|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2001 American Society for Clinical Oncology
Clinical Trial Designs for Cytostatic Agents: Are New Approaches Needed?From the Cancer Therapy Evaluation Program, Division of Cancer Treatment and Diagnosis, National Cancer Institute, Bethesda, MD. Address reprint requests to Edward L. Korn, PhD, Biometric Research Branch, EPN-739, National Cancer Institute, Bethesda, MD 20892; email korne{at}ctep.nci.nih.gov
ABSTRACT: Preclinical data suggest that some new anticancer agents directed at novel targets demonstrate tumor growth inhibition but not tumor shrinkage. Such cytostatic agents may offer clinical benefits for patients in the absence of tumor shrinkage. In addition, lower doses of some of these agents may be just as effective as higher doses, implying that toxicity may not be an ideal end point for dose finding. Because of these factors, the sequence and design of traditional phase I, II, and III trials used for cytotoxic agents (which typically shrink tumors and in a dose-dependent manner) may not be appropriate for cytostatic agents. This article discusses options for modifying trial designs to accommodate cytostatic agents. Examples are given where these options have been tried or are currently being tried. Recommendations given for choosing among the trial designs depend on what is known preclinically about the agents (eg, does one have a validated and reproducible biologic end point that can be used to guide a dose escalation?), what is known about the patient population being studied (eg, does one have a well-documented historical progression-free survival rate at 1 year for comparison with the experience of the new agent?), and the numbers of agents and patients available for participation in trials. Planned and ongoing trials will test the utility of some of these new approaches.
THE USUAL CLINICAL development of a cytotoxic chemotherapeutic agent involves phase I, II, and III trial designs and is based on the prediction that (1) the agent will shrink tumors, (2) more of the agent will shrink tumors better (if toxicity is acceptable), and (3) tumor shrinkage will lead to potential benefit in terms of lengthened survival and/or improved quality of life. In contrast, cytostatic agents may slow or stop the growth of tumors and the development of metastases, without shrinking existing tumors. In addition, the mechanism of action of some of these agents may be such that higher doses of the agents beyond a certain level may offer no additional benefit. Concerns have been expressed that these differences between cytotoxic and cytostatic agents may lead to rejecting a clinically useful cytostatic agent because it is tested using standard cytotoxic trial designs.1-4 This problem is of increasing importance because advances in molecular biology, cancer genetics, and technology are making available an increasing number of novel agents directed at targets that seem important for cancer initiation, angiogenesis, invasion, or metastasis, for example. Though some of these agents do not shrink tumors, they do inhibit tumor growth and/or decrease the number of metastatic lesions in animal models. Consequently, such agents might prove to be useful anticancer agents. Those involved in cancer drug development have struggled with the problem of developing clinical trial designs that can efficiently screen cytostatic agents.1-4 Ultimately, the clinical value of any new agent must be documented in a randomized phase III trial. However, even if phase III testing without any preliminary (phase II) efficacy evidence was an ethical and efficient approach, available patient and financial resources are insufficient to test all new cytostatic agents this way. Therefore, designs that allow early identification and discontinuation of the development of ineffective agents, without incorrectly discarding promising agents, will permit resources to be focused on more promising agents. We note that if an agent shrinks tumors (and in a dose-dependent manner), then standard cytotoxic trial designs could be used regardless of the actual biologic mechanism of action. On the other hand, if an agent does not shrink tumors but slows or stops the growth of tumors and the development of metastases (possibly with higher doses not necessarily leading to more effect), then standard cytotoxic trial designs may not be effective in identifying clinically useful agents. For ease of discussion in this article, we shall refer to these latter agents as cytostatic, regardless of their mechanism of action. This article describes options for how the usual phase I/II/III trial designs can be modified to accommodate evaluation of cytostatic agents. The options discussed and our recommendations provided at the end of the article are not meant to be definitive but are provided for consideration by those who are also contending with the challenges provided by some of the novel agents already undergoing clinical development.
The purpose of a phase I trial is to find the dose of a new agent that should be used for further testing (recommended phase II dose). For a cytotoxic agent, this is usually defined as the highest dose that can be administered with acceptable toxicity, referred to here as the maximum-tolerated dose. In theory, the maximum-tolerated dose of a cytostatic agent may not be necessary to achieve the maximum or near-maximum cytostatic effect.
Use of a Biologic End Point Other Than Toxicity to Define the Dose
Statistical trial designs for identifying a dose that maximizes the biologic response (optimum biologic dose) would likely require many more patients than are typically studied in phase I trials.8 For example, serum markers of tumor progression were studied from 312 of 415 patients treated with a matrix metalloproteinase (MMP) inhibitor to try to establish an appropriate dose for further studies.9 Trials designed to demonstrate that there is a dose-response for biologic activity would require fewer patients than trials to find the optimum biologic dose but would still require more than phase I trials. For example, a trial to demonstrate a dose response between two dose levels with true biologic response rates of 70% and 90% of the patients would require 76 patients treated at each level; biologic response rates of 40% and 90% would require 17 patients treated at each level. (If the dose levels are known to have tolerable toxicity, a randomization between the dose levels is advisable.) Trial designs for identifying a dose that yields biologic activity above some prespecified level would generally require fewer patients but potentially still more than studied in standard phase I trials. For example, a phase I trial10 of O6-benzylguanine, which targeted a dose at which
Use of Toxicity to Define Dose The definition of tolerable in a phase I trial needs to take into account that the future use of the agent may be for long-term administration. This can create some problems in that the patient population typically treated in phase I trials may have tumors that progress rapidly, not allowing the tolerability of long-term administration to be adequately evaluated. One possibility is to estimate a maximum-tolerated dose based on a shorter exposure period, recognizing that this dose may require reduction in future efficacy trials. In these efficacy trials, toxicity would require careful monitoring to assess toxicity with long-term exposure.
PHASE II PRELIMINARY EFFICACY TRIALS Single-Arm Designs: Comparisons With Historical Experience This design is the one that is typically used for cytotoxic agents. In that setting, one uses historical experience to define the true response rates necessary to generate interest in pursuing development of the agent(s). For example, one might target a 20% versus 5% response rate for a new agent in a setting where therapeutic options have limited effectiveness and historically a 5% background response rate might be expected in the absence of active treatment. For a combination with a known active agent, one might target a 50% versus 30% response rate, where 30% is the single-agent response rate of the active agent based on historical data. Based on these targeted response rates, typically a design using two stages of accrual involving a total of 30 to 50 patients is used to determine whether the agent warrants further study based on the observed response rate.18,19 A similar design could be used for a cytostatic agent, where one substitutes for response rates a clinical end point expected to be affected by the agent (eg, progression-free survival).20 For example, one could use a standard design involving 37 patients that targeted 1-year progression-free survival rates of 20% versus 5% or a design that targeted a median time-to-progression of 8 months versus 4 months. An example of the use of this design is an ongoing trial21 of the cyclin-dependent kinase inhibitor flavopiridol in patients with advanced colorectal cancer, which targeted 20% versus 5% 6-month stable or responsive disease.
A slight complication in using a two-stage design with an outcome that takes 6 months or 1 year to evaluate is that the accrual of the first stage can be completed some time before it is known whether there are sufficient positive clinical outcomes to continue on to the second stage of the trial. In this situation, one can temporarily stop accrual or allow for a slight over-accrual to the first stage.22 The chance of this problem occurring may be lessened by having an additional criterion involving tumor response that permits one to go on to the second stage, eg, if
A more serious difficulty in using these types of designs for cytostatic agents can be deciding what is an appropriate target progression-free survival. Unlike a partial or complete tumor response that occurs rarely in the absence of treatment, in some settings it would not be unusual for some patients to be progression-free at 1 year in the absence of treatment, eg, patients with renal tumors. Therefore, if one targets a 20% versus 5% 1-year progression-free survival, one should have historical data that documents that the 1-year progression-free survival of patients treated with inactive agents is One possible way to avoid the difficulty in determining an appropriate target progression-free survival is to use a quality-of-life or clinical-benefit end point in which no responses would be expected to be seen in the absence of treatment. For example, a clinical-benefit responder could be defined to be a patient with a specified improvement in pain or a specified improvement in performance status. If symptomatic treatment is already optimized and one would not expect such responders in the absence of treatment, then one could use a target of a small percentage of responders in a phase II trial. This, of course, presumes that one believes that the cytostatic agent is likely to affect these clinical parameters. A positive phase II trial of this sort would be followed by a (large) comparative randomized trial that could target the same clinical end point or progression-free or overall survival. This was the approach used in the development of the (cytotoxic) agent gemcitabine for pancreatic cancer, where after patients were reported to have improvement in disease-related symptoms in a phase II trial,26 a randomized trial targeting clinical benefit was performed.27 Single-Arm Designs: Targeted Biologic End Points If there is a biologic end point specific for the agent then one could use a single-arm design to demonstrate that the agent does modify this end point. One would require sufficient data on the reproducibility of the end point to define a statistically valid threshold of biologic response or, even if the data are analyzed without a dichotomous definition of response, to determine the sample size necessary for identifying a significant effect of the treatment. A small sample size may be sufficient (eg, 30 patients), but it may not be easy to accrue these patients because eligibility would be restricted to patients for whom the end point was measurable. In particular, patients with tumor tissue accessible for serial biopsy may be required, although it may be possible in some circumstances to assess the biologic end point using serial plasma sampled in an ex vivo assay. Another potential problem with this design is that if an agent acts by a mechanism that is different from that hypothesized, then one might discard a clinically active agent because the wrong biologic end point was assessed. In addition, the performance of biologic studies on patients is resource intensive. Lastly, even if an agent demonstrates activity in terms of the biologic end point, some may not consider this strong enough evidence in the absence of supporting clinical data to begin a large randomized efficacy trial; biologic activity for some biologic end points would be viewed as a necessary but not sufficient condition for proceeding. Even when focusing on a targeted biologic end point, one can also consider a patient with an objective tumor response (partial/complete response) or a patient who is progression-free at a designated time point to be a positive outcome with this type of design. Single-Arm Designs: Each Patient as His Own Control In this design, a single cohort of patients with progressive disease is treated with a cytostatic agent to ascertain whether the agent slows down the rate of progression for each patient as compared with his pretreatment rate of progression. Various approaches to quantifying the rate of progression may be used. One possibility is to use a standard definition of progression over a fixed time period, eg, a 25% increase in tumor volume or the presence of new metastases over a 2-month period. However, even if all the patients in a cohort progressed in the 2 months before the start of treatment with a cytostatic agent and only one half progressed after the start of treatment, there are two reasons that this result would not necessarily imply that the agent was having any effect. First, there is random variation in tumor growth. Therefore, if the average rate of growth of tumors was approximately 25% every 2 months, one might expect one-half of the tumors to grow less than 25% in any 2-month period. This might be true even if the cohort of patients was selected as having greater than 25% tumor growth in the 2 months proceeding treatment with the cytostatic agent (this is sometimes known as regression to the mean). And second, for many tumors, the doubling time lessens as the tumor becomes larger even in the absence of treatment.28 Therefore, a reduction in the number of patients with tumors progressing may not be because of treatment with the cytostatic agent.
One might attempt to avoid these problems by using a different quantification of tumor progression. For example, suppose one measures the tumor growth rate in the 2-month period before and after the start of treatment with the cytostatic agent. One could consider the before-to-after ratio of these growth rates as a measure of cytostatic effect. This approach would avoid some of the problems associated with using a standard definition of progression because relatively large ratios could be targeted. More simply, one could require a large increase in tumor size (eg, This type of approach is quite difficult to implement for the following three reasons: (1) patients must be enrolled in the study for 2 months preceding treatment with the agent, (2) rules will need to be defined for calculating growth rates in the presence of new metastases, and (3) it might be difficult to identify suitable patients who meet the entry criteria in terms of tumor growth rate. Even if these obstacles can be overcome, sufficient historical data on the natural history of untreated tumors will be required to ensure that a specified slow-down in tumor growth would be unlikely to occur without treatment. An alternative approach, which can use standard definitions of progression, is to compare the time-to-progression for each patient while being treated with the cytostatic agent as second-line treatment with his/her time-to-progression while being treated with first-line standard treatment. If his/her time-to-progression is longer during treatment with the cytostatic agent, say at least 33% longer, then that patient is considered as having had a successful treatment with the cytostatic agent. The decision as to whether to further study the agent is then based on the number of patients whose times-to-progression were longer in this manner. This approach has been used in the setting of testing cytotoxic agents in lymphoma where the quality and durations of responses for first- and second-line agents are compared.29 Recently, this design has been suggested for trials of cytostatic agents.3 A potential problem with this approach is that there may be bias in declaring progression too early on the first-line treatment because the investigators know that the patients will receive the cytostatic agents as soon as they progress; careful guidelines within the protocol for assessing progression can help minimize this potential bias. Additionally, there may be implementation difficulties in enrolling patients before their first-line treatment for a trial of a second-line treatment. Enrolling patients after they progress on first-line treatment avoids these problems but leads to potential bias in the selection of the patients included in the trial. Multi-Arm Randomized Selection Designs In this design, the choice of patient treatment is randomized among different cytostatic agents. The numbers of patients treated with each agent is relatively small, eg, 30 to 50 patients. The purpose of such a trial is to select an agent for further study. With these small sample sizes, the trial cannot ensure that the agent with the best observed clinical outcome (eg, median time to progression) is indeed the agent with the greatest clinical activity. However, such a design can ensure that an inferior agent will not be chosen if it is sufficiently inferior to the best agent.30 Selecting the agent with the best clinical outcome in such a trial does not guarantee that the agent will have sufficient activity to warrant further evaluation. Therefore, one should try to build into such trial designs the option of not studying any of the agents further if the best one does not meet some minimal criteria for activity. As with the single-arm designs, defining minimal clinical activity may not be easy.
COMPARATIVE RANDOMIZED EFFICACY TRIALS When overall survival is the end point of the trial, a reviewer mentions the possibility of continuing to treat patients with the cytostatic agent after they have progressive disease because of the possibility that the agent could inhibit tumor growth despite increasing tumor size. Although there may be theoretical reasons why cytostatic agents may continue to be efficacious after progression, as a practical matter it would be inappropriate not to offer salvage therapy to patients after they progress. If the patients on the treatment arm not receiving the cytostatic agent receive (effective) salvage therapy and the patients on the treatment arm receiving the cytostatic agent do not (because they continue to receive the cytostatic agent), this could bias the survival comparisons. Comparison of Cytostatic Agent With a Placebo This design requires a setting in which a treatment arm with only supportive care is acceptable, eg, patients who are responding or have stable disease after initial cytotoxic chemotherapy or patients for whom there is no effective therapy. A similar design is used in an ongoing trial by the Eastern Cooperative Oncology Group of marimastat versus placebo after major response to cytotoxic chemotherapy in metastatic breast cancer. A typical sample size and design for such a trial is 244 patients (122 per arm) accrued over 2 years with 1.5 years of additional follow-up to detect a 50% improvement in median progression-free survival from 8 months to 12 months (one-sided alpha = 0.05, power = 0.90). Other possible designs are listed in Table 1 and depend on the underlying assumptions concerning the effect size. Note that one could use survival rather than progression-free survival as the outcome in some settings, eg, in a trial with advanced-stage non–small-cell lung cancer patients after chemotherapy. The designs listed in Table 1 are also appropriate if one targets the same improvements in overall survival as those listed for progression-free survival, eg, an increase in median survival from 8 months to 12 months. With survival as the end point, one can use a no-further-treatment arm rather than a placebo treatment arm.
If a randomized trial is used in a screening mode in which the agent has not previously shown clinical activity and a larger definitive trial is planned if the agent shows activity in the present trial, then one might consider using a one-sided  0.10 or 0.20 instead of 0.05. (An 0.10 is frequently used in standard phase II trials of cytotoxic agents.) Using = 0.10 or = 0.20 would reduce the required sample size by approximately 23% and 47%, respectively. Using = 0.10 is probably not worth the cost of performing a definitive follow-up trial because the reduction in sample size is only 23%. A trial using = 0.20 may be an attractive option, provided that the investigators understand the implications of using such a large alpha. At the conclusion of the trial, a P  .20 should be considered sufficient evidence of activity to perform the follow-up trial. However, an inactive agent will lead one in five times to a P .20. If one uses screening randomized trials, it is important to allow for early stopping of the trial if it seems that the agent is not effective (not necessarily significantly worse). Formal statistical methods have been developed for this purpose.31 Given that there may be little or no clinical evidence that the agent is effective when beginning a screening trial, we suspect that most agents tested will not be sufficiently active to warrant further study. Therefore, the provision for early stopping may result in a considerable savings in sample size and time. A potential problem with designs incorporating a placebo is that patients may find it unattractive to participate in a trial with a treatment arm that does not receive active treatment. This problem is somewhat lessened with progression-free survival as the end point by allowing a cross-over to the other treatment at the time of progression. But with this latter design, it is important that the trials be performed double blinded to avoid the potential of investigator bias in declaring progression. Another possible approach is to use an enrichment (randomized discontinuation) design, in which all patients initially receive the agent and then patients not progressing are randomized to continue the agent or placebo.4,32 However, patients may also find this design unattractive in that they will be potentially discontinued from a therapy that seems to be working. In addition, a large sample size may be required to obtain a sufficient number of randomized patients. In theory, the number of randomized patients could be less than that required for a comparative randomized trial using a standard design because of the enrichment. However, the total sample size required to obtain the enriched population could be larger than required in a standard design. Furthermore, there are serious potential biases in using an enrichment design,33 with the main problem being in the present application that the design will lead to an effective agent being declared to be ineffective if its continued use in a patient is not sufficiently better than its initial use. We do not recommend this type of design unless it is believed to be impossible to conduct a trial with a standard design. Comparison of Cytostatic Agent Plus Cytotoxic Agent Versus Cytotoxic Agent The initial setting for this design is generally metastatic disease in which cytotoxic chemotherapy is appropriate. There may also be some high-risk adjuvant or locally-advanced disease settings where this design might be useful. For example, a trial is being conducted by the Eastern Cooperative Oncology Group of chemotherapy and radiotherapy, with or without thalidomide, for patients with stage III non–small-cell lung cancer. Typical sample sizes are the same as those listed in Table 1. A potential problem with this design is that there may be a negative interaction between the effects of the cytostatic agent and the cytotoxic agent given concurrently, which could make it seem that the cytostatic agent is ineffective. This potential problem can presumably be lessened by cycling the administration of the cytostatic and cytotoxic agent in the combination treatment arm. However, it may not be optimal to interrupt a cytostatic agent. Another alternative is to administer the cytotoxic agent first, followed by the cytostatic agent. Again, this may not be optimal if the cytostatic agent is most effective if given early. Ideally, such interactions would be explored in preclinical models before initiating the trial, and pharmacokinetic and pharmacodynamic interactions would be studied in a small number of patients in the clinic.
RECOMMENDATIONS After this dose is chosen, consider whether there are sufficient historical data on a patient population, untreated or treated with inactive agents, that is similar to the patient population being considered for treatment with the cytostatic agent. The historical data required could be the survival or progression-free survival experience for a group of patients with the same stage of disease and amount of prior treatment, similar organ function and performance status, and for whom the same procedures were used for monitoring disease progression. Preferably this historical experience would come from patients treated at the same institutions with the same referral patterns in a recent era, so that similar diagnostic methodologies and supportive care were available. If such an historical database exists, then one may use a 30- to 50-patient single-arm trial design, with the level of disinterest determined by the historical data. An agent demonstrating sufficient activity above this level would then be tested in a 200- to 300-patient comparative randomized trial.
If such a historical database does not exist, then consider performing either a small screening comparative randomized trial (with
We thank the reviewers for their helpful suggestions.
1. Harris AL: Antiangiogenesis for cancer therapy. Lancet 349: 13-15, 1997 (suppl 2) 2. Denis LJ, Verweij J: Matrix metalloproteinase inhibitors: present achievements and future prospects. Invest New Drugs 15: 175-185, 1997[Medline] 3. Von Hoff DD: There are no bad anticancer agents, only bad clinical trial designs: Twenty First Richard and Hinda Rosenthal Foundation Award Lecture. Clin Cancer Res 4: 1079-1086, 1998[Abstract]
4.
Gelmon KA, Eisenhauer EA, Harris AL, et al: Anticancer agents targeting signaling molecules and cancer cell environment: challenges for drug development? J Natl Cancer Inst 91: 1281-1287, 1999 5. Holmgren L, O’Reilly MS, Folkman J: Dormancy of micrometastases: Balanced proliferation and apoptosis in the presence of angiogenesis suppression. Nat Med 1: 149-153, 1995[Medline] 6. Brooks PC, Stromblad S, Klemke R, et al: Antiintegrin v3 blocks human breast cancer growth and angiogenesis in human skin. J Clin Invest 96: 1815-1822, 1995 7. O’Reilly MS, Boehm T, Shing Y, et al: Endostatin: An endogenous inhibitor of angiogenesis and tumor growth. Cell 88: 277-285, 1997[Medline] 8. Simon R: New statistical designs for clinical trials of immunomodulating agents, in Kresina TF (ed): Immune Modulating Agents. New York, NY, Marcel Dekker, 1998, pp 539-550 9. Nemunaitis J, Poole C, Primose J, et al: Combined analysis of studies of the matrix metalloproteinase inhibitor marimastat on serum tumor markers in advanced cancer: Selection of a biologically active and tolerable dose for longer-term studies. Clin Cancer Res 4: 1101-1109, 1998[Abstract] 10. Friedman HS, Kokkinakis DM, Pluda J, et al: Phase I trial of O6-benzylguanine for patients undergoing surgery for malignant glioma. J Clin Oncol 16: 3570-3575, 1998[Abstract]
11.
Slaton J, Perrotte P, Inoue K, et al: Interferon-mediated down-regulation of angiogenesis-related genes and therapy of bladder cancer are dependent on optimization of biological dose and schedule. Clin Cancer Res 5: 2726-2734, 1999 12. Goodman SN, Zahurak ML, Piantadosi S: Some practical improvements in the continual reassessment method for phase I studies. Stat Med 14: 1149-1161, 1995[Medline]
13.
Simon R, Freidlin B, Rubinstein L, et al: Accelerated titration designs for phase I clinical trials in oncology. J Natl Cancer Inst 89: 1138-1147, 1997 14. Wojtowicz-Praga S, Torri J, Johnson M, et al: Phase I trial of Marimastat, a novel matrix metalloproteinase inhibitor, administered orally to patients with advance lung cancer. J Clin Oncol 16: 2150-2156, 1998[Abstract]
15.
Eckhardt SG, Rizzo J, Sweeney KR, et al: Phase I and pharmacologic study of the tyrosine kinase inhibitor SU101 in patients with advanced solid tumors. J Clin Oncol 17: 1095-1104, 1999 16. Pithavala Y, Shalinsky D, Wilding G, et al: Comparison of preclinical efficacy and associated plasma concentrations of AG3340, a matrix metalloprotease (MMP) inhibitor, with plasma concentrations achieved clinically. Proc Am Soc Clin Oncol 18: 223a, 1999 (abstr 860) 17. Goel R, Hirte H, Shah A, et al: Phase I study of the metalloproteinase inhibitor Bayer 12-9566. Proc Am Soc Clin Oncol 17: 217, 1998 (abstr 840) 18. Fleming TR: One sample multiple testing procedure for phase II clinical trials. Biometrics 38: 143-151, 1982[Medline] 19. Simon R: Optimal two-stage designs for phase II clinical trials. Controlled Clin Trials 10: 1-10, 1989[Medline]
20.
Eisenhauer RA: Phase I and II trials of novel anti-cancer agents: Endpoints, efficacy and existentialism. Ann Oncol 9: 1047-1052, 1998 21. Bennett P, Mani S, O’Reilly S, et al: Phase II trial of flavopiridol in metastatic colorectal cancer: Preliminary results. Proc Am Soc Clin Oncol 18: 277 1999 (abstr 1065) 22. Herndon JE: A design alternative for two-stage, phase II, multicenter cancer clinical trials. Controlled Clin Trials 19: 440-450, 1998[Medline] 23. Zee B, Melnychuk D, Dancy J, et al: Multinomial phase II cancer trials incorporating response and early progression. J Biopharm Stat 9: 351-363, 1999[Medline] 24. Fine HA, Loeffler HS, Kyritsis A, et al: A phase II trial of the anti-angiogenic agent, thalidomide, in patients with recurrent high-grade gliomas. Proc Am Soc Clin Oncol 16: 385a, 1997 (abstr 1372) 25. Dixon DO, Simon R: Sample size considerations for studies comparing survival curves using historical controls. J Clin Epidemiol 41: 1209-1213, 1988[Medline] 26. Casper ES, Green MR, Kelsen DP, et al: Phase II trial of gemcitabine (2,2`-difluorodeoxycytidine) in patients with adenocarcinoma of the pancreas. Invest New Drugs 12: 29-34, 1994[Medline]
27.
Burris HA III, Moore MJ, Andersen J, et al: Improvements in survival and clinical benefit with gemcitabine as first-line therapy for patients with advanced pancreas cancer: A randomized trial. J Clin Oncol 15: 2403-2413, 1997 28. DeVita VT Jr: Principles of cancer management: Chemotherapy, in DeVita VT Jr, Hellman S, Rosenberg SA (eds): Cancer: Principles & Practice of Oncology, ed 5. Philadelphia, PA, Lippincott-Raven, 1997, pp 333-347 29. Liu SY, Wary JF, Petersdorf SH, et al: Follow-up of relapsed B-cell lymphoma patients treated with iodine-131-labeled anti-CD20 antibody and autologous stem-cell rescue. J Clin Oncol 16: 3270-3278, 1998[Abstract] 30. Simon R, Wittes RE, Ellenberg SS: Randomized phase II clinical trials. Cancer Treat Rep 69: 1375-1381, 1985[Medline] 31. Wieand S, Schroeder G, O’Fallon JR: Stopping when the experimental regimen does not appear to help. Stat Med 13: 1453-1458, 1994[Medline] 32. Kopec JA, Abrahamowicz M, Esdaile JM: Randomized discontinuation trials: utility and efficiency. J Clin Epidemiol 46: 959-971, 1993[Medline] 33. Leber PD, Davis CS: Threats to the validity of clinical trials employing enrichment strategies for sample selection. Controlled Clin Trials 19: 178-187, 1998[Medline] Submitted June 16, 1999; accepted July 18, 2000. This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
INTRODUCTION
90% of the patents had undetectable O6-alkylguanine-DNA alkyltransferase levels, treated up to 13 patients at each dose level, and smaller sample sizes may be possible (work in progress). 
