Oncolytics Biotech Inc. (ONCY)

[onco_investor]

Phase II Cancer Trial Design Presentation from AACR 2009

ONCY’s Phase II NSCLC trial design is a single stage Fleming design according to the 2010 ASCO Abstract – This is not randomized. A presentation by Miguel Villalona-Calero & Donn Young was made at AACR 2009 and is a good read on current trial design.

From the ASCO abstract...

http://abstract.asco.org/AbstView_74_52799.html

"We have hypothesized those patients with EGFR-mutated, EGFR-amplified, or Kras-mutated NSCLC should all have a common downstream activated Ras pathway and will exhibit response to treatment with reovirus.

Methods: This is a Fleming single stage, single-arm, phase II study to evaluate the objective response rate and 6-month progression-free survival of reovirus in combination with P and C as first-line therapy in patients with metastatic NSCLC. Eligible patients are those with ECOG PS 0-2, adequate organ function, no prior systemic chemotherapy, and tumors with the specified genotype. Reovirus (3 x 1010 TCID50) will be administered intravenously daily on days 1-5, C (AUC 6) on day 1, and P (200 mg/m2) on day 1 of each 21-day cycle."

The original Villalona Young presentation can be found at...

http://www.vailworkshop.org/files/2009/Syllabus/8-2-2009/Phase%20II%20Trials/Villalona%20syllabus%20material%20phase%202.doc

By Miguel Villalona-Calero & Donn Young

The primary objective of Phase 2 cancer clinical trials is the determination of anti-tumor efficacy in a manner that provides information to allow one to decide if further studies of the drug or regimen are warranted. These studies to test regimens or agents for suitability for Phase 3 studies address the conflicting needs to accrue sufficient patients to better define efficacy, but to minimize the number of patients treated if the drug is ineffective. Current accepted Phase 2 trial designs include a wide range of single- and multiple-stage, randomized and single-arm designs using frequentist, Bayesian, or predictive probability designs, addressing single efficacy endpoints or multiple endpoints incorporating efficacy and safety. Commonly used Phase 2 design and outcomes measures are listed below.

Study Designs

Frequentist
-Gehan 2-Stage
-Simon 2-Stage Optimal
-Simon 2-Stage Minimax
-Fleming 1-stage
-Gehan-Simon 3-stage
-Randomized Phase 2
-Constant Arc-Sine
-Randomized Discontinuation

Bayesian
-Thall-Simon-Estey Bayesian
-1-Stage Bayesian
-2-Stage Bayesian
Tan Machin
Heitian

Endpoints Outcomes Measures

RECIST/WHO Response Rate = Progression-Free Rate
CR + PR = Disease-Free Rate
CR + PR + SD = Event-Free Rate
Time Point = Based-Biological Endpoints
6-month = Safety & Adverse Events
12-month = Multiple Endpoints
Continuous = Quality of Life

The two-stage Phase 2 design to minimize the number of patients treated with ineffective regimens was initially proposed in Gehan’s (Gehan et al, J Chron Dis, 1961) “Rule of 14” design with 14 patients accrued in the first stage followed by 25 patients overall if the first stage goal is met. While this design allows early termination for ineffectual regimens, it fails to address probabilities associated with the decision to recommend that the Phase 2 regimen be considered for subsequent randomized Phase 3 studies. Regimens with a low probability for "success" were recommended for subsequent large-scale randomized clinical trials.

A decision-based two-stage design was refined by Simon (Simon et al, Cont Clin Trials, 1989) whose procedure minimizes the expected sample size given specified response rates and * and * error rates. While a large number of possible trials are generated using his procedure, the investigator usually selects either the ‘optimal’ solution which minimizes the number of patients treated if the regimen is truly ineffective or the ‘minimax’ solution which minimized the overall sample size. The majority of Phase 2 trials are based on Simon’s design.

The two-stage designs usually address an efficacy endpoint of objective clinical response, recently codified using the RECIST criteria for solid tumors. Positive responses are commonly defined as either complete [CR] or partial [PR] responses.

Unfortunately, these response determinations fail to address biological endpoints used in the evaluation of cytostatic rather than cytotoxic agents. With these molecularly-targeted agents, stable disease [SD], quality of life, or improvement in symptoms may prove to be appropriate measures of therapeutic benefit, especially when potential improvements may be subtle. As noted by Ratain and Eckhardt (J Clin Oncol, 2004), more flexible measures of antitumor activity need to be distinguishable from no treatment or a placebo.

Rather than relying on the combination of [CR+PR+SD] as demonstration of benefit, Phase 2 trials often address time-dependent endpoints of progression-free or disease-free survival translated to dichotomous alternatives at a given time point; i.e., the proportion of patients free of progression at one year following initiation of treatment. Given the time period from initiation of treatment to the endpoint, two-stage designs often prove impractical and a Fleming (Biometrics,1982) one-stage design is used. Since a larger number of patients will be treated prior to a decision to embark on additional studies and due to limited information on the toxicity profile of a new agent, the single-stage designs frequently incorporate sequential early stopping rules for adverse events.

More recently, Korn et al (J Clin Oncol, 2001) proposed single-arm Phase 2 designs with comparisons with historical control data.

The lack of well-characterized historical data with which to make comparisons often limits one’s confidence that the historical data present a reasonable baseline from which to detect therapeutic improvements. Mick et al (Cont Clin Trials, 2000) developed a novel Phase 2 design for failure-time endpoints by comparing time to treatment failure or progression on the new regimen [TTP2] with the individual patient’s failure time or TTP1 observed with their prior regimen of treatment. If the new agent demonstrated a TTP2/TTP1 ratio of greater than 1.33, it would be considered effective and worthy of further study.
Randomized Phase 2 trials provide a mechanism to determine which of two regimens should undergo further study in the Phase 3 setting (Simon et al, Cancer Treat Rep, 1985). These trials usually randomize patients between one of two regimens differing by dose level, schedule, or specific agent. Heeding the cautions of Liu et al (Liu et al, Control Clin Trials, 1999) the randomized Phase 2 trial is not to be viewed as an inexpensive Phase 3 trial since the study is not powered for inferential comparisons between the treatment arms. With both arms incorporating two-stage designs, however, the randomized Phase 2 trial offers four specific decision points for determining regimen efficacy. Very recently, Wiand (J Clin Oncol, 2005) criticized the randomized

Phase 2 design in a situation (Dark et al, J Clin Oncol, 2005) where both arms in a trial met first and second stage goals offering the investigators a “pick the winner” dilemma in which other information such as quality of life, toxicity, and costs were involved in the selection of the “winner.”

The randomized discontinuation design (Kopec et al, J Clin Epidemiol,1993), which has recently been proposed for selection of antineoplastic agents by Rosner et al (J Clin Oncol, 2002), incorporates a time-dependent endpoint such as time to progression with disease response. Patients with stable disease are randomized to either continuation with the agent or a placebo (the discontinuation). Patients subsequently showing progression on placebo are then retreated with the agent to determine if disease stability can be regained. This design allows one to demonstrate the effectiveness of a cytostatic agent by distinguishing between disease stability due to the agent versus due to a naturally slow tumor growth rate. This design is most appropriate in diseases where tumor growth rates are slow, whereas with an aggressive/rapidly progressive malignancy, few patients would quality for randomization, limiting therefore the design’s effectiveness.

While frequentist design have predominated in Phase 2 clinical trials, Bayesian Phase 2 trials have become more visible following the pioneering work of Staquet and Sylvester (Biomedicine, 1977; Cancer Treat Rep, 1980) and extensive development by Thall (Biometrics, 1994, Stat Med, 1995, Stat Med, 1998), Tan and Machin (Stat Med, 2002), Heitjan (Stat Med, 1997), and Mayo and Gajewski (Cont Clin Trials, 2004). In single and two-stage designs, Bayesian designs allow the investigator to determine the probability that the true response rate exceeds a pre-specified target response or to determine the response interval that has a 95% chance of containing the true response rate. Using prior probabilities based on the investigator’s prior beliefs about the new regimen, the study design re-computes posterior probabilities based on observed data. While many Bayesian designs use continuous monitoring, studies may be adapted to a two-stage model.

Sample Size Calculation

Prior determination of the sample size that is needed to show an important difference is essential in a well designed Phase II study. Two errors can be made in a test of a hypothesis: 1- rejecting the null hypothesis when it is true (Type I Error, ?) (false-positive); or 2- not rejecting the null hypothesis when it is false (Type II, ?) (false-negative). Another important consideration is Power, which is defined as the probability of rejecting the null hypothesis when it is false, or of concluding the alternative hypothesis when it is true. In other words, Power is the capability of a study to detect a given difference of a given size if the difference really exists. Confidence intervals give the additional information of the variability of the estimate, given the upper limit and a lower limit with an associated probability. Many statisticians prefer to see confidence intervals, since they clearly illustrate the magnitude of the difference and make clear the role played by sample size.

Before you go to your statistician it is important to figure out the following points:
1-Do you want to consider studying single versus two proportions
2-What is the desired level of significance of the null hypothesis (p0)?, e.g., what is the response rate for conventional care
3-What chance should there be of detecting an actual difference (what power) associated with the alternative hypothesis (p1) is desired?
4-How large should the difference between the proportions (p1- p0) be in order for it to have clinical importance?
5-What is a good estimate of the standard deviation in the population? The value of the null hypothesis, determines in most cases the standard deviation

Studies of biologicals and molecularly targeted agents provide substantial challenges to trial design, since traditional RECIST/WHO criteria response rate may not be adequate to comprehensively assess the clinical benefit produced by the agent, and in some cases a surrogate marker in tumor, normal tissue or cells may be required to assess if appropriate tumor/stroma concentrations or the agent are being achieved. This may have implications of individualized dosing and schedule.

Given this complexity of design and outcome alternatives, the selection of a trial design requires close collaboration between the study investigator and clinical biostatisticians to clearly define study objectives, to select appropriate endpoints, to select a trial design, and to compute the required number of patients to be enrolled.


References
1)Gehan EA, The determination of the number of patients required in a preliminary and a follow-up trial of a new chemotherapeutic agent. J Chron Dis, Apr;13:346-53.1961
2)Simon R. Optimal two-stage designs for phase II clinical trials. Control Clin Trials. 1989 Mar;10(1):1-10
3)Ratain M, Eckhardt SG, Phase II studies of modern drugs directed against new targets: if you are fazed, too, then resist RECIST. J Clin Oncol. 2004 Nov 15;22(22):4442
4)Fleming TR. One-sample multiple testing procedure for phase II clinical trials. Biometrics. 1982 Mar;38(1):143-51
5)Korn et al. J Clin Oncol, 2001 Clinical trial designs for cytostatic agents: are new approaches needed? J Clin Oncol. 2001 Jan 1;19(1):265-72.
6)Mick R, Crowley JJ and Carroll RJ. Phase 2 clinical trial design for noncytotoxic anticancer agents for which time to disease progression is the primary endpoint. Cont Clin Trials. 2000;21:343-359.
7)Simon R, Wittes RE Methodologic guidelines for reports of clinical trials. Cancer Treat Rep. 1985 Jan;69(1)
8)Liu PY., LeBlanc M, and Desai M. False positive rates of randomized Phase 2 designs. Control Clin Trials. 1999;20:343-352.
9)Wieand HS. Randomized Phase 2 trials: What does randomization gain? J Clin Oncol. 2005;23:1794-1795.
10)Dark GG, Calvert AH, and Grimshaw R Randomized trial of two intravenous schedules of the topoisomerase I inhibitor liposomal lurtotecan in women with relapsed epithelial ovarian cancer: A trial of the National Cancer Institute of Canada Clinical Trials Group. J Clin Oncol. 2005;23:1859-1866.
11)Kopec JA, Abrahamowicz M, and Esdaile JM. Randomized discontinuation trials: Utility and efficiency. J Clin Epidemiol. 1993;46:959-971.
12)Rosner GL, Stadler W, and Ratain MJ. Randomized discontinuation design: Application to cytostatic antineoplastic agents. J Clin Oncol. 2002;20:4478-4484.
13)Staquet MJ and Sylvester RJ. A decision theory approach to Phase 2 clinical trials. Biomedicine. 1977;26:262-266.
14)Sylvester RJ and Staquet MJ. Design of Phase 2 clinical trials in cancer using decision theory. Cancer Treat Rep. 1980;64:519-524.
15)Thall PF and Sung HG. Some extensions and applications of a Bayesian strategy for monitoring multiple outcomes in clinical trials. Stat Med. 1998;17:1563-1580.
16)Thall PF, Simon RM and Estey EH. Bayesian sequential monitoring designs for single-arm clinical trials with multiple outcomes. Stat Med. 1995;14:357-379.
17)Thall PF and Simon R. Practical guidelines for Phase 2B clinical trials. Biometrics. 1994;50:337-349.
18)Tan S-B and Machin D. Bayesian 2-stage design for Phase 2 clinical trials. Stat Med. 2002;21:1991-2012.
19)Heitjan DF. Bayesian interim analysis of Phase 2 cancer clinical trials. Stat Med. 1997;16:1791-1802.
20)Mayo MS and Gajewski BJ. Bayesian sample size calculations in Phase 2 clinical trials using informative conjugate priors. Cont Clin Trials. 2004;25:157-167.
21)Basic & Clinical Biostatistics Dawson-Saunders