we estimated relative risks as odds ratios (ORs) with 95% confidence intervals (Cls) by multiple logistic regression analyses. All the models were conditional on 5-year age groups.
Univariate associations between breast cancer and each of the above-mentioned potential confounders were evaluated and included in the final models with the shift-work variable, for adjusted Ors, if the p value was 0.15 (Pearson chi-square test for categorical variables and t test for continuous variables) or if inclusion of a covariate changed the odds ratio in the model with shift-work by 15%. The final models included, in addition to the shift-work variables, breast cancer in mother or sister, use of hormone replacement therapy (never), lifetime cumulative months of breastfeeding , age at first child birth (nulliparous,),age at menarche (continuous),menstrual regularity, weight fluctuations and menstrual status. The reported significance levels are two-sided.