Prospective Evaluation of Thiotepa Pharmacokinetics in Children and Young Adults Undergoing Allogeneic or Autologous Hematopoietic Cell Transplantation

Abstract

Thiotepa (TT) is often included in high‐dose chemotherapy prior to hematopoietic cell transplantation (HCT); however, exposure–response relationships of TT and its primary metabolite, N,N',N''‐triethylenethiophosphoramide (TEPA), remain undefined. This was a prospective study enrolling children and young adults undergoing autologous and allogeneic HCT from multiple centers. Time–concentration data were modeled using nonlinear mixed effects modeling along with clinical covariates potentially impacting pharmacokinetics. Cumulative area‐under‐the‐curve (cAUC) for TT and TEPA were derived using post hoc individual estimates for clearance. The combined cAUC_TT and cAUC_TEPA were also evaluated, reflecting total alkylating exposure (cAUC_total). The ratio of cAUC_TT/cAUC_TEPA was used to evaluate the efficiency of metabolic conversion of TT to TEPA. A total of 91 subjects were included with a median (range) actual body weight and age of 16.7 kg (4.6–110) and 4.5 years (0.23–22.5), respectively. The best fit model was a two‐compartment structure for TT, and a one‐compartment structure for TEPA. The conversion of TT to TEPA was found to be saturable at higher doses and may have important implications for optimizing exposure. Model‐based dosing, which accounts for age and body size, results in more predictive exposure as compared to conventional mg/kg dosing, particularly in the very young and obese. Next steps include evaluation of pharmacogenomics and exposure–response relationships to optimize dosing of TT in children.

Introduction

Thiotepa (TT) is an alkylating agent with potent immunosuppressive properties routinely incorporated into high‐dose combination conditioning prior to autologous and allogeneic hematopoietic cell transplantation (HCT) to support stem cell engraftment. Despite increasing interest and application of TT in both the pediatric and allogeneic haploidentical HCT setting, the pharmacokinetics (PK) of TT remain largely uncharacterized, particularly in infants and young children. ¹ , ² , ³ , ⁴ Understanding the relationship between dose–exposure–response is critical for the optimal use of therapeutics and particularly impactful for drugs like TT with a narrow therapeutic index and high inter‐subject variability. Unfortunately, response to TT therapy is highly variable, and the therapeutic index remains poorly defined, allowing acute and long‐term drug‐related toxicity to persist as a prominent clinical problem.

In the setting of high‐dose combination chemotherapy and HCT, TT‐containing regimens have been associated with sinusoidal obstruction syndrome of the liver, pulmonary complications, and severe mucositis. ³ More recently it has been demonstrated that when included as part of pre‐transplant conditioning prior to HCT, TT increased the incidence of transplant associated thrombotic microangiopathy (TA‐TMA). ⁵ Studies in mice evaluating neurological effects of TT demonstrate that long‐term cognitive deficits may also result. ⁶ Work by our group and others have shown PK‐guided dosing of pre‐transplant conditioning to target optimal exposure pre‐transplant conditioning agents can significantly improve clinical outcomes in the setting of pediatric HCT including the alkylators busulfan and melphalan, along with the lymphodepleting agents fludarabine and anti‐thymocyte globulin. ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³

TT undergoes metabolic conversion in the liver by the cytochrome P450 (CYP) enzymes, CYP2B6 and CYP3A4, to the primary active metabolite N,N',N'' ‐triethylenethiophosphoramide (TEPA); both TT and TEPA possess alkylating activity. ¹⁴ TT and TEPA are reportedly equivalent in alkylating potential and the resulting immunosuppressive effects on the bone marrow and circulating T and NK cells. ¹⁴ Further detoxification of thiotepa and TEPA is mediated by glutathione conjugation. ¹⁴ Although inconsistent, dose‐dependent kinetics of thiotepa and TEPA have been described over doses ranging from 180 to 900 mg/m², suggesting saturable clearance. ¹⁴ Renal elimination of TEPA is limited, accounting for less than 10% of an administered dose, with a single case report describing altered PK in the presence of moderate renal insufficiency. ¹⁴ , ¹⁵ , ¹⁶

The overarching goal of this work is to develop personalized dosing strategies for the use of TT in pediatric HCT to ensure safe and efficacious drug therapy in all patients. Patient‐specific covariates are expected to contribute substantially to inter‐individual variability in TT exposure in pediatric HCT recipients and thus clinical response and drug‐related toxicity. The primary objective of this work was to characterize the PK of systemic TT and TEPA in pediatric patients receiving TT as part of their preparative regimens prior to autologous and allogeneic HCT.

Methods

Study Population

This was a prospective PK study of thiotepa in children and young adults who underwent autologous or allogeneic HCT for a variety of malignant and nonmalignant pediatric disorders. Patients were eligible to participate in PK analysis if they were between 0 and 26 years of age, met protocol‐specific eligibility criteria for transplantation, and were to undergo an allogeneic or autologous HCT that included intravenous thiotepa as part of the conditioning regimen, irrespective of the dose or dosing regimen. Patients who received thiotepa given as a one time single dose or as several doses throughout conditioning were eligible to participate.Thiotepa PK data were collected between 2017 and 2023 at the University of California San Francisco Benioff Children's Hospital (n = 54), the University of Minnesota Masonic Children's Hospital (n = 12), UCSF Benioff Children's Hospital Oakland (n = 7), and Rady Children's Hospital of San Diego (n = 9). Additionally, nine infants were included from the multi‐center study evaluating thiotepa as part of pre‐transplant conditioning in infants newly diagnosed with severe combined immunodeficiency (PTCTC NMD 1801, CSIDE, NCT01346150). All local institutional review boards approved this study and written informed consent/assent was obtained from all patients. The study was registered at ClinicalTrials.gov as NCT03609840.

PK Sampling

All patients underwent PK sampling following a single dose of TT. For patients undergoing double or triple tandem autologous transplants, PK was performed following a single occasion only. Thiotepa was administered intravenously over 1 to 2 h in all patients depending on the type of transplant (autologous or allogeneic). For the first 17 subjects that consented to PK analysis, intensive PK was performed. For a 1‐h infusion (allogeneic conditioning), blood samples for the quantification of TT and TEPA were collected at 1, 2, 4, and 24 h after the start of TT infusion. For a 2‐h infusion (autologous conditioning) blood samples were collected at 2, 3, 4, 6, and 24 h. In the remaining subjects (n = 74) a limited sampling strategy was implemented with one PK collection 0–2 and 2–24 h post end of infusion. All blood samples for each plasma PK time point were collected in a 0.5 mL spray‐dried K₂EDTA microtainer tube, spun at 3000–4000 rpm for 10 min in a refrigerated centrifuge at 4°C, and frozen at −70°C within 30 min of drawing.

Bioanalysis of TT and TEPA

Plasma samples were analyzed for TT and TEPA using a validated liquid chromatography tandem mass spectrometry method, as previously described. ¹⁷ The TT and TEPA assay was linear in the range of 2.5 to 2000 ng/mL. During sample analysis, 17 sets of low‐, medium‐, and high‐quality control samples and 19 extra‐high‐quality control samples (for dilution) were analyzed. The mean accuracies (mean ± coefficient of variation) for TT were 98.9% ± 13.3%, 102.3% ± 7.6%, 94.9% ± 8.6%, and 99.9% ± 8.3% at low‐, medium‐, high‐, and extra‐high‐quality control levels, respectively. For TEPA the mean accuracies (mean ± coefficient of variation) were 104.0% ± 10.9%, 105.3% ± 8.6%, 94.3% ± 10.0%, and 105.6% ± 5.9% at low‐, medium‐, and high‐quality control levels, respectively.

Population PK Analysis

A nonlinear mixed effects modeling approach was used to describe the time course of TT and TEPA plasma concentrations utilizing NONMEM (v7.6) using the first‐order conditional estimation with interaction method to estimate PK parameters and variability. Data visualization and manipulation were performed in R (v4.3), and Perl‐speaks‐NONMEM was also used for tools including vpc and scm. For TT and TEPA concentrations which fell below the lower limit of quantification, the actual values were reported by the lab and utilized in model development. ¹⁸ Model selection was informed by objective function values, goodness‐of‐fit plots, a visual predictive check (VPC), and model estimation stability. Both TT and TEPA concentrations were modeled on log‐transformed concentration values and residual unexplained variability proportional to the concentration measurement was captured (additive on log‐scale).

Covariate Modeling

Clinical data were collected on the day of PK sampling prior to administration of TT to assess the influence of patient‐specific factors on TT and TEPA PK parameters. Continuous covariates included age, height, total body weight, body surface area, fat‐free mass (FFM) as estimated by the Al‐Sallami method, ¹⁹ creatinine clearance (CLcr), and markers of liver function including aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), and total bilirubin. In patients 0–16.99 years of age, CLcr was estimated by the modified bedside Schwartz equation (mL/min/1.73 m²) and, in young adults (17 years and older), by the Cockcroft–Gault equation using actual body weight (mL/min). Categorical covariates included biological sex, race, and ethnicity. Race was evaluated as indicated by the clinical site as either Asian, Black, Native American or Alaska Native, White, Other, or Not Recorded. Ethnicity was evaluated as either Hispanic or Latino, Not Hispanic or Latino, or unknown. Covariate analysis was performed using a stepwise forward additive approach followed by a stepwise backward elimination approach with an alpha of 0.05 and 0.001, respectively.

Simulations for AUC‐Specific Dosing

Based on the final population PK model, individual doses of TT were simulated to achieve the median cumulative area‐under‐the‐curve cAUC_total (cAUC_TT × number of doses) + cAUC_TEPA × number of doses) of the observed data for the corresponding total number of doses administered over the course of pre‐transplant conditioning and reflecting total alkylating potential of both parent and metabolite. Two complementary simulation analyses were conducted: (1) population‐level dose curves characterizing the model‐predicted dose as a function of patient body weight across the clinically relevant weight spectrum, and (2) a stochastic clinical simulation comparing conventional flat‐dose regimens to therapeutic drug monitoring‐guided model‐informed precision dosing (TDM‐MIPD) using the mipdeval R package. ²⁰ Run evaluation of iterative predictive performance based on historical datasets. R package version 0.1.0.9000. https://github.com/InsightRX/mipdeval).

Results

Patient Demographics

A total of 91 subjects completed PK assessments and were used for PK model building. Patient demographics are presented in Table 1. Among the study subjects, the median age of patients was 4.5 years (0.23–22.5 years) with a median body weight of 16.7 kg (4.6–110 kg).

Table 1.

Demographics and Summary of Thiotepa Dosing Regimens

	Allogeneic Transplant	Autologous Transplant	Complete Dataset
No. of Subjects	48	43	91
Sex (male/female)	26/22	25/18	51/40
Age (years)	8.4 (0.23–22.5)	3.3 (0.7–14.5)	4.5 (0.23–22.5)
Actual body weight (kg)	27.7 (4.6–110)	14.9 (7.7–32.9)	16.7 (4.6–110)
Body surface area (m2)	1 (0.26–2.23)	0.62 (0.38–1.19)	0.7 (0.26–2.23)
Race
Asian	6	5	11
Black/African American	0	2	2
Multiple	7	11	18
Native American/Alaskan	1	0	1
Other	4	1	5
Unknown	9	2	11
White	21	22	43
Ethnicity
Hispanic or Latino	14	14	28
Not Hispanic or Latino	29	27	56
Unknown	5	2	7
Thiotepa dose
4 mg/kg	9	‐	9
5 mg/kg	38	‐	38
7.5 mg/kg	1	‐	1
10 mg/kg	‐	14	14
300 mg/m2	‐	28	28
Disease
Malignant	33	43	76
Non‐malignant	15	‐	15
Regimens
Bu/Flu/TT	19	‐	19
Bu/CY/TT	4	‐	4
Mel/Flu/TT	16	‐	16
Mel/Flu/Clo/TT	4	‐	4
Carbo/Etop/TT	‐	5	5
Carbo/TT	‐	9	9
CY/TT	‐	27	27
Other a	5	2	7

^a Includes Flu/TT or TT/TBI.

Final Population PK Model

The final dataset included a total of 210 and 235 quantifiable concentrations for TT and TEPA, respectively. The final model parameters are presented in Table 2. The final model found to best fit these data was a two‐compartment structural model describing TT, a nonlinear conversion from TT to TEPA, and a one‐compartment structural model describing TEPA. Covariates retained in the final model were age (via maturation functions applied to clearance and volume of distribution parameters) and FFM (allometric scaling with exponents of 0.75 for clearance and 1.0 for volume terms). To account for age, two separate maturation function terms were evaluated: one applied to PK parameters related to CL (CL_TT, CL_TEPA, and V_max) and one applied to the volume of distribution of TT. The maturation function for CL resulted in a dOFV of −14.71 (P < .001), and the maturation function for volume of distribution of the TT further improved fit by dOFV −17.80 (P < .001). For allometric scaling based on FFM, exponents 0.75 for CL_TT and CL_TEPA, and 1.0 for V_{c, TT} and V_{p, TT}. The volume of TEPA (Vm) was fixed to 1 L as it was not independently estimable from the available data; cAUC_TEPA was derived from the cumulative metabolite compartment amount directly. The VPC of both parent and metabolite as seen in Figure 1 shows the distribution of the observed data is well captured across the majority of the PK time curve. There was some misspecification with the TT peak concentrations as demonstrated by the median raw data line and median prediction interval in the first bin, likely driven by the somewhat limited sampling strategy employed for this study. Goodness‐of‐fit plots and a sampling of individual plots are available in Figures S1–S4. The clearance equations for TT and TEPA are as followed:

$$\begin{array}{ccc}\hfill \mathbf{C}{\mathbf{L}}_{\mathbf{TT},\mathbf{i}}& =& \mathrm{C}{\mathrm{L}}_{\mathrm{TT},\mathrm{pop}}\times {\frac{\mathrm{FFM}}{15\mathrm{kg}}}^{0.75}\hfill \\ & & \times \frac{\mathrm{Ag}{\mathrm{e}}^{\mathrm{Hil}{\mathrm{l}}_{\mathrm{CL}}}}{\mathrm{T}{{\mathrm{M}}_{50,\mathrm{CL}}}^{\mathrm{Hil}{\mathrm{l}}_{\mathrm{CL}}}+\mathrm{Ag}{\mathrm{e}}^{\mathrm{Hil}{\mathrm{l}}_{\mathrm{CL}}}}\times {\eta }_{\mathrm{CL},\mathrm{TT}}\hfill \end{array}$$

$$\begin{array}{ccc}\hfill \mathbf{C}{\mathbf{L}}_{\mathbf{TEPA},\mathbf{i}}& =& \mathrm{C}{\mathrm{L}}_{\mathrm{TEPA},\mathrm{pop}}\times {\frac{\mathrm{FFM}}{15\mathrm{kg}}}^{0.75}\hfill \\ & & \times \frac{\mathrm{Ag}{\mathrm{e}}^{\mathrm{Hil}{\mathrm{l}}_{\mathrm{CL}}}}{\mathrm{T}{{\mathrm{M}}_{50,\mathrm{CL}}}^{\mathrm{Hil}{\mathrm{l}}_{\mathrm{CL}}}+\mathrm{Ag}{\mathrm{e}}^{\mathrm{Hil}{\mathrm{l}}_{\mathrm{CL}}}}\hfill \end{array}$$

Table 2.

Parameter Estimates for the Final Covariate Model and Bootstrap Analysis

Population PK Parameters	Final Model Value (RSE %)	Bootstrap Mean Value (95% CI)
Thiotepa (parent compound)
Typical value for thiotepa CL (L/h/15 kg)	6.665 (9.9)	6.60 (4.94–8.22)
Volume of the central compartment (L/15 kg)	15.837 (7.0)	17.57 (9.84–24.13)
Intercompartmental CL (L/h/15 kg)	1.876 (29.6)	1.54 (0.55–2.85)
Volume of the peripheral compartment (L/kg/15 kg)	7.139 (18.2)	6.72 (3.83–9.55)
TEPA (metabolite)
Typical value of TEPA CL (L/h/15 kg)	2.035(6.3)	2.52 (1.50–3.28)
Volume of the central compartment (L)	1–Fixed	1–Fixed
Maximum velocity for nonlinear metabolism from TT to TEPA (Vmax, mg/h/15 kg)	3.316 (12.9)	4.190 (2.648–5.339)
Michaelis–Menten constant (Km, mg/L)	0.5056 (20.4)	0.4658 (0.2657–0.6742)
Maturation Hill constant on CL (Fmat, CL, unitless)	1.056 (23.3)	1.2 (0.24–2.04)
Maturation time at 50% for CL (TM50, CL, years)	0.370 (29.9)	0.34 (0.07–0.62)
Maturation Hill constant for V (Fmat,Vc, unitless)	2.712 (30.5)	3.6 (0.3–25.5)
Maturation time at 50% for V (TM50, Vc, years)	0.3067 (17.7)	0.58 (0.13–2.18)
Between subject variability on CL (VIICL, %CV)	34.15 (11.8)	32.57 (22.82–42.58)
Between subject variability on Vmax (IIVVmax, %CV)	29.3 (12.3)	41.1 (30.37–52.99)
Between subject variability on Km (IIVKm, %CV)	96.19 (17.2)	89.84 (51.79–125.78)
Residual proportional error for thiotepa (%)	29.4 (14.9)	28.0 (21.5–39.2)
Residual proportional error for TEPA (%)	22.5 (20.7)	9.0 (0.5–13.9)

CI, confidence interval; CL, clearance; CL_m, clearance of metabolite TEPA; CV, coefficient of variation; IIV, inter‐individual variability; K_m, Michaelis constant; Q, intercompartmental clearance; RSE, relative standard error; V_c, central compartment volume; V_m, central compartment volume of metabolite TEPA; V_max, maximum velocity for nonlinear metabolism from TT to TPEA; V_p, peripheral compartment volume; TM₅₀, maturation half‐time.

Figure 1.

Visual predictive check of both TT and TEPA where the dots represent the raw observed concentrations, the solid line is the median of the observed data, the dotted lines are the 5th and 95th percentile of the observed data, and the shaded areas represent the prediction interval of the 5th, median, and 95th percentile of the model predictions. The black tick marks at the top of each plot indicate the time bin breakpoint in which the distributions are calculated.

Simulations

Exposure by Regimen

The median (IQR) observed cAUC_total for study patients receiving conventional dosing was 51.0 mg h/L (IQR: 42.0–60.1) for the 5 mg/kg every 12 h × 2 doses and 140.6 mg h/L (112.5–164.6) for the 300 mg/m² every 24 h × 3 doses, with substantial interpatient variability in both groups (Table 3). The ratio of TT‐to‐TEPA cAUC differed by regimen decreasing with dose from 1.0 to 0.5 in the 5 mg/kg every 12 h × 2 doses to 300 mg/m² every 24 h × 3 doses, respectively.

Table 3.

Comparison of Different Thiotepa Exposure (AUC_0–inf) for a Single Dose of 5 mg/kg Compared to a Single Dose of 300 mg/m². ^a Data are presented as median (interquartile range).

AUC0–inf (mg h/L)	Single Dose of 5 mg/kg	Single Dose of 300 mg/m2
No. of subjects	36	28
AUC(TT)	23.9 (21.1–32.6)	90.2 (81.3–105.8)
AUC(TEPA)	27.3 (16.7–34.0)	52.6 (36.0–69.1)
AUC(total TT +TEPA)	51.0 (42.0–60.1)	140.6 (112.5–164.6)
AUC(TEPA)/AUC(TT)	1.0 (0.7–1.4)	0.5 (0.3–0.7)

^a Does not include patients receiving 4 mg/kg, 7.5 mg/kg, or 10 mg/kg dosing of thiotepa.

Conventional Versus Exposure Targeted Dosing

In conventional weight‐ or BSA‐based dosing groups, cAUC_total demonstrated substantial interpatient variability across all three regimens, with a coefficient of variation (CV) of 30%. Median cAUC_total with conventional dosing was 54.8 mg h/L (IQR: 44.2–66.6) for 5 mg/kg every 12 h × 2 doses, 149 mg h/L (IQR: 120–183) for 10 mg/kg every 24 h × 3 doses, and 145 mg h/L (IQR: 119–175) for 300 mg/m² every 24 h × 3 doses. Target attainment within ±20% of the pre‐specified cumulative cAUC_total goal was 47% and 51% for conventional 5 mg/kg and 300 mg/m² based regimens, respectively.

The model‐predicted dose group reduced cAUC_total variability and improved target attainment across all regimens. The CV of cAUC_total was reduced to approximately 13%, and 12% for the 5 mg/kg and 300 mg/m² based regimens, respectively. Median cAUC_total for the model‐predicted doses was 53.1 mg h/L (IQR: 48.6–57.3) and 142 mg h/L (IQR: 130–153). Target attainment within ±20% of the target improved to 83% and 90% for the 5 mg/kg and 300 mg/m² based regimens, respectively (Table 4; Figure 2).

Table 4.

Model‐Based Predicted Doses of cAUC_{(TT + TEPA}) as Compared to Conventional Dosing Aimed to Achieve the Median Exposure of the Dosing Regimen Over a Complete Course of Therapy for 5 mg/kg Every 12 h for Two Doses or 300 mg/m² Every 24 h for Three Doses.

Conventional Dosing	Median Exposure Target cAUC(TT + TEPA)	Weight (kg)	Mean Model‐Predicted Dose ± SD (mg/kg)
5 mg/kg every 12 h for 2 doses	51 mg h/L	0–9.9	5.17 ± 1.57
		10–19.9	5.66 ± 1.52
		20–39.9	5.27 ± 1.55
		40–59.9	4.59 ± 1.26
		60–79.9	4.21 ± 1.28
		>80	3.37 ± 0.65
300 mg/m2 every 24 h for 3 doses	141 mg h/L	0–9.9	10.96 ± 3.21
		10–19.9	11.79 ± 3.33
		20–39.9	10.72 ± 3.23
		40–59.9	9.16 ± 2.57
		60–79.9	8.27 ± 2.58
		>80	6.48 ± 1.23

Figure 2.

Cumulative cAUC_total with conventional (mg/kg or mg/m²) versus model‐based dosing (median cAUC_total of the dosing regimen). Simulated cAUC_total in 500 virtual pediatric patients per regimen under conventional flat‐dose prescribing (dark) and TDM‐guided model‐based dosing with MAP Bayesian dose adjustment after dose 1 (light). Boxes represent the interquartile range; whiskers extend to the 5th and 95th percentiles. Dashed lines indicate pre‐specified cumulative cAUC_total targets (51 mg h/L for 5 mg/kg × 2; 141 mg h/L for 300 mg/m² × 3).

Population‐level dose curves illustrating the model‐predicted dose of TT (mg/kg or mg/m²) required to achieve each cAUC_total target across the pediatric weight spectrum are presented in Figure 3. For the 5 mg/kg every 12 h regimen, model‐predicted dose deviated substantially from the conventional flat 5 mg/kg dose across the weight range, particularly in the youngest and lowest‐weight patients, reflecting the combined influence of age‐dependent maturation of clearance and allometric scaling based on FFM. Model‐predicted doses for the 300 mg/m² every 24 h regimen similarly diverged from the conventional BSA‐normalized dose across the pediatric weight spectrum.

Figure 3.

Model‐predicted dose aimed to achieve a median cAUC_{(TT + TEPA)} exposure for a single dose of thiotepa versus conventional per‐dose thiotepa dose across the pediatric weight spectrum (5 mg/kg [Panel A], 10 mg/kg [Panel B], and 300 mg/m² [Panel C]) The dark line represents the conventional flat dose; the blue line represents the model‐based dose; the shaded ribbon depicts the 10th–90th percentile range of inter‐individual variability in dose due to covariate adjustment.

The model‐predicted dose aimed to achieve a median cAUC_total for conventional and model‐based dosing for infants and patients with obesity is presented in Figure 4. Because the final population PK model scales clearance allometrically to FFM (exponent 0.75) and applies age‐dependent maturation functions, conventional weight‐ or BSA‐based dosing markedly overshoots the per‐dose target in infants younger than 2 years, whereas clearance is reduced to approximately 15%–20% of mature values in the first months of life and increasingly overshoots at higher BMI, where total body weight diverges progressively from FFM. Model‐based dosing individualized to patient‐specific FFM and age maintained cAUC_total near the pre‐specified per‐dose target across the full spectrum of ages and body compositions evaluated.

Figure 4.

Simulated per‐dose cAUC_total comparing conventional dosing (mg/kg or mg/m²) versus model‐based dosing (based on the median cAUCtotal for the dosing regimen) in infants and patients with obesity. Per‐dose cAUC_total is shown across age 0.1–2 years using WHO 50th‐percentile growth references (Panels A, B) and across BMI 18–42 kg/m² for a representative 16‐year‐old (average of male [177 cm] and female [163 cm]; Panels C, D) for the 5 mg/kg × 2 (left) and 300 mg/m² × 3 (right) regimens. The dark line and shaded ribbon represent the conventional population‐predicted exposure and 10th–90th percentile IIV range, respectively; the teal line represents model‐based individualized dosing. The dashed line denotes the per‐dose cAUC_total target (25.5 mg h/L for 5 mg/kg × 2; 47.0mg·h/L for 300 mg/m² ×3). BMI category shading: white = normal; light orange = overweight; orange = obese class I; red = obese class II/III.

Discussion

The study population included in this analysis was purposely heterogeneous in order to provide a broad range of TT dose levels and regimens for evaluation. Our analysis aligns well with several prior studies describing the nonlinear kinetics of TT at higher doses in contrast those smaller studies evaluating a single dose level within a population. ²¹ , ²² The final structural model characterized TEPA disposition using a one‐compartment model, in contrast to the two‐compartment structure employed for TT. Although this configuration provided an adequate description of the available concentration–time data, identification of a two‐compartment model for TEPA may have been precluded by limited sampling density in the post‐distributive elimination phase. It is possible the current structural model overestimates cAUC_TEPA in some patients, which would have implications for AUC‐based dose recommendations given TEPA contributes substantially to total alkylating potential. To assess this, additional PK samples collected in the elimination phase of drug clearance with futures studies will help to confirm the TEPA compartmental structure and improve the accuracy of cAUC_TEPA estimations.

Beyond structural considerations, the covariate relationships retained in the final model also warrant discussion in the context of prior work. In an earlier population PK analysis of TT and TEPA, Huitema et al identified ALP and serum albumin as significant covariates influencing TT and TEPA clearance, respectively. In this study, inclusion of these covariates did not result in a statistically significant improvement in model fit as assessed by objective function value or goodness‐of‐fit diagnostics and were therefore excluded from the final model. While there may be little or no effect driven by these covariates, this discrepancy is more likely attributable to the limited variability within the current study population with most patients within or near the normal reference range for age at the time of PK sampling. These findings do not exclude the possibility that hepatic function or nutritional status may influence TT or TEPA disposition in patients with more markedly abnormal biochemical profiles, and this relationship warrants further evaluation in larger and more clinically heterogeneous populations.

The covariates retained in the final model, specifically FFM‐based allometric scaling and age‐dependent maturation, have direct and significant implications for considering dosing in vulnerable subpopulations. Particularly for children less than 2 years of age, the age‐maturation functions reduces clearance to approximately 15%–20% of mature values in the first few weeks of life. Conventional flat weight‐ or BSA‐based dosing, cannot account for this developmental ontogeny, and therefore may substantially overshoot the goal cAUC_total and increase the risk of regimen‐related toxicity in the youngest patients. In individuals classified as overweight or obese, the divergence between total body weight and FFM produces proportional increases in cAUC_total that go unaccounted for with conventional dosing strategies. The prospective integration of patient‐specific FFM and maturation‐informed clearance estimates into dose selection represents a clinically actionable strategy to reduce both over‐ and under‐exposure in these high‐risk subgroups and supports the broader application of model‐based dosing frameworks for TT individualization in pediatric HCT.

Several limitations of this analysis should be noted for this analysis. First, the final population PK model underlying these simulations was developed without the inclusion of pharmacogenomic data, which is a notable gap given substantial genetic variability which has been shown to contribute to the metabolism and distribution of thiotepa. Thiotepa is metabolized by CYP2B6 and CYP3A4/5 into the active metabolite TEPA and with further detoxified by glutathione S‐transferase, all of which have variants associated with intermediate to poor function that might alter the ratio of these agents. ²³ Prospective integration of these genotypic information for the variants responsible for the activation and detoxification of both TT and TEPA may reduce unexplained variability and strengthen the foundation for model‐based dosing implementation.

Second, the current conventional dosing strategy of administering thiotepa over 1–3 days was not initially derived from PK‐guided data but from early phase I studies correlating CNS toxicity with higher single dose administration. ²⁴ This clinically observed effect has not been prospectively evaluated and may be partially explained by increased dose and metabolic saturation. As concentrations approach K_m, the conversion of parent to metabolite becomes saturable, producing disproportionately larger increases in the TT/TEPA ratio. This divergence was visible across the evaluated dosing simulations and consistent with our model structure incorporating Michaelis–Menten kinetics. This mechanism was described by Ekhart et al, demonstrating that consolidating a 4‐day fractionated regimen into a 2‐day unfractionated schedule results in a 13% increase in cAUC_TT and a 21% reduction in cAUC_TEPA. ¹⁶ Given IIV on both V_max and CL_TT further increased the TT/TEPA ratio at higher doses, the current empiric schedule may be introducing PK heterogeneity that covariate‐informed dosing alone cannot resolve. A more frequent, lower per dose strategy will maintain concentrations closer to the linear range, reduce variability in the TT/TEPA ratio distribution, and create natural inter‐dose sampling windows that are currently absent from standard regimens yet essential for the potential of future adaptive dosing strategies. Prospective, PK‐driven evaluation of the total number and size of each dose is warranted.

Third, the simulations presented in this analysis reflect achievement of the median cAUC_total observed in the data for each dosing regimen. Defining the optimal cAUC_total target for thiotepa to optimize clinical outcomes was not within the scope of the current analysis with goal targets presumably variable among different subgroups of patients. Exposure–response analysis is therefore a next critical next step of a subsequent planned analysis for our group. Additionally, the extent of between‐patient variability captured by the current model may limit how precisely a single covariate‐informed dose can achieve potentially narrow exposure targets, suggesting that therapeutic drug monitoring may be necessary to fully optimize dosing. However, the potential advantage of TDM is limited by practical feasibility, including assay availability among clinical sites for drug quantification and the temporal proximity of doses in conventional TT dosing schedules. Standard 12‐ or 24‐h dosing intervals of thiotepa schedules leave little opportunity for real‐time dose adjustment.

In summary, this work highlights key PK properties of thiotepa that can be employed to prospectively evaluate exposure–response relationships and optimize dosing across a variety of patient populations. The conversion of TT to TEPA was found to be saturable at higher doses and may have important implications for exposure optimization. Model‐based dosing, which accounts for age and body size, results in more predictive exposure compared with conventional mg/kg dosing, particularly in the very young and obese. Next steps include evaluation of genetic variants for their impact on thiotepa PK and subgroup analysis of exposure–response relationships to optimize dosing in both the allogeneic and autologous HCT setting.

Author Contributions

Jordan Brooks contributed to the conceptualization, methodology, software, formal analysis, data curation, visualization, and writing ‐ original draft. Maria‐Stephanie Hughes contributed to the conceptualization, resources, and writing—reviewing and editing. Ron J. Keizer contributed to the conceptualization, methodology, software, validation, supervision, project administration, and writing—reviewing and editing. Shams Ismaeil, Maria‐Stephanie Hughes, Beth Apsel‐Winger, Ying Lu, Christine Higham, Ashish Gupta, Nahal Rose Lalefar, Eric Anderson, Michael A. Pulsipher, and Sung‐Yun Pai, provided patient data to the analysis, curated data, and writing—reviewing, editing of manuscript. Christopher C. Dvorak contributed to study design and implementation, curated data, and writing—reviewing, editing of manuscript. Kristina Nelson, Susie Long‐ Provided patient data to the analysis, curated data, and writing‐reviewing, editing of manuscript. Janel Long‐Boyle‐ Designed and implememted the study, curated data, performed analysis, and writing, reviewing, and editing of the manuscript.

Conflicts of Interest

Jordan Brooks, Maria‐Stephanie Hughes, and Ron J. Keizer are employees and stockholders of InsightRX. Christopher C Dvorak reports consulting for Alexion, Inc. and Medexus Corporation. Janel Long‐Boyle reports consulting for Medexus, Jasper Therapeutics, and Rocket Pharmaceuticals. Michael A. Pulsipher reports Advisory for Medexus, Pharming, Autolous, Stratus, Mesoblast, Cargo, Garuda, and Pfizer; educational activity for Sanofi, and study support from Adaptive and Miltenyi.

Funding

This project was supported by the National Center for Advancing Translational Sciences and National Institutes of Health, through the UCSF‐CTSI Grant Number ULTR001872. Its contents are solely the responsibility of authors and do not necessarily represent the official views of the NIH. Michael A. Pulsipher reports funding from National Institute of Allergy and Infectious Diseases (U01A1 184132 and U01A1 126612) for C‐SIDE and additional support from P30CA040214 National Cancer Institute (NCI).

Supporting information

Supporting Information

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.