Purpose: We investigated whether changes in body mass index (BMI) before a breast cancer diagnosis affected mortality and whether trajectories more accurately predict overall mortality compared to a single measure of BMI. Methods: Our prospective cohort comprised 2012 women with breast cancer who reported their weight in each decade from 20 to 50–64 years of age. We used trajectory analysis to identify groups with similar development patterns in BMI and Cox proportional hazards models to examine the association between trajectory groups and mortality, and interactions with oestrogen receptor status and smoking. We used c-index statistics to compare the trajectory model with the single measure model of BMI. Results: We identified three distinct trajectory groups, with a mean BMI at age 20 of 19, 22 and 24 increasing to 23 (normal-to-normal), 29 (normal-to-overweight) and 37 (normal-to-obese) at 50–64 years of age, respectively. Women in the normal-to-obese trajectory group experienced significantly higher overall mortality than those in the normal-to-normal trajectory group (HR 1.76, 95% CI 1.21‒2.56). The association declined to a non-significant level after adjustments for clinical prognostic factors. Although not significant, the same tendency was seen for breast cancer-specific mortality. The association was strongest in women with oestrogen receptor-negative tumours. Weight changes over time were not significantly different from a single BMI measure before diagnosis to predict survival. Conclusion: Weight gain affects overall mortality after breast cancer but clinical prognostic factors largely eliminate the association. Using trajectories of weight changes did not improve the predictive value compared to a single measure of BMI.