN3+) in patients referred to colposcopy with ASCUS or LSIL. LUMA is not intended to replace colposcopy. A thorough colposcopic evaluation with an identification or selection of biopsy sites must be performed independently and prior to the viewing of the LUMA result.” The indication has been changed from the original in two ways: the adjunctive nature of the product has been stressed, and the indication has been limited to ASCUS and LSIL subjects.
J. Thomas Cox, M.D., Director of Gynecology Services, University of California, Santa Barbara, spoke on the clinical need for LUMA. He said that the use of LUMA as an adjunct will increase the sensitivity of colposcopy and reduce its subjectivity. He differed with the NCI speakers’ analyses of the ALTS data; he believes that colposcopists always have the option to take additional biopsies, but a device such as the LUMA will help draw attention to any areas they may have missed.
The ALTS trial concluded that the imperfect sensitivity after colposcopic referral in all arms of the study is of concern. For the patients in the ALTS immediate colposcopy arm, 21 percent were found to have CIN2, 3+; of this number, about one-third were missed at initial colposcopy, a significant burden of high-grade disease, he said. Many women probably remain at risk who are receiving routine care. Other randomized trials have shown similar results. Therefore, 7 percent is probably the upper limit of possible CIN2 and 3 detection after being missed at initial colposcopy.
Ross Flewelling, Ph.D., Chief Technical Officer, MediSpectra, provided an overview of the product’s technological features. He noted that the scientific basis for tissue spectroscopy has been established and well-investigated for decades. There are currently three FDA-approved optical devices that assess in vivo tissue dysplasia in endoscopy and bronchoscopy.
The LUMA uses three optical measurements: fluorescence excitation, reflectance white light backscatter, and video imaging. The device’s noncontact probes works at a distance of about 100 mm from the cervix, outside of the speculum, and takes about 12 seconds to complete a scan. The scan area has a 25 mm diameter, with 499 points at a 1 mm resolution in a dense hexagonal path array. There is a disposable cover on the probe. The LUMA’s algorithm is composed of the three optical measurements, and was established during the 2001 clinical Pilot Study. This produced pathology-qualified measurements on 1,569 samples that could be correlated with the device’s optical measurements.
Joan L. Walker, M.D., Section Chief, Gynecologic Oncology, University of Oklahoma, worked on ALTS trial from 1995 to 2000 as a clinical center principal investigator, and has worked with MediSpectra on their trials of the device. She said that the device has helped her to become a better colposcopist. She presented a video that covered the basics of how the LUMA is used as an adjunct to colposcopy. The patient is prepped for the procedure and the LUMA is calibrated for each individual patient. The speculum is placed in the vagina and dilute acetic aced applied to the cervix. A timing device is initiated on the device using a foot pedal, to make sure the scan is performed both after the acetic acid has developed any changes in the cervix and before the effect has disappeared. The probe visualizes the cervix and then performs a scan that takes approximately 12seconds. The colposcopic examination can proceed as usual. After identifying colposcopically-indicated biopsy sites, the LUMA image is reviewed. Blue areas on the LUMA display indicate additional sites for possible biopsy.
Dr. Walker discussed the features of the LUMA device pivotal clinical studies. The two studies totaled 2,526 patients, and were conducted over 15 months at 13 U.S. practice centers, by 52 different colposcopists. The studies were fully IRB-approved, declared a nonsignificant risk, and were IDE exempt; the two studies produced similar results. Study participants were 18 years or older with abnormal Pap smears. Participants were excluded for pregnancy, heavy menstruation, Pap test within 7 days but no longer than 6 months, and a biopsy or treatment conducted since their last Pap test. All colposcopists were required to have had experience and formal training in colposcopy. The histopathology slides were reviewed by two pathologists; if they disagreed a third pathologist was consulted. The patients’ results were categorized as either with CIN2, 3 or not.
Thomas C. Wright, Jr., M.D., Director of Gynecologic Pathology, Columbia University, and Consulting Medical Director, MediSpectra, presented the results of Pivotal Study I, a dual-arm, randomized study. Participants were randomized to receive either LUMA and colposcopy (physicians saw the scan and could take additional biopsies), or colposcopy only. Patients were referred because of an abnormal Pap test result including ASCUS, LSIL, or HSIL. The two arms were well balanced, and there was strong minority representation in the trial. The study endpoints were true positive (patient had at least one biopsy diagnosis of CIN2, 3+) and false positive (patients had no evidence of CIN2, 3+ in all biopsies). The study had two hypotheses: the true positive rate would be greater for the LUMA +colposcopy group than the colposcopy-only group; and the false positive rate for the LUMA +colposcopy group would be no more than 8 percent greater than for the colposcopy-only group.
The difference between the two arms was a 1.9 percent increase in true positives overall, with a 3.0 percent increase in the ASCUS/LSIL group; this did not meet statistical significance. They recognize that examining the ASCUS and LSIL sub-group is a post hoc analysis, but their reasons for doing so were (1) there was a similar disease prevalence in each group, and (2) 80 to 85 percent of all colposcopic examinations are performed for these two cytology reports.. Ultimately, he said, while the absolute gain in the detection of true positives is 3.0 percent, the relative gain was 26.5 percent.
For false positive outcomes, there was an overall increase between the two arms of 3.1 percent (which meets the study hypothesis) and an increase of 4.0 percent for the combined ASCUS/LSIL group (which does not meet the original study hypothesis). This was a moderate increase in false positives. Dr. Wright suggested looking at the mean biopsy rates to understand further the false positive outcome. This analysis produced an overall difference between the arms of 0.27 biopsies and a difference in the ASCUS/LSIL group’s two arms of 0.31. He noted that the positive predictive value went from 10.8 percent for colposcopy only, to 18.1 percent for those biopsies that were also LUMA positive—this is a significant increase in the positive predictive value of biopsies taken on the basis of also being LUMA-positive. Because of the appearance of false negatives, Dr. Wright stressed that the role of LUMA is not to replace colposcopy but to serve as an adjunct.
Dr. Wright noted that there were no device-related effects in any of the patients; 4 patients did report cramps, bleeding, and other symptoms typically associate with colposcopy. He added that they do not see any age effect for the device; the FDA disagrees with their conclusions. If they break the age groups into the population’s true tertiles (<23, 23-33, >33 years), LUMA plus colposcopy has an increased true positive rate for identifying lesions with CIN2/3+ compared to colposcopy alone across all three age groups.
Brent Blumenstein, Ph.D., TriArc Consulting, discussed the joint inference testing used on three outcomes for the patients in Pivotal Study I: true positive, false positive, or no biopsy taken. According to his analysis, the Cochran-Mantel-Haenszel test (stratified) p-value is 0.011, and the exact test (not stratified) p-value is 0.013; these results for Pivotal Study I are significant. The joint inference testing provides evidence that there is significant difference between the two arms of the study.
Warner K. Huh, M.D., Gynecologic Oncologist, University of Alabama, Birmingham, presented the Pivotal Study II data, a single-arm study in which each patient served as her own control. A total of 227 patients were enrolled, and 197 were included in the final analysis. A LUMA scan was performed on the after acetic acid was applied to the cervix, but blinded to the investigator. Standard colposcopies were performed with biopsy commitments. The commitments were annotated on the screen and locked into place, and could not be changed later. The scan was unblinded to the investigator who was committed to take biopsies at LUMA-indicated sites; these sites were also locked into place. Biopsies were taken and sent to the reference pathologist.
The LUMA true positive rate was defined as the number of subjects with no positive biopsies recommended by the colposcopic examination (includes the no biopsy recommended group) that also have at least one positive incremental LUMA directed biopsy (one or more biopsy positive results) divided by the total number of subjects in the per-protocol population. The LUMA false positive rate was defined as the incremental false positive patients among the “colposcopy no-biopsy” patients, as a percent of the total population. The study hypothesized that the LUMA incremental true positive rate would be at least 2 percent for the overall population (the lower bound of the 95 percent confidence interval would be at least 2 percent); and that the LUMA incremental false positive rate would be less than 15 percent for the overall population (the upper bound of the 95 percent confidence interval would be 15 percent or less).
The study resulted in a LUMA increment absolute gain of 4.7 percent and a relative gain of 22.0 percent for the overall population; among the ASCUS/LSIL population, the numbers were, respectively, 3.6 percent and 25.0 percent. This showed that LUMA identified significantly more CIN2, 3+. The false positive rate increase was moderate, Dr. Huh said, and was expected given the study design. There were no device-related effects among the patients. Two patients experienced minor adverse events commonly seen in those undergoing a standard-of-care colposcopy.
Dr. Wright finished the sponsor’s presentation by stressing that the two studies were designed to measure only the adjunctive benefit of using LUMA with colposcopy. There were no device related events. There was remarkable consistency in the outcomes of both studies, including roughly the same relative gain. In addition, the 3 percent absolute gain produced in Pivotal Study I was nearly significant, and the results were statistically significant in the second study. If joint inference testing is used in the first study, there is statistical significance. In the second study, for the ASCUS/LSIL group and the overall population, they meet the statistical hypotheses for true positives, but not for false positives; he suggested that this may have been due to the instructions given to clinicians to take a biopsy whenever possible. This resulted in one additional biopsy per patient. He stressed that the LUMA device identifies the patients who need additional biopsies at the time of the examination, which the sponsor hopes will lower the rate of loss-to-follow-up of women with true cancer precursors who are missed by colposcopy.
PANEL QUESTIONS FOR THE SPONSOR
The Panel asked for clarification on several points made by the sponsor. A Panel member asked how the “eligible” number of patients was calculated for the second study. Dr. Wingrove answered that there were 193 subjects after exclusions; 152 subjects were found not to have CIN2, 3+ after colposcopy, referred to as “eligible.” A member suggested that another way to interpret the data from Pivotal Study I is that, the study is negative because the true positive failed. The joint analysis study doesn’t clarify this issue because it was done post hoc. The second study results look better, but only if arbitrary redefinitions are made post hoc. Dr. Wright suggested putting the study results into a more clinical perspective. The sponsors clarified a number of points: LUMA does not identify atypical vessels; if there was an exceptionally large cervix the central transformation zone would still be captured; LUMA produces a static image; and doing a Pap immediately before using the LUMA may produce excessive blood and could limit the area in which the device is effective.
Joyce M. Whang, Ph.D, Office of Device Evaluation, provided an overview of the devices’ FDA review and a brief regulatory history. In August 2000, the FDA determined that the LUMA™ Cervical Imaging System would be regulated as a Class III device under the premarket approval process. In February 2001, the FDA decided that the use of this device in clinical studies would be a nonsignificant risk; therefore, the clinical studies did not require regulatory approval from the FDA via the IDE process. There was no agreement letter for the first Pivotal Study, but there was one issued in February 2003 that described the single-arm Pivotal Study II. The PMA received expedited review in June 2004.
Dr. Whang described the device’s three main components: the console and accessories, the illumination probe; and the disposable probe cover. She covered what was involved in four of the six major review areas for the device. For the electro-optics, all review concerns have been adequately addressed. The FDA looked at algorithm development, especially the changes made to the algorithms between Pivotal Studies I and II. The FDA concluded that the use of the revised algorithm would not have reduced the improvement of true positives as seen with the LUMA in the first study, although it may have made the increase in false positives in the first study slightly larger. For bioresearch monitoring the FDA reported no outstanding concerns. For manufacturing, the sponsor has been found to have an acceptable program.
Julia A. Carey-Corrado, M.D., Office of Device Evaluation, reviewed the devices’ indication for use, highlighting three aspects: the device is to be used as an adjunct to colposcopy; ASCUS includes ASC-US and ASC-H; and the indication for use reflects the FDA’s “always/never rule.” This rule, or approach, states that the clinician must always identify colposcopically-directed biopsy sites prior to viewing the LUMA image, and the clinician should never eliminate a colposcopically-directed biopsy after viewing the LUMA image.
She compared the two studies’ designs, treatment groups, primary endpoints, sample sizes, and presence of agreement letter. She noted that the “always/never approach” was not followed in Pivotal Study I but was followed in Pivotal Study II. The denominators in the first study were the total number of subjects, as opposed to the total number of diseased subjects; she noted that the study designs made it impossible to know the true disease status of the subjects. As well, neither study was designed to determine the sensitivity of LUMA to detect CIN2, 3+ in patients who have the disease.
Dr. Carey-Corrado looked at subject accountability in the first study to show how many patients were lost from each study arm, and made the point that patients were excluded prior to and after randomization. The primary reasons for exclusion were device malfunction and no majority pathology diagnosis. For the first study, the only hypothesis met was the false positive endpoint for all subjects. The hypothesis was not met for the ASCUS/LSIL group. The LUMA false negative rate was 23 percent, compared with the colposcopy false negative rate of 15; this shows that the LUMA device did not identify some of the CIN2+ lesions that the colposcopist saw, and, similarly, the colposcopist did not see all of the lesions detected by LUMA.
Pivotal Study II followed the “always/never” approach. The FDA agreed with the sponsor’s design choice for the second study. The FDA wanted to see what benefit LUMA contributed after colposcopy. The study design did require additional biopsies based on the LUMA overlay if it indicated any suspicion of CIN2, 3+; there was some room, however, for individual clinician discretion. The study hypotheses required an incremental true positive rate of at least 2 percent, and an incremental false positive rate of <15 percent.
For the second study, the only hypothesis that was met was the true positive endpoint for all subjects. Neither hypothesis was met for the ASCUS/LSIL group.
Gene Pennello, Ph.D., Office of Surveillance and Biometrics, discussed additional analyses performed on the clinical trial data. He looked at how diagnostic endpoints are typically evaluated, compared with how they were analyzed in the two studies. He emphasized that it is clinically more important to detect more CIN2, 3+ subjects than CIN2, 3+ biopsies, which is why the primary endpoints are at the subject level. He also noted that nothing is known about the accuracy of the tests for those patients who tested negative, because there was no follow-up.
In addressing the design limitations of the two pivotal studies, he noted that the colposcopist in the first study was not blinded to which arm the subject was randomized. They knew, for example, when LUMA was used; this could have introduced bias because they may have been relying ultimately on the LUMA. In the second study, the subject was her own control; if the initial colposcopy was undercalled, there could have been bias in favor of finding an effect with the LUMA device.
He examined the joint analysis and the age effect in Pivotal Study I. The dual hypothesis was not met for all subjects or for the ASCUS/LSIL group; however, it is unclear whether the true positive difference is significant and whether the false positive difference is no more than 8 percent. Dr. Pennello said that there may be an age-related device effect. The FDA used Bayesian analysis shows that the device may be finding more disease in younger age groups.
In Pivotal Study II, there were two changes made in the protocol analysis: the target value for the ASCUS/LSIL group changed from 2 to 1.5 percent; and the denominator was reduced from all subjects to all subjects minus the true positive subjects already detected by initial colposcopy. For the first change, the dual hypotheses were not met for the ASCUS/LSIL group. For the second change, the analysis is invalid unless the hypotheses for the true positives and the false positives are both changed; this change does not coincide with sponsor’s agreement to use the full denominator.
Dr. Pennello briefly presented issues of concern that appeared in both studies: biopsy frequency, subgroup testing; site and colposcopist variability; and a per biopsy positive predictive value. Ultimately, neither study met both the true positive and false positive hypotheses in any of the populations. Combining data from both studies is problematic because the two designs are so different. The age effect that seems to appear in the first study does not appear in the second study, but this may be an artifact of the size difference between the two studies. He also noted that one could get a LUMA false negative rate of 23 percent based on colposcopy used as a reference for detecting CIN2, 3+ subjects, based on the Pivotal Study I results.
PANEL QUESTIONS FOR FDA
The Panel asked the FDA for some clarification about LUMA’s algorithm. Brandon Gallas, Ph.D., Office of Science and Engineering Laboratories, said that there are many components to the algorithm, and it can be very difficult to track one item in a coherent manner. Because of this, they cannot look to the algorithm to make judgments about the device’s effectiveness; they must rely on the clinical study to make that call.
PANEL QUESTIONS FOR SPONSOR
The Panel asked the sponsor to speak on the LUMA device’s sub-par ability to detect HSIL lesions, and whether the preponderance of academic medical sites limited the type of colposcopic experience in the study to highly experienced clinicians. Dr. Huh said that, according to the literature and their interpretations from the trial, HSIL lesions are typically more obvious. Therefore, the sponsor did not expect to improve detection in these cases substantially. Dr. Wright said that there were a wide range of colposcopists at the sites, but they saw no significant variability depending on colposcopists. He suggested that there may be less impact from the LUMA as the colposcopists’ skill increases.
Panel members expressed concern about the early termination of the second study. The sponsors stressed that Pivotal Study II was cancelled for strictly financial reasons, and not because of any findings from that study.
PANEL QUESTIONS FOR SPONSOR AND FDA
Panel members asked the sponsor about the lack of LUMA benefit with high-grade lesions, and whether the benefit is based on the prevalence of the disease in the population. Dr. Wright said that the prevalence of disease in patients with HSIL is different than those with ASCUS and LSIL. Colposcopists may perform differently because of the expectation of whether a high grade lesion is present; this may explain why the LUMA did not have as much of an effect as expected in HSIL in the first study.
There was also a question about older patients in the studies. Dr. Walker said that colposcopy is acknowledged to be more difficult in older women; in fact, the LUMA will not perform in older women because the transformation zone recedes into the canal of the cervix, making lesions after an abnormal Pap more difficult to find.
A Panel member asked about the longevity of the fiber optic system in the device. Dr. Flewelling said that the sponsor has done life testing on the fibers with a result of more than 10,000 shots, and that the fibers last for the lifetime of the product, more than 5 years. There is required yearly maintenance where the system is tested against standards. In addition, at run time, the device goes through a series of statistical tests to calibrate the system and verify that it is performing within normal operational parameters.
The Panel raised an additional question about patient age, noting that the study did not include patients under 18 years. Dr. Wright said that the sponsor is asking for labeling that includes only those age groups studied, so those under 18 would be excluded from intended us. Panel members were concerned that the disease was somewhat more prominent in younger women, and that the sponsor would be excluding younger women from device use. The sponsor noted that the indications for use currently do not have an age cut-off; if the Panel had a recommendation for language including younger patients the sponsor would abide by that. They expect the device to perform the same in women across all reproductive ages.
The Panel asked the sponsors how clinicians unfamiliar with the trial will be trained on the device. Dr. Wright said that the first part of the training program will be conducted on-site, during installation; the second training will consist of hands-on, interactive methods to show clinicians how to interpret the displays. He noted that the trial’s investigators were able to learn to use the device very quickly.
In response to a Panel member’s question about data for patients under age 18, Dr. Wright said that they had very few 15- and 16-year-old patients; therefore, they had no analysis specifically done for patients under age 18.
FDA QUESTIONS AND PANEL DISCUSSION
1. The two co-primary effectiveness endpoints of Pivotal Study I (PS I) pertained to true positive (TP) and false positive (FP) rates. Individual subjects were TP, FP, or negative as follows.
§ TP subject: one or more biopsies taken, at least one was CIN 2/3+ positive
§ FP subject: one or more biopsies taken, none was CIN 2/3+ positive
§ Negative subject: no biopsies taken
TP and FP rates were calculated by dividing the numbers of TP and FP subjects by the total numbers of subjects.
a. For the original study population (i.e. all per protocol subjects), the FP endpoint was met, but the difference of TP rates was not statistically significant. Please discuss the strength of these findings relative to the proposed indication, i.e. as an adjunct to colposcopy for use in patients referred with ASCUS or LSIL cervical cytology results after a thorough colposcopic evaluation with identification or selection of biopsy sites.
b. The original population in the clinical trials was women referred for colposcopy because of an abnormal Pap smear result (first-time ASC-US/-H, repeat ASCUS/- H, LSIL, HSIL, or squamous cell cervical cancer). The proposed indication is for a portion of this population, the ASCUS/LSIL Pap referrals. Please discuss the clinical significance of this proposed population.
c. For the ASCUS/LSIL sub-population, neither the TP nor the FP endpoint was met in PS I. Please discuss the significance of these findings. Note that the analysis for this sub-population was not pre-specified.
There was general agreement that one of the critical aspects of Question 1a was that the difference in true positive rates was not statistically significant. Other members expressed difficulty with the sponsor’s shift from the original population to the subpopulation, and lack of adherence to the original protocol. Members raised the issue of colposcopists receiving large numbers of false positives using the LUMA adjunctively without a corresponding significant increase in the number of true positives. Even though the increase was not statistically significant, one member said that it would still be an absolute increase in the number of women whose lives could possibly be saved. Another member pointed out that this would not be the last chance to pick up a patient’s disease; colposcopy is a secondary testing tool, and there would, at least, be repeat cervical cytologies because the patients were referred originally for an abnormal Pap. The LUMA would be a third level of testing. For Questions 1b and c, Panel members discussed whether screening fewer younger women with LUMA would further reduce the significance levels.
2. Please comment on the observed age-related device performance seen in PS I, i.e. the increase in the TP rate was observed primarily among patients in the <21 age group across all Pap strata. What are the clinical implications of this finding?
Panel members discussed the implications of the observed age differences in the first study. A Panel member said that the literature documents indicated that lesions in younger patients are more likely to regress; therefore, picking up more lesions is not necessarily a good thing in this age group. Studies are needed in patients under 18 years old. The Panel discussed the possibility, as well, of recommending upper and lower age limits for the device, although Panel members agreed that most clinicians with access to this instrument, who believe it would help a patient, would most likely use it regardless of the patient’s age. Members agreed that there may be potential problems in older patients. Another member expressed concern that clinicians may come to rely on the LUMA too heavily.
3. a. In Pivotal Study II (PS II), for the original study population, the TP endpoint was met, but the FP endpoint was not met, i.e. there was 95% confidence that the LUMA increment of TP exceeds 2%, but there was not 95% confidence that the LUMA increment of FP would be less than 15%. Please discuss the strength of these findings relative to the proposed indication, i.e. as an adjunct to colposcopy for use in patients referred with ASCUS or LSIL cervical cytology results after a
thorough colposcopic evaluation with identification or selection of biopsy sites.
b. For the ASCUS/LSIL sub-population, neither the TP nor the FP endpoint was met in PS II. Please discuss the significance of these findings. Note that the analysis for this sub-population was not pre-specified.
With regards to Question 3a, there was discussion among Panel members that the sponsor arbitrarily chose the 2 percent increment. Dr. Noller noted that the 2 percent and 15 percent were agreed upon by the FDA with the sponsor prior to the initiation of the trial. There were also concerns about why the sponsor changed the sample size but not the margin.
For Question 3b, most Panel members agreed that the ASCUS/LSIL sub-population was the correct population to look at from a clinical point of view, if another diagnostic tool is to be added.
4. The Sponsor has proposed two changes to the analysis of TP rate for PS II compared to the original protocol:
i. changing the success criterion based on observed prevalence, and assigning success criteria to each Pap substrata,
ii. changing the denominator for calculating TP rate.
As indicated below, with these changes, the study meets the revised TP requirements for both the overall population and the ASCUS/LSIL sub-population. For analogous changes to the FP rate analysis, the study does not meet the revised FP requirements for either the overall or the ASCUS/LSIL populations. Please discuss the significance of these results.
Panel members discussed the sponsor’s justification for switching to a 1.5 percent increment and changing the denominator. The greatest concern was that much of this was done post hoc, and that the change was arbitrary. There is no true way of judging the significance of these changes. One member said that colposcopists are trained to select the most abnormal area(s); however, with the LUMA device there may be a tendency to perform additional biopsies based on the LUMA without regard for the colposcopic image and impression.
5. Based on the results of PS I, LUMA’s false negative (FN) rate can be estimated to be 23% when colposcopy is used as the gold standard to determine true disease status, i.e. of all CIN 2/3+ patients found by colposcopy, 23% were not be found by LUMA. In PS II, 22% of the TP biopsies identified by colposcopy were not identified by LUMA.
These false negative rates underscore the importance of the “Always-Never” rule. Does the panel believe that this risk of false negatives is adequately mitigated by the indication statement (for use only after a thorough colposcopic evaluation with identification or selection of biopsy sites)?
Panel members discussed whether the increase in false negatives was worth the extra biopsies. One member said that this shows that LUMA is truly a complementary tool. Others said that if the test was really complementary and all of the lesions that were previously missed were picked up, there would be no question as to LUMA’s value. However, the statistical significance is simply not present, and even when the interval is changed to 1.5 percent the data are not very encouraging.
A Panel member asked the rest of the Panel to consider the final numbers for who will be affected by this device. Roughly 900,000 women are screened with CIN2, 3+. About 10,000 women will get cervical cancer each year, and about one-half of those have never been screened with any method. Of the 5,000 who have ever been screened, about 2,000 will survive. Of these, another 5 to 10 percent have adenocarcenomas or very aggressive, untreatable lesions. Assuming the use of LUMA increases the pick-up by between 2 and 5 percent, there would be between 25 and 100 lives saved by the entire process. This raises the question whether 900,000 patients should be tested to save 25 lives.
Other Panel members continued to express concern that clinicians will come to rely too heavily on the LUMA rather than being diligent in their colposcopies. Colposcopy skills could decline overall, of concern because LUMA misses about 22 percent of the lesions. With respect to the “always/never” rule, Panel members agreed that the device should be used only as an adjunct as noted in the indications. Panel members discussed the high skill level of the LUMA study’s colposcopists; some expressed concern that the trial used high-level experts, which is not representative of the real-world use of the product.
6. Are there any safety concerns associated with use of this device?
Panel members agreed that the device does not pose any electrocution or infection hazard to patients. However, some Panel members were worried that using the LUMA might make clinicians less apt to do appropriate cytology screening. A Panel member asked if a Pap test is inappropriate just prior to using the LUMA. The sponsor said that conducting a Pap immediately before the LUMA scan can result in bleeding. The LUMA can be used only if blood is able to be cleared off the cervix with acetic acid.
7. Do the planned and unplanned analyses of PS I and PS II demonstrate the safety and effectiveness of this device for the ASCUS/LSIL population, i.e.
a. Do the probable benefits to health from use of the device outweigh any probable risks?
b. Will use of the device provide clinically significant results?
For Question 7a, Panel members discussed their concerns about overly aggressive treatment management. One of the groups most highly identified by LUMA was young women with CIN2, 3+ lesions, which may increase the number of loop electrosurgical excision procedures (LEEPs) they receive. This may lead to more procedures that potentially increase chance of pre-term deliveries. Additional data on younger women who have high-grade lesions should be collected. This is an area that the Panel can address in their suggestions to the FDA and the sponsor. A member stressed the fact that it will be hard for clinicians who have LUMA in their office to not use it in a patient who is outside the proscribed age parameters.
Panel members said that they had already covered Question 7b when they looked at how many deaths may be prevented using the LUMA device. One member stressed that there are no statistically significant results in the study, so the Panel may be looking at numbers that have no scientific validity.
8. Does the panel have any comments on the labeling and instructions for use provided by the sponsor?
Panel members offered a number of suggestions, primarily related to patients’ age. A couple of members said that a lower age limit of 18 would complement the cytological evaluation recommendations. Other members also raised the issue of women evaluated between 18 and 21, and expressed concern that increased identification of lesions may lead to more treatment and that treatment may impact future fertility and pregnancy. The Panel noted that pregnancy must be clearly stated as an exclusion criterion. Another Panel member suggested that the labeling specifically state that it is not appropriate for older patients. Other members expressed the belief that age recommendations should not be made, as there is no strong evidence of age differences in the study. Inclusion and exclusion criteria for a study are typically different than for clinical use; the indications are more stringent in this study than they might be in clinical use. A final comment was made about what could happen if the label forbids use in patients of a certain age. The cost would probably not be reimbursed for those patients; this is an example of how the labeling could affect medical practice.
9. Does the panel have any concerns about the training needed to use this device safely and effectively?
Panel members generally agreed that the device appears to be easy to use, especially if the clinician is already a colposcopist. Others said that the device looks straightforward but there remained concerns about bias and the undermining of clinical skills. One member asked the sponsor why the device could not be used as a colposcope; the sponsor responded that the resolution is not sufficient to examine atypical vessels.
10. Does the panel have concerns about any issues that should be addressed in a postapproval study?
A Panel member asked if there could be changes made to the study, such as design changes, after the premarket approval. The FDA responded that the sponsor would have to submit a PMA supplement describing changes and validation to support the changes.
The FDA had a few questions and comments for the Panel. Dr. Pennello told the Panel that the FDA was hearing some support for excluding younger women from the device indications; however, he wanted to point out that much of the effectiveness from the device is in the younger age group in the first study. If that age group is removed, much of the device’s effectiveness may be removed.
A Panel discussion about risk/benefit from the device ensued. Members noted that this device may be setting up young women for more invasive procedures and complicated pregnancies. A Panel member asked if it would be possible to sort out the different CIN levels in the studies. The sponsor responded that the CIN2, 3+ figures are only in aggregate.
The FDA asked whether there had been enough discussion about the use of colors in the LUMA display. A Panel member noted that the colors were actually on a continuum from yellow to blue, and asked how study colposcopists handled a biopsy instruction if there was no precisely blue spot. Dr. Wright said that colposcopists were told to biopsy a blue site, and a yellow site if clinically indicated.
OPEN PUBLIC SESSION
Mark Spitzer, M.D., Professor of Clinical Obstetrics and Gynecology, Weill Medical College, presented during this session. The primary reason he supports the use of a device such as the LUMA is that this it should help novice colposcopists with limited clinical experience improve their skills and identify significant disease. Most studies suggest that clinicians need between 50 and 150 colposcopies to gain the best experience in this procedure, but trainees and residents don’t feel adequately skilled in colposcopy because they are receiving limited clinical experience.
Dr. Schiffman spoke during the Open Public Session on his own behalf, not on behalf of the NIH. He urged the Panel not to recommend approval of the device because the device failed to meet the requirements set forth in the letter of agreement, and could force additional biopsies.
FINAL FDA COMMENTS
The FDA had no final comments.
FINAL SPONSOR COMMENTS
Ronald Alvarez, M.D., Professor of Gynecologic Oncology, University of Alabama, Birmingham, stressed that the LUMA is intended strictly an adjunct to colposcopy, not a replacement for colposcopy. There needs to be improved methods of responding to abnormal Pap smears. Colposcopy remains the “gold standard” for evaluating these patients. The LUMA device advances the performance of colposcopy by improving the detection of CIN2, 3+ by 25 percent when used with colposcopy. Relying on patient follow-up to address deficiencies in colposcopy is challenging. The sponsor has demonstrated that the device is safe and easy to use, and especially valuable in the ASCUS and LSIL populations.
Dr. Wingrove spoke, noting that the two studies have consistently resulted in the same findings. She said that the sponsor would welcome further recommendations so that they may address the Panel’s concerns.
PANEL DELIBERATIONS AND VOTE
Dr. Bailey read the instructions to the Panel on the voting procedures, the official definitions of safety and effectiveness, and the voting options. Dr. Noller asked whether there was a motion to recommend approval, approval with conditions, or not approvable; he added that this motion would be specifically for ASCUS and LSIL indications. Dr. Cedars put forth a motion not to approve the device. Dr. Snyder seconded the motion.
The Panel discussed the motion. Dr. Cedars made the motion because she doesn’t believe there is enough scientific evidence to support the device’s claims of efficacy and because of the concerns for the youngest population in which there appeared to be the greatest benefit. Dr. Snyder said he was excited about the technology, and, while he did not want to stifle ongoing research, he said that it is not in the public’s best interest to approve a device without statistical support. Dr. D’Agostino said that he was worried about approving the device with so few data and lack of statistical significance.
Dr. Noller asked the three non-voting members of the Panel to make brief statements. Ms. Moore, the Consumer Representative said that if the device is to be used as adjunct, it could serve as another beneficial tool in the diagnostic process. She said that she would like to see it approved with conditions. The Industry member, Ms. George, said that she agreed with Ms. Moore. She would choose approval with conditions—specifically after conducting more analysis of the data. Dr. Sanfilippo said that the clinical applications and the clinical importance are absolutely paramount when looking at the LUMA. He said that he would like to see some additional data because he believes that this is a device with clinical applications at all age ranges; unfortunately, he added, he did not see that in the current data.
The panel voted 9-2 (no abstentions) to recommend to the FDA that the PMA for the MediSpectra LUMA™ Cervical Imaging System (P040028) is not approvable. The following Panel members voted for the motion: Drs. Hayes, Snyder, D’Agostino, Julian, Weeks, Cedars, Hillard, Miller, and Jiang. The following Panel members voted against the motion: Drs. Gandjbakhche and Gerbie.
Dr. Noller asked each Panel member to explain their vote.
Dr. Gerbie said that the technology has shown enough promise to get into the public arena for further development, and that the positive aspects should be highlighted rather that the statistical analysis.
Dr. Jiang said that the device’s ability to pick up additional lesions was appealing, but that he was concerned about the possibility of additional biopsies and over-treatment, as well as the impact on younger patients.
Dr. Gandjbakhche said that he wanted to approve with conditions of additional data analysis and more information about the device’s algorithm.
Dr. Miller said that he was not persuaded that the technology would provide significant benefits above what a skilled colposcopist could accomplish. He also noted concerns, including over-treatment, over-diagnosis, as well as the possible “dumbing down” of a skill set that needs improvement among many clinicians.
Dr. Hillard said that the device had potential, but she was concerned that the results did not meet the preset conditions for statistical significance. An increase in false positives could lead to over-treatments that may harm women.
Dr. Cedars said that the results of the studies did not pass the burden of proof.
Dr. Weeks said that he was not convinced by the sponsor’s statistical analyses.
Dr. Julian stated that the sponsor had not rigorously met the criteria for endpoint success.
Dr. D’Agostino said that in addition to the concerns he had about the statistics, he was worried about the clinical issues; for example, he wanted a better sense of what happened to the individuals classified as false positives.
Dr. Snyder said that he would have liked to recommend that the device be kept in the hands of clinical researchers; when something like this is FDA-approved, he said, patients will hear about it and demand it that their physicians use it.
Dr. Hayes said that the device did not meet the definition of safety and effectiveness, and there could be a high probability of over-treatment and inappropriate treatment with this device.
Dr. Noller next asked Panel members to identify what was needed to help the sponsor and the FDA make the device approvable.
Ms. Moore asked for better data on the use of the device among senior citizens and adolescents.
Ms. George wanted additional focus on training and labeling. She said that the device limitations are not clearly stated on the labeling.
Dr. Sanfilippo said that he would like to see studies for patients under 18 years old; the device in the hands of residents and attending doctors who may need some colposcopic education and training; head-to-head comparisons between the device and colposcopies; certain labeling restrictions; and statistical significance in the study.
Dr. Gerbie stressed that he sees great potential with this device. He believes that a new trial would be difficult, primarily because he believes the current trial is satisfactory; however, he acknowledged that a 4.5 percent improvement is relatively small.
Dr. Jiang said he would like to see reduced numbers of false positives, and better analysis on the age factor.
Dr. Gandjbakhche said that he saw no need for additional clinical trials, but he would appreciate more information about the algorithm.
Dr. Miller asked that the sponsor enhance that which distinguishes the LUMA assessment from the colposcopic assessment; this would ideally increase the true positives and lower the false positives. There should be better analysis about what happened in the cases where the LUMA identified a lesion that didn’t exist. He added that aggregating CIN2 and CIN 3 diluted the value of data; these should be separated.
Dr. Hillard asked for additional data, such as the age of patients’ first intercourse. She would also like to see CIN2 and CIN3 data separated and analyzed by age.
Dr. Cedars said that there are more studies that should be done, but any new study should be based on new criteria. New screening guidelines should be used, and the sponsors should set and meet hard statistical guidelines. She noted that even changing the algorithm, the study must be redone because there will not be biopsies for those spots.
Dr. Weeks said he would like to see the second study completed. He suspects that the results would be very compelling. He also supports separating CIN2 and 3. As well, the teaching material should be expanded to address the issue of avoiding multiple LEEPS in young women.
Dr. Julian said that the sponsors should demonstrate that LUMA improves outcomes in nonexpert colposcopists.
Dr. D’Agostino urged the sponsors to separate CIN2 and 3. He said that the next study should be designed on a tighter population where the LUMA would make a larger contribution.
Dr. Snyder said that he would like some changes in the algorithm, if possible.
Dr. Hayes suggested another study that operationalizes many of the previous suggestions.
Dr. Noller thanked the Panel members and adjourned the meeting at 4:31 p.m.
I certify that I attended this meeting of the Obstetrics and Gynecology Devices Advisory Panel Meeting on May 17, 2005, and that these minutes accurately reflect what transpired.
Michael Bailey, Ph.D.
I approve the minutes of this meeting
as recorded in this summary.
Kenneth L. Noller, M.D.
Summary prepared by
Susan C. Sanderson
5932 Lee Highway
Arlington, VA 22207