Assessing treatment effectiveness when no appropriate comparator exists: low-density lipoprotein cholesterol lowering therapy (LLT) for the prevention of aortic stenosis (AS) as a case example

Background: Assessing treatment effects using real-world data when no appropriate comparator exists is challenging, as a non-user comparator group may increase confounding. There are no approved medications for AS. Genetic and cohort studies suggest that LLT may prevent AS development, but clinical trials have failed to show any benefit of LLT on AS progression.

Objectives: To examine the impact of study and analytic design choices on estimates comparing new users of LLT to a non-user group on the incidence of AS.

Methods: We used Optum Clinformatics data to identify patients at high-risk of AS (defined as hypertension, kidney disease, diabetes, and dyslipidemia). Using a target trial design, patients initiating LLT (index date) during Jan 2014 and Sept 2019 were propensity score and exact matched to ≤3 controls who did not initiate LLT but had an encounter in the index month. Patients were included if they were: 65+ years old, had ≥1 AS risk factor, and ≥12 months of enrollment prior to the index date. The controls’ index date was assigned as their visit date + lag between the LLT patient’s visit and dispensation date. Patients were followed using an as-treated approach. Fine-Gray hazards models with a 1-year landmark analysis were used to compare AS incidence among LLT patients vs controls. HRs for the ASCVD subgroup were estimated from an interaction term with LLT. Sensitivity analyses using cause-specific hazard models and an ITT approach were conducted.

Results: We identified 50,357 new LLT users and 140,404 controls (mean age 73 years, 58% women, 65% White); all covariates were well-balanced (i.e., SMD < 0.1). The majority of LLT patients were on statin. Overall, there was a small increase in AS risk among the LLT group (HR: 1.08 95%CI 1.01-1.16), which was lost in the landmark analysis (HR: 1.02 95%CI 0.93-1.11). Among those with ASCVD, LLT patients had a lower risk of AS vs controls (HR: 0.90 95%CI 0.80-1.00); this association became stronger in the landmark analysis (HR: 0.79 95%CI 0.68-0.93). Among those without ASCVD, LLT patients had increased risk of AS (HR: 1.14 95%CI 1.05-1.25); this association lost significance in the landmark analysis (HR: 1.10 95%CI 0.98-1.23). Results were consistent using the cause-specific approach, with estimates closer to the null in ITT models.

Conclusions: Stratifying by presence of ASCVD leads to different conclusions regarding the benefits of LLT on AS incidence. Stratification on treatment indication may be helpful to further elucidate potential relationships and groups benefiting from treatments for conditions without approved therapies. While this analysis revealed a potential relationship among ASCVD patients, further studies are needed to evaluate the benefit of LLTs on AS incidence.