Regional Variation in Hemoglobin Distribution Among Individuals With CKD: the ISN International Network of CKD Cohorts

Abstract

Introduction: Despite recognized geographic and sex-based differences in hemoglobin in the general population, these factors are typically ignored in patients with chronic kidney disease (CKD) in whom a single therapeutic range for hemoglobin is recommended. We sought to compare the distribution of hemoglobin across international nondialysis CKD populations and evaluate predictors of hemoglobin.

Methods: In this cross-sectional study, hemoglobin distribution was evaluated in each cohort overall and stratified by sex and estimated glomerular filtration rate (eGFR). Relationships between candidate predictors and hemoglobin were assessed from linear regression models in each cohort. Estimates were subsequently pooled in a random effects model.

Results: A total of 58,613 participants from 21 adult cohorts (median eGFR range of 17-49 ml/min) and 3 pediatric cohorts (median eGFR range of 26-45 ml/min) were included with broad geographic representation. Hemoglobin values varied substantially among the cohorts, overall and within eGFR categories, with particularly low mean hemoglobin observed in women from Asian and African cohorts. Across the eGFR range, women had a lower hemoglobin compared to men, even at an eGFR of 15 ml/min (mean difference 5.3 g/l, 95% confidence interval [CI] 3.7-6.9). Lower eGFR, female sex, older age, lower body mass index, and diabetic kidney disease were all independent predictors of a lower hemoglobin value; however, this only explained a minority of variance (R² 7%-44% across cohorts).

Conclusion: There are substantial regional differences in hemoglobin distribution among individuals with CKD, and the majority of variance is unexplained by demographics, eGFR, or comorbidities. These findings call for a renewed interest in improving our understanding of hemoglobin determinants in specific CKD populations.

Keywords: anemia; chronic kidney disease; geography; glomerular filtration rate; hemoglobin; sex.

Figure 1

Bubble plot showing the mean hemoglobin for men (left panel) and women (right panel) in each iNET-CKD cohort and grouped by world region.

Figure 2

Violin plot of the distribution of mean hemoglobin values in participating iNET-CKD cohorts stratified by sex and category of eGFR (left panel), and estimates of sex-based differences in hemoglobin level from age-adjusted and multi-variable-adjusted models (right panel). A violin plot is a hybrid of a box plot and a kernel density plot. The boxes represent the median and interquartile range (IQR), and the whiskers represent 1.5 times the IQR. The kernel density estimation shows the shape of the distribution. Wider and narrower sections of the plot represent a higher and lower probability, respectively, that participating cohorts of the study will take on the given value. Cohorts with less than 10 participants in a given stratum of sex and eGFR category were excluded from the plot.

Figure 3

Pooled estimates of the difference in mean hemoglobin according to the etiology of CKD based on meta-analysis of the unadjusted and multivariable-adjusted cohort-level estimates. For the multivariable-adjusted results, estimates are presented separately in individuals with and without diabetes mellitus. In each case, diabetic kidney disease (DKD) serves as the reference group. For example, in the case of glomerulonephritis (GN), the green estimate represents the mean difference in hemoglobin in non-diabetic individuals with GN compared to those with DKD; whereas, the red estimate represents the mean difference in hemoglobin in diabetic individuals with GN compared to those with DKD. DKD, diabetic kidney disease; DM, diabetes mellitus; GN, glomerulonephritis; HTN, hypertension; PKD, polycystic kidney disease.

Figure 4

Cohort-specific and pooled estimates for the expected value of hemoglobin in 3 hypothetical patients conditional on specific characteristics generated from the extended multivariable model (n = 15 cohorts). In all cases, eGFR is fixed at 34 ml/min per 1.73 m². Unless otherwise specified, other covariates are held at their mean values. eGFR, estimated glomerular filtration rate.