Purpose: The mechanisms underlying muscle wasting in patients with cancer remain poorly understood, and consequently there remains an unmet clinical need for new biomarkers and treatment strategies. Experimental Design: Microarrays were used to examine the transcriptome in single biopsies from healthy controls (n = 6) and in paired biopsies [pre-resection baseline (weight-loss 7 and 8 month post-resection follow-up (disease-free/weight-stable for previous 2 months)] from quadriceps muscle of patients with upper gastrointestinal cancer (UGIC; n = 12). Results: Before surgery, 1,868 genes were regulated compared with follow-up (false discovery rate, 6. Ontology analysis showed that regulated genes belonged to both anabolic and catabolic biologic processes with overwhelming downregulation in baseline samples. No literature-derived genes from preclinical cancer cachexia models showed higher expression in baseline muscle. Comparison with healthy control muscle (n = 6) revealed that despite differences in the transcriptome at baseline (941 genes regulated), the muscle of patients at follow-up was similar to control muscle (2 genes regulated). Physical activity (step count per day) did not differ between the baseline and follow-up periods (P = 0.9), indicating that gene expression differences reflected the removal of the cancer rather than altered physical activity levels. Comparative gene expression analysis using exercise training signatures supported this interpretation. Conclusions: Metabolic and protein turnover?related pathways are suppressed in weight-losing patients with UGIC whereas removal of the cancer appears to facilitate a return to a healthy state, independent of changes in the level of physical activity.