Identification and evaluation of a possible signal of exacerbation of colitis during rofecoxib treatment, using Prescription-Event Monitoring data.

Abstract

Identifying previously unrecognized adverse drug reactions (signals) is an important part of post-marketing surveillance. Automated signal generation now forms part of routine surveillance of spontaneous adverse drug reactions reported to the UK Yellow Card system. The Drug Safety Research Unit (DRSU) provides an additional post-marketing drug surveillance scheme in the UK, using the non-interventional observational cohort technique of Prescription-Event Monitoring (PEM), and systematically collects data on patients who were prescribed selected newly licensed drugs in primary care clinical practice. The introduction of a new computer system in January 2000 enabled the development of an automated signal generation system to compliment the process of identification of possible safety signals in drug data collected using PEM. To use incidence rate ratios (IRRs) as a form of signal generation in the DSRU database, with particular interest in rofecoxib, plus further evaluation of any signal(s) of interest by requesting additional information from the general practioner (GP) of each relevant case. Crude IRRs were calculated for each event term of interest by comparing the incidence rate for each lower term event reported in rofecoxib users with the comparable incidence rate for the whole PEM database (77 drugs of various therapeutic classes) from the total person-exposure time up to 180 days after starting the drug. To investigate a possible class effect, IRRs were also calculated using a second comparator cohort including only those PEM study drugs within the same therapeutic class (non-steroidal anti-inflammatory drugs, NSAIDs) and used for similar indications. Cases arising out of potential signals were followed up for additional information. A potential signal of 'colitis' was identified when rofecoxib was compared with the rest of the database, IRR 5.8 (95% CI 2.7,11.3; z statistic 5.6), and when the comparator group was changed to include only the other four NSAIDs, IRR 4.3 (95% CI 1.4,14.5; z statistic 2.8). Other possible signals, compared with the rest of the database, included events deemed to be related to the underlying condition, labelled adverse events and events describing effectiveness of treatment. Further evaluation of this signal revealed that there were 11 reports of colitis (two reports for one patient) that occurred while taking rofecoxib and within 180 days of exposure. A causality assessment for these 10 patients did not confirm an association with newly developing colitis, but showed that the events were 'possible' exacerbations of pre-existing colitis during treatment with rofecoxib. The use of IRRs for signal generation using PEM data is worthwhile and enables time to be taken into account. Previously unrecognized adverse events require further evaluation to confirm that a possible safety signal exists. In this study, the number of patients reported to have colitis was small but compared with other drugs on the database, the incidence rate was relatively high. Follow-up revealed a possible relationship between exacerbation of pre-existing colitis and treatment with rofecoxib. Hypotheses arising from signal generation require evaluation and cannot be taken as a conclusive evidence for clinical differences in the safety profiles of drugs.