Acute thrombocytopenia is one of the most common serious adverse reactions to drugs. Identification of drugs that can cause thrombocytopenia may occur by publication of case reports based on clinical evidence. However there is currently no consensus for adjudicating causality of adverse drug reactions, which still requires the readers’ judgment of the case report. Detection of drug-associated thrombocytopenia may also be facilitated by statistical algorithms used by health authorities and pharmaceutical companies to screen large databases of spontaneous reporting to look for statistical associations between reported drugs and thrombocytopenia. AERS is a computerized database of suspected adverse drug reaction reports provided by pharmaceutical manufacturers by regulation and voluntarily submitted by health care professionals and consumers. We have systematically reviewed all published case reports of drug-induced thrombocytopenia through August 2004, using one set of criteria to assess a significant association of a drug with thrombocytopenia (

Ann Int Med 129:886,1999
). We compared these data to results obtained by applying two mining algorithms to data derived from the US FDA AERS database. For purposes of this analysis a significant statistical association of a drug with thrombocytopenia from the AERS was defined as a signal of disproportionate reporting (SDR) that exceeded a standard predetermined value. SDRs do not necessarily reflect causality and can result from confounding or numerous reporting artifacts that plague spontaneous reports. 203 drugs have been reported as possibly causing thrombocytopenia in both published reports and in the FDA database. Among these 203 drugs, analysis of the case reports determined that a significant association with thrombocytopenia was present for 66 (33%) drugs; for 137 drugs, the association was not significant or the data were insufficient. In the data mining analysis SDRs exceeded the standard value for 135 (67%) of the drugs; the remainder were not associated with an SDR. However there was limited agreement between these 2 methods for identifying significant evidence for a drug association with thrombocytopenia.

Significant relation by both methods 48 drugs (24%) 
Not significant by either method 50 drugs (25%) 
Significant by case reports but not by data mining 18 drugs (9%) 
Significant by data mining but not by case reports 87 drugs (42%) 
Significant relation by both methods 48 drugs (24%) 
Not significant by either method 50 drugs (25%) 
Significant by case reports but not by data mining 18 drugs (9%) 
Significant by data mining but not by case reports 87 drugs (42%) 

For the 48 drugs for which a significant association with thrombocytopenia was determined by published reports and was also statistically distinctive by data mining, we determined which method provided earlier identification of drug-induced thrombocytopenia. For 7 drugs the evidence for a significant association occurred in the same year; for 21 drugs, the significant evidence from published case reports preceded the SDR; for 20 drugs, the SDR preceded the evidence from published case reports.

CONCLUSION. Neither published case reports nor data mining of the US FDA database are sufficient to identify all drugs with significant evidence for causing thrombocytopenia. The methods reported here have not been validated and implementation may vary between analysts/institutions. Data mining is a screening tool that may be more sensitive but less specific than reported clinical evidence. Use of multiple methods may enhance post-marketing surveillance for drug-induced thrombocytopenia.

Disclosure: No relevant conflicts of interest to declare.

Author notes

*

Corresponding author

Sign in via your Institution