Development and validation of a novel score to predict adverse outcomes in paediatric congenital mitral regurgitation (PRIMARY): a multicentre cohort study

Background Paediatric patients who underwent surgery for mitral regurgitation (MR) have a high risk of recurrence or death; however, no prediction tool has been developed to risk-stratify this challenging subpopulation. Methods In this multicentre cohort study, paediatric patients undergoing surgery for congenital MR in Shanghai Children's Medical Center in January 1st, 2009–December 31st, 2022 were included for analysis while those had a combination with infective endocarditis, anomalous left coronary artery from the pulmonary artery, rheumatic valvular disease, connective tissue disease, or single ventricle were excluded. A Cox regression model predictive of the primary outcome (a composite of mortality or mitral valve [MV] re-operation) was derived and converted to a point-based risk score. This score was externally validated in a cohort of patients undergoing MR surgeries between January 1st, 2009 and December, 31st, 2022 in eight independent hospitals in China. The Harrell's c index and Hosmer–Lemeshow test was used to quantify the discrimination and calibration of the risk score. Findings In total, 2640 patients (female: 57% [n = 1505]) with a median age of 0.99 (IQR, 0.47–2.60) years were included. The primary outcome occurred in 262 patients (16.6%) over a median of 5.7-year follow-up in the derivation cohort (n = 1581; median age of 1.03 years [IQR, 0.47–2.67]) and in 130 (12.3%) over a median of 7.1-year follow-up in the validation cohort (n = 1059; median age of 0.93 years [IQR, 0.46–2.51]). The score variables included preoperative variables (age, the presence of primary MR, N-terminal pro-brain natriuretic peptide, left ventricular ejection fraction, and left atrium z score), perioperative changes in z scores of LA and left ventricular end-diastolic dimension, and the procedural variable (use of MV annuloplasty or not). This risk score, ranging from −22 to 10, yielded strong discrimination (Harrell's c index: 0.85, 95% CI, 0.83–0.87) and exhibited good calibration of postoperative 10-year primary outcome (Hosmer–Lemeshow; χ 2 = 9.85; P = 0.33) in the derivation cohort, which maintained in the validation cohort (Harrell's c index: 0.86, 95% CI, 0.83–0.89; Hosmer–Lemeshow; χ 2 = 23.80; P = 0.64). In addition, a simplified stratification of the score model (low-risk: −22 to −5, intermediate risk: −4 to 0, high risk: >0) showed good performances in predicting the primary outcome in both derivation and validation ( P s < 0.001). Interpretation The scoring system represents a crucial step towards personalised management care for paediatric MR. However, this model has not been applied in clinical practice and require validation in large and diverse cohorts of patients. Further work should aim to incorporate other novel multi-modality metrics to optimise the PRIMARY score. Funding Chinese National Natural Science Foundation of China, Science and Technology Commission of Shanghai Municipality, Science and Technology Project of Guizhou Province.