Applicabilità, performance e potenzialità prognostiche degli scores di rischio poligenico nella cardiomiopatia dilatativa: risultati da due centri europei di riferimento per le cardiomiopatie

Background Polygenic risk scores (PRSs) for dilated cardiomyopathy (DCM) have been recently developed. Yet, robust data are lacking about their real-world applicability and performance. Moreover, their potential prognostic value has never been investigated. Purpose To evaluate the diagnostic accuracy, the interplay with monogenic variants and the association with DCM-related outcomes of PRSs in real-world DCM patients. Methods This was a retrospective observational multicenter cohort study including patients from two European tertiary referral centres for DCM, analyzed by next-generation sequencing (NGS) and microarray analysis (Illumina platform-USA). We tested the main PRSs for DCM developed to date: Pirruccello et al. left ventricular (LV) body-surface-area indexed (BSAi) LV end-diastolic volume PRS (LVEDVi-PRS), BSAi LV end-systolic volume PRS (LVESVi-PRS), BSAi stroke volume PRS (Svi-PRS), LV ejection fraction PRS (LVEF-PRS); Tadros et al. DCM-PRS; Zheng et al. PGS004861 and PGS004862; Jurgens et al. DCM_GWAS_allcohorts and DCM_MTAG_allcohorts. PRSs were computed for patients using the plink2 function–score and compared with local control populations of unselected individuals accessing NGS due to health-related concerns for their children. Diagnostic accuracy was assessed using receiver operating characteristic curves-area under the curve (AUC). The following composite outcomes were assessed in an individual patient data meta-analysis of the two cohorts: 1) all-cause mortality; 2) heart failure (HF)–related death, heart transplantation, or destination left ventricular assist device implantation (DHF/HTx/VAD); and 3) sudden cardiac death/sustained ventricular tachycardia/ventricular fibrillation (SCD/VT/VF). Results 905 patients were included: 334 from Trieste (39% variant-positive, 46±13 years, 70% males) and 571 from Maastricht (19% variant-positive, 53.5±12.5 years, 64% males). More than 95% of patients and controls were Caucasians. Except for SVi-PRS, all the PRSs were found to be significantly higher in patients than controls in both cohorts (P<0.0001). PGS004861 and PGS004862 showed the best diagnostic accuracy, particularly in the Maastricht cohort (AUC 95%CI 0.71 [0.68 – 0.74] and 0.72 [0.69 – 0.75], respectively), conferring an odd-ratio of being affected > 2-fold per 1 PRS-standard deviation (SD). LVEDVi-PRS was found to be significantly higher in variant-negative (not carrying pathogenic/likely pathogenic [P/LP] monogenic variants) compared to variant-positive Maastricht patients (P<0.05), while no differences emerged for other PRSs in either cohort. In a cause−specific Cox model adjusted for age, sex and other confounders: 1) higher LVESVi-PRS and DCM_MTAG_allcohorts were significantly associated with a lower occurrence of SCD/VT/VF (pooled HR 0.79, 95% CI 0.65 – 0.96, P=0.017 and pooled HR 0.79, 95% CI 0.65 – 0.96, P=0.016, respectively, per 1-SD PRS increase); 2) LVEF-PRS was significantly associated with a higher occurrence of DHF/HTx/VAD in the variant-positive patients (pooled HR 1.89, 95% CI 1.17 – 3.03, P=0.009 per 1-SD PRS increase). Conclusions PRSs for DCM confirmed a good accuracy in predicting disease in real-world scenario. They might also influence DCM-related outcomes, with impactful —despite different— contribution in variant-positive and variant-negative patients. This work constitutes a proof of principle for clinical implementation of PRSs, improving our diagnostic and prognostic tools.