Clinical Trial P-Value Explorer (AI-Assisted)

This application allows users to input a clinical trial research idea (or a URL with relevant content) and receive an AI-generated initial estimate for the total number of participants (N) and a plausible Cohen's d effect size. Users can then interactively adjust these parameters (N and Cohen's d) to see the corresponding p-value, calculated assuming a two-sample t-test with equal group sizes.

Disclaimer: The AI's estimates for N and Cohen's d are very rough approximations and intended for exploratory and educational purposes only. They should not be used for actual clinical trial planning without consultation with a qualified statistician. The primary utility of this tool is the interactive p-value calculator.

Project Structure

• api.py - FastAPI backend application
• app.py - Streamlit frontend application
• requirements.txt - Python dependencies
• .env - For API Key (you need to create this)
• README.md - This file

Setup Instructions

1. Clone the Repository (or create files): If this were a git repo, you'd clone it. For now, create the files api.py, app.py, and requirements.txt with the content provided.

2. Create a Python Virtual Environment (Recommended):

   python -m venv venv
   source venv/bin/activate  # Linux/macOS
   # On Windows (Command Prompt): venv\Scripts\activate
   # On Windows using Git Bash: source venv/Scripts/activate

3. Install Dependencies:

   pip install -r requirements.txt

4. Set up Gemini API Key:

   • You need a Gemini API key to use the AI estimation feature. You can obtain one from Google AI Studio.
   • Create a file named .env in the root directory of the project (the same directory as api.py).
   • Add your API key to the .env file like this:

     API_KEY="YOUR_API_KEY_TO_ANY_LLM_HERE"
     

   • Replace "YOUR_API_KEY_TO_ANY_LLM_HERE" with your actual key.

Running the Application

You need to run two components separately: the FastAPI backend and the Streamlit frontend.

1. Run the FastAPI Backend: Open a terminal, navigate to the project directory, and run:

   uvicorn api:app --reload --port 8000

   This will start the backend server, typically at http://localhost:8000. Keep this terminal window open.

2. Run the Streamlit Frontend: Open a new terminal window, navigate to the project directory, and run:

   streamlit run app.py

   This will open the Streamlit application in your web browser, usually at http://localhost:8051.

How to Use

1. Open the Streamlit application in your browser.
2. Enter your research idea in the text box or provide a URL to relevant content.
3. Click "Get AI Estimate for N and Cohen's d". The application will contact the FastAPI backend, which in turn queries the Gemini API.
   • The AI-estimated Total N and Cohen's d will populate the input fields.
   • A snippet of the text processed by the AI will be shown in an expander.
   • Any references the AI identifies will be listed in a "References" section.
4. Adjust the "Total Number of Participants (N)" and "Expected Cohen's d" using the input fields or sliders.
5. The "Calculated P-Value" will update automatically based on your adjustments.
   • The tool will also indicate if the p-value is typically considered statistically significant (< 0.05).
6. A new section displays the estimated power (probability of detecting the effect) along with a progress bar.

Important Considerations for the P-Value Calculation

• The p-value is calculated assuming a two-sample, two-sided t-test with equal sample sizes in each of the two groups (N_total / 2 per group).
• It answers the question: "If a study with N_total participants observed an effect size of Cohen's d, what would the p-value be?"
• This is not a power calculation (which determines the probability of detecting an effect of a certain size).
• References listed by the AI are not guaranteed to be exhaustive or correct and should be independently verified.