Validity of U-shaped risk of estimated glomerular filtration rate for all-cause mortality

Abstract

Puddu et al. reported that serum uric acid and estimated glomerular filtration rate (eGFR) had different predictive ability for cardiovascular incidence and mortality by long-term follow-up study [1]. In their study, the authors clarified that eGFR from the Chronic Kidney Disease Prognosis Consortium (CKDEPI) could only predict all-cause mortality by a U-shaped relation. Their study is partly in concordance with a report that glomerular hyperfiltration has a risk of prediabetes and prehypertension [2]. I have a concern on their reason for concluding Ushaped relation between events and eGFR_CKDEPI. The authors presented two tables on forced Coxmodel for predicting incident events in their study. They used 10 independent variables for the prediction, and their two tables fundamentally contained the same information. Among them, eGFR_CKDEPI or its square value was selected as covariate. Although the square value of eGFR_CKDEPI was selected as a significant predictor for all-causemortality, caution should be paid to this statistical procedure. As eGFR_CKDEPI shows normal distribution, there is no reason for making square value as one of the independent variable. In addition, hazard ratio (95% confidence interval) of the square value of eGFR_CKDEPI for all-cause mortality with delta value of 100 is 1.02 (1.00–1.04), which does not explain Ushaped relation. The authors set centered group around the mean value of eGFR_CKDEPI as control, and evidence of curvilinear shape with statistical significance was not presented. I recommend the authors presenting hazard ratios for all-cause and cardiovascular disease mortality. Although there is a lack of information on albuminuria in paper by Puddu et al., their study group previously reported that combination of increased albuminuria and lower GFR from calibrated four-variable modification of diet in renal disease showed an increase of cardiovascular risk [3]. The authors explained the discrepancy of results in “Discussion”, and I fundamental accept their speculation.AsOhsawaet al. [4] presented a significant increase in the risk ofmortality in subjectswith preserved or increased eGFR_CKDEPI and albuminuria, I recommend the authors additional analysis by using information on albuminuria. Anyway, the quantitative assessment of renal hyperfiltration and hypofiltration by eGFR_CKDEPI for mortality would lead to confirm the existence of curvilinear relation. I wish to express my appreciation to the members of Hygiene and Public Health, Nippon Medical School, for the preparation of this study. The author of this manuscript has certified that he complies with the Principles of Ethical Publishing in the International Journal of Cardiology.