Sequential combinations of rapid immunoassays for prompt recognition of heparin-induced thrombocytopenia Free

• A Bayesian approach combining pre-test probability with likelihood ratio of quantitative IA results allows accurate HIT diagnostic work-up

• Sequential combinations of two rapid immunoassays for HIT-antibodies perform better in comparison to the use of a single immunoassay

Early recognition and treatment of heparin-induced thrombocytopenia (HIT) are crucial to prevent severe complications. While immunoassays offer rapid diagnosis, their sensitivity and specificity are suboptimal. Sequential combinations of quantitative immunoassay results improve their diagnostic accuracy. We aimed to validate two Bayesian approaches combining two rapid immunoassays and to compare their diagnostic performance with two other diagnostic approaches ("Hamilton", "TORADI-HIT" algorithms). We included 1194 patients with suspected HIT, of whom 6.0% had confirmed HIT. HemosIL Acustar HIT-IgG (CLIA) and HemosIL HIT-Ab (LIA) (Instrumentation Laboratory, Munich, Germany) were performed. Definite HIT confirmation or exclusion was made using heparin-induced platelet activation (HIPA) test and PF4-enhanced HIPA (PIPA). Our sequential approaches (CLIA first and LIA for unsolved cases or vice versa) correctly excluded HIT in 95.6% and 96.4%, predicted HIT in 95.8% and 97.2%, with 3.3% and 2.3% of cases remaining undetermined. There were 13 respectively 15 false positive and no false negative predictions. The modified version of the "Hamilton algorithm" correctly excluded HIT in 92.1% and predicted HIT in 97.2% with 88 false positive and two false negative results. The "TORADI-HIT algorithm" correctly excluded HIT in 97.9% and predicted HIT in 93.8% (10 false positives, three false negatives). In conclusion, a Bayesian approach sequentially employing two immunoassays is accurate for HIT diagnosis. Performing immunoassays simultaneously without considering clinical pre-test probability misses HIT cases. The "TORADI-HIT algorithm" offers better HIT exclusion with a 6% false-negative rate. Using our Bayesian approach, HIT exclusion or recognition can be achieved in ≥97% of cases within <1 hour.