While critical for efficient and successful drug development, dose selection in early phase oncology trials is very challenging. Phase 1 oncology trials have typically been designed to identify the highest, safe dose of a new drug – the maximum tolerated dose (MTD) from the responses of a small, often heterogeneous, cohort of cancer patients. The MTD identified is then used in either a phase 2 trial or an expansion of the phase 1 study to obtain preliminary efficacy data..
While phase 1 trials have historically identified the MTD using a simple, rule-based design, such as the classical “3+3” method, regulators are starting to encourage model-based and model-assisted designs, which can make MTD estimation more accurate and drug efficacy assessment more reliable in early phase oncology studies. However, selecting and implementing the optimal model-based or model-assisted design can be daunting, and requires statistical expertise. Many designs exist and vary in their approach to balancing participant safety, accurate dose selection and simplicity.
ICON has demonstrated expertise and ability to design smarter innovative early phase oncology studies through model simulation, consultation and support. For example, ICON successfully identified, modified and implemented two Bayesian model-assisted designs in a two-part, first-in-human trial of a tri-specific antibody, which targeted multiple tumour types. The first part of the phase 1 trial aimed to assess drug safety and toxicity, while subsequent dose-expansion cohorts obtained preliminary efficacy data.
Implementing BOIN in a dose-escalation study to assess drug safety and toxicity
The model-assisted BOIN design was selected for an open-label, dose-escalation study of the tri-specific antibody because initial information regarding the expected dose-toxicity curve was limited, and BOIN
-
is theoretically optimal at MTD selection
- requires less statistical processing than model-based designs
- is intuitive to investigators
Subsequent simulation assessment of the study design’s operating characteristics confirmed that the design would select the true MTD and allocate the most patients to dose levels nearest to the target toxicity rate.
After initiation of the trial, it became clear that the drug was benign, and that toxicity did not increase monotonically with dose, as is traditionally assumed. Through additional simulation, ICON identified and implemented modifications to the BOIN design that maximised safety and pharmacodynamic information without increasing patient numbers, costs or study timelines.
Implementing BOP2 in a dose-expansion cohort to obtain preliminary efficacy data
The BOP2 design was selected for dose-expansion cohorts of the phase 1 trial because it
- requires few patients
- can handle simple and complicated endpoints under a unified Bayesian framework
Since the dose-escalation study established that the drug was benign, the initial design was amended to better assess clinical activity. As with BOIN, the design's operating characteristics and analysis method for the different approaches were verified through simulation.
Smarter early phase trial designs continue to evolve
Although Bayesian-based BOIN and BOP2 methods are superior to rule-based designs, alternative strategies may be optimal based on a drug’s characteristics and the study objectives. For example, if a drug is initially known to be benign, then an alternative, more efficient design to explore safety and efficacy simultaneously, such as the BOIN-ET or BOIN 12 design, may be considered. These designs allow efficacy and safety to be studied simultaneously.
Historically, oncology drugs have been developed based on the belief that “more is better”. However, this presumption is not true for modern targeted drugs such as monoclonal antibodies. Subsequently Project Optimus, a cross industry/regulatory initiative, is encouraging sponsors to identify the optimal biological dose (OBD). Consequently, study designs such as the BOIN-ET and BOIN 12 are likely to become even more common in the future.
Studies that want to test a drug in patients with different types of cancer may also benefit from deploying basket designs within a master protocol when obtaining preliminary efficacy data. Extracting information across cohorts can also improve the efficiency of the statistical analysis using a Bayesian framework. As oncology therapeutics and early phase design models evolve, sponsors will benefit from working with a partner experienced in innovative adaptive designs for phase 1 and 2 oncology trials.
To learn more, please view our case study: Improving early phase oncology clinical trial design using Bayesian based BOIN and BOP2 designs or contact us to speak with our experts.