Serum 25-hydroxyvitamin D below 25 ng/mL is a risk factor for long bone fracture comparable to bone mineral density in Japanese postmenopausal women

There is emergent evidence for divergent associations between 25(OH)D levels and fractures by race and ethnicity, but data on Asian populations are sparse. We investigated this association in a primary care cohort of 1470 postmenopausal Japanese women followed for a mean period of 7.2 years and explored a potential threshold of 25(OH)D. Endpoints were incident vertebral, proximal femur, and long bone fractures. Rate ratios were estimated using multivariate Poisson regression adjusted for lumbar or femur bone mineral density (BMD) less than −2.5 SD of the young adult mean (YAM), age, weight, presence of diabetes mellitus, parathyroid hormone, estimated glomerular filtration rate, prior fracture, back pain, present medications and past medical history. Mean age was 63.7 ± 10.7 years and osteoporosis patients were 41.3 %. The background data of the present participants were almost identical to the subjects participating in the National Health and Nutrition Survey of 2003. Overall, 49.6 % of the subjects had a 25(OH)D value <20 ng/mL and 27.8 % had a 25(OH)D value from 20 to 24 ng/mL. The propensity score for exposure to 25(OH)D < 25 ng/mL in the present and independent community dwelling populations, namely the Miyama and Taiji cohorts, were not significantly different, suggesting no evidence for selection bias. The generalized additive models showed clear decreasing trends in incidence rates of proximal femur and long bone fractures at higher levels of 25(OH)D, and the annual incidence rate of proximal femur fracture was around 0.0005 in women with 25(OH)D > 25 ng/mL, probably leading to the decreasing trend in long bone fracture. Multivariate-adjusted rate ratios of 25(OH)D < 25 ng/mL were 1.01 (95 % confidence interval [CI], 0.84–1.22, p = 0.88) for vertebral fracture, 2.71 (95 % CI 0.94–7.83, p = 0.07) for proximal femur fracture, and 2.20 (95 % CI 1.37–3.53, p < 0.01) for long bone fracture. The respective rate ratios of a BMD level lower than −2.5 SD of the YAM were 1.61 (95 % CI 1.33–1.94, p < 0.01), 1.52 (95 % CI 0.67–3.45, p = 0.32), and 1.54 (95 % CI 1.02–2.33, p = 0.04). In conclusion, 25(OH)D is a leading risk factor for long bone fracture comparable to BMD in Japanese postmenopausal women. The contribution of 25(OH)D to fracture risks is substantial even below 25 ng/mL and is possibly site-specific. We recommend measuring the serum 25(OH)D level in primary care settings.