- Email: [email protected]
Phase I clinical trials: Adapting methodology to face new challenges
Annals of Oncology 5 (Suppl. 4): S67-S70, 1994. O 1994 Kluwer Academic Publishers. Printed in the Netherlands.
Symposium article Phase I clinical trials: Adapting methodology to face new challenges
D. J. Kerr Department of Clinical Oncology, Queen Elizabeth Hospital, Birmingham,
Background: The conventional phase I clinical trial stands as a bridge between the laboratory and the clinic in the development of novel anticancer agents. The purpose of the phase I study is primarily to determine toxicity and to define the maximum tolerated dose of the drug in man. It is assumed for this purpose that dose-response curves for toxicity and efficacy are parallel or, simply expressed, the more pain, the more gain. Novel antineoplastic drugs are being developed that are mechanistically remote from conventional cytotoxic
Introduction
The relative lack of impact of conventional cytotoxic drugs on survival of patients with common solid tumours, once they reach an advanced stage, provokes a dual academic response. There are calls to invest in research into optimizing clinical use of currently available anticancer drugs (either empirically or following rational leads based on pharmacokinetics or mechanism of action), and to keep searching for new drugs in the hope that the next compound tested could be 'the one'. The purpose of this paper is to consider the challenges facing phase I clinical trialists as more mechanistically unconventional drugs are being considered for clinical study. It is prudent, however, to set the current practise of phase I trials within a historical context.
drugs, which have DNA as their predominant target; some of these new agents have, at least in vitro, bell-shaped doseresponse curves. Conclusion: It is essential that flexible clinical trial methodologies are developed to accommodate new drugs and that attempts are made, when possible, to incorporate pharmacodynamic endpoints in addition to toxicological endpoints. Key words: anticancer drugs, pharmacodynamics, pharmacokinetics, phase I trials
mated in vitro screen, using panels of well characterized human cancer cell lines, with a view to delineating compounds that may be relatively organ specific (e.g. active against breast cancer cells but not against colon cancer cells). A wave of compounds has just rolled off this screen, and it will be fascinating to observe the predictive power of the in vitro screen as these compounds enter the clinic. Molecular roulette Random chemical synthesis of analogues of existing antineoplastic drugs occasionally produces a drug with an improved pattern of toxicity and a slightly better therapeutic ratio, this method contributes little to the concept of curing advanced solid tumours. Serendipity
Sources of novel anticancer drugs Screening
Unfortunately, screening has developed a rather bad name. It is associated with initial testing of new chemicals against murine leukaemia cells growing intraperitoneally, followed by progression through a complex decision network of murine tumours, which are relatively refractory to most cytotoxic drugs. This expensive, random screening technique has not been at all successful in identifying new, clinically active anticancer drugs. The National Cancer Institute in Bethesda, Maryland, U.S.A., has set up an anticancer semi-auto-
Accidental discoveries result from an ungovernable and intermittent force which needs to be coupled ultimately with acute observation. Cisplatin provides a good example of such a drug discovery story. Rational drug design This method depends on being able to define the phenotypic differences that separate the transformed from the normal host cell at a molecular level, with a view to synthesizing inhibitors or activators of key biochemical pathways. Surely, this is where the future of anticancer drug development lies.
Downloaded from http://annonc.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on June 20, 2015
Summary
U.K.
68 Traditional phase I model
Pitfalls of the traditional phase I approach
Phase I trials are rather poor at detecting chronic or cumulative toxicities, as relatively few patients receive multiple courses. Furthermore, few patients entered into phase I trials benefit in terms of tumour control or prolongation of survival [3]. Such patients have tumours that have proved refractory to conventional therapy or for which no conventional therapy exists; therefore, by definition, they have cancer that is likely to be unresponsive. On average, phase I studies accrue 20-25 patients, but this number may vary enormously. Studies requiring in excess of 40 patients are not uncommon [4], which represents a waste of a human resource and means that the more time spent at low dose levels, the lower the likelihood of response. Dose-response curves that deviate from the expected linear 'big is beautiful' pattern may produce aberrant results. Three rather different dose-response relationships are summarized in Fig. 1. If dose-response curves for toxicity and efficacy are parallel (Fig. la), the prac-
Therapeutic
Adverse
100 i-
CD CD
2
3 4 5 Dose (arbitrary units)
Adverse _
100
Therapeutic
CO
-
50
S. 2
,-.
Adverse
100r
o a>
Therapeutic
3=
a •§
I 3 4 5 6 Dose (arbitrary units)
50
1
2
3 4 5 6 Dose (arbitrary units)
7
8
Fig. 1. Dose-response curves showing efficacy and toxicity, a) parallel curves; b) plateau curves; c) a bell-shaped curve. Reproduced with permission from Kerr DJ, Proc R Coll Physicians Edinb 1993; 23:636-40.
Downloaded from http://annonc.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on June 20, 2015
Once the decision has been taken to proceed to clinical trial of an active compound, the next preclinical stage is to determine the toxicology of the drug in animals [1]. This is to produce descriptive data and can be predictive of the pattern of side-effects in humans, particularly myelosuppression. One of the key toxicological endpoints is the LD10 (i.e. the dose of drug that kills 10% of treated mice). The starting dose for the human study (corrected for body surface area) is 1/10 of the murine LD 10 . Historically, the maximum tolerated dose (MTD) in humans and the LD10 in mice have been comparable, hence the mathematical relationship for the derivation of an arbitrarily safe starting dose [2]. Three patients are treated at each dose level as the drug dose is increased according to the Fibonacci regimen (i.e. dose level 1 = X, (i.e. 1/10 LD 10 in mice); dose level 2 = X, + 100% X,; dose level 3 - X2 + 50% X2; dose level 4 •= X3 + 30% X3; dose level n - Xn - 1 + 30% X,, _,). The drug dose is rapidly escalated from entry, doubling initially but utilizing more modest dose increments (i.e. 30% of previous dose level) when the dose rises and toxicity supervenes. The ethical intention must be to treat, and if it appears that the drug is likely to be effective only when associated with toxicity, then it makes sense to employ rapid, initial dose increments. The MTD is defined as the dose of drug that causes unacceptable toxicity (usually common toxicity criteria or World Health Organization grade 3 or 4) in 3 out of 6 patients entered at that dose level. The dose level immediately preceding this will be scrutinized closely, and more patients will be entered at that dose level, which is likely to be recommended as the dose/schedule for phase II studies.
tice described in the phase I studies is reasonable. If the efficacy curve reaches a plateau with a point of inflexion at a significantly lower dose than gives rise to toxicity (Fig. lb), the implication is that needless side-effects are induced in the mistaken belief that there is 'no gain without pain'. The worst scenario is depicted in Fig. lc. If the dose-efficacy response is bell-shaped, then it is possible that efficacy could be pushed down the descending limb of the response curve to a subtherapeutic level; several studies have shown this type of doseresponse curve for the cytokine, interferon, both in vitro and clinically (e.g. in the treatment of Kaposi's sarcoma).
69 Is it possible to modify the practice of phase I studies to accommodate new drugs with novel mechanisms of action and so improve the efficiency of such studies? One method based on pharmacokinetic observations is possible.
Drug therapy
Adjust drug dose
>
t
Pharmacokinetically guided dose escalation (PGDE)
Mechanism of action-based approach to phase I studies
The majority of conventional chemotherapeutic drugs inhibit DNA synthesis. There is now an increasing tendency to attempt rational design of anticancer drugs based on the biochemical pathways that differentiate the normal from the transformed cellular phenotype. If a drug has a specific target, such as inhibition of a key enzyme, then it might be possible to titrate drug dose to target effects, rather than to toxicity. This major shift in emphasis from an empirical to a mechanism of actionbased methodology is an important innovation in phase I trial design. We have become accustomed to performing detailed pharmacokinetic studies during the phase I trial, and these can provide much useful information. The addition of pharmacodynamic information (Fig. 2) will allow the development of sophisticated models, which may predict whether the antineoplastic drug will be clinically active and may also test the hypothesis of the putative mechanism of action. For example, the process of metastasis consists of a series of linked steps involving cancer cell mobility, digestion of intercellular matrix proteins (e.g. collagen, laminin, fibronectin), altered vascular permeability and organ specific targeting, followed by invasion and establishment of growth. The metalloproteinases,
I Reiterative \ 1 refinement 1 Vof drug dosey
>r
>
How far is this value from thle target or threshold level for optimal activity?
Concentration effect model (Bayes theorem)
Fig. 2. Mechanism-based design for a phase I study. Reproduced with permission from Kerr DJ, Proc R Coll Physicians Edinb 1993; 23:636-40.
which include subsets of enzymes that digest collagen, laminin and gelatin, contribute to the early phase of the metastasic process. Patients with bony metastases show evidence of increased collagen degradation with increased urinary excretion of collagen breakdown products, such as pyridinium cross-links. Inhibitors of certain metallothionein proteinases have been developed and we have designed a phase I study to which we shall recruit patients with bony metastases. Dose increments will be introduced according to a Fibonacci regimen, and an HPLC assay system for pyridinium cross-links has been devised so that their urinary excretion may be monitored throughout. The end of the study will be either the MTD of the drug or the dose that causes maximum inhibition of urinary excretion of pyridinium cross-links (i.e. maximum enzyme inhibition), whichever is reached first. Pharmacokinetic studies of drug distribution will be performed simultaneously, and an attempt will be made to construct concentration-effect models, relating plasma concentration to the degree of enzyme inhibition (Fig- 2). As the search for novel anticancer drugs continues and moves away from empirical analogue development to a theory-based drug design programme, phase I trialists should continue to have a flexible and thoughtful approach to developing clinical trials to suit the drug, rather than incorporating new drugs into a rigid intellectual framework. There is considerable interest in developing inhibitors of signal transduction pathways which contribute to maintenance and initiation of the transformed phenotype. We have just commenced a phase I study of the naturally occurring compound which inhibits receptor-activated tyrosine kinase and protein kinase
Downloaded from http://annonc.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on June 20, 2015
This approach is an attempt to abbreviate phase I trials and permit rapid but safe dose escalation from the lower dose levels. The aim is to reach a target area under the plasma concentration-time curve (AUC) in humans in four to seven steps; it is assumed that the plasma AUC at the LD 10 in mice and the MTD in humans are similar. Pharmacokinetic studies are performed in mice at the LD 10 , and a target plasma AUC can be derived for clinical study [5]. Dose escalation is, therefore, aimed at a 'target AUC' and can proceed more efficiently than with the modified Fibonacci method. The PGDE method requires that an appropriately sensitive assay is available for murine and human blood samples, and that the AUC is related in a linear fashion to the administered dose. Although this approach is theoretically superior to the empiricism of the past, it has proved difficult to implement prospectively within a phase I trial. Further doubt has been cast on the usefulness of this approach with those cytotoxic drugs, such as rhizoxin, which show very large pharmacodynamic differences between species [6,7].
Estimate therapeutic target activity (if an enzyme) or concentration
70 C. The usual toxicology endpoints will be met and pharmacokinetic studies will be performed so that plasma concentrations can be compared to drug concentrations effective in enzyme inhibition in vitro. In collaboration with colleagues in the immunology department, we have devised a series of assays which will allow us to test whether the drug is inhibiting the target enzymes in activated lymphocytes. This mechanistic approach allows us to test the drug hypothesis in patients. Acknowledgements
References 1. Rozencweig M, Von Hoff DD, Staquet MJ et al. Animal toxicology for early clinical trials. Cancer Clin Trials 1991; 4: 21-8.
Correspondence to: Professor D J. KenDepartment of Clinical Oncology Clinical Research Block Queen Elizabeth Hospital Birmingham B15 2TH, U.K.
Downloaded from http://annonc.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on June 20, 2015
The author would like to thank Valerie Stephenson for typing the manuscript, and the Cancer Research Campaign for funding research within the Department of Oncology.
EORTC New Drug Developments Committee. EORTC guidelines for phase I trials with single agents in adults. Eur J Cancer Clin Oncol 1985; 21:1005-7. Estey E, Hoth D, Simon R et al. Therapeutic response in phase I trials of antineoplastic agents. Cancer Treat Rep 1986; 70: 1105-9. Kerr DJ, Kaye SB, Setanoians A et al. Phase I clinical and pharmacokinetic study of LM985. Cancer Res 1986; 46: 3142246. Kerr DJ, Kaye SB, Cassidy J et al. Phase I and pharmacokinetic study of flavone acetic acid. Cancer Res 1987; 47:6776-81. Bissett D, Graham MA, Setanoians A et al. Phase I and pharmacokinetics study of rhizoxin. Cancer Res 1992; 52: 2894-8. Graham MA, Bissett D, Setanoians A et al. Pre-clinical and phase I studies with rhizoxin to apply a pharmacokineticallyguided dose escalation scheme. J Natl Cancer Inst 1992; 84: 494-500.
Symposium article Phase I clinical trials: Adapting methodology to face new challenges
D. J. Kerr Department of Clinical Oncology, Queen Elizabeth Hospital, Birmingham,
Background: The conventional phase I clinical trial stands as a bridge between the laboratory and the clinic in the development of novel anticancer agents. The purpose of the phase I study is primarily to determine toxicity and to define the maximum tolerated dose of the drug in man. It is assumed for this purpose that dose-response curves for toxicity and efficacy are parallel or, simply expressed, the more pain, the more gain. Novel antineoplastic drugs are being developed that are mechanistically remote from conventional cytotoxic
Introduction
The relative lack of impact of conventional cytotoxic drugs on survival of patients with common solid tumours, once they reach an advanced stage, provokes a dual academic response. There are calls to invest in research into optimizing clinical use of currently available anticancer drugs (either empirically or following rational leads based on pharmacokinetics or mechanism of action), and to keep searching for new drugs in the hope that the next compound tested could be 'the one'. The purpose of this paper is to consider the challenges facing phase I clinical trialists as more mechanistically unconventional drugs are being considered for clinical study. It is prudent, however, to set the current practise of phase I trials within a historical context.
drugs, which have DNA as their predominant target; some of these new agents have, at least in vitro, bell-shaped doseresponse curves. Conclusion: It is essential that flexible clinical trial methodologies are developed to accommodate new drugs and that attempts are made, when possible, to incorporate pharmacodynamic endpoints in addition to toxicological endpoints. Key words: anticancer drugs, pharmacodynamics, pharmacokinetics, phase I trials
mated in vitro screen, using panels of well characterized human cancer cell lines, with a view to delineating compounds that may be relatively organ specific (e.g. active against breast cancer cells but not against colon cancer cells). A wave of compounds has just rolled off this screen, and it will be fascinating to observe the predictive power of the in vitro screen as these compounds enter the clinic. Molecular roulette Random chemical synthesis of analogues of existing antineoplastic drugs occasionally produces a drug with an improved pattern of toxicity and a slightly better therapeutic ratio, this method contributes little to the concept of curing advanced solid tumours. Serendipity
Sources of novel anticancer drugs Screening
Unfortunately, screening has developed a rather bad name. It is associated with initial testing of new chemicals against murine leukaemia cells growing intraperitoneally, followed by progression through a complex decision network of murine tumours, which are relatively refractory to most cytotoxic drugs. This expensive, random screening technique has not been at all successful in identifying new, clinically active anticancer drugs. The National Cancer Institute in Bethesda, Maryland, U.S.A., has set up an anticancer semi-auto-
Accidental discoveries result from an ungovernable and intermittent force which needs to be coupled ultimately with acute observation. Cisplatin provides a good example of such a drug discovery story. Rational drug design This method depends on being able to define the phenotypic differences that separate the transformed from the normal host cell at a molecular level, with a view to synthesizing inhibitors or activators of key biochemical pathways. Surely, this is where the future of anticancer drug development lies.
Downloaded from http://annonc.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on June 20, 2015
Summary
U.K.
68 Traditional phase I model
Pitfalls of the traditional phase I approach
Phase I trials are rather poor at detecting chronic or cumulative toxicities, as relatively few patients receive multiple courses. Furthermore, few patients entered into phase I trials benefit in terms of tumour control or prolongation of survival [3]. Such patients have tumours that have proved refractory to conventional therapy or for which no conventional therapy exists; therefore, by definition, they have cancer that is likely to be unresponsive. On average, phase I studies accrue 20-25 patients, but this number may vary enormously. Studies requiring in excess of 40 patients are not uncommon [4], which represents a waste of a human resource and means that the more time spent at low dose levels, the lower the likelihood of response. Dose-response curves that deviate from the expected linear 'big is beautiful' pattern may produce aberrant results. Three rather different dose-response relationships are summarized in Fig. 1. If dose-response curves for toxicity and efficacy are parallel (Fig. la), the prac-
Therapeutic
Adverse
100 i-
CD CD
2
3 4 5 Dose (arbitrary units)
Adverse _
100
Therapeutic
CO
-
50
S. 2
,-.
Adverse
100r
o a>
Therapeutic
3=
a •§
I 3 4 5 6 Dose (arbitrary units)
50
1
2
3 4 5 6 Dose (arbitrary units)
7
8
Fig. 1. Dose-response curves showing efficacy and toxicity, a) parallel curves; b) plateau curves; c) a bell-shaped curve. Reproduced with permission from Kerr DJ, Proc R Coll Physicians Edinb 1993; 23:636-40.
Downloaded from http://annonc.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on June 20, 2015
Once the decision has been taken to proceed to clinical trial of an active compound, the next preclinical stage is to determine the toxicology of the drug in animals [1]. This is to produce descriptive data and can be predictive of the pattern of side-effects in humans, particularly myelosuppression. One of the key toxicological endpoints is the LD10 (i.e. the dose of drug that kills 10% of treated mice). The starting dose for the human study (corrected for body surface area) is 1/10 of the murine LD 10 . Historically, the maximum tolerated dose (MTD) in humans and the LD10 in mice have been comparable, hence the mathematical relationship for the derivation of an arbitrarily safe starting dose [2]. Three patients are treated at each dose level as the drug dose is increased according to the Fibonacci regimen (i.e. dose level 1 = X, (i.e. 1/10 LD 10 in mice); dose level 2 = X, + 100% X,; dose level 3 - X2 + 50% X2; dose level 4 •= X3 + 30% X3; dose level n - Xn - 1 + 30% X,, _,). The drug dose is rapidly escalated from entry, doubling initially but utilizing more modest dose increments (i.e. 30% of previous dose level) when the dose rises and toxicity supervenes. The ethical intention must be to treat, and if it appears that the drug is likely to be effective only when associated with toxicity, then it makes sense to employ rapid, initial dose increments. The MTD is defined as the dose of drug that causes unacceptable toxicity (usually common toxicity criteria or World Health Organization grade 3 or 4) in 3 out of 6 patients entered at that dose level. The dose level immediately preceding this will be scrutinized closely, and more patients will be entered at that dose level, which is likely to be recommended as the dose/schedule for phase II studies.
tice described in the phase I studies is reasonable. If the efficacy curve reaches a plateau with a point of inflexion at a significantly lower dose than gives rise to toxicity (Fig. lb), the implication is that needless side-effects are induced in the mistaken belief that there is 'no gain without pain'. The worst scenario is depicted in Fig. lc. If the dose-efficacy response is bell-shaped, then it is possible that efficacy could be pushed down the descending limb of the response curve to a subtherapeutic level; several studies have shown this type of doseresponse curve for the cytokine, interferon, both in vitro and clinically (e.g. in the treatment of Kaposi's sarcoma).
69 Is it possible to modify the practice of phase I studies to accommodate new drugs with novel mechanisms of action and so improve the efficiency of such studies? One method based on pharmacokinetic observations is possible.
Drug therapy
Adjust drug dose
>
t
Pharmacokinetically guided dose escalation (PGDE)
Mechanism of action-based approach to phase I studies
The majority of conventional chemotherapeutic drugs inhibit DNA synthesis. There is now an increasing tendency to attempt rational design of anticancer drugs based on the biochemical pathways that differentiate the normal from the transformed cellular phenotype. If a drug has a specific target, such as inhibition of a key enzyme, then it might be possible to titrate drug dose to target effects, rather than to toxicity. This major shift in emphasis from an empirical to a mechanism of actionbased methodology is an important innovation in phase I trial design. We have become accustomed to performing detailed pharmacokinetic studies during the phase I trial, and these can provide much useful information. The addition of pharmacodynamic information (Fig. 2) will allow the development of sophisticated models, which may predict whether the antineoplastic drug will be clinically active and may also test the hypothesis of the putative mechanism of action. For example, the process of metastasis consists of a series of linked steps involving cancer cell mobility, digestion of intercellular matrix proteins (e.g. collagen, laminin, fibronectin), altered vascular permeability and organ specific targeting, followed by invasion and establishment of growth. The metalloproteinases,
I Reiterative \ 1 refinement 1 Vof drug dosey
>r
>
How far is this value from thle target or threshold level for optimal activity?
Concentration effect model (Bayes theorem)
Fig. 2. Mechanism-based design for a phase I study. Reproduced with permission from Kerr DJ, Proc R Coll Physicians Edinb 1993; 23:636-40.
which include subsets of enzymes that digest collagen, laminin and gelatin, contribute to the early phase of the metastasic process. Patients with bony metastases show evidence of increased collagen degradation with increased urinary excretion of collagen breakdown products, such as pyridinium cross-links. Inhibitors of certain metallothionein proteinases have been developed and we have designed a phase I study to which we shall recruit patients with bony metastases. Dose increments will be introduced according to a Fibonacci regimen, and an HPLC assay system for pyridinium cross-links has been devised so that their urinary excretion may be monitored throughout. The end of the study will be either the MTD of the drug or the dose that causes maximum inhibition of urinary excretion of pyridinium cross-links (i.e. maximum enzyme inhibition), whichever is reached first. Pharmacokinetic studies of drug distribution will be performed simultaneously, and an attempt will be made to construct concentration-effect models, relating plasma concentration to the degree of enzyme inhibition (Fig- 2). As the search for novel anticancer drugs continues and moves away from empirical analogue development to a theory-based drug design programme, phase I trialists should continue to have a flexible and thoughtful approach to developing clinical trials to suit the drug, rather than incorporating new drugs into a rigid intellectual framework. There is considerable interest in developing inhibitors of signal transduction pathways which contribute to maintenance and initiation of the transformed phenotype. We have just commenced a phase I study of the naturally occurring compound which inhibits receptor-activated tyrosine kinase and protein kinase
Downloaded from http://annonc.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on June 20, 2015
This approach is an attempt to abbreviate phase I trials and permit rapid but safe dose escalation from the lower dose levels. The aim is to reach a target area under the plasma concentration-time curve (AUC) in humans in four to seven steps; it is assumed that the plasma AUC at the LD 10 in mice and the MTD in humans are similar. Pharmacokinetic studies are performed in mice at the LD 10 , and a target plasma AUC can be derived for clinical study [5]. Dose escalation is, therefore, aimed at a 'target AUC' and can proceed more efficiently than with the modified Fibonacci method. The PGDE method requires that an appropriately sensitive assay is available for murine and human blood samples, and that the AUC is related in a linear fashion to the administered dose. Although this approach is theoretically superior to the empiricism of the past, it has proved difficult to implement prospectively within a phase I trial. Further doubt has been cast on the usefulness of this approach with those cytotoxic drugs, such as rhizoxin, which show very large pharmacodynamic differences between species [6,7].
Estimate therapeutic target activity (if an enzyme) or concentration
70 C. The usual toxicology endpoints will be met and pharmacokinetic studies will be performed so that plasma concentrations can be compared to drug concentrations effective in enzyme inhibition in vitro. In collaboration with colleagues in the immunology department, we have devised a series of assays which will allow us to test whether the drug is inhibiting the target enzymes in activated lymphocytes. This mechanistic approach allows us to test the drug hypothesis in patients. Acknowledgements
References 1. Rozencweig M, Von Hoff DD, Staquet MJ et al. Animal toxicology for early clinical trials. Cancer Clin Trials 1991; 4: 21-8.
Correspondence to: Professor D J. KenDepartment of Clinical Oncology Clinical Research Block Queen Elizabeth Hospital Birmingham B15 2TH, U.K.
Downloaded from http://annonc.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on June 20, 2015
The author would like to thank Valerie Stephenson for typing the manuscript, and the Cancer Research Campaign for funding research within the Department of Oncology.
EORTC New Drug Developments Committee. EORTC guidelines for phase I trials with single agents in adults. Eur J Cancer Clin Oncol 1985; 21:1005-7. Estey E, Hoth D, Simon R et al. Therapeutic response in phase I trials of antineoplastic agents. Cancer Treat Rep 1986; 70: 1105-9. Kerr DJ, Kaye SB, Setanoians A et al. Phase I clinical and pharmacokinetic study of LM985. Cancer Res 1986; 46: 3142246. Kerr DJ, Kaye SB, Cassidy J et al. Phase I and pharmacokinetic study of flavone acetic acid. Cancer Res 1987; 47:6776-81. Bissett D, Graham MA, Setanoians A et al. Phase I and pharmacokinetics study of rhizoxin. Cancer Res 1992; 52: 2894-8. Graham MA, Bissett D, Setanoians A et al. Pre-clinical and phase I studies with rhizoxin to apply a pharmacokineticallyguided dose escalation scheme. J Natl Cancer Inst 1992; 84: 494-500.