Infertility treatment: from cookery to science. The epidemiology of randomised controlled trials

ABSTRACT

Objectives To review the epidemiology of published randomised controlled trials in infertility treatment over the last 25 years, with special emphasis on the number and quality of trials.

Design Computer literature review by MEDLINE backed up by a manual search of 41 journals. Each trial was classified according to the methodology described and quality criteria. The results were recorded in a computer database. Odds ratios (OR) and confidence intervals (CI) were calculated where the data were sufficient.

Subjects Couples suffering from primary or secondary infertility. The trials studied 33 761 patients overall.

Setting Institute of Epidemiology and Health Services Research, Leeds.

Results Five hundred and one randomised trials in male and female infertility treatment were identified between 1966 and 1990. Pregnancy was an outcome in 291 (58%) and these were the subject of detailed analysis. Two hundred and twenty-four (77%) and 67 (23%) 'pregnancy trials' were concerned, respectively, with female and male infertility. Four per cent of the trials were preceded by a sample size calculation, and the average sample size was 96 patients (range 5–933); 700 patients per group would be required to demonstrate plausible success rates for most treatments. The method of randomisation was unstated or pseudo-randomised in 206 (71%) of trials where pregnancy was an outcome. Only 29 (5.8%) of studies were multicentre. The method of confirmation of pregnancy was omitted for 70% of papers. Cross-over design was used in 103 (21%) of cases. Meta-analysis is possible for selected topics such as the use of anti-oestrogens in idiopathic oligospermia and unexplained female infertility. Eight cases of double reporting were identified.

Conclusions Trials using randomised methodology were relatively few in comparison with other branches of medicine, although their use is important in the evaluation of treatment for infertility as treatment-independent pregnancy is common. It was encouraging to note that an exponential increase in the use of this methodology occurred during the last three years, especially in association with assisted conception techniques, and meta-analysis has become possible for selected topics. However, many trials suffer from an unrealistically small sample size, inappropriate use of cross-over design or pseudo-randomisation. The trend towards properly controlled studies should be encouraged but these studies should be of improved quality and organised on a multicentre or even international basis.