Improving the specificity of the PF4 ELISA in diagnosing heparin-induced thrombocytopenia

Abstract

Heparin induced thrombocytopenia (HIT) is a serious complication of heparin therapy. The PF4 ELISA is a serologic assay that provides laboratory support for the clinical diagnosis of HIT, but it is often positive in patients who do not have the syndrome. We examined whether the specificity of the PF4 ELISA can be improved by 1) taking antibody potency into consideration, 2) by measuring only IgG antibodies, and 3) by utilizing a high concentration heparin inhibition step. We reviewed clinical information on 116 patients whose samples were referred for HIT antibody testing and assigned each a clinical score related to the likelihood of the patient having HIT. The scores were then correlated with serologic findings. Patients with strongly positive PF4ELISA results (OD ≥ 1.0) using both versions of the assay (IgG/A/M and IgG only) had clinical scores and SRA activity that were significantly higher than those having reactive or negative results. When the IgG-only PF4 ELISA was used, only the strongly positive result group had significantly higher clinical scores and SRA release, and fewer samples were classified as weakly positive or reactive, suggesting that detection of IgG only in the PF4 ELISA improves the assay’s specificity. The heparin inhibition step identified “reactive” samples that were associated with clinical scores and SRA release indistinguishable from the “negative” result groups, confirming that this step further improves specificity of the test. This study supports utilizing these 3 modifications of the PF4 ELISA to improve specificity in supporting the clinical diagnosis of HIT.

Introduction

Heparin-induced thrombocytopenia (HIT) is a serious immune-mediated complication of heparin therapy affecting 1–3% of patients receiving the drug for at least 7 days [1–3]. Patients with HIT typically experience a fall in the platelet count at least 50% below baseline after 4 or more days of treatment. Thrombocytopenia itself is rarely severe enough to cause bleeding; however, up to 75% of cases are complicated by venous or arterial thromboembolic events [4,5]. Antibodies directed against complexes of heparin and platelet factor 4 (PF4) cause both thrombocytopenia by activating platelets via their FcγRIIa receptors and the generation of procoagulant activity [6–10]. Other mechanisms may also be operative [11].

HIT is a clinicopathological syndrome [5] requiring both compatible clinical findings and supportive laboratory testing for diagnosis. Two types of serologic tests are available to detect heparin-dependent, platelet-reactive antibodies: those that detect antibody binding to complexes of PF4 bound to heparin or a heparin-like molecule (so-called antigen tests) and those that detect platelet activation in the presence of patient antibody and heparin (activation tests). The PF4 ELISA test, an example of the former type of assay, is available in kit form and can be performed in hospital laboratories in as few as 4 hr from acquisition of specimen. The assay is highly sensitive and, as a result, detects both weak antibodies that may be clinically insignificant and strong ones that predispose to thrombosis [12,13]. The serotonin release assay (SRA), a type of platelet activation assay, is considered the “gold standard” for confirming the presence of clinically important HIT antibodies, that is, those that are capable of triggering thrombocytopenia and/or thrombosis. The chief limitations of SRA are its complexity and the requirement for a radioactive reagent, factors that together limit the availability of the assay to only a few reference laboratories in North America. As a consequence, several days are often required for SRA results to become available to the treating physician. Accordingly, physicians caring for patients with HIT must often depend on the more available PF4 ELISA test to make treatment decisions on a timely basis.

A diagnosis of HIT calls for discontinuation of heparin, institution of alternative anticoagulation (often at great cost), prolongation of the hospital stay, and, often, continued outpatient anticoagulation to address the known, persistent risk of thrombosis. Accordingly, it is important that the diagnosis be made with the greatest possible accuracy and, once made, that those patients who have the greatest risk of thrombosis be identified to the extent possible. Attempts to increase the accuracy of HIT diagnosis include using more uniform clinical criteria for identifying patients who are truly at risk and modifying the performance and interpretation of the more readily available PF4 ELISA. Application of a clinical criteria algorithm such as the 4T score has been recommended to establish the pretest clinical probability of true HIT [12,14,15]. At the laboratory level, it has been recommended that the strength or optical density (OD) value of a positive PF4 ELISA test result (a close approximation to antibody titer) be used to assess the likelihood that a patient with HIT will experience thrombosis [16,17] and that a version of the assay be used that only detects IgG antibodies, as this is the dominant, perhaps only immunoglobulin class capable of causing “true” HIT [18,19]. A third suggested enhancement is the demonstration that reactions obtained in the PF4 ELISA test are inhibited by a high concentration of heparin to confirm the specificity of positive test results [20,21].

In the context of a large referral laboratory, we sought to evaluate the extent to which these enhancements to the PF4 ELISA test improve the accuracy of HIT diagnosis by obtaining detailed clinical information on patients referred for HIT antibody testing and examining correlations between laboratory test results and clinical outcomes.

Materials and Methods

Sample/patient selection

The study was approved by the institutional review board of Blood Center of Wisconsin. For the purpose of our study, patient samples received between September 12, 2005 and January 19, 2009 were identified for further testing and follow-up. Eighty-four samples were referred to the laboratory for PF4 ELISA testing with or without additional SRA testing (Group 1). An additional 38 samples (Group 2) were referred for SRA testing only, and PF4 ELISA testing was performed by the laboratory for the study, but not reported on these samples. As a major aim of our study was to understand the relationship between different types of test results in the PF4 ELISA (i.e., negative, weakly positive, strongly positive, and reactive; definitions follow) and the probability that patients had true HIT based on clinical findings, laboratory staff were instructed to identify sequential samples of each result type until a predesignated number of samples were selected. Samples were selected without knowledge of the clinical status of the patients; however, it was assumed that all were from patients suspected of having HIT based either on clinical finings alone or preliminary laboratory testing performed locally.

Referring physicians were contacted by phone by a study investigator and asked to provide responses to a brief questionnaire (Fig. 1). Two investigators (J.M. and R.A.) then scored each patient’s survey information and calculated a 4T score based on a previously published algorithm for evaluating the clinical pretest probability of a patient having true HIT [12]. Scores that were within one point of each other were averaged for a consensus score. Scores that differed by two or more points were reviewed by the investigators in conference and were assigned a score agreed on by consensus. For analysis, clinical scores were grouped as follows: 4T score 0–3: HIT unlikely; 4–5: HIT possible; 6–8: HIT likely [12].

Figure 1.

Referring physician questionnaire from which the clinical scores (4T) were developed.

As the clinical information was collected retrospectively, the contacted physicians had access to the results of the tests they had initially ordered (PF4 ELISA, SRA, or both). Similarly, the investigators collecting the information were not blinded to the test results; however, every attempt was made to administer the survey uniformly without reference to the test results.

Test methods

PF4 ELISA: Assay description

The PF4 ELISA was based on a previously described method that detects antibodies of three immunoglobulin classes (IgG/A/M) against PF4/polyvinyl sulfonate complexes [22]. A modification of the assay substituted a human IgG-specific reagent to detect IgG antibodies only. For both the IgG/A/M and the IgG-only PF4 ELISA, a confirmatory high-concentration heparin step was performed concurrently, in which 100 U/mL (high dose) of unfractionated porcine heparin (UFH) was added to the patient’s sample. Serum samples were diluted to 1:50 in phosphate-buffered saline (pH 7.4) prior to being tested.

PF4 ELISA results were assigned to one of the three groups: OD < 0.4: negative; OD = 0.4–0.99: weakly positive; OD ≥ 1.0: strongly positive. Inhibition of reactivity by at least 49% when a high concentration of UFH (100 U/ml) was added was required for samples to be considered weakly or strongly positive. Samples producing OD ≥ 0.4 that failed to inhibit by at least 49% with high-dose heparin were designated as reactive and comprised a fourth group.

SRA: Assay description

The SRA was based on a previously described method [23]. SRA results were designated as follows: release with low-dose heparin < 20%: negative; release with low-dose heparin ≥20–74%, inhibiting to <20% with high-dose heparin: positive; release with ≥75%, inhibiting to <20% with high-dose heparin: strongly positive; and release with ≥20%, without inhibition to <20% with high-dose heparin: indeterminate.

Statistical analysis

Comparisons of demographic and clinical presentation characteristics of Groups 1 and 2 and the clinical scores and reactivity in the SRA of the different PF4 ELISA result groups (i.e., negative, weakly positive, strongly positive, and reactive) were done using the χ² test with Fisher’s exact test, when values in cells of the contingency tables were <5. Mean clinical scores and SRA reactivity for the different test result groups in the PF4 ELISAs were compared using the unpaired t-test. The fraction of samples from each PF4 ELISA result group associated with either ≥75% release in the SRA or a clinical score ≥6 were compared using the χ² test. P values of ≤0.05 were considered significant.

Results

Characteristics of the patient population

Characteristics of Groups 1 and 2 patients are summarized in Table I. The groups were similar in age, sex, procedures undergone, and type of heparin exposure. The only significant differences were found in the fraction of patients undergoing cardiac procedures (more common in Group 2) and in those receiving low-molecular-weight heparin as the only heparin exposure (more common in Group 1). As the two sample groups were from patients with similar (albeit not identical) clinical findings and demographics, they were combined for further analysis of associations with clinical scores and SRA results.

TABLE I.

Characteristics of Group 1 and 2 Patients

	Group 1 (N = 78)		Group 2 (N = 38)		χ2 test (p value)
Median age	63		62
Age range	11–94		13–86
Female sex	36	46.2%	18	47.4%	NS
Surgery/proceduresa
Cardiac	17	21.8%	17	43.6%	0.01
Vascular	5	6.4%	2	5.1%	NS
Other surgery	10	12.8%	4	10.3%	NS
Dialysis	11	14.1%	5	12.8%	NS
Type of heparin					NS
Unfractionated	60	76.9%	32	82.0%	NS
Low-molecular-weight heparin	14	17.9%	1	2.6%	0.02
Both	2	2.6%	4	10.3%	NS
None recorded	2	2.6%	1	2.6%	NS
Indication for heparinb					NS
Cardiac surgery or procedure	17	21.8%	20	51.3%	0.001
Vascular surgery	5	6.4%	2	5.1%	NS
Dialysis	11	14.1%	5	12.8%	NS
Other procedure	3	3.8%	2	5.1%	NS
Deep vein thrombosis prophylaxis	34	43.6%	12	30.8%	NS
Line flushes	2	2.6%	1	2.6%	NS
Therapeutic for deep vein thrombosis or other thrombosis	14	17.9%	6	15.4%	NS
Unknown	1	1.3%	1	2.6%	NS

^aSome patients underwent more than one procedure.

^bSome patients had more than one indication for heparin.

Of the 122 samples selected, responses to questionnaires were successfully obtained for 116 samples (95%). Of these, 39 were negative in the IgG/A/M PF4 ELISA, 26 were weakly positive, 29 were strongly positive, and 22 were reactive. The remaining six samples were excluded from analysis.

The initial 4T scores calculated by investigators were identical or within one point for 109 of 116 or 94% of the patients for whom clinical data were available. Of the seven patients for whom scores differed by more than one point, all but three differed by only two points.

IgG/A/M PF4 ELISA results versus clinical score and SRA test results

As shown in Fig. 2a, mean 4T scores were significantly higher for patients testing weakly or strongly positive in the IgG/A/M PF4 ELISA (5.0 ± 1.9 and 5.4 ± 2.1, respectively) than in patients testing negative or reactive (3.2 ± 2.0 and 2.8 ± 2.0, respectively). Sixteen of 32 samples (50%) from patients with the strongest 4T scores (6–8, HIT very likely) had IgG/A/M PF4 results that were classified as strongly positive, whereas 10 were weakly positive, and six were negative or reactive (Fig. 2b).

Figure 2.

(a) 4T clinical scores for each IgG/A/M PF4 ELISA result group are shown. Mean clinical scores for negative (3.2 ± 2.0) and reactive (2.8 ± 2.0) groups were not significantly different from one another (unpaired t-test). Mean scores for the weakly positive (5.0 ± 1.9) and strongly positive (5.4 ± 2.1) groups were similar and both were significantly greater (P < 0.00001) than those of the negative and reactive groups. Shaded areas represent ± 1 standard deviation. (b) Total numbers of samples in each IgG/A/M PF4 ELISA result group (solid columns); those from patients with clinical scores (4T) of six or greater (gray speckled columns) are shown. The fraction of samples from patients having clinical scores ≥6 in the weakly and strongly positive PF4 ELISA result groups was similar, and both positive groups had significantly more high scores than the negative and reactive groups (P < 0.002, χ² test).

SRA results were compared for each IgG/A/M PF4 ELISA result group. Mean SRA release was significantly greater in the strongly positive group (65.2% ± 39.3%) when compared with all other categories (negative: 15.7% ± 26.9%; reactive: 13.1% ± 23.8%; and weakly positive: 22.7% ± 31.2%), and there were no significant differences among the latter three (Fig. 3a). Sixteen of 26 samples (62%) associated with the greatest SRA release (≥75%) had IgG/A/M PF4 ELISA results that were classified as strongly positive, whereas the remaining 10 were classified as negative (4), weakly positive (5), or reactive (1) (Fig. 3b). When SRA indeterminate results (those with SRA release >20% not inhibiting to <20% with high-dose heparin) were excluded, 16 of 24 (67%) of the strongest SRA reactions fell into the strongly positive PF4 ELISA group.

Figure 3.

(a) SRA percent release for each IgG/A/M PF4 ELISA result group is shown. Mean SRA results for negative (15.7% ± 26.9%), reactive (13.1% ± 23.8%), and weakly positive (22.7% ± 31.2%) groups were not significantly different from one another (unpaired t-test). Mean SRA results for the strongly positive group (65.2% ± 39.3%) were significantly greater (P < 0.00001) than any of the other groups. Triangles represent indeterminate results in the SRA (release did not inhibit with high-dose heparin). Shaded areas represent ±1 standard deviation. (b) Total samples in each IgG/A/M PF4 ELISA result group (solid columns); those from patients with SRA release ≥ 75% (gray speckled columns) are shown. The fraction of samples from patients with SRA release ≥ 75% was significantly greater in the strongly positive group (P < 0.00001, χ² test) versus all of the other groups that did not differ from one another.

IgG-only PF4 ELISA results versus clinical score and SRA test results

Each of the analyses was repeated using results of the IgG-only PF4 ELISA (Figs. 4 and 5). Mean 4T scores were significantly higher only in patients testing strongly positive (6.0 ± 1.7) versus mean scores for all other groups (negative: 3.2 ± 2.0; weakly positive: 3.3 ± 1.5; and reactive: 2.9 ± 2.4), which were not significantly different from one another. Twenty-three of 32 samples (72%) from patients with the highest 4T scores (6–8, HIT very likely) had IgG-only PF4 ELISA results that were classified as strongly positive, whereas the remaining nine were classified as negative (7), weakly positive (1), and reactive (1) (Fig. 4b).

Figure 4.

(a) 4T clinical scores for each IgG-only PF4 ELISA result group are shown. Mean clinical scores for negative (3.2 ± 2.0), reactive (2.9 ± 2.4), and weakly positive (3.3 ± 1.5) groups were not significantly different from one another (unpaired t-test). Mean scores for the strongly positive group (6.0 ± 1.7) were significantly greater (P < 0.00001) than those of all other groups, including those of the weakly positive. Shaded areas represent ±1 standard deviation. (b) Total samples in each IgG-only PF4 ELISA result group (solid columns); those from patients with clinical scores (4T) of six or greater (gray speckled columns) are shown. The fraction of samples from patients with clinical scores ≥ 6 was significantly greater in the strongly positive PF4 ELISA result group when compared with all other groups (P < 0.00001, χ² test), and there were no significant differences among the remaining groups (negative, reactive, and weakly positive).

Figure 5.

(a) SRA percent release for samples in each IgG-only PF4 ELISA result group is shown. Mean SRA results for negative (12.7% ± 23.8%), reactive (8.8% ± 18.3%), and weakly positive (10.2% ± 12.8%) groups were not significantly different from one another (unpaired t-test). SRA results for the strongly positive group (67.6% ± 35.3%) were significantly greater (P < 0.00001) than any of the other groups. Triangles represent indeterminate results in the SRA (release did not inhibit with high-dose heparin). Shaded areas represent ± 1 standard deviation. (b) Total samples in each IgG-only PF4 ELISA result group (solid columns); those with release ≥ 75% in the SRA (gray speckled columns). The fraction of samples yielding SRA > 75% was significantly greater in the strongly positive PF4 ELISA result group when compared with all other groups (P < 0.00001, χ² test), and there were no significant differences among the remaining groups (negative, reactive, and weakly positive).

Again, the only group that had significantly higher release in the SRA (68.4% ± 35.4%) was the group strongly positive in the IgG-only PF4 ELISA (Fig. 5a). All other groups (negative: 12.7% ± 23.8%; weakly positive: 10.2% ± 12.8%; and reactive: 8.8% ± 18.3%) were indistinguishable with low mean SRA activity. Twenty-two of 26 samples (87%) that produced the greatest SRA release (≥75%) were classified as strongly positive, and the remaining four were classified as negative, with two of these actually having an indeterminate SRA result (Fig. 5a,b). When SRA indeterminate results were excluded, 22 of 24 samples with the greatest SRA release (92%) were classified as strongly positive in the IgG-only PF4 ELISA.

As other groups [17] have suggested that the association between the PF4 ELISA result and the clinical HIT may be further strengthened by identifying results of OD ≥ 2.0 versus those between 1.0 and 2.0, we analyzed our data accordingly. For the IgG-only PF4 ELISA, 20 of the 38 strongly positive samples had OD ≥ 2.0. The mean clinical scores and SRA results were compared for these two subsets. Although there was marginally greater SRA % release for samples with OD ≥ 2.0 (77.8% vs. 57%; P = 0.04), there was no significant difference in mean clinical scores or in the fraction of samples associated with the highest clinical scores (i.e., 6–8) or highest SRA % release (≥75%). For the IgG/A/M PF4 ELISA, there were too few samples with OD ≥ 2.0 (2 of 29) among the strongly positive group to analyze.

The high-concentration heparin inhibition step

Twenty-two samples were classified as reactive (failed to inhibit by 49% from the original ELISA OD value) using the IgG/A/M PF4 ELISA. The mean of the associated clinical scores was not significantly different from that of the negative group’s scores (Fig. 2a). Similarly, the mean SRA release of the IgG/A/M reactive samples did not differ from that of the negative group’s results (Fig. 3a). When the IgG-only PF4 ELISA was used (Figs. 4a and 5a), the mean scores and SRA results of the reactive group were indistinguishable from those of the negative group. Although eight (36%) of the IgG/A/M PF4 ELISA reactive samples produced initial OD > 1.0, a separate analysis of the clinical scores and SRA results from this subset (data not shown) revealed no significant differences from the reactive group as a whole or from the negative group.

As the comparison of the mean clinical scores and SRA release of the result groups may have obscured individual reactive samples associated with high clinical scores and/or SRA release, we evaluated each reactive sample in both assays. In the IgG/A/M PF4 ELISA, there was only one sample classified as reactive that was associated with a clinical score greater than 5 (this score was 8) and a high SRA result (95%; Figs. 2a and 3a). When this sample was tested in the IgG-only assay, it was classified as strongly positive (OD > 2.0). In the IgG-only PF4 ELISA, there were no samples classified as reactive that had SRA release > 10% [excluding one indeterminate SRA (Fig. 5a) and only one had a high clinical score (7.5; Fig. 4a)]. This one had an SRA of only 2%. The GAM PF4 ELISA classified this last as weakly positive with an OD of 0.7.

Discussion

The PF4 ELISA constitutes a major advance in providing readily available laboratory support for the diagnosis of HIT. However, a major limitation of this sensitive assay is the risk of overdiagnosing the syndrome, particularly in patients who do not have clinical findings compatible with HIT. The American College of Chest Physicians recommends limiting the use of serologic HIT assays to only those patients who demonstrate clinical findings that suggest HIT. They propose using serologic tests only for patients who are receiving heparin currently or who have been exposed within the previous 2 weeks, and who experience a decrease in the platelet count by ≥ 50% and/or a thrombotic event 4–15 days following initiation of heparin therapy [5]. They further advice against routine HIT antibody testing of patients who are receiving or who will receive heparin who do not have clinical evidence of HIT. Adherence to this guideline will improve the specificity of the serologic result, as it assures that only patients with a high pretest probability of HIT are tested. Conversely, applying the PF4 ELISAs to patients who do not have clinical findings supportive of HIT will identify positive results in patients who are unlikely to have true HIT [13]. It is known, for instance, that up to 50–60% of patients who have been exposed to UFH in the setting of open-heart surgery will develop detectable PF4 heparin antibodies, without developing clinical HIT [24,25].

Careful assessment of the clinical findings in patients suspected of having HIT is necessary but insufficient to establish the diagnosis, as there are frequently other causes of thrombocytopenia and/or thrombosis that may be present in these patients. It is vital to provide laboratory support for the clinical diagnosis, and therefore optimizing the specificity of the sensitive PF4 ELISA tests remains a worthwhile goal.

The findings of our study are consistent with previous reports showing that a strongly positive PF4 ELISA test result (OD ≥ 1.0) correlates well with a positive SRA result and with the likelihood of having HIT as assessed by the 4T score method [16,26,27]. Our analysis of the subset of IgG-only PF4 ELISA results with OD > 2.0 failed to support the value of further stratifying the strongly positive results. However, this analysis could have been impacted by the limited number of samples with OD > 1.0 (38 for the IgG-only PF4 ELISA) in our study. Additionally, as we used our own “home brew” PF4 ELISAs, it cannot be assumed that the performance characteristics of our test are identical to those of assays used by other groups who have found value in using a different cutoff to designate “strongly positive” reactions in the PF4 ELISA.

The association between a strongly positive ELISA result and high clinical scores as well as greater platelet activation is improved further when the IgG-only PF4 ELISA is used, confirming earlier reports of the enhanced specificity of the IgG-only assays [18,27]. An additional benefit of the IgG-only PF4 ELISA versus the IgG/A/M assay is the reduction in the number of samples in either the weakly positive or the reactive result groups. In this study, the samples classified as reactive decreased by 59% (from 22 to 9) and as weakly positive by 58% (26 to 11), whereas negative samples increased by 46% (from 39 to 57). These changes occurred without any loss in the ability of the IgG-only PF4 ELISA to classify the most significant platelet-activating antibodies as strongly positive (Figs. 4b and 5b).

The association between strongly positive IgG-only PF4 ELISA results and high 4T clinical scores and SRA release was highly significant; however, there were nine samples that were categorized as either negative (7), weakly positive (1), or reactive (1) in this test that were linked to high clinical scores (Fig. 4b). This finding might relate to the procedure used for collection of clinical information, which was a retrospective telephone interview of ordering physicians and not a direct review of the clinical records by the investigators, the latter not being possible due to the diversity of physicians and institutions involved. This may have led to some of the clinical information not being as complete or accurate as possible.

Alternatively, there may be non-IgG antibodies capable of causing the clinical findings in HIT that would not be detected in an IgG-only PF4 ELISA assay. The relevance of IgA and IgM heparin-dependent platelet antibodies in clinical HIT is controversial [19,27–29]. The fact that one of the nine nonstrongly positive IgG-only samples associated with high clinical scores was strongly positive in the IgG/A/M PF4 ELISA suggests that at least some non-IgG heparin-dependent antibodies may be associated with clinical disease. The importance of non-IgG antibodies that react with PF4–heparin complexes in clinical HIT remains uncertain and deserves further study.

There was also a highly significant correlation between the strongly positive IgG-only PF4 ELISA results and high SRA release. The latter detects HIT antibodies that are capable of activating platelets that may lead to increased thrombin generation in vitro and risk of thrombotic complications seen in HIT [30]. Excluding the indeterminate SRA results, that is, those that failed to inhibit with high-dose heparin, all but two of the samples generating the highest SRA release were classified as strongly positive in the IgG-only PF4 ELISA (Fig. 5b). These two samples were also negative in the IgG/A/M ELISA, suggesting that nonheparin-related platelet-activating antibodies may have been detected. The fact that one of these was associated with a low clinical score (3) suggests that this may be an example of a falsely positive SRA result.

The findings of our study confirm previous reports supporting the value of reaction strength (OD) and an IgG-only test to improve the correlation between PF4 ELISA test results and a clinical picture indicative of HIT. They also confirm the value of the high-concentration heparin inhibition step in enhancing the specificity of this assay. Two recent reports are in conflict about the value of the high-dose heparin confirmatory step in diagnosing HIT [21,25]. Selleng et al. [25], in a follow-up analysis of a cohort of 591 patients undergoing cardiopulmonary bypass surgery who were tested for HIT antibodies using an in-house IgG PF4 ELISA and a platelet activation assay designated heparin-induced platelet activation assay (HIPA), found that a high-dose heparin confirmatory step was positive in only 49.3% of HIPA positive sera and that only two of the three patients in this cohort who probably had true HIT had antibodies that were inhibited by high-dose heparin in the PF4 ELISA. In contrast, Whitlatch et al. retrospectively studied 212 patients in their large tertiary medical center who were found to be positive in a commercially available PF4 ELISA (GTI Diagnostics). Eighty-two percent of these positive samples were inhibited in the high-dose heparin confirmatory step, and they found that inhibition was a significant independent predictor of clinical HIT, along with the strength of the initial low-dose heparin reaction as measured by the OD [21].

Several differences in these two studies may have contributed to their opposite conclusions. The first study was done on a cohort of patients with a low pretest probability of having clinical HIT (all cardiovascular surgery patients), whereas in the second study, HIT antibody tests had been requested by clinicians who most likely had a clinical suspicion that their patients actually had HIT. In addition, the assays used differed: in-house IgG PF4 ELISA versus a commercially available PF4 ELISA. Finally, the criteria used to make the diagnosis of HIT differed: the former study requiring both an antigen and an activation assay to be positive as well as thrombosis or thrombocytopenia in the appropriate time frame versus clinical criteria alone in the other study.

The results of our study more closely match those of Whitlatch et al. We were dealing with a group of patients whose pretest probability of having HIT was high, we used PF4 ELISAs that closely resemble the commercially available kit, and we applied standard, nonserologic criteria for assigning patients a likelihood of having clinical HIT. On an ongoing basis, our laboratory finds between 200 and 300 samples per year positive in the IgG PF4 ELISA and ~ 20% of these do not confirm in the high-dose heparin confirmation step, very close to the 18% of samples in the Whitlatch study that did not confirm. In our experience, HIT antibodies that do not “confirm” with high-dose heparin are usually (although not always) weakly positive and may be insignificant clinically. In contrast, samples from patients with a clinical picture suggestive of HIT that yield strong positive results almost invariably inhibit with excess heparin. We agree with Whitlach et al. that the high-dose heparin inhibition step is a valuable adjunct to the PF4 ELISA and improves specificity of the assay.

As a result of the findings made in this study, we now use the IgG-only version of the test as our primary assay, and we continue to perform concurrently the high-concentration heparin inhibitions step on all samples. We have modified our interpretations to emphasize the value of a strongly positive result (i.e., OD ≥ 1.0). However, we have not changed the cutoff designating a positive result (OD ≥ 0.4). In patients who have weakly positive IgG-only PF4 ELISA results, we recommend testing in the SRA, particularly if there are clinical findings suggestive of HIT. In those patients with strongly positive ELISA results and moderate to high clinical suspicion of HIT, we consider that the PF4 ELISA result alone should be sufficient to confirm the diagnosis.