|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Originally published as JCO Early Release 10.1200/JCO.2009.23.6703 on August 24 2009 © 2009 American Society of Clinical Oncology.
Phase I Drug Combination Trial Design: Walking the TightropeDepartment of Medical Oncology, Erasmus University Medical Center, Daniel den Hoed Cancer Center, Rotterdam, the Netherlands Drug combination strategies are used in the palliative and curative settings for patients with cancer. Before a specific drug is incorporated into such a combination treatment strategy, it has to be explored using the clinical drug development route. The first critical step is the design of a combination phase I trial. As for all phase I trials, the design of a phase I drug combination trial is of utmost importance because it will dictate the course of the trial and, by its nature, it already incorporates, implicitly or explicitly, a short list of the potential answers to the primary end point question. While many consider phase I trials of drug combinations uninteresting and while there has not been much debate on their design, in reality, the complexity of the design of a phase I trial increases exponentially with the number of different drugs and treatment modalities included in the treatment strategy. If the treatment strategy to which the new drug is added already has curative intent, this adds a complicating dimension to the design of the trial. It means that the researchers will have to balance determining the feasible dose of the added experimental agent without compromising cure. In this issue of Journal of Clinical Oncology, Haddad et al1 report a phase I study of cetuximab/docetaxel/cisplatin/fluorouracil given as induction treatment to patients with locally advanced squamous cell carcinoma of the head and neck. Taking all the aforementioned factors into account, they should be applauded for performing a dose-finding study in the curative setting. Yet, the study raises some phase I methodology questions. In a series of theoretical phase I trials of drug A and drug B, considering only one dosing regimen per drug and disregarding all available drug data except for the maximum tolerated dose (MTD) and recommended phase II dose (RP2D) for each drug given as a single agent, as a result of the dose escalation track used, different theoretical MTDs and RP2Ds can be defined. For instance, a fixed dose of drug A (at its MTD) can be used while escalating drug B until the MTD of this combination has been determined. One should bear in mind that the MTD of drug B in the combination is not necessarily determined by its single-agent MTD; it could be higher than the single-agent MTD, perhaps because of drug-drug interactions with drug A. A similar trial could be designed for a fixed dose of drug B and dose escalation of drug A, potentially rendering a different MTD of the combination. A different approach would be a dose escalation in which both drugs are started at a relatively low dose and increased alternately. Depending on the escalation steps, different MTDs can be determined. Because the escalation steps can go back and forth between the two drugs whenever an MTD is determined, at least two MTDs can be determined. For example, the MTD is exceeded at dose level 4 for both drugs. If the previous dose level (drug A at dose level 4 and drug B at dose level 3) was safe, that can be called the MTD. However, one could also argue that the combination of drug A at dose level 3 and drug B at dose level 4 should be evaluated for feasibility and, if it is considered safe, that can also be called the MTD. As illustrated in this theoretical example in a simple combination phase I trial with only two drugs, the number of dose-escalating options to approach the MTD easily reaches four, potentially rendering four different MTDs. By adding drug C, this theoretical number rapidly increases to at least 18, and one does not want to think about the potential number of MTDs that are definable with four drugs, as used in the study by Haddad et al.1 Thus, one of the basic dilemmas facing investigators who design combination phase I trials is choosing the dose escalation approach that will determine the most biologically optimal and useful MTD. From a purely phase I methodology point of view, there is no right or wrong dose escalation approach, because at the end of the study, there will be an answer: one of the many possible MTDs. However, from a treatment practice point of view (even more so when there is curative intent), one should try to identify the dose escalation approach that is best supported by conceptual and actual data and will lead to dose recommendations that, after subsequent phase II and III studies, stand the best chance of changing practice by improved outcome. Guidance can be found in data on toxicity patterns, trials on comparable agents, and (if more than two drugs are used) in previously defined MTDs of the specific combinations to which a new drug is added. Guidance should not be sought in maximizing potential sales of the investigated drug or in beliefs not supported by evidence. As noted, the choice of the dose escalation approach should be determined by the best possible scientific and clinical practice rationale. Haddad et al1 elected to determine the MTD with a fixed dose of cetuximab (C), cisplatin (P), and docetaxel (D) while escalating the dose of fluorouracil (F). This approach is somewhat remarkable, because the same group of investigators reported that the current standard of induction chemotherapy in patients with squamous cell carcinoma of the head and neck consists of P + F with or without D.2 The authors' rationale for choosing their dose escalation approach is the potential overlap in the GI toxicity of C and F, but it is unclear why they compromised on the dose of the standard drug (F) and not on the dose of the investigational agent (C) in this regimen. Of note, the phase III EXTREME trial, which had already been closed for accrual and reported before the first patient was enrolled in the current phase I trial, refuted the argument of overlapping toxicity and showed that the addition of full-dose C to full-dose P + F is feasible without an increase in grade 3 to 4 toxicity in general or in GI toxicity specifically.3 In addition, using a fixed dose of the new drug (C) at its single-agent MTD in this induction treatment strategy seems to suggest that investigators were convinced that this dose and drug should be part of the MTD of the combination, no matter what. All of this leaves some level of concern, bearing in mind that the MTD from this study most likely will be the only MTD ever determined for the combination, because there may never be other phase I studies on this particular combination of drugs that identify other MTDs. Another approach could have used fixed-dose P + F (to guarantee the best currently known curative intent) combined with low doses of D and C and alternately escalating D and C. This would have addressed the problem of febrile neutropenia that the authors had identified beforehand. An even more attractive but more complex option would be to incorporate a more-than-one-dose escalation approach in these studies to define multiple MTDs in a single study.4 A crucial step in any combination phase I study design is to attempt to rationally predetermine the events contributing to the defined dose-limiting toxicity (DLT), because these criteria will determine when dose escalation will be halted and will define the MTD of that specific approach. The incidence and severity of toxicity considered acceptable have to be defined in detail and (most importantly) balanced against the overall incidence of toxicity of the standard drugs in the combination.
The classic 3 + 3 phase I design usually implies that an incidence of severe toxicity less than 33.3% is considered acceptable. Importantly, the relevant part of the previous sentence is not "less than 33.3%," because it gives a false sense of security. Far more important is the word "severe" and its definition. In a combination drug setting, and even more so in a curative setting, as in the trial done by Haddad et al1, the investigators will want to avoid coincidental interference of background toxicity from the standard combination with dose escalation, resulting in premature halting of dose escalation. Say, for example, the unknown but true incidence of severe toxicity of a drug combination is 25%. By using a classic 3 + 3 phase I design, there is a 40% chance (equation 1) that at least two DLTs are observed within one cohort that exceed the 33.3% predetermined limit, thus falsely halting further dose escalation.
There is no golden rule that successfully excludes coincidence in dose escalation decisions. One of the most obvious solutions is to enroll more patients at a specific dose level; in our opinion, it is worthwhile to consider the 3 + 3 + 3 design in combination phase I trials if the expected incidence of a specific toxicity is rather high because of one of the elements of the combination. The only difference between the 3 + 3 + 3 and the classic 3 + 3 design would be a further expansion whenever two DLTs are observed in a cohort of six patients. Applying this to the theoretical model, in which the unknown but true incidence of severe toxicity of a drug combination is 25%, will decrease the chance of a premature interruption from 40% to 30%. As mentioned before, the setting of induction therapy in the context of a curative intention strategy creates a further challenge to the phase I trial design. Curative intent should never be lost, regardless of drug doses per cycle; one should also predetermine the DLT time frame. In a standard single-agent phase I trial of a new drug, usually only the first treatment cycle is used to evaluate DLTs. In the trial designs discussed in this editorial, this time frame should most likely be longer because subsequent cycles could yield a necessity of dose delay. If the MTD defined by using only the first treatment cycle is applied to a treatment strategy using three cycles of induction systemic therapy before local treatment, cycle delays due to toxicity, in theory, could lead to considerable delay in starting local treatment. Because both systemic and local therapy are integral parts of the curative strategy, delaying local therapy may be considered unacceptable. This will definitely have to be taken into consideration when constructing DLT rules for these specific phase I trials. In case only first-cycle DLTs are considered, only a single-cycle RP2D can be provided, but not a multiple-cycle recommended phase II strategy (RP2S). In the study by Haddad et al,1 DLTs were assessed only in the first cycle. This resulted in an interesting construction: an MTD was defined that can serve as an RP2D, but a different dose is recommended for further testing. The latter recommendation was based on the achieved cetuximab dose delivery over multiple courses in this particular study. Although investigators should keep a close eye on achieved dose delivery, we need to realize that the recommended schedule has not been tested from day 1 onward, and it comes from a highly selected patient population. It is not inconceivable that in a less selected phase II or III study population the prospectively untested RP2S will not work. There also seems to be some ambiguity in the recommendation (that in the end includes a 33% reduction in cetuximab dose intensity compared with the MTD) since the investigators deliberately elected in their study protocol to apply cetuximab in all dose levels at its full single-agent dose. Finally, we should caution against jumping to conclusions in phase I trials that are not justified by the trial methodology itself. Phase I trials usually enroll a highly selected population and this likely has a major impact on dose intensity achieved and treatment activity observed. This selection phenomenon cannot be excluded in the phase I trial by Haddad et al,1 as illustrated by the considerable difference in performance status of the study population compared with that in the pivotal phase III trial.5 In the latter trial, 54% of patients had a WHO performance status of zero, while in the current phase I trial, 100% of patients had a performance status of zero.1,5 The reported response rate of 100% in the phase I trial (in contrast to 72% in the pivotal trial by using P + F + D) seems impressive but should not be overestimated, given patient selection issues. We caution against stressing reported activity in phase I trials too extensively, because readers who are not familiar with phase I trials might incorrectly interpret these data. It is well known that response rates decline from phase II to phase III drug development studies using the same treatment.6 The same is likely to be true for phase IB studies, as in the Haddad et al trial. And how should we move forward with current data? Should the MTD found in the phase I trial be used in succeeding trials or should the untested RP2D or RP2S be used? Estimating the feasibility of the latter may be preferred. Given the fact that this dose has not yet been studied prospectively, only a small study should be considered. In designing such a study, the subsequent phase III trial designs should be kept in mind, as should the fact that the therapy has curative intent. Combination phase I trials have long been considered uninteresting, easy, dose-finding studies. But in reality, their designs pose more challenges than a single-agent phase I trial. Dose escalation approaches, predetermined DLT criteria, and assessing DLT only in cycle 1 (as opposed to the full treatment period) all influence outcome and may determine whether the most rational RP2S will be found. Trial design in early drug development requires a delicate balance. Randomly selecting dose escalation approaches or DLT criteria may be hazardous. Every step should have a sound basis. A small misstep in the design can have large consequences, because it is far easier to drop from the tightrope than to reach the platform. AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST The author(s) indicated no potential conflicts of interest. AUTHOR CONTRIBUTIONS Conception and design: Paul Hamberg, Jaap Verweij Collection and assembly of data: Paul Hamberg, Jaap Verweij Data analysis and interpretation: Paul Hamberg, Jaap Verweij Manuscript writing: Paul Hamberg, Jaap Verweij Final approval of manuscript: Paul Hamberg, Jaap Verweij REFERENCES 1. Haddad RI, Tishler R, Norris C, et al: Phase I study of C-TPF in patients with locally advanced squamous cell carcinoma of the head and neck. J Clin Oncol 27:4448–4453, 2009.[CrossRef][Medline] 2. Haddad RI, Shin DM: Recent advances in head and neck cancer. N Engl J Med 359:1143–1154, 2008. 3. Vermorken JB, Mesia R, Rivera F, et al: Platinum-based chemotherapy plus cetuximab in head and neck cancer. N Engl J Med 359:1116–1127, 2008. 4. de Jonge MJA, Loos WJ, Gelderblom H, et al: Phase I pharmacologic study of oral topotecan and intravenous cisplatin: Sequence-dependent hematologic side effects. J Clin Oncol 18:2104–2115, 2000. 5. Posner MR, Hershock DM, Blajman CR, et al: Cisplatin and fluorouracil alone or with docetaxel in head and neck cancer. N Engl J Med 357:1705–1715, 2007. 6. Zia MI, Siu LL, Pond GR, et al: Comparison of outcomes of phase II studies and subsequent randomized control studies using identical chemotherapeutic regimens. J Clin Oncol 23:6982–6991, 2005.
Related Article
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2009 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
