Causal relevance of obesity on the leading causes of death in women and men: A Mendelian randomization study

Background Obesity traits are causally implicated with risk of cardiometabolic diseases. It remains unclear whether there are similar causal effects of obesity traits on other non-communicable diseases. Also, it is largely unexplored whether there are any sex-specific differences in the causal effects of obesity traits on cardiometabolic diseases and other leading causes of death. We therefore tested associations of sex-specific genetic risk scores (GRSs) for body mass index (BMI), waist-hip-ratio (WHR), and WHR adjusted for BMI (WHRadjBMI) with leading causes of mortality, using a Mendelian randomization (MR) framework. Methods and Findings We constructed sex-specific GRSs for BMI, WHR, and WHRadjBMI, including 565, 324, and 338 genetic variants, respectively. These GRSs were then used as instrumental variables to assess associations between the obesity traits and leading causes of mortality using an MR design in up to 422,414 participants from the UK Biobank. We also investigated associations with potential mediators and risk factors, including smoking, glycemic and blood pressure traits. Sex-differences were subsequently assessed by Cochran’s Q-test (Phet). Up to 227,717 women and 194,697 men with mean (standard deviation) age 56.6 (7.9) and 57.0 (8.1) years, body mass index 27.0 (5.1) and 27.9 (4.2) kg/m2 and waist-hip-ratio 0.82 (0.07) and 0.94 (0.07), respectively, were included. Mendelian randomization analysis showed that obesity causes coronary artery disease, stroke (particularly ischemic), chronic obstructive pulmonary disease, lung cancer, type 2 and 1 diabetes mellitus, non-alcoholic fatty liver disease, chronic liver disease, and acute and chronic renal failure. A 1 standard deviation higher body mass index led to higher risk of type 2 diabetes in women (OR 3.81; 95% CI 3.42-4.25, P=8.9×10−130) than in men (OR 2.78; 95% CI 2.57-3.02, P=1.0×10−133, Phet=5.1×10−6). Waist-hip-ratio led to a higher risk of chronic obstructive pulmonary disease (Phet=5.5×10−6) and higher risk of chronic renal failure (Phet=1.3×10−4) in men than women. A limitation of MR studies is potential bias if the genetic variants are directly associated with confounders (pleiotropy), but sensitivity analyses such as MR-Egger supported the main findings. Our study was also limited to people of European descent and results may differ in people of other ancestries. Conclusions Obesity traits have an etiological role in the majority of the leading global causes of death. Sex differences exist in the effects of obesity traits on risk of type 2 diabetes, chronic obstructive pulmonary disease, and renal failure, which may have implications on public health.


Introduction 122
It is increasingly evident that obesity negatively impacts human health and the prevalence of obesity is 123 increasing world-wide (1). Obesity and central fat distribution, commonly measured by body mass 124 index (BMI; obesity usually defined as BMI >30 kg/m 2 ) and waist-hip-ratio (WHR), respectively, 125 have been linked to cardiometabolic diseases and death in observational studies (2-5). However, 126 conventional observational studies can be affected by bias, confounding, and reverse causation, which 127 might lead to erroneous findings. Mendelian randomization (MR) offers an approach to circumvent 128 these issues by using single nucleotide polymorphisms (SNPs) that reliably associate with an exposure 129 as an instrument to test the causal relationship between an exposure and outcome (6). Owing to the 130 nature of genotypes and therefore genetic associations, MR estimates should be less affected by 131 confounding and reverse causation (S1 Supporting Information) (6). Previous studies have found 132 causal relationships between for example higher BMI and WHR adjusted for BMI (WHRadjBMI) and 133 type 2 diabetes (T2D) and coronary artery disease (CAD), mostly using a limited number of 134 previously known obesity-associated SNPs (7-12). However, previous studies have not thoroughly 135 investigated causal sex-specific relationships, nor have they comprehensively investigated the role that 136 obesity traits play in the leading causes of death beyond these cardiometabolic diseases. 137 Obesity traits differ between women and men-for example, regional obesity prevalence rates often 138 vary between the sexes (13,14), women have higher SNP-based heritability for WHR (15), and >90% 139 of WHRadjBMI-associated SNPs that show evidence of sexual dimorphism have larger effect sizes in 140 women than men (15). Observational studies have indicated that waist-related traits might be more 141 strongly associated with cardiometabolic outcomes in women, although previous studies are 142 inconclusive (16)(17)(18)(19)(20). Only a few studies have investigated sex differences in the effect of genetic risk 143 for obesity-related traits on disease risk (7,10,12). These studies have mostly been restricted to waist-144 related traits and T2D and CAD, using a limited number of analyses and/or SNPs, but without finding 145 evidence of differences in disease risk between men and women (7,10,12). 146 A sex difference in the effect of obesity traits on major causes of death could signify that disease 147 burden arising from obesity may be differential in women and men, allowing prioritizing of public 148 health resources and potentially, sex-specific preventative strategies. We therefore investigated the 149 extent to which obesity traits causally impact the risk of the major global causes of death, and whether 150 relationships with disease are differential between women and men, exploiting recent advances in 151 discovery of obesity-associated SNPs (15). 152

Data sources and study participants 154
The UK Biobank is a prospective UK-based cohort study, with 488,377 genotyped individuals aged 155 We evaluated several approaches to construct sex-specific genetic risk scores (GRSs) for BMI, WHR, 169 and WHRadjBMI (S1 Supporting Information, Fig A-B in S1 Supporting Information). The approach 170 with the highest ranges of trait variance explained and F-statistics for the relevant obesity trait, and 171 with no demonstrable heterogeneity between men and women, was selected as the main model. In this 172 model, GRSs were constructed by including the primary ("index") genome-wide significant (P<5×10 -173 9 ) SNPs in the men, women, or combined-sexes analyses in the largest genome-wide association study 174 (GWAS) available with sex-specific European summary statistics, a meta-analysis of the Genetic 175 Investigation of ANthropometric Traits (GIANT) (22,23) and the UK Biobank (Fig 1, S1 Supporting 176 Information) (15). Primary SNPs were identified in the original GWAS (15) by proximal and joint 177 conditional analysis using GCTA in associated loci. Associated loci included all SNPs (associated 178 with the GWAS obesity trait P<0.05) ±5 Mb around a top SNP (P<5×10 -9 ) and that were in linkage 179 disequilibrium (LD; r 2 >0.05) with the top SNP; overlapping loci were merged (15). We then kept the 180 SNP with the lowest combined-sexes P-value within each 1 Mb sliding window to limit correlation 181 between SNPs discovered in different sex-strata in each obesity trait. We excluded non-biallelic SNPs 182 (N=2), SNPs that failed quality control (N=2), and one SNP per pair with long-distance linkage 183 disequilibrium (r 2 >0.05, N=2) (S1 Supporting Information). For the combined-sexes analyses, SNPs 184 were weighted using estimates from the combined-sexes European meta-analyzed GWASs. For the 185 men-and women-only analyses, SNPs were weighted by their sex-specific European estimates. All 186 SNPs were orientated so that the effect allele corresponded to a higher level of the investigated obesity 187

204
Outcomes 205 We investigated associations between three obesity traits (BMI, WHR, and WHRadjBMI) with all 206 non-communicable diseases on the World Health Organization's (WHO) list of leading mortality 207 causes world-wide and in high-income countries (24); CAD, stroke (including ischemic, hemorrhagic, 208 and of any cause), chronic obstructive pulmonary disease (COPD), dementia, lung cancer, T2D and 209 type 1 diabetes (T1D), colorectal cancer, renal failure (including acute, chronic and of any cause) and 210 breast cancer in women (Table A in S1 Supporting Information). In addition, we included infertility, 211 non-alcoholic fatty liver disease (NAFLD) and chronic liver disease (CLD) as they have previously 212 been linked to obesity and represent important and increasing burdens of disease (25-31). For T2D 213 and T1D, we drew case definitions from a validated algorithm for prevalent T2D and T1D (using 214 "probable" and "possible" cases) and those the algorithm denoted as "diabetes unlikely" were used as 215 controls (32). For CAD, we used the same case and control definitions as a large GWAS (33). Case 216 and control criteria for the other disease outcomes were defined using self-report data, data from an 217 interview with a trained nurse, and hospital health outcome codes in discussion between two licensed 218 medical practitioners (Table A in S1 Supporting Information). For CAD, acute renal failure, chronic 219 renal failure, stroke of any cause, ischemic stroke and hemorrhagic stroke, exclusions for certain codes 220 were also made in the control groups after defining the case groups. 221 To assess potential mediation, we also investigated associations between the obesity traits and the 222 potential cardiometabolic risk factors systolic blood pressure (SBP), diastolic blood pressure (DBP), 223 fasting glucose (FG), fasting insulin (FI), and smoking status. 224 Baseline measurements were used for all continuous traits, including BMI, WHR, WHRadjBMI, SBP 225 and DBP. For SBP and DBP, the mean of the up to two baseline measurements were used. Fifteen 226 mmHg to SBP and 10 mmHg to DBP were added if blood pressure lowering medications were used 227 (defined as self-reported use of such in data-fields 6153 and 6177), as in previous blood pressure 228 GWASs and as suggested in simulation studies (34,35). These anthropometric and blood pressure 229 measurements were then standardized by rank inverse normal transformation of the residuals after 230 regression of the trait on baseline age, age 2 , assessment centre, and, if applicable, sex. This was done 231 separately in the men and women only analyses, but jointly in the combined analyses, after any sample 232 quality exclusions (S1 Supporting Information). WHRadjBMI was generated in a similar manner, but 233 with adjustment for BMI as well, as in the original GWAS (15). 234 Sex-specific summary-level data for plasma FG (in mmol/L, untransformed, corrected to plasma levels 235 using a correction factor of 1.13 if measured in whole blood in the original GWAS) and serum FI (in 236 pmol/L, ln-transformed) were kindly provided by the Meta-Analyses of Glucose and Insulin-related 237 traits Consortium (MAGIC) investigators and can be downloaded from 238 https://www.magicinvestigators.org/downloads/ (36). SNPs in chromosome:position format were 239 converted to rsIDs using the file All_20150605.vcf.gz from the National Center for Biotechnology 240 Information (NCBI) (37) (available at 241 ftp://ftp.ncbi.nih.gov/snp/organisms/archive/human_9606_b144_GRCh37p13/VCF/). All SNPs were 242 then updated to dbSNP build 151 using the file RsMergeArch.bcp.gz, also from the NCBI (37) 243 (available at ftp://ftp.ncbi.nlm.nih.gov/snp/organisms/human_9606/database/organism_data/). 244 Smoking status was defined as self-report of being a current or previous smoker or having smoked or 245 currently smoking (most days or occasionally; any code 1 or 2 in any of the data fields 1239, 1249, 246 and 20116). 247

Statistical analyses 248
The GRSs were first assessed if they were robustly associated with their respective obesity traits by 249 computing trait variance explained and the F-statistics (S1 Supporting Information, Table B in S1 250

Supporting Information). 251
We explored the associations of sex-specific GRSs with outcomes in the UK Biobank (21). For 252 disease outcomes and smoking status, logistic regression was used while for continuous traits 253 (including evaluation of the GRSs in their respective obesity traits and the blood pressure traits) linear 254 regression was used. Associations of sex-specific GRSs with outcome traits that surpassed our P-value 255 thresholds were taken forward for MR to more formally quantify the effect of the obesity trait on the 256

outcome. 257
Individual-level MR was performed using the Wald method, with the instrumental variable estimate 258 being the ratio between the outcome and risk factor regressed separately on each GRS (38). Standard 259 errors were adjusted to take the uncertainty in both regressions into account by using the first two 260 terms of the delta method (39-41). MR regressions of the risk factors on the GRSs was performed in 261 controls only for the binary outcomes. 262 Adjustments were made for baseline age, age 2 , array type, assessment centre, 10 principal 263 components, and sex if applicable, for all traits when in clinical units, and array and 10 principal 264 components if rank inverse normal transformed (where adjustment for age, age 2 , assessment centre, 265 and if applicable sex had already been performed in the rank inverse normal transformation of the 266 residuals).

Sensitivity analyses 284
We performed several sensitivity analyses to ascertain robustness; we performed (a) analyses adjusting 285 for smoking status and (b) analyses restricted to those of genetically confirmed British ancestry only 286 (S1 Supporting Information). We also (c) evaluated the robustness of the MR findings by comparing 287 different weighting strategies, including use of unweighted and externally weighted (using weights 288 from the GIANT 2015 studies (22,23)) GRSs, and (d) investigated for pleiotropy and performed more 289 pleiotropy-robust sensitivity analyses (44,45) (S1 Supporting Information). We also (e) performed 290 logistic regressions using the same number of cases and controls in men and women for the disease 291 outcomes and (f) conducted analyses using stricter T2D and T1D case definitions (S1 Supporting 292

Evaluation of genetic risk scores 302
The GRSs included 565 SNPs for BMI, 324 for WHR and 338 for WHRadjBMI. Trait variance 303 explained varied between 2.5-7.1% and the F-statistic between 4,941-26,311, depending on trait and 304 sex-stratum (Table B in S1 Supporting Information). After having assessed the associations between 305 GRSs and risk factors and disease outcomes using regression analyses, associations that surpassed 306 correction for multiple testing were taken forward for MR (Table C- WHRadjBMI Phet=0.007) (Table D in S1 Supporting Information). We therefore ran the individual-311 level MRs adjusting for smoking status to assess potential mediation. 312

Mendelian randomization of obesity with disease outcomes: all individuals 313
Obesity traits were causally implicated with diseases that represent the major causes of death (Fig 2  314 and 3). All measures of obesity were strongly causally related to risk of CAD (odds ratio (OR) ranging 315 from 1.39 for WHRadjBMI to 1.73 for WHR in the combined analyses per 1-SD higher obesity trait). 316 For stroke, both BMI and WHR conferred higher risk (ORs 1.41 and 1.33, respectively). Strong effects 317 were seen for all obesity traits with T2D (OR range 2.13 to 3.64) and BMI also associated with risk of 318 In addition to these endpoints, strong effects were seen for risk of NAFLD (OR range 1.61-2.85) and 324 CLD (ORs 1.62 for BMI and 1.83 for WHR). 325

327
The obesity trait-disease combinations brought forward for Mendelian randomization, with estimates given in 328 odds ratio (95% CI) per 1-SD higher obesity trait. Filled diamonds indicate that the P-value for the obesity trait 329 to disease endpoint surpasses our threshold for multiple testing; empty diamonds indicate that the P-value does 330 not surpass this threshold (Bonferroni-adjusted P-value-threshold set at <0.001 (=0.05/51) for 51 obesity trait-331 disease outcome combinations in the study). * denotes that the P-value for heterogeneity (from Cochran

Supporting Information). 406
Mendelian randomization of obesity with disease outcomes: sex-stratified analyses 407 Five obesity trait-disease associations differed between women and men (Fig 2). The risk of T2D from 408 1-SD higher BMI was higher in women (OR 3.81; 95% CI 3.42-4.25, P=8.9×10 -130 ) than men (OR 409 2.78; 95% CI 2.57-3.02, P=1.0×10 -133 ), with strong evidence for sexual heterogeneity (Phet=5.1×10 -6 , 410 Phet-threshold set at <0.001 (=0.05/48) for 48 male-female disease estimate comparisons, since breast 411 cancer was investigated in women only). This sexual heterogeneity could also be observed in 412 sensitivity analyses where the number of cases in women and men was similar (Phet=4.4×10 -5 ) ( Table  413 H in S1 Supporting Information). 414 WHR increased risk of COPD to a greater extent in men (OR 1.87; 95% CI 1.61-2.17, P=1.2×10 -16 ) 415 than in women (OR 1.22; 95% CI 1.10-1.36, P=2.0×10 -4 , Phet=5.5×10 -6 ), per 1-SD higher WHR. While 416 the association of WHR with smoking was greater in men than in women (Table I in S1 Supporting 417 Information) and estimates of WHR with COPD for both men and women attenuated after adjustment 418 for smoking status, the association of WHR and COPD remained higher in men after adjusting for 419 smoking (Phet=1.2×10 -4 ; Table F in S1 Supporting Information). 420 There was also evidence of WHR leading to a higher risk on renal failure in men than in women. Men 421 had a higher risk of chronic renal failure per 1-SD higher WHR, with the risk in men being OR 2.32 422 (95% CI 1.81-2.98, P=4.4×10 -11 ) and in women OR 1.25 (95% CI 1.03-1.52, P=0.02, Phet=1.3×10 -4 ), 423 with similar sex differences seen for WHRadjBMI. Men also had a higher risk of acute renal failure 424 (men: OR 1.88; 95% CI 1.49-2.36, P=8.2×10 -8 ; women: OR 1.23; 95% CI 1.00-1.53, P=0.05, per 1-SD 425 higher WHR, Phet=0.009), although the Phet-value did not pass our Phet-threshold. 426 Sensitivity analyses using different GRS weighting strategies strongly supported sex-differences in the 427 effect of BMI on T2D and WHR on chronic renal failure and COPD, but only weakly supported a sex-428 difference in the effect of WHR on renal failure of any cause (S1 Supporting Information, Fig D,E in 429 S1 Supporting Information). 430

Potential mechanisms 431
To identify potential mediators, we assessed the relationship of obesity traits with blood pressure 432 (SBP, DBP), glycemic traits (FG, FI), and smoking status (Tables I-M in S1 Supporting Information). 433 All obesity traits causally impacted risk on SBP, DBP, FG and FI. The increase in DBP arising from 434 elevated BMI was greater in women than men (Phet=3.5×10 -5 , Phet-threshold set at <0.003 (=0.05/15) 435 for 15 obesity trait-risk factor combinations). BMI and WHR both associated with higher risk of being 436 a smoker, with the magnitudes of effect being larger in men than women (BMI Phet=0.002; WHR 437 Phet=3.7×10 -14 ). WHRadjBMI was only associated with smoking status in men. 438

439
Our study demonstrates that obesity is causally implicated in the etiology of two thirds of the leading 440 causes of death from non-communicable diseases (globally and in high-income countries) (24). 441 Furthermore, we identify that for some diseases, obesity conveys altered magnitudes of risk in men 442 and women. Such sexual dimorphism could be observed in the effects of BMI on T2D and waist-443 related traits on COPD and renal failure. These findings have potential implications for public health 444

policy. 445
Obesity traits were causally related to higher risk of T2D, in alignment with previous studies (7-446 12,20,61). We could not detect a sex difference in risk of T2D from higher WHR or WHRadjBMI. 447 Even though some observational studies have suggested that WHR may be a stronger predictor of T2D 448 risk in women than in men (19,20), studies investigating the effect on T2D risk from genetic 449 predisposition to higher WHRadjBMI have not found evidence of sexual heterogeneity (7,10,12). In 450 contrast, we found that BMI conferred a higher T2D risk in women than in men. Whereas men tend to 451 be diagnosed with T2D at lower BMI than women (62), there may be a stronger association between 452 increase of BMI and T2D risk in women than in men (16,19,61,63-66). Whether this reflects a 453 stronger causal effect of BMI on T2D risk in women has hitherto been unknown. We found no 454 evidence for sexual heterogeneity of the causal effect of BMI on potential glycemic trait risk 455 mediators (FG and FI). There have been indications of higher BMI being observationally associated 456 with lower insulin sensitivity in men than in women, but this observed sex-difference may not reflect a 457 causal pathway or we are not capturing it by our glycemic measurements (67-69). We also found 458 evidence of BMI causally increasing risk of T1D. Previous observational (70) and MR (71) studies 459 have implicated childhood BMI in risk of T1D. As SNPs associated with adult BMI have also been 460 found to affect childhood BMI (71,72), our results may well reflect the consequences of childhood 461 BMI on T1D rather than adult BMI. The results were robust to use of a stricter T1D case definition, 462 minimizing risk of erroneous finding due to misclassification of diabetes type. 463 Higher BMI, WHR and WHRadjBMI increased risk of CAD in both sexes, as shown previously (4,7-464 9,11,12,16,18). Our obesity trait-CAD analyses did not show evidence for sexual heterogeneity. 465 Observational studies have indicated that waist-related traits may be more strongly associated with 466 cardiovascular disease in women and men, but have not been conclusive (16,18,73). However, a recent 467 study (12) investigated the effect of higher WHRadjBMI, lower gluteofemoral fat distribution, and 468 higher abdominal fat distribution, proxied by genetic variants, on CAD and T2D risk and found no 469 evidence that relationships differed between men and women, similar to our findings. BMI and WHR 470 have previously been observationally associated with risk of stroke (74-76) and a previous MR study 471 found a causal effect of BMI on ischemic stroke (77). However, some studies have found WHR to be 472 an epidemiological risk factor for stroke in men only (74,75). Our results confirm BMI as a causal risk 473 factor for overall stroke in both men and women. In women, the effects of WHR were directionally 474 consistent with harm, but the estimates were imprecise, probably reflecting insufficient power in the 475 sex-stratified analysis. 476 Our results also indicate that higher BMI and WHR increase risk of COPD and higher BMI the risk of 477 lung cancer; a likely common mechanism is through smoking. BMI has previously been implicated in 478 COPD, but is not an established epidemiological nor causative risk factor (8,78-80). Obesity may 479 directly contribute to COPD as its diagnosis is partly based on spirometry values, and obesity is 480 associated with lower lung function (80,81). Higher BMI also increased risk of lung cancer in our 481 study, similar to a previous MR study (82). Observational studies tend to identify associations between 482 smoking and lower body weight, but whereas smoking lowers body weight, higher BMI is associated 483 with increased smoking (82-85). We found associations between particularly BMI and WHR with 484 smoking propensity. To assess mediation, we therefore conducted analyses adjusting for smoking 485 status. This attenuated the associations between the obesity markers and risk of COPD and lung 486 cancer, suggesting that smoking status may be on the causal pathway between obesity, COPD and 487 lung cancer. This diminution does not discredit the validity of the MR analyses unadjusted for 488 smoking provided that the obesity instruments only affect smoking propensity through altered obesity 489 (86). Rather, they suggest that higher BMI impacts on disease beyond the immediate physiological 490 effects of obesity: by altering human behavior (i.e. increased smoking, likely motivated as a weight 491 loss strategy (87,88)) and this increased propensity to smoking has additional, far-reaching, deleterious 492 effects on human health, as evidenced by the higher risks of serious lung disease. Higher WHR was 493 associated with higher effects on both COPD and being a smoker in men than in women. Whereas the 494 sex difference in the effect of WHR on COPD persisted after adjustment for smoking status, we 495 cannot rule out that WHR has a higher effect on COPD in men than women through its effect on 496 smoking propensity, but that our smoking phenotype does not fully capture the life-long effects of 497 smoking in men and women. 498 Our results also provide further evidence for a role of obesity traits in both acute and chronic renal 499 failure using an MR design -previous MR studies assessing these relationships have not been 500 conclusive (7,8,89-91). Obesity may affect chronic renal disease through a number of mechanisms, 501 including structural changes in the kidney and through higher risks of mediating diseases, such as T2D 502 and renal cell carcinoma (91-95). We found central fat distribution (as measured by WHR and 503 WHRadjBMI) to have higher effects on chronic renal failure in men than in women, with evidence of 504 sexual heterogeneity. The reason for this sex difference is unclear -a recent MR study found both 505 BMI and WHR to increase risk of renal cell carcinoma but with no difference in risk between men and 506 women (95). 507 Obesity traits associated with increased risk of NAFLD and CLD (important and emerging causes of 508 chronic disease and mortality (27-30)), with the effect on CLD possibly mediated by NAFLD, since 509 CLD may be caused by NAFLD (28). A previous MR study found BMI to increase hepatic 510 triglyceride content (96). Our study confirms a role of both general obesity and central fat distribution 511 in NAFLD and CLD using an MR design. This strengthens evidence of a causal effect and emphasizes 512 the risk of increased CLD burden if the obesity prevalence continues to increase (1,27-30). 513

Strengths and limitations 514
Genetic instruments should only affect the outcome through the risk factor of interest and not through 515 any confounders (97,98). We performed sensitivity analyses (MR-Egger, weighted-median based 516 methods) more robust to such bias, which supported the main findings (44,45). 517 If instruments are weakly associated with their respective traits, it can introduce bias in MR studies 518 (99). We therefore only used instruments strongly associated with their respective risk factor, and 519 performed sensitivity analyses using a variety of SNP-selection and weighting approaches, including 520 unweighted and externally weighted scores, which also supported the main results (41,99,100). 521 Recent studies have also indicated that there may be slight population stratification in both GIANT 522 and UKBB, although such bias is likely to be minor (101,102). Our study was restricted to individuals 523 of Europeans ancestry; limiting our analyses to those of British ancestry only yielded near-identical 524 results. Associations between the obesity traits and outcomes may differ in other ancestries. 525 Finally, it is possible that our genetic instrument for WHRadjBMI might show features of collider bias 526 whereby SNPs included in the GRS associate with both higher WHR and lower BMI leading to 527 potentially spurious findings (103). We note that a recent GWAS (15) evaluated the potential for 528 collider bias in the WHRadjBMI GWAS and found limited evidence for such, although the GRS was 529 associated with higher WHR and lower BMI. The directional consistency of associations between 530 WHR and WHRadjBMI and disease endpoints in our analysis suggests that collider bias is unlikely to 531 represent a major source of error in this study. 532

Conclusion 533
Global prevalence of obesity is increasing (1). Our results implicate major obesity traits (BMI, WHR, 534 and WHRadjBMI) in the etiology of the leading causes of death globally, including CAD, stroke, type 535 2 and 1 diabetes, COPD, lung cancer and renal failure, as well as NAFLD and CLD. The risk increase 536 from obesity traits differs between men and women for T2D, renal failure and COPD. This 537 emphasizes the importance of improved preventative measures and treatment of obesity-related 538 disorders and implies that women and men may experience different disease sequelae from obesity, 539 with potential implications for provision of health services and health policy. Widenlife; and NIH (5P50HD028138-27). TF, AM, RM report no conflicts of interest. 553 We thank the UK Biobank (http://www.ukbiobank.ac.uk/; application 11867). 554 Supporting Information 855 S1 Supporting Information. 856