Head-To-Head Comparison Of The Two Currently Availabl

1
Head-To-Head Comparison Of The Two Currently
Available Continuous Monitors
John Walsh, P.A., C.D.E., and Ruth Roberts, M.A.
January, 2007

• North County Endocrine
• 700 West El Norte Pkwy
• Escondido, CA 92126
• [phone]
• The Diabetes Mall
• www diabetesnet.com
• [phone] jwalsh_at_diabetesnet.com

2
Dexcom STS Monitor
FDA release 3/27/06 Approved for 18 and
older Readings every 5 min. 3-14 days of
readings per sensor One high and two low
alerts 800 (often less) 35 per 3 day
sensor Transmitter replacement 250 every 6 mos
Transmitter 0.8 x 1.5
3
Medtronic Paradigm RT
Trend arrows Readings every 5 min. 999 35
per 3 day sensor Transmitter replacement
900 every 6 mos
FDA release 4/13/06 Paradigm 522/722 pump
Sensor connected to transmitter by wire Approved
for 18 and older One high and one low alert
NOT a closed loop
4
Continuous Monitors Compared
The Dexcom STS Paradigm RT continuous monitors
are currently available in the U.S. This study
was designed as a head to head comparison while
being worn by one person with Type 1 diabetes. An
Ultra meter was used as the standard against
which each monitors accuracy was tested. Each
monitors screen is shown above over the same 3
hour time span over a glucose range of 0 to 400
mg/dl. The Ultra meter reading at the time read
73 mg/dl compared to 93 and 122 mg/dl.
5
Research Design

• The low alert in each monitor was set at 80 mg/dl
  and the high alert at 160 mg/dl
• 262 readings from an Ultra meter were performed
  over 33 days (7.94 tests per day)
• Simultaneous readings from each monitor were
  compared to the Ultras readings
• Testing with the Ultra meter was performed
• As soon as either monitors low or high alert
  sounded
• When values between the monitors disagreed
• And at routine intervals, including calibrations

6
Calibrations

• Each monitor required 2 calibrations with an
  Ultra glucose meter each day to reduce drift or
  loss of accuracy over time.
• Calibrations were generally done when the glucose
  level was stable.
• Calibrations were usually done on both monitors
  when requested by one of them.
• In all, 3 to 4 calibrations were done per day on
  both monitors.
• Care was taken not to do more than 4 calibrations
  per day on either monitor due to the
  recommendation by one of the manufactures that no
  more than 4 be done per day. (This recommendation
  makes no sense to us, though we have discussed it
  with various engineering and sales personnel from
  the company.)
• Accuracy of the other monitor appears to improve
  when more calibrations are done from observations
  before and after the study.

7
Readings Sorted By Ultra BG Values
Sorted By Consecutive Readings Minor note
Sensor B always gives results in even numbers for
reasons we do not understand. However, an odd
number would sometimes result when a second
reading was displayed before the Ultra test was
complete and the two readings were averaged.
Red first to detect a low Yellow first to
detect a high Purple over 50 mg/dl off
8
Readings Sorted By Ultra BG Values
Low to High High to Low
Red first to detect a low Yellow first to
detect a high Purple over 50 mg/dl off
9
Plotted Readings With Trend Lines
A
Ultra
Monitor A shows linearity from 50 mg/dl up to at
least 240 mg/dl. Monitor B shows more
scattered values and flattens out with higher
low readings and lower high readings than the
Ultra.
B
Ultra
10
GlycensitTM Analysis
B
A

• Blue dotted lines ISO standards
• Yellow area where 95 of data points will fall
• Red lines minimum and maximum deviation by star
  points
• Ideally, all readings would fall between the blue
  dotted lines -- this is the standard for todays
  glucose meters

http//tomcatbackup.esat.kuleuven.be/GLYCENSIT/
11
Monitor Accuracy
44.3 31.7 12.2 4.2 3.4 1.5 0.8 1.9
19.8 23.7 17.6 13.7 10.3 5.3 2.7 6.9
76.0 19.8 4.2
43.5 41.6 14.9

• Table shows how much each monitors glucose
  values differed from a simultaneous reading on
  the Ultra meter.

12
Monitor Accuracy
Monitor A read slightly high (5) for readings
below 80, while Bs readings were 34 high for
this range. Between 81 and 160, both sensors
were within 3 of the Ulta. Between 161 and 240,
A averaged 3 lower than the Ultra, while B
averaged 17 lower. Above 240, there are too few
readings to judge accuracy, but both monitors
appear to read lower, although the difference was
more pronounced for Monitor B.
13
Monitor Accuracy
B
A
Lower than the Ultra
Higher than the Ultra
BGs lt 80
BGs 80-160
BGs gt 160
BGs lt 80
BGs 80-160
BGs gt 160
This GlycensitTM graphic summarizes data shown on
the previous slide. The average reading for
Monitor A, shown by the middle of the
spool-shaped objects for low, middle, and high,
is much closer to that of the Ultra meter (the
green line in the middle) with less variation
(the height of the spool the standard deviation)
than Monitor B. The shorter the spool and the
closer its middle is to the line the more
accurate the Monitor.
http//tomcatbackup.esat.kuleuven.be/GLYCENSIT/
14
Which Monitor Alerted First?

• One of the most important functions of a
  continuous monitor is to accurately warn of low
  and high readings. This table shows which monitor
  alerted at least 5 min earlier for lows and
  highs.
• Monitor A was first to alert for a reading below
  80 mg/dl 76 of the time, Monitor B was first 3
  of the time, and 21 were ties.
• For detection of highs above 160, Monitor A was
  first 68 of the time, Monitor B was first 5 of
  the time, and 27 were ties.

15
Time Advantge For First Alerts
Time not recorded for 2 of these events Time
not recorded for 4 of these
16
Time Delays For Low And High Alerts

• In situations where Monitor A was first to
  alert, the time delays for Monitor B to give an
  alert are shown.
• confirmed with the Ultra meter
• Readings were maintained below 80 mg/dl for
  long periods to test these delays.

17
Problems Needing Answers To Close The Loop
Glucose
Insulin

• Sensor inaccuracy
• Not reliable
• Slow insulin
• Lifestyle variability
• No control algorithm
• Sensor lag time
• Insulin overdose

Engineering / Materials Internal Checking Viaject
/ Oral insulin Rapid Prediction Software /
Testing Engineering Dual delivery / Glucagon
18
Findings

• Neither continuous monitor has any hope for use
  in a closed loop system at this time, especially
  Monitor B.
• In our single case study, Monitor A when compared
  with Monitor B was noticeably
• More accurate
• Quicker to detect changes in glucose
• Quicker to warn of low and high glucose readings
• Monitor B was slower to respond to changes in
  glucose and tended to have flat readings that
  were more likely to read high for lows and low
  for highs.
• This study involved only one individual, so each
  monitors may perform differently in others.
• Additional head-to-head comparisons of different
  continuous monitors are recommended.

19
Addendum

• To eliminate a possible equipment problem as the
  cause for poor accuracy with Monitor B, two
  different transmitters and receivers were used
  over two different periods. The first combo was
  used for 19 days in late September and early
  October of 2006, and the second for 14 days in
  December, 2006. This change of equipment did not
  improve accuracy.
• Sensors for both monitors have been shown to
  maintain accuracy when used longer than 3 days.
  Over the 33 days of the study, 6 different
  sensors were used for Monitor A and 9 sensors
  were used for Monitor B. This had no impact on
  accuracy.
• Two sensors (one for each monitor) were replaced
  within a few hours of each other in late
  September after 4 bad readings (over 50 mg/dl
  off) were received from each. These readings were
  also included with no real effect on overall
  accuracy.