`
`http://annonc.oxfordjournals.org/
`
` at University of Wisconsin - Madison on February 2, 2016
`
`Special article
`
`Annals of Oncology 3: 339-34 7. 1992.
`Q 1992 Kluu't'!r Acad~mic Publish~rs. Prim~d in th~ N~thulands.
`
`The impact of pharmacokinetically guided dose escalation strategies in phase I
`clinical trials: Critical evaluation and recommendations for future studies
`
`M. A. Graham & P. Workman
`CRC D~panment M~dica/ Oncology, University of Glasgow, CRC B~atson Laboratories, &arsd~n. Glasgow, UK and EORTC
`Phamwco/ogy and Mol~cular Mechanisms Group
`
`Summary. Phase I studies requiring multiple dose escalation
`steps have led to the development of pharmacokinetically
`guided dose escalation (PGDE) strategies to expedite the
`conduct of early clinical trials. This article critically reviews
`PGDE strategies for a number of new anticancer agents in(cid:173)
`cluding amphethinile, brequinar sodium, iodo-doxorubicin,
`the anthrapyrazoles (DuP 941, DuP 942 and DuP 937),
`
`rhizoxin, and aphidicolin glycinate. The benefits and prob(cid:173)
`lems associated with PGDE are examined. Recommenda(cid:173)
`tions are made for the optimal deployment of pharmacologi(cid:173)
`cal information in future phase I studies.
`
`Key words: pharmacokinetic strategies, phase I trials, dose
`escalation
`
`Introduction
`
`Over the last ten years, pharmacokinetic studies have
`been gaining increasing importance in the process of
`new anticancer drug development. Why is this? Un(cid:173)
`doubtedly the reason is that pharmacokinetic studies
`are widely perceived by both experimental and clinical
`oncologists as having a valuable role to play. In earlier
`days there was a tendency to regard pharmacokinetics
`as a dry science practised by aficionados who played
`their trade with HPLC equipment and computer mod(cid:173)
`els, presenting complex equations and incomprehen(cid:173)
`sible tables of data, with scarce concern for the rel(cid:173)
`evance to drug design and development. No doubt
`some felt it to be a necessary, but tedious or even
`eccentric occupation, pursued for its own sake or to
`satisfy the regulatory agencies. So what has changed?
`We believe that two significant changes in attitude
`have taken place. The first is that those involved direct(cid:173)
`ly in pharmacokinetic studies have become much more
`proficient at integrating their activities into the drug
`development process - from the laboratory to the bed(cid:173)
`side. At the same time, clinicians carrying out early
`clinical studies have become convinced of the value of
`pharmacokinetic studies - indeed many of them have
`themselves been trained in the science of pharmaco(cid:173)
`kinetics.
`There are as yet very few examples where pharma(cid:173)
`cokinetic measurements are essential to the clinical
`practice of oncology. The gold standard is of course
`methotrexate, where monitoring of drug levels can
`avoid potentially fatal toxicity. There is however a
`rapidly developing interest in therapeutic drug moni(cid:173)
`toring - the individualization of drug dosage based on
`measured drug concentrations. Examples include eto-
`
`poside and 5-fluorouracil for example [IJ. The use of
`limited sampling strategies and Bayesian statistics will
`greatly help this approach. The successful development
`of simple dose individualisation formulae for carbo(cid:173)
`platin, based on renal clearance, also owes its success
`to a pharmacokinetic approach [2, 3J.
`Without a doubt the greatest impact of pharmaco(cid:173)
`kinetics occurs further back in the drug development
`pipeline. Firstly, pharmacokinetic studies in animals are
`likely to be carried out at an earlier stage than hitherto
`- perhaps even before a lead compound emerges. Thus
`this information can feedback into the drug design
`process. Secondly, pharmacokinetic analysis is now
`carried out routinely to aid interpretation of toxicology
`studies. Thirdly, pharmacokinetics are essentially man(cid:173)
`datory to the conduct of a modem phase I clinical trial.
`Most major cancer centres participating in phase I
`studies have access to appropriate analytical equipment
`and pharmacokinetic expertise.
`It is in this general context - of a growing apprecia(cid:173)
`tion of the value of a knowledge of how the drug is
`handled by the body - that the concept of pharmaco(cid:173)
`kinetically-guided dose escalation or PGDE has been
`developed.
`
`Origins ofPGDE
`
`PGDE was originally proposed by Collins and co(cid:173)
`workers [4] in the Blood Level Working Group of the
`Division of Cancer Treatment at the US NCI. The
`impetus came from the widespread frustration that
`many phase I studies were requiring an unacceptable
`number of dose escalations before the maximum toler(cid:173)
`ated dose (MTD) could be defined. For example a
`
`INTELGENX 1035
`
`
`
`Downloaded from
`
`http://annonc.oxfordjournals.org/
`
` at University of Wisconsin - Madison on February 2, 2016
`
`ed factor of two method, has received the widest ac(cid:173)
`ceptance, probably because of concerns over using
`dose escalations greater than a doubling. It should be
`mentioned that where the AUC at the phase I entry
`dose is close to the target (mouse LDlO) AUC, then
`the escalation would actually proceed more cautiously
`than with the conventional approach.
`An attractive feature of the PGDE concept is that it
`is rationally based. Moreover, although it should be a
`very safe method, it nevertheless permits a faster dose
`escalation in many cases. The retrospective study of
`selected drugs showed that the potential savings are
`large: 5-6 steps rather than twice that number with
`doxorubicin, for example.
`
`Further development ofPGDE
`
`Stimulated by the results of the Blood Level Working
`Group, the EORTC Pharmacokinetics and Metabolism
`Group (now the EORTC Pharmacology and Molecular
`Mechanisms Group) in tum analyzed their consider(cid:173)
`able retrospective experience of pharmacokinetics in
`relation to toxicity [14J. Despite expressing certain con(cid:173)
`cerns about the quality of the data in their retrospective
`analysis, it was felt that PGDE would have been of
`value in a number of cases. Particularly good examples
`were certain anthracyclines and platinum complexes.
`Details of the excellent retrospective study with 3 pla(cid:173)
`tinum compounds were published by van Hennick et a\.
`[ 15]. In contrast the thorough study of Kerpel- Fronius
`et al. [16J showed that AUC considerations did not
`help with the explanation for the observation of a lower
`toxicity for the alkylating agent diacetyl dianhydro(cid:173)
`galactitol in humans compared to mice. Chloroethylat(cid:173)
`ing agents such as nitrosoureas and temozolomide did
`not perform well. Not surprisingly, drugs requiring
`metabolic activation also gave poor correlations (e.g.
`melamines, dacarbazine, N-methyl formamide) as did
`the antifolate MZPES where peak plasma levels pre(cid:173)
`dicted better than AUC for the acute neurotoxic effect
`[14J.
`The EORTC PAM Group report emphasized that
`particular care should be taken where there are species
`differences in plasma protein binding or metabolism.
`Another important point is that the existence of non(cid:173)
`linear pharmacokinetics will generally preclude the use
`of the 'conventional' PGDE approaches mentioned
`above. On the other hand, pharmacokinetic monitoring
`is especially important in this situation because a small
`increment in dose may produce a disproportionally
`large increase in AUC. Potential technical difficulties
`such as route of administration, formulation and assay
`sensitivity were also highlighted. Inter-patient hetero(cid:173)
`geneity was also thought to be a potential problem.
`Based on the two initial publications [4, 14J and as(cid:173)
`sociated studies, there was a strong feeling that PGDE
`should be subject to a thorough prospective analysis. In
`addition to North American investigators, the EORTC
`
`340
`
`phase I study of flavone acetic acid in this department
`required a total of 14 dose escalations [5J. What this
`means is that in such cases the MTD in man is being
`poorly predicted by animal toxicology studies. In gen(cid:173)
`eral toxicological investigations are carried out in mice,
`rats and sometimes dogs [6,7J. Overall, the LDlO dose
`in mice, when expressed in units of surface area, is a
`fairly accurate predictor for the MTD in man [4, 6,
`8-12J. However, in individual cases the human MTD
`can vary from one-tenth to ten times the mouse LD10.
`Because of the lower limit of one-tenth, the phase I
`starting dose is usually set at this fraction of the mouse
`LDlO in order to introduce an appropriate margin of
`safety without unduly restricting the initial doses to an
`unnecessarily homeopathic level. Again on average,
`about 5 dose escalations would be required to reach the
`MTD using the standard 'modified Fibonacci' scheme
`(sequential increments of 100%, 67%, 50%, 40% and
`then 30%-35%). Clearly, where the entry dose is lower
`than normal or where the mouse gives a falsely low pre(cid:173)
`diction of the human tolerance, a greater number of es(cid:173)
`calations will be needed. This inevitably means that the
`phase I study will consume more resources, time and
`patients, and in addition decreases the likelihood that
`the patients might receive benefit from the drug (the
`objective response rate in phase I trials is 4% [131).
`Collins et al. [41 recognized that disparities in drug
`tolerance between species may be due to differences in
`systemic (i.e. plasma) pharmacokinetics and/or phar(cid:173)
`macodynamics - the latter term referring to target cell
`sensitivity. They then sought to estimate the pharmaco(cid:173)
`kinetic element by comparing retrospectively the drug
`exposures at the mouse LD I 0 and the human MTD. To
`do this they used the standard exposure parameter:
`area under the plasma concentration versus time curve
`(C x T or A UC). For several drugs the A UC values
`gave better agreement than did the dose, indicating that
`pharmacokinetic differences were particularly impor(cid:173)
`tant. Doxorubicin was an especially good example of
`this, whereas with antimetabolites the AUC correla(cid:173)
`tions were poor. In the latter case differences in intra(cid:173)
`cellular handling and metabolism almost certainly pre(cid:173)
`dominate.
`Impressed by these findings, Collins et al. [4J went
`on to suggest that comparative pharmacokinetic meas(cid:173)
`urements in mice and men could be used to guide the
`phase I dose escalation. Consider the situation where
`the plasma AUC at the phase I entry dose is very small
`compared to that at the LD 10 in the mouse: the solu(cid:173)
`tion is to escalate more aggressively, monitoring the
`AUC at each stage, until the human exposure ap(cid:173)
`proaches 40% of the mouse LD 10 exposure. At that
`point the escalation can be completed by a modified
`Fibonacci method. Two possible approaches were sug(cid:173)
`gested: one involved using a single PGDE step by a fac(cid:173)
`tor equal to the square root of the ratio of mouse LD 10
`to the human entry dose (the geometric mean method)
`while the other required a progressive doubling of the
`dose. The latter method, sometimes called the extend-
`
`INTELGENX 1035
`
`
`
`Downloaded from
`
`http://annonc.oxfordjournals.org/
`
` at University of Wisconsin - Madison on February 2, 2016
`
`New Drug Development and Coordinating Committee
`and the CRC Phase 1/II Clinical Trials Committee have
`endorsed this view and are now testing the approach in
`their own trials. The main aim of the present paper is to
`review the prospective experience which has been pub(cid:173)
`lished so far. For further background the reader is
`referred to the update by Collins et al. ]17] and the
`commentary by Newellll8].
`
`Basic recommendations for PGD E
`
`Both Collins et al. II 7] and the EORTC PAM Group
`114] have emphasized that there should not be a stand(cid:173)
`ard rule book for PGDE. It is important to gear the ap(cid:173)
`proach to the pharmacology of the individual agent.
`Nevertheless some basic principles have emerged. The
`minimum requirements are:
`I. Determine the plasma AUC at the mouse LDlO,
`checking for non-linearity and also for protein
`binding in mouse and human plasma. The same
`batch of mice should be used for the toxicity and
`AUC measurements and the clinical route and
`formulation should normally be employed.
`2. Initiate the phase I study at one-tenth the mouse
`LD I 0 dose and measure the plasma A UC.
`3. Dose escalate to the MTD as appropriate, moni(cid:173)
`toring the AUC at each stage. In general this will
`be done by doubling the dose to approach 40% of
`the mouse LDlO (target) AUC, completing by a
`conventional escalation.
`
`The more extensive requirements proposed by the
`EORTC PAM Group ]14] are:
`I. Determine the presence of metabolites in mice
`and their contribution to the drug's effects.
`2. Develop an assay for the drug (and metabolites)
`sensitive to one-tenth of the mouse LD 10 dose.
`3. Determine the mouse LD10 with acceptable con(cid:173)
`fidence limits and then measure the AUC for the
`drug (and active/toxic metabolites) at the LD10,
`half the LD 10 and one-tenth the LD 10. Watch
`out for non-linear kinetics.
`4. Quantitate protein binding in mouse and human
`plasma at relevant concentrations.
`5. Initiate the clinical study at one-tenth the mouse
`LD I 0 dose and treat 3-5 patients to determine
`the AUC with acceptable accuracy.
`6. Use an appropriate escalation scheme to reach
`the target AUC with monitoring of drug and
`metabolite levels at each escalation and modify as
`required for non-linearity.
`
`Review of the prospective PGDE experience
`
`Amphethinile
`
`The spindle poison amphethinile was one of the first
`agents to prospectively undergo PGDE in a phase I
`
`341
`
`study. Preclinical toxicological studies estimated that
`the acute i.v. LD10 was 400-411 mg/m 2 from a which a
`phase I starting dose of 40 mg/m2 was derived (Table
`1 ). The pharmacokinetics of the drug were investigated
`in mice at 100,200 and 400 mg/m2 to establish a target
`AUC (313 Jlg 1- 1 h) and to check for linear pharmaco(cid:173)
`kinetics [19]. These preclinical analyses revealed that
`the pharmacokinetics of amphethinile were non-linear
`in mice as each doubling of the dose resulted in a
`3-4-fold increase in AUC ]19]. Despite non-linear
`pharmacokinetics it was decided to attempt a PGDE
`strategy in the phase I with the intent of escalating the
`dose by a modification of the geometric mean method
`]4]. Included in these proposals was the provision for a
`maximal initial dose escalation of 5n (where n is the
`phase I starting dose) providing that the AUC at the
`starting dose was <5% of the target AUC ]19]. Phase I
`studies were initiated at one-tenth the mouse LD10 (40
`mg/m 2) and pharmacokinetic monitoring performed in
`3 patients. Unfortunately the drug could not be detect(cid:173)
`ed at the starting dose, a major contributory factor
`being the relative insensitivity of the assay method em(cid:173)
`ployed (100 ng/ml) ]19]. On the basis of the failure to
`detect the drug at 40 mg/m 2 the dose was escalated by
`a large increment (5n) to 200 mg/m2 at which the
`pharmacokinetics of the drug could be estimated. In
`retrospect this aggressive dose escalation strategy is
`questionable, particularly in view of the non-linear
`pharmacokinetics of amphethinile in mice. However,
`no serious toxicity was encountered in the 3 patients
`treated at this dose level with the exception of one pa(cid:173)
`tient who experienced a grand mal convulsion follow(cid:173)
`ing a second course ]19]. It was felt that this toxicity was
`probably related to the rate of injection, a phenomenon
`which had been observed preclinically in mice. Sub(cid:173)
`sequently, the dose was progressively doubled to 400
`and 800 mg/m 2, but given as a short infusion rather
`than a bolus injection to alleviate acute neurotoxicity.
`Although the AUC values at 800 mg/m 2 ranged from
`24-81 Jlg 1- 1 h, considering that dose limiting toxcities
`were absent, the dose was escalated further to 1200
`mg/m 2• Severe nausea/vomiting and alopecia were ex(cid:173)
`perienced and two deaths occurred within 48 h after
`drug treatment at 1200 mg/m 2, at which point the trial
`was terminated. Pharmacokinetic analysis at this last
`dose level revealed that the plasma AUC values (361
`Jlg 1- 1 h) were either similar to the target AUC (313
`Jlg 1- 1 h) or substantially lower (154-195 Jlg 1- 1 h)
`(Table I). Surprisingly, the lower values were associat(cid:173)
`ed with the fatalities.
`The experience with amphethinile illustrates some
`of the difficulties with implementing PGDE studies. It
`stresses the importance of developing sensitive analyti(cid:173)
`cal procedures to detect the drug at the starting dose
`and highlights the problems of non-linear pharmaco(cid:173)
`kinetics. In retrospect, employing a smaller initial dose
`increment (2n) may have been more appropriate, al(cid:173)
`though no serious toxicities were encountered follow(cid:173)
`ing the 5n dose escalation. However, this may not be
`
`INTELGENX 1035
`
`
`
`342
`
`Table I. Prospective evaluation of pharmacokinetically guided dose escalation strategies in phase I trials.
`
`Drug
`
`Class and
`schedule
`
`Mouse
`LDIO
`(mglm')
`
`LDIO
`AUC
`
`Target
`AUC
`
`Starting Human Mean
`dose
`MTD
`AUC at
`(mg/m 2)
`(mg/m 2) MTD
`
`Protein
`bmding (%)
`
`Comments
`
`Refs
`
`Downloaded from
`
`Limtted assay senSttivtty.
`Drug not detectable at starting dose.
`Escalation accelerated.
`
`19
`
`Combination of modtfied Fibonacct 20
`and PGDE when mouse data
`became avatlable.
`Escalation accelerated.
`
`Marked species differences in
`metabolism to 1-Doxol. 1-Dox and
`1-Doxol AUCs were summed and
`doses escalated by PGDE scheme
`using origmal 1-Dox target AUC.
`Escalation accelerated.
`
`21
`23
`24
`
`Wide interpatient vanability in drug 211
`clearance at starting dose
`30
`PGDE scheme could not be used.
`3 I
`
`211
`32
`
`41 I
`
`313•
`
`313•
`
`40
`
`1,200
`
`236•
`
`ND ND
`
`Mice Human
`
`396
`
`10,070"
`
`10,070"
`
`15"
`
`2,250
`
`7854"
`
`ND ND
`
`Amphe(cid:173)
`thinilc
`
`Spindle poison
`(once every 3
`weeks)
`
`Brcqinar
`sodium
`
`Antimetabolite
`(once every 3
`weeks)
`
`Jodo(cid:173)
`doxo(cid:173)
`rubicin
`
`Anthracycline
`(once every 3
`weeks)
`
`19
`
`2
`
`80
`
`0.31<
`
`lJ6
`
`93
`
`1-Doxol Metabolite
`
`20
`
`6.W
`
`NA
`
`NA
`
`NA
`
`95
`
`DUP-941 Anthrapyrazole 52
`(once every 3
`weeks)
`
`277"
`
`II O"
`
`DUP-941
`
`ND
`
`ND
`
`277d
`
`5
`
`2
`
`50
`
`288"
`
`114
`
`95
`
`24
`
`151-399" 84
`
`95
`
`http://annonc.oxfordjournals.org/
`
` at University of Wisconsin - Madison on February 2, 2016
`
`The target AUC m this study was
`based on the AUC wtth a singlet v
`dose LD I 0 1281. Drug was not
`detected at the starting dose and
`wtdc mtcrpatient variability in
`clearance prevented the use of a
`PGDE scheme.
`
`Escalatton accelerated using a
`PGDE scheme (6-9 fewer
`pattents claimed).
`
`PGDE could not be used due to
`assay insensittvity at startmg dose
`and rapid plasma clearance. AUC
`at human MTD exceeded LD I 0
`AUC by 40"/o
`
`PGDE could not be used due
`to wide inter-patient variability m
`drug clearance.
`
`Drug not detectable at starting
`dose. Marked species dtfference' m
`plasma AUC values. Suspected
`pharmacodynamic differences in
`bone marrow stem cell sensitivity.
`PGDE scheme could not be used.
`
`Pharmacokinetics assisted do»e
`esc<tlatton. AUC predicted better
`than dose. MTD from daily x 5
`study tncreased the entry dose
`for the 24 hour contmuous
`infusion schedule.
`
`35
`
`31i
`
`37
`
`3Y
`
`40
`
`(once weekly
`for 3 weeks)
`
`DUP-942 Anthrapyrazolc 75
`(once every 3
`weeks)
`
`177'
`
`59'· 1
`
`7.5
`
`lliO
`
`126'
`
`ND ND
`
`DUP-942 (once every 3
`weeks)
`
`75
`
`300 1
`
`120 1
`
`7.5
`
`150
`
`435'
`
`ND ND
`
`DUP-937 Anthrapyrazole
`(once every 3
`weeks)
`
`Rhizoxin
`
`Spindle Poison
`(once every 3
`weeks)
`
`Aphidi(cid:173)
`colin
`glycinate
`
`DNA
`polymerase
`inhtbitor
`(once daily for
`5 days every 3
`"''eeks)
`
`(24 hour
`continuous
`infuston)
`
`31i
`
`3145•
`
`125H•
`
`3.6
`
`25.2
`
`5271•
`
`ND ND
`
`8-12
`
`71 d
`
`21\d
`
`OH
`
`2.6
`
`0.45"
`
`96
`
`97
`
`300
`
`40.1"
`
`lli-40"
`
`12
`
`2250
`
`62.5
`
`ND ND
`
`ND
`
`ND
`
`ND
`
`435
`
`4500
`
`157
`
`ND ND
`
`• J.l& 1-o h- 1: "J.l& ml 1 h: < J.IM h; " J.IM mm; <I'& ml 1 min; 'J.Imoll- 1 min; • ng ml- 1 h; • Starting dose-'/, toxic dose low in the dog;
`
`1 A UC values determined at mouse LD50 (90 mglm'). Target AUC calculated as follows: 177 I'S ml 1 min x 7 5/90 x 0.4 - 59 I'& ml- 1 min
`
`the case for other anti-tumour agents and it is re(cid:173)
`commended that dose increments do not exceed a
`factor of 2n in the absence of pharmacokinetic data at
`the starting dose or if non-linear pharmacokinetics are
`suspected.
`
`Brequinar sodium
`
`Brequinar sodium is a novel quinoline carboxylic acid
`analogue which blocks de novo pyrimidine biosynthesis
`by inhibiting mitochondrial dihydroorotate dehydro-
`
`INTELGENX 1035
`
`
`
`Downloaded from
`
`http://annonc.oxfordjournals.org/
`
` at University of Wisconsin - Madison on February 2, 2016
`
`343
`
`genase. As a prelude to clinical studies, the toxicity of
`brequinar sodium was evaluated in mice following a
`single i.v. bolus dose and the LD 10 estimated as 396
`mg/m 2 [20). A check on the toxicity of the proposed
`phase I starting dose in a second species, the dog,
`proved toxic. Hence a starting dose equivalent to one(cid:173)
`third of the toxic dose low (TDL) in the dog, namely 15
`mg/m 2
`, was derived for phase I investigations (Table 1).
`At the onset of the phase I study no pharmacokinetic
`or target AUC data were available and for the first half
`of the trial dose escalations followed a modified
`Fibonacci scheme up to 300 mg/m2 without encoun(cid:173)
`tering any severe toxcity [20). Mouse pharmacokinetic
`data (determined at the LDlO dose only) became avail(cid:173)
`able during the trial, and revealed that the plasma levels
`at 300 mg/m 2 were approximately 1/20th the target
`AUC. On the basis of these data and the lack of any
`measurable toxicity, the dose was doubled for the next
`two steps up to 1200 mg/m 2• Subsequent escalations
`were based on clinical judgement with 25% incremen(cid:173)
`tal rises in poor risk patients and 50% increments in
`good risk patients until the MTD had been established.
`This study can be acclaimed as a successful example of
`a PGDE strategy which saved at least three unneces(cid:173)
`sary dose escalation steps, as compared to a conven(cid:173)
`tional modified Fibonacci approach [20).
`
`lodo-doxorubicin
`
`4'-Iodo-4'-deoxydoxorubicin (I-Dox) is one of a num(cid:173)
`ber of new anthracycline derivatives, substituted in the
`daunosamine sugar, which possesses anti-tumour activ(cid:173)
`ity in anthracycline resisitant tumours and reduced car(cid:173)
`diotoxicity in experimental models [21, 22). In a pro(cid:173)
`spective phase I study a PGDE strategy was attempted
`on the basis of somewhat limited mouse pharmaco(cid:173)
`kinetic and toxicity data [23). A particularly thorough
`clinical study was documented by Gianni et al. [24]
`which incorporated the essential components of the
`proposals by Collins et al. [4) and the refinements of
`the EORTC Pharmacokinetics and Metabolism Group
`I 1 4). A detailed and well conducted strategy was de(cid:173)
`vised to overcome problems encountered as a result of
`marked differences in the metabolism of I-Dox in
`mouse and man [24).
`Clinical testing of l-Dox commenced 2 mg/m 2, ap(cid:173)
`proximately 1/lOth of the mouse LD10 (19 mg/m 2),
`with the intent of escalating the dose to 40% of the tar(cid:173)
`get AUC (5 J.LM h) by an extended factor of two
`method [24). However, at the starting dose it became
`apparent that there were marked species differences in
`the pharmacokinetics and metabolism of the drug. This
`disparity was principally due to the rapid and extensive
`formation of 4'-iodo-4'-deoxy-13-dihydrodoxorubicin
`(1-Doxol) due to an aldo-keto reductase which was ab(cid:173)
`sent in mice [24J. Consequently, the initial 6 dose esca(cid:173)
`lations up to 26 mg/m 2 utilised a modified Fibonacci
`scheme whilst the pharmacology of I-Dox and l-Doxol
`were investigated further. A particularly commendable
`
`aspect of this study was the series of comparative phar(cid:173)
`macological investigations on I-Dox and l-Doxol in
`vitro and in vivo. F!rstly, this study demonstrated the
`equivalent toxicity of I-Dox (IC50 50nM) and I-Doxol
`(IC50 80nM) against the growth of human bone mar(cid:173)
`row cells in vitro (CFU-GM assay). Secondly, the toxi(cid:173)
`city of the parent drug and metabolite were shown to
`have comparable LD 10 values in mice, namely 6 mg/
`kg and 6.8 mg/kg for 1-Dox and 1-Doxol respectively
`(Table 1) [24). As a result of these findings a PGDE
`scheme was re-introduced and subsequent doses esca(cid:173)
`lated by the summation of the 1-Dox and I-Doxol
`AUCs, utilising the original target 1-Dox AUC to guide
`the dose increments. At 26 mg/m 2 the mean sum of
`these AUCs was 1.4 J.LM h which was 24% of the target
`AUC. It was at this point that the authors then aban(cid:173)
`doned the Fibonacci scheme, for which escalations
`were already limited to 35%, and doubled the dose of
`52 mg/m2• At this dose grade 4 granulocytopaenia was
`seen in one patient and the Fibonacci escalation was
`therefore resumed to complete the study. Pharmaco(cid:173)
`kinetic analysis at the MTD of 80 mg/m 2 revealed a
`close match between the AUC at the mouse LDlO and
`sum of the 1-Dox/1-Doxol plasma AUC in patients,
`which lends support to the original Collins hypothesis
`(4). Use of PGDE at a relatively late stage saved a single
`dose escalation step. What is particularly encouraging
`from this study is the notion that the MTD could have
`been reached in only five steps had the comparative
`pharmacological information on l-Dox and 1-Doxol
`been available at the outset.
`
`Anthrapyrazole DuP941 (CJ-941)
`
`The anthrapyrazoles were synthesised in an attempt to
`find a non-cardiotoxic DNA binding drug which re(cid:173)
`tained or possessed superior anti-tumour activity to
`doxorubicin [25-27[. Three lead compounds were
`identified (CI-941, CI-942 and CI-937 now prefixed
`DuP), each of which displayed excellent anti-tumour
`activity against murine tumours J26J all three were sub(cid:173)
`sequently developed for phase I clinical testing. PGDE
`strategies were developed for all three agents but these
`met with mixed success due to a number of unantici(cid:173)
`pated problems.
`It was recognised from the experience with amphe(cid:173)
`thinile that sensitive analytical methods were a pre(cid:173)
`requisite for PGDE studies. A highly sensitive solid
`phase extraction and HPLC assay (limit of detection 1
`ng/ml) was developed for DuP941 which was subse(cid:173)
`quently used to describe the pharmacokinetics of
`DuP941 in mouse and human plasma J29J. The pre(cid:173)
`clinical pharmacokinetic and
`toxicity studies of
`DuP941 were integrated so that direct correlations
`could be drawn between drug exposure (AUC) and
`toxicity (i.e. LD10/LD50). These studies were per(cid:173)
`formed in the same batch of randomised animals, with
`the clinical formulation of the drug, given by the same
`route of administration to be used clinically (i.v.) [28).
`
`INTELGENX 1035
`
`
`
`Downloaded from
`
`http://annonc.oxfordjournals.org/
`
` at University of Wisconsin - Madison on February 2, 2016
`
`344
`
`Following a single i.v. bolus injection the LDlO and
`LD50 values were accurately determined with 95%
`confidence limits: the LDlO was 20 mg/kg (19-21
`mg/kg) and the LD50 was 22 mg!kg (21-23 mg/kg)
`[28). Pharmacokinetc studies were then performed at 4
`dose levels in mice over a dose range which encom(cid:173)
`passed one-tenth of the LDlO and the LD50, namely
`1.5, 10, 15 and 20 mg/kg. Analysis of these data re(cid:173)
`vealed that the pharmacokinetics were linear up to 15
`mg/kg but became significantly non-linear at higher
`doses, a feature which was probably reflected in the
`extremely steep dose-lethality relationship [28). The
`LD10 AUC was calculated at 277 11M min; however in
`view of the non-linear pharmacokinetics it was decided
`that the AUC value at 15 mg/kg (110 f.lM x min) was
`an appropriate target AUC as this figure represented
`40% of the LD10 AUC and was the highest dose at
`which linear pharmacokinetics were observed [28[. An
`additional component of this study was the comparison
`of plasma protein binding in mouse and human plasma
`which was found to be similar between the two species
`(Table 1).
`Phase I studies with DuP941 were initiated at 5
`mg/m 2 (1/10th the mouse LDlO) with the aim of esca(cid:173)
`lating the dose by an extended factors of 2 method until
`the target AUC had been achieved. If no toxicity was
`encountered at this stage the remainder of the escala(cid:173)
`tions were to follow a modified Fibonacci scheme until
`the MTD had been reached. However, at the starting
`dose there was wide interpatient variability in drug
`clearance [30, 31[ with one patient attaining approxi(cid:173)
`mately 50% of the target AUC. It was felt that the
`AUC data accrued at the starting dose could not be
`used to guide subsequent escalations and the dose was
`increased cautiously in 5 mg/m 2 increments up to 55
`: 50 mg/m 2 being the recommended dose for
`mg/m 2
`single administration schedules [30, 31[. Wide inter(cid:173)
`patient variability in drug clearance was observed at all
`dose levels and detailed investigations to identify the
`causes were undertaken [30). Measures of interpatient
`plasma protein binding, pre-treatment renal (5 1Cr(cid:173)
`EDTA clearance), hepatic (plasma alanine transami(cid:173)
`nase and alkaline phosphatase levels), or cardiac func(cid:173)
`tion (left ventricular ejection fractions) did not strongly
`correlate with drug clearance [30[. Overall, it could be
`argued that the use of pharmacokinetics and the selec(cid:173)
`tion of a fixed dose increment of 5 mg/m 2 increased the
`total number of dose escalation steps. In retrospect this
`is probably a valid criticism: adopting a modified
`Fibonacci scheme would have allowed reaching the
`MTD in 6 steps instead of 11 [31[. However, the use of
`a modified Fibonacci scheme would also have meant
`that the MTD would have been exceeded at the 60
`mg/m 2 dose level and it is likely that lower dose levels
`{50-55 mg/m 2) would also have been studied, making
`a total of 9 possible dose levels [30[. In conclusion,
`therefore,
`the use of this conservative escalation
`scheme did not severely retard the phase I study of
`DuP941. More
`importantly,
`the pharmacokinetics
`
`were helpful in the decision to employ cautious fixed
`escalations.
`Wide interpatient variability in clearance was also
`observed in a second phase I study of DuP941 given
`by an alternative schedule (once weekly for 3 weeks)
`[32). Similar variability in drug clearance was also en(cid:173)
`countered in this study and a PGDE scheme could not
`be implemented. However, pharmacokinetic analyses
`demonstrated that the drug could be given on a repeat
`administration schedule without drug accumulation
`[32) which further illustrates the value of pharmaco(cid:173)
`kinetic monitoring in phase I trials.
`
`Anthrapyrazole DuP942 (C/-942)
`
`Two phase I studies with PGDE schemes have been
`described for the anthrapyrazole DuP942 {CI-942)
`[33-36). The first study by Ames et al. [35] was based
`on a limited pharmacokinetic investigation of DuP942
`in mice. In this study the mouse LD10 and LD50 were
`calculated to be 75 and 90 mg/m 2, respectively, from
`which a starting dose of 7.5 mg/m 2 was derived. As full
`pharmacokinetic data at the mouse LD 10 were not
`available at the start of the study, the target AUC was
`estimated from the AUC at the LD50 {177 !Lg ml- 1
`min) according to the following equation:
`
`40% target AUC = 177 !J.g.min/ml x 75/90 x 0.4 =
`59 !J.g.min/ml.
`
`The underlying assumption here is that the pharmaco(cid:173)
`kinetics of DuP941 are linear between 75 and 90
`mg/m 2, which is clearly not the case for the structurally
`similar anthrapyrazole DuP941 [28[. Consequently it
`is possible that deriving the target AUC from the phar(cid:173)
`macokinetic data at the LD50 could result in an over(cid:173)
`estimation of the target AUC. This may encourage the
`use of an overly aggressive and potentially hazardous
`dose escalation scheme, particularly in the presence of
`non-linear kinetics. However, as a sufficient safety
`element was introduced into the equation, by defining
`the target AUC as 40% of the estimated LDlO AUC,
`the risk of escalating too rapidly to a supra-toxic dose is
`minimised. Nevertheless, the safest option, wherever
`possible, is to unequivocally demonstrate linear kinet(cid:173)
`ics in the mouse and then to calculate the