Serum Cystatin C May Be a Better Marker of Renal Impairment than Creatinine

To the Editor: We would like to comment on Van Den Noortgate et al.'s recent paper in the Journal of the American Geriatrics Society.1 The authors were unable to demonstrate a significant difference between serum cystatin C and creatinine as markers of renal function in older patients. They conclude that serum cystatin C is not superior to serum creatinine in the detection of renal impairment. We have some concerns regarding the statistical approach that was used to reach this conclusion. The authors report a sensitivity for renal impairment of 73% for cystatin C and 57% for serum creatinine but report that the areas under the curve using receiver operator curve (ROC) analysis were not significantly different. The choice of cutoff is critical to such analysis. The cutoff chosen for serum creatinine (0.85 mg/dL, 77 μmol/L) lies within the lower end of reference ranges typically used by laboratories (e.g., 71–124 μmol/L for men and 53–97 μmol/L for women).2 Conversely, for cystatin C, a cutoff of 1.09 mg/L was chosen; no data are cited supporting this cutoff, which is above the upper limit of normal reported by other authors using the same methodology (e.g., 0.98 mg/L3, 1.01 mg/L4). This results in overestimation and underestimation of the sensitivities of serum creatinine and cystatin C, respectively. We also have doubts about the value of ROC analysis when the total population studied was only 48; the authors' failure to observe a significant difference may well reflect a type 2 error (i.e., a false-negative result). Indeed, the description of these sensitivities in percentage terms is in conflict with the Journal's statistical requirements, because fewer than 50 subjects were studied. A more appropriate conclusion would have been that serum cystatin C may be superior to creatinine in the detection of mild renal impairment but that confirmation of this would require a larger study. A recent meta-analysis supports such a conclusion.5 Finally, we would take issue with the statement that serum cystatin C measurement is 100 times as expensive as a serum creatinine measurement. In our laboratories, a single serum creatinine measurement costs approximately $0.15, compared with approximately $3.00 for cystatin C, but when the true costs of these investigations are taken into account (e.g., including staff and overheads), the actual difference in cost becomes marginal. To the Editor: We appreciate the comments of Dr. Lamb et al. and would like to comment on the following points: serum creatinine values are dependent on the assays used for measurement.1 The reference values cited by Lamb et al. are clearly values of an uncompensated Jaffé method, which is affected by an important protein error. Reference values using a compensated Jaffé (as used by Van Den Noortgate) or an enzymatic method are much lower than those cited by Lamb: 61.9–106.1 μmol/L (0.70–1.20 mg/dL) for men and 44.2–79.6 μmol/L (0.50–0.90 mg/dL) for women.2 Recent work has shown that compensated Jaffé and enzymatic creatinine values better reflect glomerular filtration rate than uncompensated Jaffé methods.3 Nevertheless, recent meta-analyses also pay little attention to the methods of serum creatinine measurements.4,5 Moreover, receiver operating curve (ROC) analysis has the advantages of comparing different methods and not being sensitive to cutoff levels. The area under the curves for cystatin C and serum creatinine were comparable (P=.5). Including two more patients to reach the Journal's statistical requirements would most certainly not have changed the level of significance in our study. Concerning the cost of creatinine measurement, the price given by Lamb et al. is the price for an enzymatic creatinine assay. Jaffé methods cost approximately $0.02, which is 1/100th the cost of cystatin C measurements ($2.00). Lamb et al. also discuss the cost-effectiveness in the meta-analysis study.5 We also want to draw attention to the fact that we studied an old-old population admitted to a geriatric department and not a healthy young elderly population, as in previous studies.6 This may have consequences for the cutoff value of cystatin C because there is current evidence that the properties of cystatin C are made less clear by the presence of malignancy and by the use of medications such as corticosteroids.7,8 As far as we could discern, malignancy and the use of corticosteroids were excluded. In conclusion, because the assay methods for serum creatinine influence the obtained values and because the properties of cystatin C in an elderly population can be different, we are not sure that our failure to observe a superiority of cystatin C is due to a type II error. Moreover, cystatin C measurement is expensive and therefore, pending the results of a larger number of patients, should not be recommended for routine use in an elderly population. Nele J. Van Den Noortgate, MD 5 Joris R. Delanghe, MD, PhD 6 Norbert H. Lameire, MD, PhD 7 5 Department of Internal Medicine Division of Gerontology and Geriatrics 6 Department of Clinical Biochemistry 7 Department of Internal Medicine Renal Division University Hospital Ghent, Belgium