Figure 2.

Estimated % of the association between replacement of red and processed meat with other food sources of protein and risk of type 2 diabetes that is explained by serum ferritin (n total = 24,611, n cases of type 2 diabetes = 10,769). CIs were derived from a bootstrap procedure using 1,000 replicates. HRs on which these estimates are based were estimated from Prentice-weighted Cox regression models with adjustment for age (underlying timescale), sex, center, country, education, physical activity, smoking status, total energy intake, alcohol consumption, fruit, vegetables, sweets, soft drinks, coffee, tea, other dairy products, and BMI. Serving sizes were 70 g/day for yogurt, 50 g/day for red and processed meat, 30 g/day for cheese and cereals, and 10 g/day for nuts. Estimated % of association explained was calculated as follows: [(HR_without − HR_with) / (HR_without − 1)] ∗ 100.