The diagnosis of drug-induced thrombocytopenia can be supported only by the resolution of thrombocytopenia after discontinuation of therapy with the suspected drug. To avoid unnecessarily interrupting required therapy, the clinician must often decide whether to discontinue therapy with one or more of the patient's medications by assessing the probability that an agent is causing the thrombocytopenia. Reviews of drug-induced thrombocytopenia have resulted in extensive lists of drugs that have been reported to cause thrombocytopenia [2-5]. However, these lists include so many commonly used drugs that they do not help clinicians to decide which therapy to interrupt first.

To help clinicians better understand the likelihood that a drug will cause thrombocytopenia, we analyzed all published reports of drug-induced thrombocytopenia by using explicit, a priori criteria for establishing levels of evidence of a causal relation [6]. Our systematic review has two objectives: 1) to help clinicians determine which drugs are more likely to cause thrombocytopenia by distinguishing drugs for which evidence shows a definite or probable causal relation from those for which evidence is weaker and 2) to provide standardized criteria for use in reporting occurrences of drug-induced thrombocytopenia.

Methods

Top Methods Results Discussion Author & Article Info References

The MEDLINE database was searched for literature published from 1966 through 31 December 1997. Articles were sought by using the MeSH term thrombocytopenia with the attached subheading chemically-induced. Limits were set for human only and English language only. This reference list was supplemented by publications from 1989 to 1997 retrieved by using alternate search software (Reference Update, Research Information Systems, Inc., Carlsbad, California) and by cross-checking against the bibliographies of retrieved articles to identify additional reports, especially those published before 1966 (and therefore not included in the MEDLINE database). Articles reporting thrombocytopenia associated with heparin and heparin analogues were not retrieved because the etiologic relation of heparin to thrombocytopenia is well established [7].

Criteria for excluding articles from further review were 1) insufficient clinical data with which to evaluate the relation between drug administration and thrombocytopenia; 2) platelet count of 100 000 cells/[micro sign]L or more; 3) use of a known cytotoxic agent that causes marrow suppression; 4) cases in which the patient was exposed to a nontherapeutic agent or used an agent in a nontherapeutic manner (for example, environmental toxins, illicit drugs, drug overdose, and drugs not currently in use); 5) drug-induced disease that included thrombocytopenia but predominantly involved other abnormalities, such as aplastic anemia or the thrombotic thrombocytopenic purpura-hemolytic uremic syndrome; and 6) patient age of 16 years or younger. The age criterion was established because idiopathic thrombocytopenic purpura in children is typically an acute, spontaneously resolving illness that may not be distinguishable from drug-induced thrombocytopenia [1]. Current use of drugs in the United States was determined by listing in the American Hospital Formulary Service [8] and in other countries by listing in the Martindale Pharmacopoeia [9].

By using the criteria listed in Table 1, two of the authors independently reviewed each patient case report to establish the level of evidence for a causal role of the drug in thrombocytopenia. These criteria were established a priori for this study. Disagreement between the two reviewers was resolved by adjudication by a third independent reviewer. When a causal role for the drug was supported by level I or level II evidence, the clinical importance of the drug-induced thrombocytopenia was assessed by using standard criteria to define three levels of severity of bleeding [10]: 1) major bleeding (defined as intracranial or retroperitoneal bleeding) or overt bleeding (defined as visible or symptomatic bleeding) with a decrease of hemoglobin concentration by more than 2 g/dL (20 g/L) or the requirement for transfusion of two or more units of erythrocytes; 2) minor bleeding, defined as overt bleeding that did not meet the criteria for major bleeding (melena, gross hematuria, epistaxis or gingival bleeding that is prolonged for more than 30 minutes or requires medical intervention; excessive menstrual bleeding or vaginal bleeding other than menses); and 3) trivial bleeding, which included petechiae, purpura, brief epistaxis or gingival bleeding, guaiac-positive stool, or microscopic hematuria.

Selected articles are cited in this review. The full list of articles reviewed and the database established by this review are available at http://moon.ouhsc.edu/jgeorge. The authors intend to update this database annually.

Results

Top Methods Results Discussion Author & Article Info References

The literature search retrieved 581 articles after exclusion of articles reporting on heparin and heparin analogues. Twenty articles that were reviews and descriptions of laboratory studies were excluded because they contained no patient case reports. The remaining 561 articles contained 774 patient case reports, of which 259 were excluded: One hundred thirty contained insufficient clinical data with which to evaluate the criteria cited in Table 1, 19 had platelet counts of 100 000 cells/[micro sign]L or more, 14 reported on known cytotoxic drugs, 89 reported on nontherapeutic agents or use, 10 reported on multisystem disorders that included thrombocytopenia, and 48 reported on patients who were 16 years of age or younger. Some articles were excluded for more than one criterion.

The remaining 515 patient case reports involved 152 drugs. For these reports, the levels of evidence (Table 1) were I, definite (87 patient case reports); II, probable (160 reports); III, possible (172 reports); and IV, unlikely (96 reports). The initial two reviewers agreed on 452 (88%) of the 515 case reports. Adjudication was required for 63 (12%) of the patient case reports; in all cases, the third reviewer agreed with one of the two primary reviewers. Forty-eight drugs had level I evidence, and 50 other drugs had level II evidence. Table 2 lists the drugs that had level I evidence (which includes recurrent thrombocytopenia with rechallenge in the same patient) and drugs for which the causal relation to thrombocytopenia was validated by at least two reports with level II evidence.

Review of clinical outcomes found that in the 247 patient case reports with level I or level II evidence, 23 patients (9% [95% CI, 6% to 14%]) had major bleeding, including 2 patients (0.8% [CI, 0% to 3%]) who died of bleeding; 68 patients (28% [CI, 22% to 34%]) had minor bleeding; and 96 patients (39% [CI, 33% to 45%]) had trivial bleeding. Sixty patients (24% [CI, 19% to 30%]) had no bleeding symptoms. Both patients who died had quinine-induced thrombocytopenia [11]. When the sex (74% women) and mean age (54 years) of the 23 patients who experienced major bleeding were compared with the sex (57% women) and mean age (53 years) of patients without major bleeding, the difference was not significant (P > 0.1).

Analysis of the course of thrombocytopenia in the 87 patient case reports that met the criteria for level I evidence showed that the time of drug ingestion before the initial occurrence of thrombocytopenia was less than 1 day to 3 years (median, 14 days) and that the time to recovery of a normal platelet count was 1 to 30 days (median, 7 days). With rechallenge, the time to the nadir of thrombocytopenia was less than 1 to 60 days (median, 3 days) and the time to recovery was similar to the time to recovery after the initial occurrence of thrombocytopenia (<1 to 60 days [median, 5 days]).

Discussion

Top Methods Results Discussion Author & Article Info References

Reports of drug-induced thrombocytopenia are common, but our review found that most do not present evidence for the drug as a definite or probable cause of the thrombocytopenia. Because many patients are taking several drugs when thrombocytopenia is first discovered, clinicians need to know the probability of a causal role for each drug. Our systematic review provides the best available data with which to make a clinical decision about which drugs may more likely be implicated as a cause of thrombocytopenia and therefore should be discontinued. It also identifies which drugs require stronger evidence to support a causal relation and provides a structure for reporting future occurrences of drug-induced thrombocytopenia.

The drugs listed in Table 2 are similar to those found in a case-control study of drugs associated with acute thrombocytopenia [12]; in both analyses, quinidine and sulfonamides were among the most common drugs. One difference is that dipyridamole was identified in the case-control study as being associated with acute thrombocytopenia [12]; however, our systematic review identified no published case reports of dipyridamole-induced thrombocytopenia. Abciximab is not listed in Table 2 because all of the reported patients with abciximab-induced thrombocytopenia also received heparin [13] and therefore level II evidence could not be established (Table 1). However, a recent clinical trial established the causal relation of abciximab to thrombocytopenia with the observation that 35 of 630 patients (5.6%) who received abciximab developed thrombocytopenia with platelet counts less than 100 000 cells/[micro sign]L compared with 8 of 635 (1.3%) patients who received placebo (P < 0.001) [14].

Reports with level I (definite) evidence for a causal relation help to define the temporal relation between drug exposure and thrombocytopenia, because each of these patients had recurrent thrombocytopenia after rechallenge with the drug. Although the time from initiation of therapy with the drug to the initial episode of thrombocytopenia varied, the onset of thrombocytopenia after rechallenge was prompt (typically within 3 days). The time to recovery of a normal platelet count was prompt in both episodes, with medians of 7 and 5 days. This observation may have diagnostic value because prolonged thrombocytopenia after discontinuation of therapy with the suspected drug would be evidence against a causal role for that drug. An exception may be gold-induced thrombocytopenia. Many patients with presumed gold-induced thrombocytopenia had persistently low platelet counts, for many months, a condition attributed to the prolonged retention of gold salts; many had recurrent thrombocytopenia when prednisone therapy was discontinued; and some required splenectomy [15,16]. This clinical course is indistinguishable from idiopathic thrombocytopenic purpura, and it was therefore not possible to establish a causal role for gold in some reports.

In most patients, thrombocytopenia was discovered after the occurrence of minor purpura. Major bleeding occurred in 23 of 247 (9% [CI, 6% to 14%]) patients with level I or II evidence of drug-induced thrombocytopenia. Two deaths were attributable to thrombocytopenic bleeding (fatal bleeding rate, 0.8% [CI, 0% to 3%]). Both were reported in a letter to the editor describing quinine-induced thrombocytopenia [11]. Patients with major bleeding may be more likely to be recognized and reported; therefore, the occurrence of major bleeding due to drug-induced thrombocytopenia may be less than the estimate from these data. Nevertheless, the rate of 9% derived from this systematic review suggests that major bleeding associated with drug-induced thrombocytopenia is not rare and is clinically important.

In some reviews of drug-induced thrombocytopenia, demonstration of drug-dependent antiplatelet antibodies was included as evidence confirming a causal role of a drug [2,3,17]. However, there are no standard assays for drug-dependent antiplatelet antibodies, no standardized criteria for distinguishing positive from negative results, and no data on the sensitivity and specificity of these assays based on clinical criteria for a causal relation. Drug-dependent antiplatelet antibodies can be detected in patients who do not develop thrombocytopenia [7,18]. Prospective studies are needed to validate the role of assays for drug-dependent antibodies, in which results are available at the time of the clinical diagnosis and patients with negative laboratory test results continue therapy with the suspected drug with no adverse effects.

The limitations of our review are its restriction to English-language publications and the difficulty in assessing some patient case reports because they lacked detail. For example, whether other drugs were used or continued was sometimes not explicitly stated in the patient case reports (Table 1). In addition, the data provided on exclusion of other causes of thrombocytopenia varied (Table 1). These limitations, however, do not alter our major conclusion-that most reports do not present evidence for the drug as a definite or probable cause of thrombocytopenia. Future accurate assessment of a causal relation of drugs associated with thrombocytopenia will require inclusion of these explicit details.

Note: A complete list of drugs implicated in patient case reports can also be found in the electronic version of this paper at http://www.acponline.org.

Requests for Reprints: James N. George, MD, Hematology-Oncology Section, Department of Medicine, University of Oklahoma Health Sciences Center, Box 26901, Oklahoma City, OK 73190.

Current Author Addresses: Dr. George: Hematology-Oncology Section, Department of Medicine, University of Oklahoma Health Sciences Center, Box 26901, Oklahoma City, OK 73190. Mr. Raskob, Drs. Shah and Rizvi, and Mr. Osborne: Department of Medicine, University of Oklahoma Health Sciences Center, Box 26901, Oklahoma City, OK 73190.

Dr. Hamilton: Cancer Treatment Center, 230 North Midwest Boulevard, Suite 200, Midwest City, OK 73110.

Dr. Vondracek: College of Pharmacy, 1110 North Stonewall, Oklahoma City, OK 73104.