Trends in Metformin Use Among Patients With Prediabetes: 2008–2020

Since 2008, the American Diabetes Association has recommended metformin as an adjunct to intensive lifestyle intervention for people with prediabetes, especially those with BMI ≥35 kg/m², with history of gestational diabetes mellitus (hxGDM), or aged <60 years (1). Prior studies assessing trends of metformin use among patients with prediabetes yielded mixed results and were limited by small sample size, use of self-reported information, insufficient control for important patient characteristics, and lack of recent data (2,3). In this study, we used a claims-based data set to examine metformin use among >900,000 people with prediabetes from 2008 to 2020.

Optum Clinformatics is a large national database containing commercial health insurance medical/pharmacy claims data and available outpatient laboratory results. We selected patients with index prediabetes diagnosis between 2008 and 2020 according to hemoglobin A_1c (HbA_1c) 5.7–6.4%, fasting plasma glucose 100–125 mg/dL, 2-h post–stimulated glucose (oral glucose tolerance test) 140–199 mg/dL, or ICD-9 and ICD-10 diagnosis codes (4). We excluded patients <20 and ≥60 years old, those on Medicare, and those who within 1 year of index prediabetes diagnosis did not have continuous insurance enrollment or had type 1 or 2 diabetes (HbA_1c >6.4%, fasting plasma glucose >126 mg/dL, oral glucose tolerance test >199 mg/dL, or qualifying ICD-9 or ICD-10 code), prescription of nonmetformin antidiabetes medications, metformin contraindication kidney disease stage 3 or above, or alternative metformin indication polycystic ovarian syndrome (4). ICD-9 and ICD-10 were used to identify BMI ≥35 kg/m² (ICD-9 V85.35–V85.39/V85.4; ICD-10 Z68.35–Z68.39/Z68.4) and hxGDM (ICD-9 648.8/ICD-10 O24).

We estimated incidence of metformin prescription fill within 12 months after index prediabetes diagnosis, with subgroup analysis on those with BMI ≥35 kg/m² or hxGDM, via age- and sex-adjusted logistic regression with year dummy variables. Multivariate logistic regression was performed to assess the association of metformin use with age, sex, race (White (reference)/Black/Hispanic/Asian), BMI ≥35 kg/m², and hxGDM, with two-sided P < 0.05 considered significant. This study was approved by the University of Southern California Institutional Review Board.

There were 945,414 patients with index prediabetes diagnosis between 2008 (n = 63,251) and 2020 (n = 84,883), among whom 47,257 (5.0%) had BMI ≥35 kg/m² or hxGDM (610 [0.96%] in 2008 and 8,774 [10.3%] in 2020). Age- and sex-adjusted metformin use in the overall cohort increased from 1.14% in 2008 to 2.01% in 2020 (P < 0.001 for the difference) (Fig. 1A). In subgroup analysis, use increased between 2008 and 2020 from 0.87% to 4.99% (P < 0.001), with substantial increases between 2008–2009 and 2012–2017 but no significant change from 2017 (4.90%, 95% CI 4.37–5.43) to 2020 (4.99%, 95% CI 4.53–5.45; P = 0.798).

Multivariate logistic regression revealed independent, positive associations of metformin use with BMI ≥35 kg/m² (odds ratio [OR] 2.74, 95% CI 2.62–2.86), hxGDM (OR 2.52, 95% CI 2.27–2.79), and female sex (OR 1.92, 95% CI 1.86–1.99), while Asian relative to White race was associated with lower use (OR 0.83, 95% CI 0.77–0.89) (Fig. 1B). These findings are generally consistent with those of previous studies (2,3).

Despite the American Diabetes Association’s long-standing recommendation of metformin for diabetes prevention, <1 in 20 adults with prediabetes and BMI ≥35 kg/m² or hxGDM used metformin in 2020. In a recent study, investigators found the strongest predictor of whether patients receive any treatment for prediabetes (including metformin) to be their primary care physicians’ treatment rates, which varied substantially (5). Efforts to improve diabetes prevention may benefit from lessons learned with other guidelines and from implementation science more generally. There was improvement between 2008 and 2020, and some of it may have stemmed from systemic changes in U.S. health care. The Affordable Care Act has expanded health insurance coverage and promoted payment and delivery reforms, potentially catalyzing improvements in care throughout the health care sector.

It must be noted that some patients with prediabetes not taking metformin may still have received appropriate care via intensive lifestyle intervention or other diabetes medications (e.g., sodium–glucose cotransporter 2 inhibitor for comorbid heart failure or kidney disease, although patients taking such medications are excluded in our study) (1).

Our research design improves on the limitations of the existing literature including small sample size (N < 8,000), use of self-reported data, older data (before 2014), and lack of control for metformin contraindications or alternate indications (2,3). Compared with prior studies where investigators did not control for time since diagnosis, our focus on patients with newly diagnosed prediabetes and incident metformin use with a fixed follow-up period also better accounts for patient heterogeneity in disease severity and allows for clearer characterization of changes in prescription fill patterns over time. Nevertheless, study limitations include a commercially insured sample that may not generalize, lack of sensitivity with using ICD-9 and ICD-10 codes to identify patients with BMI ≥35 kg/m² or hxGDM, and inability to examine intensive lifestyle intervention participation, which may account for some patients’ lack of metformin use in our study. Future research should investigate intensive lifestyle intervention alongside metformin use and also examine patterns and determinants of care over time and across areas.