Biphasic fracture risk in diabetes: a population-based study.

Diabetes is associated with increased fracture rates but the effect size, time course and modifying factors are poorly understood. This study was undertaken to assess the effect of diabetes on fracture rates and possible interactions with age, duration of diabetes and comorbidity. A retrospective, population-based matched cohort study (1984-2004) was performed using the Population Health Information System (POPULIS) for the Province of Manitoba, Canada. The study cohort consisted of 82,094 diabetic adults and 236,682 non-diabetic matched controls. Diabetes was subclassified as long term, short term, and newly diagnosed. Number of ambulatory diagnostic groups (ADGs) was an index of comorbidity. Poisson regression was used to study counts of combined hip, wrist and spine (osteoporotic) fractures (5691 with diabetes and 16,457 without diabetes) and hip fractures (1901 with diabetes and 5224 without diabetes). Independent effects of longer duration of diabetes (p-for-trend<0.0001) and number of ADGs (p-for-trend<0.0001) were observed on fracture rates. Newly diagnosed diabetes showed a reduction in osteoporotic fractures (rate ratio [RR] 0.91 [95% CI, 0.86-0.95]) and hip fractures (RR 0.83 [0.75-0.92]). Long-term diabetes showed an increase in osteoporotic fractures (RR 1.15 [CI, 1.09-1.22]) and hip fractures (RR 1.40 [1.28-1.53]). We conclude that long-term diabetes is associated with increased fracture risk, whereas newly diagnosed diabetes shows a reduction in fractures. It is hypothesized that the opposing effects of overweight/obesity and diabetes-related complications contribute to the observed biphasic fracture risk, though causality cannot be proven from this observational study.