2. Dose‐response association between BW and cancer risk
Totally 20 studies were collected on the relationship between BW and breast cancer including 9 case-control studies 4-12and 11 cohort studies 13-23. We selected the non-linear model with three knots in the splines at the 35th, 55th, 75th percentiles, the slope of the dose‐response curve increased continuously with the BW gain (X2=5.3, Pnonlinearity<0.01). (Figure 1) The RRs (95% CIs) of breast cancer risk were 0.98 (0.96-1.00), 1.00 (0.96-1.03), 1.07 (1.02-1.12) and 1.11 (1.03-1.18) for 2.0, 3.0, 4.0, and 4.5 kg, respectively. Yet the simulation of the linear model and quadratic model were not statistically significant (Plinearity=0.44, Pquadratic=0.06). We chose 10 studies of testicular cancer 24-33 and the relationship between testicular cancer and BW satisfied the non-linear model (X2= 236.7, P< 0.001) and quadratic model (X2=235.3, P< 0.001). The fitting effect of the non-linear model was better, and the percentiles of three knots were 25th, 55th, 85th. Taking 3000g BW as the reference dose, the dose-response curve showed a U-shape. The slope decreased before 3000g and increased above 3000g. (Figure 2) The RRs (95% CIs) of testicular cancer risk were 1.12 (1.09-1.14), 1.05 (1.04-1.06), 1.00 (1.00-1.00) and 1.02 (0.96-1.10) for 1.0, 2.0, 3.0, and 4.0 kg, respectively. Yet the simulation of the linear model was not statistically significant. (Plinearity=0.15) Other types of cancer including ovarian cancer, colorectal cancer, endometrial cancer and prostate cancer were not dose relative to BW. Both linear model and no-linear model were not statistically significant.
We conducted the sensitivity analysis and Egger’s test (Pbreast=0.40; Pcolorectal=0.31; Pendometrial=0.68; Povarian=0.41; Pprostate=0.59; Ptesticular=0.75) of all included articles to assess the heterogeneity between articles, all of them pass the tests. The sensitivity analysis plots and funnel plots were showed in figure S8- figure S19.