APTTO: a clot waveform analysis-based algorithm for diagnostic triage of prolonged activated partial thromboplastin time.

BACKGROUND

Prolonged activated partial thromboplastin time (APTT) is a common laboratory finding that frequently triggers extensive diagnostic testing and may delay clinical decision-making, particularly in the preoperative setting. However, many patients have no clinically relevant bleeding disorder. Clot waveform analysis (CWA), generated during routine APTT testing, provides dynamic information on coagulation but remains underused in prolonged APTT.

OBJECTIVES

To develop and temporally validate a CWA-based algorithm to assist in the evaluation of prolonged APTT.

METHODS

In a multicenter cohort with prolonged APTT and normal prothrombin time, qualitative and quantitative CWA features were extracted. Two logistic regression models were developed: APTTO1 to identify patients unlikely to have etiological findings, and APTTO2 to differentiate lupus anticoagulant (LA) from intrinsic pathway factor deficiency or von Willebrand disease (VWD) among higher-risk cases. Discrimination, calibration, and decision-curve analysis (DCA) were evaluated using a temporally independent cohort.

RESULTS

APTTO1 demonstrated good discrimination (AUC 0.88) and calibration, providing 99.5% sensitivity and 95.2% negative predictive value (NPV) at a prespecified threshold. APTTO2 showed good discrimination for differentiating LA from factor deficiency or VWD (AUC 0.86 after recalibration). DCA demonstrated net clinical benefit of both models over investigate-all and investigate-none strategies. The two-step APTTO algorithm reduced the proportion of patients requiring etiologic testing while maintaining a high NPV, supporting its potential to streamline diagnostic pathways.

CONCLUSIONS

APTTO is a pragmatic, CWA-based, two-step algorithm that demonstrated robust performance for the evaluation of prolonged APTT and may support more efficient diagnostic triage. Prospective external validation is required before clinical implementation.