Perceived Accuracy in Continuous Glucose Monitoring

When considering the subject of device accuracy from the patient’s perspective, we are—at a fundamental level—talking about trust. Are the devices I am using telling me the truth? What about the bathroom scale that I stand on every morning? The home blood pressure monitor that I recently purchased? And what about the numbers I see on my blood glucose meter? If these devices are not perceived as trustworthy, if the patient cannot have faith in the numbers that he sees, will he feel comfortable taking action in response to what he sees? Indeed, will he be likely to continue to use the device (or devices) at all?

The common belief is that the perception of accuracy in point-of-care devices is a key variable that will influence how such devices are accepted and used by the patient. As the accuracy of devices improves, or as patient satisfaction with device accuracy improves, it might be expected that patient acceptance might grow. To date, however, no studies have directly examined the impact of the perceived accuracy of point-of-care devices on patient attitudes toward the device or patient use of the device. While broad theories, such as the technology acceptance model, include variables related to perceived accuracy—such as perceived usefulness and performance expectancy—most studies that have made use of such theories have focused on the acceptance of devices and broader health IT by health care professionals, not patients. ¹ Shaw and colleagues ² found that patients’ stated intention to begin using a portable coagulometer device was greater when, not surprisingly, patients endorsed a series of survey items indicating a more positive perception of the technology. This series included items that included patient’s perception of device accuracy, but these items were not examined separately.

In diabetes, the data are similarly limited. While many studies have examined patient attitudes toward self-monitoring of blood glucose (SMBG), we have uncovered none that have explored perceptions related to SMBG accuracy. Since SMBG is typically conducted relatively infrequently, it is likely that most patients do not notice fluctuations in blood glucose results that would cause them to question the accuracy of their monitors. Indeed, it is rare for patients to directly compare SMBG results with different meters or to perform multiple SMBG measurements over very short periods. Therefore, it is our clinical impression that most patients are unaware that current blood glucose meters may be producing data that are plus or minus 20% of the actual value.

In contrast, real-time continuous glucose monitoring (RT-CGM) produces so many values in relatively short periods of time that it is likely to be more apparent to patients when problems in accuracy exist. Also, regular calibration with SMBG readings means that patients are repeatedly reminded of potential differences between the 2 sets of results. Patients use RT-CGM data to help them determine whether they will be safe from harm when, for example, exercising, driving or sleeping. Therefore, the personal value placed on device accuracy, whether or not the device can be trusted, may be viewed as more critical than for other point-of-care devices, such as home blood pressure monitors. In the years since RT-CGM was introduced, concerns about accuracy (and the mixed feelings that result from these) have been a commonly stated frustration. As an illustration, consider this Internet post from 2006: “The CGM . . . is a buggy, first-gen product that takes a lot of work and is often inaccurate. . . . Having said that, . . . I would rather go back to 18th century bloodletting than try to treat my diabetes without it.” ³ Though the technology has advanced considerably, recent studies continue to highlight the importance of this issue. Ramchandani et al ⁴ found that sensor inaccuracy was a key reason for quitting RT-CGM in the 14 families they surveyed. In a larger survey, Chamberlain and colleagues ⁵ found that problems with sensor accuracy was the most commonly reported reason for infrequent RT-CGM use. In the largest study to date, Polonsky and Hessler ⁶ surveyed 877 current RT-CGM users (all using the Dexcom Seven Plus; DSP) and determined that satisfaction with device accuracy was an independent predictor of the 3 major quality of life (QOL) benefits associated with RT-CGM use: greater perceived control over diabetes (β = .19, P < .001), greater hypoglycemic safety (β = .19, P < .001), and better interpersonal support (β = .07, P < .05). Furthermore, unpublished data from the same data set indicate that greater satisfaction with accuracy was also associated with more frequent RT-CGM use (β = .08, P < .05) and with seeing the device as more helpful in avoiding hypoglycemia (β = .29, P < .001) and hyperglycemia (β = .32, P < .001), and in achieving better overall glycemic control (β = .24, P < .001).

Additional data from this study examined survey responses from ex-users of the DSP (n =102); compared with current DSP users, ex-users reported significantly less satisfaction with the device’s accuracy (P < .001). While 83% of current users reported art least moderate satisfaction with device accuracy, the comparable number for ex-users was only 51%. When ex-users were asked about their major reasons for quitting RT-CGM, 34% cited “the numbers couldn’t be trusted” as a major reason, 26% cited “too many false alarms,” and 22% endorsed “too often the device stopped working.” ⁷ In total, these data point to the critical importance of device accuracy, especially as it pertains to RT-CGM, especially since frequency of RT-CGM usage has been shown to be a key and consistent predictor of glycemic improvement across studies. ⁸ Patients are more likely to use RT-CGM when they believe that the device and the resulting data can be trusted and, in a related vein, when they are not being subjected to frequent alarms that turn out to be false. Alarm fatigue is now a well-documented phenomenon,^9-11 and we suspect that such complaints are reduced, though not likely to be eliminated, when patients believe that their alarms accurately reflect worrisome, out-of-range glucose levels.

In addition to more frequent usage, the perception of greater accuracy may also influence how the device is used. First, many patients may become more active users of the information provided. Following from the categories put forward by Joubert and Reznik, ¹² greater confidence in the accuracy of RT-CGM data may lead those patients who are more “reactive” users (responding only to alerts and alarms) to engage more frequently with the device and become “proactive” users, wielding the numerical data and trend data to anticipate glucose changes and respond early and aggressively. In a recent survey of 222 current RT-CGM users, Edelman and colleagues ¹³ found that the majority had increased their frequency of insulin boluses since beginning RT-CGM and were typically making changes in their insulin dosages that were more variable and aggressive than previous recommendations have suggested. ¹⁴ Second, these frequent adjustments often occurred without a confirmatory blood glucose test, thus hinting at a sense of trust and confidence in the RT-CGM data. Indeed, in an early trial of the FreeStyle Navigator, SMBG frequency fell precipitously over the 3-month study, from a mean of 4.9 tests/day at study initiation to 2.6 tests/day at study end, due—perhaps—to a growing comfort with the RT-CGM data (note that all 23 subjects elected to continue using the device after the active trial concluded). ¹⁵

It is also noteworthy that patients may perceive RT-CGM as more accurate and reliable when they are given support and guidance regarding how to use the data more effectively. This may explain, at least partially, the reported QOL benefits seen in the trial by Jenkins and colleagues, ¹⁶ where a patient-centered algorithm helped to guide RT-CGM responses. As seen in previous studies of SMBG, patients’ interest and willingness to check blood glucose levels is markedly elevated when they perceive that the data can be understood and used effectively.^17-18 Similarly, RT-CGM benefits, including patients’ enthusiasm to continue using the device as well as QOL benefits, are likely to depend on this sense of trust and efficacy—that the device is producing data that are precise and usable.

The good news is that the current generation of RT-CGM devices is now even more accurate and reliable than previously, though there still appear to be sizeable differences in accuracy between the current manufacturers. ¹⁹ Indeed, the general lack of QOL benefits seen in most RT-CGM trials to date ²⁰ may have stemmed in part from substantial differences in patients’ trust in the devices that have been provided. Satisfaction, or dissatisfaction, with accuracy—possibly resulting from differences in the devices chosen for use—may also contribute to the noted differences in how often patients have chosen to use RT-CGM and/or whether discontinuation has occurred.^5,14

In future studies, it may be worthwhile to determine whether differential responses to the currently available devices, especially in how patients may feel differently about the accuracy and reliability of their device, may allow researchers to identify behavioral and QOL benefits that have not been previously determinable. Note that the emphasis throughout this commentary has been on the perception of device accuracy and the associated sense of satisfaction with accuracy, not accuracy per se. The key is whether patients’ personal expectations for accuracy are being met and whether they feel that the technology is accurate enough to meet their needs. This points to the need for critical RT-CGM training, where patients can learn about the physiological differences between SMBG and RT-CGM, why perfect agreement between the 2 should not be expected, and how best to make use of glucose data that are never flawless. Of course, devices features other than perceived accuracy contribute to patients’ overall sense of RT-CGM satisfaction, including pain, cost of the device and—as mentioned above—alarm fatigue, but the perception of accuracy and the overall sense of trust in the device deserve additional attention as RT-CGM studies and technology move forward.

In conclusion, RT-CGM provides potential benefits relative to episodic SMBG via alerts and alarms, by providing the direction and rate of glucose change, and by detecting actual patterns of high and low glucose that occur over time. As the technology continues to mature, RT-CGM will become increasingly accurate and patients’ confidence and trust in the available devices will likely grow, leading to greater uptake and more proactive use of RT-CGM. Further studies are needed to confirm whether and how perceived accuracy leads to greater RT-CGM adherence, decreased alarm fatigue, reductions in SMBG, more confident and aggressive insulin adjustments based on the predictive trend data, and improvements in glycemic as well as QOL outcomes.