Magnetocardiography Combined With the SYNTAX Score for Exploratory Modeling of Clinician‐Selected Revascularization Category in Three‐Vessel Coronary Artery Disease: A Single‐Center Pilot Study

Background Three‐vessel coronary artery disease (3V‐CAD) often requires revascularization with either percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG). The SYNTAX score is widely used for anatomical risk stratification, but whether magnetocardiography (MCG) provides incremental information beyond anatomical assessment remains uncertain. Objectives To evaluate, in an exploratory pilot analysis, whether MCG‐derived parameters add information beyond the SYNTAX score for modeling clinician‐selected revascularization category in 3V‐CAD and to examine the noncausal association between model‐treatment concordance and major adverse cardiovascular and cerebrovascular events (MACCEs). Methods Prospective cohort data were analyzed retrospectively. Candidate MCG parameters were screened using Pearson correlation, random forest analysis, and stepwise logistic regression, and selected variables were combined with the SYNTAX score. Model discrimination, calibration, bootstrap internal validation, and LASSO penalized logistic regression sensitivity analysis were assessed. Model‐treatment concordance was explored using Kaplan–Meier analysis and multivariable Cox regression. Results Among 544 patients, 543 complete cases were available for model evaluation, including 42 CABG events. In the overall cohort, the combined MCG‐SYNTAX model did not materially improve discrimination compared with the SYNTAX‐only model (AUC, 0.853 vs. 0.847; p = 0.628). Discrimination was also similar in the low‐risk subgroup (SYNTAX < 22; AUC, 0.824 vs. 0.848; p = 0.540). In the intermediate‐high‐risk subgroup (SYNTAX ≥ 22), the combined model had a numerically higher apparent AUC, but this finding was considered hypothesis‐generating. Bootstrap internal validation did not support a maintained incremental value of the combined model (optimism‐corrected AUC, 0.834 vs. 0.848 for the SYNTAX‐only model). In LASSO sensitivity analysis using all appended MCG‐derived candidate variables, only the SYNTAX score was retained. The exploratory follow‐up analysis showed an unadjusted difference in MACCE‐free survival between concordance groups, but this association was not retained after multivariable adjustment and was not interpreted causally. Conclusions In this single‐center pilot modeling study, selected MCG parameters showed limited and unstable incremental value beyond the SYNTAX score for modeling clinician‐selected revascularization category in 3V‐CAD. These findings do not support clinical implementation at this stage and require external validation. Trial Registration: Chinese Clinical Trial Registry: ChiCTR2200066942