Drug development in oncology: classical cytotoxics and molecularly targeted agents

Abstract

There is an apparent need to improve the speed and efficiency of oncological drug development. Furthermore, strategies traditionally applied to the development of standard cytotoxic chemotherapy may not be appropriate for molecularly targeted agents. This is particularly the case for exploratory Phase 1 and 2 trials. Conventional approaches to determine dose based on maximum tolerability and efficacy based on objective tumour response may not be suitable for targeted agents, since many of them have a wide therapeutic index and inhibit tumour growth without demonstrable cytotoxicity. Instead, exploratory trials of targeted agents may have to focus on other end-points such as pharmacological effects and disease stabilization. Thus, there is an increasing interest in making the best possible use of biomarkers and pharmacogenomics in early phases of drug development.

Introduction

The strategies traditionally used for early clinical development of standard anticancer agents, which usually involve patients with advanced incurable malignancy, differ considerably from those in most other therapeutic areas involving less serious diseases. They may also be inappropriate for the expanding number of novel molecularly targeted oncology agents in development. Oncology drugs go through various phases of clinical trials to establish their use in the treatment of cancer (Table 1). A primary objective of Phase 1 trials of standard cytotoxic chemotherapeutic agents is ordinarily to define maximum tolerated dose (MTD) [1]. This dose is typically recommended for exploratory Phase 2 efficacy trials, under the assumption that the higher the dose the greater the antitumour activity. However, molecularly targeted agents are generally less toxic than standard cytotoxic drugs and treatment effects may occur at doses much lower than the MTD. Furthermore, Phase 2 efficacy trial designs that employ objective tumour response (tumour shrinkage) as the primary end-point, which is commonly used for standard cytotoxic agents, may not be adequate to identify promising activity with cytostatic targeted agents. This paper will review the strategies used in exploratory trials of conventional anticancer drugs and introduce emerging paradigms for the development of novel molecularly targeted agents. For the purposes of this review, molecularly targeted agents are defined as drugs that target growth factor receptors and signal transduction pathways.

Table 1. Traditional phases of cancer drug clinical development.

Phase 1

Trials evaluating the safety, toxicities, and pharmacokinetics as well defining the maximum tolerated dose and recommended Phase 2 dose of a drug or regimen through step-wise dose escalations

Phase 2

Trials with the primary objective of determining whether a new therapeutic agent or regimen has some minimally acceptable level of activity and to provide a rough estimate of this activity level

Phase 3

Trials in which patients are allocated (usually randomly) to one of two or more therapeutic approaches in an attempt to determine the relative role and merits of the treatment through a direct comparison

Phase 4

Postmarketing studies conducted in special populations (e.g. children, elderly, patients with renal or hepatic dysfunction, etc.) or to confirm clinical benefit

Adapted from Simon [1].

Phase 1 dose-finding oncology trials

Phase I trials of anticancer agents are relatively small dose-finding studies, usually involving about 20–60 subjects. Major objectives of Phase 1 oncology trials are to determine the safety and pharmacokinetic (PK) properties of the agent and to determine a dose for Phase 2 testing (Table 2). For cytotoxic agents, with mechanisms of action that are relatively nonselective in terms of their effects on tumour vs. normal cells, first-in-human Phase 1 trials are usually open only to patients with advanced cancers that have become refractory to standard therapy or for which no standard acceptable therapy exists. Eligibility usually requires good performance status, adequate organ function and absence of concomitant medical illnesses that could confound interpretation of adverse events or place patients at unacceptable risk (Table 3). Because of their inherent toxicity, clinical trials of cytotoxic agents are not conducted in healthy volunteers. For molecularly targeted agents that have minimal or no toxicity in animal studies, first-in-human studies may be conducted in healthy volunteers to provide some preliminary single or limited dosing PK data.

Table 2. Objectives of Phase 1 oncology trials.

• Evaluate safety and tolerance

• Determine dose-limiting toxicity

• Define maximum tolerated dose

• Define optimal biologically active dose

• Determine dose and schedule for initial Phase 2 efficacy trials

• Evaluate pharmacokinetics (ADME*)

• Evaluate effects on molecular target or pathway

• Observe for preliminary evidence of antitumour activity

^*ADME, Absorption, distribution, metabolism, elimination.

Table 3.

Phase 1 studies of cancer drugs patient population

• Involve ∼ 20–60 patients

• Usually patients with advanced cancer after failure of standard treatment options or with cancers for which there is no standard treatment

• Adequate organ function (bone marrow, renal, hepatic)

• Disease-specific Phase 1 trials – targeted agents

• Haematological malignancies if haematological toxicity is dose-limiting toxicity in solid tumour study

• Paediatric patients – initiated only after safety and tolerance evaluated in adults, and there is a potential paediatric indication

• Other special populations (organ impaired, elderly)

The starting dose for Phase 1 trials of cytotoxic oncology drugs is based on preclinical animal toxicology studies along the lines recommended by the US Food and Drug Administration (FDA) [2]. The dose is usually based on one-tenth of the dose (on a body surface area basis) that causes severe toxicity (or death) in 10% of rodents, i.e. severe toxicity dose 10 (STD 10), provided that this dose does not cause serious, irreversible toxicity in a nonrodent species. If irreversible toxicities are produced at the proposed starting dose in nonrodents or if the nonrodent is known to be a more appropriate model, then the starting dose is based on one-sixth of the highest dose tested in nonrodents that does not cause severe, irreversible toxicities. In a standard dose-escalation Phase 1 trial, cohorts of three to six patients are treated at each dose level, following an algorithm based on observed dose-limiting toxicity (DLT) similar to that shown in Table 4. The protocol should clearly define DLT, which is often described as drug-related grade 3 or worse toxicity according to US National Cancer Institute Common Terminology Criteria for Adverse Events (http://ctep.cancer.gov). Grade 4 is commonly used as the definition of DLT for bone marrow suppression (e.g. reduction in haemoglobin, white blood cells and platelets), since these events are usually considered easily manageable. Dose escalation continues until the MTD is reached. The MTD is usually defined as the dose tolerated by five of six patients and the dose level below that which results in unacceptable or DLT in two or more patients treated at that higher dose.

Table 4. Standard Phase 1 dose escalation scheme^*.

No. of patients with DLT at given dose level	Escalation decision rule
0 out of 3	Enter three patients at the next dose level
Two or more	Dose escalation will stop. This dose level will be declared the maximally administered dose (MTD exceeded). Three additional patients will be entered at the next lowest dose level if only three patients were treated previously at that dose
One out of 3	Enter at least three more patients at this dose level
	If none of these three patients experiences DLT, proceed to the next dose level
	If one or more of this group suffer DLT, then dose escalation is stopped and this dose is declared the maximally administered dose. Three additional patients will be entered at the next lowest dose level if only three patients were treated previously at that dose
≤1 out of 6 at highest dose level below the maximally administered dose	This is generally the MTD or recommended Phase 2 dose. At least six patients must be entered at the recommended Phase 2 dose

^*From NCI CTEP Phase 1 protocol template (http://ctep.cancer.gov/guidelines/templates.html). MTD, Maximum tolerated dose; DLT, dose-limiting toxicity.

Phase 1 trials of oncology drugs are initially conducted in adult subjects with solid tumours. Depending on the agent, additional Phase 1 trials in special populations may be conducted. If the DLT in the Phase 1 solid tumour trial is bone marrow suppression, a separate dose escalation trial may need to be conducted in patients with haematological malignancy, particularly acute leukaemia, since a recommended Phase 2 dose may be considerably higher in this population. Given the potential for age-related differences in PK and tolerance and to minimize risk in children, Phase 1 studies in paediatric patients usually do not proceed until data are available from adult trials. The starting dose in paediatric Phase 1 trials is usually no more than 80% of the adult MTD. Paediatric Phase 1 trials should be conducted as soon as possible with agents that have potentially promising activity in a tumour type that is relatively common in children (e.g. leukaemia, neuroblastoma, osteosarcoma, etc.). Studies in patients with organ dysfunction (e.g. hepatic or renal) are considered depending on an agent’s major route of elimination [3, 4]. These studies are usually not conducted until the PK and toxicity profile are established in Phase 1 trials in patients with adequate organ function. Since organ dysfunction trials require a considerable resources and time to complete, they should be conducted only after it is clear that the drug will probably obtain marketing approval. The FDA often requires sponsors to commit to obtaining PK and dosing information in patients with organ dysfunction at the time a New Drug Application (NDA) is reviewed or approved.

Different types of dose-escalation strategies have been utilized in Phase 1 oncology trials (Table 5). Until recently, a modified Fibonacci has been used most frequently. This approach consists of increasing the dose by fixed decreasing increments of dose, in succeeding levels, usually 100%, 67%, 50% and 40%, followed by 33% for all subsequent levels [5]. The modified Fibonacci dose-escalation design has a number of shortcomings, including the following: it requires too many patients and dose levels to reach MTD; too many patients are treated at doses well below a biologically effective level with little or no chance of benefit; and it often takes too much time to complete. In an attempt to decrease the number of patients treated at doses likely to be subtherapeutic, which is especially the case when the starting dose is far below the MTD, several different approaches have been proposed as alternatives to the modified Fibonacci design [5–9].

Table 5. Phase 1 dose escalation strategies.

• Typical modified Fibonacci scheme [5]

Starting dose: 1/10 MELD10

Fixed decreasing dosage increments in sequential cohorts of 3–6 patients

Six patients treated at MTD

• Continuous reassessment method [9]

Each patient is treated with a dose that is as close as possible to the current estimate of MTD (Bayesian approach)

Potentially provides an improved estimate of MTD

• Pharmacologically guided [6]

Utilizes information from preclinical pharmacology studies

Assuming equivalent mouse–man pharmacokinetics, MELD10 approximates MTD

Dose escalation based on target plasma AUC level

• Accelerated titration designs [5]

One patient cohorts and 100% increments until moderate toxicity is observed

Intrapatient dose escalation

Fewer patients treated at ineffective dose levels

MELD, mouse equivalent lethal dose; MTD, maximum tolerated dose.

In the mid 1980s, there was increasing interest in the concept of applying pharmacologically guided dose escalation to the design of Phase 1 clinical trials of investigational anticancer drugs [6]. This approach was based on the hypothesis that there is a direct relationship between toxicity and the plasma levels of the drug, and on the assumption that toxicity (MTD) would be produced at similar plasma levels in humans as in the animals used for preclinical studies [10]. If preclinical results provide information such as the plasma drug concentration corresponding to levels in tumours high enough to inhibit the target, it may be reasonable to increase the dose rapidly in the clinical trial until the target concentrations are approached. Dose escalations in humans could be safely and rationally based on measurement of drug levels in plasma, rather than on the modified Fibonacci or similar empirical escalation schemes. The pharmacologically guided dose-escalation design has not been widely used, perhaps because it requires near real-time PK analysis. However, the concept, if appropriately applied to selective agents, could reduce the number of patients treated at subtherapeutic doses and expedite trial completion.

Another approach that has advantages over the modified Fibonacci scheme is the application of the accelerated titration designs proposed by Simon et al. [5], now widely used for first-in-human Phase 1 oncology trials. Common elements of the accelerated titration designs are an initial accelerated dose-escalation phase, single-patient cohorts during the accelerated phase, intrapatient dose escalation and conversion to a more conservative standard dose-escalation phase after moderate toxicity (i.e. grade 2) is observed (Table 6). Two of the designs use 100% dose increments during the accelerated phase. After moderate toxicity is observed, all designs revert to the standard three-to-six-patient cohorts (Table 4). As such, the accelerated titration designs should substantially reduce the number of patients treated at subtherapeutic doses without compromising safety.

Table 6. Accelerated titration designs [5]– summary of key elements.

One-patient cohorts and 100% increments during the initial accelerated phase

Converts to 3–6 patient cohorts and 40% increments after moderate toxicity is observed

Allows for intrapatient dose escalation

Could reduce average number of patients per trial by up to 50%

Should substantially reduce the number of patients who are treated at subtherapeutic doses

Recently, the discovery of a plethora of target-based antineoplastic compounds has created major challenges in drug development. In animal models, many of these newer agents exhibit antimetastatic or growth-inhibitory effects as opposed to inducing rapid tumour regression. Thus, these compounds are commonly referred to as having cytostatic effects. Although the specific effects of each class are dependent on the target or targets inhibited, many are predicted to reduce tumour growth without producing cytotoxicity. As molecularly targeted agents (e.g. inhibitors of metalloprotease, protein kinase C and tyrosine kinase receptors) act on highly specific targets which are differentially expressed or activated in cancer cells and may result in low normal tissue toxicity, they may have a very wide therapeutic ratio. Therefore, increasing the dose in a traditional Phase 1 trial to normal tissue tolerance (MTD) may be an irrelevant end-point [11]. As such, a different approach than the usual Phase 1 trial design is required, since the standard primary end-points of toxicity may not apply [11–13]. Rather, the end-points of primary interest for dose determination are target or pathway inhibition.

For targeted agents, the primary end-point of the Phase 1 trial should be to determine the ‘biologically active dose’ required to inhibit maximally the relevant target or pathway. However, in order to define a biologically effective dose it is essential to have a reliable, validated assay to evaluate for target inhibition. There has to be understanding of the expected effects of inhibition, whether target inhibition leads to tumour regression, whether the target needs to be inhibited continuously or intermittently to inhibit tumour growth and to identify the function of the target in normal cells. Thus, the optimal development of these agents in the clinic requires extensive preclinical studies, including development of better animal models that are predictive of eventual clinical effect and thorough interrogation of the assay in animal models. Following this, the drug and its effect on tumour (as evaluated by the assay) should then, preferably, be evaluated in first-in-human pilot studies prior to definitive Phase 1 and 2 trials [14]. Some of these pilot trials could be performed much earlier than possible for definitive Phase 1 trials. If the pilot study involves a limited number of subjects and very limited drug exposure with no therapeutic intent, less preclinical toxicity data may be required in accordance with the recently issued FDA guidance for Exploratory Investigational New Drug (IND) studies (http://www.fda.gov/cder/guidance/7086fnl.htm). Examples of studies that can potentially be conducted under an Exploratory IND are those that provide important PK information (e.g. oral bioavailability), those designed to select the most promising lead agent from several candidates and those that determine whether a mechanism of action (e.g. target inhibition) observed in animal models can be observed in humans.

Assessing target inhibition in tumour tissues in clinical trials raises logistic and patient safety considerations. Therefore, in lieu of pre- and post-treatment tumour biopsies, one strategy to determine target modulation and the biologically active dose is to exploit a less invasive evaluation of the target in a surrogate tissue or by using a relevant functional imaging technique. For example, demonstrating a ‘biologically effective dose’ in peripheral blood mononuclear cells (PBMCs) may serve as a guide in dose-finding protocols. However, a dose that inhibits a target in PBMCs may not provide sufficient exposure to inhibit the target to the same degree in tumour tissue [15]. Thus, before PBMCs can be used as a surrogate for target effect in tumour tissue, validation studies comparing effects in PBMCs vs. tumour tissue should first be performed in animal models. Subsequent pilot correlative studies in the initial Phase 1 trial or in a pre-Phase 1 pilot study would then be appropriate prior to initiating definitive Phase 1 or 2 trials.

Another strategy to define the optimal biologically effective dose is to incorporate innovative molecular and functional imaging, such as magnetic resonance imaging (MRI) and positron emission tomography (PET), into early phase dose-finding trials of targeted agents [16, 17]. For example, digital contrast-enhanced MRI (DCE-MRI), which can measure tumour permeability and vascularity, may be a useful biomarker for dose determination in studies of angiogenesis inhibitors [18, 19].

Given the growing number of targeted agents in development and the complexities of determining a dose that yields optimal biological activity based on target inhibition rather than on toxicity, there is an increasing need for novel Phase 1 trial statistical designs. Integral to the process of choosing among the trial designs is an understanding of what is known preclinically about the agents and the patient population being studied. Recently, Hunsberger et al. [20], proposed Phase I designs for molecularly targeted agents based on the assumption that the target response is a binary value (i.e. positive or negative) determined in each patient. The goal of the proposed design is to find a biologically adequate (rather than optimal) dose using a minimum number of patients, with the intent to obtain sufficient data to move as quickly as possible from dose-finding to evaluating efficacy. Based on limited simulations, this design appears to perform adequately with only three to four patients treated at each dose level. The utility of the approach warrants prospective evaluation in future clinical dose-finding trials of a variety of targeted agents [20].

Phase 2 efficacy trials

Measures of efficacy

With the increasing number of anticancer agents being developed with novel mechanisms of action, there is an urgent need to change the paradigm for demonstrating anticancer agent efficacy. End-points, such as objective response rates, that were considered standard for most Phase 2 trials with traditional agents may not be adequate for targeted agents that produce growth inhibition without tumour regression.

The various parameters used to evaluate efficacy and clinical benefit in cancer clinical trials include objective tumour response, response rate (RR), time to (tumour) progression (TTP), progression-free survival (PFS), overall survival (OS), time to treatment failure (TTF) and a composite clinical benefit end-point [21].

OS is the most clinically relevant and meaningful end-point in trials of patients with malignant disease. It is easy to record, reliable and free of bias. However, there are disadvantages in requiring OS as a measure of efficacy, since the length of time to complete the study is usually prolonged due to the long-term nature of the end-point and there may be confounding variables such as those related to cross-over designs and patients receiving other therapies at the time of progression [22]. Also, OS is not a feasible end-point for Phase 2 trials, which are generally small single-arm studies primarily aimed at providing preliminary evidence of efficacy.

Tumour shrinkage provides the clearest evidence of pharmacological drug effect and objective tumour response is often used as a surrogate end-point for other measures of clinical benefit such as OS and palliation of symptoms. This is also utilized to guide decisions regarding continuation of current therapy in a given patient. Objective tumour response can be easily measured, usually by imaging studies and the use of standardized criteria that have been developed for clinical trials, including the World Health Organization (WHO) response criteria and the Response Evaluation Criteria in Solid Tumors (RECIST) [23, 24]. The more commonly used RECIST criteria define objective tumour response as a 30% reduction from baseline in the sum of the longest diameters of all target lesions. Tumour progression is defined as a 20% increase compared with baseline or best response or a new lesion.

RR has been the basis for FDA approval of a number of chemotherapeutic agents for a variety of indications. Between 1990 and 2002, RR or TTP formed the basis for 27 of the 57 approved marketing applications for oncology drugs [25]. However, objective tumour response does not always correlate with clinical benefit and improvement in OS. Also, since newer targeted agents may not necessarily cause tumour shrinkage, the traditional criteria of objective tumour response may not be adequate in assessing efficacy in exploratory trials.

PFS (duration of time alive without progression) and TTP are measures of efficacy used in cancer clinical trials that do not depend on tumour shrinkage. These end-points have limited utility outside of randomized controlled, preferably blinded, Phase 3 trials. The drawbacks of using PFS or TTP in Phase 2 studies are that their clinical relevance varies based on the underlying disease; and they are both influenced by the frequency of follow-up evaluations and subject to potential bias. However, for evaluating cytostatic agents, inhibition of progression may be the most relevant clinical end-point, as these drugs may cause disease stabilization without tumour shrinkage.

Another related clinical end-point is progression-free rate (PFR), which is a binary outcome (i.e. progression or no progression) at a specified time from treatment initiation. This end-point may be preferred over PFS or TTP for exploratory Phase 2 efficacy trials, since with the latter two end-points there is greater potential for bias in determining progression times [26]. This is particularly the case when conducting randomized unblinded Phase 2 comparator studies, because of a tendency for more frequent response assessments in patients assigned to a placebo or a control treatment arm conjectured to be inferior. In contrast, using PFR at a predetermined time from initiation of study treatment as the primary end-point minimizes bias related to the timing of the evaluations.

Cytotoxic vs. cytostatic agents

Cytotoxic agents are drugs that result in cell kill and eventual tumour shrinkage, whereas cytostatic agents inhibit tumour growth without direct cytotoxicity. Molecularly targeted agents may be cytostatic or cytotoxic. Due to their specificity and growth-inhibitory properties, different approaches may need to be applied in the clinical development of these agents from those commonly used for cytotoxic chemotherapy (Table 7). Traditionally, the dose recommended for Phase 2 trials of cytotoxic agents is based on the MTD determined in Phase 1. The inherent assumption in establishing and using an MTD is that the therapeutic effect and the associated toxicities are correlated and that the mechanism of action of both the toxic and therapeutic effects is the same – higher doses result in greater efficacy. Also, as noted previously, Phase 2 trials of cytotoxics commonly use objective RR as a measure of activity warranting further clinical evaluation. Furthermore, evidence of a therapeutic response to cytotoxic drugs is expected to occur relatively early in the course of treatment (e.g. after two cycles of therapy).

Table 7. Development of cytotoxic vs. cytostatic molecularly targeted agents.

	Conventional chemotherapy	Molecularly targeted agents
Cellular effects	Cytotoxic	May be cytostatic
Toxicity	Usually nonspecific multiple organ system; often bone marrow, gastrointestinal, hepatic	Presumably less toxic; target specific
Phase 1 primary end-points	Characterize toxicities; determine DLT and MTD; evaluate PK	Determine target inhibition, biologically active and optimal doses; evaluate PK
Patient selection	Disease histology	Molecular pathology or presence of target(s)
Phase 2 efficacy trial end-points	Objective tumour response (tumour shrinkage)	Objective tumour response or stabilization (progression-free rate)
Measures of efficacy	Anatomical imaging	Anatomical or functional imaging
Time to clinical response	Relatively short (e.g. 6–8 weeks)	Relatively late; may require prolonged dosing for therapeutic effect

The assumptions that have guided these clinical trial methodologies for cytotoxic agents may not be appropriate in the era of targeted therapies [11, 12, 22, 27]. Targeted drugs have a wider therapeutic window; the mechanism of their therapeutic effects may differ from their toxic effects; they may require dosing for longer periods at low doses to be effective; they may have maximum clinical effects in combination rather than as single agents; and they may be cytostatic instead of cytotoxic. Since cytostatic drugs may not necessarily cause tumour shrinkage, the traditional criteria of objective tumour response and RR may not be optimal in assessing efficacy [22]. Also, more than one dose or schedule may need to be evaluated in Phase 2 trials, especially in combination studies, depending on the desired biological or pharmacodynamic (PD) effects.

For molecularly targeted agents, particularly cytostatics, trial designs for exploratory efficacy studies may need to rely more on biological/PD end-points as markers or surrogates for antitumour effect than on toxicity or objective tumour response. The development of mechanism-based biomarker assays would significantly expedite drug development, since such an assay would aid in determining early on if a drug ‘hits’ the target. This approach could also help select the lead agent from a group of compounds, help determine dose, guide patient selection and assessment of response and provide a basis for combination trials [28, 29]. However, this requires having a validated target, evaluating for the presence of the target, understanding the biology of target inhibition as well as any nontargeted biological effects mediated by the study agent and defining ‘optimal target inhibition’. However, the issue of evaluating for the presence of target is demanding because it requires having a reliable assay to measure the target, as exemplified by the comparative value of its measurement of Her2neu by immunohistochemistry vs. fluoresence in situ hybridization prior to trastuzumab therapy [30].

To evaluate adequately PD end-points in tumour tissue, exploratory trials of molecularly targeted agents will need increased numbers of biopsies, which can potentially impair enrollment and raise patient safety concerns. Also, such trials require careful attention to the use of standardized and validated procedures for tissue acquisition, handling and assay methodology. This is critical, as biological processes may be altered over time, which could significantly increase inter- and intrapatient variability, making the interpretation of the results problematic. Also, different tissue-handling procedures may change the expression or level of activity of targets, e.g. enzymes, protein substrates, etc. To minimize the need for visceral tumour biopsies in subsequent trials, early exploratory studies should evaluate effects in other tissues such as skin or PBMCs as potential surrogates. Exploratory efficacy trials should also utilize newer techniques such as molecular profiling of the tumour and normal tissue, to have a better understanding of the effects of inhibiting the target in tumour and host tissue. These approaches depend heavily on the availability of clinical and laboratory-based investigators and their willingness to collaborate.

One of the other ways to optimize trial design is to evaluate for some early evidence of treatment effect after an initial course of therapy and then to continue therapy only in those with a positive effect. Such a Phase 2 design would minimize the number of patients receiving ineffective experimental treatment and improve the efficiency of the trial while individualizing therapy. This type of enrichment study design relies heavily on a treatment effect that is predictive of a favourable outcome. A good illustration of this approach is the use of DCE-MRI for blood flow measurements in patients receiving antiangiogenesis drugs [18, 19, 31]. MRI can potentially be used to evaluate relationships between PK and PD parameters in early clinical trials and to identify whether a particular drug is affecting the target or not, which could improve the efficiency and minimize the cost of drug development [14, 32]. Another emerging and promising area of mounting interest is the use of PET and functional imaging probes to visualize molecular targets and processes in cancer [28]. However, factors such as interpatient variability and tumour heterogeneity along with availability of resources need to be better examined in exploratory trials prior to wider acceptance and use of imaging as biomarker for clinical benefit.

Optimizing patient selection is an important consideration for enrollment on efficacy trials with molecularly targeted therapies. For molecularly targeted agents, including multitargeted agents, patient selection should be based primarily on the presence of the target in the tumour. To date, Phase 2 trials have focused on evaluating a drug in a given histological type, based on the presumption that a given histological type represents a uniform single disease entity. However, there is increasing evidence of tumour molecular heterogeneity within a given tumour type. Genetic variations in the tumour may significantly affect drug sensitivity/resistance. This is exemplified by the unique responsiveness of nonsmall cell lung cancer patients who have a mutation in exon 19–21 of the epidermal growth factor (EGFR) receptor gene to the EGFR receptor inhibitor gefitinib [33]. Since these mutations are observed in 8% of patients not previously exposed to gefitinib, the potential benefit of gefitinib in this small patient population could be potentially masked in a group of unselected nonsmall cell lung cancer patients, unless an extraordinary number of patients are treated [34]. Similarly, exon 11 mutations of c-kit in gastrointestinal stromal tumour [35] and T315I bcr/abl mutations in chronic myelogenous leukaemia [36] confer sensitivity and resistance to imatinib, respectively. Thus, in designing Phase 2 clinical trials for molecularly targeted agents, it will be increasingly important to address the issue of molecular heterogeneity and its effect on drug efficacy [37, 38]. For example, to maximize the chance of success in nonsmall cell lung cancer trials of EGFR tyrosine kinase inhibitors, enrollment could be restricted to patients whose tumours are EGFR+ by immunohistochemistry (IHC), those with increased numbers of copies of the EGFR gene, or the presence of EGFR mutation [38].

Other than molecular heterogeneity, differences in tumour growth rates, both within a patient and between patients (e.g. renal cell cancer), make evaluation of PFS or PFR, the primary end-point of many of the trials with cytostatic agents, difficult to interpret, particularly in relatively small exploratory studies. One trial design that has been proposed to overcome some of these issues is the Randomized Discontinuation Design (RDD), which is also an example of a method to enrich a specific patient group [39, 40]. With this design, all eligible patients are treated with a study agent for a defined period of time; then, only those patients who show evidence of stable disease are randomized to receive either the study drug or placebo. The potential to ‘enrich’ the sensitive patient population exists because only those patients without early drug failure, who are likely to respond, are continued on therapy. The application of the RDD or other types of novel screening designs in exploratory Phase 2 trials may be useful in the early development of a molecularly targeted cytostatic agent for which a dependable assay to select patients based on target expression is not available [26, 40].

Importance of drug interactions

As with medications in general, the potential for drug interactions is a key concern in oncological drug development [41, 42]. A detailed history of concurrent medications, including herbal medications, is critical in analysing efficacy and toxicity data from clinical trials (e.g. capecitabine and warfarin; irinotecan and St John’s Wort) [43, 44]. Depending on the interaction, exposure to active drug may be either increased or decreased, leading to enhanced toxicity or reduced effectiveness. Since cytotoxic drugs in general have a narrow therapeutic window, as well as complex pharmacological profiles, some interactions can have serious clinical consequences [45]. Drug interactions may be pharmaceutical, pharmacokinetic (absorption, distribution, metabolism and excretion) or pharmacodynamic. Examples of pharmaceutical interactions relate to the effects of a vehicle on drug delivery and activity (e.g. encapsulation of doxorubicin; development of nanosuspensions or submicron colloidal dispersions of pure drug particles) [46]. Thus, early clinical trials evaluating different formulations of the same drug should incorporate detailed PK and PD studies, as well as focusing on the biochemical mechanisms of drug action.

Oral administration has become the preferred route for small-molecule cytostatic agents to accommodate the continuous daily dosing that is required to achieve the prolonged exposures necessary for optimal growth-inhibitory effects. With an increase in the development of orally active chemotherapeutic agents, there is a heightened awareness that the administration of other agents may significantly affect bioavailability and inter- and intrapatient variability in PK. Cytochrome p450 (CYP450) isoenzymes and drug transporters in the intestinal epithelium can significantly influence drug uptake [41, 42]. For example, inhibition of P-glycoprotein in the gut epithelium by ciclosporin A can substantially increase the oral bioavailability of paclitaxel, a combination that has demonstrated clinical activity. Therefore, recognizing these issues is critical in drug development as they may allow identification of mechanisms of resistance and optimization of lead compounds.

When anticancer drugs that are substrates for CYP450 are given in combination with other agents that induce the production of the same CYP450 isoenzymes, the metabolism of the anticancer drug can be markedly enhanced. This is the case when irinotecan, imatinib, tipifarnib and other anticancer drugs are administered to brain tumour patients who are concomitantly receiving enzyme-inducing antiepileptic medications (EIAEDs). Patients on EIAEDs must receive substantially higher doses of these anticancer agents than patients not on EIAEDs to achieve comparable levels of exposure [47–50].

Other than the simple coadministration of drugs resulting in interaction, there are examples of drug sequencing that can affect both efficacy and toxicity of a regimen (e.g. carboplatin and docetaxel in the treatment of lung cancer) [51]. This raises the issue of evaluating different sequences of drug delivery in clinical trials of combinations of various agents. As most anticancer therapies are likely to be used in combination with other targeted agents or conventional chemotherapies, drug sequencing is a critical issue; early drug development trial designs should consider evaluating various schedules and sequences to establish the optimal recommendations for further evaluation in Phase 3 trials.

Pharmacogenomics

Another major challenge in drug development is elucidating as early as possible the role of pharmacogenomics in an agent’s PK and PD. Genetic polymorphisms in the individual’s genome may significantly affect drug disposition and tolerance [52–54]. Inherited polymorphisms in the drug-metabolizing enzymes involved in the biotransformation of anticancer agents can be a key factor with respect to toxicity, including the following enzyme systems: CYP450, UDP-glucuronosyltransferase (UGT), N-acetyl transferase (NAT1 and NAT2), glutathione S-transferase, sulfotransferase and those involved in purine/pyrimidine metabolism [52, 53, 55–57]. For example, genetic polymorphisms in the UGT1A1 gene are associated with increase risk of adverse reactions to irinotecan [53–56, 58].

Pharmacogenomics is becoming an increasing part of drug development and forming the basis for moving towards ‘individualized medicine’[53, 54]. With advances in gene expression array technology, genotyping and informatics will continue to contribute to the design of early clinical trials. Furthermore, if interrogated early in the drug development process, pharmacogenomics could help avoid premature abandonment of otherwise promising agents.

Conclusions and future directions

Approaches to early drug development in oncology are changing with the emergence of a growing number of molecularly targeted agents. The strategy for dose determination has shifted from a primary focus on toxicity to one of identifying a dose that optimally inhibits the target. Consequently, it is imperative that the assay for target modulation, both in tumour and surrogate tissue, be thoroughly interrogated in animal models and then, preferably, in a pilot clinical trial involving a limited number of patients prior to conducting a larger, more definitive Phase 1 trial. These pilot studies should focus on assay methodology and validation, as well as procedures for tissue acquisition, handling and processing. Early trials could also be used to optimize agent selection among several potentially promising analogues and provide PK and PD data to improve the design and efficiency of definitive Phase 1–2 trials. Some of these pilot trials could be conducted with less than the usual amount of preclinical toxicity data under the purview of an ‘exploratory IND’ as proposed in the recently issued FDA guidance (http://www.fda.gov/cder/guidance/7086fnl.htm).

The emergence of molecularly targeted agents also presents challenges for the development of efficacy trials. There is an increasing need for novel trial designs to evaluate these agents, since objective tumour response, the primary end-point typically used in Phase 2 trials of standard chemotherapeutic drugs, may not be adequate for many of the molecularly targeted agents that inhibit tumour growth without measurable tumour shrinkage. Consequently, it is imperative to consider carefully the mechanism of antitumour effect in the design of exploratory efficacy trials. We also need to improve our understanding of the role of pharmacogenomics in individualizing treatment and to apply this knowledge to early drug development.

There is a growing need to improve the efficiency of developing new investigational drugs. A new medicinal compound entering Phase 1 testing for any indication is estimated to have only about an 8% chance of reaching the commercial market and the success rate has not improved substantially during the past two decades, suggesting that recent biomedical breakthroughs have not improved the ability to identify successful candidates [59]. The success rate for oncology agents is equally disappointing [60]. As proposed by Von Hoff [60], many investigational anticancer agents fail because clinical trials were designed and conducted without sufficient attention to the agent’s attributes and mechanism of action. Clearly, to improve success in oncological drug development, there need to be more efficient and rational clinical approaches and close collaboration between laboratory and clinical scientists.