Our case-control study was nested within a longitudinal prospective study of patients who underwent native or transplant kidney biopsies at Johns Hopkins Hospital from 2004-2018 and at Medical University of South Carolina from 2017-2018. Eligible cases (n=24) included those with confirmed diagnosis of incident ESKD after enrollment. Controls (n=154) consisted of patients who underwent kidney biopsy from 2004-2018 who did not progress to ESKD after enrollment. This study was approved by the institutional review boards at Johns Hopkins University and the Medical University of South Carolina.
From the 616 eligible patients, 178 patients were identified through the registry of kidney biopsies excluding patients with kidney failure requiring dialysis (n=19), dialysis-dependent chronic kidney disease (CKD) stage V patients with biopsy confirmed ESKD at the time of enrollment (n=7), and those with missing biopsy or hematology data (n=411) (Figure 1). Patients were 18 years and older who underwent kidney biopsy for confirmation of their kidney diagnosis. We included patients who met the Kidney Disease Improving Global Outcomes (KDIGO) criteria for acute kidney injury (AKI) (12) with and without abnormal clinical findings of hematuria, pyuria or proteinuria. For each case, two controls were selected and matched on sex, age and duration of follow-up time since biopsy, so that 24 cases were matched to 48 controls. While we could not match on diagnoses given limited sample size, we did demonstrate the spectrum of etiologies in those with and without peripheral eosinophilia.
Incident peripheral eosinophilia was assessed at enrollment using hematology reports at the time of kidney biopsy and analyzed as a binary variable. Eosinophil percentage was used instead of absolute eosinophil count for clinical relevance and as previously described (6), no eosinophilia as <4% of peripheral blood leukocytes (WBC) and peripheral eosinophilia as >4% of WBC. We further categorized the severity of peripheral eosinophilia as 4%-10%, and ≥10%.
The primary outcome was defined as incident progression to ESKD, classified by estimated glomerular filtration rate (eGFR) <5 mL/min/1.73m2, an International Classification of Diseases Ninth/Tenth (ICD-9/10) revision code for a kidney disease-related hospitalization or death, per nephrologists’ diagnosis for patients requiring renal replacement therapy, and/or repeat kidney pathology suggesting extensive chronic, irreversible changes in the biopsy specimen.
All socio-demographical and clinical information were obtained at enrollment using Epic electronic medical records (EMR). Past medical history (hx) of atopic illness, filarial disease, asthma, and kidney transplantation were defined as binary variables. Similarly, history of hypertension (HTN), diabetes and medication use of proton pump inhibitor (PPI) were defined as binary variables. eGFR was obtained as patients’ “normal” eGFR, as measured by the CKD-Epi equation (13), prior to study entry and assessed by combination of previous medical records and laboratory chemistries. Other baseline variables measured as continuous variables at the time of enrollment included serum creatinine (Cr), serum Immunoglobulin E (IgE) levels, complements (C3, C4), and proteinuria.
The indication for kidney biopsy was characterized by four categories as per the nephrologists’ standard orders in ICD-9/10: AKI, CKD, AKI on CKD (AOCKD), or nephrotic syndrome. We characterized urine proteinuria based on the urine-protein-creatinine ratio (UPCR) and on urinalyses, as trace, +1, + 2, + 3, or +4 as reported by standard laboratory processing. Urinalyses was also assessed for the presence of pyuria, urine eosinophils and hematuria.
Tissue from kidney biopsy specimens was processed in the pathology departments using standard methods for light, immunofluorescence, and electron microscopy. The exact locations of the eosinophils were captured on the tissue specimens, and other inflammatory markers (e.g. lymphocytes and plasma cells) were documented using individual biopsy reports. The number of eosinophils were documented as per hpf and refers to the number of eosinophils per hpf in the renal interstitium. For the purpose of this study, pathologists categorized kidney tissue eosinophils as: “rare” if <5 per hpf, “few” if 5-10 per hpf, “many” if >10 per hpf, and “numerous” if >25 per hpf, as previously documented (14). Pathologists independently evaluated biopsy slides to establish primary and secondary diagnoses, including acute tubular injury, chronic changes, or other kidney biopsy abnormalities such as IN.
Analysis of variance (ANOVA) and X2 t-test were used for statistical analysis on demographics and clinical characteristics. Results were reported as proportions for binary or categorical variables and mean for continuous variables. Pearson’s correlation was used to evaluate possible correlations amongst all the variables, but since no strong correlations existed, none of the variables were eliminated. Sex and age are known, strong confounders in ESKD and therefore matched upon (15). Race was not matched in order to evaluate the independent effect of it on our outcome. Every case (n=24) was matched to two controls, of the same sex, age and follow-up time (months) from biopsy.
Matched odds ratios (OR) for ESKD, calculated, as an estimate of the hazard ratio, and corresponding 95% confidence (CIs) were estimated using conditional logistic regression. Both univariate and multivariate models were used to show associations. A final multivariate model was created through stepwise elimination of variables of interest from univariate analysis while biologically relevant variables were retained, with the intent of using one variable for every 10 outcomes to avoid overfitting of the model. Additional analyses were conducted for baseline clinical demographics and statistical significance was determined with the use of likelihood-ratio test. UPCR, eGFR, HTN were included in multivariate models because they are strong predictors for ESKD (16, 17). All analyses were performed using Stata version 15.1 (StataCorp, College Station, TX) (18).
Sensitivity analyses were performed using peripheral eosinophilia as a continuous variable. Univariate and multivariate analyses showed significant associations with higher degree of peripheral eosinophilia and ESKD. UPCR was also modeled as a binary variable and per KDIGO guidelines, normal UPCR defined as <0.5mg/dl in 24-hour urine (19). The area under the ROC (AUC) was calculated to assess the ability of peripheral eosinophilia to discriminate between ESKD progressors and non-progressors.