Welding fumes were recently classified as carcinogenic to humans and worldwide millions work as welders or perform welding operations. The purpose of this study was to identify new biomarkers of welding-induced carcinogenesis. We evaluated a panel of 91 putative cancer-related proteins in serum in a cohort of welders working with mild steel (n = 77) and controls (n = 94) from southern Sweden sampled on two occasions 6-year apart using a longitudinal analysis (linear mixed models). The significant results from the longitudinal analysis were tested for reproducibility in welders (n = 88) and controls (n = 69) sampled once during the same sampling period as timepoint 1 or timepoint 2 (linear regression models), i.e., in a cross-sectional setting. The models were adjusted for age, body-mass index, and use of snus. All study participants were non-smokers at recruitment. Exposure to welding fumes was assessed using questionnaires and respirable dust measurement in the breathing zone that was adjusted for personal respiratory protection equipment. The median respirable dust in welders was 0.7 (0.2-4.2) and 0.5 (0.1-1.9) mg/m 3 at the first and second timepoints, respectively. We identified 14 cancer-related proteins that were differentially expressed in welders versus controls in the longitudinal analysis, out of which three were also differentially expressed in the cross-sectional analysis (cross-sectional group). Namely, syndecan 1 (SDC1), folate receptor 1 (FOLR1), and secreted protein acidic and cysteine rich (SPARC) were downregulated, in welders compared with controls. In addition, FOLR1 was negatively associated with years welding. Disease and function analysis indicated that the top proteins are related to lung cancer as well as cell invasion and migration. Our study indicates that moderate exposure to welding fumes is associated with changes in circulating levels of putative cancer-related proteins, out of which FOLR1 showed a clear dose-response relationship. It is, however, unclear to which extent these changes are adaptive or potential early biomarkers of cancer.