The FDA has taken a clear position with Project Optimus in shifting toward more progressive tailored approaches while rejecting antiquated study designs to evolve clinical trial strategies to better align with newer drug classes.
The FDA released its draft guidance in support of Project Optimus in January 2023. It offers clear direction for industry stakeholders to unlock efficiencies in oncology drug and biologic development. The FDA’s draft guidance also underscores the benefits patient-centric approaches can have on quality of life, medication adherence, and long-term clinical outcomes.
Worth noting from the FDA document:
The guidance represents an increased awareness from the agency of the quickly changing oncology landscape: a shift from cytotoxic chemotherapy toward more targeted therapies, such as kinase inhibitors and monoclonal antibodies.1 It unequivocally calls for the robust use of pharmacokinetic, pharmacodynamic, and pharmacogenomic data as a core component of identifying the optimal dose. Significantly, the FDA’s draft guidance characterizes the traditional maximum-tolerated dose (MTD) approach as fundamentally unsound and inherently risky. Moreover, the draft guidance has stoked acceleration of more innovative trial designs.2
These designs should study multiple doses and be randomized. With its draft guidance, the FDA has placed adaptive trials and, more specifically, model-informed adaptive trial designs at the forefront of the minds of every oncology drug developer.
Since the FDA announced Project Optimus in 2021, a confluence of technological developments has altered the industry’s initial reaction and shaped its ability to respond. Major advances in artificial intelligence (AI) are leading to the availability of actionable and scalable solutions.
OpenAI’s initial release of ChatGPT in November 2022 and subsequent unveilings of GPT-3.5 and GPT-4 captivated our collective consciousness. Since then, ChatGPT has quickly infiltrated areas of drug development and discovery that are in desperate need of innovation to drive operational efficiency—identifying new drug targets, locating clinical trial volunteers, enhancing the informed consent process, and even reviewing scientific literature.3
ChatGPT also has put a focus on the exponential growth of deployed AI in clinical trials we’re about to experience. While the clinical trials segment of life sciences dominated in AI deployments in 2021—owing to the adoption of these technologies in clinical trial design, study adherence, and patient recruitment—the AI in clinical trials market is projected to reach $4.8 billion by 2027 at a CAGR of 25.6% during the forecast period.4
With more integral clinical trial applicability, sizable investments in capabilities such as deep learning (DL)/neural networks, machine learning (ML), and natural language processing (NLP) have begun to transform the time and effort it takes to gain meaningful insights from vast clinical, patient-reported outcomes (PRO) and real-world data sets.
How do the draft guidance and seemingly exogenous advancements in AI and ML act as catalysts for more efficient and patient-centric oncology drug development? To properly unpack this, it's worthwhile examining the rise of targeted therapies and the growing use of adaptive trials.
Targeted therapies (whether small molecule or monoclonal antibodies) have significantly impacted the landscape of oncology clinical trials by introducing more personalized and precise approaches to treatment. These therapies are designed to specifically target molecular alterations or pathways involved in cancer development and progression, offering the potential for enhanced efficacy and reduced toxicity compared to traditional chemotherapy. Drugs such as the widely lauded monoclonal antibody Herceptin target the cancer cells while largely leaving other healthy cells alone.
While targeted therapies offer great promise for patients, they create a multitude of challenges for drug developers. One of the most prominent struggles associated with successful targeted therapy involves optimal dose identification.5 Many targeted therapies have limited preclinical data, lack of disease and response characterization, and possess narrow therapeutic indices. This is further compounded by patient population heterogeneity, drug resistance mechanisms, and the absence of well-established biomarkers.
Collectively, these issues make dose optimization critical and, consequently, study design incredibly important. Dose-finding studies often employ model-based approaches that integrate pharmacokinetic/pharmacodynamic modeling (PK/PD) modeling, statistical analysis, and clinical judgment to determine the optimal dose. These complexities warrant bespoke approaches to study design, which translate into substantial manual effort.
Removing manual processes with the help of AI has the potential to unlock efficiencies and aid in making targeted therapies the engine of oncology development that Project Optimus intends. Parallel to the rise of targeted therapies, though, is the growing use of model-informed adaptive trials, which layer in their own complexities.
Unlike traditional fixed designs, adaptive designs provide the ability to make informed modifications throughout the trial. These modifications can include changes to sample size, treatment allocation, patient eligibility criteria, dosage regimens, or statistical methods.
Model-informed adaptive trial designs use mathematical models—often PK/PD models—to inform design, conduct, and analysis. PK/PD models are used to optimize dose selection, explore different dosing regimens, predict response rates, and identify patient subpopulations most likely to benefit from treatment. Overall, adaptive designs are ideal for achieving a better understanding of the dose-response or dose-toxicity relationship.6
As model-informed trial designs gain traction, the demand for rapid and efficient data analysis, interpretation, and communication to all key stakeholders will intensify. Artificial Intelligence (AI) can serve as a pivotal tool in streamlining this process. A prime example is the deployment of advanced language models and NLP technologies to expedite pharmacology analysis and simulations.
Utilizing these technological assets can dramatically reduce the programming effort required, thereby accelerating the time to actionable insights. Within the framework of Project Optimus, this has the game-changing potential to expedite our understanding of dose-response relationships and accelerate dose identification and optimization.
The injection of AI into model-informed adaptive trials empowers researchers and clinicians with data-driven tools, enabling more precise and efficient trial designs, improved patient stratification, enhanced treatment decisions, and increased overall trial success rates. AI’s ability to handle large datasets, identify complex patterns, and provide real-time insights makes it an invaluable asset in optimizing the adaptive trial process and advancing personalized medicine.
With Project Optimus, the FDA has taken a clear position. It solidifies the shift away from traditional chemotherapy and toward more progressive tailored approaches. It rejects antiquated study designs and offers a relevant primer on how to evolve clinical trial strategies to better align with newer drug classes.
The baton has now been passed to the life sciences industry to harness its resources to drive efficiencies in process and enable a patient-first mindset. AI arguably is the most essential tool at our disposal for unlocking those efficiencies and reaping the benefits of breakthrough science. It's up to us as stewards to ensure that the right drugs now reach the right patients at the right dose.
About the Authors
Sirj Goswami, PhD, is the CEO and co-founder, and Jason Rizzo, MBA, is the vice president of Global Biopharma Strategy for InsightRX.
References