Early life body size
Genetically predicted early life body size was estimated to increase risk of colorectal cancer (OR per category change: 1.12, 95% confidence interval [CI]: 0.98–1.27), colon cancer (OR: 1.16, 95% CI: 1.00–1.35), and distal colon cancer (OR: 1.25, 95% CI: 1.04–1.51), although for overall colorectal cancer, the confidence interval crosses the null, while proximal colon and rectal cancer were minimally influenced by early life body size (OR: 1.11, 95% CI: 0.93, 1.32 and OR: 1.14 95% CI: 0.93, 1.38, respectively) (Fig. 2, Additional file 1: Table S7). In the multivariable models, the direct effect estimates of early life body size were attenuated to null for colorectal (OR: 0.97, 95% CI: 0.77–1.22) and colon cancer (OR: 0.97, 95% CI: 0.76–1.25), while an estimate of similar magnitude with more imprecision was observed for distal colon cancer (OR: 1.27, 95% CI: 0.90–1.77) (Fig. 2). In multivariable models for proximal colon and rectal cancer, there was little evidence for a direct effect of genetically predicted early life body size (OR: 0.82, 95% CI: 0.61, 1.09 and OR: 1.05 95% CI: 0.76, 1.45, respectively).
For women, early life body size was estimated to increase colorectal cancer (OR: 1.20, 95% CI: 0.97–1.48) and colon cancer (OR 1.20, 95% CI: 0.95–1.51) (Fig. 3, Additional file 1: Table S7). These genetically predicted effects attenuated towards the null in the multivariable model (colorectal cancer, OR: 1.08, 95% CI: 0.81–1.45; colon cancer, OR: 0.99, 95% CI: 0.71–1.36) with similar patterns for the other subsite models. For men, little evidence for an effect of early life body size on colorectal cancer risk was observed in the univariable model (OR: 0.96, 95% CI: 0.73–1.26) with an inverse point estimate observed in the multivariable model (OR: 0.74, 95% CI: 0.49–1.12) (Fig. 3).
The effect estimates were similar across cancer subsite (P-heterogeneity ≥ 0.14) and by sex (P-heterogeneity ≥ 0.10) in both the univariable and multivariable analyses. For distal colon cancer opposing direct effect, estimates of early life body size with wide confidence intervals were observed for men (OR: 0.71, 95% CI: 0.41, 1.23) and women (OR: 1.32, 95% CI: 0.87, 2.03); P-heterogeneity = 0.08.
Adult life body size
In the sex-combined multivariable model, adult body size was estimated to directly increase the risk of colorectal cancer (OR per category change: 1.27, 95% CI: 1.03, 1.57), colon cancer (OR: 1.32, 95% CI: 1.05, 1.67), and proximal colon cancer (OR: 1.57, 95% CI: 1.21, 2.05), whereas estimates for distal colon (OR: 1.02, 95% CI: 0.76, 1.38) and rectal (OR: 1.13, 95% CI: 0.84, 1.52) cancers were of smaller magnitude (Fig. 2).
For women, adult body size was estimated to directly increase, albeit imprecisely, the risk of colon (OR: 1.20, 95% CI: 0.87, 1.65), proximal colon (OR: 1.63, 95% CI: 1.12, 2.39), and rectal cancer (OR: 1.22, 95% CI: 0.79, 1.92), whereas null estimates were observed for colorectal cancer (OR: 1.06, 95% CI: 0.79, 1.42) (Fig. 3). For men, adult body size was estimated to directly increase the risk of colorectal cancer (OR: 1.54, 95% CI: 1.06, 2.20), colon cancer (OR: 1.52, 95% CI: 1.01, 2.32), and distal colon cancer (OR: 1.75, 95% CI: 1.07, 2.89) with similar positive effect estimates found for proximal colon (OR: 1.34, 95% CI: 0.81, 2.23) and rectal cancer (OR: 1.34, 95% CI: 0.84, 2.14) (Fig. 3).
Similarly to the early life body size analysis, opposing multivariable analysis estimates were observed for the direct effect of adult body size on distal colon cancer in men and women (OR: 1.75, 95% CI: 1.07, 2.89 and OR: 0.83, 95% CI: 0.54, 1.25, respectively, P-heterogeneity = 0.02).
Sensitivity analyses
F-statistics for genetic instruments indicated sufficient strength to satisfy the “relevance” assumption—we found little evidence of weak instrument bias under the univariable framework (F statistics were > 10 for all SNPs included in the analysis). In multivariable MR Fearly life body size, Fadult body size statistics were all over 10 with the exception of the analysis in men where Fearly life body size was equal to 8, and Fadult body size was equal to 10 suggesting that weak instrument bias may influence the analyses for men (Additional file 1: Tables S1 and S8). We evaluated heterogeneity and pleiotropy to test the “exchangeability” and “exclusion restriction” assumptions (that the instrument does not share a common cause with the outcome and that any effect of the instrument on the outcome is exclusively through its potential effect on the exposure, respectively). There was evidence of heterogeneity in most of the analyses as denoted by the Q statistics (Additional file 1: Tables S7 and S8). Under the univariable framework, based on Egger’s intercept test, evidence of directional pleiotropy was found for the effect of early life body size on colorectal cancer (overall, women only) and proximal colon cancer (overall), with stronger positive effect estimates observed for the MR Egger regression models (overall colorectal cancer , OR 1.46, 95% CI: 1.09, 1.93; colorectal cancer women only, OR 1.84, 95% CI: 1.16, 2.92; and overall proximal colon cancer, OR 1.60, 95% CI: 1.09, 2.34) (Additional file 1: Table S7). Finally, the multivariable MR Egger’s intercept test identified some evidence of pleiotropy in the analyses of adult body size on colorectal, colon and proximal colon risk, but the related effect estimates replicated the positive direct effects of adult body size observed in the main analysis (Additional file 1: Table S9).