BackgroundHigher body mass index is associated with a higher incidence of colorectal cancer (CRC) but also with improved survival in metastatic CRC (mCRC). Whether weight change after mCRC diagnosis is associated with survival remains largely unknown. MethodsWe analysed individual patient data for previously untreated patients enrolled in five phase 3 randomised trials conducted between 1998 and 2006. Weight measurements were prospectively collected at baseline and up to 59.4 months after diagnosis. We used stratified multivariable Cox models to assess the prognostic associations of weight loss with overall and progression-free survival, adjusting for other factors. The primary end-point was a difference in overall survival (OS) between populations with weight loss and stable or increasing weight. FindingsData were available for 3504 patients. The median weight change at 3 months was −0.54% (IQR −3.9 … +1.5%). We identified a linear trend of increasing risk of death associated with progressive weight loss. Unstratified median OS was 20.5, 18.0, and 11.9 months (p < 0.001) for stable weight or gain, <5% weight loss, and ≥5% weight loss at 3 months, respectively. Weight loss was associated with a higher risk of death (<5% loss: aHR 1.18 [1.06–1.30], p < 0.002; ≥5% loss: aHR 1.87 [1.67–2.1], p < 0.001) as compared to stable or increasing weight at 3 months post-baseline (reference), while adjusting for age, sex, performance, and a number of metastatic sites. InterpretationPatients losing weight during systemic therapy for metastatic colorectal cancer have significantly shorter OS. The degree of weight loss is proportional to the observed increased risk of death and remains evident among underweight, normal weight, and obese individuals. On-treatment weight change could be used as an intermediate end-point. FundingThe creation and management of the database containing the individual patient data from the original randomised trials is supported by the Aide et Recherche en Cancérologie Digestive Foundation.