On December 20, 2016, the US Food and Drug Administration (FDA) approved labeling for use of the Dexcom G5 Mobile continuous glucose monitor (CGM) for making diabetes treatment decisions without a need for confirmation of the device’s readings with fingerstick or laboratory testing (ie, nonadjunctively).1 The FDA relabeling allowed the Centers for Medicare & Medicaid Services (CMS) to reclassify the device 3 weeks later as durable medical equipment able to replace blood glucose meters and therefore eligible for reimbursement.2 The device measures interstitial fluid glucose concentrations to estimate blood glucose concentrations. To mitigate the acknowledged risk from inaccurate estimates of blood glucose values, labeling for the device previously required a confirmation with a blood glucose meter whenever the device was used for treatment decisions, including the dosing of insulin.3 This requirement has been removed.
Continuous glucose monitors have proven to be effective in improving glycemic control.4 These devices provide patients with timely alerts about impending high or low blood glucose levels, with information about the direction and speed of trends in glucose levels, and with information about an individual’s response to food and exercise. However, the recent FDA decision has potential risk for patients with diabetes who might now rely only on CGM readings for dosing insulin without confirmation of blood glucose levels.
As of October 1, 2016, more than 25 000 medical device reports indicating large inaccuracies with blood glucose levels measured with the CGM device had been filed with the FDA since the start of 2015,5 shortly after FDA approval of the most recent software and sensor combination currently used by the Dexcom G5. By June 30, 2017, the number of reports of inaccuracy had increased to more than 40 000. Even though the average error of readings had steadily improved through successive generations of the device, these reports suggest that sporadic large errors still occur. Many of the reports lack detail, and approximately one-third of the concerns about inaccurate readings were attributed by the manufacturer to user error. Nonetheless, there have been detailed reports of unexplained inaccurate readings associated with loss of consciousness, seizures, car crashes, hospitalizations, intensive care unit stays, and deaths. Individually, many of the concerns raised in the reports might be regarded as unique, incomplete, or inconclusive; however, taken together, these reports represent a safety signal requiring investigation (100 examples of reports of Dexcom CGM inaccuracy appear in the Supplement).
As part of the application submitted to the FDA for relabeling of the device, no new clinical evidence was introduced to address the reports of inaccuracy from numerous users. The only clinical data considered regarding the accuracy of the device were the same data that had been used originally to gain approval for using the device for diabetes treatment decisions only with verification. The data on accuracy were based on only 44 adult patients with type 1 diabetes, 7 with type 2 diabetes, and 79 pediatric patients with type 1 diabetes.6,7 Estimates of device accuracy using such small numbers of patients observed in an in-clinic setting can be overly optimistic. For example, data from the pivotal study6 for adult users, showed that when the CGM reading was greater than 80 mg/dL, there were no occurrrences in which the true blood glucose was below 60 mg/dL.3 In contrast, the Supplement includes 20 reports in which CGM values of 115 mg/dL or greater masked hypoglycemic blood glucose readings of 45 mg/dL or less that were corroborated by medical personnel, the altered consciousness of the patient, or both.
An FDA advisory panel was convened in July 2016 to provide recommendations. The panel was not provided information about the large number of reports of device inaccuracy. Instead, panelists (1) heard testimonials for the device from some current users and physicians during a public hearing; (2) were presented with 2 sponsor-originated computer simulations ultimately determined unconvincing by FDA reviewers; and (3) were presented with the original clinical data. The panel also received a petition signed by users who had been told that labeling the device for nonadjunctive use was necessary to permit the CMS to provide reimbursement coverage, and would hasten the development of an artificial pancreas device. The panel voted 8-2 to approve the label changes.8
Lacking new clinical data, the FDA decision was based on (1) the advisory panel vote; (2) the testimonials, public comments to the docket, petition, and the original limited data used by the advisory panel; (3) the intent for postmarket monitoring; and (4) the expectation that a new label could mitigate risk.1
The Dexcom G5 user guide, for example, provides prominently displayed guidance to not use the CGM readings “if they do not match your symptoms or expectations.”8,9 However, it is the roughly 25% of patients with diabetes with impaired glycemic awareness for whom CGMs have been considered the most useful.3 The advice also does not serve to mitigate risk for very young patients or their parents who need to rely on CGM values, not symptoms, to remotely monitor young children.
Another proposed safeguard was to caution users “to develop proficiency with the use of the system and confidence in how the system performs for them prior to starting to use the information from the device to make diabetes treatment decisions.”1 Users should “consult with their healthcare provider regarding how they can best use the device to manage their diabetes.”1 This is sound advice, but in calculating an insulin dose, in some circumstances, no amount of proficiency, confidence, or diabetes caregiver advice would be able to overcome an inaccurate CGM reading.
Above all, a measuring instrument needs to be reliably accurate, particularly if used for decisions on administering a drug with a narrow therapeutic window such as insulin. Without citing supporting data, the FDA proposed to mitigate risk with a label change urging users to “become familiar with the potential for variability in device accuracy between sensors.”1 To date, there is little documentation on such variability, although it could be important for the issue of sporadic inaccuracy.
The FDA summary of safety and effectiveness data noted that the approval would allow users to “access other benefits of this device (eg, real-time alerts/alarms, trend information, continuous passive sharing of real-time glucose information, etc) which are not available in currently marketed glucose meters used for traditional blood glucose monitoring.”1 But the question facing the FDA was whether a sporadically inaccurate device should be approved as safe for use in insulin dosing without confirmation. Continuous glucose monitors had already been approved for more than 15 years.
After reviewing the benefits and listing the risks of using the G5 Mobile as a stand-alone device, the FDA summary of safety and effectiveness data concluded “the benefits of additional information gained from this device outweigh the risk of inaccurate results, rates of change, and false negative and positive alarms and alerts.”1 There was no explanation of how the risks of potential adverse effects (eg, episodes of loss of consciousness, seizures, and emergency resuscitations) in a small percentage of patients had been balanced against the cited benefits of the device.
The FDA had other options for furthering the use of CGMs besides declaring that the performance of CGMs “is now adequate to support safe and effective use of the device for replacement of fingerstick testing.”1 There was the option of temporarily withholding approval until the causes of the inaccurate values had been identified so they might be prevented, predicted, or mitigated. The FDA also could have made the information about potential inaccuracies more readily available. Now that there is Medicare coverage for at least 1 CGM, many more physicians and patients will be exploring ways to use these devices safely; they too need to be provided with complete information about device performance.
Another alternative would have been to move beyond an adjunctive-nonadjunctive dichotomy. Instead of approving labeling of the device as “able to replace blood glucose meters for diabetes treatment decisions,”1 finer distinctions regarding the type of decisions could have been made. For example, the device might be acceptably safe to use by some patients for some diabetes treatment decisions (such as carbohydrate ingestion, exercise, meal timing, and sleep), even if it is not yet demonstrated to be reliably accurate for dosing insulin.
Patients with diabetes are heterogeneous. They differ in their susceptibility to hypoglycemia; the frequency, speed, and extent of glycemic excursions; their counter-regulatory responsiveness; and their degree of glycemic awareness. The extent to which a CGM susceptible to intermittent accuracy can safely replace, instead of complement, blood glucose meters for insulin dosing is likely to differ for each patient. In particular, the number of confirmatory blood glucose readings needed to use a CGM safely will vary between individuals. The CMS and other payers will need to recognize the need for more than just 2 calibration test strips per day for many of the patients using CGMs for making diabetes decisions.
Devices such as CGMs can be essential, even life-saving tools for persons with diabetes. However, to make a useful class III device such as the Dexcom G5 CGM more accessible to the public, the FDA determined that necessary safety procedures (in particular, the confirmation of device readings before treatment interventions) were no longer indicated. This decision lacked adequate nuance and evidentiary support and may have created a potentially hazardous situation for some patients who use these devices.
Corresponding Author: Alan R. Shapiro, MD, MSPH, Department of Medicine, New York University School of Medicine, 5 Pheasant Run Rd, Pleasantville, NY 10570 (firstname.lastname@example.org)
Published Online: September 14, 2017. doi:10.1001/jama.2017.12075
Conflict of Interest Disclosures: The author has completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Additional Contributions: I thank Marion Greif, EdM, for providing valuable editorial assistance. Ms Greif was not compensated.
Complaints of Dexcom CGM Inaccuracy
Shapiro AR. FDA Approval of Nonadjunctive Use of Continuous Glucose Monitors for Insulin DosingA Potentially Risky Decision. JAMA. Published online September 14, 2017. doi:10.1001/jama.2017.12075