Searching for the causal effects of body mass index in over 300 000 participants in UK Biobank, using Mendelian randomization

Abstract

Mendelian randomization (MR) has been used to estimate the causal effect of body mass index (BMI) on particular traits thought to be affected by BMI. However, BMI may also be a modifiable, causal risk factor for outcomes where there is no prior reason to suggest that a causal effect exists. We performed a MR phenome-wide association study (MR-pheWAS) to search for the causal effects of BMI in UK Biobank (n = 334 968), using the PHESANT open-source phenome scan tool. A subset of identified associations were followed up with a formal two-stage instrumental variable analysis in UK Biobank, to estimate the causal effect of BMI on these phenotypes. Of the 22 922 tests performed, our MR-pheWAS identified 587 associations below a stringent P value threshold corresponding to a 5% estimated false discovery rate. These included many previously identified causal effects, for instance, an adverse effect of higher BMI on risk of diabetes and hypertension. We also identified several novel effects, including protective effects of higher BMI on a set of psychosocial traits, identified initially in our preliminary MR-pheWAS in circa 115,000 UK Biobank participants and replicated in a different subset of circa 223,000 UK Biobank participants. Our comprehensive MR-pheWAS identified potential causal effects of BMI on a large and diverse set of phenotypes. This included both previously identified causal effects, and novel effects such as a protective effect of higher BMI on feelings of nervousness.

Fig 1. QQ plot of 22,922 MR-pheWAS results.

Green dashed line: Bonferroni corrected threshold (p = (0.05/22922 = 2.18x10^-6). Red dash-dotted line: FDR threshold (p = 0.05 × 587/22922 = 1.28x10⁻³). Blue dotted line: actual = expected. Purple points: results of tests performed in MR-pheWAS. Red stars: results with P values < 2.23x10^-308 (the smallest number allowed in R, specified in .Machine$double.xmin).

Fig 2. QQ plot of results in UK Biobank field category ‘psychosocial factors’ (ID 100059).

Green dashed line: Bonferroni corrected threshold calculated for psychosocial traits only (0.05/59 = 8.47 x 10⁻⁴). Blue dotted line: actual = expected. Purple points: results of tests performed in MR-pheWAS.

Fig 3. Results of follow-up analysis of the associations between nervousness / worrying traits (outcomes) and genetically predicted BMI (exposure).

BMI: body mass index; FID: field identifier; SD: standard deviation; IV: instrumental variable; TS: two stage; SIMEX: simulation extrapolation; MBE: mode-based estimate. Estimates are in terms of the log odds of outcome variable for a 1SD higher BMI. TS probit analyses are adjusted for age, sex and first 10 genetic principal components (see Table F in S1 Text for results of sensitivity analysis, adjusting for age, sex and first 40 genetic principal components). TS Probit estimates of the odds ratio of outcome for a SD higher BMI are calculated by taking the exponent of 1.6 times the probit estimate [34]. TS probit, full sample, 97 SNPs: Estimates using two-stage IV probit regression using our 97 SNP score as an instrument for BMI, and our full sample (N = 334,968). TS probit, full sample, 96 SNPs: Estimates using two-stage IV probit regression using our 96 SNP score (excluding FTO SNP) as an instrument for BMI, and our full sample (N = 334,968). TS probit, full sample, FTO: Estimates using two-stage IV probit regression using FTO SNP as an instrument for BMI, and our full sample (N = 334,968). TS probit, discovery sample: Estimates using two-stage IV probit regression using our 97 SNP score as an instrument for BMI and our discovery sample. TS probit, replication sample: Estimates using two-stage IV probit regression using our 97 SNP score as an instrument for BMI and our replication sample. Discovery results are estimated on the UK Biobank sample used in the PHESANT application note usage example [27]. Sample size for discovery samples: Field 1970: 111,746; Field 1980: 111,673; Field 1990: 111,161; Field 2010: 110,451. Sample size for replication samples: Field 1970: 216,311; Field 1980: 216,301; Field 1990: 215,435; Field 2010: 214,152. F-statistics for association of 97 SNP, 96 SNP (excluding FTO) and FTO genetic instruments with BMI are 6114.52, 5200.41 and 901.91, respectively. See Table E in S1 Text for full results.

Fig 4. Participant flow diagram.