Coagulation and Transfusion Medicine / OD LEVELS AND SCORING SYSTEMS FOR HIT

Transcription

1 Coagulation and Transfusion Medicine / OD LEVELS AND SCORING SYSTEMS FOR HIT Usefulness of Optical Density Values From Heparin Platelet Factor 4 Antibody Testing and Scoring Models to Diagnose Heparin-Induced Thrombocytopenia Kim A. Janatpour, MD, 1 Robert C. Gosselin, CLS, 1 William E. Dager, PharmD, 2 Andrew Lee, PharmD, 2 John T. Owings, MD, 3 Jufen Zhou, 5 and Ted Wun, MD 1,4,6 Key Words: Thrombocytopenia; Thrombosis; Heparin; Optical density; Enzyme-linked immunosorbent assay; ELISA DOI: /RPE753J4PMG9773Q Abstract Heparin-induced thrombocytopenia (HIT) is a complication caused by antibodies directed to the heparin platelet factor 4 (PF4) complex with a seemingly paradoxical high risk of thrombosis. Discontinuation of heparin and administration of an alternative anticoagulant is important in prevention of catastrophic thrombosis. Diagnosis is challenging and based on clinical probability models (Warkentin 4 Ts and Chong scale) and, to a lesser degree, laboratory testing. Enzyme-linked immunosorbent assay (ELISA) measurement of heparin-pf4 antibodies is commonly used but has low predictive values for thrombosis. We analyzed 105 cases of suspected HIT and compared optical density values and the Warkentin 4 Ts for sensitivity, specificity, positive predictive value, and negative predictive value (NPV). The predictive value of ELISA alone was inferior to the Warkentin 4 Ts score. The sensitivity and NPV of the clinical score was improved by incorporating ELISA results. The combination of a negative ELISA result with low probability 4 Ts resulted in an NPV of 94%. Heparin-induced thrombocytopenia (HIT) is a potentially serious complication of heparin administration caused primarily by antibodies directed against the heparin platelet factor 4 (PF4) complex resulting in platelet activation, generation of thrombin, and a hypercoagulable state. Patients with HIT typically have thrombocytopenia 5 to 10 days after exposure to heparin. Seemingly paradoxical, the development of thrombocytopenia in patients with HIT is associated with venous 1 or arterial 2 thrombosis. The early recognition of HIT is vital because failure to discontinue heparin and initiate alternative anticoagulation can result in increased morbidity (limb loss) and possible mortality. The diagnosis of HIT is difficult and is primarily based on clinical suspicion along with the demonstration of heparin-pf4 antibodies by serologic testing. However, not all patients with anti heparin-pf4 antibodies develop HIT (either thrombocytopenia or thrombosis), 2 and, conversely, the absence of detectable anti heparin-pf4 does not completely exclude the disorder. 1,3 A number of laboratory tests are available for heparin-pf4 antibody detection, but they vary in sensitivity, specificity, technical difficulty, and availability. The serotonin release assay (SRA) is generally reported to have relatively high specificity but is technically difficult, uses radioactive reagents, and is not widely available. 3 ELISA tests for heparin-pf4 antibodies are widely available and highly sensitive but not as specific as the SRA for clinical HIT. Recent data from Zwicker and colleagues 4 have suggested that the ELISA anti heparin-pf4 optical density (OD) level may improve the predictive value of the test for thrombosis. The purpose of the present study was to retrospectively evaluate the predictive value of the OD value on ELISA testing for anti heparin-pf4 antibodies for HIT, used alone and Am J Clin Pathol 2007;127: DOI: /RPE753J4PMG9773Q 429

2 Janatpour et al / OD LEVELS AND SCORING SYSTEMS FOR HIT in combination with clinical probability scores developed by Warkentin and Heddle 5 and Chong and Chong. 6 Materials and Methods Patient Cohort and Scores After approval by our institutional review board on human subject testing, we retrospectively identified and reviewed the clinical and laboratory records of patients in whom the clinical suspicion of HIT was deemed high enough to initiate treatment with a direct thrombin inhibitor (DTI). Baseline and nadir platelet counts, heparin exposure history, and serologic testing for the presence of anti heparin-pf4 antibodies were recorded. Retrospective probability scores were calculated by two of us (W.E.D. and A.L.) using the Warkentin 4 Ts pretest probability model 5 and the Chong HIT classification system. 6 Anti Heparin-PF4 ELISA Each patient had at least 1 serologic test for measuring antibodies to the heparin-pf4 complex (GTI, Waukesha, WI) performed according to the manufacturer s instructions. Briefly, polyvinylsulfonate-pf4 is the surrogate target antigen, and antibodies present in patient serum will bind to microwells coated with polyvinylsulfonate-pf4. After a short incubation, the excess serum is washed and a polyclonal antihuman conjugate added that will bind to IgG, IgM, and IgA isotype antibodies. Unbound material is removed, and a chromogen is added to label the bound conjugate, producing a yellow color, read at 405 nm. The amount of yellow produced at the end point, as indicated by the OD, is proportional to the amount of antibody present. A positive result is reported if the OD reading is or more. 4 Statistical Analysis Kruskal-Wallis analysis of variance (ANOVA) on ranks was used to determine differences in platelet counts between HIT probability categories. The Pearson correlation was used to compare the Warkentin 4 Ts pretest probability scores and the Chong classification scores. Kruskal-Wallis 1-way ANOVA on ranks and multiple pairwise comparisons using the Dunn test were used to compare the heparin-pf4 antibody ELISA OD test results between clinical probability categories. Sensitivity, specificity, and positive and negative predictive values for HIT were determined for the Warkentin 4 Ts score and heparin-pf4 antibody ELISA test OD levels. The Chong categories of definite and probable or unlikely were used as the gold standard for defining or excluding HIT disease. An OD cutoff of or more and the manufacturer s OD threshold of were evaluated. Results We retrospectively analyzed the data from 114 consecutive patients who received a DTI for a presumed diagnosis of HIT. Of these, complete clinical and laboratory data were retrievable for 105 patients. There were 55 men (52.4%) enrolled in our study, with a median age of 60 years (range, years). The majority of patients were managed by the internal medicine (45 [42.9%]), cardiovascular/thoracic surgery (26 [24.7%]), or trauma (13 [12.4%]) service. According to the Chong classification, 34 (32.4%) of 105 patients were defined as having definite HIT, 26 (24.8%) as having probable HIT, 23 (21.9%) as having possible HIT, and 22 (21.0%) as unlikely to have HIT. For 23 patients, multiple laboratory tests had been done for HIT antibodies. Of these, 8 patients (35%) seroconverted from negative to positive, results for 9 (39%) remained negative on repeated testing, and results for 6 (26%) remained positive on repeated testing. Patients in whom seroconversion occurred were deemed to have positive results for purposes of determining the Chong score, which incorporates the presence of antibody and platelet recovery into the diagnosis of HIT. According to categories derived by using the Warkentin 4 Ts scoring system, there were 36 patients with low, 48 patients with intermediate, and 21 patients with high pretest probabilities for HIT. The Kruskal-Wallis ANOVA on ranks revealed no significant difference between baseline (P =.08) or nadir (P =.24) platelet counts between definite, probable, possible, and unlikely Chong classifications. There were no significant differences in baseline (P =.18) or nadir (P =.37) platelet counts between the Warkentin categories. Comparison of Chong and Warkentin 4 Ts Scores There was good correlation (Pearson correlation, R 2 = 0.62) between all Warkentin 4 Ts pretest probability and Chong posttest scores. In comparing likelihood categories, of the 60 patients with definite or probable HIT using the Chong scoring system, 21 (35%), 34 (57%), and 5 (8%) demonstrated high, intermediate, and low Warkentin 4 Ts pretest probabilities, respectively. Of the 22 patients unlikely to have HIT by the Chong classification, 20 had a low pretest probability, but 2 had intermediate pretest probability. The positive predictive value of a high 4 Ts score was 100%. Scoring Systems and Heparin-PF4 Antibody ELISA OD Test Results scoring systems and heparin-pf4 antibody ELISA OD test results are shown in Figure 1. The OD was significantly different between the 4 Ts categories of high, intermediate, and low probability (P =.003; Kruskal-Wallis 1- way ANOVA on ranks). There was a significant difference between the categories of high vs low and intermediate vs low (P <.05; Dunn test) but not between the categories of high and 430 Am J Clin Pathol 2007;127: DOI: /RPE753J4PMG9773Q

3 Coagulation and Transfusion Medicine / ORIGINAL ARTICLE A B OD, 405 nm Low Intermediate High OD, 405 nm Unlikely Possible Probable Definite Figure 1 Relationship between optical density (OD) readings with (A) Warkentin 4 Ts and (B) Chong scale scoring systems. Bars represent median OD level. A, P <.05, low vs high, and low vs intermediate probability, Kruskal-Wallis, Dunn test. B, P <.05, definite vs unlikely, definite vs probable, and definite vs possible probability, Kruskal-Wallis, Dunn test. intermediate. The OD was also significantly different between the Chong categories of definite, probable, possible, and unlikely for HIT (P.001; Kruskal-Wallis 1-way ANOVA on ranks). There was a significant difference between the categories of definite vs unlikely, definite vs probable, and definite vs possible (P <.05; Dunn test) but not between the categories of probable vs possible. Sensitivity, Specificity, and Predictive Values The specificity for definitive or probable HIT according to Chong scoring was highest at an OD level of or more but at the expense of sensitivity and predictive values Table 1. The 4 Ts probability score showed better sensitivity, specificity, and predictive values than ELISA OD levels used alone. However, the sensitivity and negative predictive value of the 4 Ts score were improved using the 4 Ts + OD threshold of less than as a negative cutoff value. Using the Warkentin 4 Ts pretest probability in combination with a positive ELISA result improved the sensitivity from 81% to 94% and the negative predictive value from 80% to 94%. Discussion The diagnosis of HIT remains challenging; it is based on clinical and laboratory parameters that are temporally associated with the administration of heparin, including the development and timing of thrombocytopenia, the presence of heparin-pf4 antibodies, evidence of new or worsening thrombosis, and exclusion of other causes of thrombocytopenia. 5 Several scoring systems have been proposed to assist in the diagnosis of HIT. 5,6 The scoring system proposed by Chong and Chong uses data obtained when the diagnosis is first being considered and later when additional clinical and laboratory data become available after heparin is discontinued. 6 Thus, the Chong system is a useful tool to define HIT in retrospect and, perhaps, to minimize the duration of alternative anticoagulant therapy, but it does not provide guidance on the decision of whether to discontinue heparin. The 4 Ts clinical probability score proposed by Warkentin and Heddle 5 is a pretest probability score originally designed to predict a positive SRA result. 7 The use of the 4 Ts score to make treatment decisions Table 1 Comparison of Sensitivity, Specificity, and Predictive Values for 4 Ts and OD Levels, Alone and in Combination * Sensitivity (%) Specificity (%) NPV (%) PPV (%) 4 Ts score (high and low) OD level Ts + OD level Ts + OD level NPV, negative predictive value; OD, optical density; PPV, positive predictive value. * For calculations, the Chong categories of definite/probable and unlikely were used to define the presence or absence of heparin-induced thrombocytopenia. Am J Clin Pathol 2007;127: DOI: /RPE753J4PMG9773Q 431

4 Janatpour et al / OD LEVELS AND SCORING SYSTEMS FOR HIT had not been prospectively validated. Prospective validation of scoring systems or laboratory tests is difficult because there is currently no gold standard for diagnosis of HIT. We found that the Warkentin 4 Ts scoring system for high and low probability correlates well with the Chong posttest categorization system. For all patients studied, anticoagulant therapy was changed to a DTI. However, 22 cases were retrospectively categorized as unlikely to be HIT according to the Chong score. Of these, 20 were also categorized as low probability by the Warkentin 4 Ts probability score. Alternative anticoagulation might have been safely withheld in these patients. Five of 60 patients with a low 4 Ts pretest probability were categorized as having definite or probable HIT based on the Chong posttest analysis. Therefore, if a low 4 Ts score is used to withhold therapy with a DTI, a small percentage (8% in our study) might have treatment withheld inappropriately. In our group with an intermediate 4 Ts pretest probability, 34 (71%) of 48 corresponded to a Chong category of definite or probable. Thus, alternative anticoagulation should strongly be considered in patients with a 4 Ts score in the intermediate range. The 4 Ts scoring system does not include anti heparin-pf4 antibody results. Although the OD levels were statistically different between the clinical categories within both scoring systems, there was significant overlap. We found that adding the ELISA heparin-pf4 test results improved the sensitivity and negative predictive value over using the 4 Ts system alone. The negative predictive value was increased from 80% to 94%. Therefore, using the combination of ELISA results and the 4 Ts score would potentially result in only 6% of patients in whom therapy should have been changed but was not. Thus, in patients with a low 4 Ts pretest probability, a negative antibody test result improves the confidence to safely withhold treatment and may minimize the unnecessary use of DTIs. Anticoagulant therapy changes are often made without the benefit of laboratory test results in patients with suspected HIT. In part, this may be due to limited availability for laboratory tests or long turnaround times. The SRA has a 100% positive predictive value for clinical HIT, 8 but because this test is offered at only a limited number of reference laboratories, the usefulness for clinical decision making is limited. ELISA methods are reported to have a lower positive predictive value 8 but are more widely available. Our findings suggest that clinical decision making could be improved by having rapid turnaround times for ELISA results. Reporting the OD value of the HIT ELISA test may contribute to the diagnostic value of the HIT ELISA test. Zwicker et al 4 found that higher ELISA OD measurements correlated significantly with thrombosis and that patients with an OD level of 1.0 or more had a 6-fold increased risk of thrombosis compared with patients who had OD values between 0.4 and In our study of patients for whom there was a strong clinical suspicion of HIT, we found that an OD level of or more had better specificity than the manufacturer s threshold of (85% vs 75%, respectively) but resulted in lower positive and negative predictive values. However, the outcome variable of our study (definite or probable Chong score) was different from the outcome used by Zwicker et al 4 (thrombosis). There are several limitations of our study. The study group is relatively small and is biased because it consists of patients who had a change in anticoagulant treatment based on a strong clinical suspicion of HIT exclusive of consideration of the 4 Ts or Chong clinical scoring systems or HIT ELISA results. This group does not include all patients who underwent evaluation or laboratory testing for HIT. It is, therefore, not possible to fully evaluate how treatment decisions might have been affected by HIT ELISA results, OD thresholds, or 4 Ts scoring or to assess whether a model using ELISA OD levels in the 4 Ts scoring system would improve the negative and positive predictive values of determining an HIT diagnosis. Conclusions The Warkentin 4 Ts pretest probability score correlates well with the Chong posttest classification. The positive predictive value of a high Warkentin 4 Ts score alone was 100%. In these cases, one would continue alternative anticoagulation regardless of the anti heparin-pf4 ELISA results. The results of ELISA alone yielded lower predictive values. The combination of a low 4 Ts score and a negative ELISA result gave the highest negative predictive value (94%). However, if this criterion were used to withhold alternative anticoagulation, 6% of patients would have had a DTI withheld inappropriately (at least by the Chong score). Similar to the use of Wells criteria and the D-dimer assay in the assessment of the risk of deep venous thrombosis and pulmonary embolism, 9 the focus of future clinical research should be to determine whether a combination of clinical probability score and the results of a single laboratory test or combination of tests has sufficient negative predictive value to exclude the diagnosis of HIT and safely withhold alternative anticoagulation therapy. From the Departments of 1 Pathology, 2 Pharmacy, and 3 Surgery and 4 Division of Hematology Oncology, University of California, Davis Medical Center; 5 Division of Biostatistics, Department of Public Health Sciences School of Medicine UC Davis; and 6 VA Northern California Health Care System, Sacramento. Supported by the Department of Pathology and Laboratory Medicine, University of California, Davis Medical Center, Sacramento. Presented in part at the International Society on Thrombosis and Haemostasis Meeting, Sydney, Australia, August Am J Clin Pathol 2007;127: DOI: /RPE753J4PMG9773Q

5 Coagulation and Transfusion Medicine / ORIGINAL ARTICLE Address reprint requests to Dr Janatpour: Dept of Pathology and Laboratory Medicine, UC Davis Medical Center, 4400 V St, Sacramento, CA Acknowledgment: We acknowledge Laurel Beckett, PhD, professor and head, Division of Biostatistics, Department of Public Health Sciences School of Medicine UC Davis, for assistance with the statistical analysis. References 1. Greinacher A, Warkentin TE. Recognition, treatment, and prevention of heparin-induced thrombocytopenia: review and update. Thromb Res. 2006;118: Warkentin TE, Greinacher A. Heparin-induced thrombocytopenia: recognition, treatment, and prevention: the Seventh ACCP Conference on Antithrombotic and Thrombolytic Therapy [published correction appears in Chest. 2005;127:416]. Chest. 2004;126(3 suppl):311s-337s. 3. Davoren A, Aster RH. Heparin-induced thrombocytopenia and thrombosis. Am J Hematol. 2006;81: Zwicker JI, Uhl L, Huang WY, et al. Thrombosis and ELISA optical density values in hospitalized patients with heparininduced thrombocytopenia. J Thromb Haemost. 2004;2: Warkentin TE, Heddle NM. Laboratory diagnosis of immune heparin-induced thrombocytopenia. Curr Hematol Rep. 2003;2: Chong BH, Chong JH. Heparin-induced thrombocytopenia. Expert Rev Cardiovasc Ther. 2004;2: Lo GK, Juhl D, Warkentin TE, et al. Evaluation of pretest clinical score (4 T s) for the diagnosis of heparin-induced thrombocytopenia in two clinical settings. J Thromb Haemost. 2006;4: Pouplard C, Amiral J, Borg JY, et al. Decision analysis for use of platelet aggregation test, carbon 14 serotonin release assay, and heparin platelet factor 4 enzyme-linked immunosorbent assay for diagnosis of heparin-induced thrombocytopenia. Am J Clin Pathol. 1999;111: Wells PS, Owen C, Doucette S, et al. Does this patient have deep vein thrombosis? JAMA. 2006;295: Am J Clin Pathol 2007;127: DOI: /RPE753J4PMG9773Q 433