A clinical development strategy is a comprehensive plan designed to establish the safety and efficacy of new therapeutics. Developing an effective plan requires multidisciplinary expertise and adapting to accumulating learning and changes in clinical practice and the market environment.
Effective clinical development strategy is adaptive by nature. Though it mainly focuses on achieving regulatory approval, it also needs to pave the way to reimbursement and integration of new therapeutics into clinical practice. Clinical development strategy is designed to meet the goals outlined by the TPP (Targeted Product Profile). TPP and clinical development strategy are interdependent: accumulating clinical data may lead to modifications of the TPP, and clinical development strategy needs to be adapted if the TPP evolves due to changes in the regulatory, financial, and competitive landscape.
Ideally, clinical development strategy is based on scientific, clinical, and regulatory considerations. However, clinical developers need to consider business and financial aspects, such as risks, runways, costs, and time. The trade-offs and their impact differ in early development and confirmatory settings.
Early clinical development: Trade-offs and decision-making
The main goal of Phase 1 trials is to assess the drug’s safety, tolerability, pharmacokinetics, and pharmacodynamics. For most therapeutics, Phase 1 trials are typically conducted in healthy volunteers, while Phase 2 studies are conducted in patients and are designed to evaluate efficacy and further evaluate safety.
However, depending on the targeted conditions and the nature of therapeutics, Phase 1 trials can also be conducted in patients. Examples of such targeted conditions include oncology and rare genetic diseases. Examples of therapeutics include gene therapies and treatments with expected serious side effects, such as cytotoxic therapies.
Early oncology development
Quantitative strategies can enable more efficient clinical development plans. For example, early oncology development often requires trade-offs between multiple indications, dose levels, combinations of therapies and costs, and times to decision driven by various milestones and runway.
In the dose-escalation stage, modern designs such as Bayesian Optimal Interval (BOIN), Modified Toxicity Probability Interval (mTPI), i3+3, or Bayesian Logistic Regression Model (BLRM) designs allow a more flexible sample size and enable a more precise estimate of toxicity.
In the dose-expansion stage, continuous evaluation of accumulated data against a historical control can inform efficient decision-making. Such monitoring is based on Bayesian probability of success and can be performed as often as desired if it is prespecified. If the data indicate that the probability of success in the expansion cohort is dropping too low, enrollment can be stopped, and resources can be reallocated to more promising cohorts.
Another way of improving the efficiency of dose-expansion studies is a modern design using Bayesian Hierarchal Models. These models account for similar hypotheses, such as with respect to disease (e.g., first- and second-line treatment in the same tumor) and the regimen (e.g., monotherapy vs. combination therapy). Bayesian Hierarchical Models enable more efficient use of data to inform decisions, can reduce sample size and provide options for cost savings or an additional cohort.
Rare disease development
In rare diseases, when studies have to be done in a small population with life-threatening or debilitating diseases, finding efficacy signals or deciding on futility as early as possible helps to protect patients from futile interventions.
In single-arm studies, Bayesian interim monitoring based on predictive probabilities against a reference efficacy benchmark can inform more efficient decision-making. For example, if the probability of observing a targeted efficacy at the end of the study is low, the study could stop for futility; if the probability is high, the decision can be made to invest in accelerating enrollment (e.g., open new sites) or accelerating the preparation of the next phase; if the probability is between low and high thresholds, no decision is made until the next interim analysis. These thresholds for predictive probabilities can be determined by simulations depending on key assumptions and the company’s strategic priorities.
In randomized studies, Bayesian borrowing for the control arm based on external data can reduce the sample size and allow more patients to be randomized to the active treatment. The Bayesian framework provides several methods that can be used for borrowing, such as robustified meta-analytic priors, power priors, and commensurate priors.
Other considerations
Independent of the therapeutic field, under the challenging financing for biotechs, there is increasing pressure to demonstrate efficacy signals as early as possible. In Proof-of-Concept (POC) studies, using the same clinical endpoints necessary to demonstrate efficacy in Phase 3 is rarely possible due to smaller sample sizes and limited study duration. Therefore, the efficacy signal finding in POC relies on surrogate endpoints, biomarkers, devices/wearables, and other signals. Modern quantitative methods — from statistical models to AI/ML — can also support your evaluation of efficacy and safety signals. For example, AI/ML can be better suited to dealing with multidimensional and non-linearity, and can be used for pattern identification, dimensionality reduction, classification, and prediction.
In some rare diseases and other debilitating conditions with high unmet clinical needs, surrogate endpoints, biomarkers, and wearable-based endpoints can be used for registrational trials. Suppose your clinical development strategy foresees registrational study with such an endpoint; it is essential that you include a generation of evidence to support such endpoint as fit-for-purpose in your plan, which can be based on external data as well as data generation in early phases of your therapeutic’s development.
Clinical development programs can benefit from optimized interim monitoring in any therapeutic field. One way to optimize interim trial monitoring is to use the Bayesian framework to inform decisions. Interim analysis in the frequentist framework usually evaluates the treatment benefit based on completed patients. However, a more relevant question is whether a trial is likely to reach a definitive conclusion by the end of the trial, which is inherently a prediction problem and can be better addressed by Bayesian predictive probability. It is important to note that Bayesian methods can be used for the interim analysis of frequentist study, i.e., the final analysis can be done using a conventional frequentist framework.
Read Part 2 of this article exploring late-stage clinical development strategy.
Interested in learning more? Download our new ebook, Adaptive Trial Design, which outlines common adaptive trial designs, the benefits of adaptive trials, how to optimize your adaptive trial, and a ten-point framework to determine if your trial should be adaptive.
About Natalia Muehlemann
Natalia Muehlemann is Vice President, Clinical Development, at Cytel. Natalia Muehlemann, MD, MBA, has over 20 years of experience in general management, clinical development, and business development in the life sciences. Dr. Muehlemann joined Cytel in 2020, and prior to Cytel, served as Global Category Head, Acute Care - Oncology - Devices at Nestle Health Sciences. She acts as an Expert Jury member for the European Commission’s Innovation Council. Dr. Muehlemann holds an MD and an MBA (IMD) and professional certifications in statistics and data science.
Read more from Perspectives on Enquiry and Evidence:
Sorry no results please clear the filters and try again
How to Ensure Your Adaptive Trial Is Appropriate for Regulatory Submission
News from ESMO: Challenging the Status Quo of Early Phase Clinical Trial Design — Moving from “Why” to “How”
Bayesian Strategies in Rare Diseases
