Reassessment of casual association between body mass index and site-specific solid cancers: A comprehensive Mendelian randomization study

doi:10.21203/rs.3.rs-3552788/v1

Download PDF

Research Article

Reassessment of casual association between body mass index and site-specific solid cancers: A comprehensive Mendelian randomization study

https://doi.org/10.21203/rs.3.rs-3552788/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Purpose

Previous Mendelian randomization (MR) studies on association between body mass index (BMI) and cancer only focused on several common cancers, and the genetic instruments varied, making the effect of BMI on other cancers unclear and direct comparison across all cancer types difficult. Our comprehensive MR aims to reinforce the established evidence of the association.

Methods

We conducted a unified MR analysis to reassess the association between BMI and 19 site-specific solid cancers. Genetic instruments for BMI were extracted from genome-wide association studies (GWAS) of 806,834 European individuals and 86,757 East Asian individuals, respectively. The associations of genetic instruments with 19 site-specific solid cancer (sub)types extracted from the FinnGen study, the Biobank Japan (BBJ) project, and international consortia were examined by using various MR methods.

Results

For lung, endometrial and breast cancers in the FinnGen study, estimation of the MR analysis aids the triangulation of the positive or inverse association between genetically predicted BMI and these (sub)types. In several large consortia in the European population, our results strengthen confidence in previously established positive findings for lung, ovarian, endometrial and breast cancer and null findings for prostate cancer. In the BBJ project, no causal association was found between BMI and cancers, although a possible increased risk for endometrial cancer may exist.

Conclusion

Our comprehensive MR estimates reinforce the causal role of BMI in some solid cancers in the European population but cannot rule out a possible causal role in some other cancer (sub)types and in other populations.

Mendelian randomization

Body mass index

Cancer

Genetic epidemiology

Over the past several decades, with a few exceptions, human society has experienced rapid development of the social economy, accompanied by the global popularization of the Western lifestyle, which is characterized by overnutrition and being more sedentary, leading to the pandemic of overweight and obesity (Blüher 2019). Based on data from the Global Burden of Disease (GBD) 2015, overweight or obesity affects over 2.1 billion people, or nearly 30% of the global population, and the prevalence will become more serious if post-2000 trends continue (The GBD 2015 Obesity Collaborators 2017; NCD Risk Factor Collaboration 2016). The epidemic of overweight or obesity is accompanied by a significant increase in the burden of obesity-related noncommunicable diseases, including cardiovascular disease, type 2 diabetes mellitus, non-alcoholic fatty liver disease, and certain types of cancer (Lin and Li 2021).

Although several metrics have been established to assess the degree of adiposity, body mass index (BMI) is still widely accepted. Therefore, the associations between high BMI and obesity-related cancers have been extensively investigated, leading to the consensus that high BMI is the third most attributable behavioural risk factor behind smoking and infection globally and the second behind smoking in Western countries (Renehan and Soerjomataram 2016). Previous well-conducted studies based on conventional observational studies have established strong evidence for an association between high BMI and increased risk of various cancers, although there are slight differences in affected tumour types among these reports (Lauby-Secretan et al. 2016; Kyrgiou et al. 2017). In GBD 2019, 13 cancer types were found to be affected by high BMI, and the estimated number of high BMI-related cancer cases accounted for 4.59% and 4.45% of all cancer-cause deaths and disability-adjusted life years (DALYs), respectively (Zhi et al. 2022). However, these conventional observational studies are susceptible to potential confounding bias, and the interpretation of the findings from these reports is complicated with reverse causality (Fang et al. 2022).

Owing to the ability to address some of these limitations, Mendelian randomization (MR) studies have received increasing attention in recent years and have been utilized to investigate the effects of high BMI on various cancers. The rationale of MR is the random assortment of genes from parents to offspring, leading to the fixation of germline genetic variants that are robustly associated with exposures at conception without being modified by subsequent factors (Davey Smith and Hemani 2014). Therefore, MR offers evidence for the causality of putative BMI-cancer associations as randomized controlled trials while remaining in an observational setting (Fang et al. 2022; Davey Smith and Hemani 2014). Despite these strengths, several points can be informed from studies (summarized in Supplementary Table 1) on associations between high BMI and cancer using MR methods. First, most MR studies included populations with European ancestry, leaving the evidence in other populations limited. Second, nearly all studies focused on several common cancers, such as colorectal cancer (CRC), lung cancer and breast cancer, and reported a significant association between high BMI and these cancer types. Whether the absence of other cancer types in the current literature is due to no associations between high BMI and these cancers or no data are available for them is not clear. Third, even though the genetic instruments were selected from the same data source, the Genetic Investigation of Anthropometric Traits (GIANT) and/or the UK Biobank (UKB), the number of single nucleotide polymorphisms (SNPs) varied across studies. In addition, genome-wide association studies (GWAS) on cancer are obtained from different large consortia, making the comparison of the effects of high BMI and its magnitude across cancer types difficult. Finally, some studies did not use sex-specific genetic instruments to match with a cancer occurring in sex-specific organs, such as breast and prostate cancer, while sex instrumental heterogeneity has been revealed to have a nonignorable impact on sex-specific two-sample MR studies, and the causal effects of exposures on sex-specific outcomes would be biased if sex-combined genetic instruments are incorrectly employed (Gao et al. 2021).

Therefore, the aim of this study is to reassess the association between high adult BMI (hereafter BMI) and solid cancers using MR in a comprehensive way to investigate the effects and compare the magnitude of high BMI across most of the cancer (sub)types. For this aim, we selected SNPs from a large collaborative meta-analysis of GWASs combining data from the GIANT consortium and the UKB with European ancestry to assess the causal effects of high BMI on various cancers from the same population recruited in the FinnGen study. In complementary analyses, we also replicate the MR analysis using the same set of SNPs for cancers whose genetic summary data from large-scale GWAS consortia are available. In addition, we applied the same methods using publicly available summary statistics in the East Asian population.

An overview of the workflow of the selection of genetic instruments and genetic summary data on cancers and analytical methods are listed in Fig. 1. This study was conducted following the reporting guidelines of the Strengthening the Reporting of Observational Studies in Epidemiology Using Mendelian Randomization (STROBE-MR, Supplementary Table 2) (Skrivankova et al. 2021). All studies providing data for our MR analysis were approved by a relevant ethical review board, and participants provided informed consent. All data used in this MR analysis are publicly available genetic summary data from anonymous databases, and no additional ethical approval from an institutional review board was needed.

Cancer data sources Genetic summary data for each cancer were obtained from the FinnGen study and several large international consortia in European ancestry and the BioBank Japan (BBJ) project in East Asian ancestry. Information on cancer GWASs is listed in Supplementary Table 3, and brief descriptions of the FinnGen study, individual cancer consortia and BBJ project are provided below.

FinnGen, which was launched in Finland in the autumn of 2017, is a large public‒private partnership aiming to collect and analyse genome and health data from Finnish biobank participants (Kurki et al. 2022). The genetic summary data that can be downloaded publicly are the last R8, which was released on Dec 1, 2022, and include 342 499 participants, with 190 879 females and 151 620 males. The R8 data release contains various types of malignant neoplasms, and the participants diagnosed with any cancer were all excluded from the control group. Excluding cancer types that have small cases (fewer than 500) and that pediatric patients have been included (such as malignant neoplasm of brain and primary lymphoid and hematopoietic malignant neoplasms), 19 site-specific solid cancers and subtypes for breast, lung and colorectal cancer were included in our MR analysis (Supplementary Table 3).

Genetic summary data were publicly available for several cancers and their subtypes, which were granted by respective large cancer consortia or reported by additional population-based cohort and case‒control studies. One of the advantages of using these data is that they included a large sample size and number of cases. These consortia include the Breast Cancer Association Consortium (BCAC, 122 977 cases and 105 974 controls) (Michailidou et al. 2017), the International Lung Cancer Consortium (ILCCO, 11 348 cases and 15 861 controls) (Wang et al. 2014), the Ovarian Cancer Association Consortium (OCAC, 25 509 cases and 40 941 controls) (Phelan et al. 2017), a genome-wide meta-analysis on endometrial cancer reported by the Endometrial Cancer Association Consortium (ECAC), the Epidemiology of Endometrial Cancer Consortium (E2C2) and the UK Biobank (12 906 cases and 108 979 controls) (O'Mara et al. 2018), and the Prostate Cancer Association Group to Investigate Cancer Associated Alterations in the Genome (PRACTICAL, 79 148 cases and 61 106 controls) (Schumacher et al. 2018). These genetic summary data for cancers and their respective consortia are listed and briefly described in Supplementary Table 3.

AGEN, Asian Genetic Epidemiology Network BMI-Consortium; BMI, body mass index; GIANT, Genetic Investigation of Anthropometric Traits; GWAS, genome-wide association study; IVW, inverse-variance weighted; LD, linkage disequilibrium; MR, Mendelian randomization; MR-PRESSO, Mendelian randomization-pleiotropy residual sum and outlier; SNP, single nucleotide polymorphism.

For the East Asian population, BBJ, one of the largest non-European biobanks, was used for outcome data extraction. From April 2003 to February 2008, the BBJ project collaboratively collected DNA, serum, and medical records with written informed consent from 12 Japanese medical institutions and comprised a large cohort of approximately 260,000 patients aged 20–89 years old who underwent follow-up for 47 target diseases (Ishigaki et al. 2020; Nagai et al. 2017; Hirata et al. 2017). The genetic summary data on 12 site-specific solid cancers in the BBJ project can be obtained from the IEU open GWAS project (Hemani et al. 2018).

Identifying genetic instruments for BMI To avoid bias by reverse causality, the samples used to derive SNP-BMI and SNP-cancer associations should not overlap (Lawlor 2016). Therefore, genetic instruments for BMI in the European population were identified using results from a large collaborative meta-analysis combining data from the GIANT consortium and the UKB, including 806 834 participants with European ancestry (Pulit et al. 2019). For sex-specific cancers, the genetic instruments were obtained from sex-stratified European-ancestry genetic data with up to 434 794 females and 374 756 males (Pulit et al. 2019). For identifying genetic instruments for BMI, the genome-wide significance level for SNP-BMI was set at smaller than 5×10 − 8 (to avoid weak instrument bias). Independent top-associated SNPs were then identified using the clumping algorithm of a linkage disequilibrium (LD) window of 10 Mb (r2 > 0.001). All SNPs were searched for their possible associations with the outcomes and specific confounders with a threshold of P < 5×10 − 8 using the Phenoscanner database (Staley et al. 2016). To estimate the instrument strength for each SNP in MR analysis, the proportion of phenotypic variance in BMI explained by SNPs and the F-statistic were calculated according to the approaches described by a previous study (Pierce et al. 2011). When the SNPs of interest were not available in the outcome data, proxies for BMI SNPs were selected using the following criteria: LD r2 with BMI SNP > 0.8 and minor allele frequency (MAF) > 0.3. Standard allele harmonization was performed to align alleles on the forward strand and to omit ambiguous SNPs with nonconcordant alleles and palindromic SNPs with a noninferrable allele frequency before any MR analysis.

Currently, for the East Asian population, the publicly available GWAS data for BMI can only be obtained from the BBJ project. As the outcome data used in our MR analysis were extracted from the same population, we searched PubMed for other studies reporting significant genetic loci for BMI in the East Asian population and found a meta-analysis of GWAS in East Asian-ancestry populations by the Asian Genetic Epidemiology Network (AGEN) BMI Consortium, although some participants from BBJ were still included (Wen et al. 2014). Because the primary genetic summary data were unavailable and a limited number of significantly associated SNPs were reported, we directly obtained SNPs that may be associated with BMI in the East Asian population from the original study by AGEN.

Two-sample MR analysis In the absence of individual-level data, we carried out causal inference analyses utilizing the two-sample MR framework, a powerful statistical approach that only requires GWAS summary-level data (Hemani et al. 2018). For each cancer (sub)type evaluated for the MR association with BMI, the inverse-variance weighted (IVW) method with random effects was first used to combine the SNP-BMI and SNP-cancer association across multiple BMI instruments. Then, sensitivity analyses were performed using weighted median, MR‒Egger, the intercept test of MR‒Egger regression and Mendelian randomization pleiotropy residual sum and outlier (MR-PRESSO) to ensure that our findings were robust against weak violation of MR assumptions. The weighted median can help triangulate causal estimates under conditions when a small proportion of invalid instruments are present (Hartwig et al. 2017). Horizontal pleiotropy can be identified and adjusted to provide MR estimates by the intercept test of MR‒Egger regression and MR‒Egger, respectively (Burgess and Thompson 2017; Bowden et al. 2016). Cochran's Q test on the IVW was utilized to test the heterogeneity among estimates of SNPs, while the MR-PRESSO method can detect and correct the effects from outliers of SNPs, providing an MR estimate that is robust to heterogeneity (Verbanck et al. 2018). The estimated MR association between BMI and cancer risk was expressed as the odds ratio (OR) and corresponding confidence interval (CI) per standard deviation (SD) increase in genetically predicted BMI.

All analyses were two-sided and performed using the Two-Sample MR package (version 0.5.6) in R software 4.2.2 (Hemani et al. 2018). As multiple testing was conducted in our MR analysis, the Benjamini‒Hochberg method was applied to control the false discovery rate (FDR). The association with an adjusted P value < 0.05 was deemed significant, while a nominal P value < 0.05 but an adjusted P value ≥ 0.05 was treated as a suggestive association.

Genetic instruments

For European ancestry, after being clumped in SNPs with genome-wide significance, 520, 306 and 254 SNPs remained as genetic instruments that together explained 5.05%, 4.38% and 4.39% of the phenotypic variance in BMI for the sex-combined, female, and male populations, respectively (Supplementary Table 4). Using the Phenoscanner database, we found that these SNPs are mainly associated with other human anthropometric traits, such as height, weight and waist circumference, which are intimately associated with BMI and are not treated as confounders. However, some SNPs were found to be significantly associated with specific cancer and cancer risk or protective factors, such as smoking, alcohol, physical activity, and menstrual and birth characteristics (Supplementary Table 4). Therefore, we searched PubMed for MR analysis or meta-analysis to confirm the associations between these factors and individual cancers (Supplementary Table 5), and SNPs that were associated with these factors were excluded from the final MR analyses. For the East Asian population, 37 BMI-associated SNPs that together explained 1.34% of the phenotypic variance were directly obtained from the original report by AGEN (Supplementary Table 4). Estimations of power for MR analyses and to detect true ORs at 1.1, 1.3 and 1.5 for various cancers based on total sample size and proportion of phenotypic variance explained by BMI instruments were estimated using an online tool and are listed in Supplementary Table 6 (Brion et al. 2013). In the FinnGen study, the power to detect a moderate effect size (OR of 1.3 per SD increase in BMI) was greater than 0.8 for 14 of 25 cancer (sub)types. Only one estimated power is less than 0.8 for 14 cancer (sub)types from large consortia, while only one estimated power is greater than 0.8 for 12 cancer types from the BBJ project.

BMI and cancer in the FinnGen study

In the meta-analysis that combined the data of GIANT and UK Biobank, a one SD increase in BMI roughly translates to a 4.8, 5.2 and 4.2 kg/m² increase in sex-combined, female, and male populations. The estimated ORs per SD increase in genetically predicted BMI on cancer risks and the evaluation of pleiotropy effect and heterogeneity are shown separately in Fig. 2 and Supplementary Table 7. In the 19 selected site-specific solid cancers, a suggestive positive association was found between genetically predicted BMI and malignant neoplasm of bronchus and lung, oesophagus, pancreas, bladder, and corpus uteri, while a suggestive inverse association was found for malignant neoplasm of breast. For the six cancer subtypes, a suggestive positive association was found for lung squamous cell carcinoma, while genetically predicted BMI was suggestively associated with decreased odds of both HER-positive and HER-negative breast cancer. After adjusting by the Benjamini‒Hochberg method, significant associations remained for malignant neoplasms of the pancreas (OR 1.457, 95% CI 1.102-1.927), breast (OR 0.846, 95% CI 0.755-0.948) and its HER-negative subtype (OR 0.759, 95% CI 0.642-0.897), bronchus and lung (OR 1.258, 95% CI 1.081-1.465) and its squamous cell carcinoma subtype (OR 1.502, 95% CI 1.128-2.000). Estimation of causal associations under weighted median and MR‒Egger models showed consistency with the majority of the original IVW estimates, and the intercept test of MR‒Egger regression did not detect any pleiotropy across the analyses except for malignant neoplasm of the breast and its HER-negative subtype, providing evidence against bias due to horizontal pleiotropy. In the 25 main analyses, 13 associations were revealed to have evidence of heterogeneity among effect sizes estimated by the BMI instruments by using the computed Cochran's Q statistics. However, the point estimates for these heterogeneous trait associations remained in the MR-PRESSO analysis after removal of the outliers.

BMI and cancer in large consortia

For the five cancer types obtained from large consortia, regardless of nominally or after Benjamini‒Hochberg adjustment, positive associations were found between genetically predicted BMI and lung cancer (OR 1.230, 95% CI 1.091-1.387), ovarian cancer (OR 1.169, 95% CI 1.064-1.284), and endometrial cancer (OR 1.797, 95% CI 1.598-2.021), while an inverse association was found for breast cancer (OR 0.880, 95% CI 0.815-0.950). There was no association between genetically predicted BMI and prostate cancer (OR 0.925, 95% CI 0.826-1.036). Considering their subtypes, the genetically predicted BMI was significantly associated with squamous cell carcinoma in lung cancer, both ER-positive and ER-negative subtypes in breast cancer, endometrioid and mucinous subtypes in ovarian cancer, and both endometrioid and nonendometrioid subtypes in endometrial cancer (Fig. 3 and Supplementary Table 8). Consistent with analyses in the FinnGen study, the heterogeneity was obvious for more than half of the 14 MR analyses. Similarly, the IVW point estimates for most associations were not meaningfully different from those estimated via various other MR models (weighted median, MR‒Egger, and MR-PRESSO).

BMI and cancer in the East Asian population

Nominally, for the 12 site-specific solid cancers in the BBJ project, a causal association was found only between genetically predicted BMI and endometrial cancer, with an OR per SD increase in BMI of 1.820 (95% CI 1.074-3.086). However, the significance was lost when the result was adjusted by the Benjamini‒Hochberg method, indicating that our IVW results did not support the causal association between BMI and the risk of any of the 12 cancer types in the East Asian population. Using weighted median, MR‒Egger, and MR-PRESSO methods, we obtained similar results to those of IVW, although obvious heterogeneity was detected for 6 out of 12 MR analyses (Fig. 4 and Supplementary Table 9).

Using genetic instruments for BMI from the currently available dataset with the largest sample size and with sex-specific data, our results strengthen confidence in previously established positive findings for lung, ovarian, endometrial and breast cancer and null findings for prostate cancer included in several large consortia in the European population (Supplementary Table 1). Applying the same genetic instruments for BMI to cancer types in the FinnGen study, estimation of the MR analysis using various methods and sensitivity analyses aids the triangulation of the positive or inverse association between genetically predicted BMI and lung, endometrial and breast cancer. In addition, we also found that genetically predicted BMI was associated with an increased risk of pancreatic cancer and possibly oesophageal and bladder cancer. The causal association was limited to specific cancer subtypes; for example, the genetically predicted BMI was significantly associated with squamous cell carcinoma but not adenocarcinoma in lung cancer and with endometrioid but not serous ovarian cancer. For the East Asian population, the genetic instruments from the AGEN consortium did not infer a causal association between BMI and cancers in the BBJ project, although a possible increased risk for endometrial cancer may exist.

Along with the global obesity pandemic in recent decades, the association between obesity (mainly presented by high BMI) and cancer has attracted the attention of many researchers and has been extensively explored by numerous conventional observational studies. Two groups have summarized this evidence, and a report from the International Agency for Research on Cancer (IARC) Handbook Working Group concluded that there is sufficient evidence that high BMI increases the risk of cancer of the colon, oesophagus (adenocarcinoma), kidney (renal-cell), breast (postmenopausal), corpus uteri, gastric cardia, liver, gallbladder, pancreas, ovary, and thyroid, as well as multiple myeloma and meningioma (Lauby-Secretan et al. 2016). Another umbrella review of 204 meta-analyses found that an increase in BMI was associated with a higher risk of developing oesophageal adenocarcinoma, colon and rectal cancer in men, biliary tract system and pancreatic cancer, endometrial cancer in premenopausal women, kidney cancer, and multiple myeloma (Kyrgiou et al. 2017). Obvious inconsistencies exist across these previous reports that may result from the inherent limitations of conventional observational studies and the inconsistent methods used to rate evidence. Therefore, using MR analysis, which is considered to have less susceptibility to potential confounding bias and reverse causality, we confirmed the associations between BMI and endometrial and pancreatic cancer, which were also supported by both previous reports with strong evidence. In our MR analysis and both previous reports, the ORs for endometrial cancer were the largest among all analysed cancer types, consistent with a report based on data from GBD 2019 that uterine cancer has the largest fraction of DALYs attributable to high BMI, indicating that obesity has a major role in the development of this disease (Zhi et al. 2022). The underlying molecular mechanism is the “unopposed oestrogen” hypothesis, which proposes that endometrial carcinogenesis is driven by excess endogenous or exogenous oestrogen levels that are unopposed by progesterone (Hazelwood et al. 2022). This may explain why the effect of BMI on ovarian cancer is mainly determined by its endometrioid subtype in our MR analysis.

Inconsistent with a previous conventional observational study, our and other published two-sample MR analyses found that BMI was positively associated with lung cancer and inversely associated with breast cancer and was mainly associated with the squamous cell subtype in lung cancer and the HER-negative subtype in breast cancer. For lung cancer, the inconsistency may be explained from two sides. On the one hand, the interpretation of the findings from conventional observational studies is complicated with reverse causality, as lung cancer and its established risk factor, smoking, are thought to reduce body weight; therefore, conventional observational studies cannot detect the effect of high BMI on lung cancer incidence, and some studies have even reported an inverse relationship (Carreras Torres et al. 2018; Smith et al. 2012; El Zein et al. 2013). On the other hand, although the instruments associated with smoking were excluded in our study, because there are complex relationships between BMI and smoking, two-sample MR analysis cannot completely remove bias caused by smoking phenotypes; therefore, when both factors were jointly modelled, the effect of BMI on lung squamous cell carcinoma was shown to be mediated by smoking (Zhou et al. 2021). Nonetheless, using SNPs only associated with BMI, our results indicated a total effect for BMI on lung cancer. For breast cancer, although both groups that summarize evidence from previous observational studies reported a strong or highly suggestive association between BMI and postmenopausal breast cancer, they also found a possible inverse association between BMI and premenopausal breast cancer, the latter of which can be explained by lower levels of progesterone and oestrogen due to the longer anovulatory cycles that they experience (Ooi et al. 2019). The genetic summary data in the FinnGen study and the BCAC consortium did not differentiate between pre- and postmenopausal breast cancer cases; therefore, our results may be explained by the assumption that the genetic portion of BMI may reflect early-life weight gain, while early-life BMI was inversely associated with both premenopausal and postmenopausal breast cancer (Guo et al. 2016). This assumption was supported by the findings that the protective effect of adult body size became null when adjusted by childhood body size (Richardson et al. 2020; Hao et al. 2023).

While the replication MR analysis with unified genetic instruments for BMI and more outcome data confirmed that BMI is causally associated with some cancer types, our MR analysis did not confirm the associations between BMI and other cancer types supported by conventional observational or previous MR analyses, such as oesophageal, colorectal, liver and biliary tract, kidney, gastric and thyroid cancer. However, our results cannot support the conclusion that there is no causal association between BMI and these cancer types due to several inherent limitations of our MR study. First, both conventional observational and MR studies found that the significant associations were determined by specific subtypes of these cancers, such as oesophageal adenocarcinoma and cancer of the gastric cardia. The null findings in our MR analysis may be the result of unavailable data of specific subtypes of these cancers. Second, while the implementation of more genetic instruments would usually translate to better trait prediction, it also has the potential to introduce more invalid instruments and heterogeneity. Therefore, other MR models that can provide valid estimates in the presence of invalid instruments revealed a possible association between BMI and colorectal, liver, and biliary tract system cancer. Third, the power of MR analysis is an important determinant for true OR detection. However, in our MR analysis, the power to detect moderate effect sizes (OR of 1.3 per SD increase in BMI) was weak for most cancer types in the FinnGen study and the BBJ project. Previous observational studies have found that the increase in the risk of developing most cancers for every 5 kg/m² (approximately equal to one SD in our MR analysis) increase in BMI is less than 30% (Kyrgiou et al. 2017); therefore, a significant association was mainly found for cancers on which the effect of BMI is obvious or of which the data are obtained from large consortia with very strong statistical powers. The best example is endometrial cancer, in whose tumorigenesis the involvement of obesity has the strongest evidence; therefore, in the BBJ project that the detection powers are very weak, a possible causal association was found only for endometrial cancer. Fourth, although sex-specific genetic instruments were available in our MR analysis, except for cancer of the breast and reproductive organs, we exploited the sex-combined effect of the SNPs on the other cancer types. However, both conventional observational and MR studies have found that there are sex discrepancies in the effect of BMI on some cancer types; for example, BMI only increases the risk of colorectal cancer in males (Bull et al. 2020).

Nonetheless, our MR analysis aiming to reassess the association between BMI and cancers has several advantages. First, the sample size of the discovery cohort used to identify BMI instruments was very large, and the selected genetic instruments combined explained 5.05%, 4.38% and 4.39% of the phenotypic variance for the sex-combined, female, and male populations, making us able to check the robustness of MR assumptions by various MR sensitivity analyses that depend on large numbers of SNPs. Second, we used the same set of genetic instruments to inform the association between BMI and various cancers, and such a unified framework allows for direct comparison across all MR findings. In addition, we included some cancer types and East Asian populations that have never been studied before and speculated that this may be due to the null findings using the currently available data. Third, sex-specific genetic instruments were used in our MR analysis to match with a cancer occurring in breast or reproductive organs, avoiding the impact of sex instrumental heterogeneity. Fourth, we included outcome data from various GWAS reported by nationwide biobanks and individual large consortia and found that the best data source for MR analysis is large consortia, which have the strongest power to detect the weak and moderate effect of BMI on most cancer types. However, such data sources are limited to a small number of cancer types and European populations, calling for joint efforts to provide and share large-scale GWAS data for other cancer types and in other populations.

In conclusion, our comprehensive MR estimates reinforce the established evidence that BMI causally increases the risk of lung, ovarian, endometrial, pancreatic cancer and possibly oesophageal and bladder cancer while decreasing the risk of breast cancer in the European population, with the causal association being limited to specific cancer subtypes. However, we cannot conclude that there is no causal association between BMI and some other cancer types and in other populations due to several inherent limitations of our MR study and call for joint efforts to improve research status in the field. Nonetheless, with the global pandemic of overweight and obesity and its established negative effects on various health conditions, including cancer, continued concerted efforts to reduce the prevalence of overweight and obesity are a major public health goal.

ACKNOWLEDGEMENTS

Summary-level data for BMI and cancer were obtained from the Genetic Investigation of Anthropometric Traits (GIANT), the UK biobank (UKB), Asian Genetic Epidemiology Network (AGEN) BMI-Consortium, FinnGen study, Biobank Japan (BBJ) and genetic consortia such as Breast Cancer Association Consortium (BCAC), International Lung Cancer Consortium (ILCCO), Ovarian Cancer Association Consortium (OCAC), Epidemiology of Endometrial Cancer Consortium (E2C2), Endometrial Cancer Association Consortium (ECAC) and the Prostate Cancer Association group to Investigate Cancer-Associated Alterations in the Genome (PRACTICAL). The authors thank all participants and investigators for sharing these data.

FUNDING

This work was supported by the Medical Research Project of The Third Hospital of Mianyang (202222).

COMPETING INTERESTS

None of the authors declared Financial and Non-Financial Relationships and Activities, and Conflicts of Interest regarding this manuscript.

AUTHOR CONTRIBUTIONS

J. Li and J. Lai conceived and designed the work that led to the submission. J. Lai and Y. Xu collected GWAS data and performed the main data analysis. J. Li and J. Lai curated the data and checked the statistical analysis. J. Li and J. Lai drafted the manuscript. J. Lai and Y. Xu provided necessary advice and revised the manuscript. All authors gave consent to the publication of this study.

DATA AVAILABILITY

Summary data for GWAS are publicly available for download and the detailed download website are listed in Supplementary Table 3. All data analysed in this study and statistical code can be obtained by reasonable request to the corresponding author.

Brion MJ, Shakhbazov K, Visscher PM (2013) Calculating statistical power in Mendelian randomization studies. Int J Epidemiol 42(5):1497-501. https://doi.org/10.1093/ije/dyt179
Bowden J, Davey Smith G, Haycock PC, Burgess S (2016) Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 40(4):304-14. https://doi.org/10.1002/gepi.21965
Burgess S, Thompson SG (2017) Interpreting findings from Mendelian randomization using the MR-Egger method. Eur J Epidemiol 32:377-89. https://doi.org/10.1007/s10654-017-0255-x
Blüher M (2019) Obesity: global epidemiology and pathogenesis. Nat Rev Endocrinol 15, 288–298. https://doi.org/10.1038/s41574-019-0176-8
Bull CJ, Bell JA, Murphy N, Sanderson E, Davey Smith G, Timpson NJ, Banbury BL, Albanes D, Berndt SI, Bézieau S, Bishop DT (2020) Adiposity, metabolites, and colorectal cancer risk: Mendelian randomization study. BMC Med 18, 396. https://doi.org/10.1186/s12916-020-01855-9
Carreras-Torres R, Johansson M, Haycock PC, Relton CL, Smith GD, Brennan P, Martin RM (2018) Role of obesity in smoking behaviour: Mendelian randomisation study in UK Biobank. BMJ 361. https://doi.org/10.1136/bmj.k1767
Davey Smith G, Hemani G (2014) Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet 23(R1):R89-98. https://doi.org/10.1093/hmg/ddu328
El‐Zein M, Parent ME, Nicolau B, Koushik A, Siemiatycki J, Rousseau MC (2013) Body mass index, lifetime smoking intensity and lung cancer risk. Int J Cancer 133(7):1721-31. https://doi.org/10.1002/ijc.28185
Fang Z, Song M, Lee DH, Giovannucci EL (2022) The role of Mendelian randomization studies in deciphering the effect of obesity on cancer. J Natl Cancer Inst 114(3):361-71. https://doi.org/10.1093/jnci/djab102
Guo Y, Warren Andersen S, Shu XO, Michailidou K, Bolla MK, Wang Q, Garcia-Closas M, Milne RL, Schmidt MK, Chang-Claude J, Dunning A (2016) Genetically predicted body mass index and breast cancer risk: Mendelian randomization analyses of data from 145,000 women of European descent. PLoS Med 13(8):e1002105. https://doi.org/10.1371/journal.pmed.1002105
Gao Y, Zhang J, Zhao H, Guan F, Zeng P (2021) Instrumental heterogeneity in sex-specific two-sample Mendelian randomization: empirical results from the relationship between anthropometric traits and breast/prostate cancer. Front Genet 12:651332. https://doi.org/10.3389/fgene.2021.651332
Hartwig FP, Davey Smith G, Bowden J (2017) Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption. Int J Epidemiol 46(6):1985-98. https://doi.org/10.1093/ije/dyx102
Hirata M, Kamatani Y, Nagai A, Kiyohara Y, Ninomiya T, Tamakoshi A, Yamagata Z, Kubo M, Muto K, Mushiroda T, Murakami Y (2017) Cross-sectional analysis of BioBank Japan clinical data: a large cohort of 200,000 patients with 47 common diseases. J Epidemiol 27(Supplement_III):S9-21. https://doi.org/10.1016/j.je.2016.12.003
Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D, Laurin C, Burgess S, Bowden J, Langdon R, Tan VY (2018) The MR-Base platform supports systematic causal inference across the human phenome. Elife 7:e34408. https://doi.org/10.7554/eLife.34408
Hazelwood E, Sanderson E, Tan VY, Ruth KS, Frayling TM, Dimou N, Gunter MJ, Dossus L, Newton C, Ryan N, Pournaras DJ (2022) Identifying molecular mediators of the relationship between body mass index and endometrial cancer risk: a Mendelian randomization analysis. BMC Med 20, 125. https://doi.org/10.1186/s12916-022-02322-3
Hao Y, Xiao J, Liang Y, Wu X, Zhang H, Xiao C, Zhang L, Burgess S, Wang N, Zhao X, Kraft P (2023) Reassessing the causal role of obesity in breast cancer susceptibility: a comprehensive multivariable Mendelian randomization investigating the distribution and timing of exposure. Int J Epidemiol 52(1):58-70. https://doi.org/10.1093/ije/dyac143
Ishigaki K, Akiyama M, Kanai M, Takahashi A, Kawakami E, Sugishita H, Sakaue S, Matoba N, Low SK, Okada Y, Terao C (2020) Large-scale genome-wide association study in a Japanese population identifies novel susceptibility loci across different diseases. Nat Genet 52, 669–679. https://doi.org/10.1038/s41588-020-0640-3
Kyrgiou M, Kalliala I, Markozannes G, Gunter MJ, Paraskevaidis E, Gabra H, Martin-Hirsch P, Tsilidis KK (2017) Adiposity and cancer at major anatomical sites: umbrella review of the literature. BMJ 356. https://doi.org/10.1136/bmj.j477
Kurki MI, Karjalainen J, Palta P, Sipilä TP, Kristiansson K, Donner K, Reeve MP, Laivuori H, Aavikko M, Kaunisto MA, Loukola A (2022) FinnGen: Unique genetic insights from combining isolated population and national health register data. MedRxiv. https://doi.org/10.1101/2022.03.03.22271360
Lawlor DA (2016) Commentary: Two-sample Mendelian randomization: opportunities and challenges. Int J Epidemiol 45(3):908-15. https://doi.org/10.1093/ije/dyw127
Lauby-Secretan B, Scoccianti C, Loomis D, Grosse Y, Bianchini F, Straif K (2016) Body fatness and cancer—viewpoint of the IARC Working Group. N Engl J Med 375:794-798. https://doi.org/10.1056/NEJMsr1606602
Lin X, Li H (2021) Obesity: epidemiology, pathophysiology, and therapeutics. Front Endocrinol 12, 706978. https://doi.org/10.3389/fendo.2021.706978
Michailidou K, Lindström S, Dennis J, Beesley J, Hui S, Kar S, Lemaçon A, Soucy P, Glubb D, Rostamianfar A, Bolla MK (2017) Association analysis identifies 65 new breast cancer risk loci. Nature 551, 92–94. https://doi.org/10.1038/nature24284
NCD Risk Factor Collaboration (2016) Trends in adult body-mass index in 200 countries from 1975 to 2014: a pooled analysis of 1698 population-based measurement studies with 19.2 million participants. Lancet 387(10026):1377-96. https://doi.org/10.1016/S0140-6736(16)30054-X
Nagai A, Hirata M, Kamatani Y, Muto K, Matsuda K, Kiyohara Y, Ninomiya T, Tamakoshi A, Yamagata Z, Mushiroda T, Murakami Y (2017) Overview of the BioBank Japan Project: study design and profile. J Epidemiol 27(Supplement_III):S2-8. https://doi.org/10.1016/j.je.2016.12.005
O'mara TA, Glubb DM, Amant F, Annibali D, Ashton K, Attia J, Auer PL, Beckmann MW, Black A, Bolla MK, Brauch H (2018) Identification of nine new susceptibility loci for endometrial cancer. Nat Commun 9, 3166. https://doi.org/10.1038/s41467-018-05427-7
Ooi BN, Loh H, Ho PJ, Milne RL, Giles G, Gao C, Kraft P, John EM, Swerdlow A, Brenner H, Wu AH (2019) The genetic interplay between body mass index, breast size and breast cancer risk: a Mendelian randomization analysis. Int J Epidemiol 48(3):781-94. https://doi.org/10.1093/ije/dyz124
Pierce BL, Ahsan H, VanderWeele TJ (2011) Power and instrument strength requirements for Mendelian randomization studies using multiple genetic variants. Int J Epidemiol 40(3):740-52. https://doi.org/10.1093/ije/dyq151
Phelan CM, Kuchenbaecker KB, Tyrer JP, Kar SP, Lawrenson K, Winham SJ, Dennis J, Pirie A, Riggan MJ, Chornokur G, Earp MA (2017) Identification of 12 new susceptibility loci for different histotypes of epithelial ovarian cancer. Nat Genet 49, 680–691. https://doi.org/10.1038/ng.3826
Pulit SL, Stoneman C, Morris AP, Wood AR, Glastonbury CA, Tyrrell J, Yengo L, Ferreira T, Marouli E, Ji Y, Yang J (2019) Meta-analysis of genome-wide association studies for body fat distribution in 694 649 individuals of European ancestry. Hum Mol Genet 28(1):166-74. https://doi.org/10.1093/hmg/ddy327
Renehan AG, Soerjomataram I (2016) Obesity as an Avoidable Cause of Cancer (Attributable Risks). In: Pischon, T., Nimptsch, K. (eds) Obesity and Cancer. Recent Results in Cancer Research, vol 208. Springer, Cham. https://doi.org/10.1007/978-3-319-42542-9_13
Richardson TG, Sanderson E, Elsworth B, Tilling K, Smith GD (2020) Use of genetic variation to separate the effects of early and later life adiposity on disease risk: mendelian randomisation study. BMJ 369. https://doi.org/10.1136/bmj.m1203
Smith L, Brinton LA, Spitz MR, Lam TK, Park Y, Hollenbeck AR, Freedman ND, Gierach GL (2012) Body mass index and risk of lung cancer among never, former, and current smokers. J Natl Cancer Inst 104(10):778-89. https://doi.org/10.1093/jnci/djs179
Staley JR, Blackshaw J, Kamat MA, Ellis S, Surendran P, Sun BB, Paul DS, Freitag D, Burgess S, Danesh J, Young R (2016) PhenoScanner: a database of human genotype–phenotype associations. Bioinformatics 32(20):3207-9. https://doi.org/10.1093/bioinformatics/btw373
Schumacher FR, Al Olama AA, Berndt SI, Benlloch S, Ahmed M, Saunders EJ, Dadaev T, Leongamornlert D, Anokian E, Cieza-Borrella C, Goh C (2018) Association analyses of more than 140,000 men identify 63 new prostate cancer susceptibility loci. Nat Genet 50, 928–936. https://doi.org/10.1038/s41588-018-0142-8
Skrivankova VW, Richmond RC, Woolf BA, Davies NM, Swanson SA, VanderWeele TJ, Timpson NJ, Higgins JP, Dimou N, Langenberg C, Loder EW (2021) Strengthening the reporting of observational studies in epidemiology using mendelian randomisation (STROBE-MR): explanation and elaboration. BMJ 375. https://doi.org/10.1136/bmj.n2233
The GBD 2015 Obesity Collaborators (2017) Health effects of overweight and obesity in 195 countries over 25 years. N Engl J Med 377:13-27. https://doi.org/10.1056/NEJMoa1614362
Verbanck M, Chen CY, Neale B, Do R (2018) Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 50, 693–698. https://doi.org/10.1038/s41588-018-0099-7
Wang Y, McKay JD, Rafnar T, Wang Z, Timofeeva MN, Broderick P, Zong X, Laplana M, Wei Y, Han Y, Lloyd A (2014) Rare variants of large effect in BRCA2 and CHEK2 affect risk of lung cancer. Nat Genet 46, 736–741. https://doi.org/10.1038/ng.3002
Wen W, Zheng W, Okada Y, Takeuchi F, Tabara Y, Hwang JY, Dorajoo R, Li H, Tsai FJ, Yang X, He J (2014) Meta-analysis of genome-wide association studies in East Asian-ancestry populations identifies four new loci for body mass index. Hum Mol Genet 23(20):5492-504. https://doi.org/10.1093/hmg/ddu248
Zhou W, Liu G, Hung RJ, Haycock PC, Aldrich MC, Andrew AS, Arnold SM, Bickeböller H, Bojesen SE, Brennan P, Brunnström H (2021) Causal relationships between body mass index, smoking and lung cancer: univariable and multivariable Mendelian randomization. Int J Cancer 148(5):1077-86. https://doi.org/10.1002/ijc.33292
Zhi X, Kuang XH, Liu K, Li J (2022) The global burden and temporal trend of cancer attributable to high body mass index: Estimates from the Global Burden of Disease Study 2019. Front Nutr 9:918330. https://doi.org/10.3389/fnut.2022.918330

No competing interests reported.

SupplementaryInformation.xlsx

Download PDF

Version 1

posted

You are reading this latest preprint version

Reassessment of casual association between body mass index and site-specific solid cancers: A comprehensive Mendelian randomization study

Status:

Version 1

Abstract

Purpose

Methods

Results

Conclusion

Figures

INTRODUCTION

METHODS

RESULTS

DISCUSSION

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1