Benchmarking metabolic surrogate indices against body mass index for left atrial diameter in atrial fibrillation: a two-cohort study

Background Metabolic surrogate indices are increasingly used in atrial fibrillation (AF), but several algebraically incorporate body mass index (BMI). We examined whether their associations with left atrial diameter (LAD) provide information beyond a simple anthropometric benchmark. Methods We analyzed 788 patients in a primary AF cohort and 813 in a replication cohort. Seven prespecified metabolic surrogate indices were compared with BMI using harmonized linear models. We evaluated adjusted associations and incremental R² (ΔR²), partitioned shared and unique ΔR² by commonality analysis, and quantified BMI-index ΔR² differences by paired bootstrap resampling. Sensitivity analyses addressed body-size indexing, additional height/BSA covariate adjustment, AF-phenotype stratification, and covariate specification. Results Indices without BMI in their formulas—TyG, TG/HDL-C, TG/G, and CHG—added little ΔR² in either cohort (all <0.4 percentage points). SPISE, TyG-BMI, and METS-IR contributed 5.66, 4.79, and 4.45 percentage points in the primary cohort and 3.57, 3.51, and 3.35 percentage points in the replication cohort, versus 6.23 and 4.29 percentage points for BMI. Most ΔR² of the BMI-containing indices was shared with BMI, whereas index-unique components were small. Bootstrap comparisons supported numerically similar or greater ΔR² for BMI, although confidence intervals did not uniformly exclude zero. Height-adjusted raw-LAD models preserved the comparative pattern, whereas additional BSA adjustment substantially attenuated BMI-related associations, supporting a body-size/allometric component. Conclusions Indices without BMI in their formulas added little ΔR² for contemporaneous LAD, whereas BMI-containing composites derived most of their ΔR² from variance shared with BMI. BMI provided numerically similar or greater ΔR², although paired bootstrap comparisons were not statistically distinguishable for all comparisons. Additional body-size covariate analyses indicated that a substantial component of these associations may reflect body-size/allometric structure. BMI should therefore be an explicit benchmark when evaluating metabolic surrogate indices for left atrial size.