Accuracy of Diagnosing Heparin-Induced Thrombocytopenia

Key Points Question Is the current diagnostic practice for suspected heparin-induced thrombocytopenia (HIT) accurate? Findings In this diagnostic study of 1318 patients suspected of having HIT, the 4Ts score produced 9.0% false negatives; chemiluminescent immunoassay produced 4.5%; and the recommended algorithm (4Ts score followed by chemiluminescent immunoassay) produced 13.5%. These same tests produced 49.0%, 6.0%, and 4.1% false positives, respectively. Meaning In this study, a substantial number of patients with suspected HIT were misclassified, which could lead to delayed diagnosis or overtreatment.


Introduction
The diagnostic utility of recommended diagnostic tests in daily practice often does not match published study results, and little is known about tests used to diagnose heparin-induced thrombocytopenia (HIT).2][3][4] Immunogenic complexes are generated from IgG antibodies targeting platelet factor 4 (PF4)/heparin-complexes. [5][6][7] In turn, these complexes activate platelets through FcγRIIA receptors and cause platelet aggregation. 8As phenotype, a triad of signs and symptoms are common: (1) thrombocytopenia, (2) severe thrombosis, and (3) a typical timing of platelet count reduction following heparin administration. 8,9HIT is uncommon and occurs in diverse settings, making it unfamiliar to the clinician involved. 10It is, however, imperative to know when to suspect HIT and how to handle it correctly to prevent severe thromboembolic complications. 8e workup of suspected HIT is challenging because the available diagnostic tools are associated with major drawbacks.The 4Ts score is a clinical scoring system aiming to determine the clinical (pretest) probability of HIT (Table 1). 11,12Despite a meta-analysis 13 stating that the 4Ts score has a high negative predictive value, the sensitivity of the score in clinical practice has been challenged by some studies. 10,14Furthermore, the positive predictive value was low, suggesting HIT in a large proportion of patients.Several enzyme-linked immunosorbent assays and rapid immunoassays have been developed to better identify and quantify PF4/heparin antibodies. 15In general, most assays have a high sensitivity and negative predictive value. 15,16Of note, not all PF4/heparin antibodies activate platelets and cause HIT, thus limiting the specificity and positive predictive value of immunoassays. 8Functional tests, eg, the washed-platelet heparin-induced platelet activation assay (HIPA) or serotonin release assay (SRA), can verify platelet-activating antibodies, thus confirming the presence of HIT. 10,17These functional tests are, however, technically challenging, and can only be performed in a few specialized laboratories.Thus, scientific societies recommend using the 4Ts score first.In case of intermediate or high risk, PF4/heparin antibodies should be determined using a validated assay. 15If PF4/heparin antibodies are also present, a functional test should then be performed to consolidate the diagnosis. 11,12However, little is known about the current practices and the performance of the recommended diagnostic algorithm in clinical practice.As part of a prospective, multicenter cohort study, we aimed to assess the diagnostic accuracy of currently recommended tests for HIT in clinical practice: (1) the 4Ts score, (2) the AcuStar HIT-IgG chemiluminescent immunoassay (CLIA), and (3) the diagnostic algorithm serially combining 4Ts score and CLIA.

Data Collection
A set of prespecified clinical characteristics and laboratory test results at diagnosis were collected by specially trained study nurses into an electronic case report form (REDCap database). 18,20,21In particular, the individual items of the 4Ts score were recorded as they were available to the treating physicians at the time of diagnosis; 4Ts scores were scored by the attending physician in conjunction with the consultant service.Specially trained study nurses transferred this data to the electronic case report form.If gross errors were found, they were corrected in consultation with the attending physicians and the principal investigator.In any case, only data available to the treating physicians at the time of diagnosis was used.The 4Ts score is an established scoring system to determine the clinical (pretest) probability of HIT (Table 1): 0 to 3 points indicates low risk; 4 to 5 points, intermediate risk; and 6 to 8 points, high risk. 22residual serum sample was obtained at the time of diagnosis, and it was frozen at −80 °C and transported on dry ice to the central laboratory Department of Clinical Chemistry, Inseptal, Bern University Hospital, Bern Switzerland.Follow-up was continued until discharge.

Determination of the CLIA
Within 1 week after arrival of the sample, PF4/heparin antibodies were quantified using a CLIA (HemosIL AcuStar HIT-IgG; Instrumentation Laboratory).CLIA was conducted on a BIO-FLASH (Inova Diagnostics) analyzer according to the manufacturers' instructions, as previously described. 23In brief, the assay was calibrated using calibrator 1 and calibrator 2 from the manufacturer, and samples were thawed rapidly at 37 °C.Internal quality controls were used before each run.The cutoff defined by the manufacturer was used, and samples with a result of 1.00 U/mL or higher were considered positive.

HIPA Test
As a reference standard, the presence of HIT was determined by a washed-platelet functional assay, the HIPA. 11,18[19]22,23 The HIPA was conducted as described in detail previously. 17

JAMA Network Open | Hematology
Accuracy of Diagnosing Heparin-Induced Thrombocytopenia

Diagnostic Accuracy
The 4Ts score correctly classified 101 patients as HIT positive and 615 as HIT negative (Table 3).The numbers of false negatives and false positives were 10 (9.0%) and 592 (49.0%), respectively.
Baseline characteristics of the individuals with a false-negative 4Ts score are available in the eTable in Supplement 1.The CLIA correctly identified 106 patients as HIT positive and 1134 as HIT negative.
The numbers of false negatives and false positives were 5 (4.5%) and 73 (6.0%), respectively.The currently recommended diagnostic algorithm (4Ts score followed by CLIA in case of an intermediateor high-risk 4Ts score) correctly identified 96 patients as HIT positive and 1157 as HIT negative.The numbers of false negatives and false positives were 15 (13.5%) and 50 (4.1%),respectively.
Sensitivities and specificities are shown in Table 3.Of note, the recommended diagnostic algorithm missed 13.5% of patients with HIT.In our dataset, the positive and negative predictive values were the proportion of false negatives, false positives, true negatives, and true positives as observed in our representative population.

Discussion
In this study, we examined the current diagnostic practice of managing clinical suspicion of HIT in a clinical setting using a prospective multicenter approach.By applying the current diagnostic algorithm, nearly half the patients would not require antibody testing.However, all tests, and especially the current diagnostic algorithm, still misclassify a concerning number of patients as either false negative or false positive.
Some studies have addressed the utility of current diagnostic instruments for HIT in clinical practice, and our results are essentially in line with these preliminary investigations.Linkins and colleagues 14 prospectively studied 526 of 1781 patients with a requested PF4/heparin assay using the SRA as reference standard (the prevalence of HIT was 6.1%).The sensitivity of the 4Ts score, representing the first step in the currently recommended algorithm, was only 81.3%, which would have missed a remarkable proportion of HIT patients. 14Similar to our results, the sensitivity of the PF4/heparin immunoassay was higher (100%; using rapid particle gel immunoassay). 14These results are, however, in striking contrast to a number of studies that report much higher sensitivity of the 4Ts score, which might potentially not reflect what happens in clinical practice. 13Similar to other studies, we have analyzed the accuracy of PF4/heparin immunoassays and found high sensitivities. 15,24,26,27In another study conducted in clinical practice, Gallo and colleagues 28 retrospectively included 319 patients in a 30-hospital US health care system, in which a previously implemented clinical decision support (CDS) fired during HIT immunoassay order entry indicating that the patient had a very low risk for HIT.Despite differences in study design, population, setting,

Figure 2 .
Figure 2. Diagnostic Performance of Diagnostic Tests for Heparin-Induced Thrombocytopenia in Clinical Practice 4Ts score A

Table 1 .
4Ts Score a 19A score of 0 to 3 points indicates low risk of heparin-induced thrombocytopenia; 4 to 5 points, intermediate risk; 6 to 8 points, high risk.Compiled from Lo et al.22detailed description of study design, setting, participants, study procedures, and collection of data are provided in previous publications that have addressed (1) the development of a decision support tool 18 and (2) the consistency of different heparin/PF4 immunoassays.19Inclusioncriteriawere(1)suspectedHIT(PF4/heparin antibodies ordered, 4Ts score applied, or consultancy service requested), (2) aged 18 years or older, and (3) written informed consent provided.Patients were included between January 2018 and May 2021.Patients were excluded in case of insufficient sample material, insufficient clinical data, or refused consent.The study was approved by all ethical committees (ie, Kantonale Ethikkommission Bern) and conducted in accordance with the Declaration of Helsinki.The manuscript was prepared following the Standards for Reporting of Diagnostic Accuracy (STARD) guideline.
25,24ief, serum was mixed with washed platelets from 4 different donors and placed on a microplate.Buffer, 0.2 IU/mL of low-molecular-weight heparin, or 100 IU/mL unfractionated heparin was added to the sample.The microplate was incubated for 45 minutes on a magnetic stirrer plate with 2 steel balls per well at 600 rpm.Platelet activation was observed every 5 minutes.The test was considered positive if aggregation occurred within 30 minutes in the presence of 0.2 IU/mL low-molecular-heparin, but not in the presence of 100 IU/mL heparin in at least 2 donors.17onpredefinedcutoffvalues.11,24Samplesizeconsiderations were reported in detail previously. 18 version 4.1.0(RProject for Statistical Computing) was used for statistical analysis and graphical illustrations of data.Sensitivity, specificity, positive predictive, and negative predictive values with 95% CIs were calculated using the epiR package.Area under the curve (AUC) on receiver operating characteristic (ROC) curves was calculated using the pROC package.25 Patient characteristics were reported grouped by HIPA test results as median with IQR because Gaussian criteria were not met.A 2 × 2 table was created, and diagnostic accuracy measures of (1) the 4Ts score, (2) the CLIA, and (3) the recommended diagnostic algorithm (4Ts score followed by CLIA)JAMA Network Open | HematologyJAMA Network Open.2024;7(3):e243786.doi:10.1001/jamanetworkopen.2024.3786(Reprinted) March 26, 2024 3/11 Downloaded from jamanetwork.comby guest on 03/31/2024 were calculated based

Table 2 .
Patient Characteristics a a Patient characteristics have previously been published.18,19bNew,recurrent, or progressive thromboembolism.

Table 3 .
Diagnostic Accuracy of the 4Ts score, CLIA, and the Recommended Diagnostic Algorithm Serially Combining 4Ts Score and CLIA Percentages are given according to the 2 × 2 table.
aReceiver operating characteristic curve of 4Ts score (A) and chemiluminescent immunoassay (CLIA; B) as with 95% CIs (shaded area).Diagnostic thresholds with 95% CIs are marked.AUC indicates area under the curve.