The Clinical and Cellular Pharmacology Of Vincristine, Corticosteroids, l-Asparaginase, Anthracyclines and Cyclophosphamide In Relation To Childhood Acute Lymphoblastic Leukaemia

Authors


E. J. Estlin, Consultant Senior Lecturer in Paediatric Oncology, Department of Paediatric Oncology, Royal Hospital for Sick Children, St Michael's Hill, Bristol BS2 8BJ, UK. E-mail: e.j.estlin@bristol.ac.uk

Recently, slow early response to remission induction therapy has been recognized as a poor prognostic factor for childhood acute lymphoblastic leukaemia (ALL) (Reiter et al, 1994; Nachman et al, 1997). Indeed, augmentation of post-remission induction therapy has been shown to improve the outcome for such cases (Nachman et al, 1998). The potential importance of primary drug resistance, as measured by the methylthiazol tetrazoliumbromide (MTT) assay, has been investigated as a determinant of potential response to remission induction therapy and overall prognosis (Pieters et al, 1991; Kaspers et al, 1997). Moreover, an association between the delivered drug dose of vincristine, l-asparaginase and anthracycline has been described for children with higher risk ALL (Gaynon et al, 1991). However, there have been no studies of the importance of the clinical and cellular pharmacology of the antileukaemic agents that form the mainstay of remission induction and intensification/consolidation therapy for childhood ALL.

The purpose of this review will be to discuss the possible optimization of vincristine, corticosteroids, l-asparaginase, anthracyclines and cyclophosphamide in the therapy of childhood ALL. For this, the discussion will focus on the known mechanisms of cytotoxicity of these agents in vitro, their activity as single agents and studies of the clinical and cellular pharmacology of these agents in the therapy of childhood ALL.

Vincristine

The vinca alkaloid vincristine induces cytotoxicity by interacting with and disrupting microtubules, especially those comprising the mitotic spindle apparatus (Rowinsky & Donehower, 1996). Vincristine may bind at low-density high-affinity sites at the ends of the microtubule, with the effect of inhibiting microtubule assembly, and at low-affinity high-density sites along the wall of the microtubule, causing disruption of microtubule architecture. For human leukaemia cell lines, vincristine causes apoptotic cell death in a manner which relates to the concentration and time of exposure to the drug (da Silva et al, 1996). For CCRF-CEM and Jurkat cells, maximum cytotoxicity was found when cells were exposed to 3 µm and 0·1 µm of vincristine for 72 h respectively. However, the cytotoxicity of vincristine is subject to the inoculum effect, in which an increase in cell density causes a reduction in the intracellular levels and therefore cytotoxicity of a given dose of the drug (Kobayashi et al, 1992). For Molt-3 cells, maximum cell kill is associated with 25% saturation of the cellular vincristine binding sites, and the lesser effect of the drug at high cell density could be explained by the lack of drug molecules to saturate cellular binding sites sufficiently (Kobayashi et al, 1998). In most experimental models, resistance to vincristine is associated with decreased drug accumulation and retention, a phenomenon which is usually associated with the cellular expression of P-glycoprotein and the MDR phenotype (Rowinsky & Donehower, 1996).

Single-agent studies of vincristine in the therapy of childhood ALL were performed in the 1960s, mainly in children with disease which had relapsed after therapy with methotrexate and 6-mercaptopurine. With doses of 0·06 mg/kg/week (Heyn et al, 1966) or 2 mg/m2/week (Karon et al, 1966), complete remission (CR) rates of 50–60% were obtained, although the study of Karon et al (1966) suggested that vincristine had no advantage over a placebo in remission maintenance. However, monthly pulses of vincristine, in association with prednisolone, have been shown to improve the disease-free survival of children with ALL (Bleyer et al, 1991; Childhood ALL Collaborative Group, 1996).

In modern protocols for the therapy of childhood ALL, vincristine forms one of the mainstays of remission induction therapy. In most protocols, vincristine is administered during remission induction as a weekly intravenous bolus at a dose of 1·5 mg/m2 (Chessells et al, 1995; Evans et al, 1998) or 2·0 mg/m2 (Veerman et al, 1996), and the maximum dose is conventionally capped at 2–2·5 mg in view of the risk of toxicity (Rowinsky & Donehower, 1996). Vincristine also plays a role in the intensification/consolidation therapy (Chessells et al, 1995; Nachman et al, 1998) and maintenance therapy (Chessells et al, 1995; Veerman et al, 1996) of ALL in many protocols.

The clinical pharmacology of vincristine has been investigated for children with ALL. After bolus intravenous administration, the pharmacokinetics of vincristine are characterized by large inter- and intrapatient variations in parameters such as clearance, volume of distribution and elimination half-life (Gidding et al, 1999). Peak plasma concentrations of 100–400 nm are only briefly achieved after an intravenous bolus of vincristine; the t1/2α (initial or elimination half-life) of approximately 8 min reflecting the rapid cellular uptake and extensive tissue binding of the drug. Indeed, in the majority of children, plasma levels fall to below 5 nm by 60 min (de Graaf et al, 1995), and the long t1/2β (half-life of the terminal phase of elimination) of approximately 14 h means that plasma levels of 1–2 nm are maintained for relatively long periods of time. Although the clearance values for children are generally greater than those for infants and adults (Gidding et al, 1999), it is not certain whether vincristine clearance decreases with age during childhood (Crom et al, 1994; Gidding et al, 1999). In addition, there is no clear relationship between vincristine neurotoxicity and systemic exposure (Crom et al, 1994). When administered as a continuous infusion of 0·5 mg/m2/24 h for 5 d after a 1·5-mg/m2 intravenous loading dose, mean steady-state plasma vincristine concentrations of 1·7 nm are maintained with acceptable toxicity (Pinkerton et al, 1988). Although the cerebrospinal fluid (CSF) penetration of vincristine has not been determined in children, adult studies have shown vincristine levels in the CSF to be 20- to 30-fold lower than concurrent plasma concentrations (Rowinsky & Donehower, 1996).

Vincristine is rapidly taken up by cells, and the ratio of intracellular to extracellular concentrations ranges from 150- to 500-fold for various human haemopoietic cell lines (Rowinsky & Donehower, 1996). For mice bearing human rhabdomyosarcoma xenografts, sensitivity to vincristine was associated with prolonged retention of the drug after a single intraperitoneal injection (Horton et al, 1988). The only studies of vincristine in relation to leukaemic blasts from children with ALL have related to in vitro drug sensitivity testing by means of the MTT assay (Klumper et al, 1995), in which a mean IC50of 0·71 µm (range < 0·05 to > 50) was demonstrated.

In summary, although vincristine has an established role in the therapy of ALL, little is known of the relationship between the clinical and cellular pharmacology of the drug in vivo for childhood ALL. In vitro studies have highlighted the importance of the degree of saturation of vincristine binding sites and time of exposure for vincristine cytotoxicity in human leukaemia cell lines. Therefore, studies are needed to determine the relationship between systemic exposure to vincristine, as determined by pharmacokinetic measurements, and intracellular levels, binding site saturation and retention in leukaemic blasts. In this respect, a loading dose of vincristine followed by a prolonged infusion (Pinkerton et al, 1988) may optimize therapy with this agent, especially in the initial stages of remission induction therapy when a high leukaemic burden could potentially generate a positive inoculum effect (Kobayashi et al, 1998).

Glucocorticoids

The glucocorticosteroids prednisone and dexamethasone bind intracellularly to the glucocorticoid receptor (GR) and are able to cause cell death by apoptosis (Thompson, 1994). The molecular mechanisms by which glucocorticoids induce apoptosis are still poorly understood, but include inhibition of interleukin 2 (IL-2) production (Arya et al, 1984), downregulation of expression of the c-myc oncogene (Thulasi et al, 1993) and repression of transcription factors such as AP-1 (Jonat et al, 1990). In addition, interaction with other transduction pathways involving Janus kinase and signal transducers and activators of transcription may play a role in cytotoxicity (Stocklin et al, 1996).

For human leukaemia cell lines, glucocorticoid (GC) receptor saturation occurs at dexamethasone concentrations of approximately 30 nm, and dexamethasone demonstrates more potent cytotoxicity than prednisolone (Norman & Thompson, 1977). However, prolonged continuous exposures are necessary for significant cell kill. For example, for CCRF-CEM and Jurkat cells, 72-h continuous exposure to concentrations of up 50 µm prednisolone had virtually no effect on cell viability for CCRF-CEM cells and only a minimal cytotoxic effect with Jurkat cells (da Silva et al, 1996). Furthermore, for CCRF-CEM cells exposed to 5 µm dexamethasone, onset of apoptosis was seen only after 36 h, with only 50% apoptosis occurring by 52 h (Wood et al, 1995). Lineage-specific differences for in vitro steroid sensitivity have been demonstrated utilizing stromal support assays, with which B-lineage cell lines are more sensitive than T-lineage or erythroleukaemic cell lines to cytotoxic effects of glucocorticoids (Ito et al, 1996). This increased resistance may be related to decreased number of GR in T lymphoblasts compared with B-lineage blasts (Quddus et al, 1985). Acquired resistance to glucocorticoids in CCRF-CEM-derived cell lines (with mutation involving the GR gene) has been shown to relate to both an absence and two- to threefold reduction in GR levels, respectively, when compared with parental cells (Geley et al, 1996).

Single-agent trials (predominantly with prednisone) in the 1950s and 1960s showed that 70–80% of children with ALL achieved a complete or good partial remission with 2–2·2 mg/kg/d of prednisone administered in divided doses (Hyman & Sturgeon, 1956; Hyman et al, 1959; Wolff et al, 1967). However, remissions were brief, lasting 1–10 weeks. As glucocorticoids became an established part of remission induction therapy, studies were designed to explore the effect of the dose and schedule of administration. High-dose therapy with 1 g/d of prednisolone did not improve the remission induction rate and was associated with serious complications (Ranney & Gellhorn, 1957). Leikin et al (1968) randomly assigned 223 children with newly diagnosed acute leukaemia to receive equal cumulative doses of prednisolone (4 mg/kg/d) in three different schedules. For children receiving 1·3 mg/kg three times daily, 82% achieved a complete remission, whereas remission induction rate was 28% and 19% for those who received 8 mg/kg on alternate days and 16 mg/kg every 4 d respectively. Therefore, single-agent studies of prednisone may highlight the potential importance of schedule dependency of steroids in treatment of ALL, suggesting that persistence of a therapeutic rather than peak plasma concentration is important for efficacy.

Glucocorticoids are now used in the remission induction, intensification and maintenance phases of therapy in many treatment protocols. However, there is considerable variation between the use of prednisone and dexamethasone world-wide.

For remission induction, children in the UK are currently randomized to receive either 6·5 mg/m2/d dexamethasone or 40 mg/m2/d prednisolone. For recent Berlin–Frankfurt–Munster (BFM) studies, a higher prednisone dose of 60 mg/m2/d has been used (Reiter et al, 1994). However, for a recent Dutch Leukaemia Study Group trial, 6 mg/m2/d dexamethasone has been used for remission induction (Veerman et al, 1996).

For intensification therapy, higher doses of dexamethasone (10 mg/m2/d) or prednisolone (100 mg/m2/d) have been used (Gaynon et al, 1993; Reiter et al, 1994) for varying lengths of time. Many protocols also include prednisone (Chessells et al, 1995; Nachman et al, 1998) or dexamethasone (Veerman et al, 1996) as part of maintenance therapy. However, the use of a corticosteroid during the maintenance phase of therapy is not universal (Evans et al, 1998). Intrathecal therapy with prednisolone (Veerman et al, 1996) and hydrocortisone (Evans et al, 1998) also forms part of central nervous system (CNS)-directed therapy in several protocols.

Although the pharmacokinetics of glucocorticoids has been extensively studied in a number of patient populations with non-malignant disease, there have been very few studies that have involved children with ALL. For children with non-malignant disease receiving 0·3–2·5 mg/kg prednisone, peak levels were attained within 2 h in most patients, with a range of 0·01–2·4 µm, and the mean t1/2 was approximately 2 h (Green et al, 1978). Moreover, clearance of prednisolone is age dependent, with an inverse correlation between clearance and age as demonstrated by Hill et al (1990). For children with ALL, prednisone pharmacokinetics has been studied for only six children (Choonara et al, 1989). Patients received oral prednisolone at a dose of 13–16·9 mg/m2, and the peak plasma concentrations achieved ranged from 0·38 to 1·5 µm (mean 0·9 µm) with the free fraction of prednisolone ranging from 0·28 to 0·55 µm. Therefore, peak plasma concentrations were achieved which would be cytotoxic to leukaemic blasts in vitro, and there was a large interpatient variation in pharmacokinetic parameters such as area under the curve (AUC) and total clearance. The CSF pharmacokinetics of prednisolone or hydrocortisone has not been studied for children with ALL. However, the CSF pharmacokinetics of prednisone has been studied in adults, and equilibrium between plasma and CSF is achieved after 6 h and mean CSF peak level are approximately one-third of peak plasma levels (Bannwarth et al, 1997).

The pharmacokinetics of dexamethasone has not yet been studied for children with ALL. However, for children receiving dexamethasone at a dose of 0·1 or 0·3 mg/kg given intravenously for croup or head injury, large interpatient pharmacokinetic variability was observed. The average plasma dexamethasone concentration was 0·36 µm (range 0·13–0·67 µm) and the t1/2 ranged from 2·3 to 9·5 h (Richter et al, 1983). Overall, dexamethasone has a lower clearance, a larger volume of distribution and thus long t1/2 when compared with prednisone. In primate models, dexamethasone has been shown to have better CNS penetration than prednisolone with a longer CSF t1/2 of 246 vs. 175 min (Balis et al, 1987).

Of the cellular determinants of sensitivity to glucocorticoids, GR levels have been the most extensively studied in relation to childhood ALL. Whereas mature lymphocytes contain 2500–5400 GRs/cell, GR levels in lymphoblasts range between less than 1000 to greater than 20 000/cell. Lineage-specific differences in GR levels have been described for children with ALL, with early pre-B and pre-B ALL having twofold higher median GR receptors levels than children with T-cell ALL and B-cell ALL (Quddus et al, 1985). Moreover, low level of GRs in ALL tend to correlate with ‘high-risk’ features such as high presenting white cell count, age < 2 or > 10 years, CNS disease and black race. Although a high GR content of leukaemic blasts has been found to be a favourable prognostic factor for childhood ALL, such experiences have not been uniformly consistent (Kato et al, 1993; Pui et al, 1983).

The prognostic significance of in vitro drug resistance, as assessed by the MTT assay, has been studied extensively for childhood ALL. With this system, the antileukaemic activity of dexamethasone is a median 16-fold higher than that of prednisolone (Kaspers et al, 1996), and steroid resistance has been related to response to initial steroid monotherapy (Kaspers et al, 1998), event-free survival (Kaspers et al, 1997) and relapse (Klumper et al, 1995). In addition, increased glucocorticoid resistance has been found for children with pro-B and T-cell ALL, and those under 18 months or over 10 years of age at diagnosis (Pieters et al, 1998). In comparison, the cytotoxicity of glucocorticoids vs. leukaemic blasts in an environment with stromal support shows dexamethasone to be only five to six times more potent (on a molar basis) than prednisolone. For this system, the median IC50/IC90 values for prednisolone and dexamethasone were 0·043 µm/0·26 µm and 0·007 µm/0·1 µm respectively (Ito et al, 1996). However, unlike the studies using the MTT assay, there was no significant correlation for steroid sensitivity and clinical characteristics such as age, sex, white cell count (WCC) and immunophenotype.

In summary, although recent studies indicate that dexamethasone may confer an advantage over prednisone in terms of CNS relapses and event-free survival (Veerman et al, 1996; Bostrom et al, 1998), and the current Medical Research Council (MRC) UKALL trial in which children are randomized to receive prednisolone or dexamethasone during their treatment may eventually support the superiority of one of these drugs, there is still a paucity of pharmacological data in this area. Studies are needed to determine fully the interpatient pharmacokinetic variability for prednisolone and dexamethasone in childhood ALL, and this information could be related to cellular measures such as GR levels and saturation in leukaemic blasts and to pharmacodynamic outcomes such as early response or minimal residual disease status. In addition, such studies may identify possible relationships between the clinical pharmacology of corticosteroids and toxicity such as avascular necrosis (Gaynon & Lustig, 1995). In this respect, the pyrazolosteroid cortivazol may play a role in the future therapy of glucocorticoid-resistant ALL, with less systemic toxicity than conventional corticosteroids (Gaynon & Carrel, 1999). Moreover, many of the early single-agent studies used prednisone at a dose of 2 mg/kg/d in divided doses for remission induction. If this is compared with the current conventional dosing [based on body surface area (SA)] of 40 mg/m2/d, all children above 10 kg weight (SA 0·5 m2) receive less than 2 mg/kg/d of prednisone. Indeed, there have been no single-agent studies to prove that less than 2 mg/kg/d of prednisone is as effective in remission induction.

L-asparaginase

l-Asparaginase catalyses the hydrolysis of l-asparagine to l-aspartic acid and ammonia, thereby depriving cells which do not have the capacity to synthesize l-asparagine of this amino acid. l-Asparagine depletion results in decreased protein synthesis and apoptosis, but the relationship between these events is not yet clear. For murine L5178Y (Ueno et al, 1997) and Molt-4 human T-lymphoblastoid cell lines exposed to asparaginase at 1·0 IU/ml for 15 h (Shimizu et al, 1992), cell cycle arrest in G1 phase occurs. In the case of murine lymphoblastoid L5178Y cells, DNA fragmentation is observed in G1 phase cells as early as 8 h after asparaginase exposure (Ueno et al, 1997). A wide range of in vitro sensitivity to l-asparaginase has been reported for human leukaemia cell lines. For the asparaginase-sensitive human T-lymphoblastoid cell lines Molt-3 and Molt-4 exposed to varying l-asparaginase concentrations for 72 h, cytotoxicity with an IC50 of < 10−4 U/ml is observed. By contrast, the asparaginase-insensitive human T-lymphoblastoid CCRF-CEM, JM and B-lymphoblastoid RPMI-8866 cell lines, are between 100- and 1000-fold less sensitive to the drug (Kiriyama et al, 1989).

Study of the determinants of sensitivity to asparaginase in murine lymphoma cells (6C3HED) has revealed that reduced sensitivity is mediated by an increased rate of asparagine synthesis in response to asparagine depletion (Broome & Schwartz, 1967). For the human leukaemia cell lines MOLT-4, NALL-1 and BALL-1, amino acid deprivation leads to a change in transcriptional regulation and increases in asparagine synthetase (AS) mRNA, protein and enzymic activity (Hutson et al, 1997). In addition, for U937 human histiocytic lymphoma cells with acquired resistance to asparaginase, a stable 80-fold increase in asparagine synthetase expression has been observed (Kiriyama et al, 1989). A further determinant of sensitivity to asparaginase is p53 status; studies in HL60/S and HL60/N3 promyelocytic leukaemia cell lines indicate that functional p53 protein is required for asparaginase to exert its effect in inducing cellular apoptosis (Fu et al, 1998; Nandy et al, 1998).

An early single-agent study assessed the effectiveness of Escherichia coli asparaginase for remission induction for children with relapsed as well as newly diagnosed ALL. Schedules of daily or thrice weekly administration were compared (10–5000 IU/kg given intravenously, intramuscularly or intrathecally). The overall remission rate was 62% with median duration of 60 d (range 15–248 d), which appeared to be dose independent. Moreover, maintenance therapy with asparaginase did not increase the duration of remission (Tallal et al, 1970). In a further study, Rausen & Glidewell (1970) reported complete remission rates of 66% for 15 children with relapsed ALL who received 5000 U/kg E. coli asparaginase i.v. weekly 3× compared with a 30% CR rate observed for 15 children who received 1000 U/kg i.v. daily 15×. A dose-dependent effect was also found for a further study of E. coli asparaginase (administered with 6-mercaptopurine to reduce sensitivity reactions) in children with relapsed ALL. In this study, complete remission rates varied from 9·5% with 300 IU/m2 to 35·1% with 3000 IU/m2, 53·5% with 6000 IU/m2 and 62·5% with 12000 IU/m2 of l-asparaginase administered three times per week (Ertel et al, 1979). Therefore, single-agent studies of E. coli asparaginase indicate both a possible dose and schedule dependency for effectiveness of this drug in the treatment of ALL.

L-Asparaginase derived from the bacteria E. coli or Erwinia chrysanthemi is currently used as a major component of remission induction and consolidation regimens for the treatment of acute lymphoblastic leukaemia world-wide. However, there is considerable variation in the dose, route and schedule of administration between various protocols. For remission induction therapy, the MRC UKALLXI trial specified 6000 IU/m2 subcutaneously three times weekly for nine doses, whereas the ALL-BFM 90 trial (Boos et al, 1996) specified eight doses of 10 000 IU/m2 intravenously administered at 3-d intervals. Moreover, for the Children's Cancer Group (CCG) 1882 trial, 6000 IU/m2 asparaginase was intramuscularly three times weekly for nine doses (Nachman et al, 1997). Similar variation in scheduling also exists for consolidation therapy in current protocols.

Differences in disposition of asparaginase which were dependent upon the source of the enzyme were recognized in early clinical studies (Schwartz et al, 1970). Pharmacokinetic studies in newly diagnosed children with ALL have shown that peak plasma concentrations after a single injection of E. coli asparaginase are related to the dose administered (Asselin et al, 1993). Concentrations of approximately 1·0 and 10 IU/ml are achieved after 24–48 h with doses of 2500 and 25 000 IU/m2 respectively. However, the half-life of both the asparaginase activity and protein (mean approximately 1·5 d) are dose independent. For Erwinia asparaginase (dose 25 000 IU/m2), peak serum activity levels are reached within 24 h and the half-life is significantly shorter (mean 0·65 d) than for E. coli asparaginase. In direct contrast, polyethylene glycol-modified (PEG) asparaginase (dose 2500 IU/m2) achieves peak levels at 72–96 h and the half-life is significantly longer (mean 5·7 d) than for the native E. coli asparaginase preparation (Asselin et al, 1993). Preparations of asparaginase from differing sources have shown differing median trough levels when measured after the same initial dose (10 000 U/m2) which may or may not (in the case of Erwinia asparaginase) be above the level required for cytotoxicity in vitro (Asparaginase Medac 528 U/l, Crasnitin 49 U/l and Erwinase < 20 U/l). These trough levels were associated with significantly differing levels of complete asparagine depletion in plasma before the next dose, i.e. > 90%, > 60% and 26% respectively (Boos et al, 1996; Werber et al, 1997). Under pharmacological control, it is possible to reduce the dose of E. coli Asparaginase Medac from 10 000 IU/m2 (intravenously every third day) to 2500 IU/m2 and still achieve complete depletion of both serum and CSF asparagine (Ahlke et al, 1997). However, Dibenedetto et al (1995) have observed that approximately 35% and 70% of children receiving intramuscular Erwinia asparaginase (at either 10 000 IU/m2 every 3 d or 25 000 IU/m2 weekly) had incomplete depletion of CSF asparagine levels after 3 and 5 d respectively. This may represent suboptimal therapy in terms of CNS protection.

The dose of asparaginase cytotoxic to human leukaemic blasts in vitro is as low as 0·0001 IU/ml of tissue culture medium (Asselin et al, 1989), and there was no dose dependency for cytotoxicity in the range 0·0001–0·1 IU/ml. At each dose, asparagine reduction by > 95% occurred in the culture medium by a maximum of 24 h after the addition of asparaginase. Moreover, a minimum of 4 d exposure to l-asparaginase seemed necessary to produce cytotoxicity consistently, and longer durations of exposure did not enhance cytotoxicity. In contrast, Pieters et al (1991), using the MTT assay, determined that the median IC50 concentration of asparaginase against lymphoblasts from children with ALL in vitro was 0·213 IU/ml after 4 d of exposure to the drug. However, differences between the techniques of Asselin et al (1989) and Pieters et al (1991) make conclusions about the cellular pharmacodynamics of asparaginase difficult and a measure of the equivalence of the techniques used is needed. Sensitivity to asparaginase in vitro has been shown to predict in vivo response and to be a risk group-independent prognostic factor for relapse after treatment for childhood ALL (Asselin et al, 1989, 1999). However, in direct contrast, no prognostic significance was attached to asparaginase (ASP) sensitivity in the treatment of ALL using the MTT assay (Pieters et al, 1991).

Formation of antibodies against l-asparaginase may result in either a decrease (reviewed by Müller & Boos, 1998) or an increase (Wahn et al, 1983) in the half-life of enzymatic activity of the drug and may or may not be associated with allergic reactions, which are more commonly encountered with intravenous administration of the drug (Müller & Boos, 1998). However, the preliminary results of one study suggest that that the development of anti-ASP antibodies was not associated with a worse event-free survival (Woo et al, 1999).

In summary, although it is now possible to reduce individual doses of asparaginase under pharmacological control and still achieve complete serum and CSF depletion of asparagine, there is as yet insufficient clinical pharmacological information with which to optimize therapy with the various l-asparaginase preparations for childhood ALL. However, the plasma asparaginase pharmacokinetic profile and maintenance of asparagine depletion associated with the use of E. coli-derived preparations may prove to confer greater antileukaemic efficacy. Indeed, preliminary results from the EORTC 58881 study of the randomized comparison between E. coli and Erwinia asparaginase in induction and subsequent intensification for newly diagnosed childhood ALL has shown significantly better 4-year event-free survival with E. coli asparaginase than with Erwinia, albeit in a complex schedule in which maintenance 6-mercaptopurine was administered intravenously (Otten et al, 1996). However, comparative studies are required to relate the pharmacokinetics of various asparaginase preparations with pharmacodynamic measures such as l-asparagine depletion in plasma, bone marrow and CSF. This information, along with the formation of anti-asparaginase antibodies, may identify whether pharmacological variation is an important determinant of early disease response, minimal residual disease status and patient outcome. Moreover, there have been no clinical studies to determine whether leukaemic blasts begin to increase asparagine synthetase expression during the early stages of remission induction therapy in response to asparagine depletion.

Anthracyclines

The anthracyclines are known to exert their cytotoxicity by a number of intracellular reactions, which include free radical formation, DNA intercalation, inhibition of topoisomerase II, disturbance of helicase function and effects on signal transduction (Doroshow, 1996). For the human leukaemia cell lines CCRF-CEM and Jurkat, the cytotoxicity of the anthracyclines doxorubicin and daunorubicin is concentration and exposure time dependent, with maximal cytotoxicity associated with prolonged (> 48 h) exposure to 10 µm of the drugs. However, the cytotoxic effect of doxorubicin is subject to a positive inoculum effect in the Molt-3 human leukaemia cell line, in which an increase in cell density from 106/ml to 108/ml has been found to result in a 95% reduction in intracellular uptake and a 40-fold decrease in cytotoxicity of the drug (Kobayashi et al, 1992). In vitro mechanisms of resistance to anthracyclines include enhanced efflux of intracellular drug by P-170 glycoprotein (MDR) and multidrug resistance-associated protein (MRP), reduced topoisomerase II activity and enhanced antioxidant defence (Doroshow, 1996).

Single-agent studies of daunorubicin were performed in the 1960s, which demonstrated antileukaemic efficacy in children with predominantly relapsed ALL. In these early studies, an overall response rate of 28% (7% complete remission) was found with a variety of treatment schedules. For example, short-lived responses were obtained for daunorubicin administered at 1 mg/kg/d 4–5× followed by weekly maintenance doses (Tan et al, 1967) and 25 mg/m2/d 3× followed by weekly maintenance doses (Holton et al, 1968). Although the patient numbers were small, the activity of daunorubicin was greater for children who had no prior therapy than for those with relapsed disease (Tan et al, 1967).

In modern protocols for the treatment of childhood ALL, anthracyclines (predominantly daunorubicin) are used in the remission induction and intensification/consolidation phases of therapy. However, the dose and schedule of administration of daunorubicin varies world-wide. For remission induction therapy, various protocols include daunorubicin administered as a short infusion of 45 mg/m2/d 2× (Chessells et al, 1995), 40 mg/m2 weekly 4× (Reiter et al, 1994) and 25 mg/m2 weekly 4× (Nachman et al, 1997). Similar daunorubicin schedules are used as part of intensification/consolidation therapy, although in some protocols doxorubicin is the drug of choice and is administered at 25 mg/m2/day weekly 3× (Nachman et al, 1998) or 30 mg/m2 weekly 4× (Reiter et al, 1994).

The pharmacokinetics of daunorubicin and doxorubicin has been determined for adults, but not for children. Pharmacokinetic parameters reflect the high level of tissue uptake and subsequent release of drug back into the circulation. The initial half-life (t1/2α) for daunorubicin and doxorubicin are 40 min and 10 min respectively. However, the terminal half-lives (t1/2β) of 20–50 h account for the majority of the systemic exposure to the drugs (Doroshow, 1996). For daunorubicin, mean peak plasma concentrations of 227 ng/ml are achieved following a bolus injection of 45 mg/m2, falling to a mean of 16 ng/ml when the same dose is administered as a 72-h infusion. Despite this, the AUCs achieved by these two schedules are similar (Speth et al, 1987a). Similarly for doxorubicin, although mean peak plasma levels of 1640 ng/ml following bolus administration of a 30-mg/m2 dose were 35-fold higher than those attained during a 72-h infusion of the drug, the mean AUCs were similar (Speth et al, 1987b). Anthracyclines penetrate poorly into the CNS (Doroshow, 1996). Daunorubicin and doxorubicin are metabolized to daunorubicinol and doxorubicinol, and these have low levels of cytotoxicity. In the case of idarubicin, substantial cytotoxicity is retained by the alcohol metabolite idarubicinol. The pharmacokinetics of idarubicin has been investigated in children with leukaemia (Reid et al, 1990). As with adult pharmacokinetic studies of anthracyclines, idarubicin demonstrates wide interindividual pharmacokinetic variability, undergoing rapid initial tissue uptake, with a large volume of distribution, and has a terminal half-life of approximately 20 h. In contrast with idarubicin, idarubicinol could be detected in the CSF of children receiving systemic therapy.

The cellular pharmacology of the anthracyclines is characterized by the ability of essentially all nucleated cells to accumulate these drugs, primarily because of DNA binding, rapid association with cell membranes and storage in several different intracellular compartments. As a consequence, the ratios between the intracellular and extracellular concentrations of daunorubicin and doxorubicin are routinely in the order of 30–1000 at the end of short-term exposures to these drugs (Doroshow, 1996). Study of the relationship between the clinical and cellular pharmacology of daunorubicin and doxorubicin has been mainly performed with respect to adult acute myeloid leukaemia (AML). For daunorubicin administered at a dose of 45 mg/m2 as a bolus or 72-h continuous infusion, similar intracellular AUCs are obtained for leukaemic blasts in vivo. In addition, although the peak plasma levels of daunorubicin administered by 72-h infusion were 14-fold lower than those found with bolus administration, the peak daunorubicin concentration in leukaemic blasts was only 1·6-fold lower. Moreover, peak intracellular daunorubicin levels and AUC values were 70- to 100-fold and 300- to 600-fold higher than the corresponding plasma values respectively (Speth et al, 1987a). Daunorubin was retained by leukaemic blasts with a terminal half-life of 13 h. A similar relationship between the schedule of doxorubicin administration and the cellular pharmacology of myeloid blasts in vivo (Speth et al, 1987b). Moreover, a possible indication of the importance of the inoculum effect has been suggested by the study of Kokenberg et al (1988), in which the concentration of daunorubicin in peripheral white blood cells (which had correlated positively with bone marrow nucleated cell levels at diagnosis) was negatively correlated with the numbers of peripheral blasts at diagnosis.

The relationship between the clinical and cellular pharmacology of anthracyclines and response has also been studied in relation to adult acute leukaemia. For daunorubicin, no relationship has been found between daunorubicin pharmacokinetic parameters and the likelihood of achieving a complete remission (Kokenberg et al, 1988; Galettis et al, 1994). However, the significance of cellular daunorubicin levels is uncertain. Whereas Galettis et al (1994) found that lower daunorubicin intracellular AUCs related to the likelihood of poor response in a study of adults with either ALL or AML, Kokenberg et al (1988) found no relationship between peak intracellular daunorubicin levels and response for adults with AML. In relation to childhood ALL, Klumper et al (1995) demonstrated that both daunorubicin and doxorubicin produced cytotoxicity in leukaemic blasts, with IC50 values of 0·117 µg/ml and 0·347 µg/ml respectively.

In summary, although anthracyclines have an established role in the remission induction and intensification therapy of childhood ALL, there is as yet insufficient clinical or cellular pharmacological information with which to optimize therapy further with daunorubicin and doxorubicin. In particular, the pharmacokinetics of doxorubicin and daunorubicin have not yet been determined in children, and such studies could be linked to cellular pharmacological parameters such as peak intracellular anthracycline levels and intracellular AUC. Indeed, given the acute toxicities of anthracycline-containing remission induction therapy (Shaw et al, 1995), and long-term cardiac toxicity associated with even moderate anthracycline exposure (Sorensen et al, 1997), studies are required to investigate the optimal schedule of administration of anthracyclines in childhood ALL. Therefore, studies are needed to determine the relationship between systemic exposure to anthracyclines, as determined by pharmacokinetic measurements, and intracellular levels, binding site saturation and retention in leukaemic blasts. In this respect, prolonged infusion of anthracyclines may optimize therapy with these agents, especially in the initial stages of remission induction therapy in which a high leukaemic burden could potentially generate a positive inoculum effect (Kobayashi et al, 1998). Indeed, the findings of Speth et al (1987a,b), whereby acute toxicities for the remission induction therapy of AML are reduced by prolonging the time of infusion of anthracyclines and without a reduction in cellular anthracycline AUC or clinical response, could be important in this regard.

Cyclophosphamide

Alkylating agents compose a diverse group of electrophiles which reacts with DNA to form various adducts, of which interstrand DNA cross-links are particularly significant. Under favourable circumstances, the presence of these adducts either initiates apoptosis or cell cycle arrest (O'Connor et al, 1991). There have been few comparative studies of the relative benefits of individual agents, and the oxazaphosphorine cyclophosphamide (CP) remains the most commonly used for childhood ALL.

Experiments performed during the early development of CP revealed no evidence of cell cycle-dependent cytotoxicity, and as a result the drug is usually administered as intermittent single-dose therapy (Bruce et al, 1966; Sensenbrenner et al, 1972). More recent studies have described both an improved therapeutic index of CP in animals and increased cytotoxicity in cell culture experiments using multiple dosing schedules (Voelcker et al, 1984; Teicher et al, 1989). Several mechanisms of resistance to alkylating agents in paediatric malignancies have been proposed, including increased levels of intracellular thiols, upregulated glutathione-S-transferase activity and modulation of apoptosis with bcl-2 (reviewed in Yule et al, 1997).

Single-agent studies were reported in the early 1960s. For dosing schedules that included CP at a dose of 150–200 mg/kg/d intravenously for up to 14 d (Sweeney et al, 1962) and 2–7·5 mg/kg/d orally as a maintenance dose after intravenous priming (Fernbach et al, 1962), response rates of 30–40% (10–20% complete remission) were found for children with relapsed ALL.

In many recent protocols for the therapy of childhood ALL, isolated doses of CP (600–1000 mg/m2) have been included in the intensification phase of treatment. Opinion is divided in the USA, where the Paediatric Oncology Group only use CP in the treatment of ‘high-risk’ disease, including T-cell ALL. In contrast, the Children's Cancer Group include CP in the treatment of all forms of the disease, both during induction and repeated consolidation cycles (Amylon et al, 1999; Reaman et al, 1999). Ifosfamide, a structural isomer of CP, has not been widely adopted into ALL therapy, although the drug is included in the late intensification block of a recent BFM study (Reiter et al, 1994). However, the comparative efficacy of these oxazaphosphorines has not been established, and there is no evidence that the substitution of CP by ifosfamide in the treatment of ALL would be of therapeutic benefit (Shaw & Eden, 1990)

CP is a prodrug which requires metabolic transformation, including an initial reaction mediated by hepatic cytochrome P450 to generate phosphoramide mustard, the active alkylating species (Yule et al, 1995). Previous paediatric studies have demonstrated a high degree of interpatient variation in CP metabolism, much of which is likely to reflect differences in the expression of the individual cytochrome P450 enzymes involved (Yule et al, 1996). Although the pharmacokinetics and metabolism of CP have not been specifically investigated in children with ALL, reports which included leukaemic patients found no evidence of differentially altered metabolism in this group. The largest study performed to date reported a median half-life of 3 h, clearance of 3 l/h/m2 and a low volume of distribution consistent with minimal protein binding (Yule et al, 1996). More recent studies have raised the possibility of saturated metabolism with a proportional reduction in the generation of active alkylating species at higher dose levels. This phenomenon could only be demonstrated at doses far in excess of those used in current ALL protocols (Yule et al, 1996; Busse et al, 1997). The relationship between the pharmacology of CP and treatment outcome is incompletely understood at present. This is partially due to difficulties in the measurement of reactive alkylating metabolites and an incomplete understanding of which species are important in mediating cytotoxicity. To date, only a single adult study reported an inverse correlation between the area under the CP concentration-versus-time curve and both treatment-related cardiotoxicity and event-free survival (Ayash et al, 1992).

In summary, there is as yet a paucity of clinical and cellular pharmacological information with which to guide the optimization of cyclophosphamide in the therapy of childhood ALL. However, the introduction of fractionated CP, i.e. 3 g/m2 delivered in six 12-hourly doses, may be advantageous in avoiding enzyme saturation and allowing autoinduction of metabolism (Motzer et al, 1993). This may be supported to some extent by clinical studies using prolonged infusions of CP in adults with ALL in first relapse with both authors claiming higher remission rates than would have been expected on the basis of historical data (Solidozo et al, 1981; Ciolli et al, 1991). In addition, the introduction of a multiagent regimen including fractionated CP has significantly improved outcome in childhood B-cell lymphoblastic leukaemia/lymphoma (Murphy et al, 1986). Therefore, there is a continuing need for direct comparisons between different administration schedules of CP in ALL, and these studies should be supported by pharmacokinetic analyses. Indeed, as CP is associated with infertility and the development of second malignancies, especially myelodysplasia and acute non-lymphoblastic leukaemia (Tucker et al, 1987; Pryzant et al, 1993), optimal delivery of the drug may be important in terms of event-free survival and late effects.

Conclusion

World-wide, multiagent treatment protocols vary with respect to the dose, schedule and even use of agents such as anthracyclines and alkylating agents (Niemeyer et al, 1991). Moreover, although the clinical pharmacology of vincristine and l-asparaginase have been the subject of recent investigation in relation to childhood ALL, there is at present a paucity of information relating to anthracyclines, cyclophosphamide and corticosteroids. Therefore, studies in relation to childhood acute leukaemia are needed to identify whether there are any important interindividual differences in the clinical pharmacology or lineage-specific differences in the cellular metabolism of these agents which may be related to measures of cytoreduction and patient outcome. In this regard, single-agent ‘window’ studies, performed at presentation or first relapse, may be an essential tool for defining these relationships.

Acknowledgment

The authors would like to thank Dr S. P. Lowis for his help and advice in the preparation of this manuscript.

Ancillary