Abstract
Estimation of GFR from serum concentrations of creatinine and cystatin C has been refined using cross-sectional data from large numbers of people. However, the ability of the improved estimating equations to identify changes in GFR within individuals over time has not been rigorously evaluated, particularly within the normal range of GFR. In cross-sectional and longitudinal analyses of 1441 participants in the Diabetes Control and Complications Trial (DCCT)/Epidemiology of Diabetes Interventions and Complications (EDIC) study with type 1 diabetes, we compared GFR estimated from creatinine (eGFR(Cr)), cystatin C (eGFR(Cys)), or both (eGFR(Cr+Cys)) with iothalamate GFR (iGFR), including changes in each over time. Mean (SD) iGFR was 122.7 (21.0) ml/min per 1.73 m(2). In cross-sectional analyses, eGFR(Cr+Cys) estimated iGFR with the highest correlation (r=0.48 versus 0.39-0.42), precision, and accuracy. In longitudinal analyses, change in eGFR(Cr+Cys) best estimated change in iGFR; however, differences between estimates were small, and no estimate accurately classified change in iGFR. Over a median 23 years of follow-up, mean rate of change in eGFR was similar across estimates of eGFR(Cr), eGFR(Cys), and eGFR(Cr+Cys) (-1.37, -1.11, and -1.29 ml/min per 1.73 m(2) per year, respectively). Associations of BP and hemoglobin A1c with change in eGFR were strongest for eGFR(Cys) and eGFR(Cr+Cys). Together, these results suggest that the addition of cystatin C to creatinine to estimate GFR may improve identification of the causes and consequences of GFR loss in type 1 diabetes, but may not meaningfully improve the tracking of GFR in clinical care.