- Email: [email protected]
Pick the Winner Designs in Phase II Cancer Clinical Trials
EDITORIAL
Pick the Winner Designs in Phase II Cancer Clinical Trials Sumithra J. Mandrekar, PhD, and Daniel J. Sargent, PhD
(J Thorac Oncol. 2006;1: 5–6)
I
n this issue of the Journal of Thoracic Oncology, Masters et al.1 present the results of a randomized phase II trial evaluating two schedules of gemcitabine and carboplatin in patients with advanced non–small-cell lung cancer using a “selection design.” Assuming a hazard rate of 1.2 for progression-free survival, the authors state that a sample of 50 patients per arm provided at least an 80% chance of picking the correct regimen using a one-sided log rank test. Unfortunately, the statistical section is rather brief, with no references cited for the selection design used; hence, the exact calculations could not be verified. This is not uncommon in the literature; only 60 of the 266 randomized phase II trials included in a recent literature review had an identifiable statistical design section.2 Given the recent increased interest in randomized phase II trials, driven primarily by the need to evaluate the efficacy of multiple regimens concurrently, a detailed description of the design, conduct, analysis, and interpretation of the trial results is vital to assess the scientific merits and improve our understanding of these designs. We present some of the nuances of a randomized selection design (also known as a screening or pick the winner design). The primary purpose of phase II clinical trials in cancer is to identify promising regimens to take forward to a larger phase III trial. Evaluating regimens by comparing outcomes among independent single-arm trials is problematic because of the confounding effects of trial outcomes with “unknown” trial effects such as patient selection, referral bias, and differences in supportive or ancillary care. Randomization in phase II studies is one method to reduce such outcome-trial effect confounding.3 For cases in which there may be several promising regimens/schedules/treatments for a patient population, a randomized selection or screening design can be used to select the most promising treatment to be further tested. In this setting, controlling type I error (false positive) is less relevant than controlling type II error (false negative), as the goal of such trials is to ensure that if one regimen is indeed superior, then there is a high probability that it will be selected. Screening designs are designed to make a selection between promising experimental regimens in a phase II setting and therefore should not typically include the standard of care or control arm.4 In other words, screening trials are underpowered for performing formal hypothesis testing or comparisons of endpoints, both primary and secondary, across selection arms. Although commonly reported in published studies, including that of Masters et al.,1 p values based on formal statistical tests of comparisons from selection designs are incorrect and should be avoided.2 The selection of an experimental treatment in a screening design can be based solely on the primary endpoint, or it can include other factors when the observed difference in the primary outcome is deemed small.3–5 The latter is known as a flexible screening design, in which other factors such as toxicity, cost, convenience, or quality of life are taken into consideration in making the selection in addition to the primary efficacy measure, much as in clinical practice.4 The wealth of opportunities in cancer drug development mandates intelligent clinical trial design. In our opinion, phase II randomized screening trials, when properly applied, can assure optimal use of limited phase III financial and patient resources. Division of Biostatistics, Mayo Clinic Cancer Center, Mayo Clinic, Rochester, Minnesota, Copyright © 2006 by the International Association for the Study of Lung Cancer ISSN: 1556-0864/06/0101-0005
Journal of Thoracic Oncology • Volume 1, Number 1, January 2006
5
Journal of Thoracic Oncology • Volume 1, Number 1, January 2006
Mandrekar and Sargent
REFERENCES 1. Masters GA, Argiris AE, Hahn EA, et al. A randomized phase II trial using two different treatment schedules of gemcitabine and carboplatin in patients with advanced non-small cell lung cancer. J Thorac Oncol 2006;1:19–24 2. Lee JJ, Feng L. Randomized phase II designs in cancer clinical trials: current status and future directions. J Clin Oncol 2006;23:4450–4457.
6
3. Simon R, Wittes RE, Ellenberg SE. Randomized phase II clinical trials. Cancer Treatment Rep 1985;69:1375–1381. 4. Sargent DJ, Goldberg RM. A flexible design for multiple armed screening trials. Stat Med 2001;20:1051–1060. 5. Steinberg SE, Venzon DJ. Early selection in a randomized phase II clinical trial. Stat Med 2002;21:1711–1726.
Copyright © 2006 by the International Association for the Study of Lung Cancer
Pick the Winner Designs in Phase II Cancer Clinical Trials Sumithra J. Mandrekar, PhD, and Daniel J. Sargent, PhD
(J Thorac Oncol. 2006;1: 5–6)
I
n this issue of the Journal of Thoracic Oncology, Masters et al.1 present the results of a randomized phase II trial evaluating two schedules of gemcitabine and carboplatin in patients with advanced non–small-cell lung cancer using a “selection design.” Assuming a hazard rate of 1.2 for progression-free survival, the authors state that a sample of 50 patients per arm provided at least an 80% chance of picking the correct regimen using a one-sided log rank test. Unfortunately, the statistical section is rather brief, with no references cited for the selection design used; hence, the exact calculations could not be verified. This is not uncommon in the literature; only 60 of the 266 randomized phase II trials included in a recent literature review had an identifiable statistical design section.2 Given the recent increased interest in randomized phase II trials, driven primarily by the need to evaluate the efficacy of multiple regimens concurrently, a detailed description of the design, conduct, analysis, and interpretation of the trial results is vital to assess the scientific merits and improve our understanding of these designs. We present some of the nuances of a randomized selection design (also known as a screening or pick the winner design). The primary purpose of phase II clinical trials in cancer is to identify promising regimens to take forward to a larger phase III trial. Evaluating regimens by comparing outcomes among independent single-arm trials is problematic because of the confounding effects of trial outcomes with “unknown” trial effects such as patient selection, referral bias, and differences in supportive or ancillary care. Randomization in phase II studies is one method to reduce such outcome-trial effect confounding.3 For cases in which there may be several promising regimens/schedules/treatments for a patient population, a randomized selection or screening design can be used to select the most promising treatment to be further tested. In this setting, controlling type I error (false positive) is less relevant than controlling type II error (false negative), as the goal of such trials is to ensure that if one regimen is indeed superior, then there is a high probability that it will be selected. Screening designs are designed to make a selection between promising experimental regimens in a phase II setting and therefore should not typically include the standard of care or control arm.4 In other words, screening trials are underpowered for performing formal hypothesis testing or comparisons of endpoints, both primary and secondary, across selection arms. Although commonly reported in published studies, including that of Masters et al.,1 p values based on formal statistical tests of comparisons from selection designs are incorrect and should be avoided.2 The selection of an experimental treatment in a screening design can be based solely on the primary endpoint, or it can include other factors when the observed difference in the primary outcome is deemed small.3–5 The latter is known as a flexible screening design, in which other factors such as toxicity, cost, convenience, or quality of life are taken into consideration in making the selection in addition to the primary efficacy measure, much as in clinical practice.4 The wealth of opportunities in cancer drug development mandates intelligent clinical trial design. In our opinion, phase II randomized screening trials, when properly applied, can assure optimal use of limited phase III financial and patient resources. Division of Biostatistics, Mayo Clinic Cancer Center, Mayo Clinic, Rochester, Minnesota, Copyright © 2006 by the International Association for the Study of Lung Cancer ISSN: 1556-0864/06/0101-0005
Journal of Thoracic Oncology • Volume 1, Number 1, January 2006
5
Journal of Thoracic Oncology • Volume 1, Number 1, January 2006
Mandrekar and Sargent
REFERENCES 1. Masters GA, Argiris AE, Hahn EA, et al. A randomized phase II trial using two different treatment schedules of gemcitabine and carboplatin in patients with advanced non-small cell lung cancer. J Thorac Oncol 2006;1:19–24 2. Lee JJ, Feng L. Randomized phase II designs in cancer clinical trials: current status and future directions. J Clin Oncol 2006;23:4450–4457.
6
3. Simon R, Wittes RE, Ellenberg SE. Randomized phase II clinical trials. Cancer Treatment Rep 1985;69:1375–1381. 4. Sargent DJ, Goldberg RM. A flexible design for multiple armed screening trials. Stat Med 2001;20:1051–1060. 5. Steinberg SE, Venzon DJ. Early selection in a randomized phase II clinical trial. Stat Med 2002;21:1711–1726.
Copyright © 2006 by the International Association for the Study of Lung Cancer