Background: Cancer cachexia leads to reduced overall survival (OS) in mCRC. Novel therapeutics targeting the myostatin/activin pathway can reverse cachexia. Here we investigate the effect of myostatin/activin pathway gene expression and SNPs in mCRC patients (pts). Methods: Blood samples from 836 pts enrolled in 3 randomized first-line trials: TRIBE (FOLFIRI bevacizumab [bev]; FOLFOXIRI bev), FIRE-3 (FOLFIRI bev; FOLFIRI cetuximab [cet]) and MAVERICC (FOLFIRI bev; FOLFOX bev) were genotyped by Illumina OncoArray. The impact on outcome of 13 SNPs from 5 myostatin/activin genes (ACVR1B, ACVR2A, ACVR2B, MSTN, INHBA) was tested. Gene expression analysis included 433 mCRC pts treated with either bev (n = 226) or cet (n = 207) in combination with first-line chemotherapy from the CALGB/SWOG 80405 trial (NCT00265850). RNA isolated from FFPE tumor samples were sequenced with HiSeq 2500 (Illumina). OS and progression-free survival (PFS) were compared by gene expression tertiles: low (L), medium (M), and high (H). Logrank test was used for univariate (UV) analysis, Cox proportional hazards regression for multivariate (MV). Results: Pts with the germline G/G variant of ACVR2B rs2276541 had worse PFS (FIRE-3 bev: 9.7 vs 12.2 months [mo]; UV P = .044, MV P = .027; HR 2.10, 95%CI 1.13 - 3.38 & FIRE-3 cet: 7.6 vs 13.2 mo; UV P = .047; HR 1.65, 95%CI 1.00 - 2.72) and tumor response (TRIBE FOLFIRI bev: 48 vs 61%, MV P = .035) compared to any A carriers. Rs3749387 C/C variant in the same gene showed longer OS (MAVERICC FOLFOX bev: 36.9 vs 24.3 mo; MV P = .005; HR = 0.41, 95%CI 0.21-0.80) and PFS (MAVERICC FOLFIRI bev: 14.5 vs 11.8 mo; MV P = .009; HR = 0.53, 95%CI 0.32-0.88) relative to G variants. Any C allele in ACVR2A rs10497025 demonstrated improved OS (MAVERICC FOLFOX bev: 25.5 vs 11.9 mo; HR = 2.38, UV P = .035; 95%CI 1.03- 5.50 & TRIBE FOLFIRI bev: 26.9 vs 22.4 mo; UV P = .017; HR = 1.90, 95%CI 1.11-3.26) and PFS (FIRE-3 - FOLFIRI cet : 32.1 vs 12.2 mo; UV P = .047; HR = 0.47, 95%CI 0.22-1.02 & MAVERICC FOLFOX bev 11.0 vs 6.8 mo; UV P = .006; HR = 2.52, 95%CI 1.26-5.00) relative to G/G. Tumor ACVR1B-H expression was associated with longer PFS (L/M/H, 9.5 vs 10.9 vs 13.5 mo, UV P = .034) and OS (L/M/H, 25.0 vs 28.2 vs 38.6 mo, UV P< .001) in all treated pts. ACVR1B-H also conferred improved OS in treatment subgroups (L/M/H; cet: 21.0 vs 27.4 vs 40.3 mo, UV P< .001; bev: 25.9 vs 29.0 vs 36.9 mo, UV P = .025). ACVR2B-L correlated with worse PFS (L/M/H, 10.3 vs 12.2 vs 11.2 mo, UV P = .043) and OS (L/M/H, 24.7 vs 33.8 vs 34.9 mo; P = .006) in the overall cohort. This association held true in cet treated patients (OS 21.2 vs 35.8 vs 37.1 mo; UV P< .001) but not bev. Conclusions: Our results suggest a role of cachexia SNPs and gene expression as a prognostic marker for mCRC undergoing first line treatment. These biomarkers may enable targeted treatments for cancer cachexia and support exploring this approach in combination with standard therapy in selected pts.