Simon's Two-Stage Design: A Smarter Way to Design Single-Arm Phase II Clinical Trials

Why Simon's Two-Stage Design?

Designing a Phase II oncology study often involves balancing two competing goals. Researchers want to identify promising treatments as quickly as possible while avoiding unnecessary exposure of patients to therapies that are unlikely to work. One statistical approach has remained the standard for more than three decades because it addresses exactly this challenge: Simon's two-stage design.

Traditional single-stage trials require recruiting the full planned sample size before any decision can be made. If the treatment is ineffective, many patients may receive a therapy with little chance of benefit.

Richard Simon introduced the two-stage design in 1989 to solve this problem. The idea is simple: enroll an initial group of patients, evaluate the observed responses after Stage 1, stop early if the treatment appears ineffective, and continue to Stage 2 only if the interim results are sufficiently promising.

This approach reduces the expected number of patients treated under ineffective therapies while maintaining predefined Type I error and statistical power.

How Does It Work?

A Simon two-stage design is defined by four key parameters: n1, the number of patients in Stage 1; r1, the early stopping boundary; n, the total sample size; and r, the final success boundary.

Before designing the trial, investigators specify the null response rate, p0; the target response rate, p1; the Type I error, alpha; and statistical power, 1 - beta.

The statistical algorithm then searches all feasible designs and identifies one or more optimal solutions.

Minimax vs. Optimal Designs

Simon's method usually produces two recommended designs. The minimax design minimizes the maximum total sample size. It is often preferred when patient recruitment is difficult, rare diseases are studied, or total trial cost is a major concern.

The optimal design minimizes the expected sample size when the treatment is ineffective. This design may recruit slightly more patients in the best-case scenario but often stops earlier for ineffective treatments.

Neither design is universally superior. The best choice depends on the study objectives and practical constraints.

Looking Beyond Sample Size

Although sample size is important, experienced trialists often consider additional operating characteristics, including probability of early termination, expected sample size under the null hypothesis, expected sample size under the alternative hypothesis, overall efficiency, practical feasibility, and recruitment timeline.

Comparing several candidate designs side by side often provides better insight than focusing on a single recommended design.

Common Challenges

While the statistical theory is well established, practical implementation can still be challenging.

Researchers frequently ask which candidate design best matches a study, why several designs are considered admissible, how large the tradeoff is between sample size and early stopping, how sensitive the design is to small changes in response rates, and whether published Simon designs can be reproduced.

Answering these questions typically requires more than simply generating one design table.

Modern Tools Can Simplify the Process

Several statistical packages can generate Simon two-stage designs, particularly within R. Modern software can perform exhaustive searches across feasible two-stage designs, compute operating characteristics exactly, and present optimal, minimax, and admissible candidates for comparison.

Interactive visualization further helps investigators understand the tradeoffs between early stopping probability, expected sample size, and maximum sample size before selecting a design.

Recently, web-based tools have started providing these capabilities in a more accessible format. For example, BioStatHub offers an interactive environment for exploring Simon two-stage designs, comparing admissible solutions, visualizing design frontiers, and exporting publication-ready reports. The emphasis is on helping researchers understand the design process rather than simply producing numerical output.

If you are planning a Phase II trial, you can explore multiple candidate Simon designs interactively, compare operating characteristics, and export publication-ready reports using BioStatHub.

Final Thoughts

More than thirty years after its introduction, Simon's two-stage design remains one of the most widely used methods for single-arm Phase II clinical trials. Its combination of statistical efficiency, ethical patient protection, and practical simplicity explains its continued popularity across oncology and other therapeutic areas.

Whether you generate designs using statistical software or an interactive web application, taking time to explore multiple candidate designs and understanding the tradeoffs behind them can lead to more informed and defensible clinical trial decisions.

Frequently Asked Questions