Clinical Impact of Maximum Plasma Concentration of Busulfan in Pediatric Patients Undergoing Allogenic Hematopoietic Stem Cell Transplantation

Abstract

Busulfan's narrow therapeutic index and high pharmacokinetic (PK) variability warrant investigation of its exposure–toxicity relationship. We retrospectively analyzed 334 pediatric and young adult patients who underwent allogeneic hematopoietic stem cell transplantation at Seoul National University Children's Hospital between 2009 and 2020 and received once‐daily intravenous busulfan over four days (3‐hour infusion, n = 122; 6‐hour infusion, n = 212). Factors associated with toxicities were identified using multivariable logistic regression, and overall survival (OS) was evaluated by Kaplan–Meier and Cox regression analyses. Despite comparable exposures, the 3‐hour infusion group showed higher C _max (4,260.2 vs 3,028.6 μg/L, P < 0.0001) and more frequent liver function test elevations (13.1% vs 0.5%, P < 0.0001). Receiver operating characteristic analysis identified a C _max cut‐off of 4269.6 μg/L as a factor for hepatotoxicity. The 4‐year OS was superior in the 6‐hour infusion group (82.3% vs 71.9%, P = 0.02) and in the low C _max group (80.8% vs 66.5%, P = 0.01). Cox regression revealed C _max as an independent predictor of OS. A population PK model was developed using NONMEM. A 1‐compartment model incorporating enzyme turnover auto‐inhibition successfully described the busulfan PK profile. Simulations of published once‐daily dosing regimens showed that extending the infusion to 6 hours successfully maintained C _max below the cut‐off threshold. These results show that maintaining busulfan C _max below 4269.6 μg/L is associated with improved safety and survival. Prolonged infusion should be considered when adopting the busulfan once‐daily regimen.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE OF THE TOPIC?

Busulfan has a narrow therapeutic index, for which systemic exposure is closely associated with both efficacy and toxicity. To improve dosing convenience, once‐daily busulfan regimens have been explored.

• WHAT QUESTION DID THIS STUDY ADDRESS?

When systemic exposure is fixed, how do the infusion rate and the resulting maximum plasma concentration (C _max) influence the efficacy and toxicity of busulfan?

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

In pediatric and young adult patients undergoing allogeneic hematopoietic stem cell transplantation, C _max over 4,269.6 μg/L was identified as a significant factor for hepatotoxicity. The incidence of both acute and chronic graft‐vs.‐host disease did not differ between patients receiving 3‐hour and 6‐hour infusions. In addition, overall survival was superior in patients receiving a 6‐hour infusion compared to those receiving a 3‐hour infusion. Simulation by a population pharmacokinetic model demonstrated that compared to the 3‐hour infusion, the 6‐hour infusion markedly increased the proportion of the population achieving the target C _max.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

The integration of real‐world data with modeling enables the development of dose optimization strategies for special populations.

Busulfan, a bifunctional alkylating agent, is widely used in conditioning regimens prior to hematopoietic stem cell transplantation (HSCT), particularly in patients with hematologic malignancies. In pediatric patients, busulfan is also administered for off‐label indications beyond hematologic malignancies. Notably, pediatric patients exhibit marked interindividual variability in busulfan pharmacokinetics (PKs), necessitating a thorough understanding of its PK characteristics in children and underscoring the importance of therapeutic drug monitoring (TDM) to guide individualized dosing strategies. ¹ , ² Busulfan is primarily metabolized through glutathione S‐transferase (GST) mediated conjugation, with approximately 30% of the administered dose excreted in the urine within 48 hours, and less than 1% recovered as unchanged drug. ³ , ⁴ , ⁵ Intravenous administration of busulfan demonstrates dose‐proportional PK characteristics up to 1 mg/kg, with clearance estimates ranging from 2.25 to 2.74 mL/min/kg. ⁴ , ⁵ However, a progressive decline in clearance over the course of repeated administration has been reported, highlighting the need for dynamic dosing adjustments in pediatric patients undergoing multiple‐day regimens. ⁶

Busulfan has a narrow therapeutic index, whose efficacy should be maximized and toxicity minimized through TDM. Currently, the FDA and EMA recommend a four‐times‐daily intravenous busulfan regimen, consisting of 2‐hour infusion every 6 hours, with a target area under the plasma concentration time curve (AUC) of 900 to 1,500 μM·min (3,700 to 6,200 μg·h/L) per administration. ⁴ , ⁵ Higher exposure of busulfan is known to be associated with an increased risk of complications. For instance, the label indicates that the risk of veno‐occlusive disease (VOD) is elevated when the AUC exceeds 1,500 μM·min per administration. ⁴ Beyond regulatory guidelines, various studies have proposed alternative AUC thresholds associated with clinical outcomes. In a multicenter study, the AUC of 78,000 to 101,000 μg·h/L during 4‐day regimen improved event free survival, while both lower and higher exposures were associated with inferior outcomes. ⁷ Also, toxicities and transplant‐related mortality were significantly more frequent in patients with higher doses or AUC. ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² In our institution, we have also investigated the exposure‐efficacy/toxicity relationship of busulfan and currently are performing TDM targeting a cumulative 4‐day AUC of 75,000 μg·h/L. ¹³ , ¹⁴

To improve dosing convenience while maintaining therapeutic exposure, recent studies have evaluated once‐daily busulfan regimens using label‐based AUC targets. Our institution has developed a population PK model in pediatric patients and proposed a dosing nomogram to achieve the target AUC (60,000 to 90,000 μg·h/L) using a once‐daily 3‐hour infusion regimen over four days. ¹⁵ Similarly, other studies have also presented nomograms for once‐daily 2‐hour and 3‐hour infusion regimens. ¹⁶ , ¹⁷ , ¹⁸ Although busulfan related toxicities are known to be associated with systemic exposure (i.e., AUC), maximal plasma concentrations (C _max) may also play a critical role. In general, a higher C _max can lead to an increased risk of adverse events and one study reported that a higher C _max of busulfan may be associated with an increased risk of VOD. ¹⁰ Once‐daily intravenous regimens may lead to higher C _max values compared to the traditional four‐times‐daily dosing schedule. Therefore, we aimed to elucidate the relationship between PK parameters including C _max and AUC, and clinical outcomes. Furthermore, a population PK model was developed to simulate PK profiles and assess how well previously proposed dosing regimens achieve target C _max and AUC thresholds, thereby providing a framework for safer and more effective busulfan dosing strategies in pediatric patients.

METHODS

Study population and transplantation procedure

This study was approved by the Institutional Review Board of Seoul National University Hospital (H‐2510‐118‐1686).

This study included pediatric patients who received intravenous busulfan as a component of the conditioning regimen at Seoul National University Children's Hospital (SNUCH) between 2009 and 2020. Eligible patients were those who underwent allogeneic HSCT and received busulfan over a 4‐day conditioning regimen. To be included in the analysis, patients were required to have complete and analyzable plasma concentration data for all 4 days of busulfan administration. Patients were classified as belonging to one of the following categories: high‐risk malignancies (leukemia, lymphoma, myelodysplastic syndrome, and solid tumors), congenital hematologic diseases, immunodeficiencies, histiocytosis, or other rare genetic diseases. Patients who underwent autologous HSCT were excluded from the study. All patients received targeted busulfan exposure based myeloablative conditioning regimen. Busulfan was combined with fludarabine, fludarabine and etoposide, fludarabine and cyclophosphamide, cyclophosphamide and melphalan, fludarabine and melphalan, cyclophosphamide, or clofarabine and etoposide as their disease status and donor type. Post‐transplant cyclophosphamide was administered to patients who underwent haploidentical HSCT. Regarding the busulfan infusion protocol, patients treated between 2009 and 2013 received busulfan via a 3‐hour intravenous infusion, while those treated from 2014 to 2020 received a 6‐hour infusion.

Definition of clinical outcomes and toxicities

Acute and chronic graft‐vs.‐host disease (GVHD) was diagnosed and graded according to established criteria. ¹⁹ , ²⁰ The diagnosis and severity of VOD were determined based on the European Society for Blood and Marrow Transplantation (EBMT) guideline. ²¹ Pulmonary toxicity was considered infectious if a pathogen was confirmed or could not be excluded, and noninfectious if no pathogen was identified despite comprehensive evaluation, including bronchoalveolar lavage. Pulmonary involvement of chronic GVHD (cGVHD) was not included in the pulmonary complication but was classified under the cGVHD category. Liver function test (LFT) elevation was defined as an increase in AST, ALT, or total bilirubin to grade 2 or higher according to the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0 following busulfan administration. Events were defined as relapse, treatment‐related mortality (TRM), or death due to another cause.

Plasma sample collection and pharmacokinetic analysis for therapeutic drug monitoring

To measure plasma busulfan concentrations, samples were collected immediately after the completion of the intravenous infusion, and at 1, 2, and 4 hours post‐infusion. In some cases, additional samples were also collected prior to the next day's dose. For PK analysis, the administered dose, infusion start and end times, and each blood sampling time point were documented. Samples were collected in EDTA tubes, and concentrations were quantified using Liquid chromatography/Tandem mass spectrometry (LC/MS/MS).

The AUC was calculated using the one‐compartment linear PK model implemented in Phoenix WinNonlin® software (Pharsight Co, Mountain View, CA, USA). Busulfan clearance was calculated as the dose divided by AUC. The recommended daily dose was adjusted to target a cumulative 4‐day AUC of 75,000 μg·h/L, accounting for daily AUC variability. The C _max was considered to be the concentration measured immediately after the end of infusion.

Population pharmacokinetic analysis and simulation

The population PK model of busulfan was developed using NONMEM software version 7.5.1 (Icon Development Solutions, Ellicott City, MD, USA). The baseline model evaluation included one‐compartment and two‐compartment models based on previous studies. ⁶ In order to explain the decrease in busulfan clearance over repeated administration, an enzyme turnover auto‐inhibition elimination model was evaluated. This model describes the elimination of busulfan incorporating the enzyme levels, where the concentration of busulfan affects the turnover rate of the enzyme. The differential equations for the amount of drug in the central compartment (A₁) and the amount of enzyme (A₂) are provided in Eqs 1 and 2 as follows:

$$\frac{\mathrm{d}{A}_1}{\mathrm{d}t}=-\frac{\mathit{CL}}{V}·A_1·A_2$$ (1)

$$\frac{\mathrm{d}{A}_2}{\mathrm{d}t}=K_{\mathit{ENZ}}·\left(1-\frac{\mathit{\text{Conc}}}{\mathit{IC}50+\mathit{\text{Conc}}}\right)-K_{\mathit{ENZ}}$$ (2)

where CL and V describe the clearance and volume of distribution of busulfan, respectively, K _ENZ the enzyme turnover rate, Conc the busulfan concentration, and IC50 the busulfan concentration causing 50% inhibition of enzyme turnover rate. The initial amount of enzyme is set to 1, and its synthesis is inhibited by busulfan in a concentration‐dependent manner, leading to a reduction in enzyme levels over time. This results in a decreased elimination of busulfan from the central compartment.

Demographic data (age, sex, height, weight, body surface area (BSA)), laboratory test results (AST, ALT, total bilirubin, serum creatinine), and the type of dosing regimen (3‐hour infusion, 6‐hour infusion) were evaluated as covariates. Covariates were included in the final model if the addition of a particular covariate decreased the objective function value by more than 3.84 and was supported by physiological plausibility. Model evaluation was conducted using bootstrap analysis, goodness‐of‐fit (GOF) diagnostics, and prediction corrected visual predictive checks (pcVPC).

Using the final population PK model, the previously published nomograms and label‐recommended dosing regimens were simulated. ⁴ , ⁵ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ The demographic characteristics of virtual patients used for the simulation were based on data from patients who received busulfan at SNUCH between 2009 and 2020. The PK parameters of the simulated data were calculated using the ncar package in R software version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria). For each regimen, the proportion of patients achieving the target C _max and AUC was calculated.

Statistical analysis

All patient characteristics and PK parameters were presented as descriptive statistics. Continuous variables between dosing regimens were statistically compared using t‐test, while categorical variables were analyzed using the Chi square test or Fisher's exact test, as appropriate. P‐value of <0.05 was considered statistically significant. PK parameters, demographic characteristics, transplantation conditioning regimen, and type of donor were included as covariates in the multivariable logistic regression model for toxicities. C _max and AUC values were evaluated up to the time point at which the particular toxicity occurred. For instance, if LFT elevation occurred on day 2, the highest busulfan concentration observed up to day 2 and the mean AUC until day 2 were used for the analysis. In contrast, in cases where toxicity did not occur or occurred after the final dose of busulfan, the PK parameters were analyzed using all available concentration data. For parameters showing statistically significant associations with toxicities, receiver operating characteristic (ROC) curves were constructed, and the optimal cut‐off values were identified using Youden's J statistic of the pROC package in R.

The overall survival (OS) and event‐free survival (EFS) were estimated by Kaplan–Meier analysis. A Cox proportional hazard regression model was used for the multivariate analysis of factors affecting OS. The factors included in the analysis were sex, donor type, conditioning regimen, and whether the cumulative AUC and C _max exceeded the cut‐off value as categorical variables, and age as continuous variable. The cumulative incidence of TRM was estimated by competing risk analysis. Statistical analyses were performed using R software version 4.4.0.

RESULTS

Patients' characteristics

A total of 334 patients who received a 4‐day intravenous busulfan regimen with allograft HSCT were included for analysis (Table 1 ). One hundred and twenty‐two patients were administered the 3‐hour infusion regimen, whereas 212 patients received the 6‐hour infusion regimen. Among the study population, only a small proportion of patients were aged ≥18 years, comprising 4 patients in the 3‐hour infusion group (3.3%) and 26 patients in the 6‐hour infusion group (12.3%). Demographic characteristics were generally comparable, although a higher proportion of male patients was observed in the 6‐hour infusion group (63.7% vs. 51.6%, P = 0.0413). Mean age, height, weight, and BSA did not differ significantly between groups (P = 0.0737, P = 0.3856, P = 0.582, and P = 0.5106, respectively). While the distribution of diagnoses was not statistically different, significant differences were observed in the distributions of donor types and conditioning regimens. No statistically significant differences were observed in both the total busulfan dose per BSA (426.9 mg/m² vs. 416.0 mg/m²) and cumulative AUC (75063.0 μg·h/L vs. 74731.9 μg·h/L). Significantly higher C _max in the 3‐hour infusion group was observed compared to the 6‐hour group (4260.2 μg/L vs. 3028.6 μg/L, P < 0.0001).

Table 1.

Patient characteristics

	3‐hour infusion (N = 122)	6‐hour infusion (N = 212)	P‐value
Demographics
Sex, n (%)			0.0413a
Male	63 (51.6)	135 (63.7)
Female	59 (48.4)	77 (36.3)
Age (year), mean (sd) [min – max]	8.8 (5.4) [0.6–22.2]	10.0 (6.6) [0.7–30.3]	0.0737b
Height (cm), mean (sd) [min – max]	126.0 (32.2) [65.2–181.9]	129.2 (33.1) [13.8–184.4]	0.3856b
Weight (kg), mean (sd) [min – max]	32.3 (18.7) [7.4–76.2]	33.5 (19.2) [7.9–91.3]	0.582b
BSA (m2), mean (sd) [min – max]	1.05 (0.43) [0.37–1.92]	1.08 (0.45) [0.26–2.13]	0.5106b
Diagnosis, n (%)			0.0881c
High‐risk malignancy	102 (83.6)	151 (71.3)
Congenital hematologic disease	5 (4.1)	9 (4.2)
Immunodeficiency	7 (5.7)	28 (13.2)
Histiocytosis	3 (2.5)	6 (2.8)
Others	5 (4.1)	18 (8.5)
Donor type, n (%)			<0.0001a
Related	28 (23.0)	35 (16.5)
Unrelated	67 (54.9)	72 (34.0)
Cord	20 (16.4)	8 (3.8)
Haploidentical	7 (5.7)	97 (45.8)
Conditioning regimen, n (%)			<0.0001a
BuFlu	51 (41.8)	47 (22.2)
BuFluVP	60 (49.2)	64 (30.2)
BuFLuCy	1 (0.8)	97 (45.8)
Others	10 (8.2)	4 (1.9)
GVHD, n (%)
aGVHD	45 (36.9)	80 (37.7)	0.9703a
cGVHD	25 (20.5)	33 (15.6)	0.3201a
Toxicity, n (%)
VOD	21 (17.2)	52 (24.5)	0.1556a
Mild	0 (0.0)	5 (2.4)	0.1628c
Moderate	8 (6.6)	17 (8.0)	0.785a
Severe	7 (5.7)	19 (9.0)	0.397a
Very severe	6 (4.9)	11 (5.2)	1a
Pulmonary toxicity	16 (13.1)	18 (8.5)	0.2469b
Noninfectious	4 (3.3)	4 (1.9)	0.4699c
Infectious	12 (9.8)	14 (6.6)	0.3956a
LFT elevation	16 (13.1)	1 (0.5)	<0.0001c
CTCAE Grade 2	10 (8.2)	1 (0.5)	0.0003c
CTCAE Grade 3	6 (4.9)	0 (0.0)	0.0022c
Dose/BSA (mg/m2), mean (sd)	426.9 (112.2)	416.0 (91.2)	0.363b
Pharmacokinetic parameter, mean (sd)
Cumulative AUC (μg/L × h)	75063.0 (3061.8)	74731.9 (2137.8)	0.2924b
C max (μg/L)	4260.2 (683.7)	3028.6 (380.8)	<0.0001b

The P‐values were calculated based on (a) chi‐square test, (b) t‐test, and (c) fisher's exact test.

aGVHD, acute graft‐vs.‐host disease; AUC, area under the curve; BSA, body surface area; Bu, busulfan; cGVHD, chronic graft‐vs.‐host disease; C _max, maximum plasma concentration; CTCAE, common terminology criteria for adverse events; Cy, cyclophosphamide; Flu, fludarabine; GVHD, graft‐vs.‐host disease; LFT, liver function test; MDS, myelodysplastic syndrome, standard deviation; VOD, veno‐occlusive disease; VP, etoposide.

GVHD, toxicities, and survival analysis

Incidence rates of acute and chronic GVHD were similar across both regimens (Table 1 ). Occurrence rates of toxicities were comparable between groups in VOD and pulmonary toxicity. However, LFT elevations occurred significantly more frequently in the 3‐hour infusion group (13.1% vs. 0.5%, P < 0.0001), with corresponding differences in CTCAE grade 2 and 3 elevations (P = 0.0003 and P = 0.0022, respectively) (Figure S1 ).

The 4‐year EFS and OS rate of all patients were 71.9% and 78.3%, respectively. When analyzed by infusion group, the 4‐year EFS rate was 68.8% and 73.7% (P = 0.4), and the OS rate was 71.9% and 82.3% (P = 0.02) in the 3‐hour and 6‐hour infusion group, respectively (Figure 1 ). The 4‐year RFS (relapse‐free survival) rate for malignant disease did not show a significant difference between the 3‐hour and 6‐hour infusion groups (82.9% vs. 75.1%, P = 0.1).

Figure 1.

Survival analysis by infusion duration and maximum plasma concentration (C _max) of busulfan: (a) Event‐free survival (EFS) of all the patients, (b) Overall survival (OS) of all the patients, (c) EFS of patients by infusion duration, (d) OS of patients by infusion duration, (e) EFS of patients in low and high C _max group, (f) OS of patients in low and high C _max group. HSCT, hematopoietic stem cell transplantation.

Relationships between clinical outcomes and busulfan pharmacokinetic parameters

When the number and proportion of GVHD and toxicities were evaluated according to the distribution of C _max, and the cumulative AUC, higher C _max values were associated with increased occurrences of pulmonary toxicity and LFT elevation (Figure 2 ). The cumulative AUC of all patients was within the therapeutic range defined by the FDA label (60,000 to 90,000 μg·h/L), and no clear association was observed between AUC and the incidence of GVHD.

Figure 2.

The relationship between graft‐vs.‐host disease (GVHD) and toxicities by (a) maximum plasma concentration and (b) area under concentration of busulfan. The green represents patients without toxicities, and the blue shows patients with toxicities. aGVHD, acute graft‐vs.‐host disease; cGVHD, chronic graft‐vs.‐host disease; LFT, liver function test; PULT, pulmonary toxicity; VOD, veno‐occlusive disease.

Logistic regression analysis identified C _max and mean AUC as significant covariates associated with LFT elevation (P = 0.0116 and P = 0.0043, respectively). In the addition, conditioning regimen “Others” was identified as a significant covariate for VOD (P = 0.0141) and infectious pulmonary toxicity (P = 0.0084). Demographic characteristics and type of donor were not identified as significant predictors of VOD, pulmonary toxicity, or LFT elevation. ROC curve analysis was performed and optimal cut‐off values were determined to evaluate the association between C _max, mean AUC, and LFT elevation. The area under the ROC curve was 0.817 for C _max, compared to 0.64 for mean AUC, indicating a stronger discriminatory power of C _max in relation to LFT elevation (Figure 3 ). The optimal C _max cut‐off value was 4,269.6 μg/L, with a sensitivity of 0.71 and a specificity of 0.86. The optimal AUC cut‐off value was 19,385.0 μg·h/L with a sensitivity of 0.53 and a specificity of 0.9.

Figure 3.

The logistic regression curve of (a) maximum plasma concentration (C _max) and (b) mean AUC of Busulfan up to the time of LFT elevation and the receiver operating characteristic (ROC) curve of (c) C _max and (d) mean AUC of Busulfan up to the time of LFT elevation vs. LFT elevation. AUC, area under the plasma concentration time curve; LFT, liver function test.

The 4‐year EFS rate was 64.9% and 73.1% in the high and low C _max groups, respectively (P = 0.2). The 4‐year OS rate was 66.5% in the high C _max group and 80.8% in the low C _max group, showing a statistically significant difference (P = 0.01, Figure 1 ). The cumulative incidence of TRM was 22.8% and 9.9% in the high and low C _max groups, respectively (P = 0.0067). Cox regression revealed that C _max and cord donor were associated with OS (Table 2 ).

Table 2.

Univariate and multivariate analyses of predictive factors for overall survival

	No. (%)	Univariate		Multivariate
		P	Hazard ratio	P	Hazard ratio
Sex
Male	199 (59.6)
Female	135 (39.2)	0.332	0.79	0.170	0.70
Age	344 (100.0)	0.169	0.97	0.397	0.98
C max
C max <4,269.6 μg/L	277 (82.9)
C max >4,269.6 μg/L	57 (16.6)	0.013	1.94	0.018	2.03
Donor type
Related	63 (18.9)
Unrelated	139 (41.6)	0.718	1.14	0.737	1.13
Cord	28 (8.4)	0.010	3.10	0.020	2.94
Haploidentical	104 (31.1)	0.151	1.71	0.076	1.99
Cumulative AUC
Cumulative AUC < 76,000	247 (74.0)
Cumulative AUC > 76,000	87 (26.0)	0.168	1.42	0.056	1.65

AUC, area under the curve; Bu, busulfan; C _max, maximum plasma concentration; Cy, cyclophosphamide; Flu, fludarabine; VP, etoposide.

Population pharmacokinetic analysis and simulation result

The population PK model for busulfan was developed using 5,357 valid plasma concentrations collected from 334 patients. A one‐compartment model with an enzyme turnover auto‐inhibition elimination model adequately described the PK characteristics of busulfan (Table 3 , Figure S2 ). The half‐maximal inhibitory concentration (IC50) was estimated to be 3,250 μg/L which falls within the observed plasma concentration range. BSA was identified as a significant covariate for both volume of distribution and clearance, while age was also found to influence clearance. Age was identified as a significant covariate describing the maturation of GST activity, and 50% maturation was predicted to be reached at 0.37 years. The model was validated using GOF plots and pcVPC, which demonstrated robust performance across a wide range of observed data (Figure S3 ).

Table 3.

Parameter estimation of the final pharmacokinetic model of busulfan

Parameter	Final model
	Estimate (RSE)	Bootstrap median (95% CI)
Structural model
Va; Volume of distribution (L)
${\theta }_1$; typical V value with BSA 1.73 m2	42 (2%)	42.1 (40.8–43.3)
${\theta }_2$; BSA exponent for V	1.21 (3%)	1.22 (1.16–1.26)
CLb; Clearance (L/h)
${\theta }_3$; typical CL value with BSA 1.73 m2	11 (2%)	11.05 (10.78–11.35)
${\theta }_4$; Age reaching 50% of adult maturation	0.374 (13%)	0.375 (0.268–0.492)
IC50; Half maximal inhibitory concentration (μg/L)	3,250 (8%)	3,245 (2752–3,620)
K ENZ; Enzyme turnover rate (h−1)	0.0404 (17%)	0.0403 (0.0327–0.0512)
Inter‐individual variability (IIV)
IIV for V (%)	17.2 (9%)	17 (14.9–20.2)
IIV for CL (%)	22.1 (5%)	21.7 (19.7–24.2)
Correlation between IIV on CL and V	0.0226	0.0219 (0.0153–0.0316)
Residual error
Proportional residual error	0.103 (3%)	0.102 (0.097–0.109)

CI, confidence interval as 5th and 95 h percentiles derived from a bootstrap analysis of 100 resampled datasets; RSE, relative standard error.

^a V = ${\theta }_1\times {\left(\frac{\mathrm{BSA}}{1.73}\right)}^{{\theta }_2}$.

^b CL = ${\theta }_3\times \left(\frac{\mathrm{BSA}}{1.73}\right)\times \left(1-{\mathrm{e}}^{\left(-\frac{0.693}{{\theta }_4}\times \mathrm{Age}\right)}\right).$

Based on the final population PK model, simulations were conducted using label‐recommended dosing regimens and published once‐daily dosing nomograms, and the probability of achieving the target C _max and AUC was assessed. The target C _max was defined using the optimal cut‐off value derived from the ROC curve analysis (4,269.6 μg/L), while the target AUC was referenced from the value recommended in the FDA label (60,000 to 90,000 μg·h/L). When simulating the EMA and FDA‐recommended regimens (i.e., every 6 hours dosing), all the simulated subjects achieved the predefined target C _max. In contrast, other regimens (i.e., once daily dosing) demonstrated lower attainment rates, ranging from 40.6% to 84.6% (Figure 4 , Table S1 ). Modifying these regimens to a 6‐hour infusion improved target C _max attainment to 86.0% to 97.7%. For AUC targets, approximately 50% of patients achieved the goal range under existing dosing strategies, and this proportion remained similar even after adjusting the infusion duration to 6 hours.

Figure 4.

Simulated maximum plasma concentration (C _max) and area under the curve (AUC) after intravenous administration of busulfan in pediatric and young adult patients under published regimens. The dashed red lines represent boundaries of the targeted C _max (4269.55 μg/L) and AUC (60,000 μg/L*h, 90,000 μg/L*h), respectively. EMA; 2‐hour infusion of 1.0 mg/kg for a body weight of <9 kg, 1.2 mg/kg for 9–16 kg, 1.1 mg/kg for 16–23 kg, 0.95 mg/kg for 23–34 kg, and 0.8 mg/kg for >34 kg, FDA; 2‐hour infusion of 1.1 mg/kg for a body weight of ≤12 kg, and 0.8 mg/kg for >12 kg, Buffery et al.; 2‐hour infusion of 4.8 mg/kg for a body weight of 10–16 kg, 4.4 mg/kg for 17–18 kg, 4.0 mg/kg for 19–22 kg, 3.6 mg/kg for 23–25 kg, and 3.2 mg/kg for >26 kg, Lee et al.; 3‐hour infusion of 80 mg/m2 for <1 year of age, and 120 mg/m² for ≥1 years, Rhee et al.; 3‐hour infusion of BSA and dosing day based regimen by 9 age categories once daily for 4 days, Yin et al.; 3‐hour infusion of 3.2 mg/k.

DISCUSSION

As far as we know, this is the first study to report large‐scale clinical data comparing 3‐hour and 6‐hour infusion durations of intravenous busulfan in pediatric and young adult patients undergoing HSCT. The major strength lies in the comprehensive evaluation of both survival outcomes and toxicity profiles with PK parameters. Furthermore, as the AUC was controlled within the therapeutic range, this allowed for the investigation of factors other than AUC on clinical outcomes.

Numerous studies have emphasized the importance of maintaining busulfan exposure within a defined AUC range to reduce GVHD and toxicities. The association between elevated busulfan AUC and the risk of VOD has been introduced in previous studies. ⁷ , ⁸ However, as all patients in our study remained within the target AUC range, no correlation was observed between AUC and VOD incidence. Pulmonary toxicity is another well‐recognized complication of high‐dose busulfan exposure. ⁹ , ¹² Although not statistically significant, a higher incidence of pulmonary toxicity was observed in the 3‐hour infusion group compared to the 6‐hour infusion group, and a positive correlation between C _max and pulmonary toxicity incidence was identified. Patients who received regimen “Others” showed a higher incidence of VOD and infectious pulmonary toxicity (35.7% and 28.6%, respectively). Nevertheless, as the number of patients in regimen “Others” was small and accounted for only 4.2% of the studied population, the possibility that this result occurred by chance cannot be ruled out. Further studies with larger patient cohorts are warranted to identify factors influencing VOD and pulmonary toxicity.

Compared to AUC, the relationship between C _max and busulfan related toxicities has been relatively underexplored. One study suggested a C _max higher than 1,880 μg/L as a risk factor for VOD using classification and regression tree analysis. ¹⁰ However, in our study, no increasing trend in VOD incidence was observed with higher C _max values. Moreover, despite all patients in our study having C _max values exceeding 1,880 μg/L, the incidence of VOD was lower than that reported in the aforementioned study. These findings suggest that setting the C _max cut‐off value at 1,880 μg/L may not be appropriate. Another study comparing twice‐daily and once‐daily busulfan regimens reported elevations in AST and ALT levels among patients with higher C _max values. ²² Likewise, a significantly higher incidence of LFT elevation was observed in our study in the 3‐hour infusion group compared to the 6‐hour infusion group. Although both C _max and mean AUC were identified as statistically significant covariates for LFT elevation in the logistic regression analysis, C _max demonstrated superior discriminatory performance compared to mean AUC, as reflected by a higher area under the ROC curve. These findings suggest that while both parameters are associated with hepatotoxicity risk, C _max may be a more clinically relevant predictor for identifying patients at risk of LFT elevation. The cut‐off value calculated using the ROC curve was determined to be 4,269.6 μg/L. In the cox regression analysis, a C _max of 4,269.6 μg/L was evaluated as a statistically significant covariate of OS, suggesting that this value may also hold clinical relevance. In contrast, although a mean AUC greater than 19,385.0 μg·h/L was identified as a covariate related to LFT elevation, this value lies within the optimal AUC range suggested by the drug label and previous studies. ⁴ , ⁵ , ⁷ Moreover, its lower predictive performance compared to C _max in the ROC analysis suggests that it holds less clinical significance.

When comparing survival outcomes, no significant differences in EFS or RFS were observed between the 3‐hour and 6‐hour groups, whereas the OS rate was 10.4% higher in the 6‐hour infusion group, with statistical significance. Similarly, the low C _max group demonstrated a significantly higher OS rate, with a 14.3% greater rate compared to the high C _max group. In this study, the 4‐day cumulative AUC was targeted at 75,000 μg·h/L, resulting in similar BSA‐based doses. Consequently, a shorter infusion time indicated a higher C _max, and the 3‐hour infusion group showed a significantly higher C _max compared to the 6‐hour infusion group. Notably, when patients were stratified by the optimal C _max cut‐off value defined in this study, all patients in the 6‐hour group had C _max below the threshold, whereas only 53.3% of patients in the 3‐hour group fell below it. Comparison of TRM by C _max level revealed that TRM was significantly higher in the high C _max group. These findings suggest that the difference in OS between the 3‐hour and 6‐hour groups, despite similar relapse rates, may be explained by the predominance of high‐C _max patients in the 3‐hour group. In our multivariate Cox regression analysis, a higher C _max remained significantly associated with inferior outcomes (Hazard ratio (HR) 2.03, P = 0.018). Among donor types, cord blood transplantation was significantly associated with a higher risk compared with related donor transplantation (HR 2.94, P = 0.020). In contrast, other covariates such as age, sex, and cumulative AUC of busulfan did not demonstrate statistical significance. These findings suggest that C _max and donor source may play more prominent roles in post‐transplant prognosis than demographic factors.

Using the population PK model, we simulated various dosing regimens proposed in regulatory guidelines and the literature. The EMA and FDA recommend a dosing regimen of relatively smaller doses administered four times daily with a 2‐hour infusion, while several once‐daily regimens reported in prior studies typically employ 2‐ to 3‐hour infusion durations with relatively larger doses. Our simulations showed that while all patients maintained C _max below the predefined threshold when busulfan was administered according to the four‐times‐daily regimens, only 40% remained below the threshold C _max with the once‐daily regimens. However, when the once‐daily regimens were modified to extend infusion duration to 6 hours, the proportion of patients maintaining C _max below the threshold markedly improved to a mean of 93.6%, with a range of 86.0% to 97.7%. These findings emphasize the critical influence of infusion duration on C _max and support the need for a longer infusion duration when applying once‐daily busulfan dosing regimens. In our institution, the 6‐hour infusion of busulfan is initiated at 11 PM and completed by 5 AM, followed by laboratory assessments to evaluate patient safety. Subsequently, a clinical pharmacology consultation is conducted to determine the amount of busulfan for the following day. This example illustrates that the 6‐hour infusion regimen is feasible in real‐world clinical settings.

Several population PK models of busulfan have been proposed to account for the decrease in clearance over the course of repeated administration, incorporating time as a covariate or using day‐specific factors. ⁷ , ²³ , ²⁴ , ²⁵ However, estimation of decreased busulfan clearance using these approaches is considerably affected by the dataset used for model building and may have limited predictability when extrapolated to alternative infusion schedules or doses. In contrast, we utilized an enzyme turnover model to provide a physiological explanation for the concentration dependent changes in enzyme activity. The enzyme turnover model incorporates the metabolic enzyme compartment whose synthesis or elimination rate is affected by drug concentration. In the present model, the synthesis rate of the enzyme representing GST was estimated to decrease as busulfan levels increased. Due to the limited information available on GST parameters in humans, it is challenging to directly compare the parameter estimates obtained from the PK model with actual values. However, given that the turnover half‐life of CYP450 enzymes is reported to range from 10 to 140 hours and GST proteins are relatively short‐lived, the enzyme turnover rate estimated in this model appears reasonable. ²⁶ , ²⁷ , ²⁸ In addition, the maturation profile of GST with respect to age indicated that 50% maturation was achieved at approximately 0.37 years, which is consistent with previous reports suggesting that GST reaches maturation at an early age. ²⁹ Taken together, the PK model of busulfan developed in this study was considered to be physiologically plausible and appropriate for simulation across various dosing regimens.

This study has several limitations. First, it was conducted as a retrospective study, which inherently limits the ability to control confounding variables and establish causal relationships. Future prospective studies are warranted to validate the observed associations between PK parameters including C _max and clinical outcomes. Second, given the extended study period, temporal improvements in therapeutics or supportive care may have influenced outcomes, potentially introducing bias. Third, there was an imbalance in sample size between the 3‐hour and 6‐hour infusion groups, as well as unequal sex distribution, which may have introduced bias and limited the interpretability of comparative analysis. Lastly, aside from LFT elevation, no other complications demonstrated statistically significant differences between the 3‐hour infusion and 6‐hour infusion groups, which may reflect limited statistical power to detect less common or more subtle toxicities. These limitations underscore the need for further well‐controlled investigations to confirm our findings and refine individualized busulfan dosing strategies in pediatric populations undergoing HSCT.

In conclusion, our findings support the clinical benefit of a prolonged 6‐hour infusion of intravenous busulfan, particularly in the context of once‐daily dosing. The C _max cut‐off of 4,269.6 μg/L was identified as a meaningful predictor of both toxicity and survival, providing a rationale for more individualized dosing strategies.

FUNDING

No funding was received for this work.

CONFLICT OF INTEREST

The authors declared no competing interests for this work.

AUTHOR CONTRIBUTIONS

S.B., J.Y.C., K.T.H., and K‐S.Y. wrote the manuscript; S‐J.R. and H.J.K. designed the research; S.B. and J.Y.C. performed the research; K.T.H., K‐S.Y., B.K.K., H.J.P., S‐J.R., and H.J.K. analyzed the data.

Supporting information

Supporting Information