Ovarian cancer and pregnancy: comment on a paper by Whittemore et al

We must concede that we were not able, with the published aggregate figures, to correct for the nuisance variables mentioned in the report, nor were we able to examine the individual "Study" results to derive estimates of "Between Study" variation, and so assess the regularity and statistical importance of the patterns observed. The tentative inferences outlined in this note are made therefore on the assumption that the analysis of these aggregate figures provides a fair reflection of the complete data base. Certainly, our analyses are numerically consistent with those in the report by Whittemore et al. (1). Because of the inconsistencies discovered between portions of the data presented by Whittemore et al. (1), we believe that the nature of the association between ovarian cancer and certain features such as nulliparity and age at first pregnancy remains unclear and demands further investigation, possibly with the very data base used by Whittemore and her colleagues.