Relationship of Estimated GFR and Albuminuria to Concurrent Laboratory Abnormalities: An Individual Participant Data Meta-analysis in a Global Consortium

Abstract

Rationale & objective: Chronic kidney disease (CKD) is complicated by abnormalities that reflect disruption in filtration, tubular, and endocrine functions of the kidney. Our aim was to explore the relationship of specific laboratory result abnormalities and hypertension with the estimated glomerular filtration rate (eGFR) and albuminuria CKD staging framework.

Study design: Cross-sectional individual participant-level analyses in a global consortium.

Setting & study populations: 17 CKD and 38 general population and high-risk cohorts.

Selection criteria for studies: Cohorts in the CKD Prognosis Consortium with data for eGFR and albuminuria, as well as a measurement of hemoglobin, bicarbonate, phosphorus, parathyroid hormone, potassium, or calcium, or hypertension.

Data extraction: Data were obtained and analyzed between July 2015 and January 2018.

Analytical approach: We modeled the association of eGFR and albuminuria with hemoglobin, bicarbonate, phosphorus, parathyroid hormone, potassium, and calcium values using linear regression and with hypertension and categorical definitions of each abnormality using logistic regression. Results were pooled using random-effects meta-analyses.

Results: The CKD cohorts (n=254,666 participants) were 27% women and 10% black, with a mean age of 69 (SD, 12) years. The general population/high-risk cohorts (n=1,758,334) were 50% women and 2% black, with a mean age of 50 (16) years. There was a strong graded association between lower eGFR and all laboratory result abnormalities (ORs ranging from 3.27 [95% CI, 2.68-3.97] to 8.91 [95% CI, 7.22-10.99] comparing eGFRs of 15 to 29 with eGFRs of 45 to 59mL/min/1.73m²), whereas albuminuria had equivocal or weak associations with abnormalities (ORs ranging from 0.77 [95% CI, 0.60-0.99] to 1.92 [95% CI, 1.65-2.24] comparing urinary albumin-creatinine ratio > 300 vs < 30mg/g).

Limitations: Variations in study era, health care delivery system, typical diet, and laboratory assays.

Conclusions: Lower eGFR was strongly associated with higher odds of multiple laboratory result abnormalities. Knowledge of risk associations might help guide management in the heterogeneous group of patients with CKD.

Keywords: CKD Prognosis Consortium; CKD stage; Chronic kidney disease (CKD); albuminuria; anemia; diabetes; glomerular filtration rate (GFR); hematocrit; hemoglobin; hyperparathyroidism; hypertension; individual-level meta-analysis; kidney function; laboratory abnormality; laboratory tests; meta-analysis; serum bicarbonate; serum calcium; serum intact parathyroid hormone; serum phosphorus; serum potassium; staging system.

Figure 1.

Associations between estimated glomerular filtration rate (eGFR) and continuous laboratory measures by albuminuria stages in chronic kidney disease cohorts: (A) hemoglobin, (B) potassium, (C) bicarbonate, (D) parathyroid hormone, (E) phosphorus, and (F) calcium. The y-axis depicts the meta-analyzed difference from the mean adjusted value at eGFR of 50 mL/min/1.73 m² and albuminuria wih albumin excretion < 30 mg/g. eGFR was modeled as a 3-piece linear spline with knots at 30 and 45 mL/min/1.73 m²; the reference point in continuous analysis was set at 50 mL/min/1.73 m².

Figure 2.

Association between estimated glomerular filtration rate (eGFR) and continuous laboratory measures by albuminuria stages in general population and high-risk cohorts: (A) hemoglobin, (B) potassium, (C) bicarbonate, (D) parathyroid hormone, (E) phosphorus, and (F) calcium. The y- axis depicts the meta-analyzed difference from the mean adjusted value at eGFR of 80 mL/min/1.73 m² and albuminuria with albumin excretion < 30 mg/g. eGFR was modeled as a 7-piece linear spline with knots at 30, 45, 60, 75, 90, and 105 mL/min/1.73 m²; the reference point in continuous analysis was set at 80 mL/min/1.73 m².

Figure 3.

Meta-analyzed adjusted prevalence (25th and 75th percentile cohort) of abnormalities (categorical laboratory measures) in chronic kidney disease by diabetes status. The adjusted prevalence of each abnormality at each estimated glomerular filtration rate (eGFR) and albuminuria stage was computed as follows: first, we converted the random-effects weighted adjusted mean odds at the reference point (eGFR, 50 mL/min/1.73 m²) into a prevalence estimate. To the reference estimate, we applied the meta-analyzed odds ratios to obtain prevalence estimates at eGFRs of 95, 80, 65, 35, and 20 mL/min/1.73 m² for each stage of albuminuria with and without diabetes. The prevalence estimates were adjusted to 60 years old, half men, nonblack, 20% history of cardiovascular disease, 40% ever smoker, and body mass index of 30 kg/m2. The 25th and 75th percentiles for predicted prevalence were the estimates from individual cohorts in the corresponding percentiles of the random-effects weighted distribution of adjusted odds. This was done separately for each abnormality. Note that the cohorts included in the analyses of each abnormality may differ based on data availability. For example, the cohort in the 25th percentile of anemia may not be the same as the cohort in the 25th percentile of hyperparathyroidism. Color coding is based on odds ratio quartile within each abnormality. Bold red font indicates the reference cell. Definitions of each abnormality are as follows: anemia: Hgb, male < 13 g/dL, female < 12 g/dL; Hct, male < 39%, female < 36%; hyperkalemia: potassium > 5 mmol/L; acidosis, bicarbonate < 22 mmol/L; hyperparathyroidism, intact parathyroid hormone > 65 pg/mL; hyperphosphatemia, phosphorus > 4.5 mg/dL; and hypocalcemia, corrected calcium < 8.5 mg/dL.

Figure 4.

Association between estimated glomerular filtration rate (eGFR) and hypertension by albuminuria stages in (A) chronic kidney disease (CKD) cohorts and (B) general population and high-risk cohorts. Abbreviations: A1, A2, A3 refer to albuminuria stages: A1, <30 mg/g; A2, 30-299 mg/g; and A3, 300+ mg/g. The y-axis refers to the meta-analyzed adjusted odds ratio and 95% confidence interval compared to a reference of eGFR of 50 (80 in the right graph) mL/min/1.73 m² in A1 (black diamond). In analyses of the general population/high-risk cohorts, eGFR was modeled as a 7-piece linear spline with knots at 30, 45, 60, 75, 90, and 105 mL/min/1.73 m²; the reference point in continuous analysis was set at 80 mL/min/1.73 m². In analyses of CKD pop-ulations, eGFR was modeled as a 3-piece linear spline with knots at 30 and 45 mL/min/1.73 m²; the reference point in continuous analysis was set at 50 mL/min/1.73 m².