Persistent left ventricular systolic dysfunction at ≥6 months after SARS-CoV-2–associated MIS-C: a systematic review and rare-event meta-analysis

Résumé

Background Multisystem inflammatory syndrome in children (MIS-C) associated with SARS-CoV-2 frequently involves acute myocardial dysfunction. However, the prevalence of persistent left ventricular (LV) systolic dysfunction at longer-term follow-up remains uncertain. We therefore performed a systematic review and meta-analysis to estimate the prevalence of persistent LV systolic dysfunction at ≥6 months after MIS-C. Methods We included observational cohort studies reporting longitudinal echocardiographic follow-up in children and adolescents aged <21 years with SARS-CoV-2–associated MIS-C. The prespecified primary outcome was persistent LV systolic dysfunction, defined as left ventricular ejection fraction (LVEF) < 55% at ≥6 months. Study-level events and denominators were extracted and harmonized using the number of children who underwent ≥6-month echocardiography as the endpoint-specific denominator. Pooled prevalence was estimated using an intercept-only binomial logistic-normal generalized linear mixed model (GLMM) fitted directly to study-level event counts and denominators without continuity correction. The originally prespecified random-effects logit model with restricted maximum likelihood estimation and Hartung–Knapp–Sidik–Jonkman confidence intervals, together with alternative continuity corrections and transformation-based approaches, was retained as sensitivity analysis under sparse-event conditions. Results Ten cohorts met the inclusion criteria and contributed to quantitative synthesis, comprising 432 children with echocardiography at ≥6 months. Persistent LV systolic dysfunction occurred in 3 children, corresponding to a crude observed proportion of 0.69% (3/432). In the revised primary continuity-correction-free binomial logistic-normal GLMM, the pooled prevalence was 0.69% (95% CI, 0.22%–2.13%). Given that only three events were observed, this model-based estimate should not be interpreted as a precise estimate of the underlying population prevalence. Continuity-corrected sensitivity models yielded pooled estimates from 1.62% to 2.23%, further demonstrating numerical sensitivity to the treatment of zero-event cohorts. Conclusion Persistent LV systolic dysfunction was infrequently observed among children with MIS-C reassessed by echocardiography at ≥6 months. However, only three events were identified, so the available evidence is insufficient to determine the exact population prevalence with precision. Inference is further limited by heterogeneous follow-up and potential selection bias from nonuniform late imaging. Larger prospective cohorts with standardized long-term surveillance are needed.