The obesity paradox for mid- and long-term mortality in older cancer patients: a prospective multicenter cohort study.

Background: Overweight and obesity are associated with adverse health outcomes. However, substantial literature suggests that they are associated with longer survival among older people. This "obesity paradox" remains controversial. In the context of cancer, the association between overweight/obesity and mortality is complicated by concomitant weight loss (WL). Sex differences in the relation between BMI (in kg/m2) and survival have also been observed.

Objectives: We studied whether a high BMI was associated with better survival, and whether the association differed by sex, in older patients with cancer.

Methods: We studied patients aged ≥70 y from the ELCAPA (Elderly Cancer Patients) prospective open cohort (2007-2016; 10 geriatric oncology clinics, Greater Paris urban area). The endpoints were 12- and 60-mo mortality. We created a variable combining BMI at cancer diagnosis and WL in the previous 6 mo, and considered 4 BMI categories-underweight (BMI < 22.5), normal weight (BMI = 22.5-24.9), overweight (BMI = 25-29.9), and obesity (BMI ≥ 30)-and 3 WL categories-<5% (minimal), 5% to <10% (moderate), and ≥10% (severe). Univariate and multivariate Cox proportional hazards analyses were conducted in men and women.

Results: A total of 2071 patients were included (mean age: 81 y; women: 48%; underweight: 30%; normal weight: 23%; overweight: 33%; obesity: 14%; predominant cancer sites: colorectal (18%) and breast (16%); patients with metastases: 49%). By multivariate analysis, obese women with WL < 5% had a lower 60-mo mortality risk than normal-weight women with WL < 5% (adjusted HR: 0.56; 95% CI: 0.37, 0.86; P = 0.012). Overweight/obese women with WL ≥ 5% did not have a lower mortality risk than normal-weight women with WL < 5%. Overweight and obese men did not have a lower mortality risk, irrespective of WL.

Conclusions: By taking account of prediagnosis WL, only older obese women with cancer with minimal WL had a lower mortality risk than their counterparts with normal weight.This trial was registered at clinicaltrials.gov as NCT02884375.