Adding bilateral ankle-brachial index (ABI) to a flexible nonlaboratory clinical model did not improve prediction of ten-year death from heart or cerebrovascular disease in 6,807 NHANES adults aged 40 and older (409 events; 139 in the 2003-2004 evaluation set). AUROC was 0.834 both with and without ABI, a paired change of -0.0007 (95% CI -0.0047 to 0.0032). Brier score did not improve either, and gradient boosting plus ABI (AUROC 0.822) performed no better than regression.
This is a null result, and a useful one. ABI is a cheap, noninvasive test that reliably flags peripheral artery disease and correlates with heart attack and stroke risk, but association is not the same as added predictive value. Once age, blood pressure, smoking and similar clinical factors are modeled flexibly, ABI appears to contribute little extra information about mortality, consistent with earlier debates over whether it reclassifies risk beyond standard scores. Limits matter here: only 139 evaluation events leave wide intervals, so the data cannot prove ABI is useless, and the authors explicitly disclaim equivalence. Mortality from cardiac and cerebrovascular causes is also a narrower target than nonfatal events or leg-related outcomes. This is a preprint that has not been peer reviewed, was not preregistered, and its findings may shift. For adults, the practical takeaway is modest: ABI remains a reasonable check for suspected leg artery disease, but a normal or abnormal reading should not be treated as a substitute for standard risk-factor assessment. The work is confirmatory and incremental rather than paradigm-shifting.