Purpose:​ Patients with end-stage kidney disease have an increased fracture risk. Whether mild to moderate reductions in kidney function is associated with increased fracture risk is uncertain. Results from previous studies may be confounded by muscle mass because of the use of creatinine-based estimates of the glomerular filtration rate (eGFRcre). We tested the hypothesis that lower eGFR within the normal range of kidney function based on serum cystatin C (eGFRcys) or both cystatin C and creatinine (eGFRcrecys) predict fractures better than eGFR based on creatinine (eGFRcre).

Methods:​ In the Tromsø Study 1994–95, a cohort of 3016 women and 2836 men aged 50–84 years had eGFRcre, eGFRcys and eGFRcrecys estimated using the Chronic Kidney Disease Epidemiology Collaboration equations. Hazard ratios (HRs) (95% confidence intervals) for fracture were calculated in Cox's proportional hazards models and adjusted for age, height, body mass index, bone mineral density, diastolic blood pressure, smoking, physical activity, previous fracture, diabetes and cardiovascular disease.

Results:​ During a median of 14.6 years follow-up, 232, 135 and 394 women and 118, 35 and 65 men suffered incident hip, proximal humerus and wrist fractures. In women, lower eGFRcre did not predict fracture, but the risk for hip and proximal humerus fracture increased per standard deviation (SD) lower eGFRcys (HRs 1.36 (1.16–1.60) and 1.33 (1.08–1.63)) and per SD lower eGFRcrecys (HRs 1.25 (1.08–1.45) and 1.30 (1.07–1.57)). In men, none of the eGFR estimates were related to increased fracture risk. In contrast, eGFRcys and eGFRcrecys were inversely associated with hip fracture risk (HRs 0.85 (0.73–0.99) and 0.82 (0.68–0.98)).

Conclusions:​ In women, each SD lower eGFRcys and eGFRcrecys increased the risk of hip and proximal humerus fracture by 25–36%, whereas eGFRcre did not. In men, none of the estimates of eGFR were related to increased fracture risk, and each SD lower eGFRcys and eGFRcrecys decreased the risk of hip fracture by 15–18%. The findings particularly apply to a cohort of generally healthy individuals with a normal kidney function. In future studies, the association of measured GFR using the gold standard method of iohexol clearance with fractures risk should be examined for causal inference. More clinical research is needed before robust clinical inferences can be made.