Heparin-induced thrombocytopenia (HIT), an immune mediated adverse drug reaction, can occur in up to 5% of patients receiving unfractionated heparin (UFH) or low-molecular weight heparin (LMWH). HIT develops due to the formation of platelet-activating antibodies to heparin and platelet factor 4 (PF4) resulting in platelet aggregation and thrombocytopenia . Although the majority of patients with HIT develop thrombocytopenia, up to half of patients can also develop deep venous thromboembolism (DVT) and pulmonary emboli, which results in longer hospital stays and an increased risk of mortality [1,2]. HIT typically develops in critically ill, surgical, and oncology patients, and diagnosis includes utilization of a validated scoring system and two distinct HIT laboratory assays [1,3]. The 4-T score remains the recommended standard scoring system, and whereas the HIT Expert Probability Score and other scoring systems are available, none have demonstrated superiority over the 4-T score . A low 4-T score has a high negative predictive value (99.8%), and is very useful to exclude the diagnosis of HIT; however, higher scores have a lower positive predictive value (PPV) necessitating further evaluation [4,5]. Of the available laboratory assays, the ELISA test is most commonly performed due to availability and low cost. It detects antibodies to the PF4/heparin complex, and is reported as positive or negative based on the optical density (OD). Although the test is highly sensitive, it is not specific for HIT [6–8]. The serotonin release assay (SRA) is highly sensitive and specific, and is the gold standard for the laboratory diagnosis of HIT. Although a positive ELISA does not confirm the diagnosis of HIT due to its low specificity, higher OD results have been associated with a higher likelihood of a positive SRA result [9,10].
As a result of the low specificity with ELISA, clinicians may be faced with the challenge of using an alternative anticoagulant pending SRA results in patients with a high suspicion for HIT. Accurate and timely diagnosis of suspected HIT plays an important role in guiding treatment to prevent thromboembolism, reduce unnecessary bleeding risks, and prevent patient harm . There is an approximately 5% daily risk of thrombosis if anticoagulation treatment is initially withheld, and a 1% daily risk of bleeding due to anticoagulant treatment in thrombocytopenic patients [1,4,6,12].
Recently, a prospective study of surgical and cardiac ICU patients identified obesity as a potential factor independently associated with the development of HIT . The authors concluded that as obesity has been linked to other immune-mediated diseases such as diabetes, rheumatoid arthritis, systemic lupus erythematosus, and Sjogren syndrome [14,15], it may also be involved in the pathogenesis and immunologic events that cause HIT. They also suggested an alteration to the 4-T score to aid in better identifying patients with HIT, recommending the addition of BMI (or thickness) to calculate a ‘5-T score’. Although obesity may magnify the immune response in HIT, confirmation of this theory requires investigation in a more diverse patient population. The goal of this retrospective, cohort study was to determine the impact of obesity on the development of HIT and to determine if obesity independently predicts the development of HIT.
This was a single-center, retrospective cohort study of patients admitted between 1 January 2011 and 31 July 2016, to any of the four adult hospitals in a large, urban healthcare system. Patients were identified from laboratory testing reports and were included if they were age 18 years or greater, had ELISA and SRA laboratory tests reported within the same hospital admission and documented receipt of heparin-containing products. Patients were excluded for previous diagnosis of HIT, incomplete medical records, an indeterminate or missing SRA result, and pregnancy. The project was approved by the local Institutional Review Board. A power analysis was not performed to determine an adequate sample size due to a lack of data describing the rate of HIT in obese patients.
Patients were categorized as obese based upon a BMI of 30 or greater, as determined by WHO criteria. Obese patients were then compared with nonobese patients (BMI < 30) for the primary outcome of HIT diagnosis confirmed by positive SRA. Secondary outcome measures included the rate of HIT-related DVT, pulmonary emboli, ischemic stroke, myocardial infarction (MI), and in-hospital mortality between groups. ELISA OD values and 4-T scores (calculated by a single investigator for consistency), were also compared between groups. When calculating the 4-T scores, heparin administration to patients within the preceding 3 months was also taken into account as part of the patient's score and risk for HIT.
In addition to the analysis of obese vs. nonobese patients, univariate analysis of SRA positive patients compared with SRA negative patients was performed to identify possible characteristics for inclusion in the multivariable logistic regression analysis. Variables with a P value less than 0.20 were entered into a stepwise multiple logistical regression model. Repeat modeling approaches were generated to include significant variables that were unable to be placed in the original model due to multicollinearity. For example, the model was initially conducted with weight included. We subsequently replaced weight with BMI, then each obesity classification. The same approach was taken for the 4-T score and the development of thromboembolism, as this is included as a component of 4-T score.
Based upon proposed suggestion by Bloom et al., we also generated a 5-T score by including one additional point for those with a BMI of 30 or more to determine if this score predicted a positive SRA result in thrombocytopenic patients with suspected HIT. A second model was generated using the same approach as our initial models with the new 5-T score included instead of the 4-T score. The logistic regression was calibrated based on a nonsignificant Hosmer–Lemeshow goodness-of-fit P value more than 0.05 and model discrimination was assessed based upon the area under the receiver operator characteristic curve. Although an additional point was added for obesity to generate a 5-T score, the 5-T score did not change the usual ranges for low (0–3 points), intermediate (4–5 points), and high (greater than 5 points) suspicion for HIT in this study.
All continuous data were analyzed using the Student t test for parametric data and the Mann–Whitney U test for nonparametric data. Categorical data were analyzed using the Chi-square test or the Fisher's exact test. Parametric data are presented as mean ± SD, nonparametric data as median (25–75% interquartile range), and nominal data as percentages. All tests were two-tailed, and P less than 0.05 was used to represent statistical significance. All analysis, other than propensity scoring, was performed using SPSS, version 23.0 for Windows (SPSS Inc., Chicago, Illinois, USA).
A total of 273 patients were included in the study, with 124 in the obese group and 149 in the nonobese group (Fig. 1). Baseline characteristics were similar between groups, with more female patients in the obese group (Table 1). Approximately half of the population underwent a surgical operation (48.4 vs. 50.3%, P = 0.75), and the majority of patients in the obese group fell into the BMI class I or II categories. In both groups the majority of patients received UFH (71 vs. 64.4%, P = 0.25), whereas a smaller percentage received LMWH (25.8 vs. 24.8%, P = 0.85).
The primary outcome of HIT diagnosis, confirmed by SRA result, was significantly higher in the obese group compared with the nonobese group (22.6 vs. 12.1%, P = 0.02) (Table 2). In regards to secondary endpoints, the ELISA OD value (0.94 vs. 0.78, P = 0.04) and the rate of pulmonary emboli (18 vs. 9%, P = 0.02) were also significantly higher in the obese group compared with the nonobese group. We found no differences in the other secondary outcomes including 4-T score, the rate of DVT, ischemic stroke, MI or in-hospital mortality between groups.
On multivariable analysis, no characteristics related to body habitus (weight, BMI, or obesity classification) were found to independently predict the development of HIT (Table 3). The model did not demonstrate any evidence of a lack of fit based on both a nonsignificant Hosmer–Lemeshow statistic (P = 0.28) and on the ability to differentiate between patients with and without HIT (area under the curve = 0.90). In our initial model, both the ELISA OD [odds ratio (OR) 8.2] and 4-T score (1.7) were found to be independently associated with the development of HIT. Although the 5-T score also was found to independently predict the development of HIT (OR 1.6), there was no increased predictive ability found when compared with the model including the 4-T score.
The 4-T score is a widely accepted and recommended tool for evaluating patients with possible HIT . Despite its widespread utilization, it has a very low PPV which ranges from 22 to 64% based on a low or high risk score for HIT [4,5]. As a result of a low PPV, even high 4-T scores do not always accurately identify HIT and may lead to misdiagnosis or overtreatment. On the contrary, other scoring systems such as the HIT Expert Probability Score and the Lillo-Le Louët model have failed to demonstrate superiority over the 4-T score or improved ability to predict HIT [16–18]. Consequently, novel indicators of HIT are needed to improve the 4-T score. Recently, Bloom et al. evaluated patients known to be at risk for HIT in a surgical and cardiac ICU as part of a retrospective single center study. They found results similar to ours in which obesity was a risk factor for HIT, and proposed that obesity be added to the 4-T score as an additional indicator to aid in HIT recognition and diagnosis . Our study provides further evidence to theirs of the increased risk for HIT in a broader obese population across a 4-hospital system not restricted to the ICU. Several studies have also suggested increasing the OD positive value cutoff to greater than 1 or even 2 to increase its predictive value [2,12]. However, this approach is not feasible in our patient population, because several patients were found to have positive SRA results with corresponding OD values less than one. Therefore, increasing the OD cutoff in our population would have resulted in under detection and under treatment of several SRA positive patients. Despite higher OD values and overall rate of HIT, the 4-T score was not different between our obese and nonobese groups, as seen in prior studies .
The current study has several strengths, including the fact that previously only UFH was reported as a risk factor for HIT in obese patients. Even though this study was not the first to show that LMWH causes HIT, it is the first study to show that LMWH is also associated with HIT in obese patients, despite having an approximately 76% lower relative risk of HIT than UFH [1,19]. In addition, previous authors have hypothesized that obesity could predict the development of HIT if included as a fifth variable to create a ‘5-T score’ . We evaluated the concept of a 5-T score model for the first time. Although we did find that obesity was a risk factor for HIT, and our model of the 5-T scoring system was predictive of HIT, it failed to perform better than the current 4-T score. Consequently, the concept of using obesity as a new variable in the score did not improve our ability to accurately identify HIT. If incorporated with our other results, obese patients should be assessed with the knowledge of their increased risk for HIT, but the 4-T score should still be utilized to evaluate and stratify overall risk. Predicting HIT remains challenging and novel markers of HIT are needed to improve the 4-T score. Although other less validated scoring systems are available, such as the HIT Expert Probability Score, they have failed to show improvement over the 4-T score. It is unknown if including obesity would improve their predictability, but further investigation is warranted.
Based on our study results, the presence of obesity may aid in identification of those most at risk for HIT. Due to the challenge of determining necessary HIT treatment prior to full SRA results, enhanced scoring systems or alternative diagnostic tools are needed to identify patients more accurately and efficiently. At our institution, clinical pharmacists monitor all patients receiving heparin-containing products for the development of HIT, and closely follow platelet trends, 4-T scores, and heparin antibody results. We also have computerized screening tools to identify potential HIT patients and a pharmacy driven protocol for alternative anticoagulation management. Patients are identified as possibly having HIT with the screening tool if they have received heparin-containing products and their platelet count has dropped by 50% or more from baseline. When this occurs, an alert is generated in the computer for the pharmacist to investigate the possibility of the patient having HIT. The investigation involves calculation of a 4-T score, having a discussion with the physician regarding the suspicion for HIT, and if necessary ordering laboratory tests and alternative anticoagulation. When suspicion for HIT is significant, the ELISA test is ordered first, and if positive then the SRA confirmatory test is ordered. Despite these advancements and a comprehensive computerized HIT protocol, we must continue to rely on clinical acumen and experience, individualized patient assessment, and the current 4-T score to identify patients with HIT.
The current article is an original work not previously published in any substantial part. It is not under consideration of publication elsewhere. This article has been read and approved for submission by all qualified authors. All authors also declare no source of funding.
Ethics committee approval: All authors confirm that this research was conducted in accordance with the Declaration of Helsinki and under the terms of all relevant local legislation.
Conflicts of interest
All authors declare they have no actual or potential conflict of interest capable of influencing judgment on the part of any author.
1. Linkins LA, Dans AL, Moores LK, Bona R, Davidson BL, Schulman S, et al. American College of Chest Physicians Treatment and prevention of heparin
: antithrombotic therapy and prevention of thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines. Chest
2. Altuntas F, Matevosyan K, Burner J, Shen YM, Sarode R. Higher optical density of an antigen assay predicts thrombosis in patients with heparin
. Eur J Haematol
3. Pierce W, Mazur J, Greenburg C, Mueller J, Foster J, Lazarchick J. Evaluation of heparin
(HIT) laboratory testing and the 4Ts scoring system in the intensive care unit. Ann Clin Lab Sci
4. Cuker A, Gimotty PA, Crowther MA, Warkentin TE. Predictive value of the 4Ts scoring system for heparin
: a systematic review and meta-analysis. Blood
5. Lo GK, Juhl D, Warkentin TE, Sigouin CS, Eichler P, Greinacher A. Evaluation of pretest clinical score (4 T's) for the diagnosis of heparin
in two clinical settings. J Thromb Haemost
6. Smythe MA, Koerber JM, Forsyth LL, Tuchscherer RM. Implications of heparin
overdiagnosis. Ann Pharmacother
7. Kelton JG, Arnold DM, Bates SM. Nonheparin anticoagulants for heparin
. N Engl J Med
8. Floresca D, Dupree L, Basile S, Tan P. Evaluation of appropriate serologic testing for suspected heparin
. Am J Health Syst Pharm
9. Warkentin TE, Sheppard JI, Moore JC, Sigouin CS, Kelton JG. Quantitative interpretation of optical density measurements using PF4-dependent enzyme-immunoassays. J Thromb Haemost
10. Chan CM, Woods CJ, Warkentin TE, Sheppard JA, Shorr AF. The role for optical density in heparin
: a cohort study. Chest
11. Marler J, Unzaga J, Stelts S, Oliphant CS. Consequences of treating false positive heparin
. J Thromb Thrombolysis
12. Harada M, Hoang D, Zaw A, Murry J, Volod O, Sun BJ, et al. Overtreatment of heparin
in the surgical ICU. Crit Care Med
13. Bloom M, Zaw A, Hoang D, Mason R, Alban R, Chung R, et al. Body mass index strongly impacts the diagnosis and incidence of heparin
in the surgical intensive care unit. J Trauma Acute Care Surg
14. Gremese E, Tolusso B, Gigante MR, Ferraccioli G. Obesity
as a risk and severity factor in rheumatic diseases (autoimmune chronic inflammatory diseases). Front Immunol
15. Janssen I, Katzmarzyk PT, Ross R. Body mass index, waist circumference, and health risk: evidence in support of current National Institutes of Health guidelines. Arch Intern Med
16. Joseph L, Gomes MP, Al Solaiman F, St John J, Ozaki A, Raju M. External validation of the HIT Expert Probability (HEP) Score. Thromb Haemost
17. Cuker A, Arepally G, Crowther MA, Rice L, Datko F, Hook K. The HIT Expert Probability (HEP) Score: a novel pretest probability model for heparin
based on broad expert opinion. J Thromb Haemost
18. Uaprasert N, Chanswangphuwana C, Akkawat B, Rojnuckarin P. Comparison of diagnostic performance of the heparin
expert probability and the 4Ts score in screening for heparin
. Blood Coagul Fibrinolysis
19. Junqueira DR, Perini E, Penholati RR, Carvalho MG. Unfractionated heparin
versus low molecular weight heparin
for avoiding heparin
in postoperative patients. Cochrane Database Syst Rev
Keywords:Copyright © 2018 YEAR Wolters Kluwer Health, Inc. All rights reserved.
heparin; obesity; platelets; thrombocytopenia; thrombus