Real-World Cost-Effectiveness: Lower Cost of Treating Patients to Glycemic Goal with Liraglutide versus Exenatide

Data were accessed through the IMS Patient-Centric Data Warehouse, an extensive US database with linked laboratory, pharmacy and medical claims. The database contains de-identified longitudinal data, as well as clinical and demographic information (e.g., sex, age, comorbidities, insurance plan type), inpatient and outpatient claims (e.g., service charges, admission and discharge dates, procedure and diagnosis codes), laboratory tests and results, and pharmacy claims data (de-identified prescribing physician, drug dispensed based on national drug codes (NDCs), quantity and date dispensed, drug strength, days’ supply, prescription cost). The warehouse is Health Insurance Portability and Accountability Act (HIPAA) compliant. The patients for this study were selected by linking the medical and pharmacy elements of the database to the laboratory data asset to augment the number of patients with A1C laboratory values, one of two key outcome variables associated with the study’s objective.

This was a retrospective cohort study, with all patients having at least one claim for LIRA or EXEN identified between January 1, 2010 and December 31, 2010. The date of the first drug identified during this time was considered the index date and the drug was designated as the index drug. To be included in the cohort, all patients had to be at least 18 years of age on the index date, be GLP-1 and DPP-4-naïve during a 6-month pre-index period, and have at least one claim for a physician visit within the 12-month study period (6-month pre-index and 6-month post-index). Only patients who were considered persistent to the index medication during the 6-month post-index period were included in the analysis to avoid introducing bias based on historic medication use. Persistent patients included those who remained on their index drug (LIRA or EXEN) until discontinuation, medication switch, or augmentation to another index medication. Patients with evidence of gestational diabetes and/or pregnancy, Type 1 diabetes, or dopamine receptor agonists in the pre-index period or the use of insulin at any point during the study period were excluded from the analysis. Please see Table 1 for details on patients’ concomitant medications.

The A1C change from baseline and glycemic goal attainment of A1C < 7% over the 6-month post-index period for each cohort relied on available laboratory data. For the follow-up A1C measurement at the 6-month post-index point in time, the laboratory value was identified during the time period ±45 days to the 6-month post-index point in time, utilizing the mean of the laboratory values available during that time interval if several measures were present. The costs that were captured included both index drug (LIRA or EXEN) pharmacy costs and the costs of other diabetes-related pharmacy costs for concomitant medications.

All patient demographic and clinical characteristics as well as diabetes-related pharmacy costs were descriptively reported. For continuous variables, findings were presented as the mean, standard deviation (SD), and median. For categorical measures, data included the frequency [number of cases (N)] and percentage (%) of patients observed in each category. P values using the Student’s t test or Wilcoxon rank-sum test for continuous variables and the Pearson Chi square test for categorical variables were produced. A P value of <0.05 was considered statistically significant.

For the outcome measures of glycemic goal attainment of A1C < 7% and total diabetes-related pharmacy costs, multivariate analyses were performed to account for baseline and post-index differences between the two treatments of interest. The likelihood of reaching A1C goal of <7% was estimated using a logistic regression model. A generalized linear model (GLM) was developed (controlling for the same independent variables as in the logistic regression model) to estimate the total diabetes-related pharmacy costs over the 6-month post-index period. Covariates in the models included gender, plan type, pre- and post-index concomitant medications, history of diabetes-related comorbidities, and patient copayment, among other explanatory variables.

Predicted values for both diabetes-related pharmacy cost per patient and A1C < 7% goal attainment over 6-month follow-up were estimated from the multivariate regression models based on the method of recycled predictions, along with constructing 95% confidence intervals from the bootstrap distribution. This method entails comparisons of two predictive margins where a particular attribute (in this case the index treatment) is assumed present or absent.

All statistical analyses were conducted using SAS^® (version 9.2, Cary, NC, USA).

This article does not contain any new studies with human or animal subjects performed by any of the authors.

There were few significant differences when comparing the clinical characteristics between LIRA and EXEN patients (Table 2). Mean A1C at baseline was 7.8% in both groups. A greater proportion of LIRA patients resided in the south compared to EXEN (67.5% vs. 52.7%; P < 0.01). In addition, a significantly greater proportion of LIRA patients had a third party health plan type compared to EXEN (94.4% vs. 73.6%; P < 0.0001) and more EXEN patients had claims for hypoglycemia in the pre-index period compared to LIRA patients (2.2% vs. 0.0%; P = 0.04). The total average outpatient pharmacy costs in the pre-index period were higher for patients initiating LIRA compared to EXEN ($704 vs. $567; P = 0.02) (Table 1).

Prior to adjusting for confounding factors, a significantly greater proportion of LIRA patients achieved an A1C value of <7% as compared to EXEN patients (64.5% vs. 54.4%; P = 0.04) (Table 3). The difference in A1C from baseline to 6 months post-index was also significantly greater for LIRA patients compared to EXEN patients (−0.99% vs. −0.68%; P = 0.02). Over the 6-month post-index period, descriptive unadjusted total diabetes-related pharmacy costs per patient were similar between the two groups ($1,993 vs. $1,924; P = 0.376) (Table 4).

The factors associated with achieving A1C < 7% were estimated using a multivariable logistic regression model (Table 5). EXEN patients were 43.5% less likely to reach A1C < 7% compared to LIRA patients (OR = 0.565; P = 0.015). Plan type as well as certain pre- and post-index concomitant medications had a significant impact on the likelihood of reaching A1C < 7%.

The factors associated with total diabetes-related pharmacy costs over the 6-month post-index period were identified via a GLM model. After controlling for covariates such as age, gender, baseline A1C level, comorbidity history and concomitant medication use, diabetes-related pharmacy costs per patient were lower for EXEN patients compared to LIRA patients (parameter estimate = −$203.1; P = 0.0002) (Table 6). As for the logistic regression model, plan type as well as certain pre- and post-index concomitant medications had a significant impact on the estimated total diabetes-related pharmacy costs in this GLM model.

The cost per patient to achieve A1C < 7% portrays a simplistic and transparent way of assessing the short-term cost-effectiveness of the treatment alternatives. The adjusted predicted diabetes-related pharmacy costs per patient were higher for LIRA compared to EXEN (all values are mean ± standard deviation) ($2,002 ± $502 vs. $1,799 ± $502, P < 0.001); however, significantly more LIRA patients reached A1C < 7% as compared to EXEN (64.4% ± 22.4% vs. 53.6% ± 23.1%; P < 0.05) (Fig. 1). In terms of cost per patient successfully achieving A1C < 7%, this translates into a lower cost of control for LIRA compared to EXEN ($3,108 ± $779 vs. $3,354 ± $936; P < 0.0001) (Fig. 2).

These results must be viewed with the typical limitations associated with studies based on administrative claims data. The correspondence between pharmacy submission of claims and patients’ receipt and consumption of the medication was assumed and not directly measured. However, prior work suggests that medication exposure can be accurately derived from pharmacy claims [21, 22]. The study also assumed that all information needed for cohort stratification was present and similar across the cohorts of interest.

Of note, this study excluded any patients that had evidence of insulin use in either the pre- or post-index periods and cohorts were limited to consist of patients being persistent on their index therapy for a 6-month post-index period. Insulin users were excluded to remove any of the potentially additive or synergistic glucose-lowering effects of such a combination regimen, thereby focusing exclusively on the ability of LIRA and EXEN to improve glycemic control.

Although the treatment effects of LIRA 1.2 mg vs. LIRA 1.8 mg were ascertained in clinical trials, delineating LIRA 1.2 mg from LIRA 1.8 mg is a challenge in a claims database study. Exclusively relying on the NDC codes or derived dosing calculations would not completely capture patient practice realities (i.e., dosing, titration), and assigning those who titrated at some point in the 6-month follow-up period made it difficult to assign them to one category or another, though it did reflect actual clinical practice. Additionally, we did not differentiate or stratify EXEN 5 or 10 μg and patients on either dose were included in our analysis.

Despite these limitations, this study has provided valuable information regarding the real-world cost-effectiveness of LIRA compared to EXEN in the US.