Diet-wide association study of foods and nutrients with hip fracture risk: a prospective cohort study of 27,318 incident cases among 541,887 postmenopausal women

BackgroundPrevious research on diet and hip fracture risk focused on selected foods and nutrients. ObjectiveTo conduct a diet-wide association study of hip fracture risk. MethodsThe study population comprised 541,887 postmenopausal women in the Million Women Study. Dietary information was assessed using a validated food frequency questionnaire in 2000-2004, and calibrated using a 24-hour dietary recall from a subset 10 years later. Cox regression was used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for associations between 99 dietary factors and risk of incident hip fracture, ascertained through linkage to hospital admission data. ResultsAfter an average of 19.7 years of follow-up, 27,318 incident hip fractures occurred. In multivariable-adjusted models, 60 dietary factors were associated with hip fracture risk after correction for multiple testing (False Discovery Rate p-value <0.05). The five foods most significantly associated with hip fracture risk (based on p-value) were chicken (HR and 95% CI = 0.84, 0.80-0.87 per 20 g/day), vegetables (0.88, 0.85-0.91 per 100 g/day), pasta (0.87, 0.83-0.91 per 20 g/day), chips (1.14, 1.10-1.18 per 25 g/day), and fizzy drinks (1.18, 1.12-1.24 per 50 g/day). The five nutrients with the strongest associations were protein (0.79, 0.75-0.84 per 15 g/day), zinc (0.81, 0.77-0.86 per 2 mg/day), carotene (0.91, 0.88-0.94 per 1000 {micro}g/day), fiber (0.89, 0.85-0.92 per 5 g/day), and niacin (0.82, 0.77-0.87 per 10 mg/day). The results were consistent across subgroups of body mass index, smoking, alcohol consumption, health status, physical activity, menopausal hormone use, deprivation status, and in analyses excluding the first five years of follow-up. ConclusionsMany "healthy" foods were associated with lower risk of hip fracture, and "unhealthy" foods with higher risk. Further research is needed to assess whether these associations reflect causal relationships, are indicators of a healthier overall diet, or represent confounding from other lifestyle factors.