Background: Heparin-induced thrombocytopenia (HIT) is a high-stakes diagnosis. A delay in initiation of appropriate therapy is associated with a 6.1% daily incidence of thrombosis, amputation, and death (Greinacher et al, Blood 2000). However, misdiagnosis exposes patients without HIT to alternative anticoagulants and their attendant risk of major bleeding. Although a meta-analysis showed that the negative predictive value (NPV) of a low-probability 4Ts score is 99.8%, the positive predictive value (PPV) of the 4Ts score is limited. This is particularly true among individuals with an intermediate-probability score, in whom the PPV is only 14% (Cuker et al. Blood 2012). It has been observed that immune-mediated causes of thrombocytopenia result in a more precipitous fall in the platelet count than other common causes of hospital-acquired thrombocytopenia (e.g. infection, disseminated intravascular coagulation, drug-induced myelosuppression). Nevertheless, the pace of platelet count fall is not currently captured in the 4Ts score or other clinical prediction models for HIT. We hypothesized that rapidity of the platelet count fall could help to discriminate HIT status among patients with an intermediate probability 4Ts score. Methods: In a cohort of 292 patients with suspected HIT and prospectively calculated 4Ts scores, we identified patients with an intermediate 4Ts score (4 or 5). Patients were classified as HIT-positive or HIT-negative by an independent panel of HIT experts as previously described (Pishko et al. Blood Advances 2018). We compared the frequency of scores in each of the 4Ts score categories between HIT-positive and-HIT negative patients using Chi-square analysis. For each day from hospital admission to day of HIT laboratory testing, we extracted the first measured platelet count, exposure to heparin, and platelet transfusions. We also extracted the dates of the following events of interest, as they may be associated with an immediate fall in platelet count: cardiovascular surgery, initiation of extracorporeal membrane oxygenation, intra-aortic balloon pump placement, Impella device placement, and left ventricular assist device implantation. Two reviewers reviewed the platelet count and clinical data for each patient to determine the day on which the platelet count began to decline, excluding platelet count falls 24 hours following one of the aforementioned events of interest. From this date, we calculated the change in platelet count per each 24-hour timeframe, expressed as a percentage of the previous day's platelet count. For each subject, we identified the maximum percentage decrease in platelet count in a 24-hour period (Fallmax). We then compared the Fallmax between HIT-positive and HIT-negative subjects using the Wilcoxon-rank sum test. We assessed the operating characteristics of Fallmax for the diagnosis of HIT at different cut-offs. We selected a cut-off that maximized specificity of the metric while maintaining a sensitivity of ≥95%. Results: Among 292 patients, 158 (54.1%) had a 4Ts score of 4 or 5. Twenty-two (13.9%) were HIT-positive and 136 (86.1%) were HIT-negative. Patient characteristics are listed in table 1. No single item in the 4Ts score exhibited a significant relationship with HIT diagnosis (table 2). The median Fallmax was 49.6% (IQR 42.3-58.9) in HIT-positive patients and 38.5% (IQR 27.8-50.4) in HIT-negative patients (p=0.009). At a Fallmax cut off of ≥ 30%, the sensitivity and specificity of this measure for the diagnosis of HIT was 95.5% (95% CI 77.2%-99.9%) and 29.4% (95% CI 21.9%-37.8%), respectively (table 3). 29.4% (40/136) of the HIT-negative patients had a Fallmax below this cut-off versus only 2.4% (1/22) HIT-positive patients. Conclusion: The maximum percentage decrease in platelet count within 24-hours (Fallmax) was significantly higher in HIT-positive versus HIT-negative patients. A Fallmax ≥ 30% may be a useful metric to discriminate HIT positivity among patients with an intermediate 4Ts score. In this cohort, nearly 30% of HIT negative patients did not meet this cut-off and thus may have been spared alternative anticoagulant exposure. The rapidity of platelet count fall holds promise as a marker for improving the PPV of the 4Ts score, though it requires further evaluation and external validation. Disclosures Cuker: Synergy: Consultancy; Alexion: Other: Institutional funding on author's behalf; Bayer: Other: Institutional funding on author's behalf; Novo Nordisk: Other: Institutional funding on author's behalf; Pfizer: Other: Institutional funding on author's behalf; Spark: Other: Institutional funding on author's behalf. Pishko:Novo Nordisk: Research Funding; Sanofi Genzyme: Research Funding.
Read full abstract