Most of the observational studies investigating types of protein did not adjust their estimates for intakes of other protein types.
Covariate-adjusted estimates for intakes of fruit and vegetables, red and processed meat, and both nutrients were estimated using general linear models.
In a large French prospective study including 1,469 breast cancer cases, ER-positive and progesterone-positive breast cancer risk was inversely related to an estimate for intake of total dietary lignans (secoisolariresinol, matairesinol, pinoresinol, and lariciresinol) [ 37].
Estimates using the categorical measures were largely similar to those using continuous estimates, except for intake of soft drinks, which were attenuated and became non-significant after adjustment for adiposity (highest intake compared with non-consumers, 1.13 (0.94, 1.36; p linear trend, 0.306), possibly because the smaller sample size reduced the power to detect differences.
In an additional sensitivity analysis, we adjusted our estimates for milk intake by fermented milk and cheese intakes, and vice versa.
Accordingly, our estimates for milk intake were independent of consumption by fermented milk and cheese (see supplementary tables F and G).
Within North American studies, the estimates for wine intake were statistically homogeneous for ovarian carcinoma overall (OR, 0.99; 95% CI, 0.89-1.10; I 2 =0%).
This may be because our estimates for fluoride intake through milk, beverages, and food do not differentiate whether these sources come from fluoridated or nonfluoridated areas.
Then, the individual estimates were pooled with the inverse variance weighting method to calculate the overall estimates for folate intake (per 100 μg per day increment) and breast cancer risk.
