Treosulfan Pharmacokinetics and its Variability in Pediatric and Adult Patients Undergoing Conditioning Prior to Hematopoietic Stem Cell Transplantation: Current State of the Art, In-Depth Analysis, and Perspectives

Treosulfan, a water-soluble dihydroxy derivative of busulfan, is authorized in some European countries for the intravenous or oral treatment of advanced ovarian cancer. The intravenous regimen usually relies on drug doses of 5–8 g/m² administered every 3–4 weeks [1‐3]. Moreover, for the last 17 years, high-dose treosulfan has been investigated worldwide as a myeloablative agent in conditioning regimens prior to hematopoietic stem cell transplantation (HSCT). In this therapy, patients receive 10–14 g/m² of the compound for 3 consecutive days. Retrospective analyses of clinical phase I and II trial results indicate that in both children and adults, treosulfan-based conditioning is associated with low-early hepato-, pulmo-, and neurotoxicity compared with busulfan-based treatment. On the other hand, the regimens containing the above drugs seem to exert comparable myeloablative, immunosuppressive, and antileukemic efficiency because the incidences of complete donor chimerism and malignancy relapse are similar [3‐9].

In opposition to busulfan, treosulfan is a prodrug. At pH > 5, it undergoes a nonenzymatic intramolecular nucleophilic substitution that sequentially leads to (2S,3S)-1,2-epoxy-3,4-butanediol 4-methanesulfonate (S,S-EBDM) and then to (2S,3S)-1,2:3,4-diepoxybutane (S,S-DEB). These epoxy transformers are believed to account for the DNA alkylation and biological effects observed after administration of treosulfan [10‐12].

Presently, prospective, multicenter, clinical phase II and III trials are ongoing that directly compare treosulfan versus busulfan in conditioning prior to allogeneic HSCT in children (1 month–18 years of age) with nonmalignant diseases, and in adults (18–70 years of age) with acute myeloid leukemia and myelodysplastic syndromes. In another ongoing phase III trial, treosulfan-based conditioning is compared not only with busulfan-based conditioning but also with fractionated total body irradiation-based conditioning in children and adolescents undergoing allogeneic HSCT for acute lymphoblastic leukemia, i.e. with the most frequent indication for HSCT in the pediatric population. The aim of the other prospective multicenter clinical phase II trial is to describe the safety and efficacy of treosulfan-based therapy in hematological malignancy pediatric patients (1 month–18 years of age) [13, 14]. All these studies include pharmacokinetic evaluation to give precise recommendations on the optimal dosing of the prodrug prior to HSCT, which is yet unclear, particularly in children [6‐8, 13, 14]. This issue warrants identifying all clinically relevant factors affecting the pharmacokinetic parameters of treosulfan, especially its total clearance (Cl_tot).

To date, only a few papers have briefly reviewed the clinical pharmacokinetics of sole treosulfan, most recently in 2014 [3‐5]. Since that time, novel data have been reported that provided substantial insight into the topic. The data covered the kinetics of the nonenzymatic epoxy transformation of treosulfan, including the quantitative description of pH and temperature effect; the pharmacokinetics of S,S-EBDM in children; the discovery of the formation-rate pharmacokinetics of the epoxy transformers; the organ disposition of treosulfan and S,S-EBDM in rats; treosulfan population pharmacokinetics in humans; and studies on the association of treosulfan exposure with the toxicity and efficacy of HSCT conditioning [10, 11, 15‐25]. This review is the first to be exclusively devoted to the pharmacokinetics of treosulfan and its epoxides. The main focus is given to the in-depth analysis of the pharmacokinetic parameters of treosulfan and S,S-EBDM, and their significance for conditioning prior to HSCT. We discuss possible physiology- and treatment-related reasons of the interindividual variability of the compounds’ Cl_tot and volume of distribution, especially in children.

As presented in Tables 1 and 2, the V_ss and Cl_tot of treosulfan reported in children undergoing HSCT were characterized by relatively high interpatient variability, especially in the recent large multicenter study by van der Stoep et al. (coefficient of variation up to 86%) [24]. Lower variability of the parameters was observed in the cohorts of adult patients, but, interestingly, quite marked differences occurred between the clinical centers [16, 22‐29].

In attempts to identify the factors affecting the V_ss and Cl_tot of treosulfan and their interindividual variability, population pharmacokinetic models were developed [20‐25]. Van den Berg et al. [22] used pharmacokinetic data from 93 adults and 23 children (0.4–17 years of age), including a few infants (exact number was not given). Among BSA, age, BW, height, renal function, and use of diuretics, only BSA was found to be a covariate for the volume of the central compartment (V₁), volume of the peripheral compartment (V₂), and Cl_tot of treosulfan. The BSA allometric scaling factor for Cl_tot was equal to 1.29 (Cl_tot~BSA^1.29) [22]. In the four other models, the parameters were allometrically scaled using BW, and the scaling exponents were fixed at 1.0 for both V₁ and V₂, and 0.75 for Cl_tot [20, 21, 23, 24]. In addition, Danielak et al. [23] found that sex was not a significant covariate for V₁, V₂, and Cl_tot, however they noted that the studied group included only three girls. Ten Brink et al. [20] and van der Stoep et al. [24] did not mention testing any covariates, except BW. Surprisingly, in the model developed by Mohanan et al. [25], none of the covariates tested, i.e. age, BW, BSA, sex, liver size, liver fibrosis, ferritin levels, liver enzymes, and hemoglobin level, were found to explain the wide interindividual variability of treosulfan pharmacokinetics in thalassemia major patients aged 1.5–25 years. Despite age, renal function (creatinine clearance), and administration of diuretics not being identified as significantly influencing the V_ss and Cl_tot of treosulfan [22, 23], these factors should not be neglected in future population pharmacokinetic modeling in larger and more homogeneous cohorts of HSCT patients. In particular, age may prove to be a significant variable in a sufficiently numerous cohort including young infants and children because of dynamic changes in the percentage of TBW and renal function occurring during the first postnatal year.

In our opinion, additional covariates worth testing in treosulfan pharmacokinetic modeling are blood pH, as a readily measurable marker of acid-base balance, body temperature, and the volume of fluids infused intravenously to patients.

According to Eq. 2, the physiological fluctuations of blood pH from 7.35 to 7.45 are associated with a 24% increase in the k_f (0.38– 0.47 h⁻¹ at 37°C). Greater changes of the k_f will obviously accompany acidosis or alkalosis [11]. A particularly high risk of developing metabolic acidosis is expected in infants withheld breastfeeding and thus deprived of the natural source of bases [43]. Moreover, in our previous work, we estimated that the body temperature change of only 1°C (from 36.5 to 37.5 °C) implicates an increase of k_f by 17% (0.42–0.49 h⁻¹ at pH 7.4). Therefore, blood pH and body temperature might be valuable determinants of the Cl_f and Cl_tot of treosulfan [15].

Following the widely accepted guidelines, children who undergo conditioning prior to HSCT in our department receive 3000 ml of fluid per m² of BSA daily, mainly 0.9% NaCl and 5% glucose supplemented with physiologically relevant amounts of potassium, magnesium, and calcium. This is associated with a continuous infusion halted only during the treosulfan infusion. Meanwhile, it has been found that during an intravenous infusion and several hours thereafter, the blood and extracellular fluid expand [45, 46]. Hahn et al. [46] reported that at the end of 15- to 80-min infusions of Ringer’s solution 25 mL/kg BW administered to healthy adults, the volume of blood and extracellular fluid was increased by 20–25% and ~ 10%, respectively. When 12.5 mL/kg of solution was infused over 30 min, the blood expansion was 8% [46]. It is therefore possible that the V_ss of treosulfan in HSCT patients is artificially increased to an extent that depends on the load and infusion rates of the fluids. Whether this effect is clinically significant remains an open question. In our opinion, the fluid infusions should affect the Cl_R of treosulfan. The general knowledge is that the large intravenous delivery of fluids provokes diuresis and thus diminishes the tubular reabsorption of drugs [45, 47]. As approximately one-half of treosulfan undergoes the reabsorption process, high-volume infusions should elevate the drug Cl_R [31]. Concluding, the volume of fluids administered to HSCT patients on a daily basis might contribute to the interindividual variability of the V_ss and/or Cl_tot of treosulfan, and also their differences between various clinical centers.

The results of the quantitative analysis of biologically active epoxy derivatives of treosulfan in patients’ plasma were published for the first time in 2012 [48]. Following 2 h infusion of the prodrug at a dose of 12 and 14 g/m² to two children, the S,S-EBDM concentrations were approximately 100-fold lower than those of the parent compound, while the levels of S,S-DEB did not exceed the detection limit (< 0.8 μM) [48]. In the later study enrolling 16 pediatric patients, the t_½ of S,S-EBDM was found to not differ statistically from the parent drug [16]. The pharmacokinetics of S,S-EBDM and S,S-DEB remained unresolved until the preformed epoxides were administered to animals (rabbits) [17]. A very fast elimination of the compounds was then observed (t_½ 0.069 and 0.046 h, respectively), which was associated with the extremely high Cl_tot (167 and 233 mL/min/kg, respectively). When treosulfan was administered to rabbits, the t_½ of S,S-EBDM did not differ statistically from that of the prodrug (1.6 h); likewise in the previous study in humans. Thus, it was proved that after administration of treosulfan, the elimination of its epoxides is limited by their formation from the parent compound (formation rate-limited kinetics). Consequently, the t_½ of S,S-EBDM and S,S-DEB is the same as that of treosulfan, despite the levels of epoxides in the body being very low compared with the prodrug due to their high Cl_tot [17]. Additional confirmation for this phenomenon was provided by an observation that the organ elimination of S,S-EBDM in rats proceeded at a similar rate as that of treosulfan (lungs, muscle, and bone marrow), except the brain, from which the epoxide was eliminated faster [19]. A clinical importance of the above facts is that once the elimination of treosulfan is completed, S,S-EBDM and S,S-DEB are also eliminated from the patient’s body. This is crucial for the graft protection in HSCT patients [17, 19].

The available data originating from in vitro and in vivo animal studies on the metabolites of 1,3-butadiene, including 1,2:3,4-diepoxybutane, suggest that the main elimination routes of the epoxy transformers of treosulfan are liver and lung metabolism via epoxide hydrolase and glutathione S-transferase, and nonspecific reactions with tissue components [49‐51]. This is further supported by the immeasurable liver levels of S,S-EBDM in rats administered treosulfan, and the rapid elimination of the epoxy transformer from the lungs compared with the other organs [19]. To date, the V_ss and Cl_tot of S,S-EBDM and S,S-DEB in humans have not been reported. A commonly known pharmacokinetic relationship, \({\text{AUC}}_{\text{metabolite}} /{\text{AUC}}_{\text{prodrug}} = {\text{Cl}}_{\text{f}} /{\text{Cl}}_{\text{metabolite}} ,\) and the AUC data of treosulfan and S,S-EBDM obtained in 16 pediatric patients prepared for HSCT, allow to indirectly estimate the Cl_tot of S,S-EBDM to be several hundred mL/min/kg [16, 52]. In these children, the AUC of S,S-EBDM correlated significantly with the AUC of the prodrug (p = 0.022 in the Spearman test), but the correlation coefficient was not high (0.57). This indicates that the S,S-EBDM to treosulfan AUC ratio can vary substantially between HSCT patients [16]. To investigate this issue, the development of a combined prodrug–transformer population pharmacokinetic model might be helpful. Apart from the covariates that were discussed for treosulfan V_ss and Cl_tot in Sect. 3, it will be worth investigating liver function as a possible factor affecting the Cl_tot of S,S-EBDM [49, 50].

The disposition of the prodrug and S,S-EBDM has been studied using a rat model to elucidate pharmacokinetic reasons underlying the organ toxicity of treosulfan. The liver/plasma, lungs/plasma, brain/plasma, and bone marrow/plasma AUC ratios obtained for treosulfan amounted to, on average, 0.96, 0.82, 0.10, and 0.82, respectively. These data demonstrated that the prodrug has similar capability to busulfan (AUC ratios 1.3, 1.0, 0.75, and 0.83, respectively) to distribute into the organs, except the brain. However, after scaling the tissue AUC results to the plasma drug concentrations of HSCT patients, which are two orders higher in the case of treosulfan, the exposure of all four organs to the prodrug turned out to be very high compared with busulfan (Table 3) [19]. This negated the hypothesis that the favorable organ toxicity of treosulfan might stem from the limited distribution of the prodrug into the respective tissues [3, 4, 53]. The levels of monoepoxide in the rat liver were unquantifiable. The average rat lungs/plasma, brain/plasma and bone marrow/plasma AUC ratios for biologically active S,S-EBDM were 0.50, 0.35, and 0.75, respectively. After scaling the organ AUC results of S,S-EBDM to the concentrations observed in the plasma of HSCT patients, the clinical exposure of the lungs and brain to the epoxide was lower than to busulfan. On the other hand, the estimates of the AUC of S,S-EBDM and busulfan in patients’ bone marrow were similar. These results provided a pharmacokinetic rationale for the clinical observations that treosulfan-based conditioning regimens demonstrate lower hepato-, pulmo-, and neurotoxicity than busulfan-based conditioning regimens, but comparable myeloablation strength [19].

The other study investigated the influence of very young age on the penetration of treosulfan and S,S-EBDM across the blood-brain barrier. Treosulfan was administered to 10- and 34- to 35-day-old rats as they corresponded to neonates/younger infants and older children, respectively, in terms of blood-brain barrier functioning. The mean brain-to-plasma treosulfan AUC ratio in younger and older animals was 0.16 and 0.08, respectively, and the tissue/plasma AUC ratio obtained for S,S-EBDM was found to be 0.5 and 0.2, respectively. These results led to the conclusion that very young patients receiving high-dose treosulfan prior to HSCT may experience higher neurotoxicity than older patients due to the increased penetration of the prodrug and S,S-EBDM across the incompletely mature blood-brain barrier. This corresponded to the results obtained in clinical studies in which seizures only occurred in patients under 4 months of age [8, 18].

Relatively high variability of treosulfan pharmacokinetics in pediatric patients may raise the need for implementing therapeutic drug monitoring and individual dose adjustment in this group. Since pharmacokinetic studies of treosulfan began, it has been assumed that plasma (serum) concentrations of the prodrug are a good representation of the alkylating activity of its epoxy transformers [16, 20, 23‐25, 31]. However, for years, a correlation between treosulfan concentrations in plasma and levels of specific DNA adducts in tissues, for example the bone marrow, or clinical effects, have not been investigated. Van der Stoep et al. [24] and Mohanan et al. [25] recently published the first results of a relationship between the exposure of treosulfan and early toxicity, as well as clinical outcome, in children undergoing conditioning prior to HSCT. In the former study, patients with an AUC > 1650 mg·h/L demonstrated a statistically higher incidence of mucosal and skin toxicity than those with an AUC < 1350 mg·h/L (odds ratio 4.4 and 4.5, respectively). The odds of developing hepato- and neurotoxicity were also higher in the former group, but the difference did not reach statistical significance. No association was found between treosulfan exposure and early clinical outcomes, i.e. engraftment, donor chimerism, acute graft-versus-host disease, treatment-related mortality, and overall survival. The authors claimed that this might stem from the heterogeneity of patients’ primary diseases, which were both malignant and nonmalignant [24]. However, in the later study [25] conducted exclusively in thalassemia major patients, no statistically significant relationship was found between the drug AUC and graft rejection, regimen-related toxicity, and treatment-related mortality. Compared with results reported by van der Stoep et al. [24], the data presented by Mohanan et al. [25] are somewhat conflicting with regard to survival outcomes only. In the thalassemia major children, Cox regression analysis demonstrated a statistically higher risk of poor overall and event-free survival at low Cl_tot of treosulfan (< 8 L/h/m²). Moreover, the Kaplan–Meier curve showed a trend towards better overall survival at high Cl_tot of treosulfan (>8 L/h/m²) and low AUC (< 1828 mg·h/L), although without statistical significance [25]. Two hypotheses are worth considering and testing to explain the above results. The first is that a relatively low borderline AUC of treosulfan provides enough myeloablation for stable engraftment and acute graft-versus-host disease control, and also for overcoming a primary nonhematological disease, whereas the greater exposure intensifies both the antimalignancy effect and extramedullary toxicities. A second explanation might be the interindividual variability of the ratio of the systemic AUCs of S,S-EBDM and/or S,S-DEB to the prodrug (Sect. 4). Verification of this hypothesis warrants monitoring of the plasma concentrations of the treosulfan epoxy transformers, which are likely to better correlate with clinical results than those of the parent drug. At present, it seems that a relationship between treosulfan exposure and early regimen-related toxicity and clinical outcome is still unresolved. Investigations should be continued in larger disease-specific cohorts of HSCT patients. Ongoing studies will reveal how treosulfan exposure relates to long-term clinical outcomes and late toxicities, in particular gonadal function [23, 24].

As far as the future introduction of therapeutic monitoring of treosulfan is considered, it is worth mentioning that 2- or 3-point limited sampling strategies have been developed for determination of the drug AUC in HSCT children. They provided an accurate estimation of the exposure in all individual patients tested (error rate < 20%), but the validation was performed in small groups (5–8 individuals). Therefore, the robustness of the above limited sampling strategies must be confirmed in larger cohorts before their use in routine practice [20, 23].

A challenge in therapeutic monitoring of treosulfan within conditioning prior to HSCT is a very brief course of treatment, consisting of three doses administered on 3 consecutive days. This allows personalization of only the second and third dose of the prodrug unless a test dose is applied prior to starting the actual regimen. The benefit of these two possible strategies will be worth examining in reference to fixed-dose conditioning in prospective studies if therapeutic monitoring comes into practice. To avoid artificial intercenter differences, the standardized protocols for treosulfan monitoring should be established, including blood sampling time, handling of the collected blood prior to bioanalysis, and mode of the AUC estimation (classic or population models). Moreover, the uniform mode of reporting volume of distribution and clearance should be implemented in terms of normalization to either BW or BSA. This is particularly essential in infants and children, who inherently demonstrate considerable anthropometric differences. Otherwise, the interpatient and intercenter comparisons of the pharmacokinetic parameters are problematic, or even senseless.

This holistic review of the currently available literature indicates that three processes contribute to the Cl_tot of treosulfan: glomerular filtration, tubular reabsorption, and nonenzymatic epoxy transformation of the prodrug. Therefore, blood pH, body temperature, and intravenous fluid delivery should not be neglected as covariates of the Cl_tot of treosulfan in dose optimization efforts in HSCT patients, particularly infants. Organ disposition of treosulfan and S,S-EBDM in rats provides support for a lack of graft exposure to the compounds after at least a 2-day washout period preceding HSCT, the low organ toxicity of treosulfan-based conditioning compared with busulfan-based treatment, and the higher odds of neurological adverse effects in infants compared with older children. In terms of future therapeutic drug monitoring, larger studies are needed to verify the association of early and long-term toxicity and clinical outcomes with systemic exposure of not only treosulfan but also its active epoxy-transformers, at least S,S-EBDM.