A systematic evaluation of laboratory testing for drug-induced immune thrombocytopenia

Authors


Donald M. Arnold, Department of Medicine, McMaster University, 1280 Main Street West, Hamilton, Ontario, Canada.
Tel.: +1 905 521 2100 ext. 76305; fax: +1 905 521 4971.
E-mail: arnold@mcmaster.ca

Abstract

Summary.  Background: Drug-induced immune thrombocytopenia (DITP) can be confirmed by the demonstration of drug-dependent platelet antibodies in vitro; however, laboratory testing is not readily accessible and test methods are not standardized. Objective: To identify drugs with the strongest evidence for causing DITP based on clinical and laboratory criteria. Patients/Methods:  We developed a grading system to evaluate the quality of DITP laboratory testing. The ‘DITP criteria’ were: (i) Drug (or metabolite) was required for the reaction in vitro; (ii) Immunoglobulin binding was demonstrated; (iii) Two or more laboratories obtained positive results; and (iv) Platelets were the target of immunoglobulin binding. Laboratory diagnosis of DITP was considered definite when all criteria were met and probable when positive results were reported by only one laboratory. Two authors applied the DITP criteria to published reports of each drug identified by systematic review. Discrepancies were independently adjudicated. Results: Of 153 drugs that were clinically implicated in thrombocytopenic reactions, 72 (47%) were associated with positive laboratory testing. Of those, 16 drugs met criteria for a definite laboratory diagnosis of DITP and thus had the highest probability of causing DITP. Definite drugs were: quinine, quinidine, trimethoprim/sulfamethoxazole, vancomycin, penicillin, rifampin, carbamazepine, ceftriaxone, ibuprofen, mirtazapine, oxaliplatin and suramin; the glycoprotein IIbIIIa inhibitors abciximab, tirofiban and eptifibatide; and heparin.Conclusions: We identified drugs with the strongest evidence for an association with immune thrombocytopenia. This list may be helpful for ranking potential causes of thrombocytopenia in a given patient.

Introduction

Thrombocytopenia is a common hematological abnormality, especially in hospitalized patients. Many factors can contribute to the development of thrombocytopenia, including infection, hemodilution and drugs; consequently, determining the cause can be difficult. Drugs may cause thrombocytopenia by a variety of mechanisms, including myelosuppression (e.g. chemotherapeutic agents), dose-dependent impairment in platelet production (e.g. linezolid) or an idiosyncratic immune reaction. The latter is a unique syndrome called drug-induced immune thrombocytopenia (DITP), which is caused by drug-dependent antibodies that bind to platelets. The thrombocytopenia that ensues is typically severe and may result in life-threatening bleeding [1].

The diagnosis of DITP is often missed [2,3]; yet, accurate and timely recognition is important to differentiate it from primary immune thrombocytopenia [4] so that future exposures to the drug and unnecessary treatments are avoided [5]. Clinical criteria have been developed to determine the likelihood of a drug being implicated in a DITP reaction. These criteria are that: the drug exposure must precede the onset of thrombocytopenia; recovery from thrombocytopenia must be complete and sustained after the drug is discontinued; other causes of thrombocytopenia, including other drugs, must be excluded; and a drug re-challenge (if done) must result in recurrence of the thrombocytopenia [6–8]. However, clinical criteria alone are often insufficient for making the diagnosis of DITP.

Laboratory testing demonstrating antibodies that bind to platelets only in the presence of the drug can confirm the diagnosis. The difficulty is that in practice, testing is rarely accessible because it is only done in a few specialized centers and real-time results are generally not available to help with patient management. Another problem with DITP testing is that over the years, a variety of test methods have been used, some of which lack standardization and are of questionable validity.

The aim of this study was to identify drugs with the highest probability of causing DITP based on both clinical and laboratory criteria. We developed a grading system to evaluate the methods of DITP laboratory testing and applied it to published reports of individual drugs.

Methods

Development of the DITP laboratory criteria

Four of the authors experienced in DITP testing (DA, JS, TW and JK) developed laboratory criteria by consensus to evaluate the validity of DITP test methods and the reliability of the results (the ‘DITP criteria’). The three validity criteria reflected the characteristics of the pathogenic antibody and the quality of test methods: (i) antibody binding had to be dependent on the drug (or its metabolite); (ii) the test had to demonstrate specific immunoglobulin binding; and (iii) platelets had to be the target of immunoglobulin binding. One reliability criterion was also incorporated: positive testing had to be obtained by two or more independent laboratories using (the same or different) patient samples. Definitions of each of these criteria were established by consensus (Table 1). The laboratory diagnosis of DITP was considered definite when all criteria were met and probable when valid test results were reported by only one laboratory (Table 2).

Table 1.   DITP criteria used to evaluate the quality of laboratory testing for drug-induced immune thrombocytopenia. Requirements for each criterion were established by consensus
CriteriaRequirement
Drug or drug metabolite dependenceDissolved drug or drug metabolite must be added to the test in vitro. Drug metabolites may be commercially available or prepared from post-ingestion urine samples.
Immunoglobulin bindingThe test must use quantitative methods to determine the amount of immunoglobulin binding (e.g. flow cytometry or a competitive immunoglobulin binding assay) or must use purified immunoglobulin preparations (instead of untreated serum or plasma) as the starting material for the test. Adequate controls with normal serum must be used to exclude the non-specific immunoglobulin binding to platelets.
Two separate positive test results obtained by two or more laboratoriesPositive test results using valid test methods must have been obtained by two (or more) laboratories. The same or different patient samples could have been used.
Platelet specificityStarting material for the test must be washed platelets, platelet-rich plasma or purified platelet antigens, or the test must detect antibodies that recognize platelet antigens.
Table 2.   Levels of evidence to support a laboratory diagnosis of drug-induced immune thrombocytopenia (DITP)
Level of evidenceDefinition
DefiniteAll validity criteria for DITP laboratory testing were met (1, drug or drug metabolite dependence; 2, immunoglobulin binding; and 3, platelet-specificity) and reliability criterion was met (positive results obtained using valid methods by two or more independent laboratories).
ProbableAll validity criteria were met but testing was not confirmed by two different laboratories.
UnlikelyOne or more validity criteria were not met.

Identification of DITP drugs associated with laboratory testing

We performed a systematic review of the literature to identify published reports of drugs that were clinically implicated in DITP reactions and that had laboratory testing done for drug-dependent platelet antibodies. We started with a list of drugs clinically associated with thrombocytopenia compiled using a comprehensive literature search [7,9] (full list of clinical drugs available at http://moon.ouhsc.edu/jgeorge/DITP.html). Drugs were included in our systematic review if they had a definite or probable association with DITP based on the following clinical criteria: thrombocytopenia followed drug exposure; thrombocytopenia improved once the drug was stopped; other causes of thrombocytopenia, including other drugs, were excluded; and a drug re-challenge (if done) reproduced the thrombocytopenia. Drugs for which a re-challenge was not documented were also included. In addition, we searched MEDLINE and EMBASE in duplicate and independently from 1940 to February 2012 for English language publications using the search terms ‘antibody’ and ‘blood platelets’ and the name of each individual drug to identify those publications that reported laboratory testing. Our search was supplemented by a recent review, which reported test results for 67 drugs from one reference laboratory [10], and by manually searching bibliographies of primary studies and review articles.

Evaluation of DITP laboratory testing

Two assessors experienced in platelet antibody testing independently applied the DITP criteria to published reports of laboratory testing. Agreement between reviewers was measured by Cohen’s kappa (κ). Discrepancies were resolved by consensus or independent adjudication when consensus was not achieved.

Results

Drugs meeting clinical and laboratory criteria for DITP

We identified 153 drugs that were clinically implicated in DITP reactions. Of those, 72 (47%) had a positive laboratory test performed for drug-dependent platelet antibodies. Laboratory testing associated with 36 drugs did not meet validity criteria because of insufficient evidence to support drug (or drug metabolite) dependent antibody binding (n = 13) or because testing did not confirm immunoglobulin binding to platelets (n = 18). Agreement between reviewers on whether drugs met validity criteria was excellent (κ = 0.9). Disagreements were mostly due to differences in the interpretation of a positive platelet-associated immunoglobulin G test, which after discussion was deemed to be adequate if drug dependence was confirmed. Third-party adjudication was needed to resolve discrepancies in validity assessments of testing associated with five drugs, all of which were ultimately excluded because validity criteria were either not met or could not be evaluated due to insufficient reporting (Fig. 1).

Figure 1.

 Selection of drugs that met clinical and laboratory criteria for drug-induced immune thrombocytopenia.

There were 36 drugs that met all validity criteria for laboratory testing. Of those, 16 were associated with positive test results obtained from two or more independent laboratories and thus had a definite laboratory diagnosis of DITP. Those drugs were: quinine, quinidine, trimethoprim/sulfamethoxazole, vancomycin, penicillin, rifampin, carbamazepine, ceftriaxone, ibuprofen, mirtazapine, oxaliplatin and suramin; the glycoprotein IIbIIIa inhibitors abciximab, tirofiban and eptifibatide; and heparin (Table 3). The other 20 drugs had positive testing performed by only one laboratory and thus met criteria for a probable laboratory diagnosis of DITP (Table 4).

Table 3.   ‘Definite’ drugs that met all clinical and laboratory requirements for drug-induced immune thrombocytopenia, including confirmation by more than one laboratory (n = 16)
DrugAuthorDrug added in vitroMethod of detecting antibody bindingEvidence of platelet specificity
  1. *Heparin includes unfractionated heparin, low-molecular-weight heparin, polysulfated chondroitin sulphate and other glycosaminoglycans. PSIFT, platelet suspension immunofluorescence test; MAIPA, monoclonal immobilization of platelet antigens; FcR, Fc receptor; MACE, modified antigen capture enzyme linked immunosorbent assay; PAIFT, platelet adherent immunofluorescence test; EIA, enzyme immunoassay

Classic (quinine) type
 QuinidineAster et al. (10)Drug added to platelets and buffersFlow cytometryWashed donor platelets used as starting material
Chong et al. (26)Quinidine (1.5 mmol/L) added to plateletsPSIFT; MAIPAWashed donor platelets used as starting material; antibodies to platelet glycoproteins detected
 QuinineBougie et al. (27)Quinine (0.4 mmol/L) added to platelets and buffersFlow cytometry, MACEWashed donor platelets used as starting material; antibodies to platelet glycoproteins detected
Chong et al. (26)Quinine (1.5 mmol/L) added to platelets and buffersPSIFT; MAIPAWashed donor platelets used as starting material; antibodies to platelet glycoproteins detected
 Trimethoprim/ sulfamethoxazoleKiefel et al. (28)Trimethoprim, sulfamethoxazole, post-ingestion urine and metabolites added to platelets and buffersPAIFT, EIAWashed donor platelets used as starting material
Aster et al. (10)Drug added to platelets and buffersFlow cytometryWashed donor platelets used as starting material
 Vancomycinvon Drygalski et al. (25)Vancomycin (0.3 mg/mL) added to platelets and buffersFlow cytometryWashed donor platelets used as starting material
Ganly et al. (29)Vancomycin (0.3 mg/mL) added to test serumPSIFTWashed donor platelets used as starting material; antibodies to platelet glycoproteins detected
 PenicillinGarbe et al. (17)Benzylpenicillin added to platelets (1 mg/mL) and to buffers (0.1 mg/mL)MAIPAAntibodies to platelet glycoproteins detected
Salamon et al. (30)Penicillin G (1 × 105 units) added to plateletsPlatelet eluate used; platelet antiglobulin testWashed donor platelets used as starting material
 RifampinBurgess et al. (31)Rifampin (70–700 μg mL) added to platelets and buffersFlow cytometry, MAIPA, immunoprecipitationWashed donor platelets used as starting material; antibodies to platelet glycoproteins detected
Aster et al. (10)Drug added to platelets and buffersFlow cytometryWashed donor platelets used as starting material
 CarbamazepineShechter et al. (32)Carbamazepine (0.001 mol/L) added to platelets and buffersPSIFTWashed donor platelets used as starting material
Aster et al. (10)Carbamazepine added to platelets and buffersFlow cytometryWashed donor platelets used as starting material
 CeftriaxoneGarbe et al. (17)Ceftriaxone added to platelets (1 mg/mL) and to buffers (0.1 mg/mL)MAIPAAntibodies to platelet glycoproteins detected
Grossjohann et al. (33)Ceftriaxone (1 mg/mL) added to plateletsFlow cytometry, MAIPA, immunoprecipitation and immunoblot assayPlatelets used as starting material; antibodies to platelet glycoproteins detected
 IbuprofenAster et al. (10)Ibuprofen metabolite added to platelets and to buffersFlow cytometryWashed donor platelets used as starting material
Meyer et al. (34)Post-ingestion urine (containing metabolites) added to plateletsFlow cytometryWashed donor platelets used as starting material
 MirtazapineLiu et al. (35)Mirtazapine (1 mmol/L) added to platelets and buffersFlow cytometry, MAIPA, immunoprecipitationWashed donor platelets used as starting material; antibodies to platelet glycoproteins detected
Garbe et al. (17)Mirtazapine added to platelets (1 mg/mL) and to buffers (0.1 mg/mL)MAIPAAntibodies to platelet glycoproteins detected
 OxaliplatinCurtis et al. (36)Oxaliplatin added to platelets and buffer in the same concentrationFlow cytometry, MACEWashed donor platelets used as starting material; antibodies to platelet glycoprotein IIbIIIa detected
Taleghani et al. (37)Oxaliplatin added to platelets (1 mg/mL) and buffers (0.1 mg/mL)MAIPA, EIADonor platelets used as starting material; antibodies to platelet glycoprotein IIbIIIa detected
 SuraminTisdale et al. (38)Suramin (0–1.3 mg/mL) added to plateletsPurified immunoglubuin G usedPlatelet aggregation and platelet serotonin release detected.
Aster et al. (10)Drug added to platelets and buffersFlow cytometryWashed donor platelets used as starting material
Glycoprotein IIbIIIa inhibitors
 AbciximabCurtis et al. (39)Abciximab (0.7 μg per 1.6 × 107 platelets) addedFlow cytometryWashed donor platelets used as starting material
Nurden et al. (40)Abciximab (5 μg/mL) added to plateletsFlow cytometry, MAIPAWashed donor platelets used as starting material; antibodies to platelet glycoproteins detected
 TirofibanBougie et al. (41)Tirofiban (1 μg/mL) added to platelets and buffersFlow cytometry, antigen capture ELISAWashed donor platelets used as starting material; antibodies to platelet glycoproteins detected
Clofent-Sanchez et al. (42)Tirofiban added to platelets and buffersFlow cytometry, MAIPAWashed donor platelets used as starting material; antibodies to platelet glycoproteins detected
 EptifibatideBougie et al. (41)Eptifibatide (2 μg/mL) added to platelets and buffersFlow cytometry, antigen capture ELISAWashed donor platelets used as starting material; antibodies to platelet glycoproteins detected
Greinacher et al. (43)Eptifibatide added to plateletsPurified IgG used, flow cytometry, ELISAAntibodies to platelet glycoproteins detected
Heparin*
 HeparinSheridan et al. (8)Heparin (0.1 and 100 U/mL) added to plateletsInhibition by FcR-blocking antibodyWashed donor platelets used as starting material
Greinacher et al. (44)Heparin (0.2 or 100 U/mL) added to plateletsInhibition by FcR-blocking antibodyPlatelet aggregation was measured
Table 4.   ‘Probable’ drugs that satisfied all criteria for drug-induced immune thrombocytopenia but testing was confirmed by only one laboratory (n = 20)
DrugTest method
  1. *Although there were two or more publications for some drugs in this list, testing for drug-induced platelet antibodies was performed by a single laboratory. Gold induces autoantibodies, not typical drug-dependent antibodies.

AcetaminophenFlow cytometry(13,45)*
AmiodaroneFlow cytometry(10)
AmlodipineEnzyme-linked immunosorbent assay and monoclonal antibody immobilization of platelet antigen assay(17)
AmpicillinFlow cytometry(10)
CephamandolePlatelet suspension immunofluorescence test (46)
CiprofloxacinFlow cytometry(10)
DiazepamCromium-51 platelet lysis test(11)
EthambutolFlow cytometry(10)
FurosemideEnzyme-linked immunosorbent assay and monoclonal antibody immobilization of platelet antigen assay(17)
GoldEnzyme immunoassay(47)
HaloperidolFlow cytometry(10)
LorazepamEnzyme-linked immunosorbent assay and monoclonal antibody immobilization of platelet antigen assay(17)
NaproxenFlow cytometry(13)
PhenytoinFlow cytometry(10)
PiperacillinFlow cytometry(10,48)*
RanitidineFlow cytometry(49)
RosiglitazoneFlow cytometry; monoclonal antibody immobilization of platelet antigen assay(50)
RoxifibanEnzyme-linked immunosorbent assay; immunoblot assay(51)
SulfisoxazoleFlow cytometry(52); Cromium-51 platelet lysis test(11)*
TranilastIndirect platelet-associated immunoglobulin G(53)

Description of laboratory methods

A variety of test methods for drug-dependent platelet antibodies were considered valid, including flow cytometry, immunofluorescence, monoclonal-based antigen capture assays and platelet-associated immunoglobulin G. The most common test was flow cytometry, which was used in 26 (74%) of 35 drugs (excluding heparin). Test methods that were not judged to be valid included the macrophage inhibition assay and the platelet agglutination test, neither of which provided sufficient evidence for platelet antibody binding. Conversely, other functional assays such as the serotonin release assay with a confirmatory step demonstrating platelet Fc-receptor binding and the chromium-51 platelet lysis test with purified immunoglobulin fractions [11] met all criteria.

Discussion

Thrombocytopenia is common in medical practice. It may occur as an idiosyncratic reaction to certain drugs that stimulate the production and binding of platelet antibodies [12]. Laboratory testing can confirm the diagnosis of DITP by demonstrating drug-dependent platelet antibodies in vitro; however, real-time results are generally not available and cannot help with the management of individual patients. Using data from published reports, we compiled a list of drugs that satisfied all clinical and laboratory criteria for DITP. To achieve this goal, we developed a grading system that incorporated three validity criteria and one reliability criterion to evaluate the quality of laboratory testing. Using this approach, we were able to reduce the initial list of 153 clinically implicated drugs to 16 drugs with the highest pre-test probability of DITP.

Drug dependence was defined as the addition of the drug or its metabolite to the test in vitro and the demonstration of antibody binding to platelets in the presence but not in the absence of the drug. This criterion highlights some of the challenges inherent in drug-dependent platelet antibody testing, including the lack of availability of certain drug metabolites for use in the assay [13], uncertainty about optimal drug concentrations for in vitro testing and the chemical properties of certain drugs, which make them difficult to solubilise in test buffers. Evidence for immunoglobulin binding required the use of a quantitative method for the detection of bound immunoglobulins, typically flow cytometry [14]; the use of a competitive inhibition assay; or purified immunoglobulins, rather than serum or plasma, as the starting material for the test. Many older functional assays failed to meet this requirement, because immunoglobulin binding was inferred but not objectively demonstrated. The criterion for platelet specificity required the use of washed platelets or platelet-rich plasma as the target for antibody binding or the demonstration of antibody specificity for platelet glycoproteins with monoclonal antibodies. Finally, the requirement for a positive test result from two or more laboratories was incorporated in our grading scheme as a measure of test reliability. Over half of the drugs (21/36; 58%) with valid test results were excluded for this reason, which may reflect diversity of test methods, lack of expertise outside of a small number of reference laboratories or inherent differences between patients in their susceptibility to this immune drug reaction. Our findings highlight the variability in current test methods for this important clinical problem, which has motivated the Platelet Immunology Scientific Subcommittee of ISTH to embark on a project to standardize methods for DITP testing.

Previous investigators have produced lists of DITP drugs. Reese et al. [15] cross-referenced an online list of clinical drugs with a US national voluntary-reporting database and a reference laboratory. We identified four drugs (other than heparin) with a definite laboratory diagnosis (ceftriaxone, mirtazapine, penicillin and suramin) and 11 others with valid test methods from a single laboratory (amlodipine, cephamandole, ciprofloxacin, diazepam, furosemide, gold, rosiglitazone, roxifiban and tanilast) that were not identified by Reese et al. Conversely, Reese et al. identified three drugs (irinotecan, valproic acid and simvastatin) that were not captured using our approach. These differences are likely to be due to the fact that we included testing from all laboratories, whereas Reese et al. restricted their analysis to one reference laboratory, and Reese et al. [15] used clinical information from a national US adverse event reporting database and published reports, whereas we only used published reports. Drug surveillance reports from France and Germany [16,17] produced more expansive lists of implicated drugs. Our study took a systematic approach to consider clinical and laboratory levels of evidence in support of the diagnosis and came up with a more selective list.

We included heparin in our list of drugs, even though heparin-induced thrombocytopenia (HIT) is associated with some unique features that distinguish it from classic DITP: HIT is associated with thrombosis rather than bleeding and the thrombocytopenia is typically less severe [18]. Heparin also provides illustrative exceptions to our DITP rules: for one, it has been associated with drug-induced but not drug-dependent antibodies in the syndrome of delayed-onset HIT [19]; it may cause antibodies that bind to platelets in a drug-dependent fashion without causing thrombocytopenia [20]; and because the pathologic heparin-dependent antibodies are transient, a heparin re-challenge only produces recurrent thrombocytopenia when administered in close proximity to an episode of HIT [21]. Nevertheless, because it has been proven to cause thrombocytopenia in many reports, heparin (which includes unfractionated heparin, low-molecular-weight heparins, polysulfated chondroitin sulphate and other glycosaminoglycans) was included in our list of DITP drugs. Similarly, the GPIIbIIIa inhibitors (abciximab, tirofiban and eptifibatide) have several unique characteristics, including the development of thrombocytopenia within hours [10]. They too were included because of their clear association with thrombocytopenia.

Clinical grading systems using specific criteria have been applied to various areas of hematology, including DITP, HIT [22] and venous thromboembolism [23]. We applied a similar methodology to evaluate complex laboratory testing. This type of approach may be used to standardize new laboratory techniques and to ensure validity of historical tests when a gold standard is lacking. Our laboratory criteria focused on the development of IgG antibodies binding to platelets, which is the most common mechanism of DITP; thus, our criteria would not capture antibodies that target megakaryocytes [24], complement-mediated thrombocytopenia and cell-mediated drug reactions. A fifth criterion may be considered for future prospective evaluations of DITP testing: the absence of antibody binding to platelets in individuals taking the drug who do not develop thrombocytopenia, as has been shown for heparin [20] and vancomycin [25]. Although the list we compiled represents those drugs with the strongest evidence for an association with DITP, such a list should be used with caution because other drugs should be considered in any patient with typical clinical features and a compatible temporal association. Usual practice would be to stop all possible drugs in a patient with new onset thrombocytopenia.

Establishing the diagnosis of DITP is important because early recognition will ensure that the implicated drug is discontinued and inappropriate treatments for other thrombocytopenic disorders are avoided. Many patients who develop thrombocytopenia are exposed to one or more culprit drugs and the challenge confronting physicians is to determine the cause, generally without the availability of sophisticated laboratory tests. Our list of 16 drugs may help identify the most likely cause of thrombocytopenia in a given patient and our use of a grading system can serve as a model for the evaluation of other complex laboratory tests.

Addendum

D.M. Arnold conceived and designed the research, collected and abstracted the data, interpreted the results, wrote the manuscript and approved the final version; S. Kukaswadia designed the research, collected and abstracted the data, analyzed and interpreted the results, edited the manuscript and approved the final version; I. Nazi analyzed and interpreted the results, edited the manuscript and approved the final version; A. Esmail collected and abstracted the data, interpreted the results, edited the manuscript and approved the final version; L. Dewar collected and abstracted the data, analyzed the results, edited the manuscript and approved the final version; J. W. Smith analyzed and interpreted the results, edited the manuscript and approved the final version; T. E. Warkentin conceived the research, analyzed and interpreted the results, edited the manuscript and approved the final version; J. G. Kelton conceived the research, interpreted the results, edited the manuscript and approved the final version.

Acknowledgements

We thank C. Kearon for his critique of the manuscript.

Funding

Supported by the Canadian Institutes for Health Research.

Disclosure of Conflict of Interests

The authors have no conflict of interests to declare.

Ancillary