Graphical abstract. Overview of study design and results. aHR, adjusted hazard ratio; AKI, acute kidney injury; ALL, acute lymphoblastic leukemia; CKD, chronic kidney disease; IQR, interquartile range.
Introduction
Acute lymphoblastic leukemia (ALL) is the most common childhood malignancy, accounting for one-quarter of pediatric cancers [1,2]. Advances in risk stratification, chemotherapy, and supportive care have increased survival to above 90% in high-income settings [3,4]. As survivorship improves, treatment-related late effects, particularly kidney dysfunction, are increasingly recognized [5]. Children with ALL are at risk for acute kidney injury (AKI) especially during induction therapy and chronic kidney disease (CKD) in survivorship via leukemia-related and treatment-related pathways. Leukemia-related factors, which are prominent at diagnosis and during induction, include leukemic kidney infiltration (LKI) and tumor lysis syndrome (TLS). Treatment-related factors include intensive multimodal chemotherapy, adjunctive nephrotoxic medications, and contrast exposure, as well as intercurrent complications such as sepsis [5,6].
AKI in pediatric ALL is common with reported incidence of 25%–84% [7-9]. Although many episodes recover, cumulative kidney insults may contribute to subsequent CKD [9-11]. CKD is increasingly recognized among survivors of childhood cancer as longevity improves, affecting more than 1% of adult cancer survivors overall [12]. A study in pediatric cancer patients reported the CKD rate of 22.6% [10]. Contemporary data suggest that the prevalence of CKD among cancer survivors is on the rise [12]. A recent review estimates that up to 20% of adult survivors may develop reduced kidney function over time [5].
Given high cure rates and the potential lifelong burden of kidney disease in children with ALL, early recognition and risk stratification are essential. We therefore conducted a retrospective cohort study of children with ALL to: (1) describe the incidence, causes, and clinical correlates of AKI during induction therapy, and (2) estimate CKD-free survival and identify predictors of CKD, including age at diagnosis and induction-phase kidney complications, during subsequent follow-up.
Methods
1. Study design and population
We conducted a retrospective cohort study at a single, university-affiliated tertiary-care center. Medical records of patients aged 2–18 years who were newly diagnosed with ALL between 2015 and 2024 were reviewed. Patients with <3-month of follow-up after diagnosis were excluded. The study was approved by the Institutional Review Board of the Faculty of Medicine, Chulalongkorn University, Bangkok, Thailand (No. 0759/66).
2. Clinical and laboratory assessment
Demographic, clinical, and laboratory data were abstracted from the initial presentation through the development of CKD or the last follow-up. TLS was classified according to the Cairo–Bishop criteria into laboratory TLS and clinical TLS. Laboratory TLS was defined by ≥2 metabolic abnormalities within a 24-hour period occurring from 3 days before to 7 days after initiation of chemotherapy, including potassium >6.0 mmol/L, phosphate >6.5 mg/dL, calcium <7.0 mg/dL, or uric acid >8.0 mg/dL and clinical TLS required laboratory TLS plus at least one clinical complication (creatinine elevation, arrhythmia, or seizure) [13]. Sepsis was defined as life-threatening organ dysfunction (respiratory, cardiovascular, coagulation, and/or neurologic) in the setting of infection [14]. LKI was defined as radiologic evidence of kidney enlargement and/or infiltrative changes attributed to ALL in the ultrasound, computed tomography, or magnetic resonance imaging.
3. Treatment protocol
Childhood ALL was treated according to national standardized protocols, with drug selection and dosing determined by risk stratification [15]. In this study, standard risk was defined as B-cell ALL in children aged 2–10 years with a presenting white blood counts (WBCs) <50×10³/μL. High risk included B-cell ALL with age ≥10 years or WBC ≥50×10³/μL, T-cell ALL, extramedullary disease, or high-risk immunophenotypic, cytogenetic, or molecular features.
Therapy comprised induction, consolidation, interim maintenance, delayed intensification, and maintenance. Induction included vincristine, L-asparaginase, and prednisolone; doxorubicin was added for high-risk patients. Interim maintenance used methotrexate (approximately 2.5 g/m² in standard-risk; 5 g/m² in high-risk). Delayed-intensification comprised vincristine, doxorubicin, dexamethasone, L-asparaginase, cyclophosphamide (total dose 1,000 mg/m²), thioguanine, and cytarabine. Several of these agents have potential kidney effects, mainly through hemodynamic changes, thrombotic or tubular injury, but is not considered the primary driver of AKI compared with the induction phase. Maintenance comprised 6-mercaptopurine and methotrexate [15].
4. TLS prophylaxis and supportive care
TLS prevention followed a risk-stratified institutional protocol. High-risk patients, defined by hyperleukocytosis (WBC≥100×10³/μL) or a diagnosis of Burkitt leukemia, received aggressive intravenous hydration (typically 2–3 times maintenance) with strict input–output monitoring and a urine output target of approximately 4–6 mL/kg/hr; loop diuretics were used as needed to achieve this target. Allopurinol was administered for urate-lowering when serum uric acid exceeded 7 mg/dL. In patients with hyperleukocytosis, a 24- to 48-hour steroid prephase before full-intensity chemotherapy was considered part of TLS prophylaxis to reduce tumor burden and TLS risk. Low-risk patients received maintenance intravenous fluids with careful monitoring of fluid balance and electrolytes. Rasburicase was not available and was not used in this cohort.
5. Outcomes definitions
The outcomes of interest were: (1) AKI occurring during induction therapy, and (2) development of CKD during longitudinal follow-up. For AKI, serum creatinine values were reviewed from the initiation of induction to the start of consolidation chemotherapy; AKI episodes occurring after this induction window were not systematically collected and were not included in the AKI analyses. AKI was classified by the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines using serum creatinine only [16]. Baseline serum creatinine was defined as the value at initial presentation if it fell within the age-appropriate reference interval. If elevated at presentation, the baseline was the lowest serum creatinine recorded in the preceding 3 months [17]. When neither measurement was available, baseline creatinine was back-calculated using the Hoste equation [18].
6. Statistical analysis
Analyses were performed in Stata v19 (StataCorp, College Station, USA). Two-sided P<0.05 was considered statistically significant. Continuous variables are presented as median (interquartile range, IQR) and compared with the Mann-Whitney U test; categorical variables were compared using chi-square or Fisher exact tests, as appropriate. For AKI during induction, baseline characteristics and exposures were compared between AKI and non-AKI groups.
CKD-free survival was estimated using Kaplan-Meier methods, with log-rank tests for group comparisons. Time to CKD was evaluated using Cox proportional hazards models. Given the limited number of CKD events, we prespecified a parsimonious model. Age at diagnosis and presenting WBC were included a priori based on clinical relevance and prior evidence. To improve interpretability, continuous covariates were scaled as WBC per 10×10³/μL. Additional covariates (e.g., sex, AKI during induction, TLS, and LKI) were considered if their univariable Cox P value was <0.10. Because age at last follow-up is postbaseline and time-dependent, it was not included in multivariable Cox models.
The proportional hazards assumption was assessed whether the effects of age, WBC, and LKI on CKD risk changed over time using Schoenfeld residual tests. Cox results are reported as hazard ratios (HRs) with 95% confidence intervals (CIs).
Results
1. Patient characteristics
Of 114 children with ALL identified during the study period, one died within 3 months and was excluded; 113 were analyzed. The median age at diagnosis was 5.6 (3.8–9.4) years. Baseline serum creatinine was available in 80 of 113 patients (70.8%). Patient demographics and induction-phase features are summarized in Fig. 1: 50.4% were male, 86.7% had B-cell ALL, and 50.4% were classified as high-risk (Fig. 1A). Two patients (1.8%) had preexisting kidney anomalies (simple cyst, n=1; duplex collecting system, n=1).
Hyperleukocytosis was present in 14 of 113 patients (12.4%). TLS occurred in 35 patients (31.0%); 16 (45.7%) had laboratory TLS alone and 19 (54.3%) had clinical TLS. All clinical TLS events were characterized by concurrent KDIGO-defined AKI with creatinine elevation. No clinical TLS episodes with documented arrhythmia or seizure were identified. TLS was more frequent in patients with hyperleukocytosis (10 of 14 [71.4%] vs. 25 of 99 [25.3%], P=0.001), whereas clinical TLS was not significantly different between groups (4 of 14 [28.6%] vs. 15 of 99 [15.2%], P=0.25).
Seventy-six patients (67.3%) had febrile neutropenia, 14 (12.4%) had sepsis, and 28 (24.8%) required intensive care (Fig. 1E). Kidney imaging, including ultrasound or computed tomography, was performed in 91 patients (80.5%). Imaging was obtained for all patients with AKI; by contrast, some patients in the non-AKI group did not undergo imaging. Among those imaged, LKI was identified in 7 patients (7.7%) (Fig. 1E).
2. Kidney outcomes during induction
During induction therapy, AKI occurred in 49 of 113 (43.4%) (KDIGO stage 1 [n=28, 57.1%]; stage 2 [n=12, 24.5 %], stage 3 [n=9, 18.4%]). Five (4.4%) required kidney replacement therapy. Median time to AKI recovery was 21 (IQR, 11–26) days. AKI frequently occurred in the setting of TLS-related biochemical derangements, and TLS was the most documented clinical context for AKI during induction (28.6%). Other identifiable causes of AKI during induction were acute tubular necrosis (24.5%), prerenal causes (20.4%), and sepsis (10.2%). LKI was recorded in a minority, at times overlapping with TLS and/or sepsis (Fig. 2).
Sex, leukemia subtype/risk, laboratory values (hemoglobin, WBC, platelets, uric acid, phosphate, calcium), and exposure to nephrotoxins did not differ significantly between AKI and non-AKI groups (Fig. 1A–D). Rates of TLS, sepsis, and intensive care were higher in the AKI group but did not reach statistical significance. By contrast, LKI was more frequent in AKI than non-AKI patients (P=0.01) (Fig. 1E). Induction AKI incidence was similar in patients with and without hyperleukocytosis (6 of 14 [42.9%] vs. 43 of 99 [43.4%], P=0.97).
TLS onset occurred early in induction: median day 2 (IQR, 1–5) in TLS patients without AKI and 2 (1–4) in those with AKI (P=0.76). In children with both TLS and AKI, AKI onset occurred at median day 2 (IQR, -1 to 9) relative to the start of induction and lasted 24 (20–31) days. Peak serum uric acid tended to be higher in TLS patients who developed AKI than in those who did not (10.5 [7.3–13.8] mg/dL vs. 8.1 [6.6–9.7] mg/dL, P=0.07), whereas peak phosphate, peak potassium, and nadir calcium were similar between groups (all P≥0.40) (Supplementary Table 1).
3. CKD outcomes
Five patients (4.4%) died during the study period. Over 644 patient-years of follow-up (median, 5.1 [IQR, 2.9–8.3] years), 15 of 113 (13.3%) developed CKD (stage 2 [n=12], stage 3 [n=3]). CKD-free survival at 1, 3, and 5 years was 99.1% (95% CI, 93.8%–99.9%), 95.3% (89.0%–98.0%), and 94.1% (87.2%–97.3%), respectively. At last follow-up, median GFR for the cohort was 121 (99–141) mL/min/1.73 m². When stratified by induction-phase AKI, median final GFR was 116 (98–136) mL/min/1.73 m² among patients with AKI and 124 (104–148) mL/min/1.73 m² among those without AKI, with substantial overlap in the distributions (Supplementary Table 2). Among 15 patients with CKD; 9 of 49 (18.4%) in the prior-induction AKI group and 6 of 64 (9.4%) in the no-induction AKI group. During follow-up, 2 patients progressed from stage 2 to stage 3 CKD, one with and one without prior induction AKI, whereas the remaining stage 2 CKD cases remained stable.
In descriptive comparisons (Table 1), patients who developed CKD were older at diagnosis and older at last follow-up; other baseline characteristics (sex, ALL subtype/risk, preexisting kidney disease) were similar. Median WBC at presentation was higher among those who developed CKD, whereas hemoglobin, platelets, uric acid, phosphate, and calcium were comparable. AKI, kidney replacement therapy, TLS, and LKI during induction were more frequent among CKD cases, though differences were not statistically significant.
4. Time-to-event analyses
In univariable Cox models (Table 2), older age at diagnosis was associated with higher CKD risk (HR, 1.27 per year; 95% CI, 1.10–1.48; P=0.002). LKI showed a higher hazard (HR, 4.34; 95% CI, 0.87–21.54; P=0.07). Presenting WBC was not associated with CKD (HR, 0.99 per 10×10³/μL; 95% CI, 0.95–1.04; P=0.83). Sex, AKI during induction, and TLS were not significant (all P>0.10).
In a prespecified parsimonious multivariable Cox model included age and WBC at diagnosis a priori, with LKI added based on univariable screening. Age at diagnosis remained independently associated with CKD (adjusted HR [aHR], 1.28 per year; 95% CI, 1.04–1.57; P=0.02), whereas WBC was not (aHR, 0.99 per 10×10³/μL; 95% CI, 0.96–1.02; P=0.69). LKI was associated with a nearly threefold higher hazard (aHR, 2.93; 95% CI, 0.87–9.89) but did not reach statistical significance (P=0.08). Proportional-hazards testing showed no violations (global P=0.70; age at diagnosis P=0.40; WBC P=0.53; LKI P=0.39), indicating that the effects of these variables were stable over time.
Fig. 3 illustrates Cox-adjusted CKD-free survival by age and LKI status (panel A: no LKI; panel B: LKI present). With WBC held at the cohort median, curves at ages 5, 10, and 15 years show ordered, monotonic separation across follow-up; the “overall” model-based curve (evaluated at the sample mean age, WBC fixed at the median) lies between the age-specific curves. The LKI panel shows generally lower survival probabilities than the no-LKI panel, consistent with the adjusted estimate for LKI, although the association did not reach statistical significance.
Discussion
Since 2015, national standardized regimens adapted from the Children’s Oncology Group have been implemented at our center, with a 5-year overall survival of 82% in pediatric ALL [15]. Against this background, kidney complications remain clinically relevant despite advances in therapy and supportive care, particularly during the early, cytoreductive phases of treatment [7-9]. Reported AKI incidence varies widely across studies, reflecting heterogeneity in definitions, study populations, designs and settings, and treatment protocols [7-9]. In a retrospective electronic-health-record analysis, AKI occurred in 84% of children with hematologic malignancies using KDIGO criteria versus 25% using CTCAE (Common Terminology Criteria for Adverse Events) [7]. Other contemporary KDIGO-based cohorts have reported AKI in 33%–45% of pediatric ALL patients, with events concentrated mainly during induction and interim maintenance [7-9]. In our cohort, AKI occurred in 43% of children during the induction phase, aligning with these reports and reinforcing induction as a high-risk period for kidney injury.
AKI in ALL arises from multifactorial mechanisms involving prerenal, intrinsic, and less commonly postrenal pathways. In hematologic malignancies, prerenal injury related to volume depletion and reduced effective circulating volume is common, whereas acute tubular necrosis predominates among intrinsic causes [22]. AKI in ALL likely reflects the combined effects of the underlying disease (tumor burden, leukostasis, TLS, LKI) and its treatment (cytotoxic agents, nephrotoxins, hemodynamic instability) [8,10,23]. In our cohort, TLS was the most frequently documented clinical context for AKI. However, definitive causal attribution is complicated by definitional overlap: the metabolic abnormalities that define TLS (e.g., hyperuricemia, hyperphosphatemia, hyperkalemia, hypocalcemia) both predispose to and arise from impaired kidney function. Thus, TLS and AKI are best viewed as mutually reinforcing components of a shared pathophysiologic cascade rather than as a simple unidirectional cause-effect relationship.
To further characterize TLS-related AKI, we examined TLS timing, severity, and biochemical patterns. TLS occurred early in induction in both children who did and did not develop AKI. In patients with both TLS and AKI, AKI onset clustered in close temporal proximity to TLS and typically resolved within the induction phase. Peak serum uric acid levels tended to be higher in TLS patients who developed AKI than in those who did not. These findings suggest that more pronounced hyperuricemia may contribute to AKI risk in the setting of TLS, but no clear biochemical threshold could be identified, and hyperuricemia alone is unlikely to explain the full AKI burden.
Hyperleukocytosis is a classic risk factor for TLS and early organ dysfunction. In our cohort, 12.4% of children had hyperleukocytosis at diagnosis. As expected, TLS was significantly more frequent in patients with hyperleukocytosis than in those with lower WBC counts. By contrast, induction-phase AKI incidence was almost identical in patients with and without hyperleukocytosis. These exploratory data support the link between hyperleukocytosis and TLS but suggest that, within the context of TLS prophylaxis and supportive care, very high presenting WBC may not translate into a clearly higher observable risk of KDIGO-defined AKI during induction.
LKI represents another important kidney complication in ALL. Leukemia frequently involves the kidneys as an extramedullary site [24], and LKI can affect any nephron segment, with predominant interstitial involvement [25]. LKI may initially manifest as AKI [25]. Proposed mechanisms include expansion of the interstitium by infiltrating blasts, increased interstitial pressure, and compression of renal microvasculature and tubules, leading to tubular injury [22], along with cytokine-mediated inflammation and fibrosis [26]. In our cohort, LKI was rare overall but more frequent among patients with AKI than among those without AKI, consistent with these mechanistic pathways and with prior reports linking LKI to kidney dysfunction [25]. Because radiologic LKI can be clinically occult, children with documented LKI warrant close monitoring of kidney function during induction.
Despite the biological plausibility of LKI-associated AKI, routine screening imaging for LKI is not recommended. In a large autopsy series, LKI was present in 54% of ALL cases [24], yet only 5% of 668 leukemia patients undergoing computed tomography had findings consistent with LKI [27]. This discrepancy underscores the low diagnostic yield of systematic imaging solely to detect LKI. In our study, kidney imaging was performed selectively, often prompted by clinical concerns. Consequently, LKI may have been missed in some patients without AKI, and imaging-based estimates of LKI prevalence should be interpreted cautiously when assessing its association with AKI or CKD.
Our secondary aim was to place these induction-phase findings within the context of longer-term kidney outcomes. With contemporary therapy, long-term kidney outcomes in childhood ALL appear more favorable than in historical series. Earlier cohorts from prior decades reported a 10%–20% prevalence of reduced GFR among childhood ALL survivors [28-30]. More recent studies describe relatively low CKD prevalence but often lack detailed time-to-event data [7-9]. One cross-sectional cohort of 45 children followed 1–5 years reported no CKD [9], whereas another identified 4 cases among 214 patients (1.9%) but did not specify follow-up duration [8]. In our time-to-CKD analysis, approximately 1 in 8 patients developed CKD during a median follow-up of just over 5 years, and 5-year CKD-free survival was 94%, providing longitudinal context to these cross-sectional estimates. At last follow-up, median eGFR remained within the normal range overall and in both AKI and non-AKI groups. Most CKD cases were stage 2. Two patients progressed from stage 2 to stage 3 during follow-up, one with and one without prior induction-phase AKI.
Determinants of CKD in ALL survivors have varied across studies. In a recent retrospective cohort, children who developed CKD were older at diagnosis and had lower WBC and platelets and higher blood urea nitrogen than those without CKD [8]. In our cohort, higher presenting WBC observed among CKD cases did not remain independently associated with CKD after adjustment. By contrast, age at diagnosis emerged as the principal independent determinant of CKD: each 1-year increase in age at ALL diagnosis was associated with an approximately 25%–30% higher hazard of CKD, supporting a graded age effect rather than a step increase at the conventional ≥10-year threshold. Presenting WBC and induction-phase AKI did not independently predict CKD in multivariable models, whereas LKI showed a suggestive but nonsignificant association with higher CKD risk, consistent with its low frequency. These findings support treating age at diagnosis as a continuous risk factor when planning CKD surveillance within survivorship care, with older children and adolescents warranting closer follow-up.
Prior pediatric oncology studies have linked recurrent or severe AKI episodes and nephrectomy to subsequent CKD [10]. By contrast, in our cohort a single episode of KDIGO-defined AKI during induction was not independently associated with later CKD. Several factors may contribute to this apparent discrepancy. First, our AKI ascertainment was restricted to the induction phase, so recurrent or later-phase AKI episodes were not systematically captured. Second, we relied solely on serum creatinine criteria, which can miss AKI defined by urine output or by biomarker-positive tubular/interstitial injury that accumulates below AKI thresholds. Third, statistical power for CKD analyses was limited by the small number of CKD events. Finally, not all CKD in ALL survivors arises from overt AKI; non-AKI pathways, including treatment-related hypertension, metabolic dysregulation, and vascular injury, likely contribute. In children with ALL, hypertension is common, reported at 42% by clinic blood pressure and 67% by ambulatory monitoring [9]. These mechanisms may be particularly relevant in older children and adolescents, consistent with the prominent age effect observed in our study [31].
Taken together, our data highlight two complementary messages. First, AKI during induction is common and multifactorial, with TLS and LKI as key contributors. Second, in the subsequent follow-up, CKD risk in pediatric ALL appears to be shaped more by age at diagnosis and possibly by LKI than by a single episode of induction-phase AKI. This pattern supports a model in which induction-phase AKI and longer-term CKD share overlapping but not identical risk architectures: early kidney injury is concentrated during induction, whereas age-related vulnerability and cumulative treatment and comorbidity exposures modulate CKD trajectories over time.
This retrospective study from a single tertiary care center may limit external generalizability. Several aspects of AKI ascertainment warrant caution. First, when a measured baseline creatinine was unavailable, we estimated baseline values, which could affect AKI classification. Second, AKI was systematically ascertained only during the induction phase; AKI episodes occurring later in therapy were not captured, and our AKI estimates therefore reflect induction-phase risk rather than cumulative AKI burden across the entire treatment course. Third, fold-change thresholds may overcall AKI in younger children with very low baseline creatinine [10], whereas creatinine-only criteria can miss AKI defined by urine output or biomarker-positive injury. Fourth, kidney imaging was not performed systematically. Selective imaging may introduce selection and misclassification bias when evaluating associations between LKI and AKI or CKD, and some LKI cases in patients without AKI may have gone undetected.
Regarding CKD evaluation, statistical power for CKD analyses was limited by the small number of CKD events. The retrospective design also precluded precise measurement of cumulative exposures across therapy. Detailed data on repeated kidney insults and nephrotoxin dosing/duration were inconsistently recorded, and time-updated metabolic factors, including hypertension, hyperglycemia, and dyslipidemia, were incompletely captured. Consequently, we could not fully quantify cumulative kidney injury or evaluate the independent contribution of these exposures to CKD. Finally, age-related transitions in GFR equations, from Schwartz in patients ≤18 years to CKD-EPI in patients >18 years, can upwardly shift estimated GFR in older adolescents, potentially reducing CKD classification in a clinically higher-risk group [32]. Any observed association between age and CKD despite this bias likely underestimates the true effect.
Close kidney function monitoring is warranted in children with LKI, given its association with AKI in this cohort and its plausible link to later CKD. Our findings also underscore the importance of prioritized TLS prophylaxis and careful fluid management, particularly in patients with high tumor burden or hyperleukocytosis, and of nephrotoxin stewardship during induction. Age-calibrated CKD surveillance especially for older children and adolescents appears justified, as does closer monitoring when LKI is identified. Finally, survivorship monitoring for CKD should extend beyond patients with documented induction-phase AKI, given the prominent age effect and the potential contribution of non-AKI pathways to kidney impairment.
In conclusion, AKI during induction was common and multifactorial in children with ALL, with TLS and LKI as key early contributors. Older age at diagnosis was the principal independent predictor of CKD, and the age effect demonstrated a linear risk gradient rather than a conventional dichotomous ≥10-year threshold. LKI was associated with induction-phase AKI and showed a suggestive link with subsequent CKD, supporting targeted monitoring when present. Hyperleukocytosis was strongly associated with TLS but not with a clearly higher incidence of induction-phase AKI. Multicenter prospective studies are needed to refine risk-stratified surveillance strategies, to capture kidney injury beyond the induction phase, and to identify modifiable risk factors for CKD in this growing population of ALL survivors.
PDF Links
PubReader
ePub Link
XML Download
PubMed
Download Citation
