Phase I Dose Escalation Study Design: "3 + 3 Design"

For the first-in-human clinical trial, the dose escalation study design is often utilized. in dose escalation study, subjects are enrolled in cohorts (batches) with increasing doses. whether or not the study goes further to the higher dose depends on the assessment of the previous dose. The assessment is mainly based on the DLT (dose limiting toxicity) - side effects of a drug or other treatment that are serious enough to prevent an increase in dose or level of that treatment.

Dose escalation study can be with placebo control or without placebo control. For a dose escalation study with placebo control, the placebo control is within each dose cohort, not across the cohorts. A dose escalation study without placebo control is often used in studies for life-threatening diseases such as cancers and AIDS. 

The most common dose escalation study design is the rule-based "3 + 3 design". The "3 + 3 design" can be depicted as below: 

  The "3 + 3 design" is clearly explained in FDA guidance "Clinical Considerations for Therapeutic Cancer Vaccines" as following: 

The traditional standard dose escalation schedule in the development of cancer therapeutics uses the so-called “3 + 3 design” to avoid selection of a phase 2 clinical trial dose that causes a treatment-limiting toxicity in more than 17% of subjects, a standard considered acceptable as an outpatient therapeutic for patients with limited options and life-threatening diseases. In a “3 + 3 design,” three patients are initially enrolled into a given dose cohort. If there is no DLT observed in any of these subjects, the trial proceeds to enroll additional subjects into the next higher dose cohort. If one subject develops a DLT at a specific dose, an additional three subjects are enrolled into that same dose cohort. Development of DLTs in more than 1 of 6 subjects in a specific dose cohort suggests that the MTD has been exceeded, and further dose escalation is not pursued.

In summary, the "3 + 3 design" is:

• Rule based design
• Start by allocating lowest dose level to first cohort
• Adaptively escalate/de-escalate based on observed DLTs
• Repeat until MTD obtained or trial is stopped

The "3 + 3 design" is very straight forward, robust, simple and can be very well understood by clinicians and investigators. However, the "3 + 3 design" has its limitations. It is limitations may be summarized as the followings: 

• Ignores dosage history other than previous cohort
• Same action under qualitatively different situations (e.g., 0/3 and 1/6 lead to same action)
• Ignores uncertainty (if true DLT rate is p = 0.5, 11% of the time we will see 0 or 1 DLT in 6 patients)
• Cannot re-escalate
• Fixed cohort sizes (either 3 or 6)
• Pre-defined dose levels to be potentially tested
• Low probability of selecting true MTD
• High variability in MTD estimates
• MTD is not a dose with any particular probability of DLT, but in the range from 20% to 25% DLT.
• Can not estimate MTD with target probability of DLT less than 20% or greater than 33%.
• Not all toxicity data of all patients are used to determine the MTD.
• Many patients are likely to be treated at low doses.

The limitations are also cited in FDA's guidance "Clinical Considerations for Therapeutic Cancer Vaccines":

Many cancer vaccine trials have used the “3 + 3 design,” and the results show that, except in very rare situations, an MTD for a cancer vaccine may not be identified. In these trials, the dose-toxicity curve may be so flat that the highest dose that can be administered is limited by manufacturing or anatomic issues rather than toxicity. Therefore, this “3 + 3 design” may not be the most suitable approach to gathering information from early phase trials of cancer vaccines, and alternative trial designs should be considered.

Given the relatively acceptable safety profile of some classes of cancer vaccines, alternative dose-escalation approaches, such as accelerated titration or continuous reassessment, may be considered instead of the standard “3 + 3 design”. When using such designs, the protocol should describe acceptable parameters for the dosing endpoint (supported by data). Irrespective of which dose-escalation approach is chosen, the study protocol should clearly define DLTs, the subject “off-treatment” criteria, and the study stopping rules that will ensure subject safety. When no DLT is expected or achieved, optimization of other outcomes, such as the immune response, can be useful to identify doses for subsequent studies.

Because of these limitations, the alternative phase I dose escalation study designs have been proposed and used in the practice. These methods are:

• CRM (Continual Reassessment Method)
• Modified CRM 
• BLRM (Bayesian Logistic Regression Method)
• mTPI (Modified Toxicity Probability Interval Design)