Patterns of research. Oral contraceptives and cervical cancer.

Original papers on the oral contraceptive use-cervical cancer relationship are analysed. The purpose of this study was to ascertain the biases of the original articles collected in relation to various characteristics of any investigation. Papers were located by using MEDLINE, reviewing the references of each article identified by MEDLINE, and then reviewing the contents of those journals in which an original could be published. Fifty-five publications (from 49 original studies) were graded as to quality and classified as biased or unbiased. Nineteen studies were considered unbiased. The most common biases identified were confounding, detection bias, and misclassification bias. The pattern of research/publication has changed since the association began to be analysed: articles shift from gynecological to cancer and epidemiological journals; the number of studies performed by gynecologists alone and pathologists alone decreases, while studies performed by epidemiologists alone or in collaboration with gynecologists increase. This collaboration produced studies with fewer biases. It is suggested that the above mentioned collaboration should be increased to improve access to, and then the application of the results obtained in the original studies on oral contraceptives and cervical cancer.