Diagnostic criteria for cancer cachexia: reduced food intake and inflammation predict weight loss and survival in an international, multi-cohort analysis

Background Cancer-associated weight loss (WL) associates with increased mortality. International consensus suggests that WL is driven by a variable combination of reduced food intake and/or altered metabolism, the latter often represented by the inflammatory biomarker C-reactive protein (CRP). We aggregated data from Canadian and European research studies to evaluate the associations of reduced food intake and CRP with cancer-associated WL (primary endpoint) and overall survival (OS, secondary endpoint). Methods The data set included a total of 12,253 patients at risk for cancer-associated WL. Patient-reported WL history (% in 6 months) and food intake (normal, moderately, or severely reduced) were measured in all patients; CRP (mg/L) and OS were measured in N = 4960 and N = 9952 patients, respectively. All measures were from a baseline assessment. Clinical variables potentially associated with WL and overall survival (OS) including age, sex, cancer diagnosis, disease stage, and performance status were evaluated using multinomial logistic regression MLR and Cox proportional hazards models, respectively. Results Patients had a mean weight change of -7.3% (+/- 7.1), which was categorized as: +/- 2.4% (stable weight; 30.4%), 2.5-5.9% (19.7%), 6.0-10.0% (23.2%), 11.0-14.9% (12.0%), >= 15.0% (14.6%). Normal food intake, moderately, and severely reduced food intake occurred in 37.9%, 42.8%, and 19.4%, respectively. In MLR, severe WL (>= 15%) (vs. stable weight) was more likely (P < 0.0001) if food intake was moderately [OR 6.28, 95% confidence interval (CI 5.28-7.47)] or severely reduced [OR 18.98 (95% CI 15.30-23.56)]. In subset analysis, adjusted for food intake, CRP was independently associated (P < 0.0001) with >= 15% WL [CRP 10-100 mg/L: OR 2.00, (95% CI 1.58-2.53)] and [CRP > 100 mg/L: OR 2.30 (95% CI 1.62-3.26)]. Diagnosis, stage, and performance status, but not age or sex, were significantly associated with WL. Median OS was 9.9 months (95% CI 9.5-10.3), with median follow-up of 39.7 months (95% CI 38.8-40.6). Moderately and severely reduced food intake and CRP independently predicted OS (P < 0.0001). Conclusions Modelling WL as the dependent variable is an approach that can help to identify clinical features and biomarkers associated with WL. Here, we identify criterion values for food intake impairment and CRP that may improve the diagnosis and classification of cancer-associated cachexia.