Medical Devices Advisory Committee




НазваниеMedical Devices Advisory Committee
страница3/16
Дата26.10.2012
Размер0.59 Mb.
ТипДокументы
1   2   3   4   5   6   7   8   9   ...   16
Agenda Item: Clinical Experience at USC

DR. MESTMAN: Good morning.

My name is Jorge Mestman. I am a consultant to MiniMed and my trip to today's meeting was paid for by the company. I have no ownership interest in the company, except for a small number of shares of MiniMed stock, purchased by my wife in the open market.

I would like to review with you in the next few minutes some aspects of the complications of diabetes mellitus and our experience with the use of the Continuous Glucose Monitoring System that eventually will help the clinician to improve glycemic control in people with diabetes.

I would like to review first some data regarding mortality due to cardiovascular disease in people with diabetes, the revelation of complications and frequency of hypoglycemic reactions, the studies conducted in our institutions, our experience and the patients' experience with the use of the system and our conclusions.

Diabetes mellitus is a very serious disease that affects over 18 million people in America. The mortality from cardiovascular heart disease in patients with diabetes is two to four times greater than people without diabetes. In this study, directed by the Joslin Diabetes Center, the mortality due to cardiovascular heart disease is increased not only in males but also in females and indirectly related to the duration of diabetes. As you can see here, is much greater in people with diabetes as compared with known diabetic patients.

In the last two decades compelling evidence has accumulated demonstrating that good diabetes control, as measured by hemoglobin A1c values, reduces significantly the risk of progress of diabetic complications. In this particular study, hemoglobin A1c at 7 percent reduced the risk of complications and this is the number that is recommended by the American Diabetes Association and other entities to achieve in the care of people with diabetes.

Unfortunately, one of the most serious complications that we have in people with diabetes trying to achieve this hemoglobin A1c of close to 6 to 7 percent is a significantly increased incidence of hypoglycemic reaction. And as you can see here, there is an inverse relationship in the incidence of severe hypoglycemic reaction as compared with the hemoglobin A1c determinations.

In order to prevent hypoglycemic reactions, we recommend patients to check their finger stick blood sugars several times a day. This is a typical portrait of a patient with Type 1 diabetes that is under intensive insulin therapy.

In this case, the blood sugar is measured by finger stick before each meal and at bedtime. The values obtained are in general acceptable in most medical settings and there is little difference between the first and the second day. However, in spite of these three meals and bedtime values, the hemoglobin A1c is still elevated and does not achieve our goal.

If we were able to check the blood sugars every half an hour or every hour, we will see that in this same patient in which the three meals, glucose values are acceptable. There is a wide fluctuations in blood sugars before meals, after meals and during the night that explain why the hemoglobin A1c is still out of the so-called normal range.

Furthermore in this particular case, the fluctuations in blood sugars go as high three times milligram percent, as low as 50 milligrams percent. The patient was unaware of these fluctuations on this particular day.

I would like to mention to you that we have performed several feasibility studies at USC Center for Diabetes and Metabolic Disease and we have studied a total of 58 patients using 118 sensors. The results from these disability studies indicate that the sensor performance was equivalent when it was inserted in the abdominal area or other subcutaneous insertion sites.

The modifications to the shape of the sensor assembly eliminated mild excoriation experienced by two subjects. Sensor performance was not affected by exercise, ingestion of ascorbic acid or acetaminophen.

Our experience with the use of this sensor in this group of patients in the feasibility studies show that there was no infections in the site of the insertion of the sensors. The adverse events and device complications were not clinically significant.

The Continuous Glucose Monitoring System is easy for patients to use both inserting the sensor and entering the data in the monitor.

For the patients involved in the study, the teaching experience was unique. For the first time, they understood the difficult in achieving target values on a daily basis. They were unaware that their glucose fluctuations vary so often and widely and without symptoms. They understand that their finger glucose values may not correlate with hemoglobin A1c results.

All of them expressed their desire to participate in future studies and we are eager to know when the system could become commercially available.

In conclusion, the Continuous Glucose Monitoring System was well-received by patients, was easy to insert and use and did not introduce safety risks, as adverse events were not clinically significant.

We provide clinicians and health care members with more data than is otherwise available. Finally, it is my personal opinion that the application of this system will profoundly improve the management of persons with diabetes.

Thank you for your attention.

I would like to introduce Dr. Gross, the next presenter.

Agenda Item: Multi-Center Study, Overview and Data Analysis

DR. GROSS: Good morning. I would like to thank the panel for their consideration of the CGMS this morning.

I will be presenting the results of the multi-center clinical evaluation of this Continuous Glucose Monitoring System, which was performed in the second half of 1997. I will briefly review the goals of the study, its design, the results that were obtained and the conclusions that follow from those results.

The CGMS has been the focus of a series of clinical studies dating back to 1994. As Dr. Mestman has discussed, the feasibility studies were conducted at USC Center. In these early studies, subjects wore glucose sensors for up to four days in a controlled setting and under the direct supervision of clinical staff.

The multi-centered study identified in this table as GSO004-II, the bottom row, marks the first time that subjects used the system at home for up to 20 days while maintaining their normal daily routines.

During the multi-center study, data were collected using the CGMS under the exact condition of its intended use. We appreciate the large amount of material the panel members have received from our submission and I would like to help organize that material and focus my presentation today.

We have submitted three reports based on the multi-center study. The first was a clinical study report that was submitted with our 510(k) submission on December 15th of 1997. That report summarized all data that had been received by us as of December 8th of that year. Following FDA review of the 510(k), the application was converted to a PMA and additional data analysis was performed.

The report of this supplemental analysis, dated October 30th, 1998, also included additional data that had been received after our 510(k) submission. A third report was submitted February 9th of this year, which focused on a linear regression calibration that John Mastrototaro has described.

This report also updated all of the safety and efficacy results from the original clinical report to include the additional data that we had received. The results I am going to discuss today are based on this final February 9th report using the regression calibration method and include all available data.

The multi-center study was designed with input from the FDA to demonstrate the safety and efficacy of the CGMS for home use. The primary efficacy hypothesis focused on the sensors' ability to accurately track blood glucose trends. In addition, the study data were used to establish the duration of function for an individual sensor and to test for changes in performance during the life of the sensor.

At the request of the Agency, both the physician and patient sensor insertions were studied. Because the CGMS does not pose a significant risk to study subjects, no IDE was required for this device. Input on the study protocol was solicited from the FDA prior to commencing the study and with their cooperation, these key protocol requirements were developed.

A minimum of five sensors were to be worn sequentially by each subject. Each sensor could be worn for a maximum of 96 hours, yielding a maximum study duration of up to 20 days. Subjects who wore at least five sensors and who provided at least 15 days of device experience were considered to have completed the study.

Each subject had sensors inserted by the study personnel, in addition to the sensors they inserted themselves. Finally, subjects performed at least 11 blood glucose measurements each study day using an Accu-Chek Advantage meter.

The criteria for participation in this study were intentionally broad. Basically, adults with diabetes, who used insulin, who have the capacity and the willingness to follow the protocol and who are free of conditions that would make it difficult for them to follow the protocol could participate.

Eligibility was confirmed by physical exams and lab results obtained prior to study participation. This time line gives the schedule of office visits and sensor insertions for a prototypical subject. At visit 1, patients had their first sensor inserted by the physician. They were then sent home to wear the sensor for up to four days.

At visit 2, the patients had the first sensor removed and the second sensor inserted by the physician or another member of the diabetes care team. At visit 3, the subject performed their first patient insertion under supervision. At this visit, the subject would be given additional sensors to insert at home. At visit 4, the final physician insertion was performed and at visit 5, study participation was completed.

At each office visit, the CGMS memory was downloaded to the computer and the subject's abdomen was examined for any signs of sensor site irritation. Between office visits, the subject was instructed to leave the sensor in place, to leave the CGMS monitor turned on and to perform all scheduled meter readings. If the CGMS produced an alarm, the subject contacted the coordinator, who helped resolve the alarm.

At this time, the site coordinator determined if the sensor should be removed and replaced. In this way, sensors were removed prior to 96 hours and more than five sensors were worn by some subjects. Conversely, if the timing of the scheduled office visit required it, investigators did instruct some subjects to leave a functioning sensor in place beyond the 96 hour maximum.

Each day of the study, subjects were asked to perform 11 meter readings. These readings served three purposes. First, readings were taken before breakfast, before lunch, before dinner and at bedtime in order to allow the subject to manage his or her diabetes.

Second, an additional reading was taken in the morning and the average of these two fasting readings were used to test the sensitivity of the sensor. Third, meter readings were taken one and two hours after each meal to capture glucose excursions.

The specific times of these meter readings were not dictated by the protocol, but rather were determined by the timing of each subject's meal and did differ from subject to subject and across study days. Twice during the study, subjects were asked to perform 2:00 a.m. and 4:00 a.m. readings and finally subjects were instructed to perform a meter reading every 15 minutes following any hypoglycemic event.

The study protocol was implemented at four investigational sites; 62 subjects were enrolled in the study; 17 at site 1 and 15 at the remaining sites. Three subjects did not complete the minimum of five sensors. A total 415 sensors were studied and worn for 1,153 days, yielding more than three years of cumulative sensor use.

Over 300,000 individual sensor readings were obtained during the study and were paired with more than 9,000 meter readings, which were obtained during the first 72 hours of each sensors' use. This slide describes the demographic and baseline characteristics for all the subjects enrolled in the multi-center study.

The sample is representative of the population of persons with Type 1 diabetes. Of particular interest is the wide range of glycemic control as reflected in HBA1c values between 5.4 and 10.6.

The study protocol stipulated that each sensor be worn up to a maximum of 96 hours or until it lost sensitivity to glucose. Sensor function was defined as the duration of valid electrical signal from the sensor. One hour after each sensor was inserted, its sensitivity was checked by entering the average of two meter readings into the monitor.

If the sensor sensitivity slope was within range, the sensor signal was marked as valid in the monitor's download file. This sensitivity check was repeated at least once each study day.

End of function was defined as the removal of the sensor, the occurrence of a calibration error alarm, which indicted the sensor's sensitivity slope was out of range or a disconnect alarm indicating the absence of an electrical signal from the sensor. During the clinical study, the medial duration of function was 69 hours.

As I will describe in a few minutes, this value is likely an underestimate as some of the sensors were removed prematurely, due to problems with an auxiliary component of the system unrelated to the sensor. We believe the corrective actions we have taken to resolve these mechanical issues have been effective and we do not expect them to recur.

Next, I will present the safety results of the study. The safety evaluation covers the entire period of time sensors were inserted and all safety events have been reported, not just those occurring the period of valid sensor function that I have just described.

The CGMS proved to be extremely safe during the clinical study. Only seven device-related adverse events were reported in the 1,153 days of sensor use. All seven of these events concerned irritation, discomfort or minor bleeding or bruising at the sensor insertion site.

These events were characterized as mild or moderate in nature by the investigator and subjects were covered fully in each case. These events are anticipated with an insertable device of this type and are similar to those seen among patients using external and slim pump infusion sets.

Based on 1,153 days of sensor experience from the study, we would expect that only 1.8 patients out of a hundred, who used the CGMS for 72 hours would experience a similar event. Subjects had an abdominal exam performed at every office visit and, again, at seven and thirty days after the end of their study participation.

In 85 percent of these exams, the sensor site was observed to be clean, dry and intact. Fifty-six exams resulted in an observation of irritation or minor bleeding and bruising. Twenty-one of these 56 observations were attributed to the tape that was used to secure the sensor assembly to the subject's abdomen. In many cases, switching to an alternate brand of tape resolved this irritation.

Here we see a breakdown of the observations. Again, these events are anticipated with a device of this type and are similar to those seen with subcutaneous insulin infusion catheters. Also, for only seven of these events did the investigator feel an adverse event report was necessary and then only mild or moderate severity was reported.

There were 62 device complications reported during the study. Device complaints were used to document any instance where a device needed to be replaced or was taken out of service prematurely. The majority of these events involved the component of the system other than the sensor itself, as indicated here by the mechanical events and the miscellaneous events.

Turning to the details of these events, the study did provide a rigorous test of the CGMS components and led us to several specific improvements in our manufacturing process. For example, the moisture, which was observed to affect the circuitry in the monitor was resolved by sealing the O ring in the monitor case assembly.

The next two categories, cable disconnections and inadequate sensor contact, were resolved by reworking the connector between the cable and the sensor. Again, we believe these corrective actions we have taken have resolved these mechanical issues and have also removed their effect on the duration of sensor function reported earlier.

This slide documents ten instances where a sensor was removed earlier than expected. The first two types of events each led to alarms that alerted the subject to the problem. Low sensor sensitivity results in a calibration error alarm and sensor dislodgment resulted in a disconnect alarm.

The two reports of discomfort seen here were also accompanied by an adverse event report and are not separate occurrences.

I would like to turn now to the efficacy data. These results are based on all sensors used in the study, regardless of any device complaints that were reported for that sensor. Because the proposed labeling for the sensor limits its use to a maximum of 72 hours, only data from the first three days of each sensor's use were included in these results.

Here we see a sensor plot from the multi-center study. The dots represent the meter readings that were taken during that study day and the blue line represents the continuous glucose monitoring profile. Glucose concentration in milligrams per deciliter is given on the Y axis and time of day from midnight to midnight is given on the X axis.

The goal of continuous monitoring is to accurately track blood glucose, which this sensor does. The continuous profile also allows identification of both high and low excursions. In order to analyze the sensor's performance, we must reduce the data to a series of paired sensor meter data points. Therefore, the statistical analysis considers only the pairing of each meter and sensor value. It is these paired data points that form the basis of the three performance measures that I will present.

First, let me clarify a distinction between the regression calibration, as it is used in the CGMS, and the way that it was applied to the clinical study data. In clinical use, the calibration is based on all of the meter values entered into the CGMS monitor.

As John has described, the CGMS monitor stores the sensor's raw electrical signal in nanoamperes. In order to use this information, the signal must be converted back into a glucose concentration in milligrams per deciliter.

This calibration to blood glucose is performed when the data is downloaded onto a personal computer. In order to produce the most accurate calibration and to reduce any effect of error in the meter measurements used to calibrate the sensor, all of the finger stick values entered into the monitor are used for calibration.

In contrast, for the clinical study data, a maximum of four meter values were used to calibrate the clinical study data. The remaining meter readings were used to evaluate the performance of the sensor. This separation of meter values into a calibration subset and an evaluation subset was done to avoid any problems with evaluating the sensor against the same data that was used to calibrate it.

Limiting the number of study calibrations to a maximum of four produces what we would feel to be the worst case analysis, as patients will perform at least four meter readings each day they wear the sensor. Put another way, the performance of the sensor in clinical use will exceed that of a clinical study, due to the use of additional meter readings for calibration.

The calibration values were selected by defining four target times during the day, 7:00 a.m., 12:00 noon, 7:00 p.m. and 10:00 p.m. A time window was then created around each target time. If one or more meter values fell within this time window, the reading closest to the target time was used for calibration. If no reading fell within the window, the calibration was performed with less than four values.

Looking again at the plot for the multi-center study, we see first of all the four black dots represent the meter readings that were used for the purposes of calibration. The remaining dots seen here as red were used to evaluate the performance of the sensor. If this data were downloaded in the physician's office, all 11 meter readings would be used to calibrate the sensor producing a more accurate calibration.

This table summarizes the number of calibration values identified for each day of the study. Fewer than four values were obtained if there was no meter reading taken during one or more time frames or for days in which the sensor was inserted or removed, where less than four time frames were available for evaluation.

On 85 percent of the study days, two or more calibration values were identified. Three measures of effectiveness were calculated from the study data. The traditional way of looking at a blood glucose meter's performance is to compare its readings to a reference value and to calculate the numerical difference.

This method due in part to Bland and Altman measures the point to point agreement between the two devices. This measure provides useful information, but it is perhaps the least appropriate to the intended use of the CGMS. A more appropriate measure, category agreement, reflects the sensor's ability to match the meter's identification of glycemic highs and lows.

This measure is more relevant to the intended use of the device. The measure most relevant to the intended use is the intra-day correlation, which reflects the accuracy of the sensor to track the blood glucose profile over time.

Each of these measures provides meaningful but distinct information about sensor performance. Here we see a summary of the numerical agreement obtained from the study. First of all, both the sensor and the meter provide glucose values from 40 to 400, spanning the operating range of the sensor.

The sensor on average showed very little bias relative to the meter. Overall, the sensor readings were 5 milligrams below the meter readings with an average percent difference of only .3 of 1 percent. Several components do contribute to the variability observed in different scores, including the precision of the sensor, interstitial glucose that leads or lags blood glucose and error in the meter readings.

Because the intended use of the CGMS is to identify glycemic excursions, in cooperation with the FDA review team, an analysis was developed that focused on these excursions. First, three glycemic categories were developed. Values between 70 and 180 were deemed to be in target. Values below 70 were considered low and values above 180 were considered high.

Each meter value and each paired sensor value was categorized, using these boundaries with one exception. Values falling within 20 percent of each boundary were considered not categorizable and were not included in the analysis.

The 20 percent zone acts as a confidence interval surrounding the boundary. Due to the inherent error in meter readings, values falling into this zone cannot be considered reliably different from the boundary itself. In the case of the sensor values, the literature suggests that interstitial glucose may lead or lag blood glucose by ten minutes or more. Although this time lag is insignificant when viewing the continuous profile, as evidenced here by a meter value that appears to fall very close to the continuous profile, it can create cases where an accurate sensor reading is one side of the boundary and the paired meter value is on the other.

Due to the fact that this meter reading is paired with a value directly below it in time and the fact that the meter reading falls into the high category, above 180, and the sensor value falls into control. Because these cases were ambiguous in terms of agreement, they were not considered in the analysis of category agreement.

This table summarizes the category agreement between the sensor and the meter. We see here first of all the three categories of meter readings, low, in the target and high across with the three categories of CGMS glycemic level, low, in control and high.

This diagonal represents agreement where the sensor and meter fell into the same category. For 87 percent of the analyzed pairs, the sensor and the meter fell into the same category. Also, there were only two instances of extreme disagreement, where the meter read high and the sensor read low and no instances where the meter read low and the sensor read high.

This particular category represents the category of greatest risk from a clinical perspective, where a clinician reviewing the sensor data might choose to administer insulin. None of these instances occurred.

Turning now to the analysis of intra-date correlation, the correlation was calculated for each calendar of sensor use. Daily correlation was used because the sensor is calibrated daily and because calculating correlation in this way helps keep differences in calibration from sensor to sensor from reducing the overall correlation. The median correlation was .92, with 75 percent of the correlations falling above a value of .75. This excellent correlation confirms the sensor's ability to accurately track glucose trends. In our submissions, we have described two calibration methods, a single point calibration and a multiple point regression calibration.

As we can see in this table, the use of the regression calibration improves all three measures of agreement, numerical, categorical and correlational. We also analyzed two factors that might influence sensor performance, day of use and type of insertion. No significant effects of time on any of the three performance measures were observed with stable performance over the three days for the numerical agreement, categorical agreement and intra-date correlation.

The lack of statistical significance is shown here at P values above .05. Finally, we compared the performance of sensors inserted by the health care team with those inserted by patients. No significant difference was seen in numeric agreement. A small advantage was seen for medical professional insertions in correlation and a marginally significant improvement with patient insertions was seen in category agreement.

In conclusion, the multi-center study provided a thorough test of the CGMS during home use. The system was extremely, confirming that individual sensors can be worn for up to 72 hours and that sequential sensors can be wore from 15 to 20 days. Only mild or moderate site irritation was observed, which is typical of that seen within infusion sets and subjects recovered fully in every case. Similar events should occur in less than 2 percent of patients using the sensor for 72 hours.

Many of the mechanical device complaints were addressed and resolved during the clinical study. The performance of the sensor was stable over time and was not consistently different for patient versus health care and staff insertions. An overall category agreement of 87 percent and the median correlation of .92 demonstrates that the sensor accurately tracks blood glucose and can be used to identify the patterns of glucose fluctuation in persons with diabetes.

Dr. Alan Marcus will now discuss the clinical use of the CGMS.

DR. NIPPER: Excuse me. Dr. Habig, after this next speaker, we will take a break.

I assume, Dr. Marcus, you will be finished within a few minutes?

DR. MARCUS: Yes.

1   2   3   4   5   6   7   8   9   ...   16

Похожие:

Medical Devices Advisory Committee iconVeterinary medicine advisory committee

Medical Devices Advisory Committee iconRanch hand advisory committee

Medical Devices Advisory Committee iconAdvisory committee on immunization practices

Medical Devices Advisory Committee iconNational Vaccine Advisory Committee (nvac)

Medical Devices Advisory Committee iconExternal Advisory Committee on Cities and Communities

Medical Devices Advisory Committee iconWildlife Diversity Policy Advisory Committee

Medical Devices Advisory Committee iconSchedule 5 Appendix c other Medical Devices

Medical Devices Advisory Committee iconPeer reviewed by the Arizona Department of Commerce Economic Research Advisory Committee

Medical Devices Advisory Committee iconFood and drug administration national institutes of health advisory Committee on: transmissible spongiform

Medical Devices Advisory Committee iconAdvisory Committee, Cuyahoga Valley School-to-Career Consortium, Broadview Heights, Ohio 1996-2002

Разместите кнопку на своём сайте:
Библиотека


База данных защищена авторским правом ©lib.znate.ru 2014
обратиться к администрации
Библиотека
Главная страница