Racial, Ethnic, and Neighborhood Disparities in Acute Kidney Injury in Pediatric Acute Leukemia Authors: Wendy C Bravo, Yimei Li, Tamara P Miller, Kelly Getz, Lusha Cao, Edward Krause, Mark Ramos, Judy Lee, Maria Monica Gramatges, Karen R Rabin, Michael Scheurer, Richard Aplenc, Michelle Denburg Published: 2026 ABSTRACT ABSTRACTIntroductionWe sought to identify racial/ethnic and neighborhood disparities in acute kidney injury (AKI) in children with acute lymphoblastic or myeloid leukemia (ALL, AML).MethodsElectronic health record data were retrospectively collected from patients treated for de novo pediatric ALL or AML in the multicenter Leukemia Electronic Abstraction of Records Network. AKI was defined by modified Kidney Disease Improving Global Outcomes criteria. Neighborhood‐level resources were defined by the Child Opportunity Level (COL). Prevalence ratios (PRs) for AKI at diagnosis by race/ethnicity and COL were calculated using modified Poisson regression. Cox regression was used to calculate hazard ratios (HRs) of first AKI after initiation of chemotherapy (i.e., “during therapy”).ResultsOf 952 ALL patients, Non‐Hispanic Black (NHB) ALL patients had significantly increased prevalence of AKI at leukemia diagnosis compared to Non‐Hispanic White (NHW) patients in the adjusted model (PR 1.50); there were no significant differences by COL. During ALL therapy overall, NHB patients did not have any significant differences in hazard of AKI by race/ethnicity or COL in adjusted models. Among 168 AML patients, there were no significant differences in AKI at diagnosis by race/ethnicity or COL in AML. During AML therapy overall, NHB patients had significantly higher risk of AKI in both unadjusted (HR 1.95) and adjusted (HR 2.68) models, but no differences by COL.ConclusionNHB patients with ALL had increased risk of AKI at the time of leukemia diagnosis, and NHB patients with AML had significantly increased risk of AKI during therapy. Further studies to identify drivers and interventions that mitigate AKI risk are warranted.Keywords: acute kidney injury, acute lymphoblastic leukemia, acute myeloid leukemia, disparities, electronic health record, pediatric Introduction Leukemia, comprising primarily of acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML), is the most common pediatric cancer diagnosis in the United States (US) [1]. Prior literature shows that children with ALL and AML experience acute kidney injury (AKI) during therapy at rates ranging from 18% to 100% in ALL and 16% to 88% in AML [2, 3, 4, 5, 6, 7]. AKI may be associated with a lower chance of complete remission and higher mortality in patients with leukemia [5, 8, 9]. AKI also increases the risk for developing chronic kidney disease (CKD) [10], which in turn decreases life expectancy [11]. Clinical risk factors for AKI identified in previous studies of pediatric patients with acute leukemia include older age, female sex, severe sepsis, prolonged exposure to nephrotoxic antimicrobials/chemotherapies, and lower estimated glomerular filtration rate at the time of leukemia diagnosis [2, 4, 5, 9]. However, only two studies to our knowledge have examined racial and ethnic disparities in AKI among pediatric patients with acute leukemia [9, 12]. One retrospective cohort study using administrative data from the Pediatric Health Information System (PHIS) found that Black patients experienced a 67% increased hazard of AKI during the first year of AML therapy [9]. A single‐center retrospective cohort study of pediatric patients treated for ALL at Texas Children's Hospital (TCH) showed that Hispanic ethnicity was associated with delayed high‐dose methotrexate clearance, and all patients with severe methotrexate‐related nephrotoxicity were Hispanic [12]. Race and ethnicity are social constructs that may be a proxy for culture, environment, systemic racism, or other conditions that affect health outcomes. Prior analyses in both children and adults have shown that adjusting for socioeconomic status attenuates racial and ethnic disparities in AKI, highlighting the importance of identifying and addressing modifiable socioeconomic risk factors to improve AKI outcomes [13, 14]. However, there are no comprehensive, generalizable studies investigating modifiable risk factors for AKI disparities in pediatric patients with leukemia. The Child Opportunity Index (COI) is a publicly available index of 29 weighted educational, health environment, and socioeconomic domain indicators at the census tract level used to assess neighborhood‐level resources impacting childhood health [15]. This study utilized the Leukemia Electronic Abstraction of Records Network (LEARN), a multicenter cohort of patients treated for acute leukemia that enables efficient collection of very granular data from the electronic health record (EHR) using the ExtractEHR+ Toolkit [16, 17, 18]. This streamlined cohort data population process includes automated capture of residential address, which can then be mapped to the COI to assess environmental factors. Combining automated extraction of laboratory and demographic data, we aimed to describe disparities in the prevalence of AKI at the time of leukemia diagnosis and risk for development of AKI during leukemia therapy. Methods Study Design, Patient Population, and Data Source This was a retrospective cohort study in LEARN, which includes manual abstraction and extracted EHR data from patients aged 1–21 years with de novo ALL or AML diagnosed or treated with at least one chemotherapy course at the Children's Hospital of Philadelphia (CHOP) from 2011 to 2015, Children's Healthcare of Atlanta (CHOA) from 2010 to 2018, and Texas Children's Hospital (TCH) from 2011 to 2014 [16, 17, 19]. Patients with therapy‐related AML, Down Syndrome AML, and infant ALL were excluded due to significant differences in treatment and clinical course as compared to the rest of our cohort. Data from incomplete therapy courses were excluded. This study was approved by the Institutional Review Board at each institution. Data were collected from the EHR using the ExtractEHR+ toolkit [18]. Data for this study included demographics (age, sex, race, ethnicity), residential addresses, start and end dates of chemotherapy courses, clinical trial enrollment, laboratory results, medication administration records, and vital signs. Residential addresses at the time of LEARN cohort entry were geocoded using ArcGIS, a process that enables census tract‐level assessment of neighborhood socioeconomic and environmental factors. Geocoded data were cross‐referenced with the publicly available COI 2.02015 database because the majority of our patients were assessed within 5 years of 2015 [15]. Our dataset did not include past medical history at the time of leukemia diagnosis and thus we were unable to determine if patients had baseline CKD; however, the vast majority of pediatric patients with acute leukemia are previously healthy and the rate of CKD in the general pediatric population in the US is estimated at < 2% [20]. Data were collected from the initial leukemia diagnosis admission or first chemotherapy course in LEARN through upfront chemotherapy completion, death, relapse, conditioning for stem cell transplant, or transfer of care, whichever came first. Exposures The primary exposure was race/ethnicity categorized as Hispanic, Non‐Hispanic Black (NHB), Non‐Hispanic White (NHW), and Other. Race/ethnicity data was determined based on the Demographics data available in the EHR. Due to low numbers of patients, other census‐derived races (e.g., Asian, American Indian) were categorized in the “Other” group. The secondary exposure was neighborhood‐level resources measured by COI as a continuous variable, with lower values indicating less opportunity/resources and higher values indicating greater opportunity/resources. The Child Opportunity Level (COL) is a categorical variable derived from the COI, ranging from Very Low to Very High. Outcomes AKI was defined as an increase in serum creatinine by 0.3 mg/dL from baseline or ≥ 1.5 times baseline and the serum creatinine must be > 0.5 mg/dL, or need for continuous kidney replacement therapy (CKRT) as determined by the receipt of CKRT dialysate in the medication administration record. This is referred to as the modified Kidney Disease Improving Global Outcomes (mKDIGO) definition throughout this manuscript [21, 22]. This definition was chosen based on prior analyses of AKI data in LEARN [19]. AKI stages were determined using KDIGO AKI guidelines [23]. Baseline creatinine at leukemia diagnosis was the creatinine at initial presentation if within the laboratory's reference range. If above the reference range, the lowest creatinine in the 90 days prior was utilized. If these values were unavailable, validated age‐ and sex‐specific normative values were used [24]. Baseline creatinine during therapy was the first creatinine value available within each chemotherapy course. AKI at diagnosis was determined using creatinine values from initial presentation to the first day of induction chemotherapy. AKI during therapy was based on creatinine values within each chemotherapy course. Additionally, the first AKI after initiation of chemotherapy overall was determined. Covariates Baseline covariates at the time of cohort entry included age, sex, leukemia risk group at the time of diagnosis, and body mass index (BMI). During therapy, covariates included baseline age, sex, baseline leukemia risk group, baseline BMI, clinical trial enrollment (considered “yes” if enrolled in a clinical trial at any point during therapy), prior AKI (if the patient had an AKI at diagnosis or during a prior chemotherapy course), and study site. BMI was categorized for those ≥ 2 years old with available height and weight data according to Centers for Disease Control (CDC) 2000 criteria [25]. As patients < 2 years old are recommended to use weight‐for‐length percentiles as opposed to BMI per CDC guidance, weight‐for‐length percentiles were calculated utilizing World Health Organization 2006 for those < 2 years old. Patients < 2 years old were considered to be in the “underweight” BMI category if their weight‐for‐length percentile was less than the second percentile, “normal” BMI category if their weight‐for‐length percentile ranged from the 2nd to 98th percentile, and in the “obese” BMI category if their weight‐for‐length percentile was > 98th percentile [26]. Course‐specific leukemia risk group was not available in the data. Estimated glomerular filtration rate (eGFR) was calculated using creatinine‐based Chronic Kidney Disease in Children U25 equation [27]. Statistical Analysis Separate analyses were conducted for patients with ALL and AML. Baseline characteristics were summarized overall and stratified by those who ever had AKI versus those who never had AKI during follow‐up (includes at leukemia diagnosis and during therapy using available EHR data). Stratification by AKI stage is reported in Tables S1 and S2. Continuous variables were reported as median (interquartile range [IQR]) and compared using the Wilcoxon rank‐sum test. Categorical variables were compared using chi‐squared or Fisher exact tests. Modified Poisson regression was used to estimate univariable and multivariable prevalence ratios (PRs), 95% confidence intervals (CIs), and p‐values for AKI at diagnosis by covariates, race and ethnicity, COI, and COL. Multivariable models for AKI at leukemia diagnosis adjusted for baseline age, sex, baseline BMI category, and baseline leukemia risk group a priori. For first AKI during therapy, time is calculated from initiation of first chemotherapy course available in LEARN to occurrence of first AKI or censored at last follow up. Data were censored at the end of available chemotherapy course data, upfront chemotherapy completion, death, relapse, conditioning for stem cell transplant, or transfer of care, whichever came first. Kaplan–Meier curves were produced for first AKI during therapy by race/ethnicity and COL and compared with log‐rank p‐values. Cox proportional hazards models were used to calculate univariable and multivariable hazard ratios (HRs), 95% CIs, and p‐values for incident AKI during therapy overall and for incident AKI during each chemotherapy course by covariates, race and ethnicity, COI, and COL. The multivariable model for AKI during therapy adjusted for baseline age, sex, baseline BMI category, prior AKI, baseline leukemia risk group, clinical trial enrollment, and study site a priori. For time to first AKI during each chemotherapy course, time is calculated from the start of the course to the occurrence of AKI or censored at the last day of the course. As a sensitivity analysis, we repeated these analyses (with the exception of course‐specific HRs) without the AKI criterion of serum creatinine > 0.5 mg/dL (referred to as the KDIGO definition in this manuscript). Since we included patients who may not have completed all chemotherapy at a LEARN site, we also performed a sensitivity analysis of adjusted HRs of first mKDIGO AKI during therapy including only patients with complete data through upfront chemotherapy completion. p‐value < 0.05 was considered statistically significant for all analyses. Analyses were performed with Stata version 17.0. Results Baseline Characteristics There were 952 patients with ALL, of whom 482 (50.6%) had AKI at least once from time of diagnosis through end of follow‐up. Nearly one‐third of patients with ALL experienced Stage 2 or 3 AKI during follow‐up (n = 296, 31.0%). Compared to patients who never had an AKI at presentation or during therapy, patients who had an AKI were older (median 9.3 vs. 4.0 years, p < 0.001), more likely to be male (60.0% vs. 50.9%, p < 0.001), have a higher BMI percentile (median 68.2 vs. 49.8 percentile, p < 0.001), in overweight and obese categories (p < 0.001), have a lower baseline eGFR (median 91.4 vs. 104.8 mL/min/1.73 m2, p < 0.001), be in the intermediate/high risk baseline leukemia group (n = 276, 57.4% vs. n = 138, 29.7%, p < 0.001), and were less likely to be in a clinical trial (n = 292, 60.7% vs. 311, 66.9%, p = 0.048). There was a higher proportion of NHB patients and a lower proportion of Hispanic patients in the Ever AKI group (NHB: Ever AKI 19.1% vs. Never AKI 13.2%; Hispanic: Ever AKI 26.8% vs. Never AKI 31.5%), which neared statistical significance (p = 0.07). Patients with AKI had a lower median COI (43 vs. 51, p = 0.003), with lower education and SES domains (Table 1, Table S1). TABLE 1Baseline characteristics.ALLAMLOverall a (n = 952)Never AKI (n = 470)Ever AKI (n = 482) p b Overall a (n = 168)Never AKI (n = 91)Ever AKI (n = 77) p b Age (years)5.8 (3.4–11.3)4.0 (2.7–6.2)9.3 (5.3–13.4)< 0.0018.1 (2.2–14.3)3.7 (1.3–8.7)13.0 (8.4–16.2)< 0.001Age category< 0.001< 0.001< 262 (6.5)49 (10.4)13 (2.7)36 (21.4)32 (35.2)4 (5.2)2–4 years346 (36.3)248 (52.8)98 (20.3)27 (16.1)23 (25.3)4 (5.2)5–9 years261 (27.4)112 (23.8)149 (30.9)31 (18.5)16 (17.6)15 (19.5)10+ years283 (29.7)61 (13.0)222 (46.1)74 (44.0)20 (22.0)54 (70.1)Male528 (55.5)239 (50.9)289 (60.0)0.00580 (47.6)45 (49.5)35 (45.5)0.61BMI (kg/m2) c 16.8 (15.1–19.6)15.8 (14.7–17.7)17.8 (15.5–22.4)< 0.00118.8 (15.7–21.6)17.3 (15.1–21.1)19.6 (17.0–23.7)0.002BMI percentile c 60.1 (25.8–87.9)49.8 (19.8–82.4)68.2 (34.6–91.6)< 0.00164.8 (27.9–87.0)58.7 (15.7–82.5)70.1 (32.4–90.0)0.09Weight‐for‐length percentile d 65.4 (40.5–95.4)65.9 (40.5–94.7)61.8 (50.4–97.5)0.6264.9 (30.5–88.4)64.9 (31.0–84.6)61.2 (20.6–94.8)0.90BMI category< 0.0010.08Underweight82 (8.6)46 (9.8)36 (7.5)12 (7.1)9 (9.9)3 (3.9)Normal597 (62.7)324 (68.9)273 (56.6)115 (68.5)66 (72.5)49 (63.6)Overweight131 (13.8)49 (10.4)82 (17.0)15 (8.9)6 (6.6)9 (11.7)Obese131 (13.8)45 (9.6)86 (17.8)23 (13.7)10 (11.0)13 (16.9)Missing11 (1.2)6 (1.3)5 (1.0)3 (1.8)0 (0)3 (3.9)Baseline eGFR (mL/min/1.73 m2)97.6 (83.0–115.2)104.8 (89.8–126.1)91.4 (75.9–107.8)< 0.001102.8 (83.4–120.0)108.6 (92.4–125.5)93.3 (75.5–111.3)< 0.001Trisomy 2123 (2.4)9 (1.9)14 (2.9)0.32———Baseline leukemia risk group< 0.0010.67Standard risk512 (53.8)318 (67.7)194 (40.2)91 (54.2)48 (52.7)43 (55.8)Intermediate/high risk427 (44.9)144 (30.6)283 (58.7)74 (44.0)42 (46.2)32 (41.6)Missing13 (1.4)8 (1.7)5 (1.0)3 (1.8)1 (1.1)2 (2.6)Enrolled in clinical trial603 (63.7)311 (66.9)292 (60.7)0.04885 (50.9)49 (54.4)36 (46.8)0.32Race and ethnicity0.070.14Non‐Hispanic White414 (43.5)205 (43.6)209 (43.4)68 (40.5)39 (42.9)29 (37.7)Non‐Hispanic Black154 (16.2)62 (13.2)92 (19.1)48 (28.6)20 (22.0)28 (36.4)Hispanic277 (29.1)148 (31.5)129 (26.8)28 (16.7)19 (20.9)9 (11.7)Other107 (11.2)55 (11.7)52 (10.8)24 (14.3)13 (14.3)11 (14.3)Neighborhood COI47 (22–76)51 (27–80)43 (19–74)0.00344.5 (19–73)44 (19–74)50 (21–73)0.68Education56 (28–82)60 (33–84)52 (24–82)0.01953 (23–79.5)52 (20–73)59 (25–83)0.27HE32 (14–58)35 (14–61)29 (13–56)0.08235 (14–56.5)31 (13–53)42 (14–59)0.44SES49 (25–74.5)51 (28–77)44 (21–70)0.00146.5 (23–70)43 (21–75)47 (23–67)0.95Neighborhood COL0.0270.73Very low227 (23.8)95 (20.2)132 (27.4)43 (25.6)24 (26.4)19 (24.7)Low194 (20.4)92 (19.6)102 (21.2)36 (21.4)21 (23.1)15 (19.5)Moderate165 (17.3)86 (18.3)79 (16.4)25 (14.9)15 (16.5)10 (13.0)High167 (17.5)83 (17.7)84 (17.4)34 (20.2)15 (16.5)19 (24.7)Very high199 (20.9)114 (24.3)85 (17.6)30 (17.9)16 (17.6)14 (18.2) Note: Data are shown as n (%) or median (interquartile range).Abbreviations: AKI, acute kidney injury; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; COI, Child Opportunity Index; COL, Child Opportunity Level; eGFR, estimated glomerular filtration rate; HE, health environment; SES, socioeconomic status. a “Overall” includes creatinine values from the time of leukemia diagnosis (i.e., “at presentation”) through the end of therapy. AKI is defined as: creatinine is ≥ 1.5 times or ≥ 0.3 mg/dL higher than the baseline creatinine and the absolute creatinine value is > 0.5 mg/dL or patient is on CKRT. b Compares Never AKI to Ever AKI. c Calculated among patients ≥ 2 years old. d Calculated among patients < 2 years old. Baseline characteristics. Note: Data are shown as n (%) or median (interquartile range). Abbreviations: AKI, acute kidney injury; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; COI, Child Opportunity Index; COL, Child Opportunity Level; eGFR, estimated glomerular filtration rate; HE, health environment; SES, socioeconomic status. “Overall” includes creatinine values from the time of leukemia diagnosis (i.e., “at presentation”) through the end of therapy. AKI is defined as: creatinine is ≥ 1.5 times or ≥ 0.3 mg/dL higher than the baseline creatinine and the absolute creatinine value is > 0.5 mg/dL or patient is on CKRT. Compares Never AKI to Ever AKI. Calculated among patients ≥ 2 years old. Calculated among patients < 2 years old. There were 168 patients with AML, of whom 77 (45.8%) had AKI at least once. Fifty‐one patients (30.3%) experienced a Stage 2/3 AKI. Patients with AKI were significantly older (median 13.0 vs. 3.7 years, p < 0.001) and had a lower baseline eGFR (median 93.3 vs. 108.6 mL/min/1.73 m2, p < 0.001). There were no statistically significant differences by race and ethnicity, overall COI, or COL comparing ever AKI to never AKI. However, there was again a higher proportion of NHB patients and a lower proportion of Hispanic patients in the Ever AKI group (NHB: Ever AKI 36.4% vs. Never AKI 22.0%; Hispanic: Ever AKI 11.7% vs. Never AKI 20.9%; p = 0.14) (Table 1, Table S2). Baseline characteristics for patients analyzed in each chemotherapy course are presented in Tables S3 and S4. AKI at Leukemia Diagnosis In ALL, 120 out of 879 (13.6%) patients diagnosed at a LEARN site had AKI at time of leukemia diagnosis. Univariable models showed that older age, overweight and obese BMI categories, intermediate/high leukemia risk group, and NHB race/ethnicity were factors associated with significantly increased risk of AKI at diagnosis. COI and COL were not significantly associated with AKI risk at diagnosis. In the multivariable model, NHB patients had a 50% increased risk of AKI at diagnosis compared to NHW patients, which was statistically significant (PR 1.50 [95% CI 1.01–2.22], p = 0.04). Multivariable models of COI and COL were not associated with increased prevalence risk of AKI. In AML, 26 out of 159 (16.4%) patients diagnosed at a LEARN site had AKI at the time of leukemia diagnosis. Older age was associated with significantly increased risk of AKI at diagnosis; no other covariates were statistically significant. Race/ethnicity, COI, and COL were not significant in univariable or multivariable models (Table 2). TABLE 2Univariable and multivariable prevalence ratios of AKI at diagnosis.ALL PR (95% CI) p AML PR (95% CI) p Univariable modelsBaseline age (year)1.11 (1.08–1.14)< 0.0011.10 (1.04–1.17)0.001Baseline age category: Reference Group—< 2 years2–4 years3.93 (0.54–28.71)0.181.48 (0.10–22.65)0.785–9 years8.81 (1.23–63.05)0.034.53 (0.54–38.62)0.1710+ years13.86 (1.86–98.03)0.0089.44 (1.32–67.91)0.03Sex—Reference group femaleMale sex1.17 (0.83–1.64)0.370.64 (0.30–1.35)0.24BMI category: Reference group—NormalUnderweight1.14 (0.62–2.14)0.67——Overweight1.74 (1.13–2.66)0.012.12 (0.91–4.91)0.08Obese1.66 (1.07–2.57)0.021.16 (0.43–3.11)0.78Missing0.8 (0.12–5.28)0.82——Baseline leukemia risk group: Reference group—Standard risk ALL or low‐risk AML as applicableIntermediate or High1.95 (1.38–2.76)< 0.0010.56 (0.26–1.23)0.15Missing2.35 (0.67–8.22)0.181.67 (0.32–8.77)0.55Race/Ethnicity: Reference group—Non‐Hispanic WhiteNon‐Hispanic Black1.72 (1.14–2.58)0.0091.14 (0.51–2.53)0.75Hispanic0.73 (0.46–1.16)0.190.65 (0.20–2.14)0.48Other1.32 (0.79–2.22)0.290.79 (0.24–2.59)0.7Neighborhood COI1.00 (0.99–1.00)0.211.00 (0.99–1.01)0.79Neighborhood COL: Reference group—Very low COLLow0.86 (0.53–1.41)0.550.95 (0.32–2.86)0.93Moderate1.12 (0.70–1.80)0.631.86 (0.68–5.11)0.23High0.85 (0.51–1.43)0.551.32 (0.47–3.72)0.6Very high0.66 (0.38–1.13)0.130.71 (0.19–2.61)0.6Multivariable models a Race/Ethnicity model: Reference group—Non‐Hispanic WhiteNon‐Hispanic Black1.50 (1.01–2.22)0.040.92 (0.42–1.97)0.82Hispanic0.67 (0.42–1.08)0.100.84 (0.26–2.70)0.78Other1.28 (0.77–2.12)0.341.33 (0.42–4.20)0.62Neighborhood COI model1.00 (0.99–1.00)0.891.00 (0.98–1.01)0.46Neighborhood COL model: Reference group—Very low COLLow0.86 (0.53–1.41)0.570.87 (0.32–2.38)0.79Moderate1.29 (0.80–2.08)0.291.57 (0.66–3.74)0.31High0.93 (0.55–1.56)0.781.01 (0.36–2.84)0.99Very high0.78 (0.46–1.34)0.370.53 (0.14–2.00)0.35Abbreviations: AKI, acute kidney injury; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; BMI, body mass index; CI, confidence interval; COI, Child Opportunity Index; COL, Child Opportunity Level; PR, prevalence ratios. a Adjusted for baseline age category, baseline BMI category, sex, and baseline leukemia risk group. Univariable and multivariable prevalence ratios of AKI at diagnosis. Abbreviations: AKI, acute kidney injury; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; BMI, body mass index; CI, confidence interval; COI, Child Opportunity Index; COL, Child Opportunity Level; PR, prevalence ratios. Adjusted for baseline age category, baseline BMI category, sex, and baseline leukemia risk group. AKI During Therapy In ALL, 393 out of 952 (41.3%) patients experienced AKI after initiation of chemotherapy. The difference in the Kaplan–Meier curves by race/ethnicity neared significance (log‐rank p = 0.06), with NHB patients having decreased time to AKI. There was no significant difference by COL (Figure 1A,B). In univariable analyses investigating first AKI during ALL chemotherapy, older age, male sex, overweight or obese BMI status, intermediate/high baseline leukemia risk group assignment, NHB race/ethnicity, and prior AKI were associated with significantly increased hazard of AKI. CHOP and TCH study sites, increasing COI, and Very High COL had a significantly lower hazard of AKI. However, multivariable models of first AKI during therapy by race/ethnicity, COI, and COL were not statistically significant (Table 3). Kaplan–Meier curves of first AKI during therapy by race/ethnicity and COL. (A) Kaplan–Meier survival curve by race/ethnicity in ALL cohort. (B) Kaplan–Meier survival curve by COL in ALL cohort. (C) Kaplan–Meier survival curve by race/ethnicity in AML cohort. (D) Kaplan–Meier survival curve by COL in AML cohort. AKI, acute kidney injury; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; COL, Child Opportunity Level. TABLE 3Univariable and multivariable hazard ratios of first AKI a during therapy.ALLAMLHR (95% CI) p HR (95% CI) p Univariable modelsBaseline age (year)1.12 (1.10–1.14)< 0.0011.11 (1.06–1.16)< 0.001Sex—Reference group femaleMale sex1.30 (1.06–1.60)0.010.76 (0.46–1.28)0.31BMI category: Reference group—normalUnderweight0.98 (0.67–1.44)0.920.46 (0.11–1.93)0.29Overweight1.58 (1.20–2.09)0.0011.28 (0.57–2.90)0.54Obese1.71 (1.30–2.25)< 0.0011.84 (0.95–3.55)0.07Missing0.78 (0.25–2.46)0.683.39 (0.81–14.22)0.10Baseline leukemia risk group: Reference group—standard riskIntermediate or high2.08 (1.69–2.55)< 0.0011.22 (0.72–2.07)0.45Missing1.80 (0.74–4.40)0.19——Clinical trial: Reference group—never enrolledEver enrolled0.86 (0.70–1.06)0.170.75 (0.45–1.25)0.27Prior AKI b 1.55 (1.17–2.06)0.0023.35 (1.88–5.96)Study Site: Reference group—CHOACHOP0.58 (0.42–0.79)0.0011.11 (0.60–2.04)0.74TCH0.78 (0.62–0.99)0.040.88 (0.44–1.79)0.74Race/Ethnicity: Reference group—Non‐Hispanic WhiteNon‐Hispanic Black1.41 (1.07–1.86)0.011.95 (1.09–3.49)0.02Hispanic1.01 (0.79–1.30)0.920.79 (0.33–1.86)0.58Other0.94 (0.67–1.34)0.740.87 (0.35–2.15)0.76Neighborhood COI0.996 (0.993–0.999)0.021.00 (0.99–1.00)0.55Neighborhood COL: Reference group—very low COLLow0.96 (0.72–1.28)0.780.70 (0.33–1.47)0.34Moderate0.74 (0.53–1.02)0.060.69 (0.29–1.67)0.42High0.83 (0.61–1.13)0.230.95 (0.47–1.94)0.90Very high0.71 (0.52–0.97)0.030.71 (0.32–1.61)0.42Multivariable models c Race/Ethnicity model: Reference group—Non‐Hispanic WhiteNon‐Hispanic Black1.23 (0.93–1.64)0.152.68 (1.43–5.05)0.002Hispanic0.97 (0.74–1.26)0.820.94 (0.35–2.50)0.90Other0.97 (0.68–1.38)0.870.96 (0.36–2.57)0.93Neighborhood COI model1.00 (0.99–1.00)0.491.00 (0.99–1.01)0.51Neighborhood COL model: Reference group—very low COLLow0.93 (0.70–1.25)0.650.54 (0.24–1.20)0.13Moderate0.79 (0.57–1.10)0.160.57 (0.22–1.45)0.24High0.94 (0.69–1.30)0.720.66 (0.30–1.46)0.31Very high0.88 (0.64–1.20)0.420.69 (0.30–1.60)0.39Abbreviations: AKI, acute kidney injury; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; BMI, body mass index; CHOA, Children's Healthcare of Atlanta; CHOP, Children's Hospital of Philadelphia; CI, confidence interval; COI, Child Opportunity Index; COL, Child Opportunity Level; HR, hazard ratio; TCH, Texas Children's Hospital. a First AKI that occurred after the initiation of chemotherapy. b Patient was considered to have a prior AKI if they had an AKI at diagnosis. c Adjusted for baseline age category, baseline BMI category, sex, baseline leukemia risk group, clinical trial enrollment (ever vs. never enrolled), study site, and prior AKI. Univariable and multivariable hazard ratios of first AKI a during therapy. Abbreviations: AKI, acute kidney injury; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; BMI, body mass index; CHOA, Children's Healthcare of Atlanta; CHOP, Children's Hospital of Philadelphia; CI, confidence interval; COI, Child Opportunity Index; COL, Child Opportunity Level; HR, hazard ratio; TCH, Texas Children's Hospital. First AKI that occurred after the initiation of chemotherapy. Patient was considered to have a prior AKI if they had an AKI at diagnosis. Adjusted for baseline age category, baseline BMI category, sex, baseline leukemia risk group, clinical trial enrollment (ever vs. never enrolled), study site, and prior AKI. Univariable models in ALL by chemotherapy course are shown in Table S5. Multivariable models of course‐specific AKI in ALL by race/ethnicity, COI, and COL were not statistically significant. HRs in the Maintenance course, when patients tend to spend the majority of their treatment time in their home environment, as opposed to inpatient, suggest a trend of higher COL groups being associated with lower AKI risk compared to Very Low COL as all HRs were less than 1, though not statistically significant (Table 4). TABLE 4Multivariable hazard ratios a of AKI during therapy by chemotherapy course.ALLInduction HR (95% CI) p Consolidation HR (95% CI) p Interim maintenance HR (95% CI) p Delayed intensification HR (95% CI) p Maintenance HR (95% CI) p Race/Ethnicity model: Reference group—Non‐Hispanic WhiteNon‐Hispanic Black1.26 (0.79–2.01)0.320.95 (0.57–1.60)0.861.06 (0.68–1.65)0.801.04 (0.52–2.08)0.921.24 (0.67–2.26)0.50Hispanic1.04 (0.66–1.64)0.860.91 (0.53–1.55)0.731.10 (0.75–1.62)0.631.68 (0.93–3.04)0.091.42 (0.80–2.51)0.24Other0.55 (0.25–1.22)0.141.28 (0.70–2.35)0.411.13 (0.68–1.88)0.641.06 (0.46–2.44)0.891.11 (0.50–2.46)0.80Neighborhood COI model1.00 (0.99–1.01)0.951.00 (0.99–1.01)0.911.00 (0.996–1.01)0.681.00 (0.99–1.00)0.250.99 (0.98–1.00)0.09Neighborhood COL model: Reference group—very low COLLow0.81 (0.49–1.34)0.411.36 (0.79–2.34)0.261.23 (0.78–1.93)0.361.19 (0.63–2.22)0.590.64 (0.34–1.19)0.16Moderate0.71 (0.39–1.26)0.240.71 (0.36–1.40)0.321.22 (0.76–1.96)0.410.88 (0.41–1.91)0.750.55 (0.28–1.08)0.08High0.89 (0.51–1.57)0.701.25 (0.70–2.25)0.451.15 (0.71–1.86)0.560.75 (0.34–1.63)0.460.50 (0.24–1.02)0.055Very high0.94 (0.54–1.62)0.810.97 (0.53–1.78)0.921.09 (0.67–1.77)0.730.74 (0.35–1.59)0.440.62 (0.32–1.20)0.16AMLInduction I HR (95% CI) p Induction II HR (95% CI) p Intensification I HR (95% CI) p Intensification II HR (95% CI) p Race/Ethnicity model: Reference group—Non‐Hispanic WhiteNon‐Hispanic Black1.75 (0.68–4.48)0.241.54 (0.39–6.02)0.544.00 (1.27–12.63)0.020.67 (0.21–2.10)0.49Hispanic0.91 (0.25–3.26)0.893.01 (0.54–16.69)0.210.63 (0.06–6.28)0.700.29 (0.03–3.02)0.30Other1.36 (0.37–5.01)0.651.25 (0.17–8.97)0.831.13 (0.20–6.19)0.890.59 (0.06–5.38)0.64Neighborhood COI model1.01 (0.99–1.02)0.160.99 (0.97–1.01)0.191.00 (0.98–1.01)0.760.98 (0.96–1.00)0.06Neighborhood COL model: Reference group—very low COLLow0.89 (0.29–2.78)0.850.69 (0.16–3.10)0.630.23 (0.04–1.19)0.080.44 (0.10–1.85)0.26Moderate1.00 (0.27–3.68)0.9980.25 (0.02–2.59)0.241.18 (0.25–5.49)0.830.12 (0.01–0.99)0.049High1.27 (0.39–4.11)0.690.36 (0.07–1.90)0.230.51 (0.12–2.08)0.350.16 (0.03–0.81)0.03Very high2.01 (0.58–6.87)0.270.60 (0.10–3.59)0.580.61 (0.15–2.43)0.490.46 (0.10–2.02)0.30Abbreviations: AKI, acute kidney injury; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; BMI, body mass index; CI, confidence interval; COI, Child Opportunity Index; COL, Child Opportunity Level; HR, hazard ratio. a Adjusted for baseline age category, baseline BMI category, sex, baseline leukemia risk group, clinical trial enrollment (ever vs. never enrolled), study site, and prior AKI. Multivariable hazard ratios a of AKI during therapy by chemotherapy course. Abbreviations: AKI, acute kidney injury; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; BMI, body mass index; CI, confidence interval; COI, Child Opportunity Index; COL, Child Opportunity Level; HR, hazard ratio. Adjusted for baseline age category, baseline BMI category, sex, baseline leukemia risk group, clinical trial enrollment (ever vs. never enrolled), study site, and prior AKI. In AML, 62 out of 168 (36.9%) patients experienced AKI during therapy. There was a statistically significant difference between Kaplan–Meier survival curves by race/ethnicity (log‐rank p = 0.04), with NHB patients again having the shortest time to AKI (Figure 1C,D). In univariable analyses of first AKI during therapy, an increased hazard of AKI was associated with older baseline age, prior AKI, and NHB race/ethnicity. COI and COL were not significantly associated with hazard of AKI. After adjustment, the hazard of first AKI during therapy in NHB patients increased compared to the unadjusted model and was statistically significant (HR 2.68 [1.43–5.05], p = 0.002). COI and COL were not significantly associated with first AKI during therapy (Table 3). Univariable models of hazard of AKI are shown in Table S6. In course‐specific multivariable analyses, the hazard of AKI during intensification I was 4 times higher among NHB patients compared to NHW patients (HR 4.00 [1.27–12.63], p = 0.02). Moderate COL had an 88% decreased hazard of AKI (HR 0.12 [0.01–0.99], p = 0.049) and High COL had an 84% decreased hazard of AKI during intensification II compared to Very Low COL (HR 0.16 [0.03–0.81], p = 0.03) (Table 4). Sensitivity Analyses Using the KDIGO definition, 820 (86%) of the ALL cohort and 129 (77%) of the AML cohort had at least one AKI episode, either at presentation or during therapy. In ALL, patients with AKI were significantly younger, had lower baseline eGFR, and had a significantly higher proportion of patients with standard risk ALL at baseline. In AML, there were no significant differences in baseline characteristics between AKI and never AKI groups (Tables S7 and S8). Hispanic patients with ALL were at significantly lower risk of AKI at the time of diagnosis compared to NHW patients in the multivariable model (PR 0.70 [95% CI 0.52–0.96], p = 0.03), while there were no differences in the hazard of AKI during therapy by race/ethnicity. There were no significant differences in AKI at diagnosis or during therapy by COL in ALL patients. In AML, there were no significant differences in AKI at diagnosis by race/ethnicity or COL. NHB patients continued to have a significantly increased hazard of AKI during therapy (HR 1.96 [IQR 1.19–3.25], p = 0.008), but there were no significant differences by COL (Table 5). TABLE 5Sensitivity analysis: multivariable adjusted prevalence ratios of AKI at diagnosis and hazard ratios of first AKI during therapy in ALL and AML cohorts using the KDIGO definition. a AKI at diagnosis PR (95% CI) b p First AKI during therapy HR (95% CI) c p ALL cohortRace/Ethnicity model: Reference group—Non‐Hispanic WhiteNon‐Hispanic Black1.28 (0.97–1.69)0.081.11 (0.89–1.38)0.34Hispanic0.70 (0.52–0.96)0.030.88 (0.73–1.08)0.22Other1.13 (0.80–1.60)0.501.20 (0.93–1.54)0.15Neighborhood COI model1.00 (0.998–1.00)0.321.00 (0.998–1.00)0.70Neighborhood COL model: Reference group—Very low COLLow0.82 (0.57–1.17)0.280.98 (0.78–1.22)0.84Moderate1.12 (0.79–1.58)0.540.78 (0.61–1.00)0.052High1.12 (0.80–1.56)0.521.07 (0.85–1.36)0.56Very high1.06 (0.76–1.48)0.721.03 (0.81–1.30)0.82AML cohortRace/Ethnicity model: Reference group—Non‐Hispanic WhiteNon‐Hispanic Black1.29 (0.77–2.16)0.321.96 (1.19–3.25)0.008Hispanic1.13 (0.56–2.28)0.721.18 (0.60–2.30)0.62Other0.61 (0.19–1.91)0.401.14 (0.52–2.50)0.75Neighborhood COI model0.998 (0.99–1.00)0.631.00 (0.99–1.01)0.93Neighborhood COL model: Reference group—Very low COLLow0.67 (0.32–1.41)0.290.87 (0.46–1.67)0.68Moderate0.88 (0.44–1.76)0.720.96 (0.46–1.98)0.91High0.84 (0.42–1.67)0.620.93 (0.51–1.71)0.83Very high0.86 (0.46–1.62)0.640.99 (0.51–1.92)0.97Abbreviations: AKI, acute kidney injury; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; CI, confidence interval; COI, Child Opportunity Index; COL, Child Opportunity Level; HR, hazard ratio; PR, prevalence ratio. a AKI is defined as: creatinine is ≥ 1.5 times or ≥ 0.3 mg/dL higher than the baseline creatinine or patient is on CKRT. b Adjusted for baseline age category, baseline BMI category, sex, and baseline leukemia risk group. c Adjusted for baseline age category, baseline BMI category, sex, baseline leukemia risk group, clinical trial enrollment (ever vs. never enrolled), and prior AKI. Sensitivity analysis: multivariable adjusted prevalence ratios of AKI at diagnosis and hazard ratios of first AKI during therapy in ALL and AML cohorts using the KDIGO definition. a Abbreviations: AKI, acute kidney injury; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; CI, confidence interval; COI, Child Opportunity Index; COL, Child Opportunity Level; HR, hazard ratio; PR, prevalence ratio. AKI is defined as: creatinine is ≥ 1.5 times or ≥ 0.3 mg/dL higher than the baseline creatinine or patient is on CKRT. Adjusted for baseline age category, baseline BMI category, sex, and baseline leukemia risk group. Adjusted for baseline age category, baseline BMI category, sex, baseline leukemia risk group, clinical trial enrollment (ever vs. never enrolled), and prior AKI. Multivariable analyses of first mKDIGO AKI during therapy including only patients who had complete follow‐up through upfront chemotherapy completion (ALL n = 458, AML n = 99) were similar to the primary analyses. There were no significant differences in AKI risk by race/ethnicity or COL in ALL. In AML, NHB race/ethnicity had a significantly increased hazard of AKI during therapy (HR 2.37, 95% CI 1.11–5.05, p = 0.02). There were no significant differences in AKI risk by COL (Table S9). Discussion In the ALL cohort, there was a higher risk of AKI at diagnosis among NHB patients after model adjustment. In univariable analyses, NHB patients had significantly increased hazard of AKI during therapy, and higher COL groups had lower hazard of AKI. However, these differences were no longer statistically significant after controlling for clinical risk factors. This is in contrast to the AML cohort, for whom there were no significant disparities noted in AKI at diagnosis, but there was a more than two‐fold higher hazard of AKI during therapy among NHB patients after controlling for covariates. Additionally, there were differences in the impact of race/ethnicity and COL depending on the specific chemotherapy course. These findings were largely attenuated in the sensitivity analyses using the KDIGO definition in which a larger proportion of the cohorts overall were diagnosed with AKI, with the exception of the persistence of significantly increased hazard of AKI during therapy in NHB AML patients compared to NHW. Sensitivity analyses also suggested that Hispanic patients were at significantly decreased risk of AKI at the time of leukemia diagnosis (ALL only). Racial disparities in AKI have been described in various adult and pediatric populations [14, 28]. These disparities are theorized to occur due to complex interactions between clinical, socioeconomic, and genetic factors. Perhaps the most noteworthy result of this study is that even after controlling for clinical factors, NHB patients had a 2.68 times higher hazard of AKI during therapy compared to NHW patients in the AML cohort, and NHB continued to have significantly increased hazard of AKI during therapy in sensitivity analyses. This finding is consistent with the only other study to our knowledge to have evaluated race/ethnicity and AKI risk in AML, which utilized the PHIS database [9]. This increased risk was not present at the time of leukemia diagnosis; there were no significant differences by COL during therapy, and AML therapy is largely given in the inpatient setting. This suggests that there may be factors that occur during treatment itself that may put NHB patients at higher AKI risk. One potential factor could be clinical trial enrollment, as NHB patients have been found to have significantly lower clinical trial enrollment in pediatric AML, and a study in adolescent/young adult ALL patients showed that clinical trial enrollment eliminated survival disparities between Hispanic and NHW patients [29, 30]. However, our study did not show that clinical trial enrollment had significant positive or negative associations with AKI during therapy in univariable analyses. Associations between clinical trial enrollment and treatment toxicities, such as AKI, were not the aim of this study but may warrant more rigorous future investigation. Associations of APOL1 mutations, which are prevalent in patients of West African descent and highly associated with the development of CKD, have not consistently been found to have associations with AKI [28], but other as‐yet‐unknown genetic factors that may lead to differential metabolism of nephrotoxic agents and affect risk of renal toxicity could exist. For example, a prior study showed that cytarabine pharmacogenomic differences, which are more prevalent in Black patients, could lead to increased mortality in Black compared to White patients with AML. This survival disparity resolved with a pharmacogenomically‐tailored treatment regimen [31]. We did not find that Hispanic patients were at increased risk of AKI in pediatric ALL or AML, and our sensitivity analysis suggested that Hispanic patients were at significantly less risk of AKI at the time of ALL diagnosis compared to NHW patients. This was an unexpected result particularly in ALL given prior studies have shown that pediatric Hispanic ALL patients are at increased risk of methotrexate neurotoxicity, which is highly associated with creatinine elevation. However, it should be noted that the actual rate of AKI among Hispanic patients in these prior studies was quite low, around 5% [12, 32]. The strengths of this study include the richness and granularity of the laboratory data and the availability of residential address data which allowed for investigation of environmental risk factors and their association with health outcomes. Additionally, the results of this study may be generalizable to many other children with acute leukemia, as this cohort is comprised of patients from multiple high‐volume pediatric centers in diverse regions of the United States. However, there were several limitations. As the sensitivity analyses showed, there could be significant differences in results based on AKI definition used. We utilized the mKDIGO definition to mitigate potential overestimation of AKI prevalence by the KDIGO definition due to low baseline creatinine values in children with acute leukemia. However, several studies have shown that both mKDIGO and KDIGO definitions of AKI may be significantly associated with the development of CKD, hypertension, and increased mortality in pediatric critical care populations [21, 22, 33, 34, 35, 36]. There is additional risk of misclassification bias as we included patients who may not have completed all of their therapy at a LEARN site. Therefore, there may be patients who developed AKI during therapy outside of LEARN who would not have been ascertained in our study. Despite this, sensitivity analyses including only patients with full follow‐up through upfront chemotherapy were similar to our primary analysis. This study is limited in that it did not evaluate the association of disparities in AKI with other important clinical outcomes, such as leukemia survival. Additionally, we were not able to account for changes in leukemia risk group throughout therapy or treatment exposures during therapy. However, controlling for these factors may inadvertently mask the presence of disparities. We also only evaluated COI based on the residential address at the time of leukemia diagnosis, which may not necessarily reflect the actual environmental exposures of the individual patient. The course‐specific results should be interpreted with caution, as false positive results may have occurred in the setting of multiple testing. Multiple testing would not impact the main findings of the significant associations of race/ethnicity with AKI at the time of ALL diagnosis and first AKI during AML therapy. Finally, in the AML cohort, we had lower power to detect disparities due to the smaller cohort size. However, as there are only about 600 new pediatric AML diagnoses per year in the United States, our cohort does represent a significant proportion of the pediatric AML population in the United States [1]. Conclusion In conclusion, this study suggests that racial/ethnic and neighborhood disparities in AKI risk may exist in pediatric patients with ALL and AML. In ALL, these disparities may be at least partially explained by differences in clinical risk factors, as the disparities were decreased after controlling for clinical risk factors. Notably, in AML, there was no evidence to suggest a significant difference in AKI at the time of leukemia diagnosis by race/ethnicity or COL, but after controlling for clinical risk factors, there was an even more pronounced increase in risk of AKI during therapy among NHB patients compared to NHW patients. Neighborhood factors were not significantly associated with the development of AKI during AML therapy, which may be due to the majority of AML treatment occurring in the inpatient setting. There may be some differences in the impact of race/ethnicity and neighborhood factors on AKI risk by chemotherapy course, though these results should be assessed cautiously as the nature of multiple testing as required for course‐specific analyses may increase the likelihood of false positive statistically significant results. Further investigation into other potential factors that may influence racial and ethnic disparities in AKI risk is warranted. Author Contributions Wendy C. Bravo: conceptualization (equal), formal analysis (equal), investigation (equal), methodology (equal), writing – original draft (equal), writing – review and editing (equal). Yimei Li: conceptualization (equal), methodology (equal), supervision (equal), writing – review and editing (equal). Tamara P. Miller: conceptualization (equal), funding acquisition (equal), methodology (equal), supervision (equal), writing – review and editing (equal). Kelly Getz: data curation (equal), writing – review and editing (equal). Lusha Cao: data curation (equal), writing – review and editing (equal). Edward Krause: data curation (equal), software (equal), writing – review and editing (equal). Mark Ramos: data curation (equal), software (equal), writing – review and editing (equal). Judy Lee: data curation (equal), writing – review and editing (equal). Maria Monica Gramatges: data curation (equal), writing – review and editing (equal). Karen R. Rabin: data curation (equal), writing – review and editing (equal). Michael Scheurer: data curation (equal), writing – review and editing (equal). Richard Aplenc: conceptualization (equal), funding acquisition (equal), methodology (equal), supervision (equal), writing – review and editing (equal). Michelle Denburg: conceptualization (equal), funding acquisition (equal), methodology (equal), supervision (equal), writing – review and editing (equal). Funding This study was funded by the Children's Hospital of Philadelphia Philanthropy, NCI award 5K07CA211956, and NIDDK award K26DK138375. Ethics Statement This study was approved by the Institutional Review Board at Children's Hospital of Philadelphia, Children's Healthcare of Atlanta, and Texas Children's Hospital. Conflicts of Interest The authors declare no conflicts of interest. Supporting information TableS S1–S9: cam471937‐sup‐0001‐Supinfo.docx. Acknowledgments This study was funded by the Children's Hospital of Philadelphia Philanthropy, NCI award 5K07CA211956, and NIDDK award K26DK138375. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NCI, NIDDK, the National Institutes of Health, the Department of Health and Human Services, or the government of the United States. Dr. Bravo was supported by the St. Baldrick's To‐Morrow's Research Fund Scholar Award and The Saban Research Institute Research Career Development Award. Data Availability Statement Deidentified data are available upon request. References Associated Data Supplementary Materials TableS S1–S9: cam471937‐sup‐0001‐Supinfo.docx. Data Availability Statement Deidentified data are available upon request.