BACKGROUND: Evidence from prospective observational studies suggests that elevated circulating C-reactive protein (CRP) concentrations are associated with cancer risk, but it is unclear whether this association is causal. In order to examine this, we investigated whether genetic variants that are associated with circulating CRP concentrations are associated with cancer risk. METHODS: We pooled data from three population-based prospective Finnish studies: FINRISK 1992 (n = 5289), FINRISK 1997 (n = 7160) and Health 2000 (n = 6299). Cancer cases were identified from cancer registrations. Thirteen CRP-associated SNPs, identified from genome-wide association studies, were genotyped. We examined the associations of the SNPs and cancer risk using Cox, probit and instrumented probit regression models. RESULTS: Compared to common allele homozygotes, individuals carrying one or two variant T alleles at rs1892534 had 1.05-fold (95% confidence interval (CI): 0.90, 1.23) and 1.2-fold (95% CI: 1.01, 1.42) increased overall cancer risk, respectively. Individuals with one or two variant A alleles at rs1169300 or rs2464196 had approximately 1.5- and 2-fold increased risk of lung cancer, respectively (p trend for both: 0.007). CRP SNPs were not associated with colorectal, prostate or breast cancer risk nor was CRP-associated with the probability of developing cancer in the instrumented probit analyses. CONCLUSIONS: We found some evidence for an association of a small number of CRP-associated SNPs with the overall cancer risk and lung cancer risk. Our instrumental variable analyses provided no clear evidence for a causal association of CRP and cancer. These findings suggest that circulating CRP concentrations are unlikely to have a causal role in cancer.