TBI conditioning prior to HSCT was pioneered by Thomas et al. in Seattle in 1970 (4). They added high-dose cyclophosphamide (120 mg/kg given over 2 days) to TBI in an effort to increase cytoreduction pre transplant and reduce relapse risk post-transplant. In a seminal report, they described the first 100 adult and paediatric patients with relapsed acute leukaemia who were transplanted in 1971–1975 following TBI-based conditioning (5). The combination of TBI and Cyclophosphamide was well-tolerated and was associated with long-term remission in 13% of patients, which was sustained in 8% (6). These results suggested that TBI-based conditioning for HSCT offered a survival advantage over chemotherapy in patients with end-stage disease, which prompted this approach to be trialled in the late 1970's in adult and paediatric patients with less-advanced leukaemia (7).

In the 1980's, attempts began to develop effective conditioning regimens that did not contain TBI, led by the John Hopkins group in Baltimore (8). They added the alkylating agent Busulfan to Cyclophosphamide to create the first chemo-conditioning regimen to be trialled. The addition of Busulfan aimed to provide equivalent myeloablation and leukaemia-free survival to TBI conditioning but with reduced toxicity. Chemo-conditioning with Busulfan 16 mg/kg and Cyclophosphamide 200 mg/kg or 120 mg/kg were used; both regimens induced long-term remission but the lower toxicity associated with Busulfan and Cyclophosphamide 120 mg/kg came at the cost of potentially increased relapse risk (9, 10). In paediatric HSCT, Busulfan and Cyclophosphamide 200 mg/kg is generally well-tolerated and so continues to be preferred over Busulfan and Cyclophosphamide 120 mg/kg as a conditioning regimen.

In the early 1990's, the first four prospective, randomised controlled trials comparing TBI-based conditioning and chemo-conditioning were published by groups in France (11, 12), Scandinavia (13), and Seattle (14). The studies involved predominantly adult patients, although a small number of children were included. The most common indication for HSCT was myeloid disease [acute myeloid leukaemia [AML] or chronic myeloid leukaemia (CML)]; a minority of patients in the Scandinavian trial had ALL or lymphoma (13). In all four trials, patients received Cyclophosphamide 120 mg/kg. Those randomised to the chemo-conditioning received Busulfan 16 mg/kg. In the TBI arms, regimens varied with most receiving 12 Gy in fractionated doses. When first published, at a relatively short follow-up of 24–42 months, DFS was superior in patients that received TBI-based conditioning vs. chemo-conditioning for AML in CR1 in the French multicentre study (72 vs. 47%, p < 0.01) (11) and for adults with advanced myeloid or lymphoid disease in the Scandinavian randomised controlled trial (68 vs. 54%, p = 0.05) (13). In contrast, chemo-conditioning with Busulfan-Cyclophosphamide achieved equivalent DFS to TBI-based conditioning in patients with CML in results published by the Seattle (14) and French group (12). A subsequent meta-analysis of these studies, and an additional randomised controlled trial comparing conditioning with Busulfan-Cyclophosphamide against that with TBI and etoposide, confirmed a non-statistically significant trend toward better overall survival (OS) and DFS with TBI-based conditioning (15).

This trend favouring TBI over chemo-conditioning, particularly in AML, was supported by the publication of the long-term data of the four trials. At a median follow-up of 10.8 years, Blaise et al. continued to show that TBI-Cyclophosphamide was associated with statistically significant higher DFS and OS and decreased relapse rates and transplant-related mortality compared with conditioning with Busulfan-Cyclophosphamide in patients with AML (TBI-Cyclophosphamide: 10-year OS 59%, DFS 55%; Busulfan-Cyclophosphamide: 10-year OS 43%, DFS 35%) (16). In the update of the Scandinavian study at 7 years of follow-up, OS was also higher in the TBI group (63% with TBI-Cyclophosphamide vs. 54% with Busulfan-Cyclophosphamide group) but this difference was not statistically significant (17). Similarly, when Socie et al. combined the data from the original four trials, a non-statistically significant 10% lower OS was observed in patients with AML who received conditioning with Busulfan-Cyclophosphamide compared with in those who received TBI-Cyclophosphamide [projected 10-year survival: 51% for Busulfan-Cyclophosphamide vs. 63% for TBI-Cyclophosphamide, 95% confidence interval (CI) 52–74%]. No statistically significant difference in OS or DFS was observed among patients with CML, as in the original studies (18).

In 2000, Davies et al. published a large study conducted in paediatric patients comparing TBI-based and chemo-conditioning regimens. This retrospective International Bone Marrow Transplant Registry (IBMTR) analysis included children with ALL who received a matched sibling HSCT after TBI/Cyclophosphamide or oral Busulfan-Cyclophosphamide. The incidence of relapse was similar between arms, suggesting that chemo-conditioning with Busulfan may not be inferior to TBI in preventing relapse. However, the higher non-relapse mortality (NRM) in the Busulfan arm led to TBI-based conditioning being associated with a superior leukaemia-free survival over Busulfan-based conditioning (50 vs. 35%, respectively; p = 0.005) (19).

The IBMTR study was shortly followed by publication of the first randomised controlled trial in paediatric patients comparing TBI-based and chemo-conditioning regimens: the Paediatric Blood and Marrow Transplant Consortium (PBMTC) study (20). This small study compared outcomes with chemo-conditioning with Busulfan, etoposide, Cyclophosphamide and anti-thymocyte globulin to those with TBI-based conditioning including Cyclophosphamide, etoposide +/- anti-thymocyte globulin. Relapse rates were similar between groups, yet NRM rates were higher in the Busulfan-Cyclophosphamide group. Bunin et al. concluded that “significant concerns regarding late effects, particularly secondary cancers, continue to make conditioning without radiation a potential attractive option, but additional studies are required to develop a safe, effective regimen.”

Despite these data, many centres replaced TBI-based protocols with Busulfan-based conditioning, particularly for myeloid diseases. However, over the ensuing decade, TBI retained its central role in conditioning for ALL. This was reinforced by evidence within in the literature. For example, a study looking at patients with ALL in CR2 concluded that TBI followed by HSCT compared to chemotherapy alone reduced the rate of relapse for children with early first relapse (21).

At the same time, there was continued recognition of the long-term burden following TBI, including an increased risk of breast cancer (22) and thyroid cancer (23). Moreover, the association between an increased risk of second solid cancers and age at the time of TBI was reported (24).

An important point is that in all the above studies, the Busulfan preparation used was oral, not intravenous (IV). The highly variable absorption rate and bioavailability of Busulfan, adding to its variable clearance, led to the development of the IV Busulfan formulation (25, 26). IV Busulfan enables better control of the cumulative exposure to Busulfan through therapeutic drug monitoring (TDM) (27, 28). In a retrospective trial that included paediatric ALL patients, Bartelink et al. reported an improved event-free survival (EFS) (83 vs. 30%, respectively; p < 0.001) and OS (83 vs. 53%, respectively; p = 0.016) accompanied with a decrease risk of veno-occlusive disease (VOD) under TDM-guided IV Busulfan compared with fixed-dose oral Busulfan (27). Although most centres have moved to the IV route, oral administration of Busulfan in paediatric HSCT is still used. Of note, a retrospective registry-based study on 460 transplanted children with leukaemia showed similar outcomes for both IV and oral formulations of Busulfan, but it was suggested that this was likely due to the routine use of Busulfan TDM (29).

With recognition of the life-long consequences of irradiation in young children, a convergence of shared thoughts and ideas led to the creation of the protocol that became the FORUM international, randomised controlled trial (Clinicaltrials.gov identifier: NCT02670564). The rationale included the following points:

The FORUM trial compared TBI (12 Gy) plus etoposide vs. chemo-conditioning with Fludarabine and Thiotepa combined with either Busulfan or Treosulfan (by country preference) in paediatric patients with ALL in CR who were between the ages of 4 and 21 years at HSCT. Twenty-one countries were involved in this large, prospective, Phase III study. The original intention was to recruit 1,000 patients over 5 years; however, the trial was stopped in March 2019 after 417 patients had been randomised due to early results indicating superiority of the TBI arm. The early results of FORUM were published in 2021 and confirmed that TBI conditioning was superior to chemo-conditioning, with a 16% higher 2-year OS (91 vs. 75%, respectively; p < 0.0001) and reduced cumulative risk of relapse (12 vs. 33%, respectively; p < 0.0001). Treatment-related mortality (TRM) was similar between the groups (34).

With FORUM showing a clear early benefit favouring TBI, we have to rethink how conditioning therapy in childhood ALL might otherwise be improved. Options include:

We now explore each of these themes in turn.

Busulfan with TDM is recommended in paediatric HSCT for several reasons. Firstly, Busulfan has a demonstrated exposure-response relationships and narrow therapeutic window, so small variations in exposure can result in poor clinical outcomes. Secondly, despite the improved predictability of PK obtained using IV formulations, due to the bypass of the unpredictable absorption phase, the inter-individual and intra-individual PK variability in Busulfan elimination and exposure remain substantial. The American Society for Blood and Marrow Transplantation recommends TDM-based dose adjustments for paediatric patients receiving myeloablative Busulfan-based conditioning therapy (35).

The association between Busulfan exposure and outcomes in paediatric patients with varying malignant diagnoses, including ALL, has been reported in many studies (Table 1) (36–55). The therapeutic window for Busulfan recommended by the European Medicines Agency (EMA) is AUC_6h 900–1,500 μM.min (daily AUC of 14.8–24.6 mg.h/L) (56, 57). This target was originally derived from studies in adult HSCT patients using oral Busulfan. Exposure higher than 1,500 μM.min has been associated with increased toxicities such as sinusoidal obstruction syndrome (SOS) and acute graft-versus-host disease (GvHD) (47, 58, 59), while exposures lower than 900 μM.min were associated with increased graft rejection and disease relapse (52, 60). This therapeutic window has been confirmed to be safe and efficacious in various studies of paediatric patients, including those with ALL (52, 61, 62). Nguyen et al. developed a dosing nomogram designed to reach this therapeutic target in paediatric patients, which the EMA has since recommended (57). One retrospective study in 138 patients, including 13 paediatric patients with ALL, investigated the impact of narrowing the EMA-recommended typical Busulfan therapeutic window to a local target AUC_6h 980–1,250 μM.min (daily AUC 16.1–20.5 mg.h/L). The efficacy (EFS and OS) and safety (SOS) outcomes evaluated in this study cohort were not improved using a narrower therapeutic window, suggesting that the EMA therapeutic window of 900–1,500 μM.min (daily AUC of 14.8–24.6 mg.h/L) is the most appropriate for children (53).

Another target for Busulfan dosing is based on steady-state concentration (Css). Css values can be expressed as AUC values by multiplying the Css value by the inter-dose interval. The reported optimal Css window of Busulfan is 600–900 ng/mL, corresponding to a daily AUC of 14.4 −21.6 mg.h/L (43, 44), which is only slightly lower than another narrowed therapeutic window recommended by the US Food and Drug Administration (FDA) (daily AUC 14.8–22.2 mg.h/L) (63). A recent meta-analysis by Feng et al. showed that the typical lower cutoff of 900 μM.min (daily AUC 14.8 mg.h/L) was strongly associated with the risk of graft failure (AUC ≥900 μM.min vs. <900 μM.min: Relative risk (RR) 3.666; CI 1.419–9.467), while the FDA cutoff (1,350 μM.min; daily AUC 22.2 mg.h/L) was more strongly associated with the risk of SOS than the EMA target (AUC ≤ 1,350 μM.min vs. >1,350 μM.min: RR 0.370; CI 0.205–0.666) (64). This study suggested that the FDA upper AUC cutoff (1,350 μM.min, daily AUC 22.2 mg.h/L) is safer in paediatric patients in terms of protection from SOS.

Much of the discussion about the Busulfan exposure metric has been superseded with the international harmonisation process to adopt uniform units of mg.L.h (65), as used in the largest retrospective study to date on the association between Busulfan exposure and outcomes in paediatric patients (36). Of the 674 patients enrolled in that study by Bartelink and colleagues, 41% were diagnosed with malignancies but only 5% had ALL (36). Based on EFS as the main criteria, the researchers found the optimal therapeutic window to be 78–101 mg.h/L, corresponding to a daily AUC of 19.5–25.3 mg.h/L. This target was shown to be optimal regardless of patients' malignant diagnoses. This new therapeutic target is included within the EMA target, with a slightly higher upper range (25.3 vs. 24.6 mg.h/L, respectively). However, it is higher than the FDA target, which was reported to be associated with a decreased SOS risk (64). This therapeutic window proposed by Bartelink et al. was also associated with acceptable acute toxicity (defined as acute GvHD and SOS) and occurrence of chronic GvHD. In response to a letter to the editor by Paci et al. (66), Bartelink et al. demonstrated that EFS was significantly reduced when targeting the lower end of the EMA threshold (AUC 59–78 mg.h/L) (41). The different studies show that there is still no consensus on the optimal cumulative exposure to Busulfan for paediatric patients due to heterogeneous data. Future well-designed, prospective investigations should further establish the optimal target window of Busulfan. However, it is widely agreed that TDM-guided dose adjustment of Busulfan is required to reach the desired target exposure in the paediatric HSCT setting, especially in neonates and small children for whom Busulfan PK is more unpredictable (67).

Studies have also shown that HSCT outcomes are not only associated with cumulative exposure to Busulfan but also with per-dose exposure. The AUC or Css of the first dose of Busulfan has been reported to be associated with toxicities of Busulfan as well as transplant outcomes. As shown in Table 1, a study from Ansari et al. reported that a first-dose Css <600 ng/mL (AUC_6h <3.6 mg.h/L) was associated with improved OS and EFS, a lower NRM and a lower incidence of relapse and acute GvHD of grade II to IV compared to patients with Css > 600 ng/mL (38). The other toxicities reported (SOS, lung toxicities, and haemorrhagic cystitis) showed trends of lower incidence in patients receiving Busulfan with a first-dose Css <600 ng/mL (AUC_6h <3.6 mg.h/L) compared to patients with Css > 600 ng/mL. A similar association between this exposure cut off and better NRM, OS, and EFS was later demonstrated in a larger multicentre population (39). In the latter study, the association between exposure and treatment-related toxicity (TRT) risk, comprising acute GvHD of grade I–IV, was shown to depend on glutathione S-transferase A1 (GSTA1) metabolic capacity (39).

Another study reported the association between SOS with the per-dose PK parameters of Busulfan in 293 patients including 125 with haematological malignancies (50). In the univariate analysis based on logistic regression, the maximal concentration after Busulfan infusion ended, and the first-dose AUC, but not the cumulative AUC, were associated with the occurrence of SOS. In the same study, engraftment only significantly associated with cumulative AUC. Interestingly, a study by Kerl et al. reported an increased risk of SOS with AUC_6h >1,500 μM.min (daily AUC >24.6 mg.h/L) in patients receiving Busulfan four times daily but not in patients receiving once daily Busulfan (48). These studies provide evidence that per-dose exposure to Busulfan could impact the outcomes and incidence of toxicity in paediatric patients. Accurately targeted first doses of Busulfan before TDM is performed should enable clinicians to avoid the toxicities and poor outcomes related to higher per-dose exposure. A planned future analysis of Busulfan PK data from the FORUM trial will enable better understanding of the association between Busulfan exposure and outcomes in a homogenous cohort of paediatric ALL patients. A similar analysis will be performed of Busulfan exposure in AML patients in the ongoing Myechild01 trial (Clinicaltrials.gov identifier: NCT02724163). The target Busulfan exposure in FORUM and TDM adjustment settings were not harmonised; rather, they depended on the local clinical practise in each transplantation centre (34). The upcoming analysis of the FORUM PK data will enable the researchers to explore a potentially heterogeneous Busulfan exposure among patients and its relationship to patient outcomes. This heterogeneity in patient exposure could partly explain the inferiority of Busulfan-based regimens to TBI, and the analysis of the Busulfan PK data from FORUM will explore this.

In HSCT, Busulfan was originally administered during 4 days of conditioning, four times daily (every 6 h). A once daily oral or IV Busulfan schedule has been reported to be safe and efficacious in paediatric patients (27, 68–72). One study in paediatric patients receiving IV Busulfan compared SOS risk between once-daily and four-times-daily dosing, finding a similar risk with each schedule (48). However, an association between exposure and SOS was only observed in patients receiving Busulfan four times a day, probably due to the presence of other risk factors. More recently, Philippe et al. showed that the risk of SOS was associated with the maximum concentration (C_max) of Busulfan. While the cumulative AUC should be equivalent between once-daily and four-times-daily dosing, the C_max obtained with once-daily dosing is systematically higher than that obtained with four-times-daily dosing.

The study by Philippe et al. included 11 patients who received once-daily or twice-daily Busulfan, among which nine (81.8%) patients experienced SOS (50). In contrast, other studies in paediatric patients have observed a lower occurrence of SOS in paediatric patients who received once-daily IV Busulfan dosing (69, 70). Further studies should address the comparison between once-daily and four-times-daily IV Busulfan dosing in paediatric patients, in terms of efficacy and toxicity outcomes.

The once-daily Busulfan dosing schedule has many advantages. Xhaard et al. showed that once-daily Busulfan dosing was associated with better patient comfort related to reduced nausea and vomiting and less infusions (73). Once-daily dosing was perceived by healthcare professionals to be safer and less error prone, in addition to reducing workload and allowing smoother treatment management. In addition, once-daily Busulfan dosing reduces transplantation-related costs (74). Dividing the total Busulfan dose over 16 doses (four times a day schedule) provides more opportunity for dose adjustments, which may make it easier to target the desired cumulative exposure. Four times daily regimen have enabled to adjust the dose of Bu from the third dose onwards during the 1st day of Bu (depending on access to a biomedical analysis laboratory), which is not feasible with once daily dosing. However, TDM-guided dose adjustment from the 2nd day of Busulfan infusion is feasible with once-daily dosing and allows cumulative exposure to be readily estimated (75). The less commonly used twice daily Bu schedule (every 12 h administration, eight doses) allows dose adjustments from the 2nd day of Bu treatment, whilst reducing the workload associated with the four times daily dosing.

When the use of TDM accounts for the interindividual PK variability of Busulfan, so allowing you to target the desired cumulative AUC, why is it important to individualise the first dose? Relying solely on TDM for dose adjustment has some limitations as well as having time-constrained limits on how quickly and how often dose adjustments can be made. Studies have highlighted the per-dose therapeutic window of Busulfan and the necessity to target early in administration the desired therapeutic window (39, 50, 66, 76). The personalization of the first dose of Busulfan should minimise the risk of overexposure and any associated acute toxicity. In combination with efficient TDM, this strategy could enable control of cumulative Busulfan exposure throughout conditioning treatment, which may optimise the outcomes. Because engraftment is associated with cumulative underexposure to Busulfan (50), first dose under-exposure seems to be less critical as it could be accounted for via TDM-guided dose adjustment. Even so, first dose underexposure could lead to the need for substantial dose augmentation, thus reaching a toxic C_max associated with SOS occurrence (50). This is particularly of concern in the case of once-daily dosing, where plasma concentrations reached are high and dose modifications are more considerable to correct the desired exposure in only four administered doses. Dividing the first dose into two half doses counteracts this risk and has been used successfully for many years in some centres (77).

The two strategies that can be implemented to personalise the first dose of Busulfan are the “test dose strategy” and the “first dose strategy.” The test dose strategy consists of the administration of a small dose of Busulfan ≥2 days before the start of the typical 4-day Busulfan conditioning course. This is particularly useful when the laboratory performing the Busulfan PK analysis is not on-site. The Busulfan PK obtained from the test dose is used to modify the first full dose according to the predicted PK and the chosen target exposure (78, 79). The first dose strategy consists of the personalization of the first dose according to the demographic and clinical attributes of the patient (age, weight, etc.). This strategy is based on dosing nomograms or algorithms derived from population PK studies. The advantage of this strategy is that it better considers each patient's individual characteristics for the recommendation of accurate first doses. As shown in Table 2, body size metrics (actual body weight, body surface area, fat-free mass, etc.) are covariates consistently reported to explain Busulfan PK variability in paediatric patients and are used for dose calculations (38, 57, 63, 66, 67, 80, 82–100).

Several studies have also included an age-based metric to describe the ontogeny and maturation of Busulfan clearance. Such a model has been shown to result in accurate PK predictions and selection of the first dose in paediatric patients (75, 101–103).

For both the test dose and first dose strategies, intraindividual (i.e., inter-day) PK variability of Busulfan mandates that repeat PK testing is needed to assess the cumulative AUC over the course of therapy (78–80, 82, 104). In this way, personalised first doses coupled with efficient TDM permits the desired Busulfan exposure to be targeted. More importantly, repeat measurements used for TDM allow the cumulative exposure to be measured: this can inform future studies, particularly as additional drugs are added to the backbone of a Busulfan-based conditioning, so optimising the outcome and minimising the risk of toxicities related to under- or over-exposure.

In recent years, in an effort to accurately predict Busulfan PK in paediatric patients, the influence of biomarkers explaining Busulfan PK became an area of interest. Table 3 summarises the studies on the association between pharmacogenetic markers and Busulfan PK in paediatric HSCT patients (39, 71, 82, 83, 89, 97, 98, 102, 105–107, 109–120).

As Busulfan is mainly metabolised by glutathione-S-transferases (GSTs) (121, 122), clinical investigations on the influence of genetic polymorphisms related to GST activity on Busulfan PK were initiated in the early 2000's (116). Table 3 shows that Busulfan PK is mainly associated with haplotypes of the promoter regions of GSTA1 (18 studies) and GSTM1 (7 studies). The association between GSTP1 and GSTT1 with Busulfan PK is scarce, probably due to their less important role in Busulfan metabolism compared with A1 and M1 isoforms (123). GSTA1-^*B haplotypes have been associated with decreased Busulfan clearance, implying an increased exposure to Busulfan. This is due to decreased GSTA1 expression with ^*B haplotypes (39, 124). Initially, ^*A and ^*B haplotypes of GSTs were determined using one single nucleotide polymorphism (SNP) (either 52G/A rs3957356 or -69C/T rs3957357, in linkage disequilibrium) (113, 116, 117). The association of these haplotypes with Busulfan PK are still being studied (97). More recently, GSTA1 haplotypes have been shown to be more complex, requiring the genotyping of at least four SNPs of the GSTA1 promoter (39, 82, 124). In fact, sub-haplotypes within ^*A and ^*B have significantly different gene expression potentials. Within ^*A haplotypes, the ^*A1 sub-haplotype has a decreased expression potential than ^*A2 and ^*A3 haplotypes. The ^*A2 haplotype has been associated with a significantly increased clearance and thus lower Busulfan exposure (108). Within ^*B haplotypes, which are all associated with poor Busulfan metabolism, patients carrying the sub-haplotype ^*B1b have significantly decreased Busulfan metabolism and clearance compared with other ^*B haplotypes (39).

These different gene expression potentials have enabled the classification of patients into three (82, 89, 102, 124) or four (39) groups according to their capacity to metabolise Busulfan. GSTA1 polymorphisms have been also associated with the clinical outcome of HSCT (SOS, acute GvHD, transplant-related mortality, engraftment, and survival) (39, 97, 107, 108). These associations are likely to be related to differing exposure to Busulfan according to the GSTA1 haplotype. More recently, genetic polymorphisms explaining the metabolising capacity of GSTA1 have been detected as a significant covariate influencing Busulfan clearance: two recent models included as significant covariates GSTA1 metabolic groups associated with Busulfan metabolic capacity, based on GSTA1 sub-haplotypes (82, 89). Predictions based on these models have enabled researchers to accurately achieve Busulfan AUC within the Busulfan EMA therapeutic window in around 80% of the patients from an independent cohort of which 13% of patients had ALL (82). The addition of GSTA1 metabolic capacity to the model seems to have improved the accuracy of first dose selection.

The pharmacogenomic-based models are likely to enable accurate targeting of Busulfan exposure from the beginning of Busulfan conditioning, limiting the need for dose adjustments. A prospective validation is still required for the implementation of this dosing recommendation, although the model was validated in an external cohort. In addition, the feasibility of the implementation of GSTA1 genotyping in routine clinical practise needs to be assessed. These aspects are being addressed in the current BuGenes01 multicentre, prospective randomised trial (Clinicaltrials.gov identifier: NCT04822532), in which paediatric patients undergoing HSCT will be randomised to either a pharmacogenomic-based first dose algorithm or the best-performing dosing algorithm currently used (86). Personalising the first dose of Busulfan in paediatric patients should enable researchers to appraise the unpredictability of Busulfan PK, thus limiting large dose adjustments that could subsequently overexpose these patients (86).

GST polymorphisms have also been associated with poor HSCT outcomes and TRT, as shown in Table 3. These associations were reported in patients carrying GST haplotypes expressing poor metabolising phenotypes, for example GSTA1^*B, GSTM1-null, and GSTP1 313^*G haplotypes. Polymorphisms of GSTA1, GSTM1, and GSTP1 were reported as risk factors for SOS (39, 107, 116, 125) and acute GvHD (39, 107, 108, 111), while GSTA1 and GSTM1 have been associated with combined TRTs (39, 107, 108). GSTM1 was associated with graft rejection and mortality within 30 days post-transplant (111), while GSTA1 was associated with neutrophil recovery and survival (97). Whether these associations are solely due to the influence of GST polymorphisms on Busulfan PK is questionable. In a study by Ansari et al., increased TRT was associated with GSTA1 polymorphisms in multivariate logistic regression even when Busulfan exposure was accounted for (39). GSTA1 seems to have a direct influence on the transplant outcomes in addition to influencing Busulfan PK. Furthermore, the same study demonstrated that, in patients within or below the therapeutic window (Css 600 – 900 ng/ml, corresponding to daily AUC of 14.4 −21.6 mg.h/L), GSTA1 haplotypes expressing poor metabolic capacity were associated with higher TRT risk (HR 4.4; p < 0.0005) (39). This association was not observed in patients overexposed to Busulfan (Css >900 ng/mL) for whom TRT rates were very high irrespective of the GSTA1 genotype. This suggests that when patients are within therapeutic exposures, the influence of the poor metabolising capacity of GSTA1 on TRT occurrence is independent of PK. GST polymorphisms could therefore influence toxicities and outcomes of HSCT independently of Busulfan exposure. This aspect should be further explored in future studies of patients receiving Busulfan.

Other genetic markers for Busulfan conditioning toxicities have been reported. In paediatric patients, CYP2B6, CTH, MTHFR, HPSE, UGT2B10, and KIAA1715 were reported as risk factors for SOS (126). The risk related to the combined presence of these markers remains to be studied further. Interestingly, CTH c.1364 TT, a gene coding for cystathionase (an enzyme that participates in the glutathione synthesis pathway), was reported to be associated with SOS risk in combination with GSTA1^*B^*B (reduced function) (127). The data from the pharmacogenomic add-on study of the FORUM study will address this question. Recent studies have reported that polymorphisms of MGMT (128), ERC1, PLEK, NOP9, and SPRED1 were associated with increased GvHD risk (129) in paediatric HSCT, both studies included ALL patients. Donor polymorphisms of genes encoding interleukins (ILs), such as IL-6, interferon γ (IFNγ), and IL-7Rα, have also been associated with GvHD in studies including adult and paediatric patients receiving HSCT, both studies including ALL diagnoses (130, 131). The inclusion of these genetic variants in prognostic models for TRTs could be useful to guide personalised interventions. Combined with other known risk factors for SOS, genetic markers for increased risk of SOS could aid the selection of reduced toxicity chemo-conditioning regimens (e.g., those composed of maximum of two alkylating agents, or/and Fludarabine based), and the administration of defibrotide prophylaxis. Furthermore, the presence of markers of increased GvHD risk could contribute to the choice of GvHD prophylaxis.

Prior to the FORUM study, published experience of Treosulfan use in patients with ALL was scarce. Wachowiak et al. retrospectively evaluated 51 children with high risk or advanced haematological malignancies (17 with ALL) transplanted between 2000 and 2005 with Treosulfan-containing conditioning regimens and found no early regimen-related fatal toxicity and a NRM of 16% at 4 years (142). In a retrospective analysis of 109 children transplanted using Treosulfan-based conditioning between 2003 and 2009, approximately half of children had malignancy and 16 had ALL. Treosulfan was combined with agents such as Fludarabine, Thiotepa, and Melphalan. Skin toxicity was frequent but mild with Treosulfan, mucosal toxicity was reduced compared with Busulfan, VOD occurred in 3%, and seizures in 4% of patients (143). Boztug et al.'s retrospective study of 193 children and adolescents with malignant haematological disorders who received HSCT after Treosulfan-based conditioning therapy included 71 with ALL. In accordance with previous studies, toxicity of Treosulfan was low and mainly gastrointestinal in this study. VOD and neurological toxicity were rare. No association of toxicity with type of disease or Treosulfan dose was found. TRM was at 14% (144).

In a Phase II, prospective, multicentre study conducted by Kalwak et al., Treosulfan-Fludarabine-Thiotepa conditioning was investigated in 65 children with a haematological malignancy (3 ALL, 29 AML, 10 myelodysplastic syndrome and 3 juvenile myelomonocytic leukaemia). Treosulfan was dosed by body surface area (BSA), with those patients ≤ 0.5 m² receiving 10 g/m²/day; those >0.5–1.0 m² 12 g/m²/day and those >1.0 m² 14 g/m²/day for 3 days. Overall, 98.5% of patients achieved engraftment, with complete donor chimerism in 92.6% at 12 months. The most frequently reported toxicities of grade 3–4 were oral mucositis (43.1%), infections (30.8%), nausea and vomiting (16.9%), skin and subcutaneous tissue disorders (12.3%), and hepatic VOD (1.4%). NRM was estimated to be low, at 3.1% (145).

To date, only preliminary results of the Treosulfan arm in the FORUM trial have been published: the most frequent early grade 3–4 toxicities included infections (65%) and stomatitis (56%), while skin toxicity of grade 3–4 was present in 9% of patients. Of concern, neither the Treosulfan nor Busulfan arm compared favourably with TBI with regards to TRM in the modified as-treated population (12, 6, and 3%, respectively; p = 0.1103). Analysing the two chemotherapy groups together, the higher TRM compared to the TBI arm (9 vs. 2%, p = 0.027) contributed to the lower overall survival, triggering the cessation of randomisation to the chemotherapy arms (34). This raises concerns of duplicating what was seen in the PBMTC study (20), with a more intensive and thus toxic combination of chemotherapy agents not comparing favourably with the well-known early toxicity profile of TBI.

Prior to the FORUM trial, children with ALL receiving Treosulfan-based conditioning therapy prior to HSCT were reported in cohorts together with non-malignant disorders (143) or with other (myeloid) malignancies (142, 145). The numbers of paediatric ALL patients included in trials did not exceed 71 in retrospective cohorts (144) or 23 in prospective trials (145). The more robust outcome data for Treosulfan-based conditioning come from studies with myeloid malignancies in adults (146–148).

In the retrospective study of Wachowiak et al. referred to above, the estimated 4-year probability of DFS was 71% for those with myeloid malignancies and 41% in the 20 patients with lymphoid malignancies (predominately ALL), with an acceptable relapse incidence of 24% at 4 years (142). Beier et al., in a cohort including 16 patients with ALL and 11 with AML, reported a 3-year EFS of 49% and with predominant cause of death being relapse (143). In the European Society for Bone and Marrow Transplantation (EBMT) Paediatric Diseases Working Party retrospective analysis of Treosulfan-based conditioning for Haematological malignancy, the 3-year EFS was 45% and disease-related mortality 32% for the 71 ALL patients (144). The addition of an additional alkylator (either Thiotepa or Melphalan) to the Treosulfan-Fludarabine backbone resulted in significantly better OS.

One should bear in mind that these early retrospective studies selected patients who were felt to be at high risk for regimen-related toxicity, especially pulmonary and hepatic (VOD) toxicity associated with standard of care myeloablative regimes (TBI or Busulfan based). In 23 prospectively studied paediatric ALL patients given Treosulfan-Fludarabine-Thiotepa, Kalwak et al. estimated the relapse/progression incidence to be 26.1%, the relapse/progression free survival to be 69.9% and OS to be 78.3% at 36 months (145). Outcomes were comparable across each of the BSA-based Treosulfan doses (10, 12, and 14g/m²).

The most valuable knowledge on the efficacy of Treosulfan-Fludarabine-Thiotepa conditioning before HSCT for paediatric ALL comes from the 99 patients with ALL randomised to this regimen in the FORUM trial (34). Outcomes in the Treosulfan arm in the modified as-treated population-−58% EFS, 77% OS, 31% cumulative incidence of relapse and 12% TRM at 2 years—were significantly lower than the TBI arm (85% EFS, 91% OS, 12% cumulative incidence of relapse and 3% TRM at 2 years), clearly not supporting the use of an unadjusted Treosulfan regimen for patients eligible for TBI.

One difference between the Busulfan and Treosulfan arms in the FORUM study is that we know a significant proportion of patients in the Busulfan arm will have had PK analysis performed, with subsequent TDM. In contrast, we do not expect any of the Treosulfan-assigned patients to have had TDM. A fundamental question remains unanswered: is there a meaningful relationship between drug exposure and clinical outcome for Treosulfan and will optimization of dose and TDM improve the EFS vs. TBI when compared with the non-TDM-targeted Treosulfan usage in FORUM?

We know that, like most of the drugs we use in conditioning, there is high inter-patient variability in exposure to Treosulfan (149). To date, most of the PK data for Treosulfan was collected in patients with non-malignant disease. Van der Stoep et al. performed a prospective multicentre study in 77 children undergoing HSCT (84.4% of whom had non-malignant disease), focussing on the PK profile of Treosulfan. Their results showed that there is a relationship between Treosulfan exposure and early toxicity. Patients with higher exposure (AUC >1,650 mg.h/L) had an increased risk of developing grade 2 or higher mucositis and skin toxicity. No correlation between Treosulfan exposure and the early clinical outcome parameters (engraftment, acute GvHD or donor chimerism) was found (149). A prospective study in two UK centres looked at Treosulfan PK and PD in children undergoing allogeneic HSCT mainly for primary immunodeficiency after Treosulfan-Fludarabine conditioning. An association between high AUC and mortality as well as low AUC and poor engraftment was shown (150).

Mohanan et al. studied 87 patients with thalassaemia major undergoing allogeneic HSCT. Treosulfan clearance of <7.97 L/h/m² was significantly associated with poor OS and EFS; where as high Treosulfan clearance (>7.97 L/h/m2) and low AUC (<1,828 mg.h/L) showed a trend toward better OS (151).

Thus, it can be postulated that there is likely to be an association of outcome and toxicity parameters with Treosulfan exposure, yet perhaps the improved safety profile of Treosulfan over Busulfan makes this more difficult to establish until we have available larger studies on more uniform populations. In most protocols, Treosulfan is administered over 3 consecutive days in doses of 10–14 g/m²/day, with the dose adjusted according to age or body weight. Despite the dose reduction to 10 g/m² in infants, admittedly with a variety of diagnoses, including many with non-malignant disease, Treosulfan exposure remained higher compared with older children receiving 14 g/m² (149). We may find that it is in these younger (and so smaller) patients where Treosulfan TDM has a role. In order to identify and quantify sources of variability in drug concentration and to predict concentrations in individual patients, PK models have been developed (152–154). Clearly, the currently available data are not sufficient to inform a practise guideline for TDM of Treosulfan in paediatric ALL—the relationship of Treosulfan exposure to leukaemia-free survival has not been described. A number of clinical trials incorporating Treosulfan PK evaluation are underway that may provide additional insights. In particular, the PK data on Treosulfan from the FORUM trial are eagerly awaited.

One advantage of Clofarabine is that it is not associated with the neurotoxicity seen with other similar nucleoside analogues. In order to reduce toxicity but sustain efficacy, studies both in vitro and in vivo have been done where nucleoside analogues replace alkylating agents. In vitro cell line studies showed the clear synergistic cytotoxicity of Clofarabine and Fludarabine, which was further enhanced by adding Busulfan. This finding led to the combination of Clofarabine, Fludarabine, and Busulfan being investigated by the MD Anderson group (161, 162).

In that randomised controlled trial, 51 adult patients with high-risk myeloid leukaemias were randomised to receive Clofarabine-Fludarabine-Busulfan conditioning across four treatment arms that differed with respect to the Clo and Fludarabine dosing used. Initial findings were encouraging with regard to safety and antileukemic activity (162). Longer follow up of this expanded cohort (n = 70) confirmed the safety, OS and PFS advantage of the arms with higher Clofarabine doses and lower Fludarabine doses (163).

The same group studied Clofarabine and Busulfan in 107 adults undergoing HSCT for ALL (164, 165). With a median follow up of 3.3 years, 2-year leukaemia-free survival was 51% (being best in CR1 patients, at 62%), and NRM was 6% at day 100 and 18% at 2 years. These outcomes compare favourably with reports of adult patients with ALL in CR1 treated with myeloablative TBI-based regimens.

There are few data published on the use of Clofarabine for HSCT conditioning in paediatric patients. A retrospective analysis in paediatric AML using a common backbone of induction chemotherapy followed by three different chemotherapy conditioning regimens suggested that Clofarabine-Fludarabine-Busulfan had good anti-leukaemic activity with low NRM. In comparison, Busulfan-Cyclophosphamide was associated with higher relapse incidence, while Busulfan-Cyclophosphamide-Melphalan was associated with higher incidence of acute GVHD (166).

In a cohort of 60 paediatric ALL patients undergoing HSCT after Clofarabine-Fludarabine-Busulfan conditioning, the 2-year estimated EFS probability was 72.0% ± 6.0, with significantly lower EFS observed in patients with MRD positivity prior to HSCT. Two-year TRM probability was low at only 5.0% ± 2.8 and no VOD was seen.

At the time of writing, there were no ongoing clinical studies of Clofarabine use in HSCT conditioning regimens.

For Busulfan, acute and chronic toxicities remain a matter of concern even when Busulfan target exposures are strictly controlled (50, 167). As shown by several studies, the use of multiple alkylating agents in conditioning regimens is a predictor of acute toxicity in paediatric patients (36, 37). For this reason and based on adult experience, the nucleoside analogue Fludarabine—an inhibitor of DNA, RNA and protein synthesis—has been introduced as an immunosuppressive agent in the replacement of Cyclophosphamide in paediatric transplantation. The majority of data comparing Fludarabine-Busulfan to Busulfan-Cyclophosphamide conditioning regimens come from adult patients, although some of these studies included children and adolescents. The meta-analysis by Ben-Barouch et al. included studies with paediatric ALL patients (168). The authors reported that a lower risk of NRM was associated with Fludarabine-Busulfan vs. Busulfan-Cyclophosphamide, while OS was similar between the two regimens. The same study found that Fludarabine-Busulfan was associated with lower risk of SOS than Busulfan-Cyclophosphamide. However, when only considering randomised controlled trials, the SOS risk was similar between the two regimens. A higher risk of microbiological infections was associated with the Busulfan-Cyclophosphamide regimen. Other assessed outcomes (GvHD, relapse, engraftment and mucositis) were similar between the two regimens. The meta-analysis concluded that Fludarabine-Busulfan and Busulfan-Cyclophosphamide regimens have similar efficacy, but Fludarabine-Busulfan regimens are slightly more favourable in terms of toxicity profile.

Two important studies have compared Busulfan-Cyclophosphamide and Fludarabine-Busulfan regimens in paediatric HSCT. In the first, Bartelink et al. compared the data of patients prospectively recruited 64 patients (9 ALL) who received Fludarabine-Busulfan conditioning with retrospective data of 50 (5 ALL) patients who received Busulfan-Cyclophosphamide. ALL patients received melphalan (Mel) in addition to Busulfan-Cyclophosphamide. Much like the picture in adults, EFS and OS were similar between conditioning groups, while the risk of TRT such as SOS, chronic GvHD, acute lung toxicity and viral reactivations were lower in patients who received Fludarabine-Busulfan (169). Rates of acute GvHD were similar between the two groups. As shown by more recent data, the use of three alkylating agents is correlated with the occurrence of acute toxicity compared to patients with two or one alkylating agent (36). Mel-containing conditioning regimens were also associated with acute toxicity risk (37). The use of Mel could therefore have contributed to the observed higher toxicity in Busulfan-Cyclophosphamide-Melphalan group in the study by Bartelink et al. A sub-analysis of that study that excluded ALL patients (for whom Mel was indicated), showed less toxicity in patients receiving Fludarabine-Busulfan compared with Busulfan-Cyclophosphamide. The comparison between outcomes of ALL patients receiving Busulfan-Cyclophosphamide-Melphalan vs. Fludarabine-Busulfan was not reported by the authors (169). The second study, by Harris et al., compared Fludarabine-Busulfan and Busulfan-Cyclophosphamide using retrospective data from 1,781 transplanted children. Post-relapse survival was inferior in patients receiving Fludarabine-Busulfan vs. Busulfan-Cyclophosphamide, leading to an inferior OS in those patients (170). In contrast to the Bartelink et al. study, this study showed no difference in transplant-related toxicity and TRM between conditioning groups (170). This suggests that one may still consider the addition of a third agent, but on the backbone of Busulfan-Fludarabine rather than Busulfan-Cyclophosphamide.

In contrast to Busulfan-Cyclophosphamide, there is evidence of a PK drug–drug interaction between Busulfan and Fludarabine. Two studies have shown a significantly decreased clearance of Busulfan when co-administered with Fludarabine (82, 171). As the effect sizes related to Busulfan co-administration reported in these studies were fairly small, the clinical significance of this interaction is likely to be minimal. As Busulfan has a narrow therapeutic window, even this small effect size should be considered for accurate dose individualisation of Busulfan. Furthermore, Busulfan-related toxicities in patients co-administered Busulfan and Fludarabine are also exposure dependent. A higher inter-dose variability was reported in patients receiving a Fludarabine co-administered with Busulfan, than that observed with Busulfan-Cyclophosphamide (172). TDM is therefore important to control for this increased PK variability observed when Busulfan is used alongside Fludarabine in conditioning regimens.

There is a small but emerging literature on Fludarabine PK in Paediatric Transplantation. Retrospective data suggested high levels were associated with more toxicity, particularly in the setting of renal impairment (173). A more recent prospective multicentre study again showed that renal impairment predictably increased AUC. In this paediatric study, it is likely that many of the patients had reduced intensity grafts, some received fludarabine alone and so the low TRM made it difficult to demonstrate if there was an relationship between exposure and TRM (174). Another paediatric study also found no association between exposure and clinically important end-points (175).

More interestingly, there has been a first attempt to look at the impact of the pharmacokinetics of Fludarabine in combination with Busulfan (176). Rather than a multivariate analysis of the impact of the PK of both Fludarabine and Busulfan independently and then looking for any interaction, the paper describes the impact of Fludarabine PK within a retrospective cohort of patients who were all given a set dose of 160 mg/m² of Fludarabine combined with what is described as a targeted dose exposure of Busulfan. In fact, although an AUC of 90 mg.h/L was targeted, the mean exposure to Buslfan achieved was 96.1, with a wide range of AUC from 59 to 120 mg.h/L. Within this large series of adult and paediatric patients, including some leukaemias, and with a consequent much higher rate of TRM (28%) than the purely paediatric studies quoted above, the authors found that higher exposure associated with more toxicity and lower levels associated with more rejections. They suggested that an optimal cumulative exposure could be targeted by refinement of the current surface area based dosing, or measured as part of a TDM strategy. Give the variability in the exposure to Busulfan, which was not explored in this retrospective study, this is an illustration of the way forward.

The chemotherapy drugs used in transplant conditioning are not given in isolation. Additional immunosuppression, depending on donor type and cell source, is added in, typically in the form of agents such as Anti-T cell polyclonal antibodies or monoclonal antibodies, such as Alemtuzumab. Their use is considered in a separate chapter of this issue.

The Busulfan-based protocol used in the FORUM study added Thiotepa (10 mg/kg divided into two doses) to the Busulfan and Fludarabine. This combination is based upon protocols mainly studied in adult patients (177, 178), umbilical cord blood transplantation (179–181), haploidentical HSCT (180, 182), and reduced intensity regimens (183). The rationale behind the addition of Thiotepa was to improve the engraftment rates in adult umbilical cord blood transplanted patients, which was insufficient under a Fludarabine-Busulfan regimen (179, 184, 185). The original protocols used only 3 days of Busulfan at 3.2 mg/mL daily, thus a lower cumulative dose than myeloablative regimens. In FORUM, this protocol was used as the Busulfan-based conditioning arm but with the standard 4 days of conditioning and myeloablative target exposures suggested. In adult AML, intensifying Fludarabine-Busulfan-Thiotepa conditioning with full myeloablative doses of Busulfan resulted in significantly lower relapse [hazard ratio (HR) 0.47; p = 0.005] but higher NRM (HR 2.69; p < 0.001) compared with a myeloablative Fludarabine-Busulfan regimen (178). Leukaemia-free survival and OS was similar between the two regimens. Fludarabine-Busulfan-Thiotepa has been reported also to result in a lower relapse rate (HR 0.6; p = 0.02) and similar OS compared with Busulfan-Cyclophosphamide in adult AML patients (177, 178). Fludarabine-Busulfan-Thiotepa had not been studied in the conventional matched donor setting in ALL paediatric patients prior to the FORUM study. It remains unknown if this combination results in optimal outcomes in paediatric ALL and should be tested against other Busulfan-based regimens in paediatric ALL patients is therefore needed.

As described in Section Optimizing the Use of Treosulfan above, the favourable toxicity profile of Treosulfan, combined with its limited activity when combined with Fludarabine alone, led to the addition of a third agent, often Thiotepa or Melphalan.

When using potentially toxic drugs at high doses for a short period of time, after gaining as much PK and PD information as possible from investigations of each single drug, it becomes important to look at the impact of the agents in combination. For ALL, we have added Thiotepa to Fludarabine partnered with Busulfan or Treosulfan, or used Clofarabine. We then have to consider the impact of the serotherapy used. It is naïve to believe that the complex relationship between disease and disease status, type of donor and cell source used after giving multi-agent chemotherapy combined with serotherapy will have a simple relationship to even complex descriptors of any one of the conditioning agents used. For the next phase of our international PK/PD effort, we should attempt to share data to integrate information regarding each element of the conditioning. In this way, we can move closer to our goal of optimising conditioning for each individual patient.