Age and association of kidney measures with mortality and end-stage renal disease

Abstract

Context: Chronic kidney disease (CKD) is prevalent in older individuals, but the risk implications of low estimated glomerular filtration rate (eGFR) and high albuminuria across the full age range are controversial.

Objective: To evaluate possible effect modification (interaction) by age of the association of eGFR and albuminuria with clinical risk, examining both relative and absolute risks.

Design, setting, and participants: Individual-level meta-analysis including 2,051,244 participants from 33 general population or high-risk (of vascular disease) cohorts and 13 CKD cohorts from Asia, Australasia, Europe, and North/South America, conducted in 1972-2011 with a mean follow-up time of 5.8 years (range, 0-31 years).

Main outcome measures: Hazard ratios (HRs) of mortality and end-stage renal disease (ESRD) according to eGFR and albuminuria were meta-analyzed across age categories after adjusting for sex, race, cardiovascular disease, diabetes, systolic blood pressure, cholesterol, body mass index, and smoking. Absolute risks were estimated using HRs and average incidence rates.

Results: Mortality (112,325 deaths) and ESRD (8411 events) risks were higher at lower eGFR and higher albuminuria in every age category. In general and high-risk cohorts, relative mortality risk for reduced eGFR decreased with increasing age; eg, adjusted HRs at an eGFR of 45 mL/min/1.73 m2 vs 80 mL/min/1.73 m2 were 3.50 (95% CI, 2.55-4.81), 2.21 (95% CI, 2.02-2.41), 1.59 (95% CI, 1.42-1.77), and 1.35 (95% CI, 1.23-1.48) in age categories 18-54, 55-64, 65-74, and ≥75 years, respectively (P <.05 for age interaction). Absolute risk differences for the same comparisons were higher at older age (9.0 [95% CI, 6.0-12.8], 12.2 [95% CI, 10.3-14.3], 13.3 [95% CI, 9.0-18.6], and 27.2 [95% CI, 13.5-45.5] excess deaths per 1000 person-years, respectively). For increased albuminuria, reduction of relative risk with increasing age was less evident, while differences in absolute risk were higher in older age categories (7.5 [95% CI, 4.3-11.9], 12.2 [95% CI, 7.9-17.6], 22.7 [95% CI, 15.3-31.6], and 34.3 [95% CI, 19.5-52.4] excess deaths per 1000 person-years, respectively by age category, at an albumin-creatinine ratio of 300 mg/g vs 10 mg/g). In CKD cohorts, adjusted relative hazards of mortality did not decrease with age. In all cohorts, ESRD relative risks and absolute risk differences at lower eGFR or higher albuminuria were comparable across age categories.

Conclusions: Both low eGFR and high albuminuria were independently associated with mortality and ESRD regardless of age across a wide range of populations. Mortality showed lower relative risk but higher absolute risk differences at older age.

Figure 1

Adjusted hazard ratios (panels A and B) for all-cause mortality and average mortality rate (panels C and D) according to eGFR (A and C) and ACR (B and D) within each age category. The filled circles note statistical significance (p<0.05) compared to the reference (black diamond) eGFR of 80 ml/min1.73m² or ACR of 10 mg/g within each age category in panels A and B and compared to age category 55–64 years in panels C and D. Plus signs and open circles on the bottom of each panel represent significantly positive (greater effect size) and negative (smaller effect size) point-wise interactions (p<0.05), respectively, compared to age 55–64 years. Gaps indicate no significant point-wise interaction. Models are meta-analysis of general population and high risk cohorts adjusted for sex, race, body mass index, systolic blood pressure, total cholesterol, history of cardiovascular disease, diabetes, smoking status, and albuminuria (panels A and C) or eGFR (panels B and D).

Figure 2

Adjusted hazard ratios (panels A and B) and average incidence rate (panels C and D) for ESRD according to eGFR (A and C) and ACR (B and D) within each age category. The filled circles note statistical significance (p<0.05) compared to the reference (black diamond) eGFR of 80 ml/min1.73m2 or ACR of 10 mg/g within each age category in panels A and B and compared to age category 55–64 years in panels C and D. Plus signs and open circles on the bottom of each panel represent significantly positive (greater effect size) and negative (smaller effect size) point-wise interactions (p<0.05), respectively, compared to age 55–64 years. Gaps indicate no significant point-wise interaction. Models are meta-analysis of general population and high risk cohorts adjusted for sex, race, body mass index, systolic blood pressure, total cholesterol, history of cardiovascular disease, diabetes, smoking status, and albuminuria (panels A and C) or eGFR (panels B and D).

Figure 3

Adjusted hazard ratio (panel A) and incidence rate difference (panel B) for all-cause mortality by categories of eGFR and albuminuria across age groups. Each number represents a pooled estimate from meta-analysis in 34 general population and high risk cohorts. All results are adjusted for covariates. Bold numbers indicated statistical significance at P<0.05 compared to the reference cell and circles (^○) indicate a significant negative interaction and stars (*) indicate a significant positive interaction at P<0.05 compared to the corresponding cell in the 55–64 year age group. Color shading indicates the strength of association (approximately one quarter of all cells are shaded in each color; Green: low; yellow: mild; orange: moderate; red: high) measured as either hazard ratios (panel A) or incidence rate difference (panel B). Confidence intervals are provided in eTable 2.

Figure 4

Adjusted hazard ratios (panels A and B) and average incidence rate (panels C and D) for all-cause mortality (A and C) and ESRD (B and D) in CKD cohorts according to eGFR within each age category. The filled circles note statistical significance (p<0.05) compared to the reference (black diamond) eGFR of 50 ml/min1.73m² within each age category in panels A and B and compared to age category 55–64 years in panels C and D. Plus signs and open circles on the bottom of each panel represent significantly positive (greater effect size) and negative (smaller effect size) point-wise interactions (p<0.05), respectively, compared to age 55–64 years. Gaps indicate no significant point-wise interaction. Models are meta-analysis of CKD cohorts adjusted for sex, race, body mass index, systolic blood pressure, total cholesterol, history of cardiovascular disease, diabetes, smoking status, and albuminuria.