NextDose: A web-based Bayesian dose forecasting tool

Last updated 15 January 2026

Busulfan

Target Concentration

TCI for busulfan is usually used as part of conditioning for bone marrow transplant. The conditioning aims to remove existing host stem cells so that transplanted stem cells can be grafted and form new cells. Initial exposure-response studies of busulfan led to an average steady state target of 0.77 mg/L (Bolinger, Zangwill et al. 2000). This is equivalent to an AUC over a 6 h dosing interval of 1125 umol/L*min which is the target recommended in the FDA label (PDL 2006) for busulfan used for bone marrow conditioning.

Busulfan is an alkylating agent which is thought to bind irreversibly to DNA thereby causing cell damage and death. This irreversible action would explain why the cumulative exposure during treatment would be a predictor of cell kill and thus bone marrow ablation. Thus the exposure target for busulfan is the cumulative AUC over the standard treatment period of 4 days. The AUC with a constant rate infusion over 4 days approaching a steady state concentration of 0.77 mg/L would be 74 mg/L*h.

A subsequent analysis proposed an acceptable range for a 4 day cumulative AUC of 78 to 101 mg/L*h (Bartelink, Lalmohamed et al. 2016). The middle of that acceptable range is 89.5 mg/L*h (usually rounded up to 90 mg/L*h) which is the target exposure equivalent to 0.94 mg/L average steady state concentration.

A 2 day cumulative AUC of 16.8 mg/L (equivalent to 0.35 mg/L average steady state) has been proposed for bone marrow conditioning for Fanconi anemia (Mehta, Emoto et al. 2019).

A dosing strategy for busulfan is to aim for a steady state average target of 0.94 mg/L with the target type option “Css avg mg/L” then measure concentrations after the first busulfan dose (Figure 1).

The predictive performance of NextDose for doses required to reach a target cumulative AUC has been evaluated and shown to have acceptable bias and precision (Lawson, Paterson et al. 2021).

“Both software tools utilising Bayesian methods provided acceptable relative bias and precision of cumulative exposure estimations under the tested sampling scenarios. Relative bias ranged from median RE of 0.1–14.6% using InsightRX and from 3.4–7.8% using NextDose. Precision ranged from median RMSE of 0.19–0.32 mg·h·L−1 for InsightRX and 0.08–0.1 mg·h·L−1 for NextDose.”

The authors note that the bias of the Bayesian dosing methods is based on comparison with a non-compartmental (NCA) method for calculating AUC. The NCA methods are known to be biased (~10%) which means the Bayesian dose predictions are probably more reliable (Bartelink, Lalmohamed et al. 2016).

Figure 1 Dose prediction options with Css avg selected

The time course of observed busulfan concentrations after a single dose of 90 mg are shown with NextDose predictions in Figure 2.

Figure 2 NextDose busulfan observed and predicted concentrations

The proposed dose to achieve an average steady state target of 0.94 mg/L is 233 mg/day (Figure 3).

Figure 3 NextDose Results with average steady state target concentration

The NextDose proposed dose with a 24 h dosing interval can be multiplied by 4 to predict the total dose required. Subsequent measurements of busulfan can be used to refine the daily dose requirement and subsequent doses calculated from the predicted 4 day total dose minus those already given.

An alternative method is to use a cumulative AUC target type e.g. “mg/L*h (4 day cum AUC)” with a target value such as 90 mg/L*h (Figure 4).

Figure 4 Dose prediction options with 4 day cum AUC selected

You may choose any dosing interval that is appropriate e.g. 24 h and NextDose will propose doses for that dosing interval which will achieve the cumulative AUC target of 90 mg/L*h. Because of the known decrease in busulfan clearance with time after starting treatment only the next dose is proposed. Further doses require repeated use of NextDose with the previously proposed dose. See example below Figure 5.

If the time remaining before the end of the cumulative AUC interval is less than 1.25 x the dosing interval a single final dose will be proposed. Otherwise the remaining dose to achieve the cumulative target will be divided among the remaining dosing intervals in the remaining time.

An example of using the cumulative AUC target is shown in Figure 5. It shows the AUC as a % of the cumulative target AUC (8.64%). The AUC% value is estimated using between occasion variability. This is followed by the proposed dose of 303 mg given once a day and the predicted remaining % of target exposure after this first dose (91.4%).

The cumulative AUC table (Table 1) provides more information to help understand how each dose is calculated for a cumulative AUC target. This table replaces the predicted dose table used with other target types.

Table 1 Meaning of variables listed in the cumulative AUC table

Column	Meaning
Tdose	Time of dose (h) relative to first dose
T+DI	Tdose plus dosing interval (h)
AMTcum	Cumulative dose (total amount given)
AUCcum	Cumulative AUC from 0 to infinity from all doses
AUCdose	AUC from 0 to infinity from the dose at Tdose
AUC%	% contribution of AUCdose to the cumulative target AUC
Drem	Total remaining dose predicted to achieve the cumulative target AUC
Dnext	Next dose predicted to achieve the cumulative target AUC
Trem	Time remaining until the end of the cumulative target AUC period
AUCrem	AUC remaining from 0 to infinity to achieve the cumulative target AUC
CL(T)	Clearance using estimated BSV and BOV at Tdose (if not AVG model)
CLavg	Clearance using estimated BSV at Tdose

Figure 5 NextDose Results with cumulative AUC target

Note that the proposed dose of 304 mg is based on the assumption that clearance will remain unchanged. The McCune model for busulfan includes a time varying change in clearance, therefore, it is recommended to repeat the NextDose calculation for each of the proposed doses. This is illustrated in the following sequence of figures.

Figure 6 Predicted time course and proposed doses of 273 mg following the second dose of 304 mg.

Figure 7 Predicted time course and proposed final dose of 266 mg following the third dose of 273 mg.

Figure 8 Predicted time course following the fourth dose of 266 mg which is within 1% of the target.

Note how the proposed dose decreased with time reflecting the decrease in busulfan clearance that has been reported over this time period (McCune, Bemer et al. 2014).

NextDose Model

The NextDose pharmacokinetic model is based on (McCune, Bemer et al. 2014). This is a large study including infants to elderly adults. It accounts for body size and composition using fat free mass, maturation of clearance in infants and a small decrease in clearance over 4 days of treatment.

Dead Space Lag Time

Arrival of drug into the central compartment after the start of a constant rate intravenous infusion is typically delayed due to the dead space volume between the drug infusion source and delivery of fluid into the bloodstream. In adults this typically ignorable but in neonates it may take an hour before drug reaches the blood. The example in Figure 9 is taken from an actual patient after several previous doses of busulfan. This reveals 3 concentration measurements taken at the start of the infusion showing a slow decline from previous doses. The next concentrations are after busulfan has reached the blood. The model without a deadspace lagtime is a poor fit to the observed concentrations.

Figure 9 Unmodelled lag time due to dead space in a neonate

The PK lag time 2017_AVG model may be used to account for the dead space volume.

Figure 10 Selection of the PK lag time 2017_AVG model

By adding Dead Space Volume (10 mL) and the busulfan Fluid Conc (0.5 mg/L) observations using the Doses and Observations page, this model can estimate the lag time due to the dead space with much improved fit to the observed concentrations (Figure 11).

Figure 11 Modelled lag time due to dead space in a neonate

Thanks to Dr Laila Nassar, Rambam Hospital, Israel, for providing this example and helping to develop the PK lag model.

Oral Absorption

The model parameters used to describe oral dosing with busulfan have been changed based on the work reported by Hassan (Hassan, Ljungman et al. 1994). The previous busulfan model assumed first-order absorption without lag time, rate constant (KA=1/h), complete oral bioavailability (Foral=1) and no between subject variability. The Hassan study reported individual estimates of KA, Tlag and Foral from 17 patients. They reported an age associated difference in the extent of bioavailability when categorized as adults and children. Using age as a continuous variable, extracted from in their published data, provided little support for this distinction. The linear correlation between age and bioavailability was 29%. Removal of one outlier value in a child decreased the trend with a remaining correlation of 17%. The parameter values used in NextDose are based on the average of adult and children combined with population estimates of Tlag=0.41, KA=6.43 /h and Foral=0.78 (Foral outlier removed) and between subject variability (square root of omega) of Tlag=93%, KA=73%, Foral=29% (Foral outlier removed).

References

Bartelink, I. H., A. Lalmohamed, E. M. van Reij, C. C. Dvorak, R. M. Savic, J. Zwaveling, R. G. Bredius, A. C. Egberts, M. Bierings, M. Kletzel, P. J. Shaw, C. E. Nath, G. Hempel, M. Ansari, M. Krajinovic, Y. Theoret, M. Duval, R. J. Keizer, H. Bittencourt, M. Hassan, T. Gungor, R. F. Wynn, P. Veys, G. D. Cuvelier, S. Marktel, R. Chiesa, M. J. Cowan, M. A. Slatter, M. K. Stricherz, C. Jennissen, J. R. Long-Boyle and J. J. Boelens (2016). "Association of busulfan exposure with survival and toxicity after haemopoietic cell transplantation in children and young adults: a multicentre, retrospective cohort analysis." Lancet Haematol 3(11): e526–e536.

Bolinger, A. M., A. B. Zangwill, J. T. Slattery, D. Glidden, K. DeSantes, L. Heyn, L. J. Risler, B. Bostrom and M. J. Cowan (2000). "An evaluation of engraftment, toxicity and busulfan concentration in children receiving bone marrow transplantation for leukemia or genetic disease." Bone Marrow Transplant 25(9): 925–930.

Hassan, M., P. Ljungman, P. Bolme, O. Ringden, Z. Syruckova, A. Bekassy, J. Stary, I. Wallin and N. Kallberg (1994). "Busulfan bioavailability." Blood 84(7): 2144–2150.

Lawson, R., L. Paterson, C. J. Fraser and S. Hennig (2021). "Evaluation of two software using Bayesian methods for monitoring exposure and dosing once-daily intravenous busulfan in paediatric patients receiving haematopoietic stem cell transplantation." Cancer Chemotherapy and Pharmacology.

McCune, J. S., M. J. Bemer, J. S. Barrett, K. Scott Baker, A. S. Gamis and N. H. G. Holford (2014). "Busulfan in Infant to Adult Hematopoietic Cell Transplant Recipients: A Population Pharmacokinetic Model for Initial and Bayesian Dose Personalization." Clinical Cancer Research 20(3): 754–763.

Mehta, P. A., C. Emoto, T. Fukuda, B. Seyboth, A. Teusink-Cross, S. M. Davies, J. Wilhelm, K. Fuller, A. A. Vinks and F. Boulad (2019). "Busulfan Pharmacokinetics and Precision Dosing: Are Patients with Fanconi Anemia Different?" Biol Blood Marrow Transplant 25(12): 2416–2421.

PDL (2006). "Busulfex Product Label http://www.ivbusulfex.com/Busulfex_Marketing.Pi.8.06.pdf."

Copyright All rights reserved | Developed by Sam Holford & Nick Holford 2012-2026