Lay Summary: Exclusion of patients from “observational” drug safety studies

Background

In clinical trials it is commonly considered necessary to exclude many patients who might use a medicine in ordinary practice – for example those who have specific diseases or who use other medicines. One potential advantage of drug safety studies which are based on information from ordinary practice is that they may better reflect real-life conditions. These studies do not involve any intervention by the researchers and are known as “observational” studies. Because there is a lack of intervention, methods are required to address the problem that observed adverse events may be caused by factors other than the medicine(s) of interest. One such method is to exclude patients with these potential “confounding” factors. Patients may also be excluded from such studies for other reasons – for example because it was not possible to be sure that the diagnosis of the event was correct. However, as more patients are excluded, it becomes less likely that a study will reflect real-life.

Excluding patients from observational studies of drug safety could therefore have important effects on their results but there seems to have been no previous research conducted to examine the types of exclusion criteria applied.

Methods

We identified 10 senior researchers who had published extensively in the field of drug safety. All their relevant study publications between 1999 and 2004 were identified and reviewed to extract the exclusion criteria which were applied. A classification system was developed to categorise the criteria. There were 5 main categories and 11 sub-categories.


Results

A total of 200 studies were included. The average number of exclusion criteria sub-categories applied was 3.8 with a range between researchers of 2.8 to 5.1. The most common categories applied were: data quality and validation (87% of studies), patient characteristics (75%) and disease-related (69%). Drug-related (38%) and miscellaneous exclusions (3%) were less commonly applied. There were marked variations between authors, particularly in the application of sub-categories “risk factors and alternative causes (0-81%) and “medication of interest”* (5-93%).


Conclusions

Exclusion criteria are frequently applied in observational drug safety studies but there are important differences in approach between researchers. It is likely that a substantial part of the variations we observed reflect differences in research practice. Further work is required to assess the impact of exclusions on study results and there may be a need to agree a standard approach to the use of exclusion criteria in such studies.


* meaning that some users of the medication being studied were excluded.