The generalized 3+3 (G3) design for Phase I trials: A design for clinicians

Abstract

Purpose

To construct a simple and software-free dose-finding design that perform comparable to modern designs relying on statistical models.

Methods

The Generalized 3 + 3 (G3) design uses a set of arithmetic rules that depend on the number of participants enrolled and the number experiencing dose-limiting toxicity (DLT). Computer simulations are conducted based on a set of scenarios published in the literature and adopted by FDA.

Results

The G3 design provides identical decisions to 3 + 3 for 3 and 6 participants, and highly similar ones to modern designs. It performs better in selecting the MTD and patient allocation than 3 + 3 and comparable to modern designs. However, modern designs rely on specialized software while the G3 design can be generated by hand.

Conclusion

The G3 design extends the widely used 3 + 3 design by enabling flexible decision-making for any number of participants enrolled at a dose level. The G3 design is straightforward, transparent, and can be implemented without specialized software—making it an ideal option for clinicians.

Limitation

The G3 design is for classical dose-finding trials based on toxicity outcome. Modern therapeutics may rely on additional outcomes like efficacy or pharmacodynamic effects to define optimal doses.