Applying Real-World Data to Inform Continuous Glucose Monitoring Use in Clinical Practice

Continuous glucose monitoring (CGM) shows promise as a way to improve glycemic control.^1-3 Examining real-world evidence is crucial when implementing CGM into clinical practice to achieve maximum benefit.^4,5 Our study aimed to examine CGM in a user population and its association with HbA1c as a guide for education and program development. Electronic health record (EHR) data were gathered on patients with diabetes using CGM between May 1, 2014 and April 30, 2019, being cared for in a diabetes clinic. The study received institutional review board (IRB) waiver and approval from the University of Pittsburgh Medical Center (UPMC) Quality Improvement Review Committee. Associations between user characteristics, CGM type, and diabetes education with HbA1c were examined.

The sample (N = 106) was White (89.6%), female (58.5%), mean age of 47.2 ± 14.0 years, and diabetes duration of 19.7 ± 14.0 years. Sixty-four (60.4%) patients had type 1 diabetes mellitus (T1DM), 34 (34.0%) type 2 diabetes mellitus (T2DM), five (4.7%) latent autoimmune diabetes in adults (LADA) and one with gestational diabetes. A total of 43.3% had diabetes-related chronic complications: 25.5% neuropathy, 22.6% retinopathy, and 13.2% nephropathy. Thirty-four patients (32.1%) had hypoglycemia and 10 (9.4%) had hypoglycemic unawareness. Comorbidities included 62 (58.5%) patients with hypertension, 36 (34.0%) with hyperlipidemia, seven (6.6%) with coronary heart disease, three (2.8%) with congestive heart failure, and one (0.9%) with the history of myocardial infarction.

Type of CGM used included intermittently scanned (n = 54, 50.9%), real-time (n = 42, 39.6%), and passive (n = 10, 9.4%). CGM brands were Libre FreeStyle (n = 66, 62.3%), Dexcom (n = 30, 28.3%), and Medtronic (n = 10, 11.8%). More T2DM patients (72.2%) used intermittently scanned CGM than T1DM patients (37.5%), while T1DM (59.4%) used more personal CGM than T2DM (5.6%) (P < .001).

There was a significant reduction in HbA1c (%) after CGM use at 3 months (mean ± SD: 7.72 ± 1.58) and 6 months (7.70 ± 1.68) compared to HbA1c at 3 months (8.34 ± 1.71) and 6 months (8.29 ± 1.80) before CGM use (P = .003) (Figure 1). There were no differences in HbA1c by diabetes type (P = .784) or CGM type (P = .298) over time. Differences in CGM time-in-range from baseline to 6 months were compared. Using one-sample nonparametric tests, a significant reduction in percent of time spent >180 mg/dL (P = .030) and significant increase in percent of time between 70 and 180 mg/dL (P = .008) was observed, while there were no significant changes in percent of time spent <54–70 mg/dL (P = .466) and <54 mg/dL (P = .317). Of note, a total of 85 (80.2%) participants received diabetes education. HbA1c differences were found for patients receiving education (vs not), but no association was found between education (vs not) and time (education × time interaction: P = .739; education: P = .037; time: P = .066).

Figure 1.

HbA1C levels before and after use of continuous glucose monitoring.

This is the first study to our knowledge that provides real-world evidence of positive association between CGM use and glycemic control regardless of diabetes or CGM type. Findings serve to inform clinical practice, particularly as it applies to diabetes education. The possibility of late complications, the risk of hypoglycemic events, and the reality that many still remain passive spectators of a disabling disease that can profoundly affect the quality of life for patients with diabetes needs to be considered as technological applications are introduced. ⁶ Using data-driven evidence to develop focused approaches for the practice and patients is essential.