Connie Lee Batlevi and Robert J. Motzer
PHASE II TRIALS
• Primary objectives: Test antitumor efficacy of investigation drug in specific malignancies to warrant further development
• Secondary objective: Continue safety assessment of new drug
• Population: Narrowly defined population in pts w/specific cancer dx (eg, 2nd line tx of met triple negative breast CA, anplastic lymphoma kinase (ALK) rearranged NSCLC)
• Variations include combination phase I/II trials which define optimal dose & determine efficacy, combination phase II/III trials which define efficacy & expanded tox & minimize trial duration while maximizing data collection
Types of Phase II Designs
• Single arm design
Nonrandomized design w/outcome comparison to historical outcomes
Binary endpoints of clinical response vs. no response. Limitations of comparison to historical outcome include difference in pt selection criteria, methods of outcome assessment, differences in supportive care
Egs: Gehan design, Simon two stage design, Fleming design
• Randomized phase II
Requires larger sample size but reduces bias compared to historical control. Limitations include inadequate power of studies to definitively compare investigational agents.
Egs: Noncomparative, “pick the winner,” randomized discontinuation
Single Stage One Arm (Biometrics 1982;38:143)
• All pts enrolled in one stage & evaluated for clinical endpoints, including short duration endpoints, ie, PFS. Generally, no early termination rules except for instances of unacceptable tox.
• Most common is Fleming one stage design—fixed sample size based on desired type I error & RR (power) allows for adequate planning but also restricts trial flexibility
Two Stage Design
• Minimizes number of pts exposed to ineffective tx by monitoring for response in first cohort & proceeding only if sufficient response is seen. RECIST response criteria generally utilized.
• Gehan two stage (J Chronic Dis 1961;16:346)—small stage 1 minimize type II error, stage 2 sample size dependent on response of stage 1
• Simon two stage (Control Clin Trials 1989;10:1)—small stage 1 sample size aim to reduce type II error, fixed stage 2 support trial planning & resource allotment
Noncomparative Randomized (Stat Med 1986;5:441)
• Two arms—multiple experimental arms compared to historical control. Limited by comparison to historical control but useful for concurrently assessing efficacy of multiple drugs.
Selection “Pick the Winner” (Cancer Treat Rep 1985;69:1375)
• Multiarm trial w/each arm compared to control & not powered to compare efficacy between different arms → most efficacious arm evaluated in phase III
• Limits sample size by adjusting type I & type II error rates & target difference. Useful for evaluation of investigational Rx vs. standard of care using PFS standpoint.
Randomized Discontinuation (JCO 2002;15:4478)
• All pts treated w/study agent followed by double blind randomization in pts w/SD to continuation of study agent vs. placebo
• Determines whether slow tumor growth is from therapeutic agents vs. dz biology. Useful to determine if continued benefit present after initial benefit.
PHASE III TRIALS
• Gold standard to compare efficacy of experimental Rx w/standard of care
• Large number of pts often requiring multi-institutional or international recruitment to reach accrual goals
• Population: Typically single cancer type w/well-defined eligibility criteria
Parallel group—pts randomized to different tx arms
Crossover—pts initially randomized to a sequence of 2 or more tx w/crossover based on tx response vs. POD
Allocation of pt to subgroups: 1:1 equal allocation most common, stratified random allocation assign pt defined by factors predicting outcome (ie, EGFR mutation), permuted blocked allocation designed to maintain balanced accrual to each arm, ie, blocks of 6 slots per defined factor for purposes of allocation, unequal allocation, ie, 2:1 design to make trial more attractive to pts. All trials need stratification & blocking to negate random imbalance.
Blinding—one sided (physician unblinded, pt blinded) vs. double blinding (physician & pt blinded)
Intent to treat—endpoint analysis include all pts enrolled regardless of duration of tx, cross over, etc.
Superiority—evaluate benefit or efficacy based on defined endpoint
Noninferiority trials—designed by formulating null hypothesis that new Rx is inferior to standard of care. Trial powered to reject null hypothesis. Allows for smaller sample size but sometimes w/smaller limit of acceptability.
Equivalence—define smallest difference of practical clinical significance then design trial w/high power to reject null hypothesis. Requires large sample size.
Interim analysis—allows for early data analysis for early termination as needed
Phase IV Clinical Trials
• Post FDA approval investigation of risk/benefit, efficacy, tox, subpop effect