Smaller sample sizes for phase II trials based on exact tests with actual error rates by trading-off their nominal levels of significance and power

Future progress in improving cancer therapy can be expedited by better prioritization of new treatments for phase III evaluation. Historically, phase II trials have been key components in the prioritization process. There has been a long-standing interest in using phase II trials with randomization against a standard-treatment control arm or an additional experimental arm to provide greater assurance than afforded by comparison to historic controls that the new agent or regimen is promising and warrants further evaluation. Relevant trial designs that have been developed and utilized include phase II selection designs, randomized phase II designs that include a reference standard-treatment control arm, and phase II/III designs. We present our own explorations into the possibilities of developing “phase II screening trials,” in which preliminary and nondefinitive randomized comparisons of experimental regimens to standard treatments are made (preferably using an intermediate end point) by carefully adjusting the false-positive error rates (α or type I error) and false-negative error rates (β or type II error), so that the targeted treatment benefit may be appropriate while the sample size remains restricted. If the ability to conduct a definitive phase III trial can be protected, and if investigators feel that by judicious choice of false-positive probability and false-negative probability and magnitude of targeted treatment effect they can appropriately balance the conflicting demands of screening out useless regimens versus reliably detecting useful ones, the phase II screening trial design may be appropriate to apply.

Download Full-text

Predictive value of phase II clinical trials in pancreatic cancer: Rethinking the road to progress.

Journal of Clinical Oncology ◽

10.1200/jco.2013.31.15_suppl.4036 ◽

2013 ◽

Vol 31 (15_suppl) ◽

pp. 4036-4036 ◽

Cited By ~ 1

Author(s):

Daniel M. Halperin ◽

J. Jack Lee ◽

James C. Yao

Keyword(s):

Clinical Trials ◽

Phase Ii ◽

Predictive Value ◽

Error Rates ◽

Phase Iii ◽

Type I ◽

Type Ii ◽

Type Ii Error ◽

Phase Ii Trials ◽

Fda Approval

4036 Background: Few new therapies for pancreatic adenocarcinoma (PC) have been approved by the Food and Drug Administration (FDA) or recommended by the National Comprehensive Cancer Network (NCCN), reflecting frequent failures in phase III trials. We hypothesize that the high failure rate in large trials is due to a low predictive value for “positive” phase II studies. Methods: Given a median time from initiation of clinical trials to FDA approval of 6.3 years, we conducted a systematic search of the clinicaltrials.gov database for phase II interventional trials of antineoplastic therapy in PC initiated from 1999-2004. We reviewed drug labels and NCCN guidelines for FDA approval and guideline recommendations. Results: We identified 70 phase II trials that met our inclusion criteria. Forty-five evaluated compounds without preexisting FDA approval, 23 evaluated drugs approved in other diseases, and 2 evaluated cellular therapies. With a median follow-up of 12.5 years, none of these drugs gained FDA approval in PC. Four trials, all combining chemotherapy with radiation, eventually resulted in NCCN recommendations. Forty-two of the trials have been published. Of 16 studies providing pre-specified type I error rates, these rates were ≥0.1 in 8 studies, 0.05 in 6 studies and <0.025 in 2 studies. Of 21 studies specifying type II error rates, 7 used >0.1, 10 used 0.1, and 4 used <0.1. Published studies reported a median enrollment of 47 subjects. Fourteen trials reported utilizing a randomized design. Conclusions: The low rate of phase II trials resulting in eventual regulatory approval of therapies for PC reflects the challenge of conquering a tough disease as well as deficiencies in the statistical designs. New strategies are necessary to quantify and improve odds of success in drug development. Statistical parameters of individual or coupled phase II trials should be tailored to achieve the desired predictive value prior to initiating pivotal phase III studies. Positive predictive value of a phase II study assuming a 1%, 2%, or 5% prior probability of success and 10% type II error rate. [Table: see text]

Download Full-text

Reducing Sample Sizes in Phase II Trials based on Exact Binomial Tests by Shifting Design Parameters

International Journal of Cancer and Clinical Research ◽

10.23937/2378-3419/2/3/1020 ◽

2015 ◽

Vol 2 (3) ◽

Author(s):

Dirk Klingbiel

Keyword(s):

Phase Ii ◽

Design Parameters ◽

Sample Sizes ◽

Phase Ii Trials

Download Full-text

Assessing pretest probability in phase II trials: One step back before two steps forward.

Journal of Clinical Oncology ◽

10.1200/jco.2014.32.3_suppl.312 ◽

2014 ◽

Vol 32 (3_suppl) ◽

pp. 312-312

Author(s):

Daniel M. Halperin ◽

Cecile Dagohoy Dagohoy ◽

J. Jack Lee ◽

James C. Yao

Keyword(s):

Phase Ii ◽

Phase Ii Trial ◽

Error Rates ◽

Pretest Probability ◽

Phase Iii ◽

Positive Phase ◽

Type I ◽

Success Rates ◽

Phase Ii Trials ◽

Fda Approval

312 Background: With frequent phase III failures, only 5% of new oncology drugs entering clinical development gain FDA approval. We hypothesize that pivotal trial failures are directly related to the poor predictive value of “positive” phase II studies, with odds of success varying by multiple factors, including disease site. We assessed success rates from phase II to allow calculation of pretest probability of eventual approval. Methods: As the median time from trial start to FDA approval is 6.3 years, we systematically searched clinicaltrials.gov for phase II trials of GI cancer therapy from 1999-2004. We reviewed drug labels for FDA approval. Drugs without FDA approval within 12 months of clinical trial start were classified New Drug Application (NDA) eligible. Drugs with prior approval were considered supplemental NDA (sNDA) eligible. Success rates were calculated as proportion of trials with drugs eventually approved. Results: We identified 280 trials; some included multiple disease sites. Approvals are shown in the table. Trials in different diseases had distinct rates of drug approval. sNDAsuccess in colorectal tumors was driven by multiple trials of drugs later gaining approval. Conclusions: Variable rates of phase II trials resulting in eventual approval for different GI cancers reflect the landscape of distinct diseases in which we test new therapies. Posterior probability of approval after a positive phase II trial varies by prior probability. For a positive phase II study with type I and II error rates of 0.1, given pretest probability of 1%, 3%, or 10%, the odds of eventual approval are 8%, 22%, and 50% respectively. With a quantitative understanding, we can tailor phase II trial design to improve phase III success rates. [Table: see text]

Download Full-text

To randomize, or not to randomize, that is the question: using data from prior clinical trials to guide future designs

Neuro-Oncology ◽

10.1093/neuonc/noz097 ◽

2019 ◽

Vol 21 (10) ◽

pp. 1239-1249

Author(s):

Alyssa M Vanderbeek ◽

Steffen Ventz ◽

Rifaquat Rahman ◽

Geoffrey Fell ◽

Timothy F Cloughesy ◽

...

Keyword(s):

Overall Survival ◽

Clinical Trials ◽

Phase Ii ◽

Primary Endpoint ◽

Effect Size ◽

Standard Of Care ◽

Error Rates ◽

Phase Ii Trials ◽

False Positive Error ◽

Interstudy Variability

Abstract Background Understanding the value of randomization is critical in designing clinical trials. Here, we introduce a simple and interpretable quantitative method to compare randomized designs versus single-arm designs using indication-specific parameters derived from the literature. We demonstrate the approach through application to phase II trials in newly diagnosed glioblastoma (ndGBM). Methods We abstracted data from prior ndGBM trials and derived relevant parameters to compare phase II randomized controlled trials (RCTs) and single-arm designs within a quantitative framework. Parameters included in our model were (i) the variability of the primary endpoint distributions across studies, (ii) potential for incorrectly specifying the single-arm trial’s benchmark, and (iii) the hypothesized effect size. Strengths and weaknesses of RCT and single-arm designs were quantified by various metrics, including power and false positive error rates. Results We applied our method to show that RCTs should be preferred to single-arm trials for evaluating overall survival in ndGBM patients based on parameters estimated from prior trials. More generally, for a given effect size, the utility of randomization compared with single-arm designs is highly dependent on (i) interstudy variability of the outcome distributions and (ii) potential errors in selecting standard of care efficacy estimates for single-arm studies. Conclusions A quantitative framework using historical data is useful in understanding the utility of randomization in designing prospective trials. For typical phase II ndGBM trials using overall survival as the primary endpoint, randomization should be preferred over single-arm designs.

Download Full-text

Phase II Stopping Rules That Employ Response Rates and Early Progression

Journal of Clinical Oncology ◽

10.1200/jco.2007.14.1044 ◽

2008 ◽

Vol 26 (22) ◽

pp. 3715-3720 ◽

Cited By ~ 12

Author(s):

John R. Goffin ◽

Dongsheng Tu

Keyword(s):

Phase Ii ◽

Progressive Disease ◽

Response Rate ◽

Error Rates ◽

Stopping Rules ◽

Early Stopping ◽

Phase Ii Trials ◽

End Point ◽

Further Development ◽

Early Progressive Disease

Purpose Phase II oncology trials traditionally have used response rate (RR) as the primary end point, but newer targeted agents require the consideration of alternative end points. High rates of early progressive disease (EPD) suggest inadequate drug activity and may be useful in the early stopping of trials. This study used a simulation to define a set of rules to assess a combined end point of RR and EPD. Methods The simulation assumed a two-stage trial with a specified α error and power. It randomly generated the true response rate, r, of the agent under study and its true rate of early progressive disease, epd, for each run of the simulation. Two pairs of parameters were specified: (rnul, epdnul) and (ralt, epdalt). A drug was considered uninteresting for further development if r was less than or equal to rnul and epd was greater than or equal to epdnul (ie, the null hypothesis) and interesting for further development if r was greater than or equal to ralt or epd was less than or equal to epdalt (ie, the alternate hypotheses). Thresholds for the required number of patients with responses, nr and EPD, np, were generated for each set of parameters. Results Thresholds for nr and np that satisfied the specified error rates were generated. There was at least an 89% likelihood that a study would be stopped at the first stage of accrual if r and epd were uninteresting. Conclusion The simulation was able to establish stopping rules by combining the RR and the EPD that achieved the desired error rates. High rates of early stopping suggest that this design could shorten phase II trials of inactive agents.

Download Full-text