5 Key Takeaways: Insights on Alternative Designs to the Traditional 3+3 Design in Phase 1 Dose Escalation Studies

Traditionally, Phase 1 trials commonly utilize 3+3 designs to determine the maximum tolerated dose (MTD) and the recommended Phase 2 dose (RP2D). Studies have shown, however, that two out of three trials employing a 3+3 design failed in identifying the MTD, and better approaches are needed.¹

During Premier Research’s recent webinar Alternative Designs to the Traditional 3+3 Design in Phase 1 Dose Escalation Studies, Abie Ekangaki, Vice President, Statistical Consulting, and Andreas Schreiner, Vice President, Medical Affairs, Neuroscience & Analgesia, discuss alternative dose-escalation paradigms introduced into the clinical trial landscape for Phase 1 trials. In this blog post, we share five of their key insights on alternative dose escalation strategies for Phase 1 studies. To watch the full recording, click here.

1. Improving the design of phase 1 studies is essential for later-stage success. Overall, the likelihood of approval from Phase 1 is less than 10 percent. The low success rate of Phase 2 trials — less than 31% across all indications — can largely be attributed to failure to identify an optimal therapeutic dose.² Research has shown that rules-based designs such as 3+3 may result in a higher than expected proportion of study participants treated at subtherapeutic doses.¹
2. Model-based strategies offer advantages over rules-based designs. Unlike rules-based designs, model-based strategies do not require fixed cohorts, allowing for greater flexibility. They also use objective, statistical approaches for determining the MTD. These approaches are based on either pre-specifying toxicity probability priors at the start of the study or defining a priori Bayesian probability parameters incorporated into the models. Moreover, model-based designs are subject to lower sampling variability and a higher probability of estimating and selecting the true MTD.³
3. The continual reassessment method (CRM) may accelerate determination of the MTD. Among the most commonly used model-based designs, CRM can be applied to any cohort size and expresses the probability of toxicity as a function of dose level. Using a pre-specified target toxicity rate, this method uses statistical model-based dose-escalation algorithms to help estimate the MTD. Progressive algorithms allow a change in dose level after each patient is treated, so dose escalations and determination of an MTD or RP2D may occur more quickly.
4. The modified toxicity probability interval (mTPI) method simplifies dose-escalation decisions. The mTPI design utilizes a Bayesian statistical framework to compute the posterior probabilities of three intervals — an under-dosing interval, an equivalence interval, and an over-dosing interval — upon which escalation and de-escalation decisions are made. This model simplifies decision-making as these decisions can be tabulated at the start of a trial. It is, however, flawed in its approach to calculating the likelihood of an interval containing the observed toxicity rate, which involves dividing the estimated probability of toxicity by the length of the interval. A modification of this model, known as mTPI-2, seeks to address this flaw.
5. The Bayesian optimal interval method (BOIN) may offer the highest chance of selecting a therapeutic RP2D. A hybrid between rules-based and model-based design, this method is a generalization of the 3+3, accelerated titration, and Rolling-6 designs. BOIN compares the observed toxicity rate among a cohort with a set of pre-specified lower and upper toxicity boundaries to determine escalation and de-escalation decisions. If the observed toxicity rate is below the lower boundary, the dose is escalated. If it is above the upper boundary, the dose is de-escalated. In addition to being simple to implement, BOIN is also useful for estimating the cumulative probability of toxicity.

To learn more about the design and implementation of model-based dose-escalation designs, watch the full webinar “Insights on Alternative Designs to the Traditional 3+3 Design in Phase 1 Dose Escalation Studies” here.

¹ Reiner E, Paoletti X, O’Quigley J. Operating characteristics of the standard phase I clinical trial design. Comput Stat Data Anal. 1999;30(3):303-315.

² BIO. Clinical Development Success Rates 2006-2015.

³ Wheeler GM, et al. How to design a dose-finding study using the continual reassessment method. BMC Med Res Methodol. 2019;19(1):18.