Under-reporting of adverse drug reactions Estimate based on a spontaneous reporting scheme and a sentinel system
- Cite this article as:
- Alvarez-Requejo, A., Carvajal, A., Bégaud, B. et al. E J Clin Pharmacol (1998) 54: 483. doi:10.1007/s002280050498
- 313 Downloads
Objective: Spontaneous reporting is the most common method used in pharmacovigilance and the best one to generate signals on new or rare adverse drug reactions (ADRs). Under-reporting is a major drawback of this system. The objective of this study was to quantify the extent of under-reporting in general practice and to assess the factors which influence it.
Methods: Details of ADRs collected through a short intensive survey were compared with primary care spontaneous reports received by the Castilla y Leon Regional Pharmacovigilance Centre during a 12-month reference period. The survey was undertaken by a random sample of 146 general practitioners (GPs), providing care to 149 487 people. The pharmacovigilance centre received reports concerning the whole regional population (2.5 million) covered by primary health care. The under-reporting coefficient (U) was estimated as the ratio between the number of effects observed by physicians in the survey and those spontaneously reported to the pharmacovigilance centre.
Results: The overall under-reporting rate was 1144 [95% confidence interval (CI): 928–1409]. Under-reporting was greater for psychiatric (2119; 945–4752) and gastrointestinal (1946; 1424–2659) disorders. Severe effects were more reported (U=605; 151–2431) than moderate (863; 473–1575) and mild (1209; 973–1503) ones. The under-reporting rate was lower for drugs recently marketed (706; 406–1230) and slightly lower for unlabelled effects (1031; 641–1657).
Conclusion: The under-reporting rate of ADRs is considerable, though not homogeneous for the different cases. This should be taken into account when comparing adverse effects (AEs) for different drugs. Under-reporting seems to be positively selective, as it involves mainly the less severe and better-known effects, preserving the value of spontaneous reporting for signal detection.