Project Optimus Series | The Role of Biostatistics

As regulatory expectations evolve under the FDA’s Project Optimus oncology dosing initiative, biostatistics is emerging as a central pillar in designing and executing trials that move beyond the traditional maximum tolerated dose (MTD) approach.

In this fourth episode of our Project Optimus series, host Dr. Wael Harb is joined by biostatistics expert X.Q Xue, PhD, Vice President and Global Head, Biostatistics at Syneos Health to explore how statistical science is transforming dose optimization in oncology drug development. Dr. Xue discusses the limitations of legacy 3+3 dose-escalation designs and introduces innovative alternatives, including Bayesian modeling, adaptive trial strategies and randomized parallel dose-response studies, which support more precise dose selection and can ultimately improve patient outcomes and trial efficiency.

Together, Drs. Harb and Xue examine how smaller biotech companies can overcome barriers to implementation, the role of simulation and AI in trial planning and how a biostatistics-driven approach may increase the likelihood of late-phase success, reduce post-marketing adjustments and support faster regulatory approvals.

The views expressed in this podcast belong solely to the speakers and do not represent those of their organization.

If you want access to more future-focused, actionable insights to help biopharmaceutical companies better execute and succeed in a constantly evolving environment, visit the Syneos Health Insights Hub. The perspectives you’ll find there are driven by dynamic research and crafted by subject matter experts focused on real answers to help guide decision-making and investment. You can find it all at https://www.syneoshealth.com/insights-hub.

Like what you’re hearing? Be sure to rate and review us!

We want to hear from you! If there’s a topic you’d like us to cover on a future episode, contact us at podcast@syneoshealth.com. 

Transcript

SPEAKER_02 0:01

Welcome to the next episode in our Project Optimus podcast series, where we explore the evolving landscape of dose optimization in oncology drug development. In the first episode, Dr. Nick Kenny and I built the foundation an overview of Project Optimus. Then in the second one, we met with our regulatory expert, Zura Lomri, and discussed the regulatory perspective. With the latest episode, I was joined by Patrick Melvin who discussed the operational challenges to implement Project Optimus. Today, we turn our focus to biostatistics, an essential pillar for dose optimization. As we all know, Project Optimus requires a more rigorous data-driven approach to dose selection, moving beyond the traditional MTD-based model. Previously, with chemotherapy-based drug development, we would dose until we get to the highest dose double, what we call maximum-tolerant dose or MTD. But with a lot of biological, that's not necessary on only add more toxicity. With that, we have to think about a comprehensive dose response evaluation and that will have biostatistical implications. Joining me is Dr. XQ Chula, a biostatistics expert who has worked extensively on designing adaptive travel strategies and optimizing dose exposure response analysis. The role of biostatistics in Project Optimus. Next on the Senius Health Podcast. XQ, thank you for joining us.

SPEAKER_01 1:33

Thank you, Dr. Hobb, for having me. I'm very excited to be part of this very important and exciting conversation. Project Optimus has significantly changed the way that how we plan and approach the Optimus selection and determination in the early oncology study from design and operation. And I'm looking forward to explore its impact on the trial design and discussing more rigorous statistical methodologies to deal with the emerging requirement to fulfill the dose optimization in the early oncology study.

SPEAKER_02 2:13

Absolutely. As you said, biostatistics plays a critical role in dose optimization under Project Optimus. Can you walk us through how biostatistics is used to inform dose selection and dose response evaluation?

SPEAKER_01 2:29

Absolutely. Traditionally, early oncology studies focus purely on looking for the maximum tolerability dose, MTD, by only monitoring the toxicity data using the very conventional and rigid WAI 3 plus 3 dose escalation algorithm, which relies a small number of cohort patients and a very simple dose escalation, de-escalation algorithm to determine the MTD. This algorithm has long been criticized with its poor performance of finding the toxic dose with higher probability treating the patient with suboptimal or overdosing, and with very limited dose response data provided from this process. With product optimus, we now are guided to incorporate more advanced biostatistical methodology to determine the optimal biological dose instead of only finding the MTD. To better fulfill the dose selection and control the BIOS, we use randomized parallel dose response trials to compare multiple dose levels head-to-head, generating robust efficacy, safety, and toxicity data. Integrating longitudinal PKPD modelings will now allow us to continuously monitor and analyze those pharmacodynamics data and pharmacokinetic data. Allow us to model those exposure relationships dynamically rather than solely rely on the fixed-dose cohort and toxicity data only. Multiple advanced adaptive designs have been developed lately that include Bayesian method. This advanced modeling enables us real-time decision making, adapting those strategies based on the emerging trial data, which can reduce unnecessary patient exposure to suboptimal doses or overdose levels.

SPEAKER_02 4:46

So you mentioned that the limitation of 3 plus 3 dose escalation model, and we are no longer only looking for the highest tolerated dose, we are looking at several factors in addition to safety. We are exploring PK, PD, and efficacy. What are the alternative designs that you would recommend to fulfill the requirement from Project Optimus?

SPEAKER_01 5:12

Again, I wanted to highlight several limitations of the traditional 3 plus 3, those escalation-de-escalation algorithm. It has been criticized for multiple years with its limitation. The overall objective of the traditional 3 plus 3 is only to look for the MTD. There is nothing related to the most effective or most safety data. And it does not have explicit targeted toxicity rate. And it's only with very rigid escalation, de-escalation rules in the process. It has per adaptability. Make it very difficult to access the full therapeutic window. And it potentially dose the patient with overdosing or under therapy dose with not necessarily high probability and its lack of dose response characterization. It does not incorporate anything related to the dose response relationship. So in the last two decades, researchers and statisticians worked together and invented multiple alternative dose escalation de-escalation algorithms to provide better performance, more precise targeting toxicity rate, and more flexible patient cohort size. While it still maintain that simple and easy adaptation into the trial operation. This method including Bayesian optimal interval, known as the Boeing design, and modified toxicity probability interval, or MTPI, or MTPI2. And Bayesian logistic regression, BRM, this algorithm provided much more precise dose selection with an overdose control. As we start to move along for the dose selection in the early oncology study away from the toxicity only focused algorithm into the multi-dimensional data, including efficacy, toxicity, PK, and PD. Multi-algorithm has been provided to do the dose selection based on the escalation and de-escalation by monitoring both toxicity and efficacy. That includes Boeing 12 design and Redus design, which was proposed initially by Dr. Yimunowa from UNC Chapakill and myself. That algorithm provided the continuous dose explorer justification based on the real-time toxicity, efficacy, and safety.

SPEAKER_02 8:06

These novel designs that implement efficacy endpoint along with safety in dose escalation are very intriguing and would help us find the optimal dose quickly. The question is: have there been any challenges using these designs and what's your experience with them?

SPEAKER_01 8:24

The algorithm by incorporating both advocacy and the safety into the dose escalation, de-escalation for the decision making definitely provide a lot of benefit. But in the same time, there are challenges, as you mentioned. One potential challenge is the longer follow-up time for the advocacy to emerge during the trial compared to the safety data or toxicity data. Another dimension of the challenge is you're gonna have a good understanding of the benefit and risk between the toxicity and the efficacy so that you can set up a more robust futility function to take into account the benefit and the risk into that algorithm for the toxicity data and the efficacy data. More importantly, so while you are monitoring both toxicity and efficacy, the idea is to find the most biological active dose for your later phases. In this type of algorithm, does not incorporate into the randomization still, your lack of the randomization, which potentially introduce the bias. So it could be challenged from the BIOS perspective to further justification of your selected dose for your recommended dose in the later phase. Those are the top level of the challenges that I think in my mind.

SPEAKER_02 10:13

You mentioned the randomization part where we were able to compare two or more dose levels in a very well-defined cohort, tumor type, and line of therapy to look at safety, efficacy along with PKPD. What are the statistical considerations here about the sample size? How many patients per cohort is that related to the type of asset or product being developed? That's a question that comes a lot from drug developer. How many patients that they will have to randomize to be able to tell the difference between the two dose levels?

SPEAKER_01 10:50

That's a great question, Dr. Harp. If you read the FDA's early on college dose optimization guidance for the industry, which published in 2024, it provides the language, guide the industry to design the trial, which stated this way: it allows a sufficient assessment of the safety and anti-tumor activity for each dose level, while it does not require the trial to be powered to demonstrate statistical superiority among the dose levels that you are comparing. My interpretation is there is no mandatory requirement to fulfill certain statistical power to compare between the dose levels. But in the same time, it challenges the sponsors to look for the robust size and the robust strategy and the proper method to come up with a methodology to provide sufficient evidence to differentiate between the dose levels so that you can make a proper recommendation for the dose level you selected for the later phase. With that said, it does not require specific sample sign and the power calculation. But typically, what I have seen somewhere between 15 patient to 30 patients has been showing up in the protocols to accommodate the requirement of advocacy data and safety data to make a proper recommendation for the selected dose.

SPEAKER_02 12:36

Very insightful. For smaller biotech companies, implementing these advanced statistical methods can be challenging due to resource limitations. How can they overcome these challenges?

SPEAKER_01 12:50

Smaller companies can leverage biostatistics-enabled CRO partners. Working with contract research organizations that specialized in the adaptive design, PKB modelings can help bridge the expertise gap. Use of AI-driven analytics for the dose optimization. Machine learning algorithms can simulate the dose response relationships, reduce the need for the large patient cohort. Open source statistic tools, platforms like RShining offers cost-effective and flexible modeling solution for the dose response analytics. Optimize the pre-trial simulation. Often we would like to know having a better estimate of number of the subjects needed and study duration so that I can answer the study objective research question, which is rather challenging as there are multiple factors involved. It is always helpful to run pre-trial simulation with the best guess of the potential scenario of dose toxicity relationships, dose response relationships coupling with the trial operating data. For instance, the distribution of the site selection window, the distribution of the site initiation window, and the distribution of the enrollment process, where you could have an intuitive realization of the potential outcome and the different assumptions, which oftentimes are very valuable to help you plan and select the best strategy to help you to find the optimal dose quicker and less cost.

SPEAKER_02 14:34

Looking ahead, what do you see as the long-term impact of biostatistics-driven dose optimization oncology?

SPEAKER_01 14:41

The impact will be very transformative. What I can see with this new paradigm requirement from the regulatory or guidance, we will have a higher successful rate in the later stage trial. With more optimized early phase dose selection, it gives us a better idea about the dose level that we want to bring into the later phase. It provides more robust safety and efficacy data. It definitely improves the probability of success for your phase 3 LART study and will have a better patient outcome and quality of life. Patients will receive the proper dose based on their PKPD profile and will improve the adherence and minimize the toxicity during the trial. And will reduce the cost. More precise dose will reduce the need for expensive post-marketing dose adjustification that typically will happen in the past. And I believe it also means the stronger regulatory confidence. While statistically driven model improves the data reliability and leading to the faster regulatory approvals and put our drug into the market for the patient access earlier, safer, and cheaper.

SPEAKER_02 16:04

These are great points. The ultimate outcome is finding the dose that is effective but less toxic for patients and also reducing failure in later development because if we have not selected the right dose. This has been an incredibly insightful discussion. Biostatistics is at the heart of dose optimization, ensuring that oncology treatments are not only effective, but also tolerable and sustainable. Thank you very much for sharing your expertise.

SPEAKER_01 16:34

Thank you, Dr. Harper, for having me. It's really an power to discuss how we use the biostatistic knowledge to optimize the dose selection. And a really, really nice opportunity to chat with you.

SPEAKER_00 16:48

That's all for today's episode of the Senior's Health Podcast. I'm your host, Nick Kenny. If you have other topical issues you'd like to hear on this podcast, please do send us a message at podcast at Senioshealth.com. And for access to more future-focused, actionable life science insights, please visit the CENES Health Insights Hub at insightshub.health. Shortening the distance from lab to life.