Introduction
In the US, an estimated 26 million people have diabetes, approximately 8% of the entire population [
1]. Diabetic complications are an important medical and public health concern, as diabetes is associated with kidney failure, limb amputation, blindness, heart disease, and stroke [
1]. Appropriate disease management, primarily maintenance of glycated hemoglobin (HbA1c) levels below 7% can decrease the risk of microvascular complications [
2]. Current treatments for type 2 diabetes mellitus (T2DM) include lifestyle and dietary modifications, oral antidiabetic medications, non-insulin injectables, and insulin [
2,
3]. For patients with mild hyperglycemia, lifestyle change and monotherapy with metformin or an acceptable alternative, such as a dipeptidyl peptidase-4 inhibitor (DPP-4i), are recommended [
3]. Sulfonylureas (SUs) and thiazolidinediones (TZDs) can also be prescribed, although it has been suggested that the additional risks associated with these drug classes should be considered [
3]. The American Diabetes Association, European Association for the Study of Diabetes, and the American Association of Clinical Endocrinologists recommend that patients who do not reach HbA1c targets should be prescribed combination therapy, followed by insulin initiation if necessary [
2,
3].
To optimize the clinical benefits of antidiabetic pharmacotherapy, patients must remain adherent to and persistent with their prescribed medications. The importance of adherence to antidiabetic medications has been recognized by Medicare and appears as a Medicare Part D plan quality measure [
4]. Increased adherence to medications is associated with a decrease in HbA1c [
5‐
8], lower odds of mortality [
7], fewer all-cause hospitalizations [
7], and lower healthcare expenditures [
9‐
12]. These improvements in outcomes may translate into substantial benefit to the healthcare system, averting a projected 699,000 emergency room visits and 341,000 hospitalizations per year, resulting in a savings of nearly $5 billion [
13]. Despite the noted impact of good adherence, a systematic review found a wide range of adherence rates across retrospective studies, with rates ranging from 36% to 93% [
14]. A second review found that when combining study results, only 58% of patients were considered adherent [
15]. To date, the majority of interventions to improve adherence to diabetes medications have focused on education and one-on-one counseling [
16].
Several factors associated with adherence to antidiabetic medications have been identified in the literature. Younger age, female gender, residence in the Southern region of the US, and high cost-sharing have been associated with lower adherence in multiple analyses [
10,
17,
18]. Additional factors include depression, polytherapy, and twice-daily regimens [
14]. Analyses have also compared patients initiating specific non-insulin antidiabetic drugs and drug classes in terms of adherence and persistence, including metformin, SUs, glucagon-like peptide-1 receptor agonists (GLP-1RAs), TZDs, and DPP-4is [
10,
17,
19,
20].
To our knowledge, no US-based analyses comparing adherence to and persistence with DPP-4is as a medication class with other classes of antidiabetic medications have been reported. Therefore, the objective of this analysis was to compare adherence and persistence among patients initiating DPP-4is, SUs, and TZDs over 1- and 2-year follow-up periods. Additional objectives were to compare adherence and persistence between patients initiating saxagliptin and sitagliptin and to describe treatment patterns following initiation.
Methods
Data Source
This retrospective, observational cohort study used administrative claims data from the Truven Health MarketScan
® (Truven Health Analytics, Ann Arbor, MI, USA) Commercial and Medicare Supplemental databases. The databases contain inpatient medical, outpatient medical, and outpatient prescription administrative claims data and enrollment records of individuals with employer-sponsored primary or Medicare supplemental insurance. The Commercial and Medicare Supplemental Databases have been utilized in numerous published analyses [
21]. The data were previously collected, de-identified, and compliant with the Health Insurance Portability and Accountability Act privacy regulations. Therefore, Institutional Review Board approval and written informed consent were not required for this study.
Patient Selection
Patients with at least one outpatient pharmacy claim with a National Drug Code (NDC) for a DPP-4i, SU, or TZD from January 1, 2009 to January 31, 2012, were selected. A rolling selection method was used, whereby each drug claim was evaluated on the criteria described below and the earliest claim meeting all criteria was selected as the index date and the medication class (DPP-4i, SU, or TZD) was considered the index drug class. Patients must have had at least 28 days of continuous supply of the index drug class and have been at least 18 years old on the index date. All patients were required to have at least 365 days of continuous enrollment with medical and pharmacy benefits prior to the index date (pre-period) and at least 365 days of enrollment following the index date (1-year follow-up period). A subset of patients was required to have 730 days of post-index enrollment (2-year follow-up period). To capture initiators only, patients were required to have no claims for the index drug class in the pre-period. Lastly, patients were required to have at least one non-diagnostic medical claim with a T2DM diagnosis (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] 250.x0, 250.x2) during the study period. Patients with medical claims with diagnoses of type 1 diabetes (ICD-9-CM 250.x1, 250.x3) or gestational diabetes (ICD-9-CM 648.8x) or with multiple index drug classes on index date were excluded from analysis.
Outcome Variables
The primary outcomes were adherence to and persistence with the index drug class. Both measures were calculated using the service date and days’ supply fields existing on outpatient pharmacy claims for DPP-4is, SUs, and TZDs identified by NDC codes. Adherence was measured as proportion of days covered (PDC), calculated by taking the number of days a patient had the index drug class on hand during the 1- or 2-year follow-up based on the days’ supply field on pharmacy claims divided by follow-up time (365 or 730 days). Days’ supply for early refills was appended to the end of days’ supply of the previous prescription. Patients with a PDC ≥ 0.80 were considered adherent. Persistence was defined as the time from index date to last day with index drug class on hand prior to a gap of >60 consecutive days without index drug class [
22]. A cut-point of 60 days was utilized as a conservative definition of discontinuation, as patients with gaps of up to 60 days were considered persistent. Switching within drug class was allowed for the drug class comparison. When comparing adherence and persistence outcomes between patients initiating saxagliptin and sitagliptin, PDC and time to discontinuation were calculated at the drug level, rather than the drug class level.
Patients were classified into the following mutually exclusive groups based on first event after index date during the 1-year follow-up period: remained on index drug class with no augmentation, augmentation with additional drug class, discontinuation of index drug class and switch to a new drug class, discontinuation of study drug class and continuation on other medication classes with no switch, and discontinuation of all antidiabetic medications. An augmentation was defined as the addition of a medication class not part of the initial regimen that overlapped with the index drug class for >30 days. A switch was defined as the discontinuation of index drug class and the addition of medication class not in the initial regimen prior to discontinuation with overlap ≤30 days or following discontinuation. Discontinuation was measured at the drug class level for all cohort comparisons.
Independent Variables
The primary independent variable of interest was index drug class: DPP-4i, SU, or TZD. When comparing within the DPP-4i medication class, the primary predictor was index drug: saxagliptin or sitagliptin. A third DPP-4i, linagliptin, was not compared within the DPP-4i–specific analysis because few linagliptin initiators had 1 year of follow-up, and no linagliptin initiators had 2 years of follow-up in the claims data. A fourth DPP-4i, alogliptin, was not available in the US during the patient selection period.
Demographic, clinical, and cost and utilization characteristics were measured to describe the study sample and to control for potential confounding in multivariable models. Demographic measures included: sex, age, geographic region, urbanicity, insurance plan type, primary payer, presence of capitated services, and year of index date. Clinical characteristics were measured during the pre-period and included use of study drugs other than index drug class, and metformin and insulin use based on pharmacy claims, overall health captured by the Deyo Charlson Comorbidity Index (CCI) [
23] and the number of unique ICD-9-CM codes appearing on a patient’s medical claims, diagnosis of macrovascular disease (acute myocardial infarction, other ischemic heart disease, congestive heart failure, cerebrovascular accident, or peripheral vascular disease) and microvascular disease (diabetic peripheral neuropathy, diabetic retinopathy, leg and foot amputation, or nephropathy), and diagnosis or procedure indicative of renal impairment. Pregnancy during the follow-up period was captured, as it may affect diabetes treatment. Cost and utilization covariates captured during the pre-period were evidence of an endocrinologist visit and cardiologist visit, total healthcare expenditures, and diabetes prescription expenditures.
Several characteristics of the index drug prescription claim were also captured: the days’ supply, mail-order status, fixed-dose combination with metformin, cost-sharing index, which was the average patient out-of-pocket cost for a 30-day supply, calculated annually for each study drug for each insurance plan in the database, and index regimen [
24]. A patient’s initial regimen was defined as the index drug class plus any concomitant use of additional antidiabetic medication classes (alpha-glucosidase inhibitors, amylin analogs, dopamine receptor agonists, GLP-1RAs, insulin, meglitinides, and metformin). Claims with service dates occurring 60 days prior to the index date to 45 days after were evaluated. A medication class was considered part of the initial regimen if (a) the patient received a prescription for the drug class within 60 days before index date and received a second prescription for that drug class within 45 days after index date or (b) the patient received a prescription for the drug within 60 days before index date or 45 days after index date and had ≥30 days’ overlap between index drug class and additional drug class between index date and index date + 45 days. This definition was purposely conservative so as not to misclassify switches as part of a patient’s initial regimen. Patients with no additional medication classes meeting this definition were considered to be on monotherapy.
Analysis
In descriptive analyses, pairwise comparisons using t tests and Chi-squared tests were conducted comparing SU initiators and TZD initiators with DPP-4i initiators and saxagliptin initiators with sitagliptin initiators. Multivariable logistic regression models were fit to compare the odds of being adherent, and multivariable Cox proportional hazards models were fit to compare time to discontinuation (persistence) between index drug classes and index drugs. To determine which covariates were included in the multivariable models, descriptive statistics and statistical tests were evaluated to identify characteristics that differed between cohorts and also may be associated with adherence or persistence (i.e., were potential confounders). Models controlled for the following variables: age, sex, presence of capitated services, payer, region, urbanicity, plan type, indicator for fixed-dose metformin index drug, indicator for index drug filled via mail order, index drug regimen (monotherapy, index drug plus additional non-insulin antidiabetic drugs, index drug plus insulin), baseline total expenditures and diabetes prescription expenditures, index diabetes medication class cost-sharing, baseline endocrinologist and cardiologist visits, baseline renal impairment, baseline macrovascular and microvascular disease, pregnancy during follow-up, baseline number of unique three-digit ICD-9-CM diagnoses and Charlson Comorbidity Index. Sensitivity analyses of monotherapy and non-mail-order patients were also conducted, as patients initiating monotherapy have the simplest treatment profile and those with an initial prescription filled via mail order may have inflated persistence due to larger days’ supply on the index claim and automatic refill. P values were compared with an a priori alpha 0.05 to determine statistical significance. Analyses were conducted with SAS version 9.3 (SAS Institute Inc. Cary, NC, USA).
Discussion
In this US administrative claims-based analysis of adults with T2DM, patients who initiated a DPP-4i had significantly better adherence to and persistence with index drug class compared with patients who initiated an SU or TZD. Among DPP-4i initiators, saxagliptin patients had significantly better adherence and persistence than sitagliptin patients. These associations were observed over a 1-year follow-up period, a 2-year follow-up period, and within sub-samples of patients initiating monotherapy and filling index prescription through a non-mail-order method.
Patients may discontinue their oral antidiabetic medication following the onset of adverse events particular to their medication, such as hypoglycemia while on an SU [
25,
26]. Additionally, clinicians may discontinue a medication when the patient develops a comorbidity that places the patient at greater risk for an adverse event with a particular medication, for example, a patient at increasing risk for fractures on a TZD [
27]. Potential explanations for the observed association between initiation of DPP-4is and better adherence and persistence include a better tolerability profile compared with SUs, which are associated with weight gain and hypoglycemia, and TZDs, which are associated with weight gain and bone fracture [
3]. Differences noted between saxagliptin and sitagliptin may be due in part to a one-step dose adjustment for patients with renal impairment for saxagliptin, compared with a two-step dose adjustment for sitagliptin [
28,
29]. However, more research utilizing additional types of data sources is needed to specifically test these hypotheses for the differences in adherence and persistence between drug classes and DPP-4i-specific drugs.
An analysis conducted by Rathmann et al. [
20] comparing DPP-4is and SUs reported results consistent with the findings presented here. Utilizing data from general practices in Germany, researchers found that the risk of discontinuation was higher for DPP-4i initiators in the first 3 months and then fell below that of SU initiators for the remainder of follow-up, a maximum of 2 years [
20]. Using US health insurance claims, Curkendall et al. [
10] evaluated adherence and persistence over 1-year follow-up between patients initiating saxagliptin and patients initiating an SU, TZD, or GLP-1RA. These investigators found that patients initiating saxagliptin had greater odds of being adherent and lower risk of discontinuation than patients initiating medications within the other three drug classes [
10]. No analyses over a 2-year follow-up or within monotherapy or non-mail-order subgroups were conducted [
10]. Beyond adherence and persistence, Rathmann et al. [
20] found that compared with SUs, initiating a DPP-4i was associated with lower risk of hypoglycemia and newly diagnosed macrovascular outcomes, including coronary heart disease, myocardial infarction, and stroke.
This study has limitations that should be recognized. Administrative claims data are not collected for research purposes, and the diagnostic coding on administrative claims is recorded by healthcare providers to support reimbursement. Consequently, misclassification of certain study variables, comorbid conditions in particular, may have occurred. Similarly, unmeasured confounding may exist, as clinical variables such as HbA1c level and duration of disease, socioeconomic characteristics such as income and education, and other factors such as provider characteristics were not available for inclusion in the claims databases. As this was a retrospective, observational analysis, causal inferences should be drawn with caution. Calculations of adherence and persistence were based on the service date and days’ supply from outpatient prescription drug claims and assume that patients took the medication as directed. It cannot be determined whether patients skipped doses or discontinued the medication prior to the end of the current days’ supply. Any prescriptions obtained without generating a health insurance claim were not captured in this analysis. The reason for discontinuation (e.g., adverse events, failure to achieve treatment goals, etc.) is not available in claims databases. Lastly, the MarketScan Commercial and Medicare Supplemental databases contain data only on patients covered through employers, private commercial insurance, or employer-sponsored supplemental Medicare. Thus, findings from the study may not be generalizable to the whole US Medicare population, Medicaid population, or the uninsured.
Acknowledgments
The authors would like to thank Jerry Kagan of Truven Health Analytics for programming and data analyses. Funding for this analysis was provided by Bristol-Myers Squibb and AstraZeneca. AstraZeneca funded the article processing charges associated with this publication. Parts of this manuscript were presented previously at the Academy of Managed Care Pharmacy 26th Annual Meeting & Expo in Tampa, Florida, and the International Society for Pharmacoeconomics and Outcomes Research 19th Annual International Meeting in Montreal, Canada. All named authors meet the ICMJE criteria for authorship for this manuscript, take responsibility for the integrity of the work as a whole, and have given final approval to the version to be published.