How Much Accuracy of Interstitial Glucose Measurement Is Enough? Is There a Need for New Evidence?

Abstract

This analytical comment discusses what standards are needed for the evaluation of the accuracy of glucose measurement systems continuously measuring glucose in the interstitial fluid. Since accuracy standards for continuous glucose monitoring (CGM)/flash glucose monitoring (FGM) systems are currently based on modeling studies or consensus of experts, we raised the question whether non-inferiority trials evaluating the safety and efficacy of CGM/FGM measurements compared to capillary blood glucose measurement with point-of-care devices could help to establish clarity about the needed accuracy standards of CGM/FGM. Such trials could also support the replacement of capillary blood glucose measurements by modern CGM/FGM systems.

Continuous glucose monitoring (CGM) and flash glucose monitoring (FGM) are novel methods for glucose control. After a glucose sensor is set in the subcutaneous tissue, the patient can get access to the current glucose level either dircetly or by scanning the current glucose value by a reader or a smartphone as often as the patient likes to do. In addition, CGM and FGM provide the course of glucose values over the past hours and allow a forecast of the future course of glucose by displaying trend-arrows. Some CGM/FGM systems have a nonadjunctive claim, which means that glucose values can be used for treatment decisions. Apart from occasional blood glucose measurements if symptoms of hypo- or hyperglycemia do not fit the current displayed glucose levels, or at extreme high or low glucose levels, there is little need for blood glucose measurements by finger sticking. Patients who are in need for multiple daily glucose measurements in order to adjust insulin doses or who want to know their current glucose level in certain situations (eg, before driving a car, or during physical exercise) will have a specific benefit from CGM/FGM systems. Therefore, CGM/FGM systems are gaining more popularity. Also, the reimbursement situation for CGM systems is improving.

With new technologies for measuring glucose in different compartments than blood, there comes the question about the accuracy of such measurements with regard to the validity of treatment decisions. Therefore, studies examining the accuracy of CGM/FGM are very welcome. In a Chinese multicenter study of Ji et al., the measurement performance of the FGM system FreeStyle Libre® was examined. This study found a mean absolute relative difference (MARD) of 10.0% to 10.7% between glucose values measured in the interstitial fluid (ISF) by a FGM device and the results of a reference blood glucose measurement performed by a YSI device. In addition, 99.9% of ISF-glucose–blood glucose pairs were within consensus error grid zones A and B. This multicenter study confirmed previous accuracy results by demonstrating slightly better accuracy results than the study of Bailey et al,¹ who observed a MARD of 11.4%. These accuracy results are encouraging for the FGM technology.

For interstitial glucose measurement systems like FGM or CGM, however, the question remains whether treatment decisions that are based on the obtained glucose measurements are valid enough to guarantee the safety of patients: How much accuracy is enough? All accuracy studies of interstitial systems are based on a comparison between glucose values obtained in the ISF and glucose values measured in the blood. For the comparison of glucose values measured in 2 different compartments (ISF and blood) there is currently no consented criterion for the evaluation of their agreement; like for example the ISO norm 15197-2013, which in Europe serves as a criterion for the evaluation of the accuracy of blood glucose meters.² This norm defines a minimum standard for accuracy and requires that 95% of all blood glucose meter measurements should deviate less than 15% from the reference methods; less than 15 mg/dl if blood glucose values are lower than 100 mg/dl, respectively. In the United States, the ISO norm 15197-2013is considered as not adequate to protect lay-users of blood glucose self-measurements, respectively protecting patients in prescription settings.^3-6

In addition, the ISO norm 15197-2013 refers only to glucose measurements within one compartment, the blood, and compares results of blood glucose meter measurements with results of a reference method. There is an agreement that this norm cannot be applied to the comparison of measurements in different compartments like glucose concentration in the blood and ISF. The main reason for this is the physiological difference between glucose concentrations in the blood and the ISF, at least during rapid changing glucose levels (eg, after a meal, during or after exercise, after the start of insulin action). Therefore, the MARD between blood glucose and interstitial glucose levels is not only due to possible measurement errors but also to this physiological difference of glucose levels in the two compartments. Furthermore, additional information on the past and future course of glucose levels achieved via continuous glucose measurements of CGM/FGM can be used for treatment decisions.⁷ Thus, this may compensate single deviations from blood glucose values that are derived from single spot measurements via finger sticks.

If the ISO norm 15197-2013 cannot be used, for good reason, for the assessment of the accuracy of CGM/FGM, the general question remains how to judge the accuracy data of these systems. This question becomes especially relevant considering the education of patients how to use the obtained information for treatment decisions. In the absence of clear standards for the evaluation of the accuracy of glucose measurements in the ISF, current decisions regarding the use of specific CGM or FGM values for treatment decision are based on modelling⁸ or consensus decisions of the regulatory bodies.⁹ Admittedly, this is the best way to justify decisions in the absence of clear empirical evidence. However, we should strive to go beyond the point of consensus- and modelling-based decisions and accumulate more evidence about the impact of the accuracy of CGM/FGM technology on treatment decisions and ultimately on diabetes outcomes.² This evidence could also provide a more definite answer to the question of how much accuracy is needed for continuous glucose measurements in the ISF.

Up to now, many evaluation studies about the efficacy of FGM and CGM systems have been published.^10-22 However, all these studies were designed as superiority trials, in which superiority of CGM or FGM was proven compared to self-monitoring of blood glucose. Superiority trails have the advantage that they are powered to minimize the statistical α-error; that is, avoiding to erroneously accept that a new treatment has superior efficacy. Thus, these trials are powered so that the sample size is just sufficient to detect clinically relevant effect sizes. The downside of superiority trails is that, due to the resulting relatively small sample size, rather rare adverse side effects are more unlikely to be detected. This is always a concern for the evaluation of the safety of new technology or new drugs. Therefore, safety data from superiority trials should be interpreted with caution. This problem could be addressed by noninferiority trials, which have the advantage that they are powered to minimize the statistical β-error; that is, avoiding to erroneously accept that a treatment is equal to another treatment. Therefore, non-inferiority trials require larger samples sizes. A non-inferiority trial, sufficiently powered to detect rare acute complications like ketoacidosis or severe hypoglycemia, could help to provide empirical evidence about the safety of interstitial glucose measurement and its use in clinical diabetes therapy. Such a trial should primarily focus on people with an intensive insulin treatment (using multiple daily insulin injection treatment or insulin pump therapy) since these patients are in need for multiple daily insulin dose adjustments and are more endangered to acute complications like severe hypoglycemia and ketoacidosis than other groups of diabetic patients.

Such a study would be more convincing when it is not carried out by just one manufacturer alone, but rather include all available CGM/FGM systems with an acceptable measurement performance from different manufacturers. A non-inferiority trial on safety and efficacy would have two main benefits: First, the safety concerns about the use of ISF glucose measurements could be settled once and for all. Second, such a study could establish ISF measurements by CGM/FGM systems as a sort of gold standard for people with diabetes on an intensive insulin treatment.