BTC1.03 Investigating the causal relevance of hypothesized risk factors for glioma using population level genetic data.

Abstract

Background

Gliomas are a group of primary brain tumours, the most common and aggressive subtype of which is glioblastoma (GBM). GBM has a median survival of just 15 months after diagnosis. Only previous exposure to ionizing radiation and certain inherited genetic syndromes are accepted risk factors for glioma: the majority of cases are thought to occur spontaneously.Observational studies have described associations between several risk factors and glioma, but studies are often conflicting and whether these associations reflect true casual relationships is unclear because observational studies are prone to confounding, measurement error and reverse causation. Mendelian randomization (MR) is an approach to causal analysis that utilises genetic variants robustly associated with risk factors of interest as instrumental variables to determine whether the associations between the trait of interest and disease are causal. We aimed to investigate the causal relevance of hypothesized risk factors for glioma incidence.

Material and Methods

Hypothesized risk factors for glioma were first identified from published literature by conducting a search of PubMed. We employed two-sample MR using summary results from genome-wide association studies (GWAS) to index 24 potential risk factors that were investigated in relation to GWAS data 4572 glioma cases and 3286 controls. The outcome data coincided with the SNPs used for exposures in the analysis. We determined the odds ratio (OR) for each risk factor with glioma using the inverse variance weighted (IVW) method. To evaluate the robustness of our results and to ensure the MR assumptions were not violated by the presence of directional pleiotropy we performed weighted median, weighted mode and MR-Egger regression.

Results

The MR analysis suggested a causal association between a genetically predicted increase in telomere length and an increase in glioma risk by IVW (OR per standard deviation [SD] increase: OR:4.46, CI:2.28–8.73), by weighted median (OR:2.74, CI:1.37–5.50), MR Egger (OR:5.77, CI:0.16–200.19) and weighted mode (OR:2.56, CI:1.28–5.11). There was no strong indication that this result could be a result of pleiotropy. An inverse association between height and glioma risk was observed by IVW (OR per SD increase in height (OR:0.82 CI:0.73–0.92), weighted median (OR:0.74, CI:0.61–0.89), MR Egger (OR:0.59, CI:0.41–0.83) and weighted mode (OR:0.57, CI:0.36–0.88). For all other risk factors evaluated, we found no evidence for a causal association with glioma.

Conclusion

These results confirm previous evidence that genetically predicted increase in telomere length, increases the risk of glioma. Our results may provide novel evidence that height is associated with a decreased risk of glioma.