To comprehensively and exhaustively assess the relationship between serum ferritin levels and incidence of prediabetes in a prospective study.

This prospective cohort study (n=7380) with a mean follow-up of 3.07 years (range: 1–7, 95% CI: 3.03–3.12) was conducted in Tianjin, China. Blood fasting glucose, oral glucose tolerance test, serum ferritin levels and other potentially confounding factors were measured at baseline and at each year of follow-up. Adjusted Cox proportional hazards regression models were used to assess the gender-specific relationship between baseline and mean serum ferritin quintiles and prediabetes.

The incidence of prediabetes was 85 per 1000 person-years among men and 44 per 1000 person-years among women during follow-up (from 2007 to 2014). After adjusting for potential confounders, hazard ratios (95% CI) for prediabetes across baseline ferritin quintiles were: for men, 1.00, 1.13 (0.90–1.40), 1.20 (0.97–1.48), 1.41 (1.14–1.73) and 1.73 (1.41–2.11); and for women, 1.00, 1.01 (0.74–1.38), 0.68 (0.48–0.96), 0.84 (0.61–1.15) and 1.07 (0.80–1.45), respectively. Similar results were also observed for mean ferritin levels.

Both baseline and mean serum ferritin levels were significantly and linearly related to prediabetes in men, whereas U-shaped relationships were observed between baseline and mean serum ferritin and prediabetes in women. The relationship between prediabetes risk and mean serum ferritin levels may be more stable than one with baseline serum ferritin levels.