Continuous Glucose Monitoring Use in Clinical Trials for On-Market Diabetes Drugs

A list of clinical trials investigating the pharmacokinetics, pharmacodynamics, safety, tolerability, and efficacy of 40 diabetes drugs currently on the U.S. market (Supplementary Appendix A) was compiled using the National Institutes of Health ClinicalTrials.gov database. Because we intentionally limited our search to trials of 40 major drugs, our approach was not intended to be representative of all clinical trials; rather, we aimed to provide broad information on the state of CGM use in diabetes clinical trials during the past 20 years. Use of CGM was determined based on the study record.

We limited our search with the following constraints: 1) date—trial start date from 1 January 2000 to 31 December 2019, 2) recruitment—completed; and 3) condition/disease—diabetes. The date restriction was chosen based on U.S. Food and Drug Administration (FDA) approval of the first CGM system on 15 June 1999. We excluded any trials for which the primary outcome evaluated diabetes technology to avoid including trials investigating CGM itself with a pre- or off-market therapy used as a comparator. We also excluded trials for which interventions did not manipulate subjects’ glycemia and outcomes did not evaluate drug pharmacokinetics, pharmacodynamics, safety, tolerability, or efficacy, eliminating trials for which inter- and intraday glycemic metrics would provide little value.

Using these criteria, we identified 2,032 clinical trials (Supplementary Appendix B). We plotted the proportion of trials using CGM over time and performed subgroup analyses of eight major drug classes plus pramlintide. These drug classes included basal insulins, rapid-acting insulins, glucagon-like peptide 1 (GLP-1) receptor agonists, sodium–glucose cotransporter 2 (SGLT2) inhibitors, dipeptidyl peptidease-4 (DPP-4) inhibitors, sulfonylureas, metformin, and thiazolidinediones (TZDs). Trials investigating basal insulin/GLP-1 receptor agonist fixed ratio combination therapy were included in the whole-field analysis but were not analyzed separately given low trial count (n = 20).

From our ClinicalTrials.gov query, we identified 2,032 clinical trials of 40 major diabetes drugs on the U.S. market, not including basal insulin/GLP-1 receptor agonist fixed ratio combination therapy. Precisely 119 (5.9%) of these trials used CGM. By drug class, CGM was used in 8.0% of basal insulin trials (55/683), 7.8% of rapid-acting insulin trials (38/486), 3.6% of GLP-1 receptor agonist trials (11/305), 4.1% of SGLT2 inhibitor trials (6/146), 1.4% of DPP-4 inhibitor trials (4/282), 8.0% of sulfonylurea trials (21/263), 3.7% of metformin trials (15/408), and 20.5% of pramlintide trials (8/39). None of the 196 TZD trials used CGM. A total of 737 trials used multiple drug classes; these were included only once in the whole-field analysis. Results are summarized in Table 2.

Clinical trials investigating diabetes drugs have increasingly used CGM as a study tool (Figure 1). The greatest proportion of trials with CGM occurred in 2018, when 13.2% of clinical trials (n = 38) included CGM as a part of the study protocol. Throughout a steady increase starting in 2004, when 2.1% of clinical trials (n = 95) included CGM, 2 years—2018 and 2019—had >10% CGM use in trials.

CGM use over time in trials for seven individual drug classes are depicted in Figure 2. By drug class, pramlintide clinical trials exhibited the sharpest upward trend in use of CGM. However, the number of pramlintide clinical trials (n = 39) was far lower than that of all other drug classes. There was only one pramlintide trial in 2013 and one in 2015, and both used CGM. Next to pramlintide, SGLT2 trials (Figure 2D) saw the highest proportion of CGM use in 2018, with 33.3% of trials using CGM. GLP-1 receptor agonists (Figure 2C) were the only analyzed class that exhibited a downward trend in CGM use. The class saw its peak CGM use in 2014, when 8.0% of clinical trials (2/25 total) used CGM.

Developments in CGM technology in the past 20 years have allowed for the inclusion of glycemic outcomes beyond A1C both in the clinic and in research, affording patients, providers, and industry a far more comprehensive view of glycemic patterns. Standardizing CGM-derived metrics and encouraging their use has been a major focus in the diabetes field during the past several years. As far as we know, no broad analysis has been conducted to review the overall implementation of CGM in pharmaceutical trials to date.

For this analysis, we looked at 2,032 studies of major diabetes drugs that were initiated after the introduction of CGM technology in 1999 to assess the degree of CGM use in clinical trials. Of the 2,032 studies, only 5.9% used CGM, although usage has increased over time from <5% before 2005 to >6% after 2011 (Figure 1), presumably as a result of increased accuracy, greater ease of use, smaller size, and decreased cost of CGM systems (12). The proportion of trials deploying CGM continues to rise, reaching >12% in 2018 and 2019.

The factors preventing greater adoption of CGM technology in trials have been well documented and include hesitation of some regulatory agencies to fully accept CGM-derived metrics as qualified end points and include CGM data in labeling. Other factors limiting widespread adoption of CGM technology include continuing education necessary for handling of devices and management of resulting data, potentially increased patient burden, and the need for consistent calibration for some of the older devices, although the latter is less of a problem with newer devices, given the availability of factory-calibrated sensors (2,19). Additional barriers include cost (for both supplies and training time) and payer unwillingness to incorporate CGM data into their policy decisions.

The most recent draft guidance from the FDA Center for Drug Evaluation and Research for glycemic outcomes in diabetes trials (released in 2008), which was withdrawn in 2020, describes A1C as “appropriate as a surrogate end point in many study designs” (20). The FDA’s Center for Devices and Radiological Health has been more open to inclusion of CGM data in labeling. The European Medicines Agency recently published draft guidance endorsing CGM and patient-reported outcomes linked to CGM data in clinical investigations (21). However, none of these bodies will approve a drug or device based solely on improvement in a CGM-derived metric. Thus, there is less incentive to include CGM in clinical trials of antihyperglycemic drugs, despite an emerging consensus in support of CGM in the research and patient communities (22).

Some FDA officials have raised valid concerns surrounding the clinical relevance of CGM metrics. Currently, there exists no evidence for the long-term benefits of a given percentage of TIR that is not confounded by simultaneous A1C improvements. However, conducting a long-term randomized controlled trial to address this gap in research would be neither feasible nor ethical, representing a major barrier in convincing regulators and payers of the value of both continuous and professional CGM. However, although there has been no study indicating causality, a growing body of research suggests that a lower proportion of TIR strongly correlates with higher A1C and is associated with greater risk of complications of diabetes, including retinopathy and microalbuminuria (17,18).

We are hopeful that industry, academia, and funding bodies will invest in studies to further validate CGM metrics as predictors of long-term health and quality of life in the interest of ultimately encouraging payer and regulator acceptance. In 2019, an international body of diabetes clinicians and academics gathered to determine targets for TIR, time below range, and time above range. For most people with type 1 or type 2 diabetes, the committee recommended >70% of time spent in range (70–180 mg/dL), <25% of time spent above range (>180 mg/dL), and <4% of time spent below range (<70 mg/dL) (23). Tailored recommendations for older and high-risk patients, as well as women during pregnancy and/or with gestational diabetes, were provided separately. The committee added that every 5% improvement in TIR is considered clinically meaningful. These recommendations may help to increase recognition of the value of CGM-derived metrics and ultimately drive greater use of CGM in clinical trials and practice.

This analysis includes several limitations. Primarily, there exists a lag between when trials begin and when they are posted on ClinicalTrials.gov, likely deflating the numbers of trials recorded for the most recent years. Only 38 trials were posted as beginning in 2018 and just 24 in 2019, which could result in overestimation or underestimation of the proportion of clinical trials using CGM in these later years. Further, this analysis only included trials that were indicated as “completed” on Clinicaltrials.gov. There is the possibility that more diabetes drug trials using CGM have started since 2000 but have not yet been completed or marked as such on the website. The study is also limited in that we included all clinical trials for the indicated drug classes beginning in 2000. Many of these trials’ purposes are not relevant to CGM. For example, trials investigating the pharmacokinetics of a new therapy would not directly benefit from CGM, although continuous glucose data could be useful as a secondary end point. Additionally, some trial sponsors may use CGM in studies for internal purposes but not list these metrics on ClinicalTrials.gov. Finally, our data set only included studies of medications that are on the market and so excluded trials carried out as part of the clinical development of newer, pre-market therapies, which may be expected to use CGM more readily.

Given that consensus statements on definitions for CGM-derived metrics were published in December 2017, we might expect to see an uptick in CGM use in 2020 and beyond. Future studies may examine whether the consensus statements had a positive impact on use of CGM in clinical trials, based on both subsequent ClinicalTrials.gov and published literature analyses. Further subgroup analysis of the trials that do use CGM could offer additional insight into how this technology is most commonly used in a clinical trial setting.

CGM-derived outcomes such as TIR are increasingly valued among patients and clinicians. However, clinical trials for diabetes pharmacotherapies have been unlikely to include CGM because the value of CGM-derived outcomes has not yet been fully recognized by regulators (i.e., the FDA) or payers (i.e., insurance companies) as an indicator of safety or effectiveness. Of the 2,032 trials that we analyzed from ClinicalTrials.gov, 119 (5.9%) used CGM. CGM use in trials has increased over time, peaking in 2018 with 13.2% of trials using CGM. Given the documented limitations of A1C, the proportion of clinical trials using CGM in recent years is lower than we would have expected. We maintain that CGM should be used in all diabetes clinical trials that have glycemic end points and are hopeful that the recent consensus statements describing targets for CGM-derived outcomes will be useful in encouraging recognition of the value of CGM metrics in clinical trials. We hope that the data from our analysis will serve as a benchmark from which continued growth in CGM use in clinical trials can be measured over time.

The authors thank Jim Hirsch, Martin Kurian, and Dr. Daniel Drucker for advice on data analysis, as well as Dr. Charles Alexander for critical advice on manuscript structure and content.

Close Concerns, a diabetes information and consulting company, funded this research.

All authors are employees or former employees of Close Concerns, which publishes a paid information service to which numerous diabetes technology, therapy, and digital health companies subscribe. A full listing is available online (https://www.closeconcerns.com/disclosure.php). K.L.C. is chair of a nonprofit organization that receives funding from many of these same companies. A full listing is available online (https://diatribe.org/editorial-policy). No other potential conflicts of interest related to this article were reported.

B.Q.F. and P.F.B. researched data, performed data analysis, wrote the manuscript, reviewed/edited the manuscript, and contributed to the peer-review process. A.A. researched data, performed data analysis, created tables, produced figures, and reviewed/edited the manuscript. E.F. reviewed/edited manuscript. S.F., E.M., and S.S.S. researched data, performed data analysis, wrote the manuscript, and reviewed/edited the manuscript. B.L. wrote the manuscript and reviewed/edited the manuscript. R.L.M. and M.R. researched data and reviewed/edited the manuscript. B.O. and M.S. researched data, wrote the manuscript, and reviewed/edited the manuscript. M.P. researched data. K.L.C. oversaw the study and reviewed/edited the manuscript. K.L.C. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.