(197) Using negative control outcomes in trial emulation to assess the comparability of patients initiating proprotein convertase subtilisin/kexin type 9 inhibitor monoclonal antibodies following myocardial infarction

Background: Proprotein convertase subtilisin/kexin type 9 inhibitor monoclonal antibodies (PCSK9i) are recommended for patients with atherosclerotic cardiovascular disease (ASCVD) and low‐density lipoprotein cholesterol (LDL‐C) exceeding target threshold. Selection of lipid lowering therapy (LLT) plays a key role of patients’ prognosis following myocardial infarction (MI) and must be accounted for when comparing treatment effects of PCSK9i in a real-world setting, by first assessing the comparability between treatment groups to obtain unbiased treatment estimates.

Objectives: To assess comparability between two PCSK9i treatment strategies: not initiating PCSK9i within 365 days post-MI (Strategy 1) and initiating a PCSK9i within 365 days post-MI (Strategy 2); and to evaluate the cardiovascular outcomes associated with PCSK9i initiation within 365 days post-MI.

Methods: MI events occurring in patients ≥18 years between 1 Jan 2015 and 31 Dec 2021 were identified in Swedish national registries. Patient characteristics and LLT use, extracted from electronic health records and the prescription registry, were analyzed by treatment strategies. Trial emulation framework was applied to compare the 3-year risk of six negative control outcomes (NCOs) with a clone-censor-weight approach to estimate the cumulative risk of each NCO between the two treatment strategies. Patients contraindicated for PCSK9i treatment (patients with cognitive disorders or with no history of statin treatment at baseline) were excluded.

Results: Of the 47,346 MI events in 42,742 patients, 561 (1.3%) patients initiated a PCSK9i within 365 days. Patients who initiated a PCSK9i were younger (median age of 67 vs. 73 years) and more likely to have dyslipidemia (68% vs. 49%), a higher baseline LDL-C (median 3.8 vs. 2.7 mmol/L) and baseline use of ezetimibe (28% vs. 7%) than patients who did not initiate a PCSK9i. The 3-year weighted risk difference (using Strategy 1 as reference group) of NCOs was -0.005 (95% confidence interval [CI]: -0.018, 0.014) for non-melanoma skin cancer, -0.006 (95% CI: -0.009, 0.000) for kidney stones, -0.018 (95% CI: -0.032, 0.001) for cancer,0.001 (95% CI: -0.008, 0.010) for glaucoma, 0.002 (95% CI: -0.027, 0.030) for fracture, and 0.005 (95%CI: -0.015, 0.027) for hip/knee arthroplasty. However, even after weighting, imbalances remained for some baseline and time-varying covariates especially for ezetimibe use and LDL-C testing.

Conclusions: The imbalances identified on the measured baseline and time-varying covariates yielded unstable estimates of counterfactual cumulative incidence for most NCOs. A comparative effectiveness study between two PCSK9i treatment strategies using the current approach would likely lead to biased results.