Causal Inference in Cancer Epidemiology: What Is the Role of Mendelian Randomization?

Abstract

Observational epidemiologic studies are prone to confounding, measurement error, and reverse causation, undermining robust causal inference. Mendelian randomization (MR) uses genetic variants to proxy modifiable exposures to generate more reliable estimates of the causal effects of these exposures on diseases and their outcomes. MR has seen widespread adoption within cardio-metabolic epidemiology, but also holds much promise for identifying possible interventions for cancer prevention and treatment. However, some methodologic challenges in the implementation of MR are particularly pertinent when applying this method to cancer etiology and prognosis, including reverse causation arising from disease latency and selection bias in studies of cancer progression. These issues must be carefully considered to ensure appropriate design, analysis, and interpretation of such studies. In this review, we provide an overview of the key principles and assumptions of MR, focusing on applications of this method to the study of cancer etiology and prognosis. We summarize recent studies in the cancer literature that have adopted a MR framework to highlight strengths of this approach compared with conventional epidemiological studies. Finally, limitations of MR and recent methodologic developments to address them are discussed, along with the translational opportunities they present to inform public health and clinical interventions in cancer. .

A functional variant near XCL1 gene improves breast cancer survival via promoting cancer immunity

“…To address this question, we applied MR analysis, which mimics a randomized controlled trial by using a genetic variant as the IV to be the proxy for exposure. The MR approach provides supportive evidence for a causal relationship between exposure, mediator and disease outcome . In this regard, we considered gene expression to be more directly regulated by genetic variants than environmental factors and biological traits and used it as the proxy for exposure, whereas environmental factors and biological traits have been more commonly considered the proxy for exposure in published MR studies .…”

“…The MR approach provides supportive evidence for a causal relationship between exposure, mediator and disease outcome. 24 In this regard, we considered gene expression to be more directly regulated by genetic variants than environmental factors and biological traits and used it as the proxy for exposure, whereas environmental factors and biological traits have been more commonly considered the proxy for exposure in published MR studies. 37,38 Notably, the eQTLs used as IVs in the first step of MR (XCL1 eQTLs rs1024176 and rs982143) and in the second step of MR (cis-eQTLs for BTLA, C1orf54, CADM1 and CLEC9A) were retrieved from the METABRIC dataset, which is completely independent of our TWBC cohort.…”

“…On the basis of the instrumental variable (IV) assumption, genetic variants, when used as IVs, should be reliably associated with the exposure of interest but not any unmeasured confounders and should affect the outcome solely through this exposure . Although many GWAS‐identified SNPs have been used as IVs to perform Mendelian randomization (MR) analysis, how these SNPs were associated with the exposure was uncertain.…”

“…On the basis of the instrumental variable (IV) assumption, genetic variants, when used as IVs, should be reliably associated with the exposure of interest but not any unmeasured confounders and should affect the outcome solely through this exposure. 24 Although many GWAS-identified SNPs have been used as IVs to perform Mendelian randomization (MR) analysis, how these SNPs were associated with the exposure was uncertain. In contrast, significant cis-eQTLs for a particular gene are thought to be more directly associated with the expression of that gene at the molecular level, which convinced us that cis-eQTLs meet the necessary assumptions to be considered as IVs for gene expression, a kind of exposure, in the conventional MR approach.…”

“…Having found the association between XCL1, DC1s and breast cancer survival, we then applied Mendelian randomization (MR) analysis to confirm their causal relationship using two independent datasets (i.e., two-sample MR). 24 In the first step, using the METABRIC dataset, 29 the top cis-eQTL for XCL1, rs982143, was chosen as the IV, by which significant associations between rs982143 and XCL1 expression and between rs982143 and ER-positive breast cancer survival were observed in TWBC cohort (p = 1.26 × 10 −4 ). It is notable that rs982143 was not associated with rs1024176 (r 2 = 0.1654, TWBC cohort), excluding the possibility of linkage disequilibrium.…”