Generalizing VEST trial results on wearable cardioverter-defibrillators to a Polish real-world cohort: an anchored transportability analysis
P M Mitkowski, M S Sterlinski, A P Przybylski, J W Wranicz, M G Grabowski, M K Kempa, P P Ptaszynski, M G Gierlotka, R G Gil, M M F Farkowski, O K Kowalski, A P Pawlak, A W WlodarczakAbstract
Background/Introduction
Randomised evidence for wearable cardioverter-defibrillators (WCDs) after myocardial infarction (MI) may not reflect patient characteristics and care pathways in Central and Eastern Europe, limiting the direct applicability of trial findings to local clinical practice.
Purpose
To estimate the 90-day effect of a WCD on mortality when evidence from the VEST randomised trial is transported to a Polish target population using anchored transportability analysis.
Methods
Individual participant data from VEST (1) (n=2,302; intention-to-treat) were combined with two Polish observational cohorts (pooled target n=507 adults) (2,3). The primary endpoint was all-cause death at 90 days and sudden death served as a secondary endpoint (arrhythmic death proxy). We computed stabilised inverse-odds-of-participation weights from a logistic selection model including age, sex, body mass index, and baseline left ventricular ejection fraction, with median imputation and missingness indicators; weights were trimmed at the 1st/99th percentiles. Transported risks, risk differences (RDs) and risk ratios (RRs) were estimated by weighted standardisation, with uncertainty quantified by nonparametric bootstrap (B=1,000) resampling both source and target datasets and refitting weights at each replicate.
Results
Covariate overlap was high (trial weight range 0.47–1.16; median 0.93). In the transported Polish population, 90-day all-cause mortality was 3.23% with WCD versus 5.03% with control (RD −1.80 percentage points [pp]; 95% bootstrap CI −4.81 to 1.14; RR 0.64; 95% CI 0.33–1.37). Sudden death occurred in 1.50% versus 2.15% (RD −0.65 pp; 95% CI −2.85 to 1.18; RR 0.70; 95% CI 0.26–2.17). A sensitivity analysis restricted to VEST participants recruited in Central/Eastern Europe (n=632) was directionally consistent (RD −2.28 pp; 95% CI −5.81 to 0.70; RR 0.52; 95% CI 0.18–1.24).
Conclusion(s)
Transporting VEST evidence to a real-world Polish cohort yielded mortality estimates directionally consistent with the original trial, supported by strong covariate overlap and stable transportability weights. Confidence intervals crossing the null reflect limited local sample size rather than absence of effect. This transportability framework enables context-specific inference and can inform WCD implementation and policy decisions where local randomised evidence is unavailable.