In this community-based cohort of older individuals, we observed substantial differences in the rates of kidney function decline using creatinine versus cystatin C based estimates of GFR. When using the Modification of Diet in Renal Disease creatinine equation, we observed an annual eGFR decline of only 0.4 ml/min/1.73 m2
over 7 years of follow-up, far less than the expected rate of decline [1
]. The rate of eGFRcrea
decline was substantially greater in women than men. In contrast, the cystatin C-based eGFR measure declined annually by 1.8 ml/min/1.73 m2
, which was more than 4-fold the rate of eGFRcrea
, and did not differ by sex or race. Increasing age independently predicted more rapid rates of both eGFRcrea
decline. The proportion of participants identified as ‘rapid decliners’ (annual loss of GFR >3 ml/min/1.73 m2
) was 50% higher when using eGFRcys
compared with eGFRcrea
, and eGFRcys
detected twice the incidence of CKD during follow-up compared with eGFRcrea
. Because participants with rapid kidney decline in this cohort had a higher mortality risk [8
], we believe that eGFR changes detected by either creatinine or cystatin C are clinically meaningful.
The most striking finding from our study was how different the eGFR slopes were when based on creatinine or cystatin C. We believe that the differences are most likely caused by creatinine's insensitivity for detecting relatively small changes in GFR in elderly persons. Creatinine generation decreases with age, and likely is even further reduced in persons with accelerated kidney function decline, as these individuals have greater impairments in diet and physical activity. Unfortunately, no study has directly measured GFR in a large cohort of elderly persons, much less repeat GFR measurements several times during follow-up. Therefore, we cannot conclude that either the creatinine or cystatin C measurements are more accurate. In the absence of such a study, clinical research (like clinical practice), must rely on indirect estimates of GFR using endogenous filtration markers. Our prior research comparing associations of creatinine and cystatin C with longitudinal outcomes suggests to us that cystatin C may have a greater ‘signal-to-noise ratio’ than creatinine. In addition, a small cohort of persons with type-2 diabetes and normal GFR levels found that eGFRcys
changed over time in close proximity with measured iothalamate GFR, whereas eGFRcrea
depicted a falsely slow rate of decline [13
]. Furthermore, the eGFRcrea
rates of decline in our study were much lower than expected based on prior studies of the elderly [1
]. Hopefully a future study will measure GFR longitudinally in elderly persons to help us distinguish the relative accuracy and precision of creatinine and cystatin C in this population.
If our data are able to be generalized and cystatin C does capture unrecognized CKD, these results could have substantial public health importance. Based on NHANES 1999–2002, the elderly population (65+) in the United States was 36 million, of whom 28 million did not have CKD. Assuming 9% develop unrecognized CKD during a 7-year follow-up, then 2.5 million (350,000/year) could be detected by serial measurements of cystatin C, but not creatinine. Whether or not this ‘early detection’ strategy would improve patient outcomes is a question for future research.
In this study, the association of sex with kidney function decline differed between analyses based on eGFRcrea
; women had a greater decline in eGFRcrea
than men, but equivalent declines in eGFRcys
. Although no prior studies have addressed this research question using cystatin C, 2 studies that used iothalamate GFR in CKD cohorts, the Modification of Diet in Renal Disease and the African-American Study of Kidney Disease and Hypertension studies, found no association of sex with progression [21
]. Among studies that evaluated longitudinal changes in kidney function by creatinine-based measurements, 8 studies found no independent differences by sex, whereas 6 found a faster progression among men [16
]. A sex-based comparison of the longitudinal associations of eGFRcys
declines with health outcomes would be helpful for determining the relative prognostic utility of cystatin C among men and women.
The incidence of ESRD in the United States is substantially higher among blacks of all ages compared with whites; yet black and white participants had similar rates of kidney function decline in this study, based on creatinine or cystatin C. Although race did not predict rapid decline by either measurement, eGFRcys
detected a greater proportion of black participants with rapid decline of kidney function. Two prior studies in non-CKD cohorts also found no association of race with kidney function decline [25
], whereas a study of nondiabetic CKD found a faster rate of progression among blacks [21
]. A United States population analysis found that blacks and whites have a similar prevalence of CKD, but that blacks have faster progression from CKD to ESRD [27
]. Future studies should evaluate whether the measurement of cystatin C adds useful information for understanding racial disparities in kidney disease.
This study has additional limitations beyond the absence of a ‘gold standard’ for GFR measurements. Many subjects in the CHS had only 1 measurement of kidney function as they died or were otherwise unable to provide adequate sera for a repeat measurement. The absence of these subjects could have biased our results toward an underestimate of kidney function decline by either measure, as the excluded subjects had higher baseline creatinine and worse comorbidity. Both GFR estimating equations used in this study were derived from cohorts that were predominately comprised of nonelderly persons with CKD [12
]. At this time, we do not have GFR estimating equations that have been validated in either elderly persons or in populations without kidney disease. However, the observed differences in eGFRcrea
in this study are attributable to the much larger relative change over time in cystatin C compared with creatinine. Furthermore, less than 15% of our cohort was African-American, and we did not have any Asian or Hispanic participants.
In conclusion, GFR estimates from cystatin C indicated a more than 4-fold rate of decline and a 2-fold incidence of CKD compared with creatinine-based estimates. We also found twice the annual rate of GFR decline in elderly persons than had been reported in prior literature. Given the growing burden of CKD and ESRD, determination of the optimal method for GFR estimation in the elderly, whether by creatinine, cystatin C or a combination of both, should be a high priority for future research.