Association between Time-Updated Eosinophil Counts and Progression of CKD

doi:10.21203/rs.3.rs-2003296/v1

Download PDF

Article

Association between Time-Updated Eosinophil Counts and Progression of CKD

https://doi.org/10.21203/rs.3.rs-2003296/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Patients with chronic kidney disease (CKD) have high blood eosinophil count but its clinical implication is uncertain. Since eosinophils may induce tubulointerstitial injury and arteriosclerosis, eosinophilia might be related to poor clinical outcomes. This retrospective cohort study included 2,877 patients whose estimated glomerular filtration rate (eGFR) was 10–60 mL/min/1.73 m². The exposure was time-updated blood eosinophil counts. The outcomes were 1) initiation of renal replacement therapy (RRT) and 2) cardiovascular events and mortality. We analyzed the associations between eosinophil counts and outcomes using marginal structural models (MSM). Over a median follow-up of 6.5 years, eosinophil counts were measured a median of 22 times per patient (4 times a year per patient). There was a negative correlation between eosinophil count and eGFR. In total, 433 patients initiated RRT, 275 developed cardiovascular events, and 165 died. In MSM, higher eosinophil counts (≥ 289/µL) showed a 1.83-fold (95% confidence interval:1.33–2.51) higher rate of RRT initiation than lower eosinophil counts after adjustment for time-dependent confounders. Higher eosinophil counts were also associated with a higher rate of cardiovascular events and mortality in MSM (hazard ratio, 1.71 [95% confidence interval:1.30–2.25]). In conclusion, patients with CKD who had higher eosinophil counts showed worse kidney outcome.

eosinophil

chronic kidney disease

mortality

cardiovascular events

marginal structural model

Eosinophils are multifunctional leukocytes involved in an array of pathological processes¹. Besides their well-known roles in allergic reactions, parasite defense, and autoimmune diseases, eosinophils are also implicated in atherosclerosis^2–4. Marx et al. showed that activated eosinophils in atherosclerotic lesions accelerate plaque formation in concert with platelets by secreting eosinophilic granule proteins and extracellular traps⁵. Population-based cohort studies reported increased cardiovascular risks among those with higher levels of plasma eosinophilic cationic protein (ECP), a marker of eosinophil activity and degranulation^6,7.

Patients with advanced chronic kidney disease (CKD) have high blood eosinophil count^8,9 although its clinical implication is uncertain. In addition to their proatherogenic property, eosinophils might also contribute to the progression of kidney disease. For example, renal complications sometimes develop in idiopathic hypereosinophilic syndrome, where tubulointerstitial infiltration of eosinophils is typically observed^10,11. Eosinophilic granulomatosis with polyangiitis, characterized by the interstitial infiltration of eosinophils, is also evidentiary to the involvement of eosinophils in kidney injury¹². Furthermore, interstitial eosinophilic aggregates are found in common kidney diseases, such as diabetic nephropathy, IgA nephropathy, and membranous nephropathy, which are related to interstitial fibrosis and inflammation^13,14. Since there is a positive correlation between eosinophil counts in the blood and renal interstitium^14,15, we hypothesized that increased blood eosinophil counts in patients with CKD reflect eosinophilic inflammation in the kidney and thus indicate a risk of CKD progression. In the current study, we examined the association between time-updated blood eosinophil counts and the risk of kidney failure among patients with advanced CKD.

Ethical considerations

This study was conducted in accordance with the principles of the Declaration of Helsinki. The Ethics Committee of Osaka University Hospital approved the study protocol and waived the need for a written informed consent given the retrospective nature of the study (no: 20352, 22047).

Study design and participants

This retrospective cohort study included all patients referred to the outpatient department of nephrology at Osaka University Hospital from January 2005 to January 2018 who met the following inclusion criteria: 1) aged 20 years or older, 2) an estimated glomerular filtration rate (eGFR) of 10–60 mL/min/1.73 m², and 3) not receiving renal replacement therapy (RRT). Patients were excluded if they 1) were followed up for < 1 year or 2) received corticosteroids.

Follow-up period was from the first visit to the hospital to death, RRT initiation, loss to follow-up, or February 28th, 2019, whichever occurred first.

Data collection

The detailed methods have been described elsewhere^16–18. Demographics and comorbidities were extracted from a chart review of patients’ electronic medical records by nephrologists. These included age, sex, body mass index (BMI), blood pressure, diabetes mellitus (DM), cardiovascular comorbidities, chronic respiratory diseases (chronic obstructive pulmonary disease and bronchial asthma), a prior history of arterial catheterization (cardiac catheterization and endovascular treatment for peripheral artery diseases and carotid artery stenosis), and cholesterol embolism. Cardiovascular comorbidities included coronary artery disease, congestive heart failure, valvular heart disease, aortic disease, and stroke (cerebral infarction or intracranial hemorrhage).

Time-series data on laboratory measurements and prescriptions were collected using an automated data extraction system of Osaka University Hospital. Laboratory data included serum albumin, creatinine, sodium, potassium, C-reactive protein (CRP), hemoglobin, white blood cell (WBC) counts, and urinary protein-to-creatinine ratio (UPCR). eGFR was calculated using the equation for the Japanese population¹⁹. The prescription data included loop diuretics, thiazide diuretics, mineralocorticoid receptor antagonists (MRAs), angiotensin-converting enzyme inhibitors (ACEIs), angiotensin II receptor blockers (ARBs), nonsteroidal anti-inflammatory drugs (NSAIDs), proton pump inhibitors (PPIs), histamine H₂ receptor antagonists (H₂ blockers), and corticosteroids. We assessed the number of prescription drugs, not limited to those mentioned above. The time-series data were collected at a monthly interval during the study period.

We collected information regarding arterial catheterization performed during follow-up (cardiac catheterization and endovascular treatment for peripheral artery diseases and carotid artery stenosis); these procedures may have contributed to the development of eosinophilia via cholesterol embolism. These data were obtained from diagnostic procedure combination (DPC) codes²⁰.

Exposure

The exposure was time-updated blood eosinophil counts (per µL), which were collected at a monthly interval during the study period. Blood eosinophil counts were calculated by multiplying the total WBC count by the percentage of eosinophils as measured using an automated WBC differential counter. Blood eosinophil counts were categorized into quartiles.

Study outcomes

The study outcome was RRT initiation, defined as the initiation of chronic dialysis or kidney transplantation. We additionally evaluated cardiovascular outcomes, which were a composite of myocardial infarction, stroke, hospitalization for heart failure, and mortality. The dates of these clinical events were ascertained by a chart review of the patients’ electronic medical records by nephrologists.

A post-hoc analysis of a randomized controlled trial of oral carbon adsorbent

To explore the involvement of uremic toxins in elevated eosinophil counts in CKD, we analyzed the data in a previous randomized controlled trial of the oral carbon adsorbent AST-120²¹, which reduces serum uremic toxin levels such as indoxyl sulfate. This two-year, open-label, randomized, controlled trial enrolled 125 patients with stages 3–4 CKD. Among them, 123 were randomized to either receive AST-120 (6 g/day) or not, in a 3:2 ratio. In this post-hoc analysis, data on eosinophil counts at baseline, 3, 6, and 12 months were added to the original dataset.

Statistical analyses

The relationship between eosinophil counts and eGFR at baseline was depicted using a restricted cubic spline curve with three knots (10th, 50th, and 90th percentiles of eGFR).

The multivariable association between log-transformed eosinophil counts and covariates was assessed by a linear regression analysis with robust standard errors. The following variables were included: age, sex, BMI, systolic blood pressure, DM, cardiovascular comorbidities, a prior history of arterial catheterization, chronic respiratory diseases, ACEIs/ARBs, loop diuretics, thiazide diuretics, MRAs, PPIs, H₂ blockers, NSAIDs, number of drugs prescribed, hemoglobin, albumin, eGFR, CRP, UPCR, and WBC.

In the post-hoc analysis of the randomized trial of AST-120²¹, eosinophil counts were compared between the AST-120 and control groups using a linear mixed-effects model for repeated measures with an unstructured covariance matrix.

To analyze the longitudinal relationship between time-updated blood eosinophil counts and kidney outcomes, time-dependent confounding should be considered. This is because eosinophil counts increase as kidney function declines. As a result, time-dependent confounding could occur owing to a potential bidirectional relationship between eosinophil counts and kidney function in terms of the development of kidney failure. In order to appropriately account for time-dependent confounding, we used a marginal structural model (MSM). We also performed baseline Cox model, time-average Cox model, and group-based trajectory model (Fig. 1).

1) Baseline Cox model

Association between baseline eosinophil quartiles and outcomes was analyzed using multivariate Cox proportional hazards models. The following baseline covariates were adjusted in this model: age, sex, BMI, systolic blood pressure, DM, cardiovascular comorbidities, chronic respiratory diseases, a prior history of arterial catheterization and cholesterol embolism, hemoglobin, albumin, eGFR, sodium, potassium, CRP, WBC, UPCR, loop diuretics, thiazide diuretics, MRAs, ACEIs, ARBs, NSAIDs, PPIs, and H2 blockers. The proportional hazards assumption was checked graphically based on the scaled Schoenfeld residuals.

2) Time-average Cox model

The average eosinophil count during the first 12 months of follow-up was calculated for each patient. Association between time-average eosinophil quartiles and outcomes was analyzed using a multivariate Cox proportional hazards model adjusted for the same covariates as in the baseline model. In this model, the onset of survival time was set at 12 months.

3) Group-based trajectory model

Group-based trajectory model was used to assess the association between eosinophil count trajectories during the first 12 months and subsequent rates of outcomes (STATA command, traj). All available data on eosinophil counts during the first 12 months were used to identify eosinophil count trajectories. In this analysis, the eosinophil counts were log-transformed to normalize their distribution.

The group-based trajectory model is a method of data clustering that assumes that a population is composed of a mixture of distinct groups characterized by their longitudinal trajectories^22–25. Potential trajectory groups were estimated from individual longitudinal eosinophil data, using the maximum likelihood estimation method based on the finite mixture model theorem. The patients were divided into one of the trajectory groups according to their estimated probability of group membership. We selected the optimal number of trajectory groups, as well as a function of each trajectory, based on the Bayesian information criterion (BIC), with at least 5% of all patients being in the smallest group.

After deriving the eosinophil trajectory groups, multivariate Cox proportional hazards models were used to analyze the association between the trajectory groups and outcomes, adjusting for the same covariates as in the baseline model. In this model, the onset of survival time was set at 12 months.

4) MSM

MSM was employed to 1) assess the time-varying eosinophil counts throughout the study period and 2) deal with time-dependent confounding between eosinophil counts and eGFR.

MSM is a statistical method that can account for time-dependent confounding^26–29. In the current study, eGFR was considered to be the main time-dependent confounder because it influenced both exposure (eosinophil counts) and renal outcomes, while being possibly affected by previous eosinophil counts. We derived time-varying inverse probability weights (IPWs) from the inverse probability of treatment weights (IPTWs) and the inverse probability of censoring weights (IPCWs). IPTWs were the reciprocal of the predicted probability of each patient having their own exposure history (i.e., high eosinophil count or not). The probability was predicted by a logistic regression model at each of the 1-month follow-up periods, conditional on both baseline and time-dependent covariates, as described below. Two different definitions of high eosinophil counts were adopted: 1) eosinophil count ≥ 289/µL (the top 25th percentile in our cohort) and 2) eosinophil count ≥ 500/µL³⁰. Similarly, IPCWs were the reciprocal of the probability of being uncensored, as predicted by a logistic regression model, conditional on both baseline and time-dependent covariates. IPTWs and IPCWs were stabilized by multiplying them with the predicted probabilities based on baseline covariates alone. The IPWs were the product of the stabilized IPTWs and IPCWs, calculated at baseline and for each month. The IPWs were truncated at the 1st and 99th percentiles to reduce the influence of extreme weight values.

Baseline covariates included were the same as in the baseline model. Time-dependent covariates included arterial catheterization performed during follow-up, hemoglobin, albumin, eGFR, sodium, potassium, CRP, UPCR, loop diuretics, thiazide diuretics, MRAs, ACEIs, ARBs, NSAIDs, PPIs, H2 blockers, and corticosteroids.

MSM created “pseudo-populations” using IPWs, comparing the rate of events if all patients had been continuously exposed to high eosinophil counts with the risk of events if they had never been exposed to it. In MSM, there was no association between measured time-dependent confounders and future exposure. We estimated the hazard ratio (HR) and 95% confidence interval (CI) using an IPW-weighted pooled logistic regression model that produced equivalent estimates to the Cox proportional hazards model.

Effect modification was evaluated by incorporating cross-product terms between eosinophil counts and a priori specified baseline covariates into the MSM, including age (< 70 vs. ≥ 70), sex, BMI (< 22 vs. ≥ 22), systolic blood pressure (< 130 vs. ≥ 130 mmHg), DM, cardiovascular comorbidities, hemoglobin (< 12.4 vs. ≥ 12.4 g/dL), albumin (< 3.8 vs. ≥ 3.8 g/dL), CKD stage (stage 3 vs. stage 4–5), UPCR (< 1.0 vs. ≥ 1.0 g/gCr), and ACEIs/ARBs use.

Missing data at baseline were imputed using the multiple imputations by chained equation method based on all baseline covariates. Continuous variables with missing data (BMI, systolic blood pressure, eGFR, hemoglobin, sodium, potassium, UPCR, albumin, and CRP) were imputed based on linear regression imputation. We created ten imputed datasets that were analyzed separately and combined using Rubin’s rules. Missing data during follow-up were imputed using the last-observation-carried-forward method.

Two sensitivity analyses were conducted. First, the association between eosinophil count and RRT initiation was assessed after excluding patients with chronic respiratory diseases or cholesterol embolism. Second, we reanalyzed MSM after excluding patients who were followed up for less than three months.

Statistical analyses were performed using Stata/IC software (version 16.0; Stata Corp, College Station, TX, USA).

Study population

Among 2,889 patients who met the inclusion criteria, 2,877 (99%) had available eosinophil count data (Figure S1). Over the median follow-up period of 6.5 years (interquartile range, 3.5–9.9), eosinophil count was measured a median of 22 (interquartile range, 7–46) times per patient (4 [interquartile range, 2–6] times a year per patient). The clinical characteristics according to eosinophil quartiles are presented in Table 1. Patients in the highest eosinophil quartile were more likely to be men and to have diabetes mellitus, cardiovascular comorbidities, a lower eGFR, and a higher UPCR. There was a monotonic negative correlation between eosinophil count and eGFR (Fig. 2).

Table 1

Baseline characteristics according to eosinophil counts quartiles
			Eosinophil counts quartiles: range (/µL)
	Total	Missing data	Q1: < 90	Q2: 90–170	Q3: 170–289	Q4: > 289
Characteristics	n = 2,877	n (%)	n = 668	n = 684	n = 736	n = 769	P-value
Age, year	63(14)	0	63(14)	64(14)	63(14)	63(16)	0.99
Male	1,873(65%)	0	356(52%)	435(64%)	509(69%)	573(75%)	< 0.001
Diabetes mellitus	1,173(41%)	0	243(35%)	264(39%)	331(45%)	335(44%)	0.002
BMI, kg/m²	23(4)	84 (3%)	22(4)	23(4)	23(4)	24(4)	< 0.001
SBP, mmHg	131(21)	108 (4%)	130(21)	132(20)	131(20)	131(21)	0.56
Cardiovascular comorbidities	512(18%)	0	78(11%)	118(17%)	136(18%)	180(23%)	< 0.001
Prior history of catheterization	415(14%)	0	71(10%)	89(13%)	106(14%)	149(19%)	< 0.001
Chronic respiratory diseases	43(1%)	0	7(1%)	6(1%)	9(1%)	21(3%)	0.005
ACEIs/ARBs	485(17%)	0	83(12%)	103(15%)	124(17%)	175(23%)	< 0.001
Loop diuretics	299(10%)	0	48(7%)	63(9%)	68(9%)	120(16%)	< 0.001
Thiazide diuretics	101(4%)	0	11(2%)	24(4%)	30(4%)	36(5%)	0.001
MRAs	176(6%)	0	29(4%)	41(6%)	39(5%)	67(9%)	0.001
NSAIDs	63(2%)	0	16(2%)	12(2%)	24(3%)	11(1%)	0.59
PPIs	276(10%)	0	52(8%)	56(8%)	61(8%)	107(14%)	0.001
H₂ blockers	285(10%)	0	68(10%)	72(11%)	73(10%)	72(9%)	0.65
Hemoglobin, g/dL	12.3(2.1)	1 (0%)	12.1(2.0)	12.5(2.0)	12.5(2.1)	12.2(2.1)	0.68
Sodium, mEq/L	140(3)	466 (16%)	140(3)	140(3)	140(3)	139(3)	0.002
Potassium, mEq/L	4.4(0.5)	356 (12%)	4.4(0.6)	4.4(0.5)	4.5(0.5)	4.5(0.5)	< 0.001
Albumin, g/dL	3.8(0.6)	695 (24%)	3.8(0.6)	3.9(0.6)	3.8(0.6)	3.7(0.6)	< 0.001
eGFR, ml/min/1.73m²	37(14)	50 (2.7%)	39(14)	38(14)	37(14)	35(14)	< 0.001
C-reactive protein, mg/dL	0.1[0.0-0.9]	760 (26%)	0.1[0.0-0.7]	0.1[0.0-0.7]	0.1[0.0–1.0]	0.2[0.0-1.1]	< 0.001
UPCR, g/gCre	0.7[0.0-2.4]	1070 (37%)	0.4[0.0-1.8]	0.6[0.0-2.2]	0.8[0.0-2.5]	0.9[0.0-2.7]	< 0.001
White blood cells, ×10³/µL	7.2(3.4)	0	6.7(3.5)	6.4(2.2)	7.3(3.5)	8.3(3.7)	< 0.001
Data presented as mean (standard deviation), number (%), or median [25th‒75th]
Abbreviations: BMI, body mass index; SBP, systolic blood pressure; ACEIs, angiotensin-converting enzyme inhibitors; ARBs, angiotensin II receptor blockers; MRA, mineralocorticoid receptor antagonists; NSAIDs, non-steroidal anti-inflammatory drugs; PPIs, proton pump inhibitors; H₂ blockers, histamine H₂ receptor antagonists; eGFR, estimated glomerular filtration rate; UPCR, urinary protein-to-creatinine ratio

Factors associated with higher eosinophil counts

Eosinophil counts were positively associated with male, BMI, cardiovascular comorbidities, chronic respiratory diseases, ACEIs/ARBs use, and WBC, and negatively correlated with eGFR (Table 2).

Table 2

Multivariable linear regression analysis for the association between log-transformed eosinophil counts and clinical factors
Variables	β [95% CI]	P-value
Age, per 10 years increase	-0.01 [-0.04, 0.02]	0.52
Male	0.34 [0.24, 0.43]	< 0.001
BMI, per 1 kg/m² increase	0.02 [0.01, 0.03]	< 0.001
SBP, per 10 mmHg increase	-0.01 [-0.03, 0.01]	0.37
Diabetes mellitus	-0.02 [-0.11, 0.06]	0.58
Cardiovascular comorbidities	0.22 [0.11, 0.32]	< 0.001
Prior history of catheterization	0.02 [-0.10, 0.14]	0.74
Chronic respiratory diseases	0.40 [0.08, 0.71]	0.01
ACEIs/ARBs use	0.17 [0.07, 0.26]	0.001
Loop diuretics use	0.06 [-0.09, 0.20]	0.46
Thiazide diuretics use	0.07 [-0.10, 0.24]	0.40
MRAs use	0.01 [-0.16, 0.18]	0.91
PPIs use	-0.01 [-0.16, 0.13]	0.86
H₂ blockers use	-0.11 [-0.23, 0.02]	0.09
NSAIDs use	-0.07 [-0.31, 0.16]	0.54
Number of drugs, per 1 drug increase	-0.003 [-0.01, 0.01]	0.62
Hemoglobin, per 1 g/dL increase	0.01 [-0.02, 0.03]	0.58
Albumin, per 1 g/dL increase	0.03 [-0.08, 0.14]	0.60
eGFR, per 10 ml/min/1.73 m² increase	-0.07 [-0.11, -0.04]	< 0.001
C-reactive protein, per 1 mg/dL increase	-0.01 [-0.03, 0.01]	0.46
UPCR, per 1 g/gCre increase	0.01 [-0.01, 0.03]	0.32
White blood cells, per 1000/µL increase	0.05 [0.03, 0.07]	< 0.001
Abbreviations: CI, confidence interval; BMI, body mass index; SBP, systolic blood pressure; ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker; MRA, mineralocorticoid receptor antagonist; PPI, proton pump inhibitor; H₂ blockers, histamine H₂ receptor antagonist; NSAID, non-steroidal anti-inflammatory drug; eGFR, estimated glomerular filtration rate; UPCR, urinary protein-to-creatinine ratio

Baseline and time-average Cox models

RRT was initiated in 433 patients (2.1 per 100 patient-years; 95% CI, 1.9 to 2.3). In the baseline Cox model, there was a dose-dependent association between eosinophil quartiles and the rate of RRT initiation (Table 3). Patients in the highest eosinophil quartile had a 2.12-fold (95% CI: 1.44 to 3.10) higher rate of RRT initiation than those in the lowest quartile. Similarly, higher time-average eosinophil quartiles were associated with an increased rate of RRT initiation (Table 3). The clinical characteristics according to time-average eosinophil quartiles are summarized in Table S1.

Table 3

Association between eosinophil quartiles and clinical outcomes
Outcome: RRT initiation	Baseline Cox model				Time-Average Cox model				Group-based trajectory model
Eosinophil quartiles	Q1: < 90 (n = 688)	Q2: 90–170 (n = 684)	Q3: 170–289 (n = 736)	Q4: > 289 (n = 769)	Q1: ≤ 108 (n = 696)	Q2: 105–183 (n = 669)	Q3: 183–302 (n = 723)	Q4: ≥ 302 (n = 789)	Low (n = 324)	Middle (n = 1,430)	High (n = 1,123)
No. of events	71	79	139	144	62	93	141	134	26	196	208
Incidence rate, 100 p-y (95% CI)	1.4 (1.1–1.7)	1.5 (1.2–1.9)	2.7 (2.3–3.2)	3.0 (2.5–3.5)	1.1 (0.9–1.5)	1.9 (1.5–2.3)	2.9 (2.4–3.4)	2.8 (2.4–3.3)	1.0 (0.7–1.5)	1.9 (1.6–2.1)	3.0 (2.6–3.4)
Hazard Ratio (95% CI)	1.00 (ref)	1.18 (0.77–1.78)	1.77 (1.20–2.61)	2.12 (1.44–3.10)	1.00 (ref)	1.42 (0.95–2.11)	1.98 (1.34–2.94)	2.07 (1.40–3.09)	1.00 (ref)	1.74 (1.06–2.85)	2.30 (1.38–3.84)
P-value	-	0.43	0.004	< 0.001	-	0.09	0.001	< 0.001	-	0.03	0.001
Outcome: CV events and death	Baseline Cox model				Time-Average Cox model				Group-based trajectory model
Eosinophil quartiles	Q1: < 90 (n = 688)	Q2: 90–170 (n = 684)	Q3: 170–289 (n = 736)	Q4: > 289 (n = 769)	Q1: ≤ 108 (n = 696)	Q2: 105–183 (n = 669)	Q3: 183–302 (n = 723)	Q4: ≥ 302 (n = 789)	Low (n = 324)	Middle (n = 1430)	High (n = 1123)
No. of events	87	89	104	122	82	70	91	120	37	161	165
Incidence rate, 100 p-y (95% CI)	1.7 (1.4–2.1)	1.7 (1.4–2.1)	2.0 (1.7–2.5)	2.5 (2.1-3.0)	1.5 (1.2–1.9)	1.4 (1.1–1.8)	1.8 (1.5–2.3)	2.5 (2.1-3.0)	1.4 (1.0–2.0)	1.5 (1.3–2.8)	2.3 (2.0-2.7)
Hazard Ratio (95% CI)	1.00 (ref)	0.94 (0.63–1.40)	1.17 (0.77–1.76)	1.26 (0.85–1.87)	1.00 (ref)	0.77 (0.54–1.09)	1.10 (0.79–1.52)	1.30 (0.95–1.79)	1.00 (ref)	0.97 (0.65–1.44)	1.32 (0.86-2.00)
P-value	-	0.8	0.47	0.25	-	0.15	0.58	0.1	-	0.87	0.19
The models were adjusted for age, sex, diabetes mellitus, body mass index, systolic blood pressure, chronic respiratory diseases, cardiovascular comorbidities, prior history of catheterization, angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, loop diuretics, thiazide diuretics, mineralocorticoid receptor antagonists, proton pump inhibitors, histamine H₂ receptor antagonists, non-steroidal anti-inflammatory drugs, hemoglobin, sodium, potassium, albumin, estimated glomerular filtration rate, C-reactive protein, urinary protein-to-creatinine ratio, and white blood cell count.
Abbreviations: RRT, renal replacement therapy; p-y, person-years; CI, confidence interval; CV, cardiovascular.

Group-based trajectory model

Three distinct trajectories of eosinophil counts were identified: low (n = 324), middle (n = 1,430), and high (n = 1,123) (Fig. 3). The clinical characteristics in each trajectory group are presented in Table S2. In a multivariable Cox model, patients in the high-trajectory group showed a 2.30-fold (95% CI: 1.38 to 3.84) higher rate of RRT initiation than those in the low-trajectory group (Table 3).

Msm

In MSM, high eosinophil counts (≥ 289 /µL) were associated with a 1.83-fold (95% CI: 1.33 to 2.51) higher rate of RRT initiation than normal eosinophil counts (Table 4). There was no significant effect modification by a priori defined baseline covariates, i.e., age, sex, BMI, systolic blood pressure, DM, cardiovascular comorbidities, hemoglobin, albumin, CKD stage, UPCR, and ACEI/ARB use. High eosinophil counts were significantly associated with a higher rate of RRT initiation when using the other definition of high eosinophil count (≥ 500 /µL).

Table 4

Marginal structural models for the association between high eosinophil counts and outcomes
	RRT initiation		CV events and mortality
High eosinophil counts (/µL)	HR (95% CI)	P-value	HR (95% CI)	P-value
≥ 289 (vs.༜289)	1.83 (1.33–2.51)	< 0.001	1.71 (1.30–2.25)	< 0.001
≥ 500 (vs.< 500)	1.41 (1.11–1.80)	0.006	1.35 (1.06–1.73)	0.02
The models were adjusted for the baseline and time-dependent covariates. Baseline covariates included age, sex, body mass index, systolic blood pressure, diabetes mellitus, chronic respiratory diseases, cardiovascular comorbidities, prior history of catheterization, hemoglobin, albumin, estimated glomerular filtration rate, sodium, potassium, C-reactive protein, urinary protein-to-creatinine ratio, loop diuretics, thiazide diuretics, mineralocorticoid receptor antagonists, angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, non-steroidal anti-inflammatory drugs, proton pump inhibitors, and histamine H₂ receptor antagonists. Time-dependent covariates included catheterization performed during follow-up, hemoglobin, albumin, estimated glomerular filtration rate, sodium, potassium, C-reactive protein, urinary protein-to-creatinine ratio, loop diuretics, thiazide diuretics, mineralocorticoid receptor antagonists, angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, nonsteroidal anti-inflammatory drugs, proton pump inhibitors, histamine H₂ receptor antagonists, and corticosteroids.
Abbreviations: HR, hazard ratio; CI, confidence interval; RRT, renal replacement therapy; CV, cardiovascular.
Abbreviations: eGFR, estimated glomerular filtration rate.

Cardiovascular events and mortality

A total of 275 patients developed cardiovascular events (1.4 per 100 patient-years; 95% CI: 1.2 to 1.5) and 165 died (0.8 per 100 patient-years; 95% CI, 0.7 to 0.9). The association of eosinophil count with cardiovascular events and mortality was not significant in any statistical models except for MSM, where higher eosinophil counts were significantly associated with higher rates of these outcomes. (Tables 3 and 4).

Sensitivity analysis

After excluding patients with chronic respiratory diseases or cholesterol embolism, higher eosinophil counts were still associated with a higher rate of RRT initiation (Table S3). A similar result was obtained when including patients followed up for at least 90 days.

Effects of oral carbon adsorbent, AST-120, on eosinophil counts

This post-hoc analysis of the randomized controlled trial included 123 patients with stage 3–4 CKD (70 in the AST-120 group and 53 in the control group). The mean (SD) baseline eosinophil counts were 288 (442) /µL and 454 (1,364) /µL in the AST-120 and control groups, respectively. At 12 months, the mean (SD) eosinophil count was 340 (442) /µL and 456 (1,447) /µL in the AST-120 and control groups, respectively. A linear mixed-effects model for repeated measures showed no significant difference in the eosinophil counts between groups (P = 0.44) (Figure S2).

We found a dose-dependent relationship between eosinophil counts and the risk of RRT initiation among patients with advanced CKD. The results were consistent when longitudinal alterations in eosinophil counts were modeled using the group-based trajectory modeling and MSM. Although future mechanistic studies are required to validate our findings, our data suggest the possible involvement of eosinophils in the progression of CKD.

Studies that examined the association between eosinophils and CKD progression are sparse^9,15. These studies were limited by the small sample size, highly selective patient population, and insufficient adjustment for relevant confounders. More importantly, they did not consider longitudinal alterations in eosinophil counts despite the fact that eosinophils increase as kidney function declines^8,9. We demonstrated a significant association between eosinophil counts and kidney outcomes in the group-based trajectory modeling and MSM that could capture longitudinal changes in eosinophil counts. Furthermore, MSM revealed that this association was independent of time-dependent confounding factors such as eGFR. Thus, although causality cannot be proven, our study provides plausible evidence regarding the link between eosinophils and CKD progression.

Several factors may confound the association between eosinophilia and CKD progression. First, chronic obstructive pulmonary disease and bronchial asthma were strongly associated with high eosinophil counts. They might accelerate CKD progression through hypoxia and inflammation^31–33, and thereby confound the association between eosinophilia and kidney outcomes. Second, drugs such as PPIs, diuretics, and NSAIDs induce both kidney injury and eosinophilia^34–37, and thus could be potential confounders. Finally, arterial catheterization sometimes causes cholesterol embolism, characterized by progressive kidney injury and eosinophilia³⁸. Nevertheless, we did not find an association between catheterization and eosinophil count, most likely because cholesterol embolism is a very rare event. It should be emphasized that we demonstrated a significant association between eosinophil count and the kidney outcome after adjustment for these confounders.

One of the putative mechanisms linking eosinophils to CKD progression is their pro-atherogenic properties⁶. Since intrarenal arteriosclerosis/arteriolosclerosis leads to glomerulosclerosis and interstitial fibrosis and tubule atrophy (IFTA)^39–41, and is associated with worse renal prognosis⁴², eosinophils might affect kidney outcomes by accelerating nephrosclerotic lesions. Future studies are needed to clarify the effect of eosinophils on intrarenal arteriosclerosis/arteriolosclerosis.

Another possible explanation is that eosinophils may be involved in tubulointerstitial injury. Interstitial eosinophilic infiltration is typically observed in drug-induced tubulointerstitial nephritis, eosinophilic granulomatosis with polyangiitis¹², tubulointerstitial nephritis with uveitis⁴³, and idiopathic hypereosinophilic syndrome^10,11. Interestingly, interstitial eosinophilic aggregates are also found in common kidney diseases, and are associated with IFTA, interstitial edema, and eosinophilic tubulitis, suggesting that they may aggravate tubulointerstitial inflammation and fibrosis^13,14. The clinical implications of infiltrating eosinophils in the renal interstitium in terms of renal prognosis requires further detailed investigation.

A novel perspective has been proposed that eosinophils exert a tissue-protective effect. Liu et al. reported a cardioprotective role of interleukin-4 and mEar1 (human ECP ortholog) produced by eosinophils in a mouse model of myocardial infarction, suggesting that an increase in eosinophils in the heart and blood after myocardial infarction represents a compensatory protective response⁴⁴. Similarly, eosinophils prevent transverse aortic constriction-induced cardiac hypertrophy by inhibiting cardiomyocyte apoptosis and cardiac fibrosis⁴⁵. A protective role of eosinophils against liver injury has also been reported⁴⁶. Thus, higher eosinophil counts related to an increased risk of CKD progression may indicate a protective response to a more active disease status. However, these studies evaluated the effect of eosinophil deficiency, which may not be extrapolated to that of eosinophilia. Moreover, whether eosinophils play a protective or harmful role may depend on the target organs and the specific pathological context. The precise role of eosinophils in the progression of CKD requires further investigation.

Although the exact mechanisms remain elusive as to why eosinophils increase in advanced CKD, evidence has implied a role of uremic toxins. Interleukin-5, a master cytokine for eosinophil development, is elevated in nephrectomized mice⁴⁷. Uremic toxins, such as indoxyl sulfate and p-cresol, upregulate intercellular adhesion molecule-1 and vascular cell adhesion molecule-1 in endothelial cells^48,49, which promote eosinophil migration and degranulation^50,51. Thus, these studies support the involvement of uremic toxins in eosinophil proliferation and activation. Nevertheless, we did not find a significant change in eosinophil counts after AST-120 administration in CKD patients. Although this neutral result may be due to insufficient removal of uremic toxins by the drug, it might indicate that the clinical impact of uremic toxins on eosinophils is trivial.

Several factors were related to higher eosinophil counts in our study. Eosinophilia is a well-known side effect of ACEIs, and we confirmed that ACEIs/ARBs users showed higher eosinophil counts. Male had higher eosinophil counts than female. This sex difference has also been observed in healthy individuals⁵². Estrogen inhibits eosinophil production in the bone marrow and induces eosinophil apoptosis⁵³, which might explain higher eosinophil counts in men. The positive association between BMI and eosinophil counts may be explained by this sex difference, while there seems to be a complex relationship between body weight and eosinophils⁵⁴. In our study, the association between eosinophil count and kidney outcome was independent of sex and BMI. In addition, there was no significant effect modification by gender or BMI on this association.

Notably, even modest elevations in eosinophil counts were associated with CKD progression. This is consistent with previous studies showing that a modest increase in eosinophil count, below the definition of eosinophilia (≥ 500/µL)³⁰, is associated with atherosclerotic plaques⁵⁵ and albuminuria⁵⁶. Therefore, physicians should be aware of the clinical implications of subclinical eosinophilia, especially among patients with CKD.

Higher eosinophil counts were also associated with an increased risk of cardiovascular events and mortality in MSM. However, this association was not confirmed in the other statistical models. This discrepancy may be because only MSM could capture time-series changes in eosinophil counts throughout the study period. Conversely, statistical models other than MSM used eosinophil counts only at baseline or during the first 12 months, and thereby may have introduced a misclassification bias. Additionally, most cardiovascular events in our study were hospitalizations for heart failure. Given that eosinophils contribute to atherosclerotic lesions⁵⁷, the predominance of non-atherosclerotic cardiac events would have compromised the sensitivity to detect the impact of eosinophils. Indeed, a previous cohort study reported that eosinophils are not associated with incident heart failure in the general population⁵⁸. Further exploration of the association between eosinophils and cardiac events in patients with CKD is required.

The strengths of this study include large sample size, long-term follow-up period, and abundant data on eosinophil counts repeatedly measured within individuals. Despite the retrospective study design, missing data on eosinophil counts were < 1%. We assessed clinically-meaningful hard outcomes. A variety of covariates potentially related to eosinophilia were adjusted in MSM.

Our study had several limitations. First, the observational study design precludes causal inferences between eosinophil and renal prognosis. Second, some patients had a small number of measurements for eosinophils, which might have reduced the accuracy of the exposure. Third, since we enrolled advanced CKD patients in Japan, the generalizability of our findings to patients with more preserved kidney function or other ethnic groups is unknown. Finally, because we did not have data on tissue eosinophils, we could not directly link blood eosinophil counts with kidney eosinophils. However, previous studies showed a correlation between blood and kidney eosinophil counts^14,15.

In conclusion, higher eosinophil counts were associated with an increased risk of RRT initiation in CKD patients. This association was robust after adjusting for time-dependent confounders. Even a modest increase in eosinophil count was associated with poorer kidney outcomes. Our findings highlight the possible involvement of eosinophils in the pathogenesis of CKD, which has largely been ignored in this field. Further mechanistic studies are required to elucidate the exact role of eosinophils and their potential as therapeutic targets for CKD.

Acknowledgements

Nothing to disclose

Author Contributions

KH and YS conceptualized the study, were responsible for methodology, data curation, formal analysis, validation, and visualization, and wrote the original draft; TO, TK, SK, and YA were responsible for data collection and revision of the draft; JYK, IM, MM and YI were responsible for supervision and revision of the draft.

Data Availability Statement

The data underlying this article cannot be shared publicly due to the privacy of individuals that participated in the study. The data will be shared on reasonable request to the corresponding author.

Funding

Nothing to disclose

Conflict of interest statement

Nothing to disclose

Rosenberg HF, Dyer KD, Foster PS. Eosinophils: changing perspectives in health and disease. Nat Rev Immunol. 2013; 13: 9–22.
Riegger J, Byrne RA, Joner M, et al. Histopathological evaluation of thrombus in patients presenting with stent thrombosis. A multicenter European study: a report of the prevention of late stent thrombosis by an interdisciplinary global European effort consortium. Eur Heart J. 2016; 37: 1538–1549.
Niccoli G, Ferrante G, Cosentino N, et al. Eosinophil cationic protein: A new biomarker of coronary atherosclerosis. Atherosclerosis. 2010; 211: 606–611.
Niccoli G, Montone RA, Sabato V, et al: Role of Allergic Inflammatory Cells in Coronary Artery Disease. Circulation 2018; 138: 1736–1748.
Marx C, Novotny J, Salbeck D, et al. Eosinophil-platelet interactions promote atherosclerosis and stabilize thrombosis with eosinophil extracellular traps. Blood 2019; 134: 1859–1872.
Uderhardt S, Ackermann JA, Fillep T, et al. Enzymatic lipid oxidation by eosinophils propagates coagulation, hemostasis, and thrombotic disease. J Exp Med 2017; 214: 2121–2138.
Sundström J, Söderholm M, Borné Y, et al. Eosinophil Cationic Protein, Carotid Plaque, and Incidence of Stroke. Stroke 2017; 48: 2686–2692.
Ishii R, Fujita S, Kizawa S, et al. Association between absolute blood eosinophil count and CKD stages among cardiac patients. Heart Vessels 2016; 31: 198–205.
Agarwal R, Light RP. Patterns and prognostic value of total and differential leukocyte count in chronic kidney disease. Clin J Am Soc Nephrol 2011; 6: 1393–1399.
Shehwaro N, Langlois AL, Gueutin V, et al. Renal involvement in idiopathic hypereosinophic syndrome. Clin Kidney J 2013; 6: 272–276.
Dong JH, Xu ST, Xu F, et al. Clinical and morphologic spectrum of renal involvement in idiopathic hypereosinophilic syndrome. Clin Exp Nephrol 2021; 25: 270–278.
Durel CA, Sinico RA, Teixeira V, et al. Renal involvement in eosinophilic granulomatosis with polyangiitis (EGPA): a multicentric retrospective study of 63 biopsy-proven cases. Rheumatology (Oxford) 2021; 60: 359–365.
Dai DF, Sasaki K, Lin MY, et al. Interstitial eosinophilic aggregates in diabetic nephropathy: allergy or not? Nephrol Dial Transplant 2015; 30: 1370–1376.
Wu WY, Zhou XJ, Sun PP, et al. Interstitial eosinophilic infiltration in diabetic nephropathy is indicative of poor prognosis, with no therapy benefit from steroid. J Diabetes. 2020; 12: 881–894.
Tariq A, Okamato K, Tariq A, et al. Eosinophilia and risk of incident end stage kidney disease. BMC Nephrol 2020; 21: 14
Oka T, Sakaguchi Y, Hattori K, et al. Mineralocorticoid Receptor Antagonist Use and Hard Renal Outcomes in Real-World Patients With Chronic Kidney Disease. Hypertension. 2022; 79: 679–689
Asahina Y, Sakaguchi Y, Kajimoto S, et al. Association of Time-Updated Anion Gap With Risk of Kidney Failure in Advanced CKD: A Cohort Study. Am J Kidney Dis 2022; 79: 374–382.
Kajimoto S, Sakaguchi Y, Asahina Y, et al. Modulation of the Association of Hypobicarbonatemia and Incident Kidney Failure With Replacement Therapy by Venous pH: A Cohort Study. Am J Kidney Dis. 2021; 77: 35–43.
Matsuo S, Imai E, Horio M, et al. Revised equations for estimated GFR from serum creatinine in Japan. Am J Kidney Dis. 2009; 53: 982–992.
Hayashida K, Murakami G, Matsuda S, et al. History and Profile of Diagnosis Procedure Combination (DPC): Development of a Real Data Collection System for Acute Inpatient Care in Japan. J Epidemiol. 2021; 31: 1–11.
Sakaguchi Y, Hamano T, Obi Y, et al. A Randomized Trial of Magnesium Oxide and Oral Carbon Adsorbent for Coronary Artery Calcification in Predialysis CKD. J Am Soc Nephrol. 2019; 30: 1073–1085.
Nagin DS, Odgers CL. Group-based trajectory modeling in clinical research. Annu Rev Clin Psychol. 2010; 6: 109–138.
Nagin DS, Odgers CL. Group-Based Trajectory Modeling (Nearly) Two Decades Later. J Quant Criminol. 2010; 26: 445–453.
Herle M, Micali N, Abdulkadir M, et al. Identifying typical trajectories in longitudinal data: modelling strategies and interpretations. Eur J Epidemiol. 2020; 35: 205–222.
Mori M, Krumholz HM, Allore HG. Using Latent Class Analysis to Identify Hidden Clinical Phenotypes. JAMA. 2020; 324: 700–701.
Burne RM, Abrahamowicz M. Adjustment for time-dependent unmeasured confounders in marginal structural Cox models using validation sample data. Stat Methods Med Res. 2019; 28: 357–371.
Xie D, Yang W, Jepson C, et al. Statistical Methods for Modeling Time-Updated Exposures in Cohort Studies of Chronic Kidney Disease. Clin J Am Soc Nephrol. 2017; 12: 1892–1899
Lukowsky LR, Mehrotra R, Kheifets L, et al. Comparing mortality of peritoneal and hemodialysis patients in the first 2 years of dialysis therapy: a marginal structural model analysis. Clin J Am Soc Nephrol. 2013; 8: 619–628.
Shinozaki T, Suzuki E. Understanding Marginal Structural Models for Time-Varying Exposures: Pitfalls and Tips. J Epidemiol. 2020; 30: 377–389.
Shomali W, Gotlib J. World Health Organization-defined eosinophilic disorders: 2022 update on diagnosis, risk stratification, and management. Am J Hematol. 2022; 97: 129–148.
Lai CC, Wu CH, Wang YH, et al. The association between COPD and outcomes of patients with advanced chronic kidney disease. Int J Chron Obstruct Pulmon Dis. 2018; 13: 2899–2905.
Sumida K, Kwak L, Grams ME, et al. Lung Function and Incident Kidney Disease: The Atherosclerosis Risk in Communities (ARIC) Study. Am J Kidney Dis. 2017; 70: 675–685.
Huang HL, Ho SY, Li CH, et al. Bronchial asthma is associated with increased risk of chronic kidney disease. BMC Pulm Med. 2014; 14: 80.
Xie Y, Bowe B, Li T, et al. Proton Pump Inhibitors and Risk of Incident CKD and Progression to ESRD. J Am Soc Nephrol. 2016; 27: 3153–3163.
Zhan M, Doerfler RM, Xie D, et al. Association of Opioids and Nonsteroidal Anti-inflammatory Drugs With Outcomes in CKD: Findings From the CRIC (Chronic Renal Insufficiency Cohort) Study. Am J Kidney Dis. 2020; 76: 184–193.
Cholin L, Ashour T, Mehdi A, et al. Proton-pump inhibitor vs. H2-receptor blocker use and overall risk of CKD progression. BMC Nephrol. 2021; 22: 264.
Perazella MA, Rosner MH. Drug-Induced Acute Kidney Injury. Clin J Am Soc Nephrol. 2022 Mar 10: CJN.11290821. doi: 10.2215/CJN.11290821. Epub ahead of print.
Scolari F, Ravani P. Atheroembolic renal disease. Lancet. 2010; 375: 1650–1660.
Hill GS, Heudes D, Jacquot C, et al. Morphometric evidence for impairment of renal autoregulation in advanced essential hypertension. Kidney Int. 2006; 69: 823–831.
Hommos MS, Glassock RJ, Rule AD. Structural and Functional Changes in Human Kidneys with Healthy Aging. J Am Soc Nephrol. 2017; 28: 2838–2844.
Denic A, Rule AD, Glassock RJ. Healthy and unhealthy aging on kidney structure and function: human studies. Curr Opin Nephrol Hypertens. 2022; 31: 228–234.
Srivastava A, Palsson R, Kaze AD, et al. The Prognostic Value of Histopathologic Lesions in Native Kidney Biopsy Specimens: Results from the Boston Kidney Biopsy Cohort Study. J Am Soc Nephrol. 2018; 29: 2213–2224.
Yanagihara T, Kitamura H, Aki K, et al. Serial renal biopsies in three girls with tubulointerstitial nephritis and uveitis syndrome. Pediatr Nephrol. 2009; 24: 1159–1164.
Liu J, Yang C, Liu T, et al. Eosinophils improve cardiac function after myocardial infarction. Nat Commun. 2020; 11: 6396.
Yang C, Li J, Deng Z, et al. Eosinophils protect pressure overload- and β-adrenoreceptor agonist-induced cardiac hypertrophy. Cardiovasc Res. 2022 Apr 8:cvac060 doi: 10.1093/cvr/cvac060. Online ahead of print.
Wang Y, Yang Y, Wang M, et al. Eosinophils attenuate hepatic ischemia-reperfusion injury in mice through ST2-dependent IL-13 production. Sci Transl Med. 2021; 13: eabb6576.
Andres-Hernando A, Dursun B, Altmann C, et al. Cytokine production increases and cytokine clearance decreases in mice with bilateral nephrectomy. Nephrol Dial Transplant. 2012; 27: 4339–4347.
Jing YJ, Ni JW, Ding FH, et al. p-Cresyl sulfate is associated with carotid arteriosclerosis in hemodialysis patients and promotes atherogenesis in apoE-/- mice. Kidney Int. 2016; 89: 439–449.
Tumur Z, Shimizu H, Enomoto A, et al. Indoxyl sulfate upregulates expression of ICAM-1 and MCP-1 by oxidative stress-induced NF-kappaB activation. Am J Nephrol. 2010; 31: 435–441.
Chihara J, Yamamoto T, Kayaba H, et al. Degranulation of eosinophils mediated by intercellular adhesion molecule-1 and its ligands is involved in adhesion molecule expression on endothelial cells-selective induction of VCAM-1. J Allergy Clin Immunol. 1999; 103: S452-456.
Ueki S, Kihara J, Kato H, et al. Soluble vascular cell adhesion molecule-1 induces human eosinophil migration. Allergy. 2009; 64: 718–724.
Takami A, Watanabe S, Yamamoto Y, et al. Reference intervals of white blood cell parameters for healthy adults in japan. Int J Lab Hematol. 2021; 43: 948–958.
Douin-Echinard V, Calippe B, Billon-Galès A, et al. Estradiol administration controls eosinophilia through estrogen receptor-alpha activation during acute peritoneal inflammation. J Leukoc Biol 2021; 90: 145–154.
Calco GN, Fryer AD, Nie Z. Unraveling the connection between eosinophils and obesity. J Leukoc Biol. 2020; 108: 123–128.
Kitano T, Nezu T, Shiromoto T, et al. Association Between Absolute Eosinophil Count and Complex Aortic Arch Plaque in Patients With Acute Ischemic Stroke. Stroke 2017; 48: 1074–1076.
Fukui M, Tanaka M, Hamaguchi M, et al. Eosinophil count is positively correlated with albumin excretion rate in men with type 2 diabetes. Clin J Am Soc Nephrol. 2009; 4: 1761–1765.
Verdoia M, Schaffer A, Cassetti E, et al. Absolute eosinophils count and the extent of coronary artery disease: a single centre cohort study. J Thromb Thrombolysis. 2015; 39: 459–466.
Shah AD, Denaxas S, Nicholas O, et al. Low eosinophil and low lymphocyte counts and the incidence of 12 cardiovascular diseases: a CALIBER cohort study. Open Heart. 2016; 3: e000477.

No competing interests reported.

EOSsupplementalSR.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Association between Time-Updated Eosinophil Counts and Progression of CKD

Status:

Version 1

Abstract

Figures

Introduction

Methods

Ethical considerations

Study design and participants

Data collection

Exposure

Study outcomes

Statistical analyses

4) MSM

Results

Study population

Factors associated with higher eosinophil counts

Baseline and time-average Cox models

Group-based trajectory model

Msm

Cardiovascular events and mortality

Sensitivity analysis

Effects of oral carbon adsorbent, AST-120, on eosinophil counts

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1