Authors' Reply: Concerns about Hospitalization on Cystatin C– and Creatinine-Based eGFR

We thank Yang et al.1 for their interest in our recent JASN article “Differential Effect of Hospitalization on Cystatin C– and Creatinine-Based Estimated GFR.”2 We appreciate the opportunity to further discuss the analysis. First, Yang et al.1 ask whether days of hospitalization regardless of cause have an equivalent effect on GFR estimated from cystatin C (eGFRCys) minus GFR estimated from creatinine (eGFRCr) (eGFRdiff) (i.e., might days of infectious hospitalization have a different effect than days of cardiovascular hospitalization). The Chronic Renal Insufficiency Cohort (CRIC) unfortunately did not collect granular clinical data about what transpired during hospitalizations. Other work has shown similar effects on eGFRdiff of hospitalization with critical illness3 and acute decompensated heart failure4 in the same direction as the effects we found of all-cause hospitalization.2 In a prior publication, we found that AKI itself does not seem to be a driver of eGFRdiff.5 Yang et al.1 also bring up the possibility that increased inflammation associated with hospitalization could result in increased cystatin C production, thereby falsely decreasing eGFRCys. Although this is a reasonable hypothesis, prior work on changes in eGFRdiff during critical illness,3 which included directly measured GFRs, did not support this hypothesis. Rather, the preponderance of evidence is that the error in GFR estimation associated with hospitalization is greater for eGFRCr than for eGFRCys, compared with measured GFR (i.e., the error is not in eGFRCys). Unfortunately, serial C-reactive protein measurements are not available in CRIC to perform the additional analysis that Yang et al. propose.1 We did investigate whether measured GFRs in CRIC could be used here, but we were unable to do so since only four participants had measured GFRs at an annual study visit after 28 or more days of hospitalization. Finally, Yang et al.1 point out that we did not take into account longitudinal changes in the use of angiotensin-converting enzyme inhibitors and diuretics or changes in nutritional status. Regarding the former, we are unaware of either data or a pathophysiologic reason that angiotensin-converting enzyme inhibitors and diuretic use would have a differential effect on eGFRCys versus eGFRCr. Regarding the latter, we do not view the possibility of longitudinal changes in nutritional status as a confounder but rather as a likely mechanism for the effect we observed. Namely that increasing hospitalization days may be associated with decreased creatinine production due to muscle atrophy and poorer nutritional status, and therefore, eGFRCr masks to some extent the true GFR decrease observed in eGFRCys.