Association Between Change in A1C and Use of Professional Continuous Glucose Monitoring in Adults With Type 2 Diabetes on Noninsulin Therapies: A Real-World Evidence Study

This was a retrospective, observational, database study using de-identified administrative health claims and linked laboratory data from Optum’s Clinformatics Data Mart (CDM) database. The CDM is statistically de-identified under the expert determination method consistent with the Health Insurance Portability and Accountability Act and managed according to Optum customer data use agreements in compliance with Code of Federal Regulations title 45, section 164.514(b) (1)13 (37). CDM administrative claims submitted for payment by HCPs and pharmacies are verified, adjudicated, and de-identified prior to inclusion. These data, including patient-level enrollment information, are derived from claims submitted for all medical and pharmacy health care services with information related to health care costs and resource utilization.

The population included adult patients ≥30 years of age who had type 2 diabetes identified using International Classification of Diseases, 9th/10th Revisions (ICD-9/ICD-10), were enrolled in commercial or Medicare Advantage health plans, and had no prior personal CGM or p-CGM use. Patients with poor glycemic control, defined as an A1C value between 7.8 and 10.5%, and using two or more noninsulin therapies were included in the cohort. p-CGM users were identified using Current Procedural Terminology (CPT) codes 95250 and 95251 between 1 January 2018 and 31 October 2020 (the study identification period). Patients without claims for use of p-CGM who had a pharmacy claim for an oral antidiabetic drug (OAD) during the study identification period were selected as the control cohort (nonusers of p-CGM). An index date was set as the earliest observed claim for p-CGM or an OAD.

Included patients were continuously enrolled in a health plan from at least 6 months pre-index (baseline) to at least 6 months post-index date (follow-up) and had no claims for a CGM device (personal or professional) during the baseline period.

Type 2 diabetes diagnoses were confirmed by patients having one or more inpatient hospital or emergency department medical claims or at least two ambulatory medical claims at least 30 days apart with an ICD-9 or ICD-10 diagnosis code for diabetes in any position on the claim during the baseline period. Additionally, patients were required to be using at least two noninsulin medications and not taking insulin (basal or bolus) in the baseline period. Finally, patients were required to have one or more laboratory A1C test results with values between 7.8 and 10.5% during the baseline period (including the index date) and at least one A1C laboratory test value during the follow-up period.

The primary outcome measure was change in A1C determined from the average of available A1C values during the baseline and follow-up periods. Change in A1C was computed as the average A1C follow-up value minus the average A1C baseline value, with negative values indicating improvement in A1C. Secondary outcomes were change in the number of medications by class, insulin use in follow-up period, and change in A1C for patients starting insulin and not starting insulin during the follow-up period.

Descriptive statistics, including percentages, means, and SDs, were calculated for patient characteristics and study outcomes and presented by patient cohort categories (i.e., p-CGM users and nonusers). Bivariate differences between the p-CGM user and nonuser groups were tested with independent sample t tests for continuous measures and χ² tests for categorical measures. Within-cohort comparisons of continuous measures were tested with paired t tests. The difference-in-differences (DiD) estimate was calculated as the difference in change in A1C values between the p-CGM user and nonuser cohorts. The DiD estimate indicates the magnitude and direction of change in outcome between the two groups. The association between p-CGM use and antidiabetic medication changes was tested using a logistic regression model. A χ² test was used to compare the difference in insulin use in the follow-up period between the two groups. For all analyses, statistical tests were two-tailed, with P ≤0.05 considered statistically significant. Analyses were performed using Instant Health Data software (Panalgo, Boston, MA) and R, v. 3.2.1, software (R Foundation for Statistical Computing, Vienna, Austria).

A total of 15,481 patients were identified during the study identification period, including 707 p-CGM users and 14,774 nonusers. Demographic characteristics were similar between cohorts as shown in Table 1. The majority in both groups were older (≥65 years of age), there were slightly more males, and most were Caucasian and predominantly covered by Medicare Advantage plans. Among patients using a p-CGM, an endocrinologist was the prescriber for 282 patients (39.9%), whereas for 275 patients (38.9%), the prescriber was a family or internal medicine physician or nurse practitioner. The remaining 21.2% of encounters at which p-CGM was prescribed were with some other type of HCP. Thus, ∼40% of patients received their p-CGM through primary care.

The p-CGM group reduced their A1C by a mean 0.83%, from 8.70 to 7.87%, while the nonusers had a reduction of 0.32%, from 8.56 to 8.23%. The DiD estimate was a −0.51% change in A1C and was statistically significant (P <0.0001), as shown in Table 2.

As shown in Table 3, 140 p-CGM users (19.8%) started using insulin during the follow-up period, with most (112, or 15.8%) using basal insulin. About 10% of nonusers started insulin during the follow-up period. The A1C change among insulin users in the p-CGM group was −0.57%, from 8.90 to 8.34%, whereas the nonuser group had a slight increase in A1C of 0.13%, from 8.77 to 8.90%. The DiD estimate for change in A1C was −0.71% and statistically significant (Table 2). Similarly, for patients not starting insulin, we found a significant DiD change in A1C of −0.53% for nonusers of insulin during follow-up, reflecting a larger A1C decrease for p-CGM users compared with nonusers (Table 2).

Table 3 shows the diabetes medication use by cohorts during the baseline and follow-up periods. The most frequently used medications in both cohorts were biguanides followed by sulfonylureas, glucagon-like peptide 1 (GLP-1) receptor agonists, and dipeptidyl peptidase 4 (DPP-4) inhibitors in the baseline and follow-up periods. Biguanide, sulfonylurea, and DPP-4 inhibitor use decreased slightly from baseline to follow-up in both groups, with odds ratios (ORs) showing that the p-CGM group was less likely to use these medications at follow-up, adjusting for baseline, compared with the nonuser group. GLP-1 receptor agonist, sodium–glucose cotransporter 2 (SGLT2) inhibitor, and meglitinide use increased in both groups, with ORs showing that the p-CGM group was more likely to use these medications at follow-up, adjusting for baseline, compared with the nonuser group. Overall, 447 patients (63.2%) had an additional p-CGM use during the 6-month follow-up period.

The findings from this study suggest that there is a glycemic benefit of p-CGM use for adults with type 2 diabetes who are not on insulin therapy and are taking multiple OAD and/or noninsulin injectable medications with poor glycemic control. Use of p-CGM was associated with a clinically meaningful 0.51% reduction in A1C compared with the cohort of nonusers.

These findings align with other studies demonstrating a significant reduction in A1C values for p-CGM users with type 2 diabetes compared with control subjects (32,34,35). Previous studies have shown that p-CGM enables HCPs to make decisions such as adding insulin to patients’ treatment regimens based on recorded sensor data, helping patients reach glycemic targets and promoting positive behavior change (35,36,38). Our findings suggest that p-CGM use is also a means to assess and monitor glycemia for noninsulin-treated patients and can be used as a tool for therapy modification.

Our analysis showed that ∼20% of patients initiated insulin therapy in the follow-up period after p-CGM use, which was more than the 10% who started insulin in the nonuser group, but this finding might be considered lower than expected given that the cohort started with an average A1C >7.5% and may reflect HCP and/or patient inertia regarding initiating insulin therapy (12,14,15).

Even among patients who used p-CGM and did not start insulin therapy, there was significant improvement in A1C. This finding suggests that p-CGM can help with medication management and behavior change that might improve glycemic control and delay the need to start insulin therapy. The more granule daily patterns of glucose excursions that can be viewed through CGM, compared with average glycemic control reflected in A1C values, can help HCPs target diet and physical activity counseling to patients, which in turn could delay or prevent microvascular and macrovascular diabetes complications (8,10,38).

Associations were found between p-CGM use and changes in the use of noninsulin diabetes medications. The most prominent change was a 33% increase in the use of GLP-1 receptor agonists, which seems warranted given the cardiovascular risk-reduction benefits of this drug class and the cohort’s baseline A1C and average age >65 years. There was also a 14% decrease in sulfonylurea use in the p-CGM group, which may have been precipitated by p-CGM detection of previously unknown time spent in the hypoglycemic range.

There are limitations to our study. The number of identified p-CGM users with A1C values was relatively small compared with the nonuser cohort. For this analysis, we included only patients who had A1C values in both the baseline and follow-up time periods. Although the cohorts were comparable with regard to demographic variables and baseline A1C, there was the chance of confounding by indication, through which patients using p-CGM could have been different from nonusers on unmeasured variables that prompted the use of p-CGM.

Although the database provided crucial information about participants’ health status, comorbidities, medications, and use of p-CGM, it did not provide other information such as socioeconomic status or participation in diabetes health coaching programs. We also cannot determine from the data whether and to what extent HCPs and patients engaged in shared decision-making about diabetes and lifestyle management. Similarly, changes in physical activity or other positive behavioral changes and their impact on overall glycemic improvement could not be assessed. Without this information, it is not possible to fully understand how p-CGM may help individuals to improve their diabetes management.

p-CGM use is documented in health care claims with CPT codes 95250 and 95251; therefore, it was not possible to distinguish the specific p-CGM brands or products used. Three p-CGM systems were available in the United States during this study period: the Dexcom G4/G6 Pro, the FreeStyle Libre Pro, and the Medtronic iPro2. Differences among these brands may influence users’ experience and adherence to wearing the devices (34).

A key strength of our analysis was its use of comprehensive, standardized health information contained within a large database. Access to the database allowed us to use quantitative data to verify participant eligibility and identify patients’ treatment regimens during the baseline and follow-up periods. An additional study strength was that the resulting cohorts were ethnically diverse, with nearly half being non-Caucasian, which supports the generalizability of the findings beyond Caucasian individuals.

Although randomized controlled trials are considered the gold standard for demonstrating the effects of therapeutic interventions, results from retrospective, observational studies of large health data systems can inform clinicians and regulatory agencies about the real-world efficacy of treatment interventions. Our findings suggest that p-CGM use in adults on noninsulin therapies is associated with improved glycemic control and thus may be a tool HCPs can use to help patients reach their glycemic goals.

The authors thank Dr. David Price for his feedback on the data analysis and critical review of the manuscript.

The authors are employees of Dexcom, Inc. No other potential conflicts of interest relevant to this article were reported.

P.M.N. researched the data, conducted data analyses, and wrote the manuscript. K.L.H. and G.J.N. researched the data and reviewed and edited the manuscript. P.M.N. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Portions of this article were presented at the 15th annual Advanced Technologies & Treatments for Diabetes meeting in Barcelona, Spain, 27–30 April 2022.