Background: Overweight and obesity are associated with numerous adverse health outcomes. However, among older adults, substantial literature suggests an improved survival among overweight and obese patients. This phenomenon, referred to as the “obesity paradox” remains controversial. In the context of cancer, the association between overweight/obesity and mortality is complex due to the concomitant weight loss and cachexia. We aim to assess the impact of high Body Mass Index (BMI) on mortality in a large population of older cancer patients. Methods: We studied patients aged ≥70 from the ELCAPA prospective open-cohort (2007-2016; 10 geriatric oncology clinics, Great Paris urban area). Endpoints were 12- and 48-months mortality. A variable combining BMI at cancer diagnosis and weight loss (in the 6 months preceding the diagnosis) was created. BMI categories considered: underweight, (BMI < 22.4kg/m²), normal weight (BMI 22.5-24.9), overweight (BMI 25-29.9), and obese (BMI ≥30); weight loss (WL) categories: < 5%, 5- < 10%, ≥10%. Univariate and multivariate Cox proportional-hazards analysis were conducted in males and females. Results: A total of 2071 patients were included (mean age, 81; female, 48%; metastases, 49%; main localizations: digestive (37%), urinary (26%), breast (16%); underweight (30%), normal weight (23%), overweight (33%), obese (14%)). After adjustment for age, smoking, inpatient status, cancer site, metastasis, performance status and comorbidities, overweight women with < 5% WL were at lower risk of 12- and 48-months mortality compared to normal weight women with < 5% WL (aHR = 0.56; p = 0.043 and 0.65; p = 0.031, respectively). Obese women with < 5% WL had a lower mortality risk only at 48-months (aHR = 0.63; p = 0.036). The obesity paradox was stronger in overweight metastatic women. Overweight or obese women with 5- < 10% or ≥10% WL had no reduced risk of mortality compared to normal weight women. Overweight and obese men had no reduced risk of mortality irrespective of weight loss. Conclusions: By taking into account initial weight loss, we did not found evidence for obesity paradox in older patients with cancer except in the subgroup of women with minimal weight loss. Clinical trial information: NCT02884375