The aim of this systematic review and meta-analysis is to evaluate whether the design and methodology of drug-treatment studies of premature ejaculation affect the efficacy outcome differently. Therefore, methodological, design and efficacy data from 79 studies (3034 males), published between 1943 and 2003, are reviewed. A meta-analysis is performed on 43 selective serotonin reuptake inhibitors (SSRIs) and clomipramine studies (1514 males), published between 1973 and 2003; these studies were pooled to provide a summary variance-weighted effect size. The antidepressant-induced percentage increase of the intravaginal ejaculation latency time (IELT) was calculated and examined against various methodological items. A significant difference in efficacy between SSRIs was observed. Using daily treatment, paroxetine appeared more effective than the other SSRIs. Retrospective use of a questionnaire, subjective reports, single-blind and open study designs generate far greater variability of ejaculation time both at baseline and during active drug treatment than real time assessment by stopwatch. In conclusion, at daily treatment, the overall efficacy of paroxetine, clomipramine, sertraline and fluoxetine is comparable, but paroxetine exerts the strongest ejaculation delay. Only eight (18.5%) studies on antidepressant treatment fulfilled all criteria used in evidence-based medicine, for example, randomised, double-blind studies with prospective real time (stopwatch) assessment of the IELT at each intercourse. Single-blind studies, open designs, retrospective reporting, or the use of a questionnaire to assess ejaculation time should be avoided.