Head to Head Comparison

The Dexcom STS and the Paradigm RT continuous monitors are currently available in the U.S. with a prescription. In this study, they are compared head to head while being worn by one person with Type 1 diabetes. Over 33 days, 262 simultaneous readings were compared between a One Touch meter and the two continuous monitors. The meter was used to calibrate both monitors and as the standard against which their accuracy was evaluated.

Is one monitor better than the other? How close are the monitor readings to the Ultra fingerstick readings? Find out by viewing the comparison slideshow, downloading the Powerpoint presentation or reading below.

A Method For Comparison Of Two Continuous Monitors

by John Walsh, P.A., C.D.E.

Various approaches are available to measure the accuracy and utility of continuous monitors and to compare one monitor with another. The Clarke error grid, developed in 1987, has been widely used to verify the accuracy of fingerstick meters and was initally used also to validate continuous monitors. This grid compares a glucose value from a lab value, YSI instrument, or glucose meter with a simultaneous glucose value from a continuous monitor. Unfortunately, this comparison does not show glucose trend information, nor the timeliness that high and low glucose warnings provide.

Other attempts measure values like area under the curve (AUC) within glucose graphs and compare this to frequent standard glucose tests, done every 5 minutes or so. Some information is provided by comparing.

I bought a Dexcom STS sensor and have been wearing it ever since. When the local Medtronic representative asked if I would like to try the Paradigm RT pump/sensor combo, I jumped at the chance and have worn both sensors side by side for over three weeks now. This article covers my initial impressions of the pros, cons, and accuracy of each sensor. For comparison to my actual blood sugar readings and calibration, I used a One Touch Ultra meter properly coded for the strips in use. During this time, I have averaged just under 10 fingerstick tests a day (136 tests in the last 14 days) with an average reading of 130 mg/dl.

I usually leave my Dexcom sensors in place for 6 to 9 days which is longer than the company’s recommended 3 days. My experience shows that these sensors work this long. However, most of the Paradigm RT sensors were used for only 3 to 5 days because I had not tested them sufficiently to trust that they would perform longer than 3 days (Others have reported that they will.). In the first column of the spreadsheet below, the teal color shows when the Paradigm RT sensors were changed. The Dexcom sensors were changed less often as indicated by the light blue color in the second column. In this limited sample, length of use did not appear to affect accuracy in either sensor.

The Dexcom is slickly packaged and the Paradigm comes in its traditional pump packaging. The combination of a pump and sensor make the Medtronic more appealing in terms of one less device needed. The Dexcom currently has no connection with any of the current insulin pumps. Glucose results from the Paradigm RT are not used for any bolus or other dose calculations in the Paradigm pump, so the Dexcom is as useful whether combined with injections or a pump.

The con mon readings of individual blood sugars and the trend lines from the Paradigm RT appear on the Paradigm pump screen. This means only one device is needed on the belt or in the pocket. But this con mon data does not enter the pump’s data. Readings still have to be manually entered in the pump so the pump can recommend bolus doses.

Accuracy Of Readings And Trends

In general, today’s continuous monitors are better at helping the wearer see the trend in their blood sugar than in giving an accurate glucose value at a particular time. None of today’s monitors is capable of nor approved for relying on the readings for treatment decisions such as a correction bolus or eating carbs. In this test, numerous daily fingersticks were done to compare the two sensors to actual glucose values.

For testing, I induced both high and low readings to find out how each system responds to changes in glucose and to verify the speed and accuracy of their responses. As can be seen in my blood sugar spreadsheet on the right, neither sensor can yet be used to replace old-fashioned fingersticks, although the Dexcom comes close to achieving this.

For onscreen data, the Dexcom provides a larger screen display (1.19 versus 0.86 square inches for a pixel to pixel display) and a larger glucose graph display (1.38 versus 0.47 square inches). The Dexcom display screen with its white background also has more brightness and contrast than the green background of the Medtronic. The upside for the Medtronic is that it has a smaller profile (see photo) and weighs only one more ounce (roughly 4 oz. versus 3 oz.) with its built-in pump.

Because of the previous basal and bolus adjustments I made over several weeks while wearing the Dexcom, I set the alerts on each sensor to 80 mg/dl for lows and 160 mg/dl for highs to trigger low and high alerts without needing to induce too much excursion in my blood sugars. I set up the experiment as a head to head test to find out which sensor would alert first when my blood sugar was going high or low. I would then read the other sensor and compare each to the actual glucose at the time. At times, I would let my blood sugar continue to fall or rise until the other sensor gave an alert to see how much delay was involved, and to better estimate their accuracy at low and high readings.

Note for comparison purposes in the picture above that both sensors show horizontal lines for the selected targets of 80 and 160 mg/dl. The readings of 93 and 122 compare to an actual blood sugar of 73 mg/dl on the Ultra meter at the time of the picture. Note also how differently the same 3 hours of glucose values appear in the two sensors. For reference, an hour and 43 minutes earlier, about midway back on the graph, my blood sugar had risen to 290 mg/dl (actually 289 and 291 on two tests took less than a minute apart to verify accuracy), after starting at 125 mg/dl some 3 hours and 16 minutes before the picture was taken.

As in the picture above, the Dexcom turned out to be the hands-down winner for accuracy and trending. It proved to be much closer to actual blood sugars throughout the three weeks of testing. As can be seen in the spreadsheet, the Dexcom readings reflect the actual blood sugar more accurately with faster tracking of the speed and direction of blood sugar changes. It was not uncommon to have the Dexcom give a low or high alert 15 to 60 minutes before the Paradigm RT when the blood sugar was rising or falling.

In the detection of lows, the Dexcom STS won by a landslide. The STS was the first sensor to detect the blood sugar going below 80 mg/dl, as indicated by the red colored boxes, on 14 occasions compared to only one for the Paradigm RT. On three additional occasions, both sensors gave an alert at approximately the same time. When the blood sugar rose above 160 mg/dl, as indicated by the yellow-colored boxes, the Dexcom was the first to detect the change on 12 occasions compared to two occasions for the Paradigm RT. There were five additional occasions where both sensors gave an alert at about the same time.

Glucose errors are indicated by the purple boxes in the spreadsheet. I crudely estimated errors as any sensor reading that differed by more than 50 mg/dl from the One Touch Ultra reading. Again in a landslide win, the Dexcom made only 4 errors with 3 of these by one sensor, while the Paradigm RT committed 18 errors that were about evenly spread among 6 different sensors.

The Paradigm system would certainly be convenient in incorporating a continuous monitor readout into an insulin pump so one less device would need to be carried around. But in this limited sample of one, this convenience appears to be more than offset by the RT’s inaccuracy and its slow response to changing readings. After a few days of simultaneous use, I found I was automatically checking the Dexcom to estimate my current glucose and referring to the RT mostly out of curiosity. From these limited results, I cannot at this time recommend the RT system. I was, however, pleasantly surprised by the relative accuracy of the Dexcom STS sensor. (Spreadsheets are a wonderful thing — I, for one, would tend to miss the overall positive results after mentally cursing the occasional misses!)

As far as disclosure, I have no stock or direct financial ties to either company. I have one indirect connection to Dexcom in that I work part-time for North County Endocrine in Escondido, CA, where Dr. Timothy Bailey has done a limited amount of clinical trial work on the Dexcom sensor. I am listed as a “sub-investigator” in some of these studies, but I have an insufficient benefit and no bias as a result of this indirect connection. I paid full price out of my own pocket for my Dexcom continuous monitor and sensors, while the Paradigm RT pump and sensors were kindly loaned/donated to me by Medtronic. Medtronic’s 522/722 Paradigm pumps have been very reliable insulin delivery systems, but the current RT sensor appears to need additional work. Both companies, and others such as Abbot with their Navigator system, are making rapid progress in sensor technology and monitoring systems. Look for more head to head comparisons as soon as these improvements become available.