Busulfan is the backbone of myeloablative conditioning for hematopoietic stem cell transplantation (HSCT) in children. Getting the exposure right matters enormously: insufficient AUC leads to engraftment failure; excessive AUC causes sinusoidal obstruction syndrome (SOS), seizures, and pulmonary toxicity. The therapeutic window - typically targeting a cumulative AUC of 900-1350 micromol-min over four days of IV dosing - is narrow, and inter-patient variability in busulfan clearance is among the highest of any commonly used cytotoxic.
Three well-regarded academic medical centers have published pediatric busulfan dosing nomograms in the past decade. The disagreement between them, at the same body weight, is up to 19%. That is not a rounding difference. It is a clinically meaningful dose gap that places children at different ends of the therapeutic window depending on which institution's nomogram the transplant center adopted. This article examines why the nomograms diverge and what the implications are for clinical TDM practice.
Why Busulfan Clearance Is Particularly Variable in Children
Busulfan is eliminated primarily by conjugation with glutathione, catalyzed by glutathione-S-transferase (GST) enzymes. GST activity is strongly age-dependent in the pediatric population: neonates and infants have substantially lower GST activity than older children and adults. This age-dependent maturation means that busulfan clearance is not simply proportional to body weight or BSA - it follows a maturation function that accelerates through approximately age 3-4, with most patients approaching adult-typical clearance by school age.
The three published nomograms differ in how they parameterize this maturation function. One uses chronological age, one uses postnatal age adjusted for prematurity, and one uses a sigmoid maturation curve referenced to a postmenstrual age parameter. At a body weight of 10 kg, these different parameterizations produce recommended initial doses that differ by 15-19% depending on the child's age distribution - and neonates and infants are the patients where correct dosing matters most, because their transplant-related mortality is highest and their margin for exposure error is smallest.
The Model Behind the Nomogram: What Each Published Chart Assumes
A busulfan nomogram is ultimately a lookup table derived from a population PK model. The nomogram converts body weight and age into an initial dose recommendation. Its reliability depends entirely on the underlying population model's validity for your patient population.
Examining the model behind each published nomogram reveals systematic differences. The Seattle Children's nomogram (Bleyzac et al. version) was derived from a predominantly Caucasian population with a specific age distribution at their center. The St. Jude Children's Research Hospital dataset had a higher proportion of African American patients, which matters because GST alpha-1 polymorphisms differ in frequency between ancestral populations and these polymorphisms affect busulfan clearance. The European dataset (BFM group) included more patients treated with phenytoin as seizure prophylaxis, which induces CYP enzymes and may have incidentally affected the clearance parameter estimates even for busulfan (whose CYP contribution is minor but not zero).
The practical implication: if your transplant center's population differs systematically from the nomogram's derivation dataset - in terms of age distribution, ancestry, or concomitant anticonvulsant use - the initial dose recommendation from that nomogram will be systematically biased. TDM after the first dose is not optional; it is how you correct for that bias prospectively.
What TDM After Dose 1 Can and Cannot Fix
In a four-day IV busulfan conditioning schedule, the standard approach is to measure AUC after dose 1 (typically using 4-6 PK samples collected over 6 hours) and adjust doses 2-16 (busulfan is usually given every 6 hours for four days, totaling 16 doses) based on the observed versus target AUC. This adaptive dosing approach has been shown in multiple prospective studies to improve target AUC attainment compared to fixed weight-band dosing.
What TDM cannot fix is a missed first-dose seizure or acute SOS from a severely overdosed first administration. In neonates, the initial nomogram dose matters more than in older children because there is no opportunity to "correct" after the first dose if neurotoxicity has already occurred. For very young patients (under 6 months), the case for a conservative initial dose - deliberately targeting the lower end of the therapeutic window and adjusting up if the measured dose-1 AUC supports it - is stronger than at any other age group.
Sparse Sampling for Busulfan TDM: A Validated Protocol
Full busulfan PK sampling requires 6-8 blood draws over 6 hours per dose. In young children and infants, this represents a significant blood volume and patient burden issue. Sparse sampling approaches using 3-4 samples have been validated for busulfan with acceptable AUC estimation precision (CV% around 10-15%) using MAP Bayesian methods, provided the sampling schedule is optimized for the two-compartment busulfan model.
The most widely validated sparse sampling schedule for pediatric busulfan places samples at approximately 30 minutes, 1 hour, 2 hours, and 4 hours post-infusion start. At institutions with laboratory constraints, a 3-sample schedule (1h, 2h, 4h) performs acceptably in children over age 2 but is less reliable in the youngest patients where volume of distribution variability is highest. The considerations for sparse sampling design discussed in our article on MAP Bayesian estimation from sparse samples apply directly to the busulfan context.
The Allometric Scaling Controversy
Allometric scaling in pediatric PK models uses the relationship clearance = CL_typical x (weight/70)^0.75 (based on the three-quarter exponent from metabolic scaling theory) to relate typical clearance to body weight. This scaling works reasonably well in children over age 2, where clearance is primarily weight-driven. In neonates and infants, the maturation factor dominates over the allometric factor, and using allometric scaling alone - as some older nomograms do - overestimates clearance in the youngest patients.
Two approaches are used in modern pediatric PK models to handle this: the Anderson-Holford maturation function (MF = PMA^Hill / (TM50^Hill + PMA^Hill), where PMA is postmenstrual age, TM50 is the PMA at 50% mature activity, and Hill is the shape parameter), and empirical age-band correction factors based on observed clearance distributions by age group. The Anderson-Holford approach is more mechanistically grounded but requires knowing the patient's PMA accurately, which is straightforward for premature neonates but ambiguous for full-term infants born slightly pre- or post-term.
Institutional Validation Before Clinical Use
Any transplant center that is using a published busulfan nomogram should have conducted an internal validation study before trusting the nomogram as their clinical dosing standard. The validation question is: in our patient population, at what fraction of patients does the nomogram-derived initial dose produce a day-1 AUC within the target range? A fraction below 60% suggests the nomogram's underlying model is not well-matched to your patients, and the initial doses should be adjusted systematically.
Internal validation does not require a formal research study - it can be done retrospectively by pulling 30-50 consecutive transplant patients for whom dose-1 PK data were collected and comparing the observed AUC against the nomogram's prediction. Most major transplant centers have this data in their pharmacy or transplant databases but have not analyzed it explicitly. Running this analysis is a low-effort way to identify whether your nomogram is biased for your patient population before you put it in front of a clinician making a dosing decision.
The Case for Bayesian Dose Individualization from Dose 1
The argument for replacing nomogram-based initial dosing with a Bayesian approach even for the first dose is strongest in busulfan because the between-subject variability in clearance is large enough (50-70% CV% in the literature) that a population-based starting point carries substantial uncertainty. With a one-compartment Bayesian prior and a single trough concentration measured 6 hours post-infusion start, it is possible to generate a rough individual clearance estimate that outperforms the nomogram's predicted clearance in approximately 40% of patients.
The challenge is that a single trough cannot differentiate well between variation in clearance and variation in volume of distribution. Two samples - one at 1 hour post-start (capturing distribution phase) and one at 5-6 hours (capturing elimination) - provide sufficient information to estimate both parameters with meaningful individual precision for dose-1 Bayesian adjustment, even in the youngest patients. This is the approach DoseMind implements for busulfan conditioning protocols: two required samples from dose 1, Bayesian dose recommendation available before dose 2, with credible intervals that explicitly account for the maturation function uncertainty in the youngest age groups.
Conclusion: Nomograms Are a Floor, Not a Ceiling
Published busulfan nomograms are a starting point for initial dose selection, not a substitute for individual PK monitoring. The inter-nomogram disagreement documented here is not a reason to abandon them - it is a reason to use them with TDM rather than instead of TDM. The goal is to enter the therapeutic window on dose 1 and stay there for all 16 doses. Nomograms help with entry; adaptive Bayesian dosing keeps you there.
For transplant programs seeking to implement or improve their busulfan TDM workflow, DoseMind supports pediatric-specific population PK models with validated maturation function parameters and provides sparse sampling schedule recommendations optimized for age and weight. Reach the team at hello@dosemind.com.