The quality of the methodology and reporting of studies on the treatment of early cervical cancer published in English and French language over the period 1975-1985 were examined using an explicit, pre-defined protocol aimed at assessing their internal validity and generalizability. One hundred and fifty-two articles reporting results on over 40,000 patients treated with surgery, radiotherapy, or the combination of the two, were examined. The astonishing lack of formal comparative studies together with the poor quality of those actually carried out were the two major findings of our study. More than half of the reviewed papers (54%) were single series studies. Among the remaining 46% only a few formally compared the two treatments (i.e. surgery vs. radiotherapy), the remainder dealing with comparisons of specific surgical or radiotherapeutic techniques. With reference to study quality, the existence of a pre-specified research protocol could not be ascertained in most studies. A description of patients' characteristics and information on the source population were deficient in most papers reviewed; information on the two aspects was in fact satisfactorily reported in only 7% and 47% of the papers, respectively. Finally, the lack of standardization of follow-up methods (i.e. type and modalities) and of information on treatment compliance were two other severe methodological deficiencies. In view of this poor quality and of the intrinsic difficulty of drawing firm scientific conclusions from non-experimental investigations, the reliability of this literature remains highly dubious. Another possible caveat is that what is published is a biased sample of the overall evidence because of the well known tendency of authors to write about and editors to publish positive results more frequently than negative ones.