An important area of pharmacoepidemiology is the evaluation at the population level of the results of the introduction of evidence-based medicine into practice. The use of databases for this purpose, accumulating in a standardized and formalized form the results of the practical activities of many doctors, is very useful. Abroad, there are quite a number of these databases, including those that can give them access to a large circle of researchers.

This availability may however have a drawback. Misuse of data can lead to biased estimates and erroneous conclusions. This, in turn, leads to incorrect decisions and invalidates the research itself. The relevance of the creation and the reliability of the information contained in the databases is also questioned in this case.

In this regard, the review and editorial commentary of two publications published three months apart in the British Medical Journal (British Medical Journal) are very revealing. The title of the review of these publications, chosen by the review author, prof. D. Skegg from the Department of Preventive and Social Medicine, University of Otago, New Zealand - Pharmacoepidemiology traps - "Pharmacoepidemiology traps". We are talking about pharmacoepidemiological studies on thromboembolic complications in women using different generations of oral contraceptives.

Two research groups led by Farmer and Jick decided to determine whether the incidence of these complications had changed after a warning from the British Committee on Drug Safety. In October 1995, the latter reported that the risk of venous thromboembolism in women using the third generation of contraceptives (TPOK) was about twice as high as in women using drugs containing levonorgestrel.

Both research groups used the same general practitioner database for the analysis. However, their results contradict each other.

A research team led by Farmer did not detect a decrease in the number of thromboembolism cases following a change in the prescription rules for oral contraceptives in 1995 and concluded that the incidence of thromboembolic complications with TPOK is not different from that with other generations of oral contraceptives. In contrast, a research team led by Boston, led by Jick, has confirmed that there is a double risk of thromboembolism when using TPOK compared to drugs containing levonorgestrel. In addition, the frequency of complications reported after a warning from the British Committee on Drug Safety has been found to be lower.

In the examination of prof. Skegg provides possible reasons for the differences in the results of these studies. Farmer and his colleagues did not fully take into account the age of the patients and did not analyze the disruptive factors affecting the frequency of thromboembolism, limiting themselves to the study of correlation time relationships.

According to the critic, the study by Jick and his co-authors was more correct. They chose an adequate design - cohort studies and case-control nesting. Consideration of the interfering factors has been very important: a decrease in the frequency of use of TPOK mainly concerns young women at low risk of venous thromboembolism and the tendency of doctors to avoid the use of TPOK for women overweight or smokers.

Evaluating the statistical approaches which have led to the opposite conclusions, the critic draws attention to the fact that "simple studies have a rhetorical power which exceeds their scientific quality". The “null” result of two studies devalues their scientific significance, undermines the credibility of pharmacoepidemiological studies in general and the database of general practitioners in particular. The latter seems even more dangerous, since the aforementioned database has been used in more than 100 well-known studies, including on drugs that suppress appetite, heart valve diseases, pain relievers, gastrointestinal bleeding, antidepressants, etc. Probably the owner of the database - the Agency for Drug Control - should conduct a new analysis on this issue in order to eliminate the contradictions that have occurred. The mentioned database is widely available and, probably, potential researchers should be encouraged to be as strict as possible when using the data contained in the studies. Otherwise, an increased risk of bias and bias in publications will be inevitable.

Such a strategy will avoid false concerns about drug safety and will also help identify unexpected benefits associated with the use of drugs, such as the ability of statins to reduce the risk of fractures.

BMJ editors, represented by editor R. Smith, pleaded guilty to the Farmer et al. was accepted for publication, despite the presence of significant gaps. The version corrected by the authors after the initial examination differed significantly from the original (for example, the search volume was reduced by almost a third). However, the revised article was accepted for publication without being resubmitted to the reviewer, as the authors' arguments seemed convincing.

Already after the article was published by the editors, the inappropriate use of statistical procedures was identified, both with regard to the age assessment and the statistical power of the study to verify the increased risk of complications thromboembolism.

Revised Farmer et al. should be considered a beginner. This was not done due to the excessive length of the review process. In this regard, the publishers apologized to the authors, critics and readers of the review.

The example above shows that pharmacoepidemiology is a powerful tool that can only benefit patients and healthcare when its methods are applied appropriately.