Comparison of four equations for estimation of glomerular filtration rate in predicting cardiovascular events and subclinical vascular disease in patients with type-2 diabetes
Résumé
Aims: Chronic kidney disease (CKD), defined by a low glomerular filtration rate (GFR), is a predictor of cardiovascular disease in patients with type-2 diabetes (T2D). We aimed to compare four GFR equations in predicting future cardiovascular events in T2D and the presence of subclinical vascular disease.
Methods: Four equations were used to estimate GFR in asymptomatic T2D patients consulting our centre for cardiovascular assessment. Follow-up was performed to collect cardiovascular events. Cox proportional hazard ratio (HR) was used to build and compare prediction models, and the incremental value of the addition of GFR with any of the 4 formulas was evaluated. The ability to triage patients with and without CVD events according to GFR were assessed by comparing the receiver operator characteristics (ROC) curves with the 4 models.
Results: Among 829 asymptomatic T2D patients, the CKD prevalence was 20.2% for Modification of Diet in Renal Disease (MDRD), 17.3% for Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI), 20.7% for Lund-Malmö Revised (LMR) and 21.4% for Full Age Spectrum (FAS). All the estimated GFRs were well correlated from one formula to another, with stronger agreement to define CKD (GFR <60 mL/min/1.73 m2) between MDRD and CKD-EPI, and between LMR and FAS. The 5-year incidence of cardiovascular events was 8% (n = 63). After adjustment on covariables, CKD was significantly associated with cardiovascular events when defined by MDRD (HR = 2.04; 1.15-3.60) and CKD-EPI (HR = 1.90; 1.05-3.41) but missed statistical significance when using LMR (HR = 1.74; 0.97-3.14) or FAS (HR = 1.71; 0.94-3.14). Only the prediction models including MDRD and CKD-EPI provided a significant incremental information to the predictive model without GFR, but the area under the ROC curves were similar with the 4 models: 0.60 [0.54-0.68] for MDRD, 0.61 [0.49-0.65] for CKD-EPI and 0.62 [0.55-0.69] for LMR and FAS, without any significant difference among formulas.
Conclusion: In asymptomatic T2D patients, MDRD and CKD-EPI may be preferable when more specificity is desired (stronger association between GFR and CVD events), while LMR and FAS appear more sensitive by including a higher number of patients with GFR <60 mL/min/1.73 m2.