Patient selection and enrollment

Patients at a single center were selected from those enrolled as part of an ongoing study of pharmacokinetics and pharmacodynamics of beta-lactam antibiotics in PICU patients. The study was approved on 9/2/2018 by the Cincinnati Children’s Institutional Review Board under study number 2018–3245 as “BetaLactamPKSepsis.” It was granted a waiver of consent and was conducted in compliance with the ethical standards of the responsible committee on human experimentation and with the Helsinki Declaration of 1975. Briefly, patients who received at least one dose of cefepime in the PICU during the study collection periods were sequentially screened for inclusion. Patients were included if they met the following criteria: (1) at least one dose of cefepime administered in the PICU with a full 24 h of pharmacokinetic data, (2) PICU stay of duration ≥ 24 h, and (3) at least two opportunistic plasma samples available and processed within 72 h of collection. Once enrolled, patients could not be re-enrolled during the same hospitalization but were eligible for repeat enrollment with subsequent admissions. Prior or concurrent administration of other antibiotics did not exclude patients from enrollment. Patients who were on extracorporeal support such as extracorporeal membrane oxygenation (ECMO) or continuous kidney replacement therapy (CKRT) at the time of sample collection were excluded from this analysis. To focus on the impact of acute kidney injury, patients with baseline eGFR < 60 mL/min/1.73 m2 (i.e., CKD stages 3–5) as identified by chart review of the 6 months prior to study enrollment were excluded in the primary analysis. Demographics of the population with five CKD patients included may be found in Supplemental Table S1.

Sample collection and processing

Blood samples were obtained using scavenged residual samples with an opportunistic sampling strategy, as described previously [26, 27]. After completion of clinically ordered testing, residual blood collected in ethylenediaminetetraacetic acid (EDTA) or lithium-heparin tubes was requisitioned from the clinical laboratory. Samples collected within 30 min of initiation of cefepime administration were excluded, as a 30-min infusion is standard at our institution. Samples were requested daily until cefepime discontinuation, hospital discharge, or death, up to a maximum of 7 days. For the second 50 patients (enrolled 9/2021–12/2021), sample collection was discontinued upon transfer out of the ICU, up to a maximum of 7 days. Samples were stored at 4 °C for up to 72 h from the time of collection until processing and then centrifuged for 10 min at 2060 × g, after which supernatant was isolated and frozen at − 80 °C until cefepime concentration measurement. Total cefepime concentrations were measured using linear gradient high-performance liquid chromatography as described previously [26].

Chart review

Electronic medical records (EMR) were reviewed, and clinical data were collected including patient demographics, medical conditions, features of hospitalization, and identified infections. Body mass index (BMI) categories were assessed based on 2000 Center for Disease Control weight-for-length growth charts for children 0–3 years, BMI growth charts for patients 3–20 years, and a Center for Disease Control BMI calculator for patients over 20 years. Following CDC categorizations, underweight was defined as patients < 5th percentile for age, while overweight/obese was defined as > 85th percentile for age. All creatinine measurements were obtained via chart review. Creatinine measurements used for clinical care are performed by our CLIA- and CAP-certified Cincinnati Children’s Clinical Laboratory via a validated biochromatic assay. Baseline serum creatinine was defined as the lowest recorded creatinine within 3 months prior to hospitalization. For patients with prolonged hospitalization, the lowest creatinine within current hospitalization prior to ICU admission was used. For patients without prior creatinine data in our system, the baseline was the lowest value of either lowest creatinine during current hospitalization or imputed creatinine for a glomerular filtration rate (GFR) of 120 mL/min/1.73 m2 using the modified bedside Schwartz equation with k = 0.413 [28, 29]. Estimated GFR was calculated using Chronic Kidney Disease in Children (CKiD U25) calculating equations via the published shiny app [30]. Daily patient data were reviewed starting from the first day on which cefepime was administered in the ICU and continuing until cessation of cefepime sample monitoring.

Modeling

Bayesian PK analysis was performed with MwPharm +  + (Mediware, Prague, Czech Republic). Using a previously published two-compartment model with allometric scaling describing cefepime population PK in critically ill pediatric patients [14] along with the measured cefepime concentrations, Bayesian estimation was used to generate a concentration–time profile for each patient. Estimates for the clearance, volume of distribution, area under the concentration–time curve (AUC), and minimum/maximum concentrations in each dosing interval (Cmin/Cmax) were extracted. The volume of distribution was linearly scaled, and clearance was allometrically scaled with a power of 0.75 to patient body weight, as described previously [14].

Acute kidney injury analysis

AKI stage for each day of study observation was staged using Kidney Disease Defining Global Outcomes (KDIGO) creatinine criteria only [31]. The highest serum creatinine measured within each 24-h period was chosen.

Dosing and dose adjustments

The electronic health record was used to collect the cefepime doses ordered for all patients with severe (KDIGO stages 2–3) AKI. Cefepime dosing and adjustments were managed by the clinical team and were not influenced by study inclusion. Institutional dosing recommendations are included in Supplemental Table S3 [32]. Intra-operative dosing was not reported due to different medication reporting contexts within the EMR.

Statistics and analysis

A pre-dose trough concentration (Cmin) of 30 mg/L was chosen as the threshold for “elevated” cefepime exposure based on a previous adult analysis [22]. Continuous data were compared using Mann–Whitney rank sum tests, and categorical data were compared using chi-squared analysis. For analysis of the AKI effect on cefepime PK parameters and exposure, groups were analyzed by one-way ANOVA for continuous data and chi-squared analysis for categorical data, with pairwise comparisons using Dunn’s methods. For these analyses, patients were categorized by the highest degree of AKI achieved at any point during the study observation period, with the categories being no AKI, stage 1 AKI, and severe AKI (stages 2 and 3). To account for potential confounders, a mixed-effect model was used to investigate the effect of the AKI stage on cefepime exposure measured both by Cmin and AUC24. Three patient-level variables were added to the mixed model: age in years, Pediatric RISk of Mortality (PRISM) score, and baseline eGFR. As the data were not normally distributed, analysis was performed with both original and log-transformed data. The outcomes of the analysis were similar for both, so non-transformed results are presented for ease of interpretation. Statistical calculations were performed using SigmaPlot V.15 (Inpixion) and SPSS (IBM).