Diabetes Care
American Diabetes Association
Effect of Cost and Formulation on Persistence and Adherence to Initial Metformin Therapy for Type 2 Diabetes
DOI 10.2337/dc19-2426, Volume: 43, Issue: 6,

Table of Contents



Flory and Mushlin: Effect of Cost and Formulation on Persistence and Adherence to Initial Metformin Therapy for Type 2 Diabetes

Patients taking metformin adhere to their prescribed regimen, on average, about 65% of the time (1,2). Given metformin’s central role in type 2 diabetes management, optimizing adherence is important. Common causes of nonadherence include cost and side effects. Metformin is inexpensive but causes gastrointestinal side effects in up to 25% of patients (2). Extended-release (ER) formulations reduce side effects but may cost more, and they have not been clearly shown to improve adherence in routine practice (3). As a result, although it is common practice to change from immediate release (IR) formulations to ER once patients develop side effects, there is not a consensus on whether ER formulations should routinely be used when metformin is first prescribed (3). We undertook a retrospective cohort study of how cost and initial choice of formulation affect adherence and persistence to metformin.

The study used merged de-identified claims data from commercial insurance carriers in the U.S. from 1 January 2012 to 31 December 2016 (4). A cohort was identified of patients age 18 years or older who filled an initial 30-day prescription for metformin monotherapy with a baseline type 2 diabetes diagnosis, at least 1 year of baseline and follow-up data, and no prior antidiabetes drug use. Baseline covariates included age, sex, insurance type, comorbidities, medications, median income at the 5-digit ZIP code level, and total and out-of-pocket (OOP) cost of the initial metformin prescription as well as its formulation (ER or IR) and dose.

The primary outcome, “persistence,” was defined as at least one metformin prescription claim during the 6–12 month window after the initial prescription. Secondary outcomes were “adherence,” the percentage of days for which the patient had filled sufficient prescriptions to be taking metformin as prescribed, and “adequate adherence,” defined as adherence of 80% or more, both during the 1st year of follow-up (5). Univariate statistics and logistic and linear regression were used to describe covariates, outcomes, and their association. The Weill Cornell Medical College Institutional Review Board ruled this research exempt.

The final study population was 81,406 individuals, of whom 78% had commercial insurance, 19% had Medicare Advantage, and 2.5% were dual-eligible for Medicare and Medicaid. The population was 46% female. Twenty-eight percent were age 45–54 years, 31% were 55–64, and 16% were 65–74. Rates of major comorbidities ranged from 0.4% for dementia to 6.4% for diabetes complications.

Metformin prescription costs were low, with median total cost for the first prescription of $5 (interquartile range [IQR] 4–8) and median OOP cost of $4 (IQR 3–7). Nearly all metformin prescriptions (99%) were generic. Brand prescriptions were associated with higher median total ($305) and OOP cost ($55). Twenty-six percent of initial prescriptions were for ER metformin, which compared with IR had higher median total ($6 vs. $5) and OOP cost ($5 vs. $4) (P < 0.01 for all differences).

Persistence was higher with ER metformin than with IR metformin (75% vs. 73%), as was adherence (62% vs. 59%) and adequate adherence (42% vs. 37%). This relationship was consistent after stratification by prescription OOP cost. In multivariable logistic regression including adjustment for cost, dose, demographics, and comorbidities, the odds ratio (OR) for persistence was higher for ER formulations (OR 1.14 [95% CI 1.10–1.18], P < 0.001) (Table 1). The adjusted relationship between persistence and cost was nonlinear, with the highest persistence in the lowest cost quartile and lower relative persistence in all higher cost quartiles (OR range 0.81–0.95, P < 0.001 for all contrasts). Results of multivariable modeling for the other outcomes were comparable, as were results of sensitivity analyses.

Table 1
Results of multivariable logistic regression with persistence as a binary outcome.
Baseline variableOR95% CI lower bound95% CI upper bound
Commercial insuranceRef
Dual eligible1.080.961.20
Medicare Advantage1.121.061.18
Female sex0.910.880.94
Age (years)
Diabetes complications0.950.891.01
Liver disease0.930.871.00
Renal disease0.800.730.89
Baseline prescription drug use
 1–3 drug classesRef
 4 drug classes1.061.011.11
 5–6 drug classes1.071.031.12
 >6 drug classes1.051.011.10
ZIP code median income
 $42,000 to $54,0001.061.021.11
 $54,000 to $72,0001.121.071.17
ZIP code demographics
 Majority percentage white1.081.021.15
 Majority percentage black1.010.931.09
Metformin formulation
Metformin pills daily
 1 pillRef
 2 pills1.080.991.18
 >2 pills1.131.011.27
Metformin daily dose
 ≤500 mgRef
 501–1,000 mg0.990.901.08
 >1,000 mg0.960.861.06
Metformin prescription cost (OOP)
CHF, congestive heart failure; MI, myocardial infarction; PVD, peripheral vascular disease; Ref, reference.

This analysis found that, despite slightly higher cost, initial use of ER metformin was associated with better adherence. This study is consistent with prior publications but is the first to show this effect after adjustment for key covariates including cost and dose (2). While modest, the association between ER metformin and improved adherence (equating to a 2.5% absolute increase in adherence and a 13% relative increase in persistence) is meaningful if it is causal. For perspective, a program that lowered copays for medication and provider visits as well as providing access to wellness coaching improved metformin adherence by 5% (1).

We believe these findings provide the strongest evidence to date that routine initial use of ER metformin is a simple intervention that may improve adherence. This is consistent with cost-effectiveness analyses suggesting that routine use of ER metformin could be cost-saving (5). However, these data only provide preliminary evidence for such a causal claim. Prospective studies of efforts to improve adherence to diabetes medications should evaluate use of ER metformin as a potentially important component.

Article Information

Funding. This work was supported by a career development grant to J.H.F. from the Agency for Healthcare Research and Quality (K08HS023898).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.H.F. conceived the study and conducted the data analysis. A.I.M. contributed to the study design, data interpretation, and manuscript preparation. J.H.F. 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.



    Duru OK, Turk N, Ettner SL, . Adherence to metformin, statins, and ACE/ARBs within the Diabetes Health Plan (DHP). J Gen Intern Med2015;30:, pp.1645–1650


    Donnelly LA, Morris AD, Pearson ER. Adherence in patients transferred from immediate release metformin to a sustained release formulation: a population-based study. Diabetes Obes Metab2009;11:, pp.338–342


    Gout-Zwart JJ, de Jong LA, Saptenno L, Postma MJ. Budget impact analysis of metformin sustained release for the treatment of type 2 diabetes in the Netherlands. Pharmacoecon Open. 18 September 2019 [Epub ahead of print]. DOI: 10.1007/s41669-019-00179-6


    Newman D, Herrera CN, Parente ST. Overcoming barriers to a research-ready national commercial claims database. Am J Manag Care2014;20:eSP25–30


    Raebel MA, Schmittdiel J, Karter AJ, Konieczny JL, Steiner JF. Standardizing terminology and definitions of medication adherence and persistence in research employing electronic databases. Med Care2013;51(Suppl. 3):, pp.S11–S21
https://www.researchpad.co/tools/openurl?pubtype=article&doi=10.2337/dc19-2426&title=Effect of Cost and Formulation on Persistence and Adherence to Initial Metformin Therapy for Type 2 Diabetes&author=James H. Flory,Alvin I. Mushlin,&keyword=&subject=e-Letters: Observations,