Epidemiologic Characteristics of Acute Kidney Injury During Cisplatin Infusions in Children Treated for Cancer

This cohort study examines the rate of and risk factors associated with acute kidney injury in children treatment who receive cisplatin infusions.


Study Procedure
Children are typically hospitalized to receive cisplatin. Most cisplatin protocols comprise 2 to 8 cycles given every 3 to 6 weeks; dosages vary by tumor type (eg, single high-dose bolus [70-100 mg/m 2 ] vs 2-5 days of low-dose bolus [20-60 mg/m 2 daily]). 25,26 The first prospective cisplatin study visit occurred shortly after recruitment at the first or second cisplatin cycle of cancer treatment (early visit [EV]). The late visit (LV) occurred at the last or second to last cycle. At the EV and LV, blood (3 mL) and urine (30 mL) samples were collected on the same day but before cisplatin infusion (day 1), the morning after infusion (day 2), and just before hospital discharge (at most day 5 after infusion). The study protocol is shown in eFigure 1 in the Supplement.

Specimens and Measurements
Sites centrifuged blood samples (1000g for 10 minutes at 21°C). Serum and urine samples were frozen (-80°C) until shipment to the McGill University Health Centre. Specimen collection, freezing, and shipment were monitored using Research Electronic Data Capture (REDCap). 27 At the McGill University Health Centre, urine samples were thawed and processed (1000g for 10 minutes); serum and urine samples were measured for creatinine (isotope dilution mass spectrometry-traceable), potassium, magnesium, and phosphorus levels.

Data Collection
Data collection is summarized in eFigure 1 in the Supplement. Baseline (precisplatin) data were collected retrospectively (demographic characteristics [race/ethnicity, classified by site coordinators; options were investigator defined], cancer details, kidney history [hypertension, CKD, AKI, kidney anomalies, and family history]; medications taken 2 weeks before cisplatin treatment; precisplatin treatment measured GFR; and 24-hour creatinine clearance [CrCl]). Baseline GFR was ascertained using measured GFR if available or CrCl if unavailable; if both were unavailable, GFR was estimated using the CKD in Children (CKiD) equation (using the lowest 3-month precisplatin SCr level). 28 Decreased GFR was defined using age-normative GFR thresholds. 29 From 3 days before to 10 days after study cisplatin infusions, prospective data collection included the following: routinely measured SCr and electrolyte levels, nephrotoxic medications (acyclovir, aminoglycosides, amphotericin B, ifosfamide, nonsteroidal anti-inflammatory drugs, and acetylsalicylic acid), diuretics, and electrolyte supplements (magnesium, phosphorus, and potassium). Immediate precisplatin treatment estimated GFR was calculated using the CKiD equation. 28 Data collected throughout cisplatin treatment included dialysis, intensive care unit (ICU) admissions, infections (positive blood culture results and documentation), and AKI episodes between study visits. Completed standardized case report forms were sent to the Epidemiology Coordinating and Research Centre (Edmonton) for data entry (REDCap) with regular data quality queries. 27 Quality assurance measures were taken to avoid missing data.

Outcomes
Two cisplatin-associated AKI definitions were ascertained. First, SCr-based AKI was defined and staged using a KDIGO SCr-based definition (primary outcome of stage 1 or higher AKI) (eTable 1 in the Supplement). 21 Because cisplatin-associated AKI is nonoliguric, KDIGO urine output criteria were not used. KDIGO timing criteria (48 hours or 7 days) were not used because these were primarily designed for ICU populations. 21 Similar to a previous study, 30 a 10-day period was used to ascertain AKI, allowing for variable routine follow-up times of laboratory measurements after cisplatin treatment. For EV, the baseline SCr level was the lowest 3-month precisplatin treatment level; for participants with previous cisplatin exposure, 3-day pre-EV infusion values were also used. For LV, the baseline SCr level was the lowest 3-day pre-LV level. Peak SCr level was the highest SCr level 10 days after infusion. Severe SCr-based AKI was KDIGO stage 2 or higher.
Second, eAKI was defined using the National Cancer Institute Common Terminology Criteria for Adverse Events, version 4.0 (eTable 1 in the Supplement). 31 The novel term eAKI was chosen as a secondary outcome to reflect that electrolyte losses occur because of kidney tubular injury. Using nadir electrolyte values within 10 days after infusion, eAKI was graded based on serum magnesium, potassium, and phosphorus concentrations (any grade 1 or higher abnormality) (eTable 1 in the Supplement). 31 We explored a secondary eAKI definition in which oral or intravenous electrolyte (potassium, magnesium, or phosphorous) supplementation was also considered. Composite AKI was both SCr-based AKI and eAKI. SCr and electrolyte levels measured for routine care and the 3 studyspecific measures were used to ascertain AKI at the EV and LV. We decided a priori to evaluate routine vs study-measured laboratory values and exclude measures with evident bias (addressing potential issues with serum storage effects).
We calculated the number of days with SCr-based AKI and eAKI, which was the number of days fulfilling SCr-based AKI and eAKI criteria, respectively, within the 10 days after cisplatin infusion.
SCr-based AKI recovery at last SCr level measurement was defined as the last SCr level available in 10 days after infusion less than 1.5 times the baseline SCr level. 32,33 At all EV and LV sample points, fractional excretion (FE) of electrolytes was calculated as follows 34

Statistical Analysis
Sample size was predetermined based on a convenience sample to balance feasibility and power. 23 Variables were compared between groups using distribution-appropriate tests (Mann-Whitney test, Wilcoxon signed-rank test, 2-tailed, unpaired t test, χ 2 test, or Fisher exact test). Spearman correlation between routine and study-measured laboratory values was examined. Percent agreement between AKI definitions was evaluated. Undocumented events and data were considered absent (negative).
Variables associated with AKI were evaluated separately for EV and LV using multivariable logistic regression. Candidate risk factors were categorized as baseline (before cisplatin administration), immediately before or on the day of cisplatin administration, and after cisplatin administration. Purposeful selection and manual backward selection were used to select variables for regression models. Variable associations, correlations, and collinearity were assessed. On the basis of the literature and clinical rationale, we a priori forced age into models. Variables associated with AKI in univariable analyses (P < .10) were initially included in multivariable models. Models were reduced (P < .05, for variable retention) to obtain parsimonious models. Variable deletion was confirmed using the backward likelihood ratio test. In final multivariable models, variables not significantly associated with AKI (P > .05) but considered significant confounders (ie, elimination significantly affected other β coefficients) were retained. Model fit was assessed (Hosmer-Lemeshow goodness of fit and area under the receiver operating characteristic curve [AUC]). Odds ratios (ORs) with 95% CIs were calculated. Separate regression models were used for SCr-based AKI and SCr plus eAKI. A 2-tailed P < .05 was considered statistically significant unless stated otherwise.

Study Cohort
Twelve of the 17 Canadian pediatric cancer centers (71%) were represented. Of 209 eligible patients approached, 159 were enrolled (76% consent rate (mean [SD] age at early cisplatin infusion, 7.2 [5.3] years; 80 [50%] male) ( Figure). Late visit data were not evaluable for 16 participants because the EV occurred at the last cisplatin cycle (Figure). Specimens were available for 158 patients (99%) at the EV; specimens from 129 of 143 (90%) were available at the LV (Figure). Sample collection was successful (EV, Ն92%; LV, Ն83%) (eTable 2 in the Supplement), and the rate of missing data was low (eTable 3 in the Supplement). At the EV, 91 of 159 participants (57%) were cisplatin naive. The LV occurred at a median of the third cisplatin cycle (interquartile range [IQR], second to fourth cycles);

Figure. Flow Diagram of Study Participation and Data and Specimen Collection at Early and Late Cisplatin Study Visits
Cancer type and a positive kidney medical history were associated with SCr-AKI (SCr-AKI: 42% CNS tumors, 38% neuroblastoma and 4% osteosarcoma, and 17% positive kidney medical history; no SCr-AKI: 34% CNS tumors, 23% neuroblastoma and 28% osteosarcoma, and 5% positive kidney medical history; other baseline characteristics were not associated with SCr-AKI) ( Table 1). Two weeks before cisplatin treatment, nephrotoxic medication administration was uncommon (nonsteroidal anti-inflammatory drugs or acetylsalicylic acid [11%] and methotrexate [7%] were most common) (eTable 5 in the Supplement).

EV and Cisplatin Treatment Characteristics
Age younger than 3 years (eFigure 2 in the Supplement) and higher previsit estimated GFR were associated with SCr-AKI at the EV (SCr-AKI: 54% aged <3 years and median eGFR of 166 mL/min/1.73 m 2 ; no SCr-AKI: 23% aged <3 years and median eGFR of 132 mL/min/1.73 m 2 ) ( Table 1). Past SCr-AKI was more common in participants who did vs did not develop SCr-AKI (SCr-AKI: 23% with previous AKI; no SCr-AKI: 7% with previous AKI); ICU admission, infections, concurrent nephrotoxins, radiotherapy (planned in treatment protocol or given), and total cisplatin cycle dose did not differ by SCr-AKI status (Table 1). Participants with SCr-AKI (vs without) had a longer hospital stay (median of 5 days in patients with SCr-AKI and median of 4 days in those without SCr-AKI; Table 1).
l All routine and study-measured SCr values measured before cisplatin infusion at the EV or LV were used. Peak SCr level was the highest SCr level occurring before the EV and LV infusion but occurring after the baseline SCr measurement. For SCr-AKI before the EV, the baseline SCr level was the lowest routine value in the 3 months before cisplatin initiation. For AKI between the EV and LV, the baseline SCr level was the lowest routine or study-measured value in 3 months before the first cisplatin infusion. m Acyclovir, amphotericin, aminoglycosides (gentamycin, tobramycin, or amikacin), ifosfamide, or chemotherapy protocol indicates that a nephrotoxin (aldesleukin, busulfan, carboplatin, dinutuximab, gemcitabine, ifosfamide, lomustine, melphalan, methotrexate, radiotherapy, rituximab, stem cell transplant, or temsirolimus) was given within 24 hours of cisplatin infusion.
At the EV and LV, phosphorus concentrations decreased significantly from before infusion to after infusion and remained low or decreased at discharge (eFigure 3 in the Supplement). Serum magnesium and potassium concentrations increased from before infusion to after infusion and remained higher than before infusion at discharge (eFigure 3 in the Supplement). Fractional extraction of all electrolytes increased significantly from before infusion to after infusion (eTable 8 in the Supplement).

Characteristics by Cancer Type
Participants with neuroblastoma and hepatoblastoma were younger and had high SCr-AKI rates (eTable 9 in the Supplement). Those with osteosarcoma and germ cell tumors had low SCr-AKI rates; none developed SCr-AKI at the LV (eTable 9 in the Supplement). Cisplatin dosage and administration schedule varied by cancer type (eTable 10 in the Supplement). Treatment protocols of participants with germ cell tumors, osteosarcoma, and hepatoblastoma included fewer potential nephrotoxins (eTable 10 in the Supplement). Treatment protocols of 2 of the 159 participants (1%) included amifostine, and treatment protocols of 13 of the 159 participants (8%) included mesna.

Discussion
To our knowledge, this is the first multicenter, prospective, pediatric study using a KDIGO-based AKI definition to study cisplatin-associated AKI characteristics and risk factors. SCr-AKI was more common at the EV compared with the LV. Concordance between SCr-AKI and eAKI was poor; eAKI was more common than SCr-AKI.
The AKI definition affected the AKI rate. The SCr-AKI rate at the EV was comparable to previous studies (27%-75%), and the LV SCr-AKI rate was comparable to adult estimates. 11,14,16,20,30,[35][36][37] However, most studies 11 Similar to other populations, young age was associated with SCr-AKI. 1,15,38 Interindividual variability in cisplatin pharmacokinetics and pharmacodynamics and standard practice determine dose of cisplatin based on body size in young children may partly explain this finding. 15,22,39 Higher baseline estimated GFR was independently associated with AKI. This somewhat paradoxical finding has been consistently described in pediatric AKI studies, without any satisfying explanations. 1,20 One theory is that children with higher baseline estimated GFR may have lower muscle mass and poor nutrition, although this has never been reported. Another theory is that excessive hydration may dilute SCr; however, this has not been proven, and we did not collect hydration protocol information or mannitol use. In addition, we were unable to assess whether some patients had syndrome of inappropriate antidiuretic hormone secretion, contributing to SCr dilution. Further research should determine the effect of hydration protocols on SCr and cisplatin-associated AKI.
Cisplatin dosage and medications may be associated with AKI. In adults, single-bolus cisplatin infusions are more nephrotoxic than infusions administered for 2 to 5 days. 40,41 In our study, participants with neuroblastoma received the highest total cisplatin cycle dose, and their protocols contained the highest number of nephrotoxins. These factors could contribute to high AKI risk for neuroblastoma and central nervous system tumors. Synergistic nephrotoxic effects with other drugs remain unclear. 1 Genetic differences in cisplatin metabolizing enzymes or alterations in cisplatin metabolism from other medications are possible. 42 In addition, the lower cisplatin infusion dose at the LV may explain the low SCr-AKI rate at LV. We were unable to assess whether there were changes in cisplatin dose, treatment protocol, or fluid administration because of nephrotoxic effects.
Furthermore, despite receiving the highest pre-LV cumulative cisplatin dose, no participants with osteosarcoma developed SCr-AKI, likely explaining associations between low prior cumulative cisplatin dose and SCr-AKI at LV. High cumulative dose is typically associated with nephrotoxic effects near cancer treatment end and may be more important in long-term or chronic nephrotoxicity. 8,19,43 However, the LV infusion dose was higher in participants with SCr-AKI, suggesting that during later individual cisplatin cycles, planned higher dose may be associated with AKI during that cycle. After receiving multiple cisplatin doses, patients may experience unresolved tubular injury, increasing susceptibility of further injury. 44,45 Although further research is needed with larger sample sizes to differentiate the contributions of risk factors and their associations, identified risk factors should be confirmed for identifying future individual cisplatin-associated AKI risk with a goal of early management. It remains unclear how cisplatin dose should be adjusted because of AKI.