Large Discordance between Creatinine-Based and Cystatin C-Based eGFRs is Associated with Falls, Hospitalizations, and Death in Older Adults

Abstract

Key Points:

1. A large eGFR discordance (i.e., cystatin C–based eGFR >30% lower than creatinine-based eGFR) is common in older adults and increased with age.

2. A large eGFR discordance was associated with increased risk of falls, hospitalization, and death, independent of kidney function.

3. There are multiple ways to measure differences in creatinine and cystatin C; all produce similar associations with aging-related adverse outcomes.

Background: eGFR calculated using creatinine and cystatin C often differ in older adults. We hypothesized that older adults with cystatin C–based eGFR (eGFRcys) values significantly lower than creatinine-based eGFR (eGFRcr) values may have higher risk of aging-related adverse outcomes, independent of kidney function.

Methods: We conducted a longitudinal cohort study of adults 65 years and older from the Health and Retirement Study, a cohort of older American adults, to determine the relationship between eGFR discordance and aging-related adverse outcomes. We calculated eGFRcr and eGFRcys using baseline creatinine and cystatin C measurements. A large eGFR discordance was defined as eGFRcys >30% lower than eGFRcr. We assessed four aging-related adverse outcomes over a 2-year follow-up: falls, hip fractures, hospitalizations, and death. We fit separate multivariable regression models to determine the association between having a large eGFR discordance and each outcome adjusting for confounders, including kidney function.

Results: Of 5574 older adults, 1683 (30%) had a large eGFR discordance. Those with a large eGFR discordance were more likely to be older, female, and White. The prevalence of a large eGFR discordance increased with age, from 20% among those 65–69 years to 44% among those 80 years and older. Over a 2-year follow-up, there were 305 deaths (5.5%), 2013 falls (39.2%), 69 hip fractures (1.3%), and 1649 hospitalizations (32.2%). In adjusted analyses, large eGFR discordance was associated with a higher hazard ratio for death (hazard ratio, 1.43; 95% confidence interval [CI], 1.12 to 1.82) and significantly higher odds of falls (odds ratio, 1.32; 95% CI, 1.16 to 1.51) and hospitalizations (odds ratio, 1.32; 95% CI, 1.15 to 1.51). A large eGFR discordance was not associated with hip fractures.

Conclusions: In a large, nationally representative cohort of older adults, prevalence of eGFR discordance increased with age and was associated with higher risk of falls, hospitalization, and death, independent of kidney function.

Graphical abstract

Figure 1

Participant flow. BMI, body mass index; CKD-5, CKD stage 5; CRP, C-reactive protein; HRS, Health and Retirement Study; eGFRcr-cys, creatinine–cystatin C–based eGFR; VBS, venous blood study.

Figure 2

Prevalence of discordant (>30%) eGFR by age group. eGFR discordance was defined as eGFRcys more than 30% lower than eGFRcr. eGFRcr, calculated by the CKD-EPIcr equation (2021); eGFRcys, calculated by CKD-EPIcys equation (2012). CKD-EPI, CKD Epidemiology Collaboration; eGFRcr; creatinine-based eGFR; eGFRcys, cystatin C–based eGFR.

Figure 3

Forest plot for aging-related adverse outcomes of discordant (>30%) eGFR. Forrest plot depicting effect sizes and 95% CIs for clinical outcomes adjusted for age, sex, smoking status, CKD stage, BMI, hypertension, diabetes mellitus, atherosclerotic disease, osteoporosis, history of cancer, cognitive dysfunction, and albumin level <3.5 g/dl. For death, time-to-event modeling was performed and reported as HR and 95% CI. Falls, hip fractures, and hospitalizations were analyzed by logistic regression modeling, reported as OR and 95% CI. BMI, body mass index; CI, confidence interval; HR, hazard ratio; OR, odds ratio.

Figure 4

Magnitude of eGFR discordance and incidence of aging-related adverse outcomes. LOESS plot depicting the association between eGFR discordance and death (A), falls (B), hip fractures (C), and hospitalizations (D). X axis displays ventile of eGFR discordance (eGFRcys/eGFRcr*) representing 5% of the cohort; ventile 1 indicates eGFR discordance from 1.12 to 2.37 (i.e., eGFRcys 12%–37% higher than eGFRcr), and ventile 20 represents discordance from 0.22 to 0.53 (i.e., eGFRcys 47%–78% lower than eGFRcr). Open dots represent the incidence per ventile of each adverse outcome. The area shaded in gray represents the ventiles that correlate with eGFRcys >30% lower than eGFRcr. *eGFRcr, calculated by the CKD-EPIcr equation (2021); eGFRcys, calculated by CKD-EPIcys equation (2012). LOESS, locally estimated scatterplot smoothing.

Figure 5

Twenty-four–month survival in participants with discordant eGFR. Kaplan–Meier curves comparing 24-month survival in participants with an eGFRcys >50% lower than eGFRcr (blue), an eGFRcys 30%–50% lower than eGFRcr (green), and concordant eGFR (red). All curves were significantly different (P < 0.001). eGFRcr, calculated by the CKD-EPIcr equation (2021); eGFRcys, calculated by CKD-EPIcys equation (2012).