Authors' Reply: Integer Cystatin C Values: Impact on Discordance Group Assignment and Accuracy of GFR-Estimating Equations

Journal of the American Society of Nephrology, vol. 34, pp. 1915–1916

Abstract

We thank Groothof et al.1 for their letter and the opportunity to clarify certain aspects of our analysis.2 In our calculation of eGFR based on creatinine (eGFRcr), eGFR based on cystatin C (eGFRcys), and eGFRcr-cys, we followed current recommendations for reporting cystatin C (using two decimals for values in mg/L) and for serum creatinine (using two decimals for values in mg/dl and integers for values in micromole/L).3,4 Thus, we did not round cystatin C values to the nearest integer before calculations as asserted by Groothof et al. In our baseline table, we rounded cystatin C to the nearest integer; here, we present them using 2 decimals for the interested reader: 1.28 (0.97 to 1.80) mg/L in the overall population, 1.57 (1.23 to 2.12) mg/L for eGFRcys eGFRcr. We did not implement a stratified bootstrap but performed bootstrapping followed by stratification by discordance group. Reanalysis using a stratified bootstrap leads to virtually identical results: For the eGFRcys>eGFRcr discordance group, median biases (95% confidence interval [CI]) using a stratified bootstrap are −4.5 (−5.3 to −3.8) ml/min per 1.73 m2 for eGFRcr, 8.4 (7.3 to 10.0) for eGFRcys, and 1.8 (1.1 to 2.5) for eGFRcr-cys. For the eGFRcys eGFRcr discordance group (compared with 713 in our main analysis that used two decimals). When rounding cystatin C to one decimal before eGFR calculation, median biases (95% CI) in the eGFRcys>eGFRcr discordance group would be −4.4 (−5.3 to −3.8) ml/min per 1.73 m2 for eGFRcr, 9.5 (7.9 to 11.2) for eGFRcys, and 2.0 (1.5 to 2.8) for eGFRcr-cys. For the eGFRcys<eGFRcr group, median biases would be 15.0 (14.5 to 15.5), −8.7 (−9.0 to −8.3), and 0.7 (0.4 to 1.0), respectively. When rounding cystatin C to the nearest integer, median biases (95% CI) would be −1.0 (−1.3 to −0.5) for eGFRcr, 18.4 (17.4 to 19.3) for eGFRcys, and 10.1 (9.2 to 11.0) for eGFRcr-cys. For the eGFRcys

Authors 4

1. Edouard L. Fu corresponding Aachen Department of Clinical Epidemiology

   Brigham and Women's Hospital · Harvard University · Leiden University Medical Center · Karolinska Institutet

   Affiliation as printed

   Department of Clinical Epidemiology, Leiden University Medical Center, Leiden, The Netherlands

   Department of Medical Epidemiology and Biostatistics, Karolinska Institute, Stockholm, Sweden

   Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts

2. Andrew S. Levey

   Tufts Medical Center

   Affiliation as printed

   Division of Nephrology, Department of Internal Medicine, Tufts Medical Center, Boston, Massachusetts

3. Lesley A. Inker

   Tufts Medical Center

   Affiliation as printed

   Division of Nephrology, Department of Internal Medicine, Tufts Medical Center, Boston, Massachusetts

4. Juan Jesús Carrero

   Karolinska Institutet

   Affiliation as printed

   Department of Medical Epidemiology and Biostatistics, Karolinska Institute, Stockholm, Sweden