Antihyperglycemic Medications: A Claims-Based Estimate of First-line Therapy Use Prior to Initialization of Second-line Medications

The total annual cost of diabetes in the U.S. is $245 billion (1,2), including ∼$21 billion for antihyperglycemic agents and diabetes supplies (2). Pharmacologic glycemic control lowers rates of microvascular complications, stroke, and myocardial infarction (3,4) and improves quality of life (5,6). The American Diabetes Association and the European Association for the Study of Diabetes recommend metformin as the most cost-effective (7) first-line therapy for type 2 diabetes, barring specific contraindications (8–10). Recent meta-analyses have supported metformin as a relatively safe and effective first-line therapy for type 2 diabetes (11,12). The 2009, 2012, and 2015 guidelines, which were in effect during the period of our study, recommend combination therapy with additional antihyperglycemic agents if monotherapy with metformin fails to achieve the specified hemoglobin A_1c target over 2–3 months (8–10). However, nonadherence to antihyperglycemic medication is a common reason for inadequate diabetes control (13,14) and has been linked to poor outcomes (13,15) and diminished quality of life (16). Because a clinician could confuse nonadherence with metformin failure, we explored the treatment patterns of patients who receive second-line therapy for diabetes, specifically by estimating their previous claims-based use of metformin.

We performed a retrospective cohort study by using unidentifiable member claims data from Aetna, a U.S. health benefits company that covers millions of beneficiaries across the country. Forty-six million unique individuals with medical claims between July 2010 and March 2015 were included in the study. Patients between 18 and 65 years old who were given a new diagnosis of and treated for type 2 diabetes between July 2010 and March 2015 were included. In the data set, each encounter was coded with up to six ICD-9 codes. Prescription drugs were reported by using the National Drug Code. We defined a washout period of 365 days to attempt to include only patients who were newly diagnosed and truly initiating antihyperglycemic medication. Therefore, we considered only those with type 2 diabetes diagnosed ≥365 days after enrollment in the insurance program. Patients were excluded if they had <180 days of follow-up after the first recorded type 2 diabetes diagnosis. The Boston Children’s Hospital Institutional Review Board (Boston, MA) approved the study, granting a waiver of consent.

Patients were included in the study according to published and validated criteria (17). They were required to have two physician claims at least 1 day apart within 2 years or one hospitalization with the relevant diabetes ICD-9 codes (250.0x–250.7x and 250.9x, where x = 0 or 2) submitted in a claim between 1 July 2010 and 31 March 2015. To focus on patients treated for type 2 diabetes, we considered only those with claims for at least one dispensed antihyperglycemic medication for >2 weeks. Patients were excluded if they had at least one claim for pregnancy (including gestational diabetes mellitus [ICD-9 codes 640.0–679.0]) within 5 months of their first treatment claim for diabetes. Because inpatient medications are handled differently in payer claims data sets, patients with inpatient stays during the recommended use measurement periods were excluded from further analysis.

The American Diabetes Association and European Association for the Study of Diabetes both recommend metformin as first-line pharmacologic therapy for type 2 diabetes, and all other antihyperglycemic medications are classified as second line (8–10). For each beneficiary, we defined the first-line medication as metformin and second-line medication as other antihyperglycemic agents (Supplementary Table 1). If two second-line medications were dispensed, the original and additional second-line medications were defined on the basis of the dispensed medication date. If more than one antihyperglycemic medication other than metformin was dispensed on the same day, the medication prescribed for the longer period was defined as the first second-line medication. Guideline-recommended contraindications to metformin, including heart failure, chronic obstructive pulmonary disease, liver diseases, renal diseases, and gastrointestinal adverse effects (8–10), are defined in Supplementary Table 2.

We used a standard metric for adherence, proportion of days covered (PDC), defined by the Centers for Medicare & Medicaid Services as the number of days that the patient was covered by medication divided by the number of days in the measurement period (18) to assess the recommended use of first-line antihyperglycemic medication. Days of overlap between two fills counted toward the PDC but only if the two fills were of the same antihyperglycemic medication. A PDC of at least 0.80 is the accepted standard indicator of good adherence to antihyperglycemic medication (18) and has been applied in previous studies (19–21). For each individual, we calculated the PDC over a guideline-recommended 60-day period (8–10) that ended with the fill date (index date) for the first prescription of a second-line medication (PDC60) (Fig. 1). The period covered by each metformin prescription fill was defined by supply days. If patients never received metformin prior to second-line treatment or PDC60 <0.8, they were considered as having received second-line treatment without evidence of recommended use of metformin.

Because insurance claims rarely are submitted when patients obtain metformin through a low-cost generic prescription medication program (22), we performed a sensitivity analysis across a range of usage. Among patients who received second-line treatment without evidence of recommended use of metformin, we varied the percentage of patients obtaining metformin through a low-cost generic prescription medication program and adherence to metformin prior to initialization of second-line therapy from 0 to 45% on the basis of published estimates (22), examining the effect of the possible missing claims in the claims-based medication use measurement.

Student t test was used for continuous data and Pearson χ² test for categorical data. All statistical tests were two-sided. P < 0.05 was considered statistically significant. All analyses were performed by using the statistical package R version 3.2.3 (R Foundation for Statistical Computing, www.r-project.org).

A total of 1,500,560 individuals met the case definition of type 2 diabetes. After applying exclusion criteria, 52,544 individuals remained eligible (Fig. 2) for the retrospective cohort analysis. Of them, 3,887 (7.3%) were included on the basis of having one hospitalization with the relevant diabetes ICD-9 codes. Table 1 shows the cohort demographic characteristics. Patients who received only metformin and no other antihyperglycemic medication during the study period (29,588 [56.3%]) were older (standardized mean difference 0.123) than those who received at least one second-line antihyperglycemic medication (22,956 [43.7%]). Among patients who received at least one second-line antihyperglycemic medication, 6,441 (28.0%) had no prior claims evidence of having taken metformin.

Of all patients given second-line medications for type 2 diabetes, only 1,875 (8.2%) had claims evidence of recommended use of metformin in the prior 60 days. Table 2 shows these patients’ demographic characteristics. From 2010 to 2015, the percentage of patients with evidence of recommended use of metformin during the prior 60 days increased from 7.7 to 9.4% (P < 0.001). Among the 21,081 patients who received second-line medication without evidence of recommended use of metformin for the prior 60 days, 6,441 (30.6%) were not given metformin or metformin combination medications (Supplementary Fig. 1). Only a minority of these 6,441 patients had claims evidence of contraindications to metformin, including heart failure (185 [2.9%]), chronic obstructive pulmonary disease (202 [3.1%]), liver diseases (275 [4.3%]), and renal diseases (261 [4.1%]). Approximately one-third of patients (7,382 [35.0%]) received some metformin before initiation of second-line medication, but the duration of their metformin treatment prior to second-line treatment was <48 days (PDC 0.8), indicating that second-line therapy was initiated earlier than the 2 months recommended by current guidelines. Of these patients, 3,872 (52.4%) were prescribed both metformin and the second-line medication on the same day. Among patients without evidence of recommended use of metformin during the prior 60 days, only 2.6% had ICD-9 codes for metformin-related gastrointestinal adverse effects between initiating first-line and initiating second-line therapy. Patients classified as nonadherers were more likely to be male (P < 0.001) and were dispensed additional second-line medication (P < 0.001).

We conducted a sensitivity analysis among patients who received second-line medication for type 2 diabetes without evidence of metformin use for the guideline-recommended length of treatment of 2 months. We ranged our assumption about the percentage of patients obtaining metformin from a low-cost generic prescription medication program and having evidence of recommended use of metformin prior to initialization of second-line therapy from 0 to 45%. From the low to the high end of the range of use of low-cost generic metformin, between 8.2 and 49.5% of patients were estimated to have evidence of guideline-recommended metformin use before receiving second-line medication.

Beneficiaries without evidence of recommended use of metformin were more likely to be given insulin than those with evidence of recommended use (Table 2). Dipeptidyl peptidase 4 (DPP-4) inhibitors and their combination drugs were the most common branded second-line oral antihyperglycemic medications among all patient subgroups. Patients were more likely to be given an additional second-line antihyperglycemic medication or insulin (Table 2) if they were given their initial second-line antihyperglycemic medication without evidence of recommended use of metformin (P < 0.001).

A substantial fraction of patients given second-line pharmacologic treatment for type 2 diabetes did not have evidence of prior use of metformin for the guideline-recommended length of treatment of 2 months. The precise number cannot be directly measured because prescriptions filled through a low-cost generic prescription medication program are probably not represented in this data set. A previous study reported that ∼30% of patients who received metformin had at least one fill through a low-cost generic prescription medication program in 2007–2011 (22). However, even at our upper estimate of 45% of patients filling prescriptions through these low-cost programs (22), less than one-half of patients started on second-line therapies have evidence of use of metformin as recommended. At the lower bound of sensitivity for use of low-cost generics, <10% have evidence of recommended use of metformin. Patients without evidence of recommended use of metformin are more likely to be male and to receive insulin or an additional second-line antihyperglycemic medication.

More than 10% of all beneficiaries starting on an antihyperglycemic medication had no claims evidence of having been prescribed metformin as a single agent or in combination with other medications. A previous study found that ∼35% of patients initiating an oral antihyperglycemic medication were not started on metformin (23). Almost 15% of beneficiaries starting oral antihyperglycemic medication were initially adherent to metformin, but were started on second-line therapy prematurely (8–10). Although the guideline allows patients with a high baseline HbA_1c to start with a combination of two noninsulin agents (HbA_1c ≥9%) or with insulin (HbA_1c 10–12%) (9,24), prior studies have shown that 55% of newly diagnosed patients have an HbA_1c <6.5% (25), yet only 13% of patients with diabetes in the U.S. have an HbA_1c >9% (26). We have found a much higher proportion of patients prescribed a second-line antihyperglycemic medication than expected (24–26).

Informatics infrastructure encouraged under the Meaningful Use program for using certified health information technology should be leveraged to boost adherence (27). Data on medications recently dispensed to the patient are readily available in real time from companies that manage pharmacy benefits or execute e-prescribing transactions. Delivery of this information to population health managers or to clinicians at the point of care through electronic health record services or third-party applications (28,29) could raise awareness of nonadherence to guidelines prior to initiation of second-line therapy. A number innovative approaches from short message service messaging (30) to electronic medication packaging (31) can improve guideline adherence.

Retrospective claims-based analyses have several limitations, including the exclusion of uninsured patients, lack of detailed clinical or behavioral information, and no out-of-pocket medication cost data (23,32). Claims data include tests and treatments billed for and medications dispensed but not the motivation for testing or prescribing medications. The medication switches or choices might represent physician or patient choice. Patient-specific clinical data and nuances that might have played a substantial role in individual clinician’s choices and timing of prescription medications are not included in claims data analyses. Although, gastrointestinal adverse effects related to metformin therapy might lead to guideline nonadherence and early second-line medication initiation, we did not find evidence for gastrointestinal upset in the claims data. Because we relied on claims data, diagnosis codes of gastrointestinal adverse effects and contraindications may vary in accuracy (32). Furthermore, previous studies have found that only ∼5% of patients treated with metformin discontinued metformin because of an adverse event (33–35). Importantly, although claims data indicate that certain medications were dispensed and paid for, the patient may not have been actually taken them (36). Nonetheless, a substantial amount of literature supports the use of prescription fills to monitor adherence (37).

In conclusion, as few as 8% but no more than one-half of patients given second-line pharmacologic treatment for type 2 diabetes have prior evidence of use of metformin for the guideline-recommended length of treatment of 2 months. Patients without evidence of recommended use of metformin are more likely to be male and prescribed insulin or an additional second-line antihyperglycemic medication. What may be taken as evidence of treatment failure by clinicians may instead represent failure of adherence to established treatment guidelines, which in turn may lead to the use of insulin or additional second-line medications. Point-of-care decision support and population health–level approaches should focus on improving adherence to first-line therapy.

Funding. This study was funded by a grant from the National Institute of General Medical Sciences, National Institutes of Health (R01-GM-104303). Y.-J.T. was funded by grants from the Ministry of Science and Technology, Taiwan (MOST 105-2632-H-182-001 and MOST 106-2221-E-182-072), and Chang Gung Memorial Hospital (CMRPD3F0051).

The sponsors played no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

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

Author Contributions. Y.-J.T. analyzed and interpreted the data and performed the experiments. Y.-J.T. and K.D.M. designed the study and wrote the manuscript. G.S., K.P.F., J.A., and K.D.M. reviewed and edited the manuscript for important intellectual content and provided administrative, technical, or material support. K.D.M. obtained funding and supervised the study. Y.-J.T. 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.