Heparin-induced thrombocytopenia (HIT), which is characterized by thrombotic events, is a serious complication of heparin use. Its diagnosis is primarily clinical but can be supported by several laboratory tests. ELISA for anti-PF4/heparin antibodies, which is the most widely-available technique, is expressed in terms of optical density (OD) results. This test was shown to have good sensitivity but poor positive predictive value. The goal of this study is to correlate OD levels with the probability of HIT diagnosis. Contrarily to previous studies where the diagnosis of HIT was mainly based on laboratory findings, we have defined HIT based on the strength of the original diagnosis, the retrospective adjudication performed by one or two clinicians familiar with HIT diagnosis, the absence of thrombosis and the absence of a clearly identified alternative diagnosis for the thrombocytopenia.
We conducted a retrospective study involving 104 patients with a positive ELISA for anti-PF4/heparin antibodies (Stago Asserachrom HPIA essay) between 2008 and May 2012. For all patients who were hospitalized at the CHUM, an extensive chart review was performed from the day of admission and for a period of 3 months following the positive ELISA assay. For each patient that was included in the study, the Greinacher clinical score was calculated. According to the clinical evolution and the laboratory results, a final, clinical, retrospective diagnosis was made for each patient (which was either HIT-positive or HIT-negative). The OD result was collected only after diagnosis was made.
In our study, 28.8% of the patients were HIT-positive and 71.2% HIT-negative. There was a statistically significant difference in ELISA results between these two groups (Figure 1). Mean OD was 0.83 (SD ± 0.62) for HIT-negative patients, versus 2.15 (SD ± 0.76) for HIT-positive ones (P< 0.001). Figure 2 shows a distribution of patients according to Greinacher score and final diagnosis. In patients with low clinical probability, HIT-positive patients had statistically higher mean OD than HIT-negative patients (3.0 ± 0.14 versus 0.66 ± 0.36, P<0.01). This was also true for patients with intermediate probability score (1.88 ± 0.78 versus 0.67 ± 0.29, P<0.01). In high probability patients, there was only one patient with negative diagnosis.
Patients were divided into 3 groups according to their OD result. There were 59 patients who had an OD of less than 1.0. Of these, only 2 (3.4%) had a positive diagnosis for HIT. In comparison, the percentage of HIT-positive patients was 45.5% (10 patients out of 22) for the group with OD 1.0–2.0, and 78.2% (18 out of 23) when OD was more than 2.0. A Roc curve (figure 3) showed that specificity of the test increased from 28% to 78% when the cut-off for OD was moved from baseline threshold to 1.0 unit. In this situation, the sensitivity of the test was only decreased to 93%.
This study is a clinical confirmation that ELISA OD results are directly correlated with the probability of a clinical diagnosis of HIT. An OD of less than 1.0 was shown to be rarely associated with clinical HIT. To the opposite, OD values above 2.0 are highly correlated with the diagnosis, and should be considered as strong evidence for clinical HIT. Furthermore, this study showed that OD results are useful when used in addition to the clinical scores. In fact, OD could discriminate between positive and negative diagnosis in patients with low and moderate clinical suspicion. Finally, the use of baseline threshold (which corresponds to OD of 0.4) was associated with a very low specificity and likelihood ratio for HIT diagnosis. By increasing the threshold to 1.0, the specificity was increased to 78.4% with a sensitivity of 93.3%. ELISA specificity and likelihood ratio were directly proportional to optic density level.
No relevant conflicts of interest to declare.
Asterisk with author names denotes non-ASH members.