Author Affiliation: Department of Surgery, University of South Florida, Tampa (email@example.com).
To the Editor: Dr Peralta and colleagues1 categorized adults into 8 chronic kidney disease (CKD) groups based on 3 measured analytes that are altered in CKD: creatinine, cystatin C, and urine albumin-to-creatinine ratio (ACR). The 3 analytes, which are continuous variables, were dichotomized using published criteria for the cutoff points. This approach eliminated concern about the actual probability distribution function of the variables, but the choice of cut points raises a separate set of concerns. While it is appropriate to use published values for these cut points, the robustness of the model predicting mortality and end-stage renal disease would be enhanced by a sensitivity analysis determining how much the results would vary using different cut points. In the article referenced to substantiate the choice of cutoff values,2 hazard ratios demonstrated significant overlap, suggesting that the specific cutoff points are less than definitive.
Fabri PJ. Biomarkers for Detecting and Risk-Stratifying Chronic Kidney Disease. JAMA. 2011;306(6):611-612. doi:10.1001/jama.2011.1110