Salicylic acid and risk of colorectal cancer: a two sample Mendelian randomization study

Background Salicylic acid (SA) is a metabolite that can be obtained from the diet via fruit and vegetable ingestion, of which increased consumption has observationally been shown to decrease risk of colorectal cancer (CRC). Whilst primary prevention trials of SA and CRC risk are lacking, there is strong evidence from clinical trials and prospective cohort studies that aspirin (acetylsalicylic acid) is an effective primary and secondary chemopreventative agent. Since aspirin is rapidly deacetylated to form SA, it follows that SA may have a central role for aspirin chemoprevention. Through a Mendelian randomization (MR) approach, we aimed to address whether levels of SA affected CRC risk, and whether aspirin intake as a proxy for increased SA levels was required to identify an effect. Methods and Findings A two sample MR analysis was carried out using genome-wide association study summary statistics of SA from INTERVAL and EPIC-Norfolk (N= 14,149) and CRC from Colon Cancer Family Registry (CCFR), Colorectal Cancer Transdisciplinary Study (CORECT), Genetics and Epidemiology of Colorectal Cancer (GECCO) consortia and UK Biobank (55,168 cases and 65,160 controls). The Darmkrebs: Chancen der Verhutung durch Screening (DACHS) study (4,410 cases and 3,441 controls) was used for replication and stratification of aspirin-users and non-users. Single nucleotide polymorphisms (SNPs) for SA were selected via three methods: (1) Functional SNPs that influence aspirin and SA metabolising enzymes9 activity; (2) Pathway SNPs, those that are present in the coding regions of genes involved in aspirin and SA metabolism; and (3) genome-wide significant SNPs associated with levels of circulating SA. No association was found between the functional SNPs and SA levels, therefore they were not taken forward in an MR analysis. We identified 2 pathway SNPs (explaining 0.03% of the variance in SA levels and with an F statistic of 1.74) and 1 genome-wide independent SNP (explaining 0.05% of the variance and with an F statistic of 7.44) to proxy for SA levels. Using the pathway SNPs, an inverse variance weighted approach found no association between an SD increase in SA and CRC risk (GECCO OR:1.03, 95% CI: 0.84-1.27 and DACHS OR:1.10, 95% CI:0.58-2.07) and no association was found upon stratification between aspirin users and non-users in the DACHS study (OR:0.93, 95% CI:0.23-3.73 and OR:1.24, 95% CI:0.57-2.69, respectively). Wald ratio results using the genome-wide SNP also showed no association between an SD increase in SA and CRC risk (GECCO OR: 1.08, 95% CI:0.86-1.34 and DACHS OR: 1.01, 95% CI:0.44-2.31) and no effect was observed upon stratification by aspirin use (users OR:0.66, 95% CI: 0.11-4.12 and non-users OR: 1.12, 95% CI: 0.42-2.97). Conclusions We found no evidence to suggest that an SD increase in genetically predicted SA protects against CRC risk in the general population and upon stratification by aspirin use. However, based on the calculated variance explained by the SNPs and the F statistic, we acknowledge the possibility of weak instrument bias and the need to find better instruments for SA levels.