Insulin pump and continuous glucose monitoring system

Chapter 15

Icon table
 
Do not reuse
Attention: See Instructions for Use
Manufacturer
Date of manufacture (year - month)
Batch code
Use by (year - month)
Catalogue number
Device serial number
Storage temperature range
Magnetic Resonance (MR) unsafe
Fragile product
Type BF equipment (Protection from electrical shock)
Conforms to IEC60601-1 sub-clause 44.6 and IEC60529 standard
For more information about IPX7, see the Glossary on page 229.
Recycle
Radio communication
Pump specifications
204

Signifies compliance to Australian EMC and Radio communications
requirements
Signifies compliance with Industry Canada EMC and Radio communications
requirements
Pump specifications
205
Chapter 15

Warranty
Medtronic Diabetes warrants the Medtronic MiniMed Insulin Pump against defects in materials and
workmanship for a period of 4 years from the date of purchase. 
During the warranty period, Medtronic Diabetes will, at its discretion, either repair or replace (with a new
or recertified pump, at Medtronic Diabetes’ discretion) any defective pump or motor, subject to the
conditions and exclusions stated herein. In the event that a pump is repaired or replaced, the warranty
period will not be extended.
This warranty is valid only if the Medtronic MiniMed Insulin Pump is used in accordance with the
manufacturer’s instructions. This warranty will not apply:
•
If damage results from changes or modifications made to the pump by the user or third persons after
the date of manufacture.
•
If damage results from use of non-Medtronic reservoirs and/or infusion sets.
•
If damage results from service or repairs performed by any person or entity other than the
manufacturer.
•
If damage results from a Force Majeure or other event beyond the control of the manufacturer.
•
If damage results from negligence or improper use, including but not limited to: improper storage,
submersion in water or physical abuse, such as dropping or otherwise.
This warranty shall be personal to the original user. Any sale, rental or other transfer or use of the product
covered by this warranty to or by a user other than the original user shall cause this warranty to
immediately terminate. This warranty does not apply to batteries, infusion sets, reservoirs, and other
accessories.
The remedies provided for in this warranty are the exclusive remedies available for any breach
hereof. Neither Medtronic Diabetes nor its suppliers or distributors shall be liable for any incidental,
consequential, or special damage of any nature or kind caused by or arising out of a defect in the
product.
Pumps purchased through our Pathway Program will assume the warranty stated in our Pathway
Program agreement.
All other warranties, expressed or implied, are excluded, including the warranties of merchantability
and fitness for a particular purpose.
Pump specifications
206

Sensor accuracy
NOTE: 
You should review the information in this section with your healthcare professional.
 
The Medtronic MiniMed Paradigm 522/522K/722/722K pumps use a glucose sensor to continuously monitor
your glucose levels. The 522/522K/722/722K pumps use the same algorithm as the Guardian RT
®
. The
Guardian RT was evaluated in two clinical studies and the following information in this section describes
the findings. The remainder of this section discusses the results of this pivotal study, and uses the term
Guardian RT.
NOTE: 
The Guardian RT is not an abbreviation for the Guardian REAL-Time CGM System. The
Guardian RT is a continuous glucose monitoring system that was developed before the Guardian REAL-
Time CGM System. The Guardian RT was originally named TGMS II, and later renamed Guardian RT.
 
The Medtronic Guardian RT uses a glucose sensor to continuously monitor your glucose levels. The
Guardian RT sensor is calibrated using your home blood glucose meter. Once calibrated, the Guardian RT
reports glucose values every five minutes. These values were compared to reference laboratory blood
glucose measurements to check the Guardian RT’s performance characteristics in two clinical studies.
1, 2
Although presentations to characterize performance of the Guardian RT are given below, there is no
commonly accepted statistical approach for capturing the performance of continuous glucose monitors
such as the Guardian RT. Performance may be best characterized by viewing graphs called time-elapsed
plots. In these plots, the values from Guardian RT for one subject over time are overlaid with values at the
same time from the glucose reference method. Three representative time-elapsed plots are shown at the
end of this section of sensors that exhibited excellent performance, average performance and poor
performance.
1. Medtronic Diabetes, A Frequent Sample Accuracy Evaluation of the Medtronic Diabetes Telemetered Glucose Monitoring
System II (TGMS II) in Subjects with Type 1 Diabetes Mellitus, August 2004.
2. Medtronic Diabetes, An Accuracy Evaluation of the Medtronic Diabetes Guardian RT Glucose Monitoring System in Pediatric
Subjects with Type 1 Diabetes Mellitus, February, 2006.
Sensor accuracy
207
Appendix A

Performance results in adults
The performance of the Guardian RT in adults was evaluated in a clinical study. Guardian RT results were
compared to plasma glucose values from a reference method, the YSI 2300 STAT Plus™ glucose analyzer
(referred to as YSI). Sixteen subjects with Type I diabetes participated in a single-site in-clinic study.
Subjects ranged in age from 18 to 65 years old. Each subject wore two Guardian RT systems
simultaneously. One Guardian system was calibrated an average of 3.5 times per day, and the other was
calibrated approximately five times per day using the Paradigm Link blood glucose meter. YSI
measurements were taken every thirty (30) minutes.
Users and their healthcare professionals should consider that performance in this study may be idealized,
and that performance may be worse when the Guardian RT is used in a less controlled home setting. For
example:
•
The mean Hemoglobin A1c among the 16 participants was 8.2 percent. As hemoglobin A1c levels rise,
conditions often occur which are most challenging to test systems measuring glucose in interstitial
fluid, i.e., higher glucose levels, more rapid changes in glucose concentrations, and often more
hypoglycemic episodes.
•
Subjects saw, on average, between 4 and 5 fingerstick values per day. This enables subjects to better
manage their diabetes when compared to those who perform less fingersticks per day. Agreement
between Guardian RT and YSI values is shown to be closer at mid-range glucose levels, as compared to
agreement at low or high glucose concentrations.
•
Subjects were more limited in their activities than what may exist in home use, and they were
provided with all their meals. Sensors were also inserted by clinic staff rather than the subjects
themselves. Subjects who are more active, or with poor eating habits, may create more challenging
conditions for the Guardian RT.
•
Performance of the Guardian RT may vary depending on the glucose meter used and how well the
meter is maintained. It is important to carry out quality-control checks on the meter and code the
meter according to the manufacturer's instructions to optimize performance of the Guardian RT.
Appendix A
208

Accuracy of Guardian RT readings
In this study, YSI measurements (taken every half hour) were paired with the corresponding Guardian RT
reading (taken every five minutes). Pairing was done by selecting the Guardian RT value closest in time to
the YSI test result. Agreement was analyzed by comparing paired glucose measurements.
Agreement between the matched pair was estimated by evaluating the difference between the Guardian
RT reading and the YSI measurement. The difference between them was calculated as a percentage of the
YSI (Mean Absolute Percent Difference). The bias was also calculated, and it is defined as the overall
difference between the Guardian RT glucose values and the YSI values. The paired glucose measurements
are summarized below.
Number of Paired Glucose Measurements
3941
Mean Absolute Percent Difference (± SD)
19.7 ± 18.4%
Bias
-0.8 mmol/L (-15.0 mg/dL)
The accuracy of the Guardian RT was also evaluated by calculating the percentage of Guardian RT readings
within 20 percent and within 30 percent of the YSI reading (or within 1.1 mmol/L (20 mg/dL) in the low
glucose range). Results are shown below.
Plasma Glucose
Range (mg/dl)
Plasma Glucose
Range (mmol/l)
Number of Paired
Readings
Percent Within
20%
Percent Within
30%
Overall
3941
62%
79%
40-80
a
2.2-4.4
356
68%
68%
>80-120
>4.4-6.7
769
60%
77%
>120-240
>6.7-13.3
2362
62%
81%
>240
>13.3
454
61%
82%
a. For the Low glucose range, 2.2-4.4 mmol/L (40-80 mg/dL), the value shown is the percent within 
1.1 mmol/L (20 mg/dL).
Sensor accuracy
209
Appendix A

The Clarke Error Grid was used to assess the clinical relevance of the differences between the Guardian RT
readings and the comparative YSI measurements. The Clarke Error Grid divides a correlation plot into 5
zones. See following table.
Results in zones A and B are considered clinically acceptable, while results in zones C, D, and E are
potentially dangerous and, therefore, clinically significant errors. The Clarke Error Grid zones are labeled
on the correlation plot.
Zone
Description
A
Clinically accurate, would lead to correct treatment decisions
B
Would lead to benign decisions or no treatment
C
Would lead to over-correction of normal glucose levels
D
Would lead to failure to detect and treat high or low glucose levels
E
Would lead to erroneous treatment decisions
Appendix A
210

The plot below is a correlation plot of Guardian RT readings versus readings from the reference method,
the YSI 2300 Glucose Analyzer. It is overlaid with the Clarke Error Grid.
The total number of paired data points is 3941.
Meter BG (mg/dL)
Sensor BG (mg/dL)
Sensor accuracy
211
Appendix A

The percent of Guardian RT readings in the previous graph are presented in the following table according
to the percentage of points falling within each zone (A-E). Results are further broken down (stratified)
according to the range of glucose concentrations.
Glucose
Range
(mg/dl)
Number and
(%) of Data
Points
Evaluated
A + B
A
B
C
D
E
40-80
356 (9)
271 (76.1)
214 (60.1)
57 (16.0)
2 (0.6) 80 (22.5) 3 (0.8)
81-120
769 (20)
768 (99.9)
463 (60.2)
305 (39.7)
1 (0.1) N/A
a
N/A
121-240
2362 (60)
2352 (99.6) 1476 (62.5) 876 (37.1)
4 (0.2) N/A
6 (0.2)
>240
454 (11)
394 (86.8)
277 (61.0)
117 (25.8)
N/A
59 (13.0) 1 (0.2)
Overall
3941 (100)
3785 (96.0) 2430 (61.7) 1355 (34.4) 7 (0.2) 139 (3.5) 10 (0.2)
a. N/A means that the Clarke Error Grid does not consider the possibility of these zones in that concentration range.
Appendix A
212

Precision of Guardian RT readings
This study was also designed to look at the reproducibility of two sensors worn simultaneously at different
locations on the body. Precision was estimated by comparing the glucose readings from the two Guardian
RT systems. In this study 11,475 paired sensor Guardian RT values were obtained. On average, they were
different by 17.2 percent. The following figure is an example of how data was paired in this study. In
the graph there are two tracings of Guardian RT values. Each tracing comes from a different Guardian RT
unit worn by one subject during a one-day period.
Time of day
Glucose Conc
 (mg/dL)
The above chart, Guardian RT device #2 is represented by a solid line, and the Guardian RT device #4 by a
light-gray line.
Sensor accuracy
213
Appendix A

Low and High Alerts in adults
The ability of the Guardian RT to detect high and low glucose levels was measured in the same clinical
study. Since it is important to set the alert levels in a conservative fashion, the Low Glucose Alert should
be set at a value slightly higher than the value of blood glucose you want to detect, and the High Glucose
Alert should be set at a value slightly lower than the value of blood glucose you want to detect.
NOTE: 
Please ask your healthcare professional which low and high alert setting is best for you.
 
The Low Glucose Alert
The Low Glucose Alert was evaluated for its ability to detect glucose levels at 70 mg/dL (3.9 mmol/L), or
below, using the YSI 2300 STAT Plus glucose analyzer. As a reference, with the Low Glucose Alert set at 70
mg/dL (3.8 mmol/L), 49% (100/205) of low glucose events were detected by the Guardian RT. Better
detection of low blood glucose can be obtained by setting the Low Glucose Alert level higher. For
example, setting the Low Glucose Alert at 90 mg/dL (5.0 mmol/L), instead of 70 mg/dL (3.9 mmol/L),
increases the ability to detect low blood glucose levels from 49% to 82%.
Sometimes the Guardian RT will alert when the blood glucose levels are not low. When the Guardian RT
Low Alert was set at 70 mg/dL (3.9 mmol/L) in this study, 43% of the results were considered false alerts
(actual blood glucose values are greater than 85 mg/dL (4.7 mmol/L)). This percentage may be
exaggerated because blood glucose may be dropping when the Guardian RT alerts. The table below shows
the percent of Low Glucose readings correctly identified by the Guardian RT for specific settings.
Guardian RT Low Alert
Setting (mg/dL)
Guardian RT Low Alert
Setting (mmol/L)
True Alert Rate*
False Alert Rate**
70
3.9
49%
60%
80
4.4
68%
64%
90
5.0
82%
75%
100
5.6
90%
79%
* True Alert Rates are the % of times when the glucose level was at or below the alert setting and the alert
sounded.
** False Alerts Rates are the % of times when the Guardian RT Sensor alarmed but the blood glucose level
was greater than the alert setting.
Appendix A
214

Increasing the Low Alert settings will improve the ability to detect low blood glucose events, but it will
also increase the frequency of Guardian RT false alerts for blood glucose levels not below the target value.
You should consider this trade-off between the improved ability to detect true low blood glucose versus
the increased number of false alerts when setting the low alert threshold.
The High Glucose Alert
The High Glucose Alert was evaluated for its ability to detect glucose levels at 250 mg/dL (13.8 mmol/L),
or above, using the YSI analyzer. As a reference, with the High Glucose Alert set at 250 mg/dL (13.8
mmol/L), 53% (195/365) of high glucose events were detected by the Guardian RT. Better detection of
high blood glucose can be obtained by setting the High Glucose Alert level lower. For example, setting the
High Glucose Alert at 190 mg/dL (10.6 mmol/L), instead of 250 mg/dL (13.8 mmol/L), increases the ability
to detect high blood glucose levels from 53% to 85% (see the table below).
Sometimes the Guardian RT will alert when the blood glucose levels are not high. When the Guardian RT
High Alert was set at 250 mg/dL (13.8 mmol/L) in this study, 7.2% of the results were considered false
alerts (actual blood glucose values are less than 225 mg/dL (12.5 mmol/L)). This percentage may be
exaggerated because blood glucose may be rising when the Guardian RT alerts. The table below shows the
percent of High Glucose readings correctly identified by the Guardian RT for specific settings.
Guardian RT High Alert
Setting (mg/dL)
Guardian RT High Alert
Setting (mmol/L)
True Alert Rate*
False Alert Rate**
190
10.6
85%
64%
200
11.1
81%
58%
225
12.5
67%
40%
250
13.8
53%
25%
* True Alert Rates are the % of times when the glucose level was at or above the alert setting and the alert
sounded.
** False Alerts Rates are the % of times when the Guardian RT Sensor alarmed but the blood glucose level
was lower than the alert setting.
Decreasing the High Alert settings will improve the ability to detect high blood glucose events, but it will
also increase the frequency of Guardian RT false alerts for blood glucose levels not above the target value.
You should consider this trade-off between the improved ability to detect true high blood glucose versus
the increased number of false alerts when setting the high alert threshold.
Sensor accuracy
215
Appendix A

Guardian RT sensor performance and calibration stability as a function
of time
The Guardian RT sensor may be worn for up to 3 days (72 hours) and must be calibrated at least twice a
day. Two sets of data, approximately equal in number, were collected during the clinical trial. One data
set was generated when the frequency of calibrations averaged 3.5 per day (Data Set A), and the other
averaged 5 times a day (Data Set B). During the study, a total of 38 sensors were evaluated in 16
individuals.
As per the stratified Clarke Error Grid analysis above, agreement between Guardian RT values and YSI
values tends to be less at low and high glucose concentrations when compared to other concentration
ranges.
Guardian RT performance in the hypoglycemic range, as a function of sensor insertion time, is
characterized below. Results from the two different data sets are presented. The two populations were
separated according to the number of calibrations per day. This table represents the percentage of Data
Points in the 2.2 - 4.4 mmol/L (40-80 mg/dL) range that fell within 1.1 mmol/L (20 mg/dL). Data is
presented in 12-hour increments.
Data Set
0-12 hrs
12-24 hrs
24-36 hrs
36-48 hrs
48-60 hrs
60-72 hrs
A
78%
81%
73%
65%
56%
41%
B
67%
70%
93%
60%
75%
38%
Appendix A
216

An analysis of the mean percentage of Absolute Relative Error (ARE percentage) and standard deviations,
across 12-hour increments of wear periods, appears in the table below. Both data sets are pooled together
in this data.
Hours From Insertion
Mean ARE (%)
Std. Dev.
0-12 hrs
24.84
20.04
12-24 hrs
19.66
16.17
24-36 hrs
16.43
15.62
36-48 hrs
18.23
19.27
48-60 hrs
16.59
14.25
>60 hrs
22.95
23.51
The median sensor life from Data Sets A and B were 57.5 hours and 72.9 hours, respectively. Twenty-one of
the sensors operated for 72 hours, while the others were removed for a variety of reasons, most often
because of calibration errors.
Sensor accuracy
217
Appendix A

The percentage of Guardian RT readings within 20 percent and 30 percent of YSI readings was analyzed
according to time after sensor insertion and according to the glucose-concentration range (as determined
by the YSI analyzer). See the following table.
Percentage of Guardian RT values
within 20% of YSI laboratory readings
Percentage of Guardian RT values
within 30% of YSI laboratory readings
Glucose Range
(mg/dl)
During first 60
hours of sensor
wear
After 60 hours of
sensor wear
During first 60
hours of sensor
wear
After 60 hours of
sensor wear
40-80
a
62-82%
39%
62-82%
39%
81-120
57-66%
48%
72-84%
66%
a. Agreement to within ± 1.1 mmol/L (20 mg/dL) for glucose readings ≤ 4.4 mmol/L (80 mg/dL)
Performance of the Guardian RT was evaluated according to the length of time since calibration. This data
is not conclusive because of the limited number of data points during the final 3 hours of the 12-hour
calibration cycle, i.e., 10. In contrast, 3-hour time bins, earlier in the 12-hour cycle, contained hundreds
of data points. This may suggest that calibrations are often required prior to the 12-hour calibration cycle.
Appendix A
218

Effects of calibration frequency
The average bias when Guardian RT was calibrated ~ 3.5 times a day was -20.5 ± 41 mg/dl (LL: -22.40
mg/dl and UL: -18.63 mg/dl). In those calibrated ~ 5 times a day, the bias was -10.2 mg/dl ± 36 mg/dl (LL:
-11.74 mg/dl and UL: -8.66 mg/dl). When comparing Guardian RT units that were calibrated less often to
those calibrated more often, the following alarm performance was observed:
•
Specificity increased 2-4 percent in the hypoglycemic range and decreased 0-2 percent in the
hyperglycemic range
•
Sensitivity increased between 5-9 percent across the hyperglycemic range, and decreased 7-16
percent when the alarm was set to 80 mg/dl or below, and decreased 3-7 percent when set between
85 and 100 mg/dl
Stratified error grid analysis also shows better performance in the hypoglycemic range when fewer
calibrations are performed, i.e., 62 percent of data points are in Zone A when fewer calibrations were
performed, whereas 58 percent were in Zone A when more calibrations were performed.
Sensor accuracy
219