3+3 Design Example
Despite the widespread use of conventional 3+3 designs, new adaptive trial design methods may offer better operating characteristics. For example, the accelerated titration and modified continual reassessment designs allow for smaller cohort sizes and more flexible enrolment decisions.
These designs also provide additional information about pharmacokinetic interpatient variability. However, these methods can lead to accrual of patients to dose levels below the recommended dose for phase II trials.
Dose escalation is the process by which a clinical trial increases the dose of a drug or vaccine. Typically, three patients are enrolled into a given dose cohort, and if no dose-limiting toxicities are observed, the trial proceeds to enroll additional patients.
Once the MTD is identified, the trial can be stopped when specified stopping criteria are met. These criteria include: a defined number of subjects treated at a certain dose level; the width of the Bayesian credible interval for the MTD at a particular dose level; or the estimated probability that the next patient will not experience DLT at the current dose level.
The 3+3 design is the most popular phase 1 design and is widely used in oncology trials. However, it is not without its problems. It exposes a large number of patients to subtherapeutic doses, and it does not provide a reliable estimate of the MTD (Hansen et al., 2014). Alternatives to the conventional 3+3 design, such as accelerated titration and continuous reassessment, have been proposed.
Many phase I clinical trials use the traditional 3+3 design, but it may not be optimal for every trial. In addition to being a memoryless design, it may not allow for adequate titration of doses that are below the probability of severe toxicity threshold. Other design methods have been developed, such as the accelerated titration designs and the pharmacologically guided dose escalation (PGDE) method. The latter uses a prespecified plasma drug exposure and a patient’s pharmacokinetic data to determine the next dose level.
The PGDE method is computationally simpler than the 3+3 design and can be used to titrate drugs without measuring their blood levels. However, it requires the patient to attend the site and provide pharmacokinetic information in real time, which can be difficult for some patients. This has led to concerns that the PGDE design is not ethically appropriate. However, simulations have shown that it performs well when compared to other model-based designs.
The traditional 3+3 design is the most common dose-finding method used in oncology clinical trials. Its popularity is due to its simple implementation and relative safety. However, there are several drawbacks to this tried-and-true design method. For example, a large proportion of patients are exposed to low (potentially subtherapeutic) doses and few are treated at or near the recommended phase 2 dose. Moreover, the 3+3 design lacks a strong statistical foundation and results in lower precision. Despite these disadvantages, many oncologists continue to use the 3+3 design because it is easy to implement and clinically friendly.
To avoid this problem, researchers can apply a model-based design. These designs are based on the assumption that the dose-toxicity curve remains flat and that the time to a DLT is constant. However, these models require more assumptions than a 3+3 design and can take longer to complete. These models have also been shown to outperform the 3+3 design in terms of safety and reliability.
The traditional 3+3 design is the prevailing dose-finding method for phase I oncology trials. This rule-based design does not require modeling of the dose-toxicity curve and is relatively simple to implement. In addition, the accrual of three patients per dose level provides information about pharmacokinetic interpatient variability. However, the use of this trial design has some significant drawbacks. For example, it leads to a large proportion of patients being treated at low (ie, potentially subtherapeutic) doses while few receive the recommended phase II dose.
The use of the 3 + 3 design has been questioned due to its poor statistical properties, its lack of flexibility to target alternate DLT rates and its limited duration of observation. As a result, the conventional approach often selects a dose that can be unacceptably toxic when it is studied in larger groups of patients. Alternative trial designs are needed to identify the optimal dose for cancer vaccines. These include the interval i3+3 and mTPI-2 designs.