A prospective multipractice study based on well-defined data collected by a number of general practitioners (GPs) over a period of time is often used in primary health care research. However, problems of research methodology such as selection bias are common in such studies when GPs either refuse to participate or interrupt their participation. A method is described whereby selection bias is avoided. In the present study, a medical laboratory made a daily record of all abnormal liver function tests requested by GPs over a period of six months. When mailing these results to the GPs, a questionnaire was enclosed in order to obtain a number of clinical data from them. Thirty of 31 possible GPs participated. During the study period, questionnaires concerning 134 patients were mailed, and 118 of these were returned without reminders. Following one reminder the response rate was 100%. Comparing the number of requested tests during the study period with the number requested during a control period before and after the study period (same six month periods of the year), no difference was seen in the pattern of requests. Thus, this method, unbiased with respect to both selection and request-behaviour, is recommended for similar studies.