Key Considerations for Drug Development Pipelines in Early Phase Clinical Trials

Drug development is a long and costly process. On average it takes 10-15 years for each new drug to be approved.

The average cost is $1 billion to $2 billion and just 1 in 10 drugs make it to final approval after the preclinical stage.¹ With the stakes so high, the early stages of clinical development are critical. Decisions made at this stage can determine the efficiency, cost, and ultimate success of an entire program.

In an evolving landscape of early phase drug development, sponsors need flexible yet robust statistical approaches to ensure promising compounds are evaluated effectively and efficiently. Getting this stage right offers the opportunity to not only accelerate timelines but also maximize the chances of bringing safe, effective treatments to patients.

The Aims of Phase I Trials: What Has Changed?

Historically, the primary aim of Phase I clinical trials was to find the maximum tolerated dose (MTD). This assumed more treatment equaled better efficacy and more toxicity.

To determine the MTD, studies considered dose-limiting toxicities (DLTs) within a pre-specified DLT window. Studies might also have considered tolerability and efficacy, but the primary focus was safety, with particular importance placed on minimizing the number of patients treated at ineffective or toxic doses.

However, statistical methodology has evolved beyond traditional rule-based designs such as the 3+3. In 1990, the first model-based design, the continual reassessment method (CRM) was proposed.²

This was followed by the appearance of further model-based and model-assisted designs. Over the past 15 years, we have seen growing uptake of model-based and assisted designs, more “how-to” papers appearing in the literature and increased understanding among the clinical trials community.

In 2021, the FDA determined a need to move away from considering only the MTD. This led to the launch of Project Optimus, designed to reform the dose optimization and selection paradigm in oncology.³

Resulting guidance recommends evaluating and selecting doses based on all available clinical data and a preliminary understanding of dose and exposure response, such as for safety, tolerability and activity.⁴ Alongside the changes in methodology, and naturally very closely linked, are changes in treatment.

If we look at the example of oncology, over the past 25 years, cancer treatment has evolved from generalized—often toxic—approaches to personalized and targeted therapies, including precision medicine and immunotherapy.⁵ These changes mean the assumptions we used to make do not necessarily hold true and our methodologies need to change.

The primary aim of a modern Phase I clinical trial is to find the optimal biological dose (OBD). We are no longer assuming more treatment equals better efficacy and more toxicity or only thinking about DLTs.

However, it is still crucial to limit the number of patients being treated with ineffective or toxic doses within a study. The key issues now are how to determine what makes a dose “optimal” and where to start when there are so many trial designs to choose from.

Types of Phase I Trial Design

There are three main categories of trial design—rule-based, model-based, and model-assisted. In rule-based designs, decisions are made on a pre-specified set of rules.

Examples include 3+3, Rolling 6 and i3+3.^6,7 Rule-based designs are the easiest designs to implement, but commonly found to be inferior to model-based and model-assisted designs in terms of their results.

In model-based designs, decisions are made based on the results of a model of the dose-toxicity, or efficacy, relationship. This is run at the time a new participant or group of participants requires dosing.

Examples include the CRM, Escalation with Overdose Control (EWOC) and the Bayesian Logistic Regression Model (BLRM).^8,9 Model-assisted designs are a combination of the two.

Decisions are made based on a pre-specified set of rules, but these are governed by an underlying model, with decisions laid out in advance. Examples include Keyboard design, similar to Modified Toxicity Probably Interval (mTPI-2), and Bayesian Optimal Interval design (BOIN).^10-12

The advantage of model-assisted designs is that they are easier to implement during the study than model-based designs, but they can also have excellent properties. With model-based and model-assisted designs, several valuable extensions are also possible.

These include going beyond safety alone by incorporating efficacy and other relevant endpoints, as well as time-to-event (TiTE) approaches that accommodate delayed toxicities or fast recruitment. In addition, combination designs can evaluate multiple doses, schedules, or treatments, and prior information can be formally incorporated to further improve trial efficiency and decision-making.

Designing an Early Phase Trial

When starting to design an early phase trial it is crucial to clearly define what you aim to achieve. Clinicians should expect and welcome detailed input from statisticians, as these discussions help refine objectives and guide the choice of the most appropriate design.

Effective communication and close collaboration are therefore more important than ever in early phase clinical research. The key inputs for Phase I trial designs are:

• study objective
• safety considerations and definition of DLTs
• the intervention and dose levels or schedules to be considered
• participant population and sample size
• stopping rules
• statistical methods
• simulations

Together, these elements help narrow down the most appropriate design options. The trial objective and safety considerations guide what is considered optimal.

This is now a balance of multiple factors depending on the context, treatment, and patient population, with considerations such as tolerability, efficacy, pharmacokinetics, and both short- and long-term toxicity.

Next, it is important to consider what is being tested and in whom. This begins with defining the intervention and dose levels, including whether outcomes occur early or late, whether doses can be ordered, and how the starting dose should be chosen.

These decisions further refine the design choices. We then think about subjects and sample size.

Key considerations here include the population of interest, how many patients can reasonably be included considering recruitment is typically slow in Phase I studies, and whether the cohort size is fixed or flexible. Protecting trial participants is also a central consideration, primarily addressed through stopping rules.

As well as not treating at toxic or ineffective doses, stopping rules can ensure efficiency by not enrolling patients unnecessarily and by reaching reliable conclusions as quickly as possible to enable progression to later-phase trials. Finally, statistical considerations are addressed, including dose escalation strategies, model specification, prior elicitation, and the use of simulations to evaluate and optimize the design.

Choosing the Right Phase I Design

Finding the right design is an iterative process. Once the key information above is assembled, you can identify one or more candidate designs for your study.

Simulation studies are then conducted to evaluate the operating characteristics. If these are acceptable, the design can be finalized; if not, the design is refined and the simulations repeated until a suitable approach is achieved.

By performing simulations of plausible scenarios—ranging from pessimistic to optimistic—we can assess how a trial will run. Thousands of simulations are run on each scenario to assess the operating characteristics, which include the probability of correct selection, the expected number of patients treated at each dose level, and the expected toxicity/efficacy at each dose level, among others.

Comparing the operating characteristics can help us to choose and refine our study design.¹³ We could, for example, compare four different designs across a range of scenarios and use the totality of that data to decide what is acceptable in practice.

Recent Developments in Early Phase Clinical Trials

It is an exciting time for early phase clinical trial design, with several recent developments which are changing our expectations and methodologies. The Methodology for the Development of Innovative Cancer Therapies (MDICT) Taskforce now recommends all Phase I trials of monotherapy agents plan to define a recommended dose range (RDR).¹⁴

The MTD can still be defined but it may not be the most useful outcome. The RDR includes the concept of an OBD but should include two or more dosages.

The recommendation is that it ranges from the lowest level where activity has been observed, to a higher active dosage, which may be the MTD. Another hot topic has been back-filling, particularly since it was recommended by Project Optimus.^15,16

This involves assigning subjects to existing dose levels to collect additional information on efficacy, pharmacokinetics, pharmacodynamics, etc. This is similar to expansion cohorts but can be done during dose-escalation.

It can help to identify patients on a dose efficacy curve and save time by avoiding pauses in recruitment. Project Optimus also refers to the use of randomization in order to minimize bias and promote comparability of patients.

This should not be used to replace a formal, well-powered, Phase II trial; however, it can be incorporated in different ways including randomized expansion of selected dose levels and randomized backfilling. Patient-reported outcomes (PROs) are currently not commonly incorporated into dose-finding trials.

Project Optimus recommends the consideration of PRO inclusion to enhance assessment of tolerability. A key challenge is that sponsors may not know which adverse events (AEs) to expect and therefore will be unable to select relevant measures or items, even when using established tools such as the PRO-CTCAE questionnaire.¹⁷

Conclusion

Early phase trial design is a rapidly changing area. In the past five years, we have seen this change accelerate even further as different trial designs are applied and Project Optimus published its recommendations.

Novel and innovative methods allow us to tailor our design to each specific study, to ensure we are getting the right answers. However, with so many options to choose from, it can be difficult to know which design is best for your trial.

This makes collaboration between clinicians and statisticians more important than ever. By working together, we can maximize our chances of success in early phase clinical trials and ultimately improve access to safe and effective treatments for patients.

About the Author

Sam Hinsley is a Statistics Manager with over 12 years of experience in the design, conduct, and analysis of phase I, II and III clinical trials. Sam’s expertise lies in oncology, with specialist knowledge in phase I trial design methodology. Sam is a member of the NIHR early phase clinical trials statistics group and has reviewed grant applications for CRUK Clinical Research Committee and NIHR Evaluation, Trials and Studies Co-ordinating Centre, as well as journal articles for numerous statistical journals such as the Lancet and Lancet Oncology. Sam has delivered UK-wide training on the Continual Reassessment Method and published multiple papers regarding the use of phase I trial design methodology.

References

1. Sun D, Gao W, Hu H, Zhou S. Why 90% of clinical drug development fails and how to improve it? Acta Pharm Sin B. 2022;12(7):3049-3062. doi:10.1016/j.apsb.2022.02.002
2. Sheiner LB. Learning versus confirming in clinical drug development. Clin Pharmacol Ther. 1997;61(3):275-291.
3. U.S. Food and Drug Administration. Project Optimus. Accessed March 26, 2026. https://www.fda.gov/about-fda/oncology-center-excellence/project-optimus
4. U.S. Food and Drug Administration. Optimizing the dosage of human prescription drugs and biological products for the treatment of oncologic diseases: guidance for industry. Published January 2023. Accessed March 26, 2026. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/optimizing-dosage-human-prescription-drugs-and-biological-products-treatment-oncologic-diseases
5. Rise in approved new cancer treatments in the last two decades. Airfinity. Published 2023. Accessed March 26, 2026. https://www.airfinity.com/articles/300-rise-in-approved-new-cancer-treatments-in-the-last-two-decades-with-us
6. Parulekar WR, Eisenhauer EA. Phase I trial design for solid tumors: the dose escalation paradigm. J Clin Oncol. 2004;22(7):1360-1365. doi:10.1200/JCO.2004.07.132
7. Le Tourneau C, Lee JJ, Siu LL. Dose escalation methods in phase I cancer clinical trials. J Natl Cancer Inst. 2009;101(10):708-720. doi:10.1093/jnci/djp079
8. Storer BE. Design and analysis of phase I clinical trials. Biometrics. 1989;45(3):925-937.
9. Rogatko A, Babb JS, Wang H, Slifker MJ, Hudes GR. Patient characteristics compete with dose as predictors of toxicity in early-phase trials. J Clin Oncol. 2007;25(6): 627-633.
10. Yuan Y, Hess KR, Hilsenbeck SG, Gilbert MR. The keyboard design: a novel Bayesian toxicity probability interval design for phase I clinical trials. Clin Cancer Res. 2017;23(15):3994-4003. doi:10.1158/1078-0432.CCR-17-0264
11. Iasonos A, O’Quigley J. Adaptive dose-finding studies: a review of model-guided phase I clinical trials. J Clin Oncol. 2016;34(20):2396-2403. doi:10.1200/JCO.2015.65.0218
12. Liu S, Yuan Y. Bayesian optimal interval (BOIN) design for phase I clinical trials. MD Anderson Cancer Center. Published 2015. Accessed March 26, 2026. https://odin.mdacc.tmc.edu/~yyuan/Software/BOIN/paper.pdf
13. Zhou H, Yuan Y, Nie L. Accuracy, safety, and reliability of novel phase I trial designs. Front Oncol. 2021;11: 684853.
14. Early phase cancer research ready for a change. European Society for Medical Oncology. Published 2023. Accessed March 26, 2026. https://dailyreporter.esmo.org/esmo-targeted-anticancer-therapies-congress-2023/esmo-tat-congress-2023/early-phase-cancer-research-ready-for-a-change
15. Brock K, Billingham L, Copland M, Siddique S, Sirohi B, Yap TA. Implementing innovative dose-finding designs in early-phase trials. Br J Cancer. 2024;130: (pagination pending).
16. U.S. Food and Drug Administration. Project Optimus: reforming the dose optimization and dose selection paradigm in oncology. Published 2023. Accessed March 26, 2026. https://www.fda.gov/media/164555/download
17. Iasonos A, Wilton AS, Riedel ER, Seshan VE, Spriggs DR. A comprehensive comparison of the continual reassessment method to the standard 3+3 design. J Clin Oncol. 2008;26(20): 3476-3481.