Sample-Size & Randomization Strategy

Determine sample size and recommend a randomization strategy for a clinical trial.

View Source YAML

---
name: Sample-Size & Randomization Strategy
version: 0.1.0
description: Determine sample size and recommend a randomization strategy for a clinical trial.
metadata:
  domain: scientific
  complexity: medium
  tags:
  - biostatistics
  - sample-size
  - randomization
  - strategy
  requires_context: false
variables:
- name: dropout_rate
  description: The dropout rate to use for this prompt
  required: true
- name: response_rate_active
  description: '`{{response_rate_control}}`'
  required: true
- name: response_rate_control
  description: '`{{dropout_rate}}`'
  required: true
model: gpt-4o
modelParameters:
  temperature: 0.2
messages:
- role: system
  content: 'You are a senior biostatistician at an international CRO following ICH E9(R1) and regulatory guidance.


    Reason step by step internally but present only the final answer.'
- role: user
  content: '1. Review trial specifics such as indication, phase, and primary endpoint.

    2. Calculate the minimum total sample size to achieve at least 90 % power given assumed response rates and drop-out rate.

    3. Recommend a stratified block-randomization scheme with block size range, stratification factors, and generation method.

    4. Explain any sensitivity or re-estimation options.

    5. Provide R code using `pwr` or `power.prop.test` and `randomizeR` with inline comments.

    6. Summarize key references to statistical guidance.


    Inputs:

    - `{{response_rate_active}}`

    - `{{response_rate_control}}`

    - `{{dropout_rate}}`


    Output format:

    Executive summary (≤150 words) followed by two tables: sample-size scenarios and randomization parameters. Conclude with
    a fenced R code block.'
testData:
- vars:
    response_rate_active: example_response_rate_active
    response_rate_control: example_response_rate_control
    dropout_rate: example_dropout_rate
  expected: 'Executive summary (≤150 words) followed by two tables: sample-size scenarios and randomization parameters. Conclude
    with a fenced R code block.'
evaluators:
- name: Output starts with 'Executive summary'
  string:
    startsWith: Executive summary