Abstract
Many patients require glucagon-like peptide-1 receptor agonists (GLP-1 RA) and basal insulin for glycemic control. Only two fixed-ratio combination products are available: insulin degludec/liraglutide (iDergLira) and insulin glargine/lixisenatide (iGlarLixi). Our prior study using real-world data found that iGlarLixi was associated with fewer cardiovascular events versus iDegLira. This study sought to estimate the clinical and economic outcomes in a primary prevention cohort. Data were assessed through TriNetX (Cambridge, MA, USA), a global health research network. Two cohorts were created to assess patients with type 2 diabetes using either iDegLira or iGlarLixi. Patients with prior myocardial infarction (MI), heart failure (HF), stroke, or stage-4 or worse chronic kidney disease were excluded. The index date was defined as a prescription generated for either iGlarLixi or iDegLira after diagnosis of type 2 diabetes, and the observation window for the outcomes was set to 1,460 days. 1:1 propensity score matching was used to balance the cohort using gender, race, prandial insulin, glycated hemoglobin, and body weight. The primary clinical outcome was the composite of MI, stroke, and HF. We calculated total disutilities and cost by year. Propensity score matching resulted in two well-balanced cohorts (n = 2,805 each). The primary composite outcome occurred more frequently with iDegLira versus iGlarLixi (risk ratio = 1.87 [95% CI: 1.58, 2.22]). Disutilities and costs favored iGlarLixi versus iDegLira. Based on discounted utilities and costs, the average per-patient-year disutilities for iGlarLixi were roughly half that versus iDegLira (0.004 vs. 0.010), and total costs were roughly one-third less for iGlarLixi.In patients with type 2 diabetes, iGlarLixi was associated with a lower disutility and cost versus iDegLira. The study confirms and aligns with our prior real-world analysis of iGlarLixi versus iDegLira, though herein limited to cardiovascular primary prevention. A head-to-head clinical trial would be needed to confirm these results.
Introduction
Fixed-ratio injectable combinations of basal insulin and glucagon-like peptide-1 receptor agonists (GLP-1 RA) simplify adherence while achieving complementary clinical effects in the management of type 2 diabetes.1,2 Insulin glargine and insulin degludec have been compared head-to-head in a cardiovascular outcome trial, and insulin degludec was found to be noninferior. However, cardiovascular outcome trials for liraglutide and lixisenatide were completed separately, each versus placebo.1,3 These trials demonstrated a cardiovascular benefit with liraglutide versus placebo and a neutral effect with lixisenatide versus placebo. No head-to-head trials exist that compare iGlarLixi (insulin glargine U-100/lixisenatide) and iDegLira (insulin degludec U-100/liraglutide). Our previous real-world study found iGlarLixi was associated with fewer cardiovascular events versus iDegLira. 4 The difference between our real-world findings and prior placebo-controlled cardiovascular outcome trials may be due to study design. In addition, real-world studies are subject to more variability in adherence, prescribing, and dosing such that each GLP-1 RA studied may not reach the same exposure in a real-world study as in a controlled trial. The present study sought to confirm our previous real-world findings comparing iGlarLixi and iDegLira in a primary prevention cohort and to estimate the economic outcomes between these agents.
Materials and Methods
Data Source
Data were assessed through TriNetX (Cambridge, MA, USA), a health research network of electronic medical records. Data included demographics, diagnoses/procedure codes, medications, and laboratory values. TriNetX aggregates counts and statistics of de-identified data. No protected health information is available to investigators; thus, institutional review board approval was not required.
Patient Selection
Two cohorts were created for adults with type 2 diabetes, age ≥25 years, using iGlarLixi or iDegLira. Exclusion criteria were type 1 diabetes; prior use of tirzepatide or GLP-1 RA other than liraglutide or lixisenatide; myocardial infarction (MI), heart failure (HF), stroke, or stage-4 or worse chronic kidney disease prior to index. The iGlarLixi cohort excluded prior use of insulin degludec, and the iDegLira cohort excluded prior use of insulin glargine. The query covered 21 November 2016, the date iGlarLixi and iDegLira received Food & Drug Administration approval, through 13 December 2023. Index date was defined as the prescription for iGlarLixi or iDegLira after the type 2 diabetes diagnosis. Baseline data were generated from the most recent data in a 6-month look-back. The observation window for outcomes was 1,460 days post-index.
Outcomes
The primary clinical outcome was the composite of the first event of MI, HF, and stroke. For economic outcomes, cohorts were assumed to have the same baseline utilities. We calculated total disutilities by year across patients based on incident events in that or prior years. Disutilities are based on the concept of Quality of Life Scales, where patient well-being or utility is measured from zero (dead) to one (perfect health). This measure allows for the comparison of health states across different diseases and treatments and is used in cost-utility analyses to assess the economic value of alternative treatments. In our analysis, disutilities reflect reductions in quality of life associated with MI, HF, and stroke. We assumed the same disutility in the year in which the event occurred and following years. 5 For patients experiencing more than one event in our analysis, we followed a “minimum utility” approach where the maximum disutility of the events in question was used. Disutilities were based on prior research evaluating type 2 diabetes. 6
Management costs were those of iGlarLixi or iDegLira. Direct costs were those associated with events. Management cost reflected the annual cost of the drug based on the first year of our analysis (2016) and the wholesale acquisition cost (Tables 1 and 2). Given our assumption of no patient deaths during the analysis period, every patient is assumed to incur the management cost in each year. Direct costs were based on 2010 prices and are specified as an initial cost that reflects the year in which the event occurred and a fixed follow-up cost in each of the remaining years of the analysis period. 8 We assumed follow-up costs did not vary regardless of whether a patient experienced multiple events of the same type in a single year. The costs were inflated to 2023 dollars based on the medical care component of the Consumer Price index. A discount rate of 3% was applied to disutilities and costs (Tables 1 and 2).7,9
Cost Inputs and Assumptions Associated with Complications
costs were based on the prices of year 2010
HF, heart failure; MI, myocardial infarction; $, United States dollars; %, percentage.
Utility-Disutility Input Values and Assumptions Associated with Complications
HF, heart failure; MI, myocardial infarction.
Statistical Analysis
Analyses were conducted within TriNetX. To mitigate confounding, a 1:1 propensity score matching was used to balance cohorts. Propensity scores were calculated using logistic regression and a greedy nearest neighbor matching algorithm with a caliper of 0.1 pooled standard deviations. Variables included in the match were gender, race (White, Black, or African American), insulin aspart, insulin lispro, regular insulin, glycated hemoglobin, and body weight. Clinical outcomes were calculated using logistic regression to generate risk ratios (RRs) with a 95% confidence interval. The statistical methodologies and software suites employed through TriNetX include Java 11.0.16, R 4.0.2, and Python 3.7. A p < 0.05 was considered statistically significant. Utility estimates for clinical outcomes were calculated using aggregate counts from TriNetX and then analyzed in Microsoft Excel using descriptive statistics.
Results
Baseline Characteristics
A total of 114,997,604 patients from 80 health care organizations were available on the TriNetX research network when the analysis was conducted. The query identified 6,757 patients meeting the inclusion and exclusion criteria (iGlarLixi, N = 3,374; iDegLira, N = 3,383). Propensity score matching resulted in a sample size of 5,610 (N = 2,805 in each cohort).
The population was predominantly White and had a mean glycated hemoglobin of ∼9%, a systolic blood pressure of ∼130 mmHg, and a diastolic blood pressure of ∼75 mmHg (Table 3). Body mass index was slightly higher in the iGlarLixi cohort, though body weight was higher in the iDegLira cohort. These differences were not significant. About half were prescribed metformin, a statin, and either an angiotensin-converting enzyme inhibitor or an angiotensin receptor blocker (Table 1).
Comparison of baseline patient characteristics before and after propensity score matching
ACE, angiotensin converting enzyme; BMI, body mass index; DBP, diastolic blood pressure; DPP-4, dipeptidyl-peptidase 4; HbA1c, hemoglobin A1c; iDegLira, insulin degludec/liraglutide; iGlarLixi, insulin glargine/lixisenatide; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; SBP, systolic blood pressure; SD, standard deviation; SGLT-2, sodium-glucose cotransporter 2.
Cardiovascular Outcomes
The primary composite outcome occurred more frequently with iDegLira n = 345 (12.2%) versus iGlarLixi n = 184 (6.5%) (RR = 1.87 [95% CI: 1.58, 2.22]). All three components of the primary outcome favored iGlarLixi, for HF iDegLira n = 232 (8.2%) versus iGlarLixi n = 108 (3.8%) (RR = 2.14 [95% CI: 1.72, 2.68]), for MI iDegLira n = 78 (2.7%) iGlarLixi n = 44 (1.5%) (RR = 1.77 [95% CI: 1.22, 2.55]), and for stroke iDegLira n = 92 (3.2%) iGlarLixi n = 54 (1.9%) (RR = 1.70 [95% CI: 1.22, 2.37]).
Cost and Disutilities
Table 4 shows the estimated costs and utilities for the two cohorts. The disutilities, management costs, and direct costs were lower for the iGlarLixi cohort as compared to the iDegLira cohort. Based on discounted utilities and costs and 2023 prices, the average per-patient-year disutilities for iGlarLixi were roughly half that as compared with iDegLira (0.004 vs. 0.010). Total costs were roughly one-third less for iGlarLixi, while the contributions of management and direct costs to total costs were similar across the two cohorts.
Estimated costs and disutilities for iDegLira and iGlarLixi
iDegLira, insulin degludec/liraglutide; iGlarLixi, insulin glargine/lixisenatide.
Discussion
Our findings indicate lower disutility and cost with iGlarLixi versus iDegLira, suggesting greater real-world cost-effectiveness with iGLarLixi versus iDegLira. In addition, iDegLira versus iGlarLixi was associated with more MI, strokes, and new diagnoses for HF, confirming our prior analysis, though herein a primary cardiovascular prevention cohort. 4
Prior cost-effectiveness analyses of iDegLira versus iGlarLixi have been evaluated from the Czech Republic and Italian perspective.10,11 Results from the Czech Republic payer perspective found that iDegLira was associated with a gain of 0.14 QALYs as compared with iGlarLixi, with clinical benefits being driven by a reduction in incidence and delayed onset of diabetes-related complications. 10 Similarly, in the Italian perspective, iDegLira was associated with a gain of 0.13 QALYs relative to iGlarLixi, also driven by a delayed onset of diabetes-related complications. 11 In a United Kingdom analysis comparing iGlarLixi with iDegLira, total QALYs gained were found to be identical, and given that iGlarLixi has a lower annual acquisition cost, it suggests iGlarLixi represents a good value. 12 A key difference is the present analysis did not account for cost or disutility associated with microvascular outcomes. Differences in the observed outcomes may also be due to differences in dose, exposure to each agent, or patient characteristics. Furthermore, some studies may utilize modeled simulations rather than direct real-world outcomes.
Limitations include the data being de-identified and aggregated, making it impossible to follow individual patients longitudinally. Therefore, estimates are based on by-year cohort-level calculations to account for the proportion of patients with each possible combination of outcomes: none, MI-only, stroke-only, HF-only, MI-stroke, MI-HF, HF-stroke, and MI-HF-Stroke. While this approach optimizes the utility of a restricted dataset, applying a fixed disutility in every year following an event likely overestimated distutilties, as it precludes utility recovery after the event. Our assumption that each event had its own sequence of follow-up costs may overestimate costs, as some patients may have multiple events concurrently, and their true follow-up costs were less than the sum of the events had the occurred separately. However, given the benefit with iGlarLixi identified in clinical outcomes, not accounting for multiple of the same event or death may underestimate the difference between iGlarLixi and iDegLira. Finally, while the observed effect of propensity score matching balancing cohorts is encouraging, residual confounding by indication cannot be excluded, and patients at higher cardiovascular risk may have been prescribed iDegLira rather than iGlarLixi. Confounding by indication, if present, may have occurred through imprecision in measured baseline characteristics or differences in unmeasured baseline characteristics not accounted for in propensity score matching. In future studies, inclusion of microvascular outcomes may allow for a more comprehensive cost-effectiveness estimate.
Conclusions
In conclusion, among patients with type 2 diabetes, iGlarLixi was associated with lower disutility and cost versus iDegLira. This real-world finding aligns with our prior findings that iGlarLixi versus iDegLira is associated with a lower risk of MI, HF, and stroke, though now also in a primary cardiovascular prevention cohort. 4 A prospective, head-to-head, randomized controlled trial would be needed to confirm these findings.
Authors’ Contributions
Conceptualization, K.C. and N.W.C.: Methodology, all authors; formal analysis, T.Q. and N.W.C.: Investigation, all authors; data curation, N.W.C.: Writing—original draft preparation, T.Q. and N.W.C.: Writing—review and editing, all authors; funding acquisition, K.C. and N.W.C. All authors have read and agreed to the published version of the article.