Heparin-induced thrombocytopenia following hip and knee arthroplasty


Correspondence: Mr Johnathan Craik, Department of Orthopaedics, Epsom General Hospital, Dorking Road, Epsom, Surrey KT18 7EG UK.

E-mail: jcraik82@gmail.com


Until recently, both the British Society for Haematology and American College of Chest Physicians recommended platelet monitoring in all surgical patients receiving prophylactic low molecular weight heparin (LMWH) for the early diagnosis of heparin-induced thrombocytopenia (HIT). These guidelines were reversed in 2012 based upon an analysis considering resource expenditure, assay result timeframes, and complications relating to HIT treatment. However, there are no large studies reviewing lower limb arthroplasty patients on an individual basis to determine the incidence of HIT in this patient group. This study investigated 10 797 patients who underwent primary hip or knee arthroplasty with LMWH prophylaxis over a 5 years period. 32·6% of patients (= 3515) had platelet counts recorded up to 14 d postoperatively with 13 patients (0·37%) developing thrombocytopenia. Platelet counts recovered spontaneously in five patients, and two patients had other identifiable causes. Only one of the remaining six patients developed thrombosis indicating an incidence of HIT-related thrombosis of 0·03%. The potential for identifying HIT with platelet monitoring in patients receiving LMWH prophylaxis is low and therefore routine monitoring for HIT is not justified.

The National Institute for Health and Clinical Excellence (NICE) recommend chemical thromboprophylaxis in patients following hip and knee arthroplasty (NICE, 2010). However low molecular weight heparins (LMWH) carry a low but serious risk of heparin-induced thrombocytopenia (HIT). This paradoxically pro-thrombotic and potentially fatal drug reaction was first recognized in 1958 (Weismann & Tobin, 1958). It is caused by IgG antibody interactions with heparin-platelet factor 4 (PF4) complexes resulting in platelet activation and thrombin release with both arterial and venous thrombotic events (Amiral et al, 1992). HIT is rare in patients under the age of 40 years (Stein et al, 2009), is more common in women (Warkentin et al, 2006) and is associated with a 30-fold increase in risk of thrombosis (Girolami et al, 2003). Its frequency is dependent on both the heparin preparation and clinical setting (Warkentin & Greinacher, 2004), as well as the duration of heparin exposure (Smythe et al, 2007). The risk of developing HIT following exposure to unfractionated heparin (UFH) has been estimated to be 2·6% compared with 0·2% with LMWH (Martel et al, 2005). In addition, the risk in surgical patients is three times that of medical patients (Warkentin et al, 1995, 2000, 2006; Linkins & Warkentin, 2011). The risk of HIT in postoperative orthopaedic patients receiving LMWH has been estimated to be between 0·15% and 0·9% (Warkentin et al, 1995, 2000; Linkins & Warkentin, 2011).

Features of HIT include ulcerating skin lesions at injection sites together with thrombotic complications, such as deep vein thrombosis (DVT), pulmonary embolism (PE), embolic limb ischaemia, cerebrovascular accident (CVA) and myocardial infarction (MI). Three patterns of HIT have been described. Classically, HIT occurs between days 5 and 10 following initial exposure, reflecting the time required to generate an immune-mediated humoral response. Occasionally however, HIT presents within 24 h of heparin exposure in patients who have been sensitized to heparin during the proceeding 100 d. Finally, a rare delayed presentation of HIT can occur after heparin has been stopped. This tends to be a severe variant where antibodies activate platelets in the absence of heparin (Warkentin & Kelton, 2001).

Several clinical scoring systems have been developed to assist in the diagnosis of HIT based on the timing and magnitude of the drop in platelet count, and the presence of clinical signs of thrombosis. One system known as the 4 T's (thrombocytopenia, timing, thrombosis, alternative cause possible) includes either unexplained thrombocytopenia or a fall in platelet count of more than 50% from preexposure or postexposure peak levels between days 5 and 10. Prospective studies have shown that this clinical tool has a high negative predictive value (Lo et al, 2006; Bryant et al, 2008; Bakchoul et al, 2009; Warkentin & Linkins, 2010; Cuker et al, 2012). However some authors have found this test to be unreliable (Berry et al, 2011; Junqueira et al, 2011). More recently, a HIT expert probability (HEP) score was developed to improve diagnostic accuracy further with a reported sensitivity rate of 100% (Cuker et al, 2010). When HIT occurs, a fall in platelet count may precede thomboembolic complications by several days (Warkentin et al, 2009). As a result, although there are reports of HIT in patients with only a modest fall in platelets (Greinacher et al, 2005) or without frank thrombocytopenia (Greinacher et al, 2005; Arepally & Ortel, 2006; Warkentin et al, 2009), the platelet count is currently the only readily available haematological test of value for making an early diagnosis of HIT.

Several laboratory studies are also available to aid in HIT diagnosis. Enzyme-linked immunosorbent assays (ELISAs), which detect anti-PF4/heparin IgG, IgA, and IgM complexes, are most readily available. These assays are highly sensitive. However lack of specificity may lead to the over-diagnosis and over-treatment of patients with suspected HIT (Pauzner et al, 2009; Cuker, 2011). Their value therefore is where a diagnosis of HIT can be excluded with a negative result. Functional assays, such as heparin-induced platelet activation (HIPA) tests or the serotonin release assays (SRA) have higher specificity and are therefore useful adjuncts for diagnosis confirmation (Pouplard et al, 1999). Nevertheless, the interpretation of these assays is not purely dichotomous, but rather depends on the strength of the result together with the clinical picture (Warkentin, 2005). In addition, these tests are often time consuming, and rarely readily available to assist in the clinical decision process. As a result, the diagnosis and treatment of HIT is often initiated on clinical findings alone, which includes stopping all heparin agents and initiating direct thrombin inhibitor therapy.

Until recently, both the British Society for Haematology (BSH) and American College of Chest Physicians (ACCP) recommended platelet count monitoring in all surgical patients receiving LMWH where the risk of HIT was greater than 0·1% (Keeling et al, 2006; Warkentin et al, 2008). Where practical, monitoring was recommended every 24 d until either day 14 of treatment or until heparin was stopped (Warkentin et al, 2008). However with many patients being discharged from hospital on prophylactic LMWH, platelet count monitoring during this critical period poses logistical challenges as well as significant demands on health resource expenditure. In addition, studies estimating the incidence of HIT from which these recommendations were devised have produced inconsistent results and a recent meta-analysis has called for a more accurate determination for the risk of HIT (Martel et al, 2005). Currently there are no large studies reviewing lower limb arthroplasty patients on an individual basis to determine the incidence of HIT in this patient group.

The aim of this study was therefore to estimate the incidence of HIT and determine the value of postoperative platelet count monitoring for patients undergoing hip and knee arthroplasty.

Patients and methods

Data from four separate database systems (theatre manager, pharmacy prescriptions, pathology results and clinical follow-up questionnaire results) were collected and merged to generate a master database from which all subsequent data analysis was performed. All patients undergoing hip (replacement and resurfacing) and knee (total and unicompartmental) arthoplasty were identified from the operative database (Theatre Manager, Bluespier International Ltd, Droitwich, UK) in an elective orthopaedic centre over a 5 years period from 2007 to 2011. Data was collected regarding operative procedure, date of surgery, and postoperative thromboprophylaxis protocol. Information regarding patient discharge thromboprophylaxis prescriptions was collected from the hospital pharmacy database system over the same period. Patients were included in the data analysis if they could be identified from this system as having been prescribed 5000 units of subcutaneous Dalteparin (Fragmin®, Pfizer Inc., New York, NY, USA) once daily for a minimum of 10 d.

All patient hospital numbers and dates of surgery were processed through a custom pathology database search program to return the latest preoperative platelet count and all platelet counts taken between the date of surgery and two months postsurgery. Patients were selected to be at risk of HIT if their platelet counts demonstrated either a drop in platelet count of more than 50% from preoperative baseline values between days 5 and 14 postoperatively, a drop in platelet count of more than 50% from postoperative peak platelet count between days 5 and 14, or any new onset of thrombocytopenia (absolute platelet count less than 150 × 109/ml) detected between postoperative days 5 and 14. The notes of patients identified as high risk were reviewed to determine the likelihood of HIT and a pathology system search performed for any positive HIT assays. Additional data from the pharmacy database was reviewed to identify all prescriptions of direct thrombin inhibitors in order to detect any patients diagnosed with HIT that may or may not have been identified by their platelet counts alone. In addition, the clinical follow-up database (routine 6 week patient follow-up questionnaire) was reviewed for all reported incidents of thrombotic complications (DVT, PE, CVA, MI, death) and the results correlated with all patients included in the study.

Data analysis was performed using Microsoft Office excel 2003 (Windows XP with Visual Basic Macro support) and Microsoft Office excel 2008 (Mac OSx 10.6.8) All statistical analysis was performed using Pearson's chi-squared test. A P-value of less than 0 01 was considered significant.


From January 2007 to December 2011, 13 195 patients received treatment with either primary total hip arthroplasty, hip resurfacing, uni-compartmental knee arthroplasty or primary total knee arthroplasty. 10 816 patients were discharged with prescriptions for 5000 units subcutaneous Dalteparin once daily for at least 10 d. Of these, 10 797 patients were successfully matched with their peri-operative platelet counts and 9140 (84·7%) responded to a 6-week follow-up questionnaire, which identified 212 patients (2·32%) with thromboembolic complications (DVT, PE, CVA, MI or death) within 6 weeks of surgery (Fig 1).

Figure 1.

Percentage of patients developing thromboembolic complications for monitored versus unmonitored patients, and hip versus knee replacement surgery with P-values. DVT, deep vein thrombosis; PE, pulmonary embolism; CVA, cerebrovascular accident; MI, myocardial infarction.

The majority of patients were female (F = 6801, M = 3996) and the mean patient age was 69·7 years. Primary total knee arthroplasty was the most frequent operation (= 5385, 49·9%), followed by primary total hip arthroplasty (= 4622, 42·8%), uni-compartmental knee arthroplasty (= 504, 4·7%) and hip resurfacing (= 286, 2·6%).

3515 patients (32·6%) had platelet counts between postoperative days 5 and 14. The incidence of thromboembolic complications in this group was 4·4% compared with 1·4% in the unmonitored group (< 0·01). Following a change in policy in June 2009, patients were discharged without a specific request for platelet estimation, and the percentage of patients who were assessed fell from 52·5% to 14% (Fig 2).

Figure 2.

Percentage of patients whose platelet count was monitored between postoperative days 5 and 14.

Analysis of the mean change in postoperative platelet counts demonstrated a progressive fall up to postoperative day 3 by 27% of the preoperative level (Fig 3). A recovery of platelet numbers then continued to a maximum at day 13 of 205% of the preoperative level. Mean platelet counts remained elevated for at least 2 months postoperatively.

Figure 3.

Mean postoperative platelet counts for all 10 797 patients following hip or knee arthroplasty. The shaded area indicates the range of ± 2 Standard Deviations (SD). The standard lower threshold for a normal platelet count of 150 × 109/l is also shown.

Of the 3515 patients who had platelet estimations between postoperative days 5 and 14, 13 (0·37%) developed either new onset thrombocytopenia, or more than a 50% drop in baseline or peak postoperative platelet counts, and were considered at risk of HIT. The platelet count spontaneously resolved in five patients. Two other patients had an alternative explanation for a drop in platelets; one patient had known idiopathic thrombocytopenia managed with postoperative prednisolone, and the other had a postoperative massive blood transfusion secondary to wound bleeding from anticoagulation therapy for acute coronary syndrome. All of the remaining six patients completed a 6 week postoperative questionnaire and only 1 reported a thromboembolic complication (0·03%) and was therefore identified as a case of HIT-related thrombosis. This was a female patient who was diagnosed with a DVT 9 d postoperatively followed by PE on day 16. Three of the six patients, including the one confirmed case of HIT-related thrombosis, received surgery between January 2007 and June 2009 (clinical HIT incidence 0·04%). None of the patients at risk of HIT were referred for HIT antibody assays and no patients in the study were prescribed direct thrombin inhibitors.


This is the largest study performed to date estimating the incidence of HIT in postoperative orthopaedic patients on LMWH thromboprophylaxis. The results indicate that HIT-related thrombosis is rare and the potential for platelet count monitoring to identify at-risk patients is low. Guidelines released by the ACCP in 2008 suggested that platelet count monitoring should be performed where the risk of HIT was greater than 0·1% (Warkentin et al, 2008). These guidelines were based on the greater importance of early recognition and treatment of HIT over the costs of platelet monitoring. However several studies have shown that these recommendations were not widely implemented, possibly as a combined result of the burden on limited health resources and lack of awareness (ten Berg et al, 2009; Riggio et al, 2009; Rogers & Cowie, 2010). In addition, our study suggests that even when efforts to monitor patient platelet counts are made within the confines of existing health care services, the potential to identify patients at risk prior to the onset of thomboembolic complications is low. This is particularly pertinent when considering that in an earlier study, two out of five patients diagnosed with HIT-related thrombosis did not demonstrate a significant drop in platelet counts until the time of, or following the development of thromboembolic complications (Warkentin et al, 2003). A revision of both the ACCP and BSH guidelines were published in 2012 recommending that platelet monitoring would only be justified if the risk of HIT was greater than 1% (Linkins et al, 2012; Watson et al, 2012). The rationale for this decision was based upon an analysis considering resource expenditure, assay result timeframes, and complications relating to HIT treatment. In conclusion it was estimated that approximately 1000 patients would need to be monitored to prevent 1 case of HIT-related thrombosis at the cost of one major bleeding event as a consequence of initiated direct thrombin inhibitor therapy. This cost relating to major bleeding events increased where ELISA was the only available HIT assay to institutions. In accordance with these guidelines, this study demonstrates that platelet count monitoring is not justified in postoperative orthopaedic patients on prophylactic LMWH.

This study has several limitations. Firstly, the results may only be relevant for postoperative elective orthopaedic patients discharged with a prescription of 5000 units of Dalteparin daily. Nonetheless, a meta-analysis indicated that the incidence of HIT was more related to duration of exposure rather than dose and therefore these results may be more widely applicable (Stein et al, 2009). In addition, because of the retrospective nature, the quality of the results is dependent on the information collected by the respective database systems. All patients who had at least one platelet count performed between 5 to 14 d postoperatively were included and therefore potentially some patients may have developed HIT following their latest available blood test. Nonetheless, this seems to be unlikely given the clinical consequences and likely investigations entailed if HIT-related thrombosis were to occur. In addition, acknowledgement of this limitation is also a reflection of the practical limitations of platelet count monitoring following hospital discharge for the purpose of HIT identification. A further study limitation relates to the position of the Elective Orthopaedic Centre in relation to the patient population it serves. Some patients' primary care physicians may send their blood samples to other regional laboratories and therefore these results will have been missed from the analysis. This would certainly account for the identification of only 52·2% of patients receiving platelet counts from days 5–14 between January 2007 and June 2009. It is also important to recognize that patients were identified with HIT in this study if they demonstrated thrombocytopenia or a fall in platelet count in keeping with the timing of developing classical HIT followed by thromboembolic complications, without support from concomitant anti-PF4/heparin antibody assays. Nonetheless, these laboratory tests are often not readily available to physicians and therefore decisions regarding treatment must be made on clinical findings and platelet levels alone. In addition, for the purpose of calculating clinically relevant risk, our methodology would serve to over-diagnose rather than underestimate the incidence of the condition. An additional factor potentially contributing to over-estimating the true incidence is the evidence of selection bias whereby our monitored group experienced a significantly greater thromboembolic complication rate compared with the control group. This difference in incidence between the groups is likely to be a consequence of investigations performed as part of thromboembolic disease diagnosis.

Rarer forms of HIT exist that may not have been identified. A rapid onset form of HIT may develop within 4 d of starting heparin, usually as a consequence of heparin exposure within the proceeding 100 d. In addition, delayed onset HIT has been described whereby thromboembolic complications occur after cessation of heparin treatment. However for the purpose of evaluating the appropriateness of current guidelines for platelet count monitoring, which is directed to detect the classical presentation, this study has shown that this form of HIT is extremely rare.

A meta-analysis of studies estimating the incidence of HIT in both medical and surgical patients has shown the risk with UFH to be 2·6% compared with 0·2% for LMWH (Martel et al, 2005). Interestingly however, once established, the severity of complications from HIT as a result of UFH or LMWH is no different (Linkins & Warkentin, 2011). Specifically in postoperative orthopaedic patients on LMWH the risk of HIT is thought to be greater, possibly due to the combination of heparin exposure together with the postoperative pro-inflammatory state (Warkentin et al, 1995, 2000). Several studies have attempted to identify the risk of HIT specifically in orthopaedic patients following lower limb arthroplasty. Warkentin et al (1995) failed to identify any cases of HIT-related thrombosis in 333 patients receiving LMWH following hip arthroplasty compared with nine cases in 332 patients receiving UFH. Their initial definition of HIT was based on a drop in platelet count below 150 × 109/l at or after day 5 postoperatively. However a subsequent anaylsis that used a revised HIT definition of more than a 50% drop in platelet count from postoperative peak values identified one patient receiving LMWH who developed HIT-related thrombosis revealing an incidence of 0·3% (Warkentin et al, 2003). This second analysis was performed to correct for the postoperative thrombocytopenia followed by thombocytosis commonly seen in surgical patients and highlights the importance of identifying trends in postoperative platelet counts rather than relying on absolute values for HIT diagnosis. Our results demonstrate the postoperative thrombocytosis normally seen following hip and knee arthroplasty, supporting the use of this revised HIT diagnostic criteria (Fig 3).

In another study Warkentin et al (2000) investigated the occurrence of HIT in 439 patients receiving LMWH following hip arthroplasty and estimated the incidence to be 0·9%, with 3·2% of patients developing anti–PF4/heparin antibodies. In contrast Greinacher et al (2005) were unable to identify any cases of HIT in a series of 271 patients receiving LMWH following hip and knee arthroplasty. HIT in postoperative elective orthopaedic patients is therefore rare, although these studies have been limited by small sample group sizes. A further study estimated the incidence of HIT to be 0·15% in a combination of trauma and elective orthopaedic patients, based on a sample population of approximately 4000 (Warkentin et al, 2005). Interestingly, the detection of thrombocytopenia together with a positive assay for anti–PF4/heparin antibodies in at risk patients was only able to effectively identify and protect one patient out of six from the development of thromboembolic complications, again demonstrating the limited benefit of platelet count monitoring (Warkentin et al, 2005). The estimate of HIT incidence reported by Warkentin et al (2005) is greater than ours, which may partly reflect the inclusion of trauma patients in their cohort in whom the risk of HIT is known to be greater (Lubenow et al, 2010). However the study reported by Warkentin et al (2005) was also based on an estimated population size and, in contrast to ours, did not evaluate patients on an individual basis.

In conclusion, HIT is a serious adverse reaction to heparin products. Although platelet monitoring has been recommended for surgical patients treated with prophylactic LMWH between days 5 and 14 postoperatively, this study has demonstrated that HIT-related thrombosis following lower limb arthroplasty, as identified through postoperative platelet count monitoring, is extremely rare. Routine platelet monitoring for HIT is not justified in these patients.


Both named authors directly contributed to the study design, data collection, analysis and write up. We would also like to acknowledge contributions made by the following: Richard Field (Consultant Orthopaedic Surgeon, Elective Orthopaedic Centre, Epsom); Alexander Kidd (Consultant Anaesthetist, Epsom General Hospital); Gaye Hadfield (Research Business Manager, Elective Orthopaedic Centre, Epsom); Simon Easton (Pathology IT Manager, Epsom General Hospital); Ann Davies (Chief Pharmacist, Epsom General Hospital); Ian Barnard (Drug Information Pharmacist, Epsom General Hospital); Janet Gearey (Clinical Audit Co-ordinator, Epsom General Hospital); Gary Moore (Consultant Biomedical Scientist, Guys and St Thomas' Hospitals); Jeremy Archer (Haemostasis Laboratory Manager, St Georges Hospital).