UNITED STATES OF AMERICA

P R O C E E D I N G S

8:02 A.M.

DR. LISBON: Good morning, welcome to the meeting of the Anesthesiology and Respiratory Therapy Devices Panel of the CDRH Medical Devices Advisory Committee.

As I said before, I'm Alan Lisbon and I'd like to call this meeting to order. I'd now like to have the Executive Secretary make some introductory remarks.

Neel?

MR. PATEL: Thank you, Chairman Lisbon. My name is Neel Patel, the Executive Secretary of the Panel.

Allow me to introduce the members of our Panel. Please raise your hand as I call your name. The Chairman of the Anesthesiology and Respiratory Devices Panel is Dr. Alan Lisbon. Chairman Lisbon is an anesthesiologist and is Vice Chair for Critical Care at the Beth Israel Deaconess Medical Center and Associate Professor of Anesthesia at Harvard Medical School, both in Boston, Massachusetts.

Joining him are the following panel members: Dr. Charles J. Coté is an anesthesiologist and is Vice Chairman and Director of Research of the Department of Pediatric Anesthesiology at Children's Memorial Hospital and Professor of Anesthesiology, Pediatrics at Northwestern University, both in Chicago, Illinois.

Dr. Kenneth Drasner is an anesthesiologist and Professor of Anesthesia at the University of California, San Francisco General Hospital in San Francisco, California.

Dr. Babatunde Otulana is the Industry Representative and is Vice President of Clinical and Regulatory Affairs for Aerodyne Corporation, Haywood, California.

Ms. Carolyn Petersen is the Consumer Representative and is a Web Communications Consultant for the Mayo Clinic in Rochester, Minnesota.

Dr. Avery Tung is an anesthesiologist and Associate Professor int he Department of Anesthesiology and Critical Care at the University of Chicago in Chicago, Illinois.

Joining the Panel Members are the following consultants. Dr. David J. Birnbach is an anesthesiologist and is Professor of Anesthesiology and Obstetrics and Gynecology at the University of Miami School of Medicine, Miami, Florida.

Ms. Andrew Kline is a Pediatric Critical Care Nurse and Nurse Practitioner at Children's Memorial Hospital in Chicago, Illinois.

Dr. Jacqueline M. Leung is an anesthesiologist and Professor of Anesthesiology and Perioperative Care at the University of California, San Francisco in San Francisco, California.

Dr. Robert A. Mueller is an anesthesiologist and Professor of Anesthesiology and Pharmacology at the University of North Carolina in Chapel Hill, North Carolina.

Joining us at the table is Dr. Chiu S. Lin, Director of FDA's Division of Anesthesiology, Infection Control, General Hospital and Dental Devices.

Next, I'll read into the meeting the conflict of interest statement for this meeting. The following announcement addresses conflict of interest issues associated with this meeting and it's made a part of the record to include even the appearance of impropriety.

To determine if any conflict existed, the Agency reviewed the submitted agenda for this meeting and all financial interests reported by the Committee participants. The conflict of interest statutes prohibit special Government employees from participating in matters that could affect their or their employers' financial interests. However, the Agency has determined that participation of certain members and consultants, the need for whose service outweighs the potential conflict of interest involved is in the best interest of the Government.

A waiver has been granted for Dr. Robert Mueller for his interest in a firm that could be impacted by the Panel's deliberations. Copies of this waiver 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 certain matters regarding Doctors Charles Coté and Jacqueline Leung. Each of these panelists reported current and/or past interest in the firms at issue, but in matters not related to today's agenda. The Agency has determined therefore, that they may participate fully in today's deliberations.

In the event that these discussions involve any other products or firms not already on the agenda for which an FDA participant has a financial interest, the participant should excuse himself or herself from such involvement and 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 may wish to comment upon.

I would like to request that everyone in attendance at this meeting take the opportunity to sign the attendance sheet that's available at the door.

Before we begin the meeting, we have two presentations by the FDA, one on the critical path initiative by Dr. Sousan Altaie, the other on the condition of previous studies by Dr. Thomas Gross.

Dr. Altaie, please proceed with your presentation.

DR.WEININGER: I'm having minor computer difficulties at the moment. It will come up in a minute, I'm hoping.

(Pause.)

DR. ALTAIE: Good morning. Well, I've never done it without slides, so I'll try to wing it and see how it goes.

I'll give you -- as far as an overview concern, I'll give you a brief history about how the critical path came about at the FDA and then define some terms in the critical path and then I'll give you an opportunity where devices can play in the critical path and I will give you some contact information at the end to get involved, actually, and help the FDA with this initiative.

FDA critical path started with a white paper that was initiated at the Center for Drugs and the paper analyzes the hurdles of the medical product development and calls for collaboration between government research institutions and manufacturers to promote the public health by getting safe and medical products in the public domain.

Critical path is actually a path where most people refer to as the bench to bedside and it includes basic research then into prototype designs and going into preclinical/clinical studies and then finally going through the FDA review and getting on the market.

Critical path is in the view of FDA, it's a critical path rather than a translational path. And because it's critical and not translational because you think about it if you cannot pursue the path all the way through, the devices won't make it to the market or the medical products per se don't make it to the market.

Critical path is actually a serious attempt to bring attention and focus to the need for more scientific effort and publicly available information on evaluative tools. Now we're talking only about evaluating tools in the critical path and these are techniques and methodologies needed to evaluate the safety, efficacy and quality of medical devices as they move down the critical path into the market.

There are three areas in the FDA's view where critical path research can impact and if you work in three dimensions, you're looking at safety and you're looking at medical utility of medical products and you also are looking at industrialization of those medical products.

So the safety point of critical path, you're looking at material selection, structure activity relationships and then you go into the in vitro animal models and then human models and then finally you follow the safety into the market.

The second dimension that critical path, the tools can influence or impact the development of medical devices is the medical utility. These tools can be in vitro and computer models that you can utilize and simulate physiology in humans and then go into animals and humans and further down the path.

Also, industrialization impact is doing -- dealing with the tools that can participate in physical design of these medical devices. So what critical path research is is basically studies, the ways -- critical path basically studies the ways that medical products' community can leverage basic science knowledge and cumulative experience to bring products to the patients faster without compromising the level of safety and effectiveness that the public deserves.

So moving on, one might think why FDA is taking the lead on this issue and trying to identify tools to get devices faster to market. Well, FDA has a broad perspective of why some products fail and some products do succeed and get to the market. And companies in this competitive world and secrecy of proprietary things does not have this ability. And if you look at academia, centers like NIH and the academics, do not face the problems of device development and manufacturing. So FDA sounds like a logical place to pursue these critical path tools.

Since the critical path -- I'm on Slide 9 -- if you can go to that. Right.

So critical path is different for devices than it is for the drugs and I said originally that the paper, the white paper was initiated in the Center for Drugs and it lacks a little bit of modeling for devices and we are working to include that model for devices in that critical path white paper. But this is why it's different. Device regulations are totally separate than the drug regulations. We're dealing with the least burdensome provisions of the FDA Modernization Act. We're dealing with quality systems and design controls.

As far as the device innovation process is also different than drugs. You produce a drug, test it and it's on the market forever for its life. It's not an iterative process while the devices are. The devices are constantly changing and getting better. You have a user learning curve, how the device is used while it's on bedside or by users and also you have performance and durability issues and biocompatibility because most of these devices are implanted in humans.

And also device industry is not this conglomerate, huge pharmaceuticals and they're small companies, mom and pop operated. And so the critical path for device development is certainly different.

Next slide, please.

This is an example of the devices in the Center. We can go anywhere from a tongue depressor to a contact lens to a CAT scan to a biopsy, so there is a variety of devices with ways to regulate as far as safety and efficacy is concerned. So we are different than the drugs.

Next slide, please.

The critical path projects currently being pursued at the Center for Devices deal with establishing a pedigreed and credentialed serum samples that could be used for assessing the sensitivity and specificity of new hepatitis assays. Those panels lack currently in the market and developing those tools can be getting this in vitro assays quicker to the market.

Another project we're dealing with is to develop computer models of human physiology that allows testing and predicting failure of peripheral vascular stents before the animal and human studies.

Also, we're working on developing a clear regulatory path with consensus from obstetric community of intrapartum fetal diagnostic devices and those devices have been not innovative for a long time. And we're looking as far as reasons why and try to facilitate that progression.

Next slide, please.

We're also looking at establishing agreed ways for statistical validation of surrogate markers. Surrogate markers can be used in selection of the patients in clinical trials. You can use them as end points and so there are quite a number of them in the science arena, but they are not fairly validated as to how they should be used and that's a huge effort currently in the Center. We're working to start with the cancer markers and also some markers related to stents and peripheral vasculars.

So working with medical specialty organizations, we also are developing practice guidelines for appropriate monitoring of permanently implemented devices, implanted devices.

The last project that we currently are working in the Center for is to obtain consensus on the extent of neurotoxicity testing for neural tissue contacting materials.

Next slide, please.

And then these projects are actually quite much more extensive than this. The reason we are limited to this project at this time is lack of funding. We were expecting appropriated funds and those funds didn't come in and so now we are dealing with running projects with no funding and the process is slow, but they're all active and alive and kicking.

These are the contact information for you to get involved. If you have tools in your area of expertise that you think can facilitate putting material on the market faster, please contact us. You can sent your comments to the dockets as you see there and add to the white paper and let us know how you can help to facilitate getting devices faster on the market.

With that, if there are no questions, I can go on and you can go on with the rest of the meeting.

MR. PATEL: Thank you, Dr. Altaie.

Dr. Gross?

DR. GROSS: Good morning. I am Tom Gross. I'm the Director of the Division of Post-Market Surveillance in the Office of Surveillance and Biometrics in the Center. And I'd like to take a few minutes of your time to talk to you about recent changes in our condition of approval study program. And before I do that and to put it into context, I'd like to tell you a little bit about the functions that the Office of Surveillance and Biometrics serves for the Center.

Next slide.

The office has several functions, first and foremost is to provide support for pre-market review. We have a large staff of statisticians who address statistical aspects of the pre-market submissions, whether they're 510(k)s or PMAs. We have a large staff of epidemiologists who have been recently incorporated into the pre-market review process to look at the risk benefit of PMA products and to an eye towards designing conditional approval studies and I'll mention that a bit more in a minute.

We're also responsible for monitoring a nationwide passive adverse event reporting system, looking for potential public health problems, signals of potential health problems through out medical device reporting system or MDR and also a companion system of a nationwide network of health care facilities all tolled about 350 called MedSun or the Medical Device Safety Network.

Our epidemiologists also analyze safety issues. They characterize the risks through literature reviews, design of studies and applied research. We also coordinate the Center response on important potential public health issues by convening committees of Center experts to further investigate these issues, deliberate these issues and then submit recommendations for actions to our Center senior management.

And lastly, we're responsible for interpreting the medical device reporting regulation. This lays out the mandatory reporting requirements for manufacturers, user facilities and importers.

Next slide.

Now what about condition of approval studies? As you are well aware, these studies are ordered as a condition of approval for PMA devices and the regulations clearly stipulate that post-approval requirements can include the continuing evaluation and periodic reporting on the safety, effectiveness and reliability of the device for its intended use. This gives us our broad authority to levy condition of approval studies.

Next slide.

In 2002, we decided to take a good look at this program. to do that, we looked at PMAs that were approved between 1998 and 2000. All tolled, there were 127 PMAs. We focused on those PMAs that had clinical condition of approval orders. It amounted to 45 PMAs.

The bottom line was this, that CDRH had limited procedures for tracking study results and study progress. Our IT and other systems were found to be very deficient. There's a large turnover of lead reviewers that resulted in lack of follow up. Forty percent of those individuals that were lead reviewers at the time the PMA was submitted were no longer the lead reviewers in 2002, again, stressing the lack of continuity.

The lack of pre-market resources, the pre-market resources were prioritized elsewhere to analyze pre-market submissions and limited resources were available to look at these incoming reports.

Next slide.

So we developed a strategy for change, based on some simple goals, to obtain timely and useful post-market information as the device enters the post-market period and to get information on real world use, to better characterize the risk benefit profile of the product, for instance, it's long-term performance and add to our ability to make sound, scientific decisions increasing the rigor and quality of these studies.

Next slide.

So what do we do? We transfer the condition of approval study program from the pre-market side of the house, the Office of Device Evaluation to the post-market side of the house, the Office of Surveillance in Biometrics. OSB has the resources and we've got the resident expertise in terms of a staff of epidemiologists who are expert at designing observational studies.

We developed and instituted an automatic tracking system for these study commitments, first and foremost to acknowledge the receipt of study reports and to follow-up when reports were not received.

Next slide.

I alluded to this before. We added epidemiologists to a PMA review team. We started this as a pilot two years ago. It was expanded to the rest of the Center. And the epidemiologists were tasked with several items. First and foremost was to develop a post-market monitoring plan during the pre-market review of these products, how best to monitor these products i the post-market period including CoA studies; second, to take the lead in developing well-formulated post-market questions, the lead in the design of these studies because of their expertise and observational studies, the lead in the evaluation of the study progress and results, and throughout this process to continue to collaborate with the lead reviewers, the medical officers, the statisticians as part of the PMA review team and also to negotiate with the companies.

Next slide.

This speaks to the motivation for good study conduct, not only on our part but also on industry's part. We have to address important post-market safety questions and those have to be addressed through good protocol study design. As I stated before we have to be able to track these studies and give feedback on the interim reports. To be more transparent, we plan on posting the study status of these results on a CDRH website. This is currently done at the Center for Drugs and the Center for Biologics. And lastly, we may have to utilize some enforcement strategies, if there's extreme lack of due diligence in the conduct of these studies. We can turn to our so-called Section 522 authority, levy a similar study. If that is not done, we can misbrand the product and it may result in hefty monetary penalties. Again, this is a last resort. We hope not to have to use this.

Next slide.

What is the impact on the Advisory Panel? We will attempt to lay out the important post-rule public health questions at the time of panel presentations and possible approaches for the panel to consider. And also on a periodic basis, FDA and industry will update the Advisory Panel on the progress and results of these studies.

That concludes my remarks. Thank you very much. I'll entertain any questions.

MR. PATEL: Thank you, Dr. Gross.

Dr. Lisbon?

DR. LISBON: All right, I note for the record that the voting members present are constituting a quorum as required by 21 CFR 14.22 Section D and we'll now get started with the agenda.

What we're going to start with is four presentations by the FDA. The first is by Ann A. Graham. I would just ask that everybody identify yourself for the record, please.

MS. GRAHAM: Good morning, again. I'm Ann Graham, the Branch Chief in the Anesthesia and Respiratory Devices Branch. And again, I'd like to welcome everyone to this panel meeting this morning.

Before we get started, I would like to introduce the members of the branch to you. And if you could just raise your hand as I call your name: Justin Guay, Lisa Lavelle, Mike Husband, Bill Maloney is on vacation right now and Neel you've met and Dr. Joydeb Roy. Great, thanks.

We also have three anesthesiologists that we have recently hired through our Medical Device Fellowship Program, one of whom you will meet later this morning, Julian Goldman and Bill Norfleet and Eric Pierce were unable to be here this morning. Bill is an anesthesiologist at Yale and Eric is 50 percent with FDA and will be soon with Mass. General Hospital in Boston.

The subject of the meeting this morning is pulse oximeters and I would just like to briefly go through some of the regulatory aspects related to pulse oximeters.

As you heard yesterday in your orientation, they are Class 2 devices and they are subject to 510(k) which is a pre-market notification. The two classification regulation numbers you see here are the authority under which we regulate these devices. 2700 regulation is the general category for oximeters used to transmit radiation at a known wavelength through the blood and to measure the blood oxygen saturation based on the amount of reflected or scattered radiation. It may be used alone or in conjunction with a fiber optic oximeter catheter.

Under Regulation 8702710, we separately regulate the ear oximeter. It's an extravascular device used to transmit light at a known wavelength through blood in the ear. And again, the amount of reflected or scattered light, as indicated by the device, is used to measure the blood oxygen saturation.

The intended use for pulse oximeters is noninvasive, transcutaneous, continuous or spot checking monitoring of oxygen saturation of functional arterial hemoglobin and pulse rate. And I've highlighted the second bullet because it's on point to one of the questions that we've asked you to think about today. They are currently all prescription use devices for medical indications.

The patient populations are adult, pediatric, infant and neonate for transmission pulse oximeters. The site of application, depending on the optical design of the pulse oximeter can be the finger, the earlobe, the forehead or the back. Those are the four primary locations.

In the environment for use is in the operating room, critical care, post-anesthesia recovery room and some question of morphing into other locations, other environments such as home use.

The recommended elements of a 510(k) for a pulse oximeter includes these four bullets. There's obviously much more and these four are embellished, depending on the complexity of the device. But in general, the description of the device should include the operating characteristics, the design, the comparative performance of the subject device, the device under consideration in the current 510(k), compared with a legally-marketed predicate. This includes the desaturation studies that are performed to validate the saturations.

And finally, we look at the subject device and the predicate device labeling to ensure uniformity.

The accuracy specification of pulse oximeters is somewhat different, depending on the patient population and the type of oximeter sensor. As you can see in the chart, for transmittance, wrap and clip, which is the finger, all patient populations require an accuracy spec of 3 percent or less. Neonates have 4 percent. There is a 1 percent additional accuracy degradation added to the allowed spec in adults. This comes from the Agency's agreement to accept adult data for neonatal use and add a 1 percent degradation factor and Sandy and Julian will embellish that philosophy later in the presentation.

The transmittance earclip has a slightly higher accuracy spec of 3.5 percent and we currently do not have any submissions cleared in the infant or neonatal population for either the earclip or the reflectant sensors.

We have been clearing in the last couple of years roughly thirty-five 510(k)s a year, so it's a substantial part of our workload. Most are for transmittance sensors or systems, actually, because the 510(k) not only includes the sensor, but the oximeter and the patient cable. So they're cleared as a system, unless the 510(k) is just for the sensor and in that case it must be shown to have been validated with a previously approved oximeter.

Most are transmittance technology, single use and nonsterile. Last year, we have several more than our usual number of 510(k)s because we requested 510(k) submissions from manufacturers who were reprocessing single use sensors.

And the next slide just shows over the regulatory history of pulse oximeter at CDRH, you can see before 1985, I think we had five or six files come in and a big blip from 1986 to 1990 when pulse oximeters began to morph into clinical practice. And as I mentioned earlier, you can see in the last five years, we've had a substantial increase in the number of files for pulse oximeters.

The next two presentations will -- the first one will be from Dr. Sandy Weininger, in the Office of Science and Engineering Laboratories. Sandy is the chair, co-chair of the ISO Committee that has been developing the standard for pulse oximeters and this standard has recently been published, I think, in January of 2005.

And Julian Goldman is, as I mentioned earlier, is a medical device fellow and Julian is going to ask you to think about clinical conditions for reflectance and transmittance sensors. He'll ask you to think about certain considerations relating to neonatal validation and also to explore your thoughts and prepare recommendations for over-the-counter use of pulse oximeters.

We have had in the past two documents that have -- well, actually one document that has served the branch as guidance. One is our 1992 guidance document and we realize that this document needs to be amended and specifically for this panel meeting, we ask that you consider the questions related to neonatal labeling in OTC use.

The standard that I mentioned, the ISO standard that was published in January, does not call out a single test method for accuracy, surface temperature, motion or low perfusion, nor does it distinguish between reflectance and transmittance technology. So while it is a very good source document for us, it doesn't give FDA all the guidance that we seek to supplement with your advice today.

With that, I would like to introduce Sandy Weininger.

DR. LISBON: Thank you, Ms. Graham. Can you entertain a few questions now?

DR. BIRNBACH: I'd like to know how the FDA is defining neonatal?

MS. GRAHAM: Oh, that's a good question.

DR. BIRNBACH: It's used pretty often and --

MS. GRAHAM: Right. We have actually a guidance that I believe is under 30 days and I believe that there is a weight, upper margin, on that as well, like under 10 kilo.

DR. BIRNBACH: And it excludes fetal use of pulse oximeter?

MS. GRAHAM: Yes, it does. I should have mentioned that. There are two other oximeter categories that we have at FDA. One is for fetal use and fetal oximeters are all class 3 devices requiring PMA and the other are in-dwelling tissue oximeters. The fetal oximeter, I think, is regulated under a separate classification, but the tissue oximeters are regulated under the same 870.2700. So we differentiate those when we receive them from manufacturers with a different product code and they go to a different group, the cardiovascular group.

Let me get source documentation for you though on your question about neonates so that I give you the correct answer and I'll get that to you before the end of the day.

Thanks.

DR. LISBON: Dr. Coté.

DR. COTÉ: Dr. Coté. Are we considering fetal oximeters today or just neonatal use?

MS. GRAHAM: No, neonatal only.

DR. COTÉ: Okay.

DR. LISBON: Avery?

DR. TUNG: Pulse ox. technology is almost 20 years old. I might just, without knowing anything, expect a machine today to be much, much better than the ones in 1985 when it was first introduced.

When you say comparative, do you mean compared to a standard, a machine built in 1985 or do you keep evolving that standard as newer and better machines come out? Or are they not any better than they were?

MS. GRAHAM: Well, I would like to think that, in general, they are much better than they were in 1985. It's the responsibility of the manufacturer to identify the comparative oximeter system that FDA would look at to compare their new device with an existing or a legally marketed predicate device. So some manufacturers do take a very new technology to do that comparison. And this is outside of oximeters. This is in general, some manufacturers take very old predicates and you can imagine that in 1978 or in 1982, we asked for I think it's fair to say less data than we do today, so the test to meet substantial equivalence for those older predicate devices is less than it would be in comparing with a device approved or cleared in 2002, for example.

Does that help? Okay.

DR. LISBON: Do any of the other panelists have questions?

Okay. Thanks, Ms. Graham.

DR. WEININGER: Good morning, Panel Members, my name is Sandy Weininger from the Office of Science and Engineering Laboratories. And I recently have come off a detail to the Office of Device Evaluation so I'm relatively familiar with how oximeters have been regulated. I've been following them since roughly about 1990 when we tried to incorporate them into apnea monitor standard.

Let me briefly address the two questions which was neonatal applications and how oximeters are regulated. Currently, the manufacturers specify what they believe to be neonatal and that's different than what FDA considers and so that's just a statement of fact, let's keep that clear and that's something that needs to be cleared up, obviously.

And oximeters are regulated as Class 2 devices and so the term substantial equivalence which I'm sure you heard is what we -- is how we regulate them. So if a manufacturer has to demonstrate or if a manufacturer has to demonstrate that his pulse oximeter system is substantially equivalent to a legally marketed device, how good or how bad that device is is up to the predicate that they choose.

And that's actually a good segue into the regulation of pulse oximeters. Next slide. There we go.

In 1992, we came out with the draft guidance document and that was roughly after about 10 or 15 years of oximeters being on the market. You have that in your package and you have perhaps had a chance to look through there. It is 15 years old and I'll talk to you about what the elements of the guidance document are, what are the recommendations and I'll also try to give you an insight into what our current regulatory policy is because as you know, over the last 15 years, there's been a lot of new scientific and technological developments and our regulatory policy tried to keep pace with it.

Next, please.

So the obvious things to ask for in a guidance document or from a 510(k) are what's the configuration? We know that pulse oximeters have to have a probe-monitor combination in order to determine what its accuracy is. If that probe-monitor combination isn't identical to the predicate, then we ask -- we typically ask for clinical validation data to demonstrate accuracy.

We also ask for information about the accessories. For example, extender cables and some other things. They too have been shown to affect accuracy and so we want to make sure that the system, the pulse oximeter system is controlled.

We need to know what the device is so we ask for engineering drawings, center dimensions. We want to know what the functional elements are, both electrical and optical, anything that's in the signal path from the probe through to the oximetry. Saturation algorithms we consider to be essential components and if you make modifications to them we need to know that it doesn't adversely affect the performance of your device.

We obviously want to know about alarms and alarm limits and what defaults you have. And this information we use and we put into a comparison table to compare to the predicate. So if you have performance that's commensurate with the predicate device, they're deemed to be substantially equivalent and we do this comparison based on the features, specifications and accuracy of materials and also we look at intended use and we look at the indications for use. You've heard Ann talk about the indications for use; the target population, the use environment and site of application.

We go through a great deal of evidence or we recommend that you provide a great deal of -- manufacturers provide a great deal of evidence to demonstrate functional verification. We want to know that your box is going to work and we want to look at the test method to make sure that the test method is appropriate. We want to look at the acceptance criteria to make sure that it's a reasonable acceptance criteria and we want to make sure that the results support the conclusions that you're coming to. So we ask for a rationale for the test method and some others.

So probably the most important piece of information that we get on a pulse oximeter is the accuracy and as I said, if you make a modification to any of the components in the optical chain or the signal processing, we typically ask for laboratory testing based on human subjects. There are no adequate simulators that can represent the optical characteristics of the pulse oximeter, and so we need to do human subjects and these studies require both an IRB approval and informed consent. They are subject to the IDE, Investigational Device Exemption regulations but for nonsignificant risk devices.

Let me give you some elements or some details on the desaturation study. The desaturation study compares the performance of the pulse oximeter to the co-oximeter and we typically require a minimum of 10 healthy subjects that consent to induced hypoxia as part of the experimental procedure. The subject characteristics should range in age, gender and skin tone and we report -- we recommend that the manufacturer report what's called the root mean square error, A_RMS, and that actually is defined in the ISO standard which we'll get to in the second part of the presentation.

I'll note two of the panel questions deal with comparative, the comparison between a transmittance and reflectance sensor and if we're looking to see whether I can use a reflectance or transmittance sensor as a predicate, for comparing substantial equivalency on a transmittance sensor, perhaps the calibration method is something that needs to be considered and also for neonatal performance we need to consider what types of calibration studies can actually be done which brings us to neonatal use.

The guidance document recommends collecting convenience samples and recognizes the limitations of doing that. Clearly, you can't desaturate a neonate which leads to all kinds of problems and wider variances, greater uncertainty in the calibration studies. So our current practice has been to grade the adult accuracy by 1 percent and our rationale for doing that is evolved, but it's typically spoken about if talking about the fetal hemoglobins and other error sources which preclude us from getting those very accurate results as we do with a controlled desaturation study on an adult. And as I said, certainly the panel questions can revolve around that.

The data characteristics, we typically require 200 or we recommend that the manufacturer supply 200 data points over the entire range of 70 to 100 point saturation. These data points are paired observations. That's pulse oximeter and co-oximeter values. They consist of -- it's an individual pulse oximeter sensor combination with the simultaneous blood draw. And I'll note the asterisk. The standard does not call out -- it does not recommend a specific number of data points and we'll get into some of those issues later.

Again, as pointed out, if you use a different pulse oximeter sensor with a monitor and put them on a different finger for your desaturation study you've got to demonstrate that the data is poolable, that it comes from the same type of population because we know oximeter sensors, when you change them, have different calibration curves. And so that's part of the statistical analysis.s

Next slide, please.

The environmental factors, surface temperature for testing the pulse oximeter is very important. Oximeters, as you know, take optical energy and beam it through a finger and you can't have optical energy unless you have heat and the objective is to prevent burns from occurring. So there's a limit of 41 degrees C. which has been around for a long time. That's for the applied part and just as a side note the case also has a recommended maximum temperature of 50 degrees C. and those are pretty much recommended or have been identified in the international standards community as appropriate temperature limits.

Also for environmental factors, electrical safety is very important. Electromagnetic compatibility, mechanical and environmental testing of the device are also recommended.

Pulse rate, because it's a pulse oximeter, and that's one of the differences between tissue oximeters and tissue oximeters typically don't measure pulsatile blood when look at tissue where as pulse oximeters measure arterial pulsatile blood, but we allow or we recommend that manufacturers demonstrate the performance of their pulse rate detection system using in vitro calibrators. They're much more reproducible to human subjects, but we do ask that the simulator be set to their lowest values to represent the weakest pulses to make sure that the oximeters are capable of detecting pulses in weak patients.

A very important part of our review consists of looking at the labeling. The labeling includes the sensor specification, that's the pulse rate and saturation accuracy claims, what the temperature and humidity specifications are and importantly, what the pulse oximeter sensor monitor combination is. And we ask that the patient population be identified, as well as the indications for use.

An important aspect of labeling is the application time. The 1992 guidance document recommends that the sensor application site be inspected and repositioned every four hours. I'll note that the standard doesn't have that absolute limit, but instead requires disclosure of the application time and evidence to show that that application time is appropriate, so it's a slightly different approach.

Ann talked about in the intended use continuous versus spot checking. That needs to be disclosed and if your device is continuous or spot checking and you -- if you are -- for continuous use you need to have both low and high SpO2 alarms. An interesting note -- well, not an interesting note -- compared that the standard calls out whether -- or asks to disclose whether you have a physiologic alarm versus an equipment alarm as opposed to continuous versus spot checking and it's kind of the flip side of the same coin. If you use a continuous use device, you have to have a physiological alarm. If you have a physiological alarm, you must have a low SpO2 alarm. That's a slight variation from what the guidance document calls out.

For reusable probes you must demonstrate that you can clean the probe up and return it to its normal existing safe conditions and so we ask the manufacturer to demonstrate how they're going to that and to verify that they can actually get that done.

And we ask them to do that if their probe is going to have a one year life span or five year life span. We say well, show us some evidence or show us a test method where you actually do that after 30, 50, 100 cleaning cycles.

Dr. Tung, you raised the specter of oximeters improving and certainly the first oximeters that came out in the early 1980s, late 1970s, by today's standards had very simple microprocessors. The oximeters that are coming out today have multiple microprocessors that are extremely powerful devices. They run multiple algorithms and do a lot of signal processing which means that they're very software intensive.

And so software safety is an issue with pulse oximetry. We ask that manufacturers demonstrate or use our software safety guidance documents to demonstrate that their software is safe and effective. Oximeters are moderate level devices from the perspective of this guidance document. That means that they are not life-supporting or life-saving, and the guidance document talks about levels of documentation, types of documentation. Importantly, software development, as you know, you can't test software to make sure that it works alone. You need to make sure that the software development process is appropriate, as well as the outputs of those processes. And that's a fairly intensive and technological review.

Next slide, please.

Oximeters touch tissue and so we need to make sure that their biocompatibility requirements are there, that the materials used are biocompatible and so we ask that they list the materials used to construct the oximeter and demonstrate biocompatibility for all the materials used. And there's various different ways to do that and we don't want to get into that at the moment.

Next, please.

When the guidance document came out in 1992, there were no reprocessed single use devices. We now have them and they're getting to be a larger part of our review activities. And what does a manufacturer of a single use device have to -- a reprocessed single use device have to demonstrate? They have to demonstrate that their probes are accurate after the number of use or cleaning cycles. So if they say that their probe can be reprocessed five times, they have to go through typically, it's a simulated re-use protocol and then demonstrate validity which is also a human desaturation site.

Importantly, the reprocess probes do not have to show electrical leakage or electromagnetic interference, susceptibility. They don't have to redemonstrate -- I'll use the term redemonstrate. They don't have to show that they are mechanically safe because the original probe, original submission 510(k) demonstrated that.

Next, please.

So I've gone through the guidance document and showed you what the original document was and what some of our updated thinking is. Let me now take you to the standard. And --

DR. LISBON: Do you want to take any questions on the first part of your talk at this point? Are there any questions?

Yes.

DR. COTÉ: How does a manufacturer track how many times a particular probe has been cleansed? In our practice, we just take them and throw them in a big box and --

DR. WEININGER: Are you referring to single use devices that are reprocessed or are you referring to clam shells that --

DR. COTÉ: No, single use devices that are reprocessed.

DR. WEININGER: It's up to the single use device manufacturer to have a tracking system in place and we do review that.

DR. COTÉ: So there is a requirement for that?

DR. WEININGER: Yes, I call it a recommendation because it's a Class 2 device.

DR. LISBON: Are there any other questions?

DR. TUNG: One more. Because they exist, I've seen them. There are machines that don't have signal sensitivity meters. They just give you a signal and they don't tell you how strong that signal is, nor is there any requirement that -- and I'm just seeing if I get this right, that there's any requirement that the machine be able to sense a certain amplitude of signal. Is that right?

DR. WEININGER: Correct. The standard used to call out or have a requirement for a signal amplitude indicator. We've modified that to be signal adequacy because signal amplitude is very difficult to get from the box and portray in the front panel and a lot of front panels have the plethysmogram which you see and which is typically normalized so that it looks pretty on your display, but might not represent the actual pulse amplogy.

DR. COTÉ: You said that the 200 samples spread out between paired data between 70 and 100 percent , how well distributed are those 200 samples. In other words, it doesn't do much good if 190 of them are in the 95 to 100 percent range and only 5 are in the 70 percent range.

DR. WEININGER: The way it's stated is that it has to be equally distributed across the entire range, 70 to 100.

That's not a statistically rigorous statement. And we're working on that. The standard and FDA is attempting to make that a more testable recommendation.

DR. OTULANA: Babatunde Otulana. You said tat the -- for the neonates, obviously for ethical reasons studies cannot be done in that population. Are there other criteria that you use to determine when that application will be approved for both adults and neonates or just neonates?

DR. WEININGER: That's really the subject of the question for deliberation, for the Panel.

DR. OTULANA: Sure, I'm asking the current practice.

DR. WEININGER: The current practice -- I'll give you the ideal practice. The ideal practice would be that a manufacturer demonstrates that his probe fits on the neonate and it works on the neonate and demonstrating that it works on the neonate is very difficult because you can't desaturate a neonate like you can do a human. You can convenience samples and Dr. Goldman is going to get into some of those issues later on, so let's hold off on that discussion, but that is front and center what I think Question 2 is about.

DR. LISBON: Are there any other questions? Please proceed then.

DR. WEININGER: So let me give you an overview of the pulse oximetry standard and again I'll try to highlight the differences between the standard and the guidance document. FDA uses standards as much as possible. We believe they can do multiple different things. The standard can take the place of just test data as a test method standard. We do, however, we do also recognize that standards within their scope can represent safety and effectiveness for particular types of areas, for example, electrical safety. We typically recommend 601 and so let me get into this and we'll see how that plays out.

Prior to 1992, ASTM built a standard called F1415 which leads to ISO 9919 and the European standard 865 and a high use of acronyms which you don't fully understand -- please stop me and I'll fill them in.

So the history and the reason I'm bringing this up is that the ASTM group which is predominantly manufacturers and clinicians and it's a consensus development process puts a standard together and historically in pulse oximetry, that standard that domestic ASTM -- American Standards Testing -- it's a domestic national standards institute. Typically, that standard has been adopted by the international community and so it has great weight and now in 1996 with the revision of that standard from ASTM it got broadened to ISO and then ISO got broadened to both IEC Joint Working Group and all those acronyms simply mean that it gets a very wide exposure, large number of clinicians and engineers participate in the development of the standard. And that's good, because we think that it holds great weight. And as Ann said, it was completed just in January and FDA is in the process of recognizing.

The title of the standard is particular requirements of the basic safety and essential performance of pulse oximeter equipment for medical use. It is a particular standard which is a vertical product standard under the IEC 60601 family. 60601 is the general requirements for safety. ISO 9919 with the pulse oximeter standard gives specific requirements for pulse oximeters that expand on the general requirements as well as requirements that re unique to pulse oximeters.

So let me just compare what the standard talks about which is basic safety and essential performance to what happens in a 55(k) process. 55(k) process is a process of substantial equivalence so you're trying to show by objective evidence that your device in a 510(k) compares to a predicate device that has the same performance and safety and effectiveness as the predicate. The standard calls out, if you will, absolute requirements for safety and effectiveness. I say those are absolute requirements. When FDA uses a standard, they are still recommendations.

If the manufacturer can demonstrate that they have an alternate way of meeting a recommendation, they can certainly provide that. But it does give us a consensus opinion about what is basic safety. Basic safety is protection against direct physical hazards such as shock vibration and mechanical injury and importantly, essential performance which is functional safety. And that's the performance necessary to achieve freedom from unacceptable risks. Importantly in pulse oximetry that's the accuracy of the monitor. That's an essential performance criteria.

And I'll mention the standard talks about equipment when it's used under normal and reasonably foreseeable conditions. I don't want to -- you guys are the experts in figuring out what is an unreasonable use of a pulse oximeter.

The standard -- next slide, please.

The scope of the pulse oximeter standard is that it applies to all original equipment manufacturer probes. It applies to reprocessed probes and extender cables, so the standard doesn't make a distinction between a reprocessed probe or a brand new probe. They are both new medical devices and have to be treated as such.

There are extensive requirements for documentation and labeling in the standard and there are extensive informative annexes that detail the rationale for each requirement so someone implementing or trying to declare conformance or use the standard can go to the rationale and understand what the purpose of that clause is and how the test methods should work and there is also detailed educational material, particularly about functional testers because they have a history of being used as calibrators.

I'm sure you've all seen in a hospital somebody take their pulse oximeter and slap it on a functional tester and say I know this device is working when, in fact, it might work from an electrical perspective, but that doesn't guarantee the 95 percent saturation on your screen is a 95 percent saturation. We did address that issue.

Next, please.

I said that the guidance document references the standard to call out what the accuracy specification is. In this case, it's a root mean square error and that's the difference between what the pulse oximeter reads and what the referenced co-oximeter, the blood based co-oximeter. It has to be defined over the range of 70 to 100. The data must be evenly distributed and we spoke about this and unfortunately, that's one of the areas of improvement. We have some pictures and I'll show you the pictures in a moment on what that means, but we don't have a recommendation or a requirement to prove, if you will, or a testable requirement to show that it's evenly distributed.

The standard calls out a maximum uncertainty or a maximum accuracy of 4 percent for pulse oximeters. You saw and Ann showed you what the clear devices are and highest accuracy that we've cleared to date and that's 4 percent and that's in a neonate. And so the standard attempted to essentially mirror that and as far as we know there are no unsafe oximeters on the market, at least due to their accuracy claims and so the highest accuracy out there must be safe. So that was the upper limit that was selected.

Next slide, please.

Here's a picture of the recommended desaturation profile. The X axis is time and the Y axis is the co-oximeter saturation. You can see that there are a number of points taken at each plateau and there are a number of plateaus that stand the range. Again, the exact number of plateaus and the number of points for each plateau are not specified.

There are other protocols that have been used in the past, interestingly enough. There is some evidence that if you use a different protocol you get a different ARMS and so the standard tried hard to focus or to select a single protocol. Unfortunately, we didn't get there yet and so this protocol, that's why I said it's a recommended desaturation profile. I'm hoping that in the next version of the standard that it becomes the desaturation profile.

And again, the -- one last bullet on the last one. The important thing about having a single desaturation protocol and I'm doing these ideal human desaturation studies is so we can compare Monitor X to Monitor Y. If everyone is using a different desaturation protocol, then we have no objective way to compare it and since -- particularly in a substantial equivalency determination where you're looking at comparison to a predicate, you need to have your test method, you need to have a single test method to support that comparison.

Next, please.

The standards body spend a great deal of time trying to understand what the shocking vibration environment is, particularly for home and hospital. But you can see we also went into vehicles. There's four different standardized types of tests for shock and vibration and there's shock, vibration. There's drop testing and bump testing and you can conceive of in your hospital, your home where you're hitting steps or dropping monitors, these are the different types of tests that is out there.

The original guidance document had similar types of testing levels, but used what I'll call nonstandard tests. And there's nothing wrong with using nonstandard tests, except that most of the testing industry has centered on using what I'll call standard test levels and so the standard attempted to update those, one, to make them more realistic and relevant to what the perceived environments are and two, to make them standardized levels so that they can get done in a practical and efficient manner.

Next, please.

Alarms and indicators, clearly that's an important part of pulse oximeters and as I said before the standard calls out or requires that if you have physiologic alarms that you must have a low SpO2 alarm present. The original 1992 versions of the standard called out a low SpO2 limit of 80 percent saturation. That was the default and you, as the operator, can apply a higher limit, but the default was 80. That's been raised to 85. There are all kinds of arguments around false alarm rates which we can discuss later. And I'll compare that again to the guidance document which calls out continuous versus spot checking and has requirements for both or recommendations for both a high and low SpO2 alarm limit.

And again, here's the disclosure for the description of the signal adequacy indicator. It used to be called signal strength, but many of the boxes these days it's more than signal strength. It's actually a quality indicator and it does morphology checking to make sure that the pulse actually looks like a pulse, because there's lots of different noise and motion artifact and electromagnetic interference and those typically make the pulse not look like a pulse. And if your oximeter can detect that because it's got plenty of processing power, that's important information that the clinician should know.

And so the important message here is that the standard calls for a description of that so that you, the clinician, when you're purchasing the oximeter, can make an informed decision, based on your particular application.

Next slide, please.

Patient characteristics. The patient population, age, weight and the application area of the body, the types of tissue you're going to apply it to and the environment, frequency of use and location all have to be called out.

Next slide, please.

Both use and human factors are certainly important. The recommended maximum application time has to be disclosed as well as rationale on the manufacturer's part on why they chose that limit and evidence to support that. There is also from a human factors perspective a requirement for a variable pitch tone. Most of the oximeters out there when saturation decreases you can hear the pitch go down. It would be bad if half the oximeters, the pitch went up. It would be counter intuitive and present a real patient safety issue and so the standard addresses this by having a requirement. Compared to the 1992 guidance document, it doesn't address this at all because in 1992, things were just getting started and everybody did it their own way.

Next slide, please.

I mentioned safe surface temperature in the guidance document. I talked about the applied part which is the probe and I talked about the monitor itself. I want to give you just a five second history of where that comes from and talk about what the updated version of the standard talks about. The general standard, 60601, the third edition is about to come out this year or next year. It raises the limit, decreases the safety margin from 41 degree C. to 43 C. It does, however, require a risk analysis to demonstrate that that's still a safe surface temperature.

The ISO or the ASTM and the ISO Committee actually had a large amount of input into that decision. We were the driving force and did extensive literature and there's actually some experimental studies to show that 43 C. is relatively safe. Interestingly enough though the pulse oximeter standard adds further constraints. It only allows 43 C. for four hours. It allows 42 C. for eight hours and it allows 41 C. essentially for the maximum exposure time, but that's in adults. For neonates, children, pediatrics, you have to stick with 41.

Also, the standard requires that there be an automatic setback. If the clinician decides that they need the higher temperature and I'll state the obvious. The reason they need the higher temperature is to get more light because they're not getting good signal. They're not getting good signal adequacy and they want to -- they need to get the saturation on those types of patients. But they set it at 43 C. and they walk away. The standard requires that the monitor automatically be set back to a safe limit, 41 C. after four hours. And I'll note here that there's been lots of, if you will, burns reported in the literature. Many of those burns are actually pressure contusions. The pulse oximeter can be off and you can still see the same type of injury, so even though you're not delivering any heat, the mere fact that you're taping something and constricting the perfusion and the flow and some oximeters I've seen get taped on with many layers of tape and they're actually squeezed on pretty tightly to try to get a good signal.

Next, please.

Oximeters in the early days were primarily trend devices and their response time, if you wanted to see what your oximeter is doing you could set -- if you wanted to get rid of some of the noise in the oximeter, you could actually raise the averaging time or the number of beats that the oximeter averaged over and over the last few years people have discovered -- that's not quite the right word, but there's been new clinical applications, for example, some sleep studies where there's been evidence of a very rapid desaturations and resaturations and we heard testimony from many clinicians that that's an important performance aspect of pulse oximeters.

So the standard body sat down and we said well, what's the best way to talk about that to the manufacturers and the clinicians? And we came up with a disclosure requirement of response time and you can see on the curve that percent saturation is on the Y axis and the X axis is time. The very pretty idealized desaturation profile is the solid line and you typically do that with the simulator, but you can see there's some delay and the dotted line is the oximeter response, the displayed oximeter response. And so there is some delay as to when the oximeter determines that the saturation is falling and there's also some infidelity in the depth of the trough that the oximeter picks up.

All this goes to show that the oximeter needs to disclose how it performs, so that clinicians can select the best monitor for their particular application.

Next, please.

In a similar vein, the alarm system also has delays in its response time and so you can see in this type of curve, there's an alarm condition threshold. And we use again, the standard recommends the use of a simulator input because you're not interested in the patient variability here, you're interested in characterizing the performance of this particular signal processing device and there are many delays and the standard goes into great detail about the sources of the delays in the hopes that the person using the box, clinicians, can adequately understand how their device is going to work and select the best monitor for their particular task at hand.

Next, please.

I mentioned that the standard has several informative annexes. It goes to a tremendous amount of rationale, particularly for safe surface temperature and for the shock and vibration, but each of the test methods and each of the requirements has a rationale. There's informative annexes that talk about how to determine the accuracy of the pulse oximeter or techniques to do that and it talks about how to calibrate them and what the desaturation profile should look like and how to set up the desaturation profile. And importantly, Annex FF talks about simulators, calibrators and functional testers and what those are appropriate to be used for.

In the early days, we had several oximeters that -- or manufacturers made 510(k)s where they tried to demonstrate or clear their devices based on the response to the simulator and we said that well, the simulator can demonstrate or verify a particular calibration curve, but it is not an independent assessment of accuracy.

Also, because this is an international standard, well, so Annex GG, we get into, the concepts of equipment response time and the alarms which I just showed you the pictures. Annex HH is a reference to the essential principles. That's the European Union has I think it's 53 essential requirements which lay out in an objective fashion what they consider to be safe. And because this is an international standard, it cross references the requirements to that and also it has environmental aspects of pulse oximeter, what

you're supposed to do with it when you're done with your oximeter, should you just throw it in the garbage or should you attempt to reprocess it?

Next slide please.

So, let me touch base for a minute on what's not in the standard. The standard does not have specific validation or I'll call them qualification requirements for neonates, that's clearly question number two on your list of questions. It doesn't have a requirement for disclosure of possibly poor performance on neonates. You saw that the guidance document talks about the degradation factor, the 1 percent degradation factor. The standard is mute on that subject. There is no requirement for, or a test method for motion artifact or low perfusion. I'm sure you've all heard about motion artifact in pulse oximeters as the hot topic of the late 90's and the early 2000's.

The standard talks about disclosure of your performance on motion as well as the guidance document also talks about that. But there isn't a test method for that.

The standard also talks about, or the standard has a recommended desaturation profile, but it's not a requirement and that's from a standard's perspective. So the next generation asks for the assessment of accuracy or calibration. And, so as I said, if would be nice if the standard, and the standard will address that in the next go-around.

The standard also has two test methods for, two recommended test methods for assessing safe surface temperature. One is a human body model, one is actually a piece of plastic which is calibrated or represents the human, so it would be nice if the standard had a single test method so that everybody could agree upon the one way to do it. But we took the approach of let's get it in the standard this time and the next go-around, which should occur in five years or less, will hopefully adopt one of those test methods.

Next slide please.

Let me say a few words about laboratory-idealized human subject studies versus clinical performance or clinical assessment on patients. It's a very important distinction between doing your laboratory, your accuracy assessment on a human subject versus a patient. And the importance of this is we need, we want to compare apples to apples, so when manufacturers report their performance on oximeters, we want the test method to be as similar as possible, so that it's the performance of the oximeter and the saturation algorithms and the probe configuration that is contributing to that accuracy as opposed to the types of subjects that were used or the experimental set-up.

And so I'll just run through some of the things that are very well controlled in an idealized laboratory study, which you can infer are not very well controlled in an actual clinical performance assessment.

So, clearly in an idealized laboratory study, they're performed on healthy adult subjects. There is typically no motion or electromagnetic compatibility. And I use the word compatibility because not only are we interested in making sure that the pulse oximeter itself is not interfered with, but there's also very sensitive measure equipment going on in your laboratory suite which could be interfered with and so you want to make sure that the electromagnetic compatibility is broad and across all your electronic devices.

The room temperature and the subject temperature a very stable and so if you will, hopefully perfusion is relatively constant. The treatment of the blood sample and the co-oximeter is very well controlled. You get the blood out as soon as you can and you zip it over to the co-oximeter and in fact, you might have multiple co-oximeters and take averages or have some other statistical scheme for making sure that the results that you're getting from your co-oximeter are as accurate as possible. There's been some studies to show that under poorly controlled conditions, co-oximeters can have differences in bias of up to 4 percent which completely swamps out the accuracy of your pulse oximeter that you're trying to measure.

The probe is typically held on and they're brand new probes and they're taped up very well and they're glued on and their attachment is assured and you're getting good signal strength and you're making sure that the pulse oximeter is in its optimal state and if you're a manufacturer you can know that because you've got your vision into the inside of the box as well as just what's displayed -- when I say inside the box -- you have vision or you can get vision into what the oximeter is doing as well as just to what it's displaying.

When you do a controlled desaturation study, you're able to calibrate or at least to identify what the transport delay is. If you're taking a blood sample from up top of the arm and you're looking at pulse oximeters down at the fingers and I'm an engineer, so I've now exhausted my clinical knowledge. But you can calibrate that transport delay whereas in a patient all bets are off. You're not really sure what the transport delays are. There might be some abnormalities going on and so you need to know those things.

And finally, particularly for low saturations, the saturation jumps all over the place in the desaturation profile and in a controlled subject, you are able to do some averaging and to try to figure out what the real value of the saturation is whereas in the clinic, it's much more difficult to get that value.

I've already alluded to most of the future work of the Committee. You can tell where the weaknesses are and those are issues that the pulse oximeter standards committees are actively working on. A selection of a single desaturation protocol and treatment of co-oximeter accuracy and statistics are front and center, more rigorous treatment of statistical power. The number of subjects is important and the number of sample points is important, but there shouldn't necessarily be a fixed number for that if you have a higher quality process and you get tighter variances, then you should need fewer data points.

And that's the ethical thing to do is take as few samples as you can. The statistics to support convenience samples has to be worked out, particularly in the neonatal setting and that's question 2 on your list of panel questions.

Building a test method promotion which we're working on, having a single test method for calibration and temperature and trying to figure out how to define low perfusion and assess that are important. And language to disclose that the accuracy under actual clinical use will be different from the ideal clinical accuracy which you measure in a laboratory.

Next, please.

And so we've gone through the guidance document and shown that with -- that we're actually keeping -- the guidance document lays out the basis for recommendations for substantial equivalence as they were put out in 1992 as well as our current thinking to bring the guidance document up to date. And I've gone through the standard to show that it's a reasonably comprehensive treatment of safety and essential performance. It needs some improvement and those things are being worked on, but I think it's a valuable tool that the FDA will leverage in the future to support some of our substantial equivalency determinations.

Are there any questions on that?

DR. LISBON: Charles.

DR. COTÉ: I had one question on the tone of the oximeter pari parsu with saturation. And apparently, there's no requirement that a pulse oximeter change tone as the saturation drops which means to me that those oximeters that are not configured that way, there's a delay in response time for the clinician to realize something is developing. And I could just go by my own experiences of a pediatric anesthesiologist. Years ago, when we used to use -- do bronchoscopies in babies, we had to wait until the heart rate dropped to 60 and then the surgeon would stop their bronchoscopy and we'd ventilate the patient and get them pink again and the heart rate up. But as soon as oximetry became available, as soon as that oximeter went doot, doot, doot, is everything okay up there? And we no longer had these arguments or discussions in the operating room.

So it seems to me it makes no sense to have an oximeter that doesn't take advantage of that wonderful monitor that tells you something is developing and you don't have to be looking at it.

DR. WEININGER: The standard currently has a requirement for that, for the variable pitched tone.

DR. COTÉ: But only going up.

DR. WEININGER: And going down, yes.

DR. COTÉ: It doesn't have one for going down. There are oximeters out there that don't have that change in tone.

DR. WEININGER: Conformance with the standard is voluntary, so what I would say is that those manufacturers that make that product are not producing a quote standard oximeter. So there's no legal requirement that a manufacturer conform to the ISO standard. And there's currently no recommendation in the 1992 guidance document to address that.

DR. COTÉ: Is that something we should be considering today or is that a separate issue?

DR. WEININGER: The ISO standard does address that. It does have a requirement for that tone and in fact, now that the Nellcor patent has expired, the actual frequencies can be put into the standard which they've agreed to, but if you think that that's important that the FDA consider that in its guidance document, you can make that recommendation.

DR. COTÉ: I think it's very important justin looking at demographics of accidents that occur outside of the operating room. There are a lot of these oximeters out there that don't change the tone and there's a delay in response that something is occurring.

DR. WEININGER: I think that's a very good question to address to Dr. Goldman. He's been dealing with alarms in the Patient Safety Foundation and Anesthesia Patient Safety Foundation and his presentation, you can ask him that question and he can give you a very good clinical perspective on that.

DR. COTÉ: Another question I had for you is this temperature, surface temperature being based on adult data. Obviously, the skin of an premature baby and a term infant is very fragile, very thin. I guess I don't know enough about that to see if this 43 or 41 even would be harmful to a baby that size.

DR. WEININGER: In considering an appropriate maximum safe surface temperature, the ISO and the ASTM Standards Committee tried to identify the worse case situation and we believe or the standards body that that was a neonate in an incubator which has the highest ambient temperature that the incubator can be set to which is roughly 35 or 37 and we said all right, we don't want the probe to go above 41 in that particular case. There's been some work in the '50s and the '60s and granted, it was done on adults and some others, but the evidence to date has shown that 41 is pretty safe.

DR. LISBON: Are there other questions? Yes.

DR. DRASNER: I'm a bit confused on one point. Does approval rest not just on establishing substantial equivalence, but also meeting the standards that have been set for these devices now?

DR. WEININGER: From an FDA perspective, and this is semantics, but we're talking about clearance as opposed to approval. Approval has to have objective evidence of safety and effectiveness, substantial equivalence is a comparison to a legally marketed predicate.

DR. DRASNER: Okay, clearance then.

DR. WEININGER: The way the regulatory policy unfolds is that we can use the standard in our decision as to whether a device is substantially equivalent. It's not a requirement.

DR. DRASNER: But when you set a standard that now is higher than what existed for a predicate device, what happens to that predicate device? Does that still exist?

DR. WEININGER: Yes. Once cleared, unless you can identify a safety hazard with devices that are currently on the market, those devices stay on the market.

DR. DRASNER: So can you still get clearance by establishing substantial equivalence to a predicate device even though you don't meet these new standards that have been set by the FDA?

DR. WEININGER: Yes. But the flip side is again that if there's an obvious safety risk, I mean latex is a perfect example. Until we figured out the latex, that there were very serious latex allergies, there's all kinds of products on the market with latex and then once you determine that latex is a safety hazard, you can say well, those products have to be pulled from the market.

DR. TUNG: How do you seek and process the clinical feedback on the standards that you've set?

DR. WEININGER: I'm sorry?

DR. TUNG: How do you seek and process the clinical feedback on the standards that you set? Is there a --how does that --

DR. WEININGER: From a standards perspective, clinicians have a seat at the table. The standards bodies, particularly ASTM require a balanced committee of clinicians, manufacturers and others to sit at the table and determine consensus standards.

DR. TUNG: And when you sat down in '96 to revise this, was there a pattern to the clinical commentary?

DR. WEININGER: I'm not sure I understand.

DR. TUNG: Consistent in the kinds of comments that you were getting back, the kind of clinical feedback or not really any pattern that you could identify. Say everybody was asking for a tone or something like that.

DR. WEININGER: I'm sorry?

DR. TUNG: For example, everybody was asking for a tone or everybody wanted this or that. Was there any kind of pattern to the clinical feedback or not really?

DR. WEININGER: Not that I can immediately recall, but I have to think about that for a while. I mean clearly the clinicians and I'll say FDA, their primary concern is patient safety. So in that sense their pattern is the requirements that have to be set have to support patient safety.

And I guess I would even emphasize that they were very adamant about patient safety, particularly when it comes to safe surface temperature and some of the other disclosure requirements.

DR. LISBON: Are there other questions?

DR. BIRNBACH: It may be semantics, going back to what Charlie had asked before, I'm a little confused about what you can require versus recommend. You made a distinction that seems to exist.

DR. WEININGER: Now you've eclipsed my pay grade.

(Laughter.)

I'm going to let the Branch Chief address the differences between requirements and recommendations.

MS. GRAHAM: In a 510(k), we make recommendations. We recognize certain voluntary consensus standards. We give guidance to industry and to FDA staff. When we recognize certain standards, if we recognize the entire standard or certain parts of the standard and we state in that recognition whether we adopt, recognize it in its entirety or not. If we have exceptions, we state them and we may point you to additional guidance on those standards.

There are requirements, regulatory requirements in a 510(k) that are set out in the CFR. But those are -- there are some exceptions, but in general, it's a broad list of requirements for any 510(k). It has to be labeling and manufacturer's -- certain manufacturer disclosures, etcetera. So were you thinking of something in particular?

DR. BIRNBACH: How about we go back to the alarms. Some of these are not requirements per se.

MS. GRAHAM: Right, right.

DR. BIRNBACH: Even though the expectation is that they will be present, is that correct?

MS. GRAHAM: That's correct. It has evolved over time as the indication for us has changed. In the beginning, and this is my understanding, I wasn't in the branch in 1978 when they first started to be submitted, but when we first received 510(k) files, they were for spot checking.

And because there was not a great reliance, minute by minute, on these devices, the necessity for an alarm was not felt to be there. But these devices have become primary clinical decision making tools, certainly within the last 10 or 15 years. They are now continuously being used.

And so in 1992 we added or developed our original guidance document to include in vitro validation of the pulse rate, for example, and other aspects that we expect the manufacturer to submit to us. Again, that's a guidance document and it's a recommendation. It's not a requirement.

So in theory, a manufacturer may choose to submit a 510(k) to us without any of the requirements, recommendations in their 510(k). As a practical matter, FDA would be hard pressed to fine that device substantially equivalent to a predicate, particularly if it's a newer predicate.

DR. BIRNBACH: But they can intentionally use an older predicate?

MS. GRAHAM: Yes, they can. Yes, they can.

And that's one of the difficulties when are faced with a new application that points to an older device where there has been little or no performance data provided, no alarms, the labeling is cryptic and these present challenges to us. And that's where we very often enter into a lengthy dialogue with the manufacturer.

DR. LISBON: Are there other questions?

DR. DRASNER: Another question. Isn't there a mechanism to impose standards?

MS. GRAHAM: Impose standards?

DR. DRASNER: I mean when you set these standards that would supersede this idea that you can go and go against an early predicated device, can you set standards that all devices must meet?

MS. GRAHAM: Well, when we develop a guidance document and we have two different levels of guidance documents, level 1 and level 2. A level 1 guidance document basically tells the world we're changing the way we think about a particular device or a particular regulatory decision making paradigm. And we will recognize that there are predicate devices already on the market, but from the effective date of that guidance document that would serve as a special control for manufacturers, this is the way we're going to do business. But in the absence of that, Sandy is going to embellish my remarks, and correct me -- but in the absence of that we do have to acknowledge that previously cleared predicate.

DR. WEININGER: Let me try to put this in some historical perspective. In '76, when the medical device amendments were written, Congress said to FDA thou shall develop mandatory performance standards for all class 2 medical devices, pulse oximeters.

Safe Medical Device Act of 1990 and 1992 recognize that that was an impossibly difficult task and relaxed the requirement to allow us to use special control guidance documents, that's one big word which are recommendations for what -- I don't want to say constitute safety and effectiveness, but what shows is substantially equivalent to a safe device already, as well as recognizing the use of voluntary consensus standards. So FDA over the last 5 to 10 years has invested a great deal of resources in upgrading and participating in the development of these voluntary consensus standards so that we can use them in our regulatory process. They serve several things. There are test method standards which you say all right, here's a test method and everybody is using the same standard so we can get some idea that these apples are -- the results of comparable.

There are other standards like, as I said, 601, which talk about electrical safety and FDA has recognized 601 that if you conform with 601, then your device is electrically safe. So standards have a lot of different uses. They can substitute for pieces of the substantial equivalency determination. They don't -- if you look at the Venn diagrams, they don't cover the entire sphere because there's lots of other things in a 510(k) which the standards don't address, but for the pieces that are appropriate from the scope of the standard, we can use those standards in our regulatory decisions.

I sense that there was some confusion about the requirements of a standard. Those are not, if you will, imposed -- FDA cannot require those types of tests or requirements. FDA has recommendations.

Is that clear as mud yet?

DR. LISBON: Are there other questions for our previous two speakers?

Dr. Weininger, can you just give me a little background on where this 1 percent degradation for the neonates came from?

DR. WEININGER: Yes, I can give you a little background. It's not well captured, if you will, in FDA policy as to where that 1 percent comes from. So to the best of my ability to cipher this out, in the mid '90s, there was some question as to what the influence of fetal hemoglobin was on actually the co-oximeter. There were some studies shown that the co-oximeter is inaccurate, particularly at low saturations with a large percentage of fetal hemoglobin. Of course now we have to ask the question of what's the percentage of fetal hemoglobin in neonates and then you have to ask how quickly it clears and how quickly, what patients were actually applying these to, so you go through all those reps. But that was kind of the basis of them.

On the flip side in roughly '89, I believe, NIH had a consensus panel which talked about the influence of fetal hemoglobin and they said yeah, there's clearly an influence, but we don't think it's clinically significant. And so we've got several articles that show it is significant and some articles that show that it isn't significant and so the middle of the road compromise was to allow this 1 percent degradation and recognizing that you can't do a desaturation study in a neonate and that the convenience samples you typically get only at higher saturation values and Dr. Goldman is going to address many of these issues shortly.

DR. LISBON: Are there any other questions? Yes?

MS. GRAHAM: I would like to clarify the neonatal question earlier. May 14, 2004, FDA published a document with a catchy tune of FDA Guidance for Industry and FDA Staff, Pre-market Assessment of Pediatric Medical Devices. And in that document, we defined neonates as from birth to one month; infants from greater than one month to two years; children from greater than two years to 12 years; and adolescence from 12 to 21 years.

DR. COTÉ: And is that required of industry to follow that?

MS. GRAHAM: It's a recommendation.

DR. LISBON: In the interest of keeping things moving, I'd like to move on to our next speaker, Dr. Goldman.

DR. WEININGER: Let me just give you a brief introduction for Dr. Goldman. Dr. Goldman, I think is uniquely qualified to talk about the clinical aspects of the three questions in your panel because he's been on both sides of the fence. You can see now he's with Mass. Gen and Harvard. He was previously -- he worked for one of the manufacturers as their medical officer. And so I think he's going to give you good aspect. He is a Visiting Scholar under FDA's Medical Devices Fellowship Program.

So I will turn it over to Dr. Goldman.

DR. GOLDMAN: Thank you, Sandy. Good morning. I'd like to clarify something before we move into here. I know that some of this is confusing to everyone because it's a complicated topic and it asks and invokes for concepts in the standards world which uses one set of terminology and the FDA regulatory world which uses another set of terminology sometimes. But I know that it's a hot topic right now, the issue of pulse oximetry variable saturation tone because of some of the discussions that are recurring within the world of the Anesthesia Patient Safety Foundation and the ASA. And so I'd like to clarify something that Sandy said and that is the standards that are published, the standards, that is, the ASTM and the ISO standards, do not require the use of the variable saturation tone. What they state that if present, the pitch of the tone should increase as the saturation increases and should decrease as it decreases, because obviously it would be terrible if some manufacturers decreased the tone when the saturation went up and others did it the other way. So that's what it's about. That's one important point.

And the other is that question about alarms and things like that and we have to keep in mind that some pulse oximeters are designed for what you could call spot checking and don't have a need for physiological alarms, it wouldn't make sense. It would add to the cost and others, if necessary.

And another principle that is int he standards world is typically that of minimum safety and performance; to leave room for differences in features and models and costs, but to make sure that at a minimum there is something that's safe. So I'd rather not spend more time on that, since that was someone else's material.

So I have -- I will try and cover some highlights, some key issues with that address specifically the questions that you've been asked to ponder today.

The first concept I'd like to comment on is the importance of understanding the underlying model of pulse oximetry. There's a model that encompasses assumptions about the physics and the physiology of pulse oximeter and some or, in fact, much of the improvement in pulse oximetry over the last decade or two decades is related to improving the -- is deeper understanding of the model and then the ability to design and manufacture an instrument that could accommodate for the variability in that in clinical conditions.

One of the important assumptions in the model we can show graphically and that is that the thing that we care about in pulse oximetry is the change in finger blood volume over time and that change in finger blood volume over time is changing the incident light on the photo detector and the light decreases as the blood volume increases.

That's the signal that we are attempting to measure i pulse oximetry and sometimes it's successful and sometimes it isn't and any other signals that are produced for any other reason are things that probably confuse the picture and fortunately there have been great strides in that area.

So let's -- so in thinking about the model, we should keep in mind that the clinical use of the pulse oximeter will be most successful if the use conditions match the model assumptions. We should also keep in mind that calibration is out of necessity performed on healthy adults under laboratory conditions and not on sick patients under clinical conditions. And again, the model is an ideal model usually, although that's changing.

Now the problems creep in in terms of clinical performance when there's divergence of clinical conditions from the ideal lab conditions and that may contribute to inaccuracy. Examples would include dyshemoglobinemias, that is higher than the minimum amount that would be present in the laboratory condition. Venous congestion, pulsatile venous blood because the model assumed that primarily it's the arterial blood pulsation that's being detected by the photo detector most of the time. And extreme vasoconstriction, probably for complex reasons.

In the clinical setting, just a quick list to help jog the memory and expand the picture, in no specific order a list of errors in the clinical setting may be due to device failure and typically high readings in pulse oximetry are not infrequently due to a problem with a sensor or the cable, more so than a physiological statement.

Motion conditions, very low profusion with poor signals and noise ratio. The presence of pulsatile absorbers that are not the arterial signal that we care about. That would include venous pulsation from a wide open tricuspid valve, severe tricuspid valvular regurgitation, particularly when measurements are being made on the head and the venous pulsation is transmitted in a strong manner right to the measurement site.

Large artery pulsations. It has been written in some places that you should put the pulse oximeter over an artery, let's say it's on a neonate and if someone is having trouble getting a signal, people have erroneously instructed in the literature to put it over an artery. Well, the artery -- the reason it's called a pulse oximeter, as you well know, if you put it over an artery you end with bad data. The light cannot penetrate a larger artery. And so you're not actually measuring the blood in the artery. But that happens. And some people are misinformed and actually do that.

Any other conditions that are different from that lab ideal calibration setting, that venous congestion and also poor sensor fit would fit under the category. And then there are the environmental conditions which produce those conditions that are listed above and you can -- some people think of challenging states and describe environmental conditions instead.

So let's go to Question 1 and what I'd like to do is go through the Questions 1, 2 and 3 without spending time on the questions because you already know the questions, but maybe talk a little about some of the issues that will affect how you think about the answers.

So Question 1, in general, we are addressing the differences between transmission and reflectance pulse oximetry and technology. Well, the first thing is to recognize that reflectance pulse oximetry and transmission are based on the same model and the same assumptions about light absorption and the signal that we're interested in.

But once you get beyond that point, there are other effects that may be different. For example, with reflectance pulse oximetry, one may have to consider reflection from underlying structures such as underlying bone which could overwhelm the photo detector if there's a light reflectance.

Another factor would be venous pooling. Now we know venous pooling can be a problem with transmission pulse oximetry. With reflectance pulse oximetry, it may occur under different conditions. For example, if the sensor is placed on the head and the patient is tilted in Trendelenburg's position and that's why we typically see the application of pressure over the reflectance measurement site.

In addition, transmission pulse oximetry probe design constrains placement and may reduce application error. A finger sensor has been carefully designed to fit a finger. So it's less likely for that sensor to be put intentionally to be put on incorrectly, such as sideways, whereas a reflectance sensor, if it's a flat sensor, could be put in the wrong place. It could be moved a few inches to the right or to the left which may be an incorrect measuring site. And that's another thing that we should consider and what should we do perhaps to minimize the likelihood of that type of use error.

Continuation of the issues related to Question 1, do reflectance sensors need better instructions for use or operator feedback from the instrument itself? In the example, for example, with the placement, when one is using a forehead sensor, it's designed for use on a specific area of the forehead. That area does not include the temple where it would be placed over an artery that's pulsating and introduce the problem that we mentioned.

So what do we do about that? How can that be addressed? How can we minimize the likelihood that clinicians will make that error? We know how hard it is to train clinicians, all clinicians in all environments. It's absolutely impossible. So what we can do from the design or the educational perspective?

Another example is that some recommendations are made both by manufacturers and in the literature by clinicians that one can use a signal strength or what is sometimes called the perfusion index to select an optimal measuring site. So perhaps a finger probe is put on a finger and the signal is poor and the reading isn't very good or stable and one sees a low signal strength, so one can move into another digit and perhaps get a stronger signal and may get a more reliable measurement.

But if you use that same knowledge, that same trouble shooting and clinical diagnostic knowledge and apply that to a reflectance sensor, and move it around until you get the best signal, that may be the wrong place to make the measurement. It may perhaps be over an artery with a great signal, but it's not giving the right data.

So again, the question continues, how do we transfer that knowledge, how do we educate, perhaps with the instruments. Do they need to be designed to provide more information? Is that even possible.

A related question to these issues in Question 1 is should reflectance probes be specified for use only on sites that have been validated? So let's take again the example of the forehead probe. Normally what would be done is the forehead probe would be calibrated on adult laboratory volunteers under ideal conditions and then if it was used, if the probe was used on a neonate, specified for use on a neonate, we would expect to add that one percent and the technical term for that one percent is the fudge factor. And that's the way it would be used. And that would be the typical state today.

So one question that we would ask, if you asked about the use of the sites is would a convenience sample, would samples that could be obtained in the neonatal population on a convenience basis in a clinical environment, would it help validate that that concept was right?

Similarly, we would ask, if there a significant difference between the same site in adults and neonates? Is the neonatal forehead essentially the same as the adult forehead? Is it appropriate to make this transfer of the technology of the measurement and just perform that with that one percent fudge factor or are there anatomical or physiological differences between the adult and the neonate that would lead us to be concerned about just transferring that measurement?

One possibility, for example, and I think the data -- I don't know how much data there is in this area is that it has been mentioned that in children with cyanotic heart disease, there may be a higher peripheral venous pressure and we know that if there is an increase in peripheral blood volume, that may affect measurement accuracy and perhaps that would be the case as well in these patients. And if so, should there be data collected on those patients to validate the concept that you could transfer the adult data to the neonatal setting?

Another question is is it clinically acceptable to specify use of probes for other sites? For example, if we have adult laboratory validation data, if we have the adult forehead and if we've agreed that it's acceptable to then use that probe on the neonatal forehead, is it also okay to use the probe on other areas of the neonatal body, such as the abdomen or the back. Currently, the FDA would expect that there would be some validation data for the site for which it would be used. But there are people who do use clinically and report it and there's data in the literature to -- for use of probes on other sites. And we should at least think about is that acceptable practice or not, how would one approach that.

Question 2 relates to the use of that 1 percent accuracy reduction factor. I really want to say fudge factor, but it may be more appropriate. Accuracy reduction factor, is that clinically appropriate? And as Dr. Weininger said, he talked a bit about the history which is a bit nebulous, but the current practice to add 1 percent to the specified accuracy from the adult laboratory setting to estimate performance in neonates under ideal lab conditions. So keep that in mind. That 1 percent would be, would express the expected performance on neonates if they were in that same setting which is the ideal laboratory conditions.

So as Sandy said, the rationale is confusing, but it looks like the clinical accuracy data that does exist in the literature now in numerous studies of the devices under clinical use does seem to suggest that the current approach when reflecting on that, does make sense and that it is an appropriate approach, it seems to be the case.

So let's ask some other questions. Would post-market convenience samples be useful to support the clinical assumptions?

Should there be a need? Should there be thinking of collecting neonatal clinical data, but doing that under a post-market setting, after FDA regulatory approval and clearance of the device?

Why would that -- what would the value be? Well, current regulatory practice as has been discussed at some length, does not require testing on neonates for neonatal probes. Although it may be a commonly performed practice, so at most and potentially possibly all manufacturers may test neonatal probes in some way. They have to make sure that it fits correctly or they should make sure that it fits correctly, that if it's an adhesive it sticks in place. If there's a wrap type sensor that it's comfortable and so forth.

A lot of that is being done currently, it appears and one should ask should there be a requirement for that? What is the expectation clinically? When a clinician opens a package and applies a probe to a patient, are they assuming that there has been clinical testing or not? And if not, is that an issue of education or should there be a change in the requirements?

The other interesting thing is that manufacturers are ideally suited to perform some testing and they can and generally do a much better job than the clinician who has access to clinical material and may have excellent clinical knowledge, but doesn't have access to the underlying engineering information about the instrument or the data collection tools. Manufacturers can collect data that would include data about the amplifier settings in the instrument and the emitter settings and so forth.

Question 3 is with regard to over-the-counter use of pulse oximetry. Now if pulse oximeters were cleared for over-the-counter use, it would require pre-market submission and an assessment of the instructions for use for lay people. One of the key parts to the decision has to be can a lay person go to the drug store, by a pulse oximeter and use it safely and appropriately?

Well, let's look at some of the concerns. One of the obvious concerns would be if there's incorrect application of the device leading to bad data, could that result in bad outcome? Well, that's reasonable. Kind of self evident. How about errors in interpreting and using the data? Is it possible that someone will see a low saturation value, 50 percent and think that that's normal? Maybe, who knows. Maybe they'll see it in the 70s and think it's a passing grade. It's probably a letter C grade and the patient is okay.

How will they interpret the information? And will having a device like that provide a false sense of security? Okay, reasonable questions, I think, for you to consider.

Let's look at the mitigating factors that would mitigate those concerns. What if there are use errors in the pulse oximeter? Well, as we know, based upon the historical use of pulse oximetry, most errors produce erroneously low saturations which would produce false positive results. Therefore, to be a false positive detection of hypoxemia, not a false negative and it's more likely that the consumer will see the value and especially with the instructions for use are clear, would call a clinician for advice, just because of the nature of the technology. It's less likely that they will see a falsely high value and then ignore a problem. So I think that's one of the mitigating factors that we should think about.

Now let's think of the benefits. If pulse oximetry becomes available to the population at large in their local drug store, the benefit would be empowerment of patients to detect potentially seriously medical problems. So when their neighbor is complaining that they're short of breath and they call their friend who has a pulse oximeter who has a low saturation, perhaps they'll seek medical attention much sooner than they would have otherwise. That might be a good thing.

It also starts to prepare for innovation in health care in this country in terms of moving measurement ability into the home and in our institution and Partners Healthcare, we are seeing development of things which are occurring elsewhere in the country like the ambulatory practice of the future which are ways of disseminating, distributing medical devices of some type to patients at home, finding ways of acquiring the data and interpreting it remotely and so forth. This is the movement int he country. And so as we start to think about over-the-counter use of these devices, it's probably an important question in terms of public health and safety in that we may be empowering people to take better care of themselves.

And probably more applications will develop as we start to deploy technology like this nationally. For example, there's clearly a need for easy and safe titration for patients who are home oxygen. And they're generally very knowledgeable consumers and they may use technology for things that we never considered once it out there in the home such as biofeedback, assessment to athletic conditioning and things of that sort.

So finally, just some other validation issues to think about. Should probes be tested in ill patients prior to marketing? Right now, as has been discussed at length, that isn't a requirement.

On the one hand, from a clinical perspective, from a clinician's perspective, one would say of course I want these devices tested on sick patients before I buy it. I want to know that it works on ill patients. But although that sounds reasonable, it's very simplistic and probably things to consider is the variability in illness, the implications for the stability of measurement data, what metrics would be used, how many convenience samples might be required, could those patients be found, would they be in the target saturation range of interest? So important questions and ones that have really challenged people for a while.

Next question. Would post-market convenience sampling provide a needed reality check? Post-market samples may be easier to obtain due to less stringent IRB requirements because if the device has been cleared and it's available for sale in the country, that alters some of the way that's handled through the IRB. So would it be reasonable to start collecting and expecting a report from manufacturers of the performance of the instrument after it's released for marketing and sale. Would that help with the reality check on the design and on the use of the instrument that it censor?

And finally, the question that we alluded before and this gets to the issue of how does one get good convenience samples and in which populations. It's very tempting to go to the low saturation range of children with cyanotic heart disease. But there is an interesting issue with those children and we probably need more information about the nature of their peripheral circulation. Do they represent the ideal patients for which to get low saturation data or are they somehow different than other children who are experiencing a transient event such as a respiratory event and a transient desaturation? For those children, they probably don't have the same peripheral circulation or may not have. A question for you to consider and perhaps you have the expertise among you to think about that.

And finally, I think we have to recognize and we all know that the knowledge, the clinical knowledge of the average and potentially even expert clinician about the correct and ideal use of pulse oximetry, it's strengths and weaknesses and interpretation of the data is probably inadequate. And perhaps it would be reasonable to talk about, think about, what can be done to support effective clinical use, perhaps education or what other things might really help.

I'd be happy to answer questions.

DR. LISBON: We've got about five or so minutes for questions.

Dr. Birnbach.

DR. BIRNBACH: I'll make it quick. Two questions about your mitigating factors in no particular order. While I understand that it might be nice for someone on home oxygen to also have a pulse oximeter, isn't home oxygen based only on a prescription and you need to be followed? It's not like I can say I want to use home oxygen today.

DR. GOLDMAN: Well, in reality, the patients remove and apply their oxygen and they're instructed to, frequently, based upon how they feel. And so they come back to the office and the clinician has essentially no data, no objective data. They have the patient's report on how they felt, but wouldn't be helpful to know that when they felt bad their saturation was low and -- or maybe their saturation wasn't low and they felt bad for other reasons. It's another area where we don't have all the information.

DR. BIRNBACH: And right now, can a physician prescribed a pulse oximeter, for example, purchase one and give it to a patient?

DR. GOLDMAN: To the best of my knowledge, yes. I believe they certainly can.

So it gets to the area of the cost and the distribution of tools to the population, more so than the core ability of being able to get to that measurement. Absolutely, you're right. You can get the data if you want it today.

DR. BIRNBACH: And the second question is is regards to device failure. Did you suggest early on that a device, if it failed, was more likely to give you a high pulse oximeter reading?

DR. GOLDMAN: What I said was or what I intended to say was that a high, an erroneously high reading is more likely due to a device failure, that is, equipment failure, as opposed to a physiological reason for that.

DR. BIRNBACH: And do we have data about device failures and how often that gives you a falsely high reading?

DR. GOLDMAN: I wish I had that data.

DR. BIRNBACH: Because if you are describing mitigating factors, and suggesting that anybody can carry around their own pulse oximeter, they might not have the clinical basis to say this reading doesn't make sense.

DR. GOLDMAN: That's a very good point and perhaps that should be in the consideration for OTC use, what is the likelihood of reaching -- of having values that are erroneously high in patients who have a physiological low saturation.

DR. LISBON: Dr. Otulana?

DR. OTULANA: Let me just answer one of the questions you raised about use of home oxygen. Yes, we do give, we do prescribe pulse oximeters, pulmonologists do that -- for these patients to have spot checking of their SpO2 at home, so that is done often.

The question I have, you showed a number of factors that would affect the accuracy of the pulse oximeters at home. This is an interesting question in terms of the extrapolation from adults, healthy adults to sick adults on one hand and extrapolation from healthy adults to sick neonates on the other.

Do you, just your own estimate, do you think the extrapolation from healthy adults to sick adults is the same level of risk when we extrapolate from healthy adults to sick neonates or have I confused you?

DR. GOLDMAN: No, no. I think your question is a good question. I think that's frankly one of the questions that is being presented to the panel to consider today is those clinical issues.

DR. OTULANA: Right.

DR. GOLDMAN: And I think the answer to that will -- what will influence that to some degree is the kind of illness, the kind of conditions we're talking about.

DR. OTULANA: Right.

DR. GOLDMAN: And how it affects the peripheral circulation and the stability of the values.

I think that one of the -- a few of the really interesting problems here are we all have a clinical sense of the performance of pulse oximetry. With just three years of use, we think we understand it and we think we understand when the values are right. But how do we know? How do we get that information, that clinical sense? We're getting it from the pulse oximeter itself and we don't have, for example, another gold standard that we can use. The only gold standard we have is co-oximetry and that doesn't provide continuous data. It doesn't even provide real time data.

So one of the core problems with understanding pulse oximetry and it's true performance in those challenging populations is that we don't have a gold standard to use and that's why this is, I think, an unusual monitor in a particularly challenging setting.

If we had a $50,000 instrument that could provide continuous reliable co-oximetry data, then at least it could be used for research purposes and we would know what the performance of pulse oximeters was.

DR. LISBON: Dr. Leung and then we'll break.

DR. LEUNG: I have a question which could be brought up again during the discussion later with the Panel. My question is what would be the real advantage of recommending post-marketing measurement versus pre-marketing measurement. And I'm thinking of the equivalent in the drug. You're requesting only Phase 1 data currently for marketing and we're not looking at Phase 2 or Phase 3 data.

The question is is there really and obviously to the manufacturers it's advantageous, but in terms of public health concerns and a false sense of security to the users and for pulse oximeter it will be clinicians and nurses that is FDA approved. And the liability, so I think that might be something to consider.

DR. GOLDMAN: I can give a quick answer although I'll be going out on a limb here. I think the answer relates to the fact that we have a few decades of historical use of performance of pulse oximetry. Generally speaking, the instruments seem to work. They seem to have saved many more lives than they put at risk. The approaches that the manufacturers are using seem to be generally okay, although there's room for improvement. And I suppose if we were all here today discussing a new instrument that no one ever heard of before and wanted to use called a pulse oximeter, that probably would be exactly what we would be talking about I would think is getting the data in that manner.

So I think that analysis is being colored by the historical use that we have and then the question is is it worth the additional cost, expense and delay to market given that history. Those are the questions. And the answer is, I think, subjective.

DR. LISBON: We will have time for questions as we go further on during the day. I have about 10:17. We'll reconvene at 10:30. It's time for a short morning break.

DR. GOLDMAN: Thank you.

(Off the record.)

DR. LISBON: All right, if I could get started again. Next on our agenda are presentations by our industry stakeholders. There will be four presenters, actually three presentations. However, I would ask that the presenters limit themselves to 15 minutes with five minutes for questions, please, and please identify yourself for the record.

I believe that we're going to start with Dr. Paul Batchelder from Clinimark and I apologize if I've mispronounced your name.

MR. BATCHELDER: No, that was exactly right, Paul Batchelder.

Thank you for allowing the time for me and my colleagues to speak. I am the -- I'm with Clinimark. As you had mentioned, we are an independent physiological monitoring and research facility with one of the few desaturation or hypoxia laboratories in operation today.

We recently separated our labs from G.E. Healthcare, but for the last 14 years have been the head of Omeda Clinical Research in G.E. Health care.

The important piece there, the reason I tell you that is that Omeda with Nellcor were one of the two pulse oximetry developers that started out in the early '80s, '82, '81. And so we've got a lot of experience validating and testing these devices. The desaturation laboratories, they do desaturation studies that you've been hearing about this morning. They were instituted and developed in our types of laboratories way back in the 1980s.

The question that we're going to deal with today, that you're dealing with is how can pulse oximetry be effectively evaluated in neonates. What we've done is we've evaluated around 43 published peer-reviewed journals, papers, over the years that dealt with accuracy of pulse oximetry in the clinical environment in adults and neonates. I believe that you received a copy of the letter that we sent to Neel earlier today. Is that true? Does everybody have a copy? Yeah, yeah.

Unfortunately, you won't have time to read that and digest all of the information during my talk. I'll talk to that and refer to various sections there.

The papers that we evaluated, we evaluated the ones that reported bias in precision in such a way that we could calculate the ARMS accuracy of the pulse oximeter. Now the ARMS accuracy is similar to a standard deviation. It's what accuracy specification is published in the pulse oximeter manual and it's what is submitted to the FDA for regulatory concerns.

Now Ann's table, Ann Graham, that mentioned, she mentioned earlier about the -- that she displayed, that had accuracy of less than or equal to 3 percent for adults and I think the greatest inaccuracy of acceptable was less than 4 percent, those are the maximum acceptable limits. Generally, pulse oximetry that you use today has an accuracy specification that was developed on healthy adult humans in the desaturation laboratory of plus or minus 2 percent.

And then you've heard about the 1 percent fudge factor. I'll use that technical term that has been applied to the values from the adult laboratory to claim accuracy for the neonatal environment. The key points that I'm going to go over today are the fact that we believe, the literature suggests the accuracy of pulse oximetry in the hospitalized population is not dissimilar between adults and neonates. We'll talk about that in a little bit more detail in a moment.

The theoretical and published influence of fetal hemoglobin is clinically insignificant and the 1 percent adder, is based on anecdotal evidence and is incorrect. It took quite a bit of work with our -- the best retrospectroscope that we have to look back and try to determine where that adder came form. I'll talk about that a little bit.

Then we surveyed the neonatologists. That survey is not in your -- in the letter that we sent to get a spot check, a sanity check on some of the questions and to try to determine, well, how can we validate accuracy in the neonates or verify the accuracy as the case may be? And all of the feedback that we got from our neonatologists on both seaboards and in the middle of the United States was pretty much unanimous. They said collecting a statistically meaningful number of simultaneous SpO2 from the pulse oximeter and SaO2 from the co-oximeter over the full range is unfeasible. But there is hope. There is a feasible approach to effectively evaluate pulse oximetry in neonates and provide that recommendation also.

Now what we're dealing with here is accuracy in various areas of this matrix. Currently accuracy specifications are determined and developed in the adult laboratory and the 1992 review of guidance document states that we can use the data on adults to make claims in pediatrics, but not neonates.

Then the question that we're going to talk about today is well, how does the accuracy in the hospital environment relate to that laboratory determination? We will not be talking about neonatal desaturation in laboratory because that's unethical and for other reasons that's not available at all.

So our question is how does the neonate in the hospital environment, accuracy of a pulse oximetry look. How does the adult accuracy look? And how do both of those compare to the adult accuracy in the laboratory?

Let me just run over quickly, this is a picture from a laboratory before our study was started, many years ago back in the early 1990s. They've improved quite a bit since then, but the point here is is that we use consented healthy adults and our purpose is to provide reproducible results.

In the controlled laboratory setting, we screened for interferences like total hemoglobin, high blood pressure, dyshemoglobins, perfusion instabilities and that's a very important piece and the purpose of all of this is to establish a stable plateau to provide stability in the arterial tree, to reduce lag time because we sample at one point of the co-oximeter value and we measure the pulse oximeter values at another point. In addition to that, the co-oximeter is a point in time measurement as pulse oximeters are real time continuous measurement.

Now the way we do attempt to maintain stability in the laboratory is by controlling the title volume of the subject, the respiratory rate movement and dynamically control the inspired gases. We actually have a computer-controlled gas mixing system that allows us to fly the subject saturation to appropriate levels, remembering that the purpose of the physiology of the human is to maintain homeostasis and that is not homeostasis at a low saturation. The body is always trying to pull the saturation up. So it's quite a feat, even in healthy, stable volunteers, to maintain a stable saturation.

So laboratory to clinical. The clinical is subject to inherent noise, the saturation differences between the A line and the pulse oximeter due to hemodynamic instabilities and the unstable oxygen saturation.

Now there are other sources of clinical inaccuracy also, but the biggest difference can be created quickly by hemodynamic instability.

So the actual saturations at the arterial sample site and the pulse oximeter site are often not precisely the same due to time lag, fluctuation in the saturations, differences between the reading types. One is a static point in time reading and the other is a dynamic reading. Oftentimes, the blood that bathes the arterial catheter is not the same saturation as the blood that's bathing the pulse oximetry sensor site and therein lies the rub.

If your ruler is bounding around, you cannot measure the length of your device very well. If you co-oximeter says it's 12 inches and exactly at that moment the co-oximeter sees the 12-inch or a saturation of 85 percent, but the pulse oximeter is still being bathed in blood that is of a higher saturation, then it's going to give a wrong reading.

In other words, to kind of put a point to it, the arterial catheter in adults is usually placed in the radial artery and the -- and in infants, in the neonates, the arterial sample site is usually the umbilical artery and then the sensor site of the pulse oximeter is placed on a finger in the neonate. Most often, it's on the foot, depending on the type of sensor, of course.

And as the blood oxygen level changes, it doesn't reach both sites at the same time. And as a matter of fact, there are several studies recently showing that the delay in time from the arterial sample site to the finger site can be as much as three minutes in some cases. Oftentimes, low perfusion will cause shunting of the blood and the actual highly perfuse or a change in the saturation blood reaching the sensor site will be several seconds or minutes later.

They're both, the co-oximeter and the pulse oximeter could both be reading correctly, but until they both are bathed in the same saturation of blood, they're going to give inaccurate results.

So, what did we do? We reviewed the papers on adults that evaluated the accuracy of pulse oximetry in the clinical environment and we evaluated papers that did the same in neonates. Twenty-eight papers in adults and 28 papers in neonates and I see one of the authors of some of the neonatal papers is here.

What we found after calculating from the bias and precision ARMS of all of the papers, was that in the adults we see about 3.26 percent plus or minus accuracy in the hospital; when compared to the laboratory, very carefully controlled stable values. And interestingly enough, this is the important piece here.

The neonates showed a 3.44 percent. Very similar data. And when we overlaid the data, looked at the distributions and the T tests and F tests, we could not show that they were different. What we think this means and these data, this particular slide is not in your -- not in the paper, but both of the graphs that make up this, the red for neonates and the blue for adults are in your handouts and in the letter.

The bottom line here is what this means is that the laboratory is a very good place to measure things. It's a stable ruler, but the hospital, remembering back to the distribution here. We have a fairly large distribution in the hospital which might indicate that it's much harder to maintain a stable saturation that is bathing both the arterial catheter sample site and the pulse oximetry sample site at the same time.

What this means is that the hospital environment, in the hospital, the adults and the neonates show about the same accuracy precision and the laboratory though is much tighter. So maybe the 1 percent adder that I'm going to talk about next, more reflects the laboratory to hospital difference than neonates to adult difference.

Well, where did the 1 percent adder come from? We started searching back and looking at when it first showed up in the accuracy specifications and I think it's associated, we think it's associated with the time when the Nellcor N 100 was evaluated.

Jennis, his work is in your letter that you just recently received. On page 20, there is a graph, graph number 23, or paper number 23, that shows a bias and since we at that point in time we did not have the wealth of experience that we have now in pulse oximetry and a large number of clinical papers, to be safe, we thought it might make sense to add a 1 percent inaccuracy adder, just so that we didn't provide the clinician with inaccurate information.

Now Jennis showed though that the accuracy in his study was 2.6 percent which is pretty close to the 2 percent that we see in the laboratory. The bias, however, was correlated with fetal hemoglobin, but their literature review concluded that the fetal hemoglobin should not be a factor.

Then another paper looking at the effect of total hemoglobin, showed that the error was not related to fetal hemoglobin and in their study they've shown that fetal hemoglobin levels did not have significant effect on accuracy.

Pologe and Raley, one of whom is my colleague at Clinimark took data from Dr. Ziltra who lysed adult and fetal red blood cells and measured the extinction coefficient of the hemoglobin of both adult and fetal hemoglobin over the range in which pulse oximetry works, about 660 and 940.

And the thickness of this line actually depicts the difference of the pulse oximetry of error that could possibly be resultant from adult and fetal hemoglobin. So basically, interpolating from the data, the inaccuracy would go from laboratory 2.0 to 2.1 or a one-tenth increase with fetal hemoglobin.

Why not go to the ICU, collect data to support accuracy claims in the neonates? We've all heard I've got lots of kids with low sats. Use data from blue babies. This is our starting point. This is what we required in adults. The neonatologists that we spoke with were all very animated and very unanimous in -- and we also checked with some data collection nurses. Basically, the kids with cyanotic hearts usually do not have A lines. They're usually larger kids and they often are hypertensive with high venous pressures.

The reason for the A lines, they're unstable. Things change quickly. And most lower sats are transitory. The low sats the clinicians are talking about is below 90, not where we need.

So I'll just jump through this since we've got to move on, but the limited availability of neonates with A lines who are within the lower saturation range and inability to obtain data under stable, controlled conditions makes collecting data in those kids unfeasible.

Well, what can we do? The most common saturation range of the kids is 85 to 95. It's up to 100 percent in the older kids, but of course they don't have A lines. That doesn't help us. The vast majority are being kept between 85 and 89, and 95 percent of all the patients in ICU C range is between is 80 and 98 percent.

So what do we do? These are the recommendations. I'd be happy to entertain questions about these or we can talk about it further later. But to continue to use the best router we have to determine pulse oximetry accuracy in the adult laboratory and then add, which we do not have at this point in time, physical testing on the intended population for form, fit and function testing of the sensor design to demonstrate that it's the right thing for these neonates, to evaluate the adhesive and so forth and then do convenience samples and discontinue use of the adder and clarify clinical performance in labeling.

DR. LISBON: Thank you, Mr. Batchelder. WE do not have time for questions. I am going to keep on a pretty tight schedule so that the next speaker is Mr. Paul Mannheimer from Nellcore and Tyco Healthcare. I would recommend, so that we do have time for questions, that you keep your presentation to 15 minutes so that people can have a chance to ask you questions.

DR. MANNHEIMER: Thank you. I want to address the issue of reflectance pulse oximetry and I essentially want to cover three important points as I step through this. One, I'd like to demonstrate to you that reflectance and transmission oximetry really ought to be treated equivalently. There is some misconception on what reflectance oximetry is. I'd like to address that.

I think there's value in our ability as manufacturers to provide products to clinicians, whether it's reflectance or transmission geometries and I'll talk a little bit about what additional value might be afforded by having access to reflectance sensors.

And ultimately, I think the appropriate path is to treat reflectance oximetry just like Mr. Batchelder described, to test it in the laboratory, validate it in the laboratory and do the verification testing of its appropriateness on neonates in the clinic to demonstrate safety, appropriateness and do some sampling versus blood.

First of all, let's talk about pulse oximetry, in general. This first bullet point that I grabbed, actually lifted from ISO 9919, although I paraphrased to fit within the confines of a slide. It's noninvasive estimate of SaO2 from light signals. It interacts with tissue using the time-dependent changes in tissue, optical properties that occur with the pulse of a blood flow. I think Julian did a nice job of describing where that's coming from.

That statement is not specific to transmission or reflectance. It's based on the principles of the modified Lambert-Beer Law, light diffuses through tissues, true for both reflectance and transmission. The detected signals travel through about a centimeter of tissue. I'll show you that in a minute. And it's empirically calibrated or validated on signals from human tissues compared to co-oximetry.

All of these principles are independent of where the emitter in the detector are located.

Let's take a look at the most conventional transmission design. It's intended to go across the fingertip or other opposable tissue surface that in all intents and purposes, the only reason we limit ourselves to the fingers, the toes, the foot or the earlobe and the nose, those sites are thin enough where we can get enough light through the tissue bed to detect and run into the pulse oximeter. We would do transmission oximetry across the head, if we could get enough light through, but most people, that doesn't work.

(Laughter.)

The photograph at the bottom is intended to show you and I'm sorry when I took this, I didn't preview that there was a wire running back behind here. This is the emitter of a pulse oximetry probe sitting on the pad of the finger. I've over-exposed it a little just so that you can see the light coming out.

This is the red light from the pulse oximetry probe. Typically, we would place the detector on the opposing surface, but note that whether you're driving a light from the end or the side or even on the pad, all of that light is scattered through the tissue.

It doesn't really matter. All of the information that the pulse oximeter is seeing to calculate a saturation exists in every photon that's coming out of there. The one that counts is where the photon detector is located.

Let's take a look a reflectance probe and I would like to shoot the person that coined the phrase reflectance and we were talking about this at breakfast this morning. I'm not sure where it came from because I think there's a lot of misconception of what that means. There's been some discussion, I think in some of your handout that reflectance probe requires some surface to bounce off of and that's incorrect.

I like to think of reflectance probe as a transmission probe, only it's sideways. So whether you're going north and south or east and west, we're transmitting light through a bulk of tissue.

An example of this, I have my handy demo here, if I shine my laser pointer into a glass of water which has just a little bit of creamer in it, we see that the light is scattering back, scattering sideways. If we were to collect the light that is over here, it hasn't reflected off of any surface. It's scattered through this bulk scattering media. Scattering length of light and tissue is about one millimeter. So as long as the emitter and detector are located along a surface, more than a couple of millimeters apart, you're measuring the transmission of light through that bulk surface. Had we put the detector directly over or right immediately, adjacent to the emitter, within a scattering length, that might be considered reflectance, but even then it's a bulk scattering through a depth of tissue. It doesn't require a mirror on the other side to reflect that light back.

How do we know this? Well, I'm going to refer to some modeling work that I have done and published in the past where I've used a random walk model through a homogeneous tissue bed in this example. A slab of tissue, about 12 scattering lengths thick, that's comparable to about 12 or 13 millimeters thick, but the light emitted at the surface and collected at the bottom. These gradations here are depicting where the photons had the highest visitation probability.

So the region where the photons traveled the most is up and down through this middle. But there is still a lot of scattering signal that's coming from the periphery, so this cigar shape, if you will, is describing the tissues that would influence the readings of the pulse oximeter.

In a reflectance mode, I'll continue to use that bad word, but really what I'm referring to is a transmission east and west. Once again, the light is traveling primarily through this region a few millimeters deep within the tissue. Very little is traveling along the surface. The light that scatters here tends to escape. It doesn't contribute. Photons that travel very deep into the tissue or in this case on the transmission case very wide, they have a greater likelihood of being absorbed and not being detected. Photons that aren't detected don't matter very much. They're not going to affect what the pulse oximeter reads.

Is the calibration the same for reflectance and transmittance? We can use this type of modeling to characterize what the signal of the probability of collecting a photon if the detector is at a systolic and a diastolic equivalent absorption in a tissue bed and from that generate a correlation between modulation ratio, the pulse oximeter measures and the SAT and we see that a reflectance probe and a transmission probe behave generally the same, but their calibration may be slightly different from one another. That's not a limitation, if we're designing our pulse oximeter system to behave as a system to work with the reflectance probe or work with the transmission probe. We simply need to build that calibration curve into the monitor, into the design. And virtually every manufacturer I'm aware of, does something along those lines.

I went back and looked through the FDA website of cleared pulse oximetry systems that specific a reflectance probe and I came across 11 510(k)s. There's 10 listed here and one PMA and Sandy, correct me if I'm wrong, there may be others that I missed, but reflectance oximetry has been available since the mid-1980s. There were a number of products cleared through the early 1990s. There's been a recent flurry of new products cleared. All of these in blue are designed and intended for use on adults. As has bene mentioned earlier, the only clearance for 510(k) products, pictures of them is shown at the top.

One product that I want to include here is the Nellcor fetal oximeter system. It was a PMA and I do want to share a little bit of insight with you from our experience developing that product.

One of the things I just want to mention because I'll get to it, this flurry of recent activity is coming about because I think there's a renewed interest in reflectance oximetry from what's been available in the past, one of the things I'll show in a later slide.

The monitors that were cleared through the early 1990s, I think it was alluded to a little bit earlier, were older generation devices. They couldn't tolerate the very small pulses. They might be a little bit more sensitive to movement and other artifacts. We've learned a lot in the intervening 15 years. Reflectance sensors aren't inherently low signal products, but they're placed in locations in the body where signals are generally weaker. The strongest signals that we typically see are on finger tips in a warm, well perfused patient. We try to place the sensor on the chest or the arm or the forehead, it's usually a substantially weaker pulse.

Monitors today are one or two orders of magnitude more sensitive in picking up those weaker pulses, so the performance we saw in the past, maybe we need to revisit today because we can do better with our monitors.

I'll talk about that a little bit more in a minute.

This is data we collected for the Nellcor fetal oximeter system back in the late '90s and it's not to say that manufacturers can't go out and gather this data. We can collect data on neonates, although when we started to target the true neonatal population, we do become challenged.

This is data to validate the accuracy or the performance of a Nellcor system. It's a reflectance probe designed for use in fetuses, but since we couldn't take blood draws from fetuses any more readily than we can do bleed downs on neonates, we had to go to the closest model we could find. We found 27 cyanotic infants and children. Only a few of them would fit this definition of a neonate. Many of these kids were outside the scope of less than 30 days. We were able to gather 72 observations over a sat range of 41 to 93. In fetal oximetry, we're really very interested in determining what the accuracy is down in this 30 to 50 percent range, less so at the upper end. We place the sensors by hand or with a head band on the cheek, temple or forehead. Some of those locations we now -- actually, we knew then, are not ideal from the standpoint of pulse oximetry, but for the fetal environment where you don't know precisely where the sensor is located, we want to make sure we included those less than ideal sites as well.

And we used arterial samples that were taken from the same circulation that was feeding sensor site, particularly in these cyanotic kids or shunting kids, the feet and the hands and the head may be very different.

Just to point out, the study took us more than a year to accumulate these 72 observations using three sites and a full-time research nurse on call to get on an airplane and go to the facility that had the patient that met our criteria. It's possible to do these things, but it's very time consuming and expensive to do so and as Mr. Batchelder presented to you, the value of the results that we get don't seem to add a lot to the equation.

Let's take a look at the vulnerabilities of transmission reflectance oximetry because ultimately what I'm trying to demonstrate to you is that there's really no practical difference between reflectance and transmission.

One of the things that can occur with transmission sensors and I'll grab another one of my samples here, is something that was coined -- I hadn't heard this phrase other than reading in Kelleher's paper on penumbra. We all refer to it as shunting. But a sensor properly applied to a fingertip places the emitter and detector so that the light that's detected necessarily has gone through good blood perfused tissues. If the sensor is malpositioned so that some of the light can bypass the tissue, it disrupts the pulse amplitudes that we measure and will give us an erroneous reading.

So a sensor that's improperly applied, it may be applied sideways and I think Julian referred to this or on the neonate's foot improperly could potentially shunt and this does occur today. Every one of your institutions, I'm sure there are a number of subjects or patients being monitored where the sensor has slid off a little bit or it wasn't properly applied to begin with.

Sensors can be applied too tightly. Particularly wrap style sensors, it can be wrapped very snugly, use additional tape. Sandy referred to earlier. The consequences to that is it diminishes the pulse amplitude. It does affect and can affect accuracy, generally not greatly, but it does have some impact. The comment about burns usually being some form of necrosis, a sensor that's applied too tightly, especially with the lumpy emitter and detector components in there can occlude blood flow to that local tissue and over many hour period of time the tissue can be damaged.

Venous pulsation can occur in a transmission probe as well as reflectance. If the sensor is placed too tightly, that can capture some venous blood at the distal end of the fingertip, for example and cause some venous pulsations.

And sensors can be improperly located. I think in the handout that you received, I'm not going to show this slide, I did a study of what happens if you take a digit probe that's designed for transmission across the fingertip and place it in transmission across an earlobe, you get a very different answer and it's not accurate. But sometimes when patients are peripherally shut down, clinicians will move the sensor to some site where you can get a pulse. If we've not validated it for that location and have not labeled it that for location, I can't, as a manufacturer, tell you that it will work properly.

If we move on and say what are -- well, how do we mitigate all of those vulnerabilities? Obviously, we'd like to make our products foolproof. To the extent we can, we will. We also want to make them easy to use. And so to the extent that we can balance those two, and I don't have answers for foolproofness, we have to mitigate that with labeling and education. And Julian raised the issue of how do we better educate the clinicians that re using the product. I'm on the same bandwagon as Julian. Whatever we can do to educate the users is going to go to great lengths. It's going to be quite beneficial. So we deal with these issues, these limitations, vulnerabilities with labeling and education.

Well, there's almost a one to one correspondence although the details may vary just slightly when we talk about reflectance technology. We get shunting if the emitter and detector aren't cleanly placed against the tissue. The sensor could be applied too tightly, particularly if you're using a band to hold it in place with similar consequences that you get with transmission. You can properly locate a sensor, as Julian pointed out, over larger vessels and that's not good, placing a reflectance probe on the temple is not an appropriate place to put it.

And we can get venous pulsations, I think it was discussed in Trendelenburg positions where the head is below the heard. We've got a continuous fluid column between the right side of the heart and the pulse oximeter site. Venous pulsations will cause the pulse oximeter to read low.

We deal with all of these things, once again, with labeling and education. For example, the Trendelenburg effect or the Trendelenburg position is contraindicated with some of our products at least.

So the bottom line here is that the light is transmitting through the tissues, not necessarily north and south, but perhaps it's east and west and there's an equivalence between reflectance and transmittance. It works on adults and kids. All of these vulnerabilities are present in both. It's not foolproof. Both are similar, however and any differences, I didn't really touch upon this, any differences that we might be concerned of between adult and neonatal tissues might not affect pulse oximetry is true whether we're shining a light through a tissue bed vertically or horizontally. All of those tissue differences are present.

So is there value in reflectance sensors? Why should we be interested in thinking about this problem? And I'd like to equate pulse oximetry to real estate. It's all about location, location, location.

What we're looking for is a site to place the probe where we've got a strong pulse. Ideally, we're also measuring a core circulation and what reflectance probes give us is additional access to some of the sites. One of the things we've learned is that the forehead, for example, is a site absent vasoconstrictor response, so that the pulsable signal levels are fairly consistent, even if you're peripherally shut down. And circulation lag time. I'm going to address these two issues in a second.

And in the neonates, we also have an additional site for pre-ductile circulation on the right hand. It can be placed on a flat body part. We have access to non-hand and non-feet sites. These are things that we hear from clinicians who are looking for help in their sensor designs.

Let's take a look at what the vasoconstrictor response is all about. These are signal amplitudes. It's the pulse amplitude of the wiggle size that the oximeter is tracking and healthy adult volunteers, the red bars represent what the pulse size is from a finger and ear and a forehead sensor. And the finger is clearly the strongest pulse.

The gray region in the bottom half is what I call weak pulses. And this is the size signal that the 1990 era technology struggled with, signals weaker than this were hard to monitor, although some could do it, they didn't do it gracefully.

When we take these same subjects and cool them off so that they're -- they've been exposed to a cold room environment for 45 minutes, they peripherally basoconstrict. And the pulse amplitude that was on the order of 5 percent drops down to about a half a percent. Ear sensors, we found, dropped from a little under 1 percent to about a half of that. But the forehead lacks the vasoconstrictor response and its pulse amplitude was unaffected.

Another advantage of monitoring a core circulation site and this is an example on the head. This is the circulation lag time slide that I've taken from the MacLeod group at Duke, where they've desaturated a healthy volunteer that's been exposed to a cold saline drip and a cooling blanket to induce hypothermia and at time zero they induced a hypoxic episode by lowering the FiO2 to 11 percent. They had monitors placed on the forehead, the ear, and the fingers, as well as a radial arterial blood catheter for sampling. The circles, or these little dots here are the blood draws.

And we see that in about 20 seconds in their protocol, the ear and the forehead drop down below their test threshold of 95 percent sat., but it's not until 3 minutes later that the fingers finally begin to fall and on the resaturation, the same 3-minute lag time tends to occur.

One of the things I want to point out in here is that two fingers were a good 15 or 20 seconds different from one another. So one of the things that Paul Batchelder was referring to is the importance of having a stable bed so that the location of the catheter and the saturation at the sensor site, whether it's an index finger or middle finger or foot or a hand, they all need to be at the same level for us to compare those two and draw an accuracy.

They may all be completely accurate. In fact, each of these oximeters was completely accurate, I believe, for the saturation in the tissue bed that it was measuring, but the saturation wasn't the same at the various tissue sites.

When we do laboratory controlled studies, we generally don't do accuracy studies in this environment of a peripheral vasoconstriction.

DR. LISBON: Dr. Mannheimer, can I ask you to finish up in about two minutes or so?

DR. MANNHEIMER: Yes. So lastly, this is my last slide, my recommendation in 510(k) submissions for neonatal reflectance pulse oximetry is to treat it as any other pulse oximetry, independent of sensor geometry. We should validate and under controlled laboratory conditions from 70 to 100 percent on healthy adults, using a site that's representative of whether the site we would recommend for use on the neonate. For specifying a reflectance probe for the chest of a neonate, I'd recommend we test it on the chest of an adult, on the back, we test it on the back and the forehead, the forehead. If one wanted to design a sensor to go over the temple, I wouldn't recommend it, but that's where it should be tested.

We can verify the appropriateness for the neonate with form, fit and function testing to demonstrate the physical safety and appropriateness of the probe. An adhesive, for example, might be too aggressive for a neonate. This is something that we can verify with testing. I'd rather verify the contra to that, the appropriateness, rather than the inappropriateness.

And lastly, we would recommend that convenience samples taken on neonates, spanning a range around 90 percent sat where neonates do exist in the clinical environment, to confirm the system accuracy. I want to make this point fairly clear, because I don't think it was made very well by Paul previously. He didn't amplify on this. What we'd recommend doing is taking a predicate device and simultaneously testing it and the new device to show that whatever new product we're creating or introducing, is performing at a level that's comparable to what has been used in clinical practice for many years and has been shown to be safe and effective.

So it may not be possible to get an accuracy spec with a laboratory equivalent, but we can show that what the new device is performing at is comparable to what current practice is using.

Thank you.

DR. LISBON: Thank you very much.

DR. MANNHEIMER: Do we have time for some questions?

DR. LISBON: You have time for one question.

Does anybody have a question?

DR. COTÉ: Why wouldn't it be possible to do studies, as you said, comparing with currently available oximeters, but put multiple different sites at the same time. I mean the baby would look like you wallpapered him with sensors, but this might give you a better feeling of the equivalency of putting something on the interior chest, the back, on the shoulder, on the upper arm, and then you have real time data of multiple devices at once.

DR. MANNHEIMER: In the clinic or in the laboratory?

DR. COTÉ: In the clinic.

DR. MANNHEIMER: I think that's not unreasonable. It fits within the practice of caring for that patient. I think that's open to discussion.

DR. COTÉ: It's obviously not going to be within the care of the patient, but it's going to be

-- there are parents that will consent to do research on their children if they understand what it's for.

DR. MANNHEIMER: But we can't make them hypoxic.

DR. COTÉ: No, no. I didn't say that. I said I'm agreeing there are clinical situations that develop. How well do these devices track compared to each other and compared to different locations on the body.

DR. MANNHEIMER: If the saturation is all around 98 percent, then we're only testing one little region and that tends to be where inaccuracy is least -- where the performance is.

DR. COTÉ: Correct, but you'll be able to generate some information as opposed to no information.

DR. MANNHEIMER: That's true.

DR. COTÉ: And these children typically desaturate every time they're suctioned, so you've got a very predictable event that can be examined.

DR. MANNHEIMER: Do they have an A line placed?

DR. COTÉ: Well, you wouldn't necessarily. You've already said in the adult patients that you have equivalence in terms of performance, but now what you could do is track the old performance with the new performance, time response and you're right, the gold standard is to get a blood gas and that's not practical except in --

DR. MANNHEIMER: I think that's worth considering and should be part of your discussion amongst yourselves.

DR. LISBON: Thank you very much, Dr. Mannheimer. We have a presentation by the Nonin Medical Group. You gentlemen each identify yourself, that would be great.

MR. ISAACSON: Good day. I'm Philip Isaacson. I'm Managing Director and the founder of Nonin Medical.

MR. PEDERSON: I'm Brodie Pederson. I am a Senior Design Quality Engineer. I work in the quality regulatory department assuring the safety and effectiveness of our devices.

MR. ISAACSON: And I'm going to start out talking about pulse oximeter applications and specifically some of the sports applications. Brodie will finish up talking about, again, consumer application and aviation usage.

And he'll discuss a little more of the difference that we consider for consumer, over the counter, and prescription. Let's see, myself, I have been involved in design of pulse oximeters for over 20 years.

And Nonin has been selling pulse oximeters since 1986. And, at one time, a few years back I gave a talk about the ubiquitous pulse oximeter.

It's showing up in many places everywhere. And there's going to be a number of applications that are not discussed at this meeting, a number of applications which are outside of the scope of this meeting.

And I'm just going to talk about a couple of the applications we have which are consumer as opposed to medical applications. But, I think the lessons were learned there, are applicable, and we're thinking about in the discussion of over the counter.

And also, in terms of policies devices we have and medications, we've had the patients call up to buy a pulse oximeter. We told them, go back to your doctor, get a prescription, fax us the prescription, we'll sell you the pulse oximeter.

And this has been typically patients with CHF, COPD, and asthma. And I've had some anecdotal stories about the home use of these oximeters, that the patients have requested prescription from their doctor.

It has significantly changed their quality of life. And, in one case, the patient wouldn't be alive if she hadn't gotten a personal pulse oximeter.

Okay. Nonin has offered pulse oximeter for sports use. And the sports use people that are most interested in that right now are people working at high altitudes and such.

Again, it's not over the counter in that we make no medical claims. That is, we make, quote, no clinically significant diagnostic indications for use.

And, I can show you, the device that we have -- I've got one in my hand here. The device that we offer for the sports staff and the flight staff, which Brodie will talk about, is in fact identical to our medical device, although it need not be.

It differs in the packaging and labeling. And some of the key applications -- let's see -- is a mountain climber. And, basically, about every expedition to Mount Everest will have a number of these along with them.

Again, there's a case where a number of years back Public Broadcasting had a NOVA episode on Mount Everest. It showed the guy trudging up the trail.

He pulls out one of these out of his pocket. Let's see what you're doing. He said, well, you're at 70 percent saturation. Well, let's rest a little bit and see what you're at.

Anybody in this room that'd be at 70 percent saturation, you'd be rushing them to the emergency room right now. But, the application we have, again, for hikers, again, not just the normal hikers, but people hiking in the mountains at altitude.

They use it quite often to check what their saturation is and adjust their breathing patterns. So, we're looking at the normal desaturations with altitude and exertion, and breathing.

One of my formal partners went hiking with his son. He brought the pulse oximeter along. He saw he was desaturating, breathed a little more heavily.

You know, you're doing great. The son wasn't doing that and was exhausted very quickly. So, other places, of course, in the health club. I go to work out at the health club all the time here now.

I'm not exercising. I've been doing mostly strength exercises where I don't go to the point of exertion. And typically what happens under those circumstances is really for a handy way of measuring the heart rate very quickly and easily.

And my oxygen saturation as I exercise rises a little bit, which says I'm not an elite athlete. Typically an elite athlete, as they push themselves to the level of exhaustion, will slowly desaturate, while somebody like me that's not in such great shape rises until it hits the limit and then drops rapidly.

But it's, you know, useful, again, in health clubs, health institutions, people exercising. They can judge when they are really hitting the limit of their body's ability to provide oxygen -- or the heart and lungs to provide oxygen to the body.

And this one was designed for use -- there are many applications, high altitudes, light, dark, cold, in fact, many of the cases the hands are very cold but it's being used.

But the thing that's quite often done is really trying to avoid the systemic anaerobic condition. In other words, it doesn't tell whether muscles are getting anaerobic.

But it tells you whether your body as a whole is getting anaerobic. Again, usefulness, particularly at high altitude, is to recognize when the hiker is getting anaerobic to slow down, to prevent getting the altitude sickness.

And so, again, the -- for instance, going to Mount Everest. At the base camps they will be checked. Typically they'll check them. People will adjust their breathing, wait to see that they can adapt to altitude before they go on to the higher altitudes.

And the biofeedback usage SpO₂ information to just see the effects of changing breathing patterns. Again, at sea level here there's not a difference.

You get up into mountains, you see a difference. I see a difference when I'm flying in an airplane. If I'm just sitting there not doing anything, particularly the high altitude long flights, I'll saturate into the upper 80's.

Take three quick breaths and I get back up to 97 percent normally. So, is that helping me avoid some of the ill effects of travel? I don't know.

But, again, it's a non-medical type of application to see what the effect of breathing or not breathing in an airplane is and breathing. That affects me as a passenger in a commercial jet.

Brodie's going to talk more about the usage of the pulse oximeter by pilots where it's far more critical. All right.

MR. PEDERSON: Much of the success of these consumer products is based upon the many guidance and articles written by physicians and other lay members of these consumer markets that have found usability of these devices for their application, some of which the articles I'll be noting in my presentation.

The FlightStat is used primarily by pilots to detect hypoxia when flying without oxygen. When flying at high altitudes can impair the ability to reason, think, see, talk.

And these cognitive functions are very important when you're flying a plane and to protect yourself and your passenger's safety. Depending on the age and overall health conditions, hypoxia can occur as low as 5,000 feet.

This is documented in several articles. And, particular importance has been placed on night vision by organizations such as the FAA. I'll show you another reference later on.

Why do these consumers need a pulse oximeter? For both pilots and sports enthusiasts, pulse oximeters can detect potential problems caused by decrease of oxygen in the blood caused by thinner air at higher elevations.

The decrease in the blood, you know, we're all aware, impairs judgment, decreases physical stamina, causes dizziness and nausea, and other symptoms that could cause potential harm.

I'll turn to some published information that I've pulled out, some articles that I have quoted. The Federal Aviation Regulation, which is more than 20 years old, and is showing its age, requires oxygen use above 14,000 feet for any time the pilot exceeds that altitude, and continuously if you extend above 12,500 for more than 30 minutes.

This is to protect the pilot, of course, and the integrity of the aircraft. The FAA also recommends -- they've revised and recommended oxygen use in the last ten years for any time you're flying above 10,000 feet during the day and 6,000 feet at night.

And, again, this is to address the concern of loss of night vision with hypoxia. Pulse oximeters may be used in these situations to help pilots or sporting enthusiasts identify earlier that they need oxygen to prevent hypoxia from setting in.

Even trained individuals may not be aware of the effects of the hypoxia on their own. And, you know, the guidelines just show that there is an increase in understanding of the effects of hypoxia on individuals in the broader public, and not just in the medical community.

One article said you can ask pilots about oxygen usage and they'll tell you what the regulation states, which doesn't really say a whole lot.

It tells them that they should be using oxygen, but doesn't tell what flow rates, how much, how often, how to adjust breathing patterns. So, you know, most of them don't know.

And most of those pilots, again -- this article credited at the bottom of the page there stated that most people don't recognize the effects of the hypoxia that they're experiencing.

Personal tolerance to altitude can vary from day-to-day. So, one person who may have gone through training and testing in that field of application may not have the same effects on a different day.

Or, if they're having some sort of other condition or lack of sleep, or other factors that might play into the role. One source has stated that in the clinical environment the patient is not considered to be of sound mind unless their SpO₂ is above 90 percent.

Given that, you must decide whether you want less for yourself when you're flying a plane. So, it just kind of underscores the importance of knowing where your body is at at various elevations, especially when you're controlling an aircraft.

To assure proper oxygen flow, a pulse oximeter can provide an almost instantaneous read-out of one's blood oxygen level and heart rate. Just to cite the references, you know, this quote, hypoxia equals stupidity.

Once you're there you're too stupid to save yourself and problems can occur. And, there's lots of articles and literature which on the following -- at the end of my presentation I've listed out several articles and literature sources that provide information further to these points.

So, this just goes to show there are other uses for pulse oximeters other than in the medical community. And they are being used on a daily basis by many individuals to protect their life and livelihood as well as their -- the passengers on their aircraft.

There's, you know, there's the threat even to the passengers at these higher elevations as well, not just the pilot of the aircraft. And so, a lot of these pilots that are now more aware of the effects of hypoxia have been increasing, the use of oxygen among their passengers as well.

To summarize pulse oximeter application various market controls, today we have prescription devices and we have consumer devices. The prescription provides safe, effective devices with controlled distribution when doctors need to get data and information about their patient's percent oxygen level.

In the consumer market, there are no regulations. The FAA does not regulate the use or the production or manufacture of a device, or any requirements on the device's performance for use in the aircraft because the device can be taken out of the aircraft.

It's not affixed to the aircraft, and therefore the testing and regulations don't apply. So, some of those consumer products, the SportStat, FlightStat, are very useful for altitude and exercise in those environments.

But there are other products out there that are less rigorously designed than ones that we present. And that can yield somewhat of a risk. The over the counter market does still control safety and accuracy and effectiveness -- are all regulated.

It just allows more people to have access to the technology and provides added reliable health information to the public and possibly to physicians that, you know, as Dr. Goldman had suggested, may provide earlier diagnosis or assistance in diagnosis of certain conditions.

So, you know, again, here are the -- some additional references.

CHAIRPERSON LISBON: Thank you, Mr. Peterson. We have time for one or two questions.

DR. BIRNBACH: I'd like to start by way of confession, middle age is a little over rated. And I have a hard time reading the published references without a microscope.

But it doesn't appear that there's any data there about the use, validity, accuracy. Have these things ever been tested and compared with either the quote, unquote gold standard or with clinical conditions.

MR. PEDERSON: I can speak to our devices. Since our FlightStat and SportStat are technically identical to our medical device, it has been tested and is reliable and accurate.

And we have done that testing according to the standards and guidances in the regulatory arena.

DR. BIRNBACH: So, the device has been tested. What about the users to see whether users actually use it appropriately or know what they're doing, or how they're reading these things?

MR. PEDERSON: Do you have anything on that?

MR. ISAACSON: There are papers published about the need for it and the -- there have definitely been papers published about the need for it, use for it.

They are not rigorous scientific studies. They are more the anecdotal type of published papers.

MEMBER COTE: On your website you say it's operating altitude up to 30,000 feet. Have you tested it at that altitude to see if in fact it's accurate at that altitude?

MR. PEDERSON: We've tested the effect of altitude on the device and --

MEMBER COTE: On the device in what way though?

MR. PEDERSON: On its operation at altitude with simulators.

MEMBER COTE: Oh, with simulators. Okay.

MR. PEDERSON: Yes.

MEMBER COTE: So, you put it in a pressurized compartment and --

MR. PEDERSON: Yes.

MEMBER COTE: -- and a patient or a volunteer in a pressurized compartment?

MR. PEDERSON: The simulator on a device at altitude to verify its performance.

MEMBER COTE: I guess we're not communicating. I'm trying to find out, does it work on a human being accurately at 30,000 feet? Or, does the accuracy of the device drop of --

MR. PEDERSON: No, the device is not -- the electronics and the operation of the device are not affected by the altitude.

CHAIRPERSON LISBON: All right. I'd like to keep moving. We'll have time, I believe, as catch-up for questions right before lunch. Thank you very much.

MR. PEDERSON: Thank you.

CHAIRPERSON LISBON: We now go to the open public hearing. This is the first of two open public hearing sessions for this meeting. There will be a second one that follows the panel discussion this afternoon.

At these times the public attendees are given opportunity to address the panel to present data or views relevant to the panel's activities. We've been given advance notice of one person who wishes to address the panel.

That's Dr. Dale Gerstmann from Utah Valley Regional Medical Center. Are there other people that wish to address the panel that are here? So, just raise your hand. All right.

I just want to remind Dr. Gerstmann that you have ten minutes for your presentation and about five minutes for questions. Speak into the microphone because the transcriptionist is dependent upon hearing it.

And also, we're requesting that all persons making statements during these open public hearings disclose if you have any financial interest with the sponsor or products under consideration.

And, before making your presentation to the panel, in addition to stating your name and affiliation, please state the name of your financial interest in the product under consideration, who is paying for your attendance in this meeting.

DR. GERSTMANN: My name is Dale Gerstmann. I'm a neonatologist from Utah Valley Regional Medical Center. I'm here on my own. I will state that the hospital I work for has a contract to evaluate experiment sensors for Nellcor/Tyco Healthcare.

And I don't have any financial interest in any companies.

CHAIRPERSON LISBON: Thank you.

DR. GERSTMANN: What I would like to do is comment on functional performance, which relates to your discussion on neonatal validation. Next slide, thank you.

And, in particular I would like to make some comments about what I have been able to see about bios, precision, and resolution on neonatal oximetry.

Next. Just a short comment to say that this is a level three -- information from a level three nursery, which is a very typical kind of nursery, 40 beds, 500 emissions per year.

We do high frequency jet ventilation, inhale nitric oxide. But we do not take care of babies with complex congenital heart disease who have complex pediatric surgical problems or do ECMO.

The data covers a period of 2002 to 2005. Next. This information covers 114 patients. The birth weight and gestational age distributions reflect very similarly to what our admission characteristics are.

About a quarter of the infants are under 1,500 grams or under 32 weeks. Next. All of these data have been collected under two -- actually there's three RMB approved protocols.

But all of the SpO₂, AO₂ pairs where the AO₂ is functional saturation determined co-oximetry is done under informed consent. There are about 1,250 pairs of data that go into the following slides.

First let me talk about bias. Bias is the difference between the pulse oximeter and the functional arterial saturation. And here it's plotted in Bland-Altman plot for data from an older generation device, Ohmeda Biox, with a disposal sensor.

On the Y axis is the bias or SPO₂ minus SAO₂. On the X axis is the average. And, as you can see, the slope of this information, or the data plotted, is quite negative.

Even though the bias is a plus 2.2, the precision is here depicted as 95 percent confidence limit with a precision of six. And the A_rms is 3.7, which, under your current standards, would be acceptable.

However, this kind of bias would not be clinically acceptable. As you can see, in a saturation of 85, the pulse oximeter is actually reading six or seven higher typically.

Thank you, next. This is similar data using a non-disposable generic probe for the Ohmeda Biox. The negativity of the slope is not as severe. The precision is slightly better in the A_rms under three.

Next. This is current information, 250 samples using the Masimo Radical with the neonatal sensor. And, as you can see, this device and sensor combination also has a very negative slope.

Although the bias, which is the average bias for all of that data is only slightly positive, the precision is 4.4 and the A_rmsis 2.3. This negative slope, however, does continue to indicate that as the saturations drop, the pulse oximeter reads higher than actual.

Next. And, just to point out that for some sensor monitor combinations the bias can actually be appropriately zero, and that would be ideal.

That is a device sensor combination that does not have a bias. This is, you know, for SoftCare sensor information with a precision of 5.2 and a A_rms of 2.7. Next.

Similarly, this is the infant MAX-I sensor from Nellcor with 100 samples. It also has a -- essentially zero bias and an A_rms of 2.3. Next. And the MAX-N, which is the neonatal MAX sensor for Nellcor shows a similar negative slope.

The bias is slightly for positive. And the A_rms is 2.7. Here is a table summarizing the previous graphics indicating the sample size and the bias.

And, as you can see, for the most part, there are only three devices where the bias is close to zero. Otherwise it's positive, indicating a negative slope to the Bland-Altman curve.

What I would like to do is just illustrate all of that information in this graphic and make a statement saying that, as you can see fairly easily, the average bias by saturation is not uniform across the range for which we typically see saturations in sick neonates that are on ventilators and whose oxygen is being controlled.

And, as the saturation drops for almost all of these device-sensor combinations, the bias increases. It would be much more clinically acceptable to have a device that has flat performance, even in this narrow range of normal saturations.

We don't have any low saturation data because we don't keep babies with saturations under 80. We attend to them immediately. And the other point is that blood gasses are being done more and more infrequently in neonates.

There are some nurseries who don't draw any blood gasses any more on ventilated neonates who have uncomplicated courses. So, convenience data is going to be harder and harder to obtain.

Next. My preference is to actually plot the regression slopes on a graphic like this that shows the 95 percent confidence intervals and very clearly demonstrates the two devices that have regression slopes that include zero in their 95 percent confidence limits.

And these two devices would obviously be preferable -- the device-sensor combinations to the other ones, which have negative slopes. Next. Now I'll comment on precision.

This is a similar table summarizing the data off of the previous graphics. And, as you can see, the A_rms values are all of them under which you would currently accept as four percent or less.

But, please note that, except for the very old combination of Ohmeda Biox and disposable sensor, all of them are under three. And most of them, some of them are very close to two.

So, I would concur that the addition of the one percent fudge factor is absolutely unnecessary. And, in fact, in functional performance in NICU, the A_rms values are well within the context of less than three percent.

My bias, however, is not to use the A_rms, but to use the precision measured as 1.96 standard deviation, which is in fact the 95 confidence interval.

Next slide. To illustrate that the A_rms as well does not have stable values over the range of 80 to 100 saturation, this graphic illustrates that point.

Note that the precision worsens as the saturation goes up. And, for data that I don't have on this graph from older data, the precision also worsens as the saturation goes from 97, 98, 99, to 100.

We don't keep babies at high saturations. And so, I don't have any blood gas values with saturation at 99 or 100. And that's why they're not there.

The best range of A_rms is obviously for most devices somewhere between 92 and 95 or 97, somewhere in that narrow range. Next. To illustrate the point on the 95 percent confidence intervals, this is simply a graphic of SaO₂, versus SpO₂.

And, from the data that I'm just showing you, I've extracted the 95 percent confidence intervals for the Nellcor plus SoftCare over the range from about 86 to about 99 percent.

The line in the center -- I guess that is not a pointer -- is the 50^th percentile. And the upper and lower lines are the 95 percent intervals. And I would submit that clinically it is very difficult to expect to manage a baby that in this example the arterial saturation would be 90, but the pulse oximeter could be reading anywhere between 88 and 96.

And that's very typical performance for all of the neonatal -- the current neonatal oximeters. Next. This is the Masimo device. And, on this graphic you can see the effect of the bias as the 50 percentile line is really laid over very horizontal.

It should be following that green line. The same situation with the 95 percent confidence interval exists across a wide range of saturation where it is not uniform. Next.

And finally, I'd like to make a comment about resolution. Resolution is an instrumentation term in my thinking. And it relates to what you see on the display.

If you look at a pulse oximeter, it says 90, it says 91, it says 89. The precision of the monitor itself or the resolution is plus or minus one.

It doesn't show you anything more than that. Next slide. And, if you would clock that against what the 95 percent confidence intervals are on the measurement, there's this wide discrepancy between what's actually shown on the monitor and what that data really means.

I can tell you that the clinician looks at the monitor and sees it go 90, 91, 89, and thinks that's how accurate this device is. That is what I think is happening to this patient.

When, in fact, if you look at the data, the pulse oximeter reading could actually be plus or minus four, plus or minus five. And the clinician is not seeing that on the monitor. Next.

What I would like to see in terms of device response is something much more close to this where both the precision and the bias is acceptable and is very close to what the resolution of the monitor actually shows.

That would be a clinically acceptable device. Next. So, let me summarize. The bias as a summary statistic actually does not reflect a dependence on SaO₂.

But, for neonatal bias you would want it to be zero. And I presume that would be an achievable result for adults as well. The only way to actually look at this and to analyze it in a comprehensive way is to look at the slope of the Bland-Altman curves.

In terms of precision, a summary statistic there in terms of A_rms also does not reflect the SaO₂ dependence of the precision as it changes across saturations.

The 95 percent confidence interval are really fairly large for any clinical specificity. And it would be much better if the A_rms was close to one as opposed to less than three or four.

And finally, the final comment on resolution it would be ideal to have the 95 percent confidence interval or the accuracy of the device reflected in the device resolution as it's displayed to the clinician. Thank you.

CHAIRPERSON LISBON: Thank you very much. Are there questions for Dr. Gerstmann?

DR. BIRNBACH: Is any of this data published?

DR. GERSTMANN: A smaller data set has been published in the Journal of Prenatalogy two years ago. And I'll be working on a manuscript here. This is brand new information.

CHAIRPERSON LISBON: Are there other questions? Jackie?

DR. LEUNG: I have a question, kind of taking the physical data down to the real life situation. I mean, I think one of the myths about pulse oximetry is that we, you know, transfer the wavelength differences into numerical value, they're continuous, giving one an impression that it's a continuous scale.

In fact, clinically, the way we use pulse oximeter is either it's normal or abnormal. So, really beyond a certain level, we don't really know what it means.

And your data adjust that, because you don't have a lot of points really below 80, as you said, because we don't keep people in that situation for a long time.

CHAIRPERSON LISBON: Charlie?

MEMBER COTE: I think you are to be congratulated for trying to get some real information that the manufacturers keep telling us they can't get. Thank you.

DR. GERSTMANN: Thank you.

CHAIRPERSON LISBON: Avery?

DR. TUNG: It is obviously disturbing to see positive bias at low sets. It opens the door for false negatives. Did you see any in your study? Did you have any adverse events related to false negatives?

If not, how did you avoid, you know, mis-interpreting a set of 95 -- really setting in the mid 80's?

DR. GERSTMANN: I have to acknowledge that I had no part in the clinical care of these babies. Okay, so any decisions -- in fact, there were -- I can't answer that question, whether or not there were decisions made off of these values that might have made an effect or a difference to the child.

DR. TUNG: Is it possible to debrief nurses and understand how they adjusted, as they must have done in three years, to, you know, what seems like very large biases at low sets?

DR. GERSTMANN: They have been educated to know that a saturation of 89 may actually be an arterial saturation much lower than that. So, the caution is, look at the patient or do some physical evaluation to try and ascertain whether that saturation is meaningful or not.

CHAIRPERSON LISBON: Dr. Otulana?

DR. OTULANA: Going back to the transmitters, the reflectance issue that was mentioned earlier, was any particular type of probe --

DR. GERSTMANN: These are all transmission.

DR. OTULANA: They are all transmission? All right.

CHAIRPERSON LISBON: All right. We have about ten minutes before lunch. And what I'd like is that if there any -- some of the speakers were cut off.

If there are questions for any of those previous industry speakers, we could entertain those at present. So, just ask who you'd like to ask a question to.

And then we need that speaker to come up to a microphone so that we can record that.

DR. MUELLER: I have a question of Pederson and Isaacson. Let's say that you have a group of six U.S. citizens who are going to tackle K2 for the first time.

They're going up there for a four week visit. Would you recommend they purchase a number of monitors equal to the number of people, or twice as many, or three times as many?

MR. PEDERSON: I wouldn't -- you know, not knowing how they organize their trip, I don't know what I would recommend. There are always --

DR. MUELLER: What I'm trying to get at -- excuse me for interrupting. What I'm trying to get at is what's the reliability of this under the conditions, let's say, of the climb to a high altitude where the device might get wet, it might get knocked about and so forth?

Do they have to take several spares? Is there some --

MR. PEDERSON: From that point of view, you know, one is sufficient. I mean, these things very rarely fail. It would be more operation, less of a chance you're going to lose one, something like that, rather than is it going to fail?

But, if they want to be 100 percent certain, they better have two of them.

DR. MUELLER: Do you have a program so they can return it to be fixed? And, if so, what's the return rate on these for repair per year?

MR. PEDERSON: Let's see, our medical devices, I think we've got a three year warranty. And what is it per year?

MR. ISAACSON: That would require a bit of research and consulting with our return department to get accurate numbers. I'm not certain what the return rate is, especially divided between SportStat, FlightStat, and the medical device.

DR. MUELLER: Sure. Okay. The man is up on the mountain. Does he have anything? He looks at his machine and it tells him that, you know, there's a number there that's too good to be true.

What do you have in terms of a, quote, pocket restandardization of your device so that it could be reset or reprogrammed or re something in order to become reliable and predictive again?

MR. PEDERSON: None of our pulse oximeters have any means to recalibrate. They are calibrated by design. And, our particular design -- again, I'll get into some of the technical details in the way we implement it.

But, we run both the red and the infrared channel signals through the same signal chain. So, we're not going to have to worry about drift of one channel relative to the other.

They both run identical. So, most -- except for, we've got some analog front end, everything else is done digitally, and that doesn't drift.

But, I say -- from the very beginning, there never has been a means to adjust any pulse oximeter that Nonin has ever made.

CHAIRPERSON LISBON: When these devices fail, in which direction do they typically fail? Or is it just completely random?

MR. PEDERSON: The typical failure is something like, well, it doesn't automatically turn off when it should, as a common failure.

CHAIRPERSON LISBON: But they don't go falsely high or falsely low when they break?

MR. PEDERSON: I'm not aware of that having happened. I'm going to have to check all of our records.

CHAIRPERSON LISBON: Charlie?

MEMBER COTE: These are calibrated down to 70 percent plus or minus two. I'm not a climber, maybe somebody here has experience. But, the saturations at 4,500 meters are 70 plus or minus 38 percent.

So, how can you -- if the device is not accurate under 70 percent, how can you use this to make a decision at the very point where the accuracy falls off dramatically?

MR. PEDERSON: Okay. Well, our device is, you know, we have certainly tested our devices down to 50 percent.

MEMBER COTE: Only down to 70 according to your --

MR. PEDERSON: We don't --

MEMBER COTE: Your package insert says 70.

MR. PEDERSON: We haven't been specifying it below 70. But we do have data down to 50 percent. It's rather sparse data. But, generally, again, I have to look at the -- I don't have the numbers with me.

You know, what sort of data we can support, but since the data we have is -- we have tested then down to 50 percent.

MEMBER COTE: I guess, how is the climber who is not a necessarily medically knowledgeable person to make a decision as to what's dangerous and what isn't?

And, at least this one paper that I found, suggested that the only predictor of high altitude sickness was desaturations while they're asleep. If they're asleep they're not using the device.

It has no prediction while they were awake. So, I guess I'm kind of missing something here. Part of it I'm sure is my lack of knowledge of high altitude climbing.

MR. ISAACSON: A lot of the usage in this area relies upon the individual's experience level and conditioning and experience in that environment. Typically these expeditions, they're going to these extreme conditions, spend many, many months training and working and becoming knowledgeable about the tools and usages that they're going to be using during the expedition to the highest of elevations.

And so, they're going to gain a knowledge and understanding of their own performance and comfort levels as far as how they feel and other things and can use the oximeter as an adjunct to that other information.

MEMBER COTE: So, they're using themselves as their own control? Is that what you're saying?

MR. ISAACSON: Essentially. And that's part of the, you know, the conditions of use of the device. We make no claims medically on the SportStat device as to how they should use it.

That's up to them and their experience level and their understanding of the technology.

DR. TUNG: Dr. Cote mentioned the need to actually validate on a human at high altitude. I've got to agree with him. The primary physiologic response to altitude hypoxemia is to hyperventilate.

People can push the ph's, their blood ph up past 7.6, almost to 7.7. I don't know if validation of pulse oximetry at that kind of Ph is, you know, the world might be different there.

And, you know, there may be a need to validate that.

MR. PEDERSON: We have not validated at such. To my knowledge, the Ph does not affect the color of the hemoglobin. And that's what we're looking at.

So, I would have no expectation that there should be any difference. But we have no data to prove that.

CHAIRPERSON LISBON: Carolyn?

MS. PETERSON: I had two questions for Mr. Pederson and also one for Dr. Goldman with regard to the statement that the pulse oximeters were used by many individuals to protect their live and livelihood.

I'm wondering if we have any outcomes data that match readings with the decisions that individuals actually made based on those ratings, either for aviation or sport uses?

MR. ISAACSON: I guess my statements were based off the articles that I read and cited in my presentation as the articles that provided information anecdotally.

The pilots who have since started using pulse oximeters have become more judicious about their oxygen usage in that environment. And they made statements in their papers regarding that statement that I made.

MS. PETERSON: Okay. But there's no sort of recorded instances, readings --

MR. ISAACSON: Not that we have, no.

MS. PETERSON: -- decisions. Okay. And my second question had to do with the type of training that you provide or require your purchasers to undergo so that they can use the product properly.

MR. ISAACSON: In the consumer market there's no requirement for training.

MS. PETERSON: But, do you provide any or do you require your purchasers --

MR. ISAACSON: We provide instructions in the manual as to how the device is operated. And then it's up to them to do research and investigation to figure out how to interpret those results.

MS. PETERSON: And, do you clearly state that they need to do research to understand how to interpret the results your device --

MR. ISAACSON: I'd have to review the indications for use. I can't recall off the top of my head, sorry.

MS. PETERSON: All right. Thank you. Dr. Goldman? In your presentation you mentioned that when there is an error it tends to read low in saturation rather than high, or to very quite a bit.

Is there a general trend that is accepted among users so we can say it's five percent low, it's ten percent? Or is it variable?

DR. GOLDMAN: I think we know that, historically, with pulse oximetry certainly the older conventional pulse oximetry designs that some of use have come to hate, despite the fact that they have saved lives and made practice much easier, we know from clinical practice and from much data that typically they read low in the presence of low profusion and motion conditions, for example.

That's -- and we have a deeper understanding now of why that happens and probably partially because of that that's been corrected in some of the newer designs.

So, it's pretty clear that a shivering patient is going to read low, they won't read high, and that because of the physiological effect. That's that part.

So, if an instrument has been -- an improved instrument is available that's been designed to deal with that situation, it's probably going to read accurately.

If it isn't designed to do that, it may read low. The likelihood that it will read high is much less, based upon what we know about the physiology.

It seems that -- and there isn't anything published on this aspect of it. But, it seems that, based on when you try to troubleshoot high readings with pulse oximeters, which are quite uncommon I think for most of us to observe, it seems to be due to an equipment problem more frequently than not.

It's difficult to figure out physiologically why an instrument would read high. There isn't another substance or absorber really that's pulsitile that we know of that produces a saturation ratio or color ration, absorption ratio, that would drive the number high.

We just don't know of things like that. So, it would have to be something else in the optical path. Or it would have to be a fault with the device.

In my experience, I've only seen this a few times, it's usually been a cable problem, a physically damaged cable. You can usually change it and you can see an improvement in the value.

But there isn't much data on this. So, generally speaking, I think it's not commonly found to have high values.

MS. PETERSON: But, the question that I'm trying to get at pertains to how a recreational user might interpret the data if we know that -- or if it's generally common knowledge among say climbers that if they're untrue values, they read low, and people perceive, oh, it's always five percent low, or it's ten percent down, so really my saturation is X amount, so I'm okay. That's the concern I'm getting at.

DR. GOLDMAN: Oh, I see.

MS. PETERSON: If users may -- themselves to what they're reading assuming that it's too low.

DR. GOLDMAN: Well, I think we have to take all the data into consideration. And I think we saw some very interesting data today that shows a positive bias in the lower range that might be -- at least in that patient population.

I think that's the challenge, is to identify -- to some degree this cuts to a very important part of the issue which is, just how much can you rely on laboratory testing to indicate the performance in the real world?

And, under ideal conditions we see a set of certain limited -- we have limited information in a set of performance. And, if we don't ever test on the population of attended use, there probably will continue to be surprises like this.

So, I think that's a very reasonable question to ask. And, the challenge is getting this kind of data. I certainly am all for getting data. I'm for getting data in the clinical conditions of use and the population of use.

But, part of what should be discussed and considered is when can you, how necessary is it to get started, should it be required before clearance of a device, or should it be considered afterwards?

Those are all part of the discussion. It's a good point.

CHAIRPERSON LISBON: I think our morning session is done. I'd like to thank all of the participants for a very lively and informative discussion.

Lunch is in the restaurant out towards the front. And I gather there are reserved tables for the panel members. And we will start again precisely at 1:00 p.m. Thank you very much.

(Whereupon, at 12:03 p.m. the above-entitled matter recessed for lunch.)

A-F-T-E-R-N-O-O-N S-E-S-S-I-O-N

1:04 p.m.

CHAIRPERSON LISBON: All right. I would like to call this meeting back to order. We're now going to continue with the agenda. And this part is the panel discussion.

What we're going to do here is that we're going to put each of the questions up and we'll discuss each of them. Dr. Cote has graciously agreed to start the discussion of each of these questions.

And then we'll go around the table and get everybody's opinion. And, hopefully we'll be able to wrap this up on time. So, may we have the first question up, please?

MS. GRAHAM: You may as soon as the projector gets warmed up.

CHAIRPERSON LISBON: All right.

(Pause.)

CHAIRPERSON LISBON: All right. So, the first question -- do you want me to read this into the record?

(No verbal response.)

CHAIRPERSON LISBON: All right. Pulse oximeter sensors may be implemented in either transmittance or reflectance configuration. In both configurations, light is scattered by blood, which has time dependent characteristics and bone and other tissue structures which are not time dependent.

Transmittance sensors are configured in a manner where the emitter outputs light, which travels through tissue, finger, toe, or an ear and is received on the opposite side by the detector.

Reflectance sensors are configured with an emitter and detector in the same plane. Emitted light must reach the detector by reflection off of the surface, which typically results in smaller signal strengths in comparison to transmittance sensors.

And, the FDA would like us to discuss the clinical differences between transmittance and reflector senses and in the discussion comment specifically on any differences in performance between the two sensor types, whether the differences in performance would lead you to recommend different pre-marker evaluation methods and standards and, if so, what these would be and whether the differences in performance would exclude certain indications for use for one type compared with the other and if so, what would those be? Charlie?

MEMBER COTE: Okay. I think we heard very nice presentations this morning indicating that in the laboratory situation the ideal adult volunteer patients, that the two types of oximeter probes have reasonable parallelness in terms of function and accuracy.

But I think where it falls down is the actual clinical application. And this is why I think these devices do require a different kind of evaluation as part of the process, because it's clear that if a surface oximeter is applied to different parts of the head over an artery versus not over an artery over an area that's like in the back of the patient if they're lying on it.

It might be different than if you put it on the abdomen. We need to have very clear data to support where these devices need to be placed and how they need to positioned in order to achieve the same accuracy as the device when it's used in the perfect ideal laboratory situation.

There have been a number of public studies indicating, at least one paper looking at five or six different locations on a baby's head for the transmittance oximetry demonstrating that there are certain areas that are very good in terms of getting a nice pulse plethysmograph.

But, because they're near an artery, it creates an artifactually low reading. And then, other areas where there's no signal at all, and then other areas like the cheek where there's a reasonably good sensor, cartilage between the pulse plethysmograph and the reading of the oximeter.

There's indication in the literature and it was described by several people that if there's a problem with venous congestion, perhaps in, for example, in children with congenital heart disease where they have a high CVP, that there's simultaneous reading of venous and arterial blood, which, again, results in an artifactually low reading.

So, from my standpoint, I think that before we can say the device is equivalent, we have to define more precisely where on the body, at what age patients this device works in the same way as it does in the ideal laboratory situation.

CHAIRPERSON LISBON: Charlie, number C, or letter C, what differences in performance would exclude certain indications? Can you list any of those indications where one type would be preferable to another?

MEMBER COTE: Well, I think it goes both ways. There are situations where you have patients that -- for example, I used to take care of burn children.

And there was no digit that you could apply such a sensor to. But there may have been a surface that we could apply it to. That might be one situation.

The reverse would be a patient who's very edematous or, as someone mentioned -- position, where the surface electrode would not work as well as a transductant silicone.

CHAIRPERSON LISBON: Great. Thank you very much. I can either take volunteers, or I can go around the room. Does anybody want to volunteer to go next?

(No verbal response.)

CHAIRPERSON LISBON: In that case, why don't we start down at the end of the table and just work back towards me?

MEMBER DRASNER: I have very little to add to what he's saying. I think the variability and where the device is applied really makes this little different than previous devices.

And I think, as he stated so clearly, I think we need some data on how these perform specifically in specific locations on specific population.

DR. MUELLER: Yes. I would only concur that the clinical situation, simply the probes, one type won't be appropriate, they won't stick, or they're in the way of the nursing care so that it's a practical matter as to where to put them.

And I think that's best solved by the practitioner at the time of use.

DR. TUNG: I'd agree as well. You know, I feel myself like clinicians aren't very sensitive to the idea of signal quality. You know, you can tell just by the way we put pulse ox probes anywhere we can find a place.

And that's really the primary issue when it comes to where these reflectance probes are located on the body and how well they work. And, you know, we have to learn all that.

You know, and so, the result is that more data would probably help us get started in learning about how to use that kind of information in assessing the quality of a signal.

CHAIRPERSON LISBON: I get to go last.

DR. LEUNG: The one comment that I have is, currently, when you open pulse oximeter probe, you know, it's in a little plastic bag. And there's really no instruction to the user.

It's kind of implied that when you have access to that you know how to use it. And that may not be true based on what we learned today. And that may be one.

You know, one consideration is to include that information as almost like a drug insert with respect to the different sensors. There are many conditions in which the sensors wouldn't work.

And that might help the users to decide whether the information is real or not. I mean, frequently we don't change our clinical decision based on one number.

But, what I'm hearing is that, you know, more and more we're using this device to drive our clinical decision making. We're no longer using blood gasses.

And so, I think we need to know the limitations.

DR. BIRNBACH: Taking what Jackie said one step farther, we haven't gotten the information of how, what percentage of anesthesiologists actually use one type or the other.

My gestalt would be that we almost always use transmission devices in the operating room environments. Anyway, that may be different. It's clearly different in the neonatal experience and probably different in the ICU experience.

It would probably be helpful to have some kind of analysis of what different positions you're using, when, why to have that data. And then the next step would be o have better instructions with warnings, for example, which I don't think exist right now.

DR. OTULANA: I think I agree with Dr. Cote that based on the presentation we saw today, there's probably no technical differences between the two forms of pulse oximeter.

But, the clinical use, I think, is where needs to be focused in terms of the application of the sensor to the different parts of the body. So, in response to the FDA question then, perhaps the focus has to be the testing of the oximeter on the -- of the sensor rather, on the relevant part, where it would be recommended for use.

So, if it's going to be on the forehead, if it's going to be on the back, I think data needs to be generated specifically for these areas.

MS. KLINE: I too practicing in the pediatric ICU world, am not familiar with, have not used the reflectant technology. But I agree with the comments of the other panelist and Dr. Cote as far as looking at these devices further and looking at their performance, as well as having more information on exactly where to place these on the patient's body.

MS. PETERSON: I concur with the other panelists' comments and with Dr. Cote.

CHAIRPERSON LISBON: Okay. Charlie, go ahead.

MEMBER COTE: Yes, I think we are particularly concerned with the so-called neonatal and the premature neonate, again, because the skin thickness is so dramatically different in those children.

Perhaps the skin profusion might be different in those children. But, in fact, many moment-to-moment decisions are made in terms of oxygenation because the concern that we have in taking care of pre-term babies is an excess amount of oxygen could lead to the development of retrolental fibroplasias.

So, there's this general recommendation that you keep the saturation somewhere between 88 and, you know, 95 or 93 percent, it depends which nursery you're in.

But, there's always this narrow range that you like to keep it in. And you don't want the baby at 100 percent. If this oximeter is under-reading and says that the baby's saturation is 80 percent when in fact it's 87 percent, that could have significant implications for the baby, because then you're going to give them an excess of oxygen and actually be causing a situation that could potentially cause retinal injury thinking that you're providing optimal care.

And I guess that's one of my concerns in terms of the use of this device in pre-term babies. In full term babies it's not an issue.

CHAIRPERSON LISBON: All right. I guess I get to make the last note on this question. I agree that there needs to be further comparative data on these two devices, these two techniques for obtaining a pulse oximetry, particularly the lower oxygen saturations.

They are used under different clinical situations and signal -- ratios, and those such things are important in the various devices. I'd also like to see more data out there as to how these should be placed and where they should be placed.

And I'd like to see more validation data for that. And, with that, I believe we're done if none of the panelists have further comments on question number one