Rapid and accurate Bayesian diagnosis of heparin-induced thrombocytopenia.

Prompt diagnostic evaluation of suspected heparin-induced thrombocytopenia (HIT) is critical for guiding initial patient management. We assessed the performance of 3 immunoassays detecting anti-platelet factor 4 (PF4)/heparin antibodies, derived a diagnostic algorithm with a short analytical turnaround time (TAT), and prospectively validated the algorithm. Plasma samples were analyzed by Zymutest-HIA-IgG, HemosIL-AcuStar-HIT-IgG, and ID-H/PF4-PaGIA in retrospective (n = 221) and prospective (n = 305) derivation cohorts. We calculated likelihood ratios of result intervals and cutoff values with 100% negative (NPV) and positive (PPV) predictive values for a positive gold standard functional assay (heparin-induced platelet activation [HIPA]). A diagnostic algorithm was established based on the Bayesian combination of pretest probability and likelihood ratios of first- and second-line immunoassays. Cutoffs with 100% PPV for positive HIPA were >3.0 U/mL (HemosIL-AcuStar-HIT-IgG) and titer ≥16 (ID-H/PF4-PaGIA); cutoffs with 100% NPV were <0.13 U/mL and ≤1, respectively. During the prospective validation of the derived algorithm (n = 687), HemosIL-AcuStar-HIT-IgG was used as unique testing in 566 (82.4%) of 687 cases (analytical TAT, 30 minutes). In 121 (17.6%) of 687 unresolved cases, ID-H/PF4-PaGIA was used as second-line testing (additional TAT, 30 minutes). The algorithm accurately predicted HIT in 51 (7.4%) of 687 patients and excluded it in 604 (87.9%) of 687 patients, leaving only 20 (2.9%) cases unresolved. We also identified 12 (1.7%) of 687 positive predictions not confirmed by HIPA: 10 patients with probable HIT despite negative HIPA and 2 possible false-positive algorithm predictions. The combination of pretest probability with first- and second-line immunoassays for anti-PF4/heparin antibodies is accurate for ruling in or out HIT in ≥95% of cases within 60 minutes. This diagnostic approach improves initial management of patients with suspected HIT.