QSM predicts haemorrhage risk in brainstem cavernous malformations: a multicentre prospective study

Background This study investigates the predictive value of baseline quantitative susceptibility mapping (QSM) metrics for assessing the risk of future symptomatic haemorrhages in patients with brainstem cavernous malformations (CMs). Methods From July 2020 to September 2023, a prospective multicentre cohort of 155 patients with brainstem CMs was enrolled from 12 institutions. We analysed baseline QSM metrics, including lesional mean, median, IQR and maximum susceptibility. Propensity score matching was adjusted for baseline confounders, and Cox regression models assessed haemorrhage risk. Risk stratification was performed based on thresholds determined from planned receiver operating characteristic (ROC) analyses. Results Postmatching cohorts (56 haemorrhage-free vs 30 haemorrhage cases) showed balanced baseline characteristics. Over a mean follow-up of 22.6 months, the baseline QSM metrics, particularly the median susceptibility (QSMmedian) (HR 58.896, 95% CI 8.544 to 405.989, p<0.001; Bonferroni-adjusted p=0.0001, k=4) and IQR of susceptibility (QSMIQR) (HR 29.754, 95% CI 6.101 to 145.119, p<0.001; Bonferroni-adjusted p=0.0001, k=4) were associated with prospective haemorrhage after adjusting for age, gender, lesion volume and prior haemorrhage. QSMmedian (area under curve (AUC)=0.759) and QSMIQR (AUC=0.740) demonstrated modest predictive performance. Risk stratification based on QSMmedian and QSMIQR demonstrated 2-year haemorrhage-free survival rates of 83.3%, 62.8% and 35.7% for the low-risk, intermediate-risk and high-risk groups, respectively. High-risk patients showed a 7.7-fold greater risk of haemorrhage compared with the low-risk group. Conclusions This study explored the predictive value of QSM metrics for future symptomatic haemorrhage, suggesting that QSM may serve as a complementary imaging biomarker to existing prognostic models. Further validation in larger, independent cohorts is warranted.