Pitfalls of pharmacoepidemiologyBMJ 2000; 321 doi: https://doi.org/10.1136/bmj.321.7272.1352 (Published 25 November 2000) Cite this as: BMJ 2000;321:1352
- Richard Farmer, professor,
- Tim Williams, research officer,
- Alison Nightingale, research officer
- Pharmacoepidemiology and Public Health, Postgraduate Medical School, University of Surrey, Guildford, Surrey GU2 7DJ
EDITOR—Skegg1 is concerned about the differences between our study2 and that of Jick et al.3 We share his concerns but are disappointed that he did not discuss the crucial differences between the studies and how these might affect the findings.
The general practice research database comprises data from general practitioners that are collated, quality controlled, and distributed by the Medicines Control Agency. We normally use all available data or exceptionally use well-defined subsets. Our recent study focused on practices that contributed data continuously between 1993 and 1998.2 Jick et al used an undefined subset.3 They asserted that “over the past 12 years we have removed over half of the original practices because of inadequate data quality” without defining the criteria used. We estimate that they have used at most 55% of the database. Studies by Jick et al are difficult to interpret because they use different subsets in different studies. Their paper on venous thromboembolism and combined oral contraceptives reported on two studies.4 The first used a population of 303 470 women and the second 238 470 women, and …