The difference in estimates of odds ratio for reproductive factors and premenopausal breast cancer in case-control studies may be due to selection bias from the use of different types of controls. In a national case-control study in Norway a total of 105 patients (92% of those eligible) under 40 years of age were interviewed. Three different series of controls were selected: two population controls (71% of all eligible) matched on date of birth were drawn from a central register covering all residents in Norway; one neighborhood control (72%) matched on month and year of birth and postal address; and one hospital control (89%) matched on a 5-year age group. Matched, adjusted analysis gave similar estimates of odds ratio for population and neighborhood controls for age at menarche, age at first full term pregnancy, and parity. Use of hospital controls yielded different estimates for age at meanarche and age at first full term pregnancy; the latter was found to be a stronger risk factor with use of hospital controls. The study indicates that the use of hospital controls can give estimates of odds ratios for reproductive histories which differ from estimates based on population controls.
Supported by a grant from the Norwegian Cancer Society (Landsforeningen mot kreft).