Breast Cancer MCQ Validation

Overview

This repository contains 200 expert-level multiple-choice questions (MCQs) on breast cancer, developed for a benchmarking study comparing human experts vs. LLMs. Before distributing the questions to 25 expert reviewers, we need a clinical deep-dive validation to ensure every question is accurate and up to date.

Your task: Review each MCQ as a clinician and validate whether the classification, treatment, and overall question quality are correct.

---

Files

File	Description
MCQ_Validation_Interface.html	Interactive validation tool — open in your browser (all 200 questions are embedded, no separate file needed)
20042026_expert_questions.json	Raw question data (200 MCQs) — reference only; the HTML already contains this data
README.md	This file

---

How to Use the Validation Tool

You can use the tool in two ways:

• Online (recommended): Open https://cathiwi.github.io/mcq-validation-interface/MCQ_Validation_Interface.html in your browser.
• Offline: Download MCQ_Validation_Interface.html and double-click to open it in your browser (Chrome, Firefox, or Edge). No installation needed — everything runs locally.

1. Enter Your Name

• In the left sidebar, enter your name in the Reviewer field
• Your name will be saved automatically and included in the export

2. Review Each Question

For each MCQ, the interface shows you:

• Question stem — the full clinical scenario
• Answer options (A–E) — correct answer highlighted in green, wrong options in light orange
• Rationale & References — click to expand and see why the correct answer is right, why the others are wrong, and the supporting references

3. Validate with 3 Quick Clicks

After reviewing each question, answer these three things:

#	Question	Options	What to check
1	Correct answer is right?	Yes / No	Is the marked correct answer actually correct? Check subtype classification (TNBC, HR+, HER2+, HER2-low), biomarker cutoffs (HR%, HER2 IHC/ISH), and staging
2	Wrong options suitable?	Yes / Partially / No	Are the distractors (wrong answers) clearly wrong? Or could any of them be arguably correct given current guidelines?
3	MCQ usable?	OK / Revise / Reject	Overall verdict — can this question go to the expert panel as is?

If something is wrong, please leave a comment explaining what needs to change.

4. Navigate Efficiently

• Use the sidebar to jump to any question
• Use arrow keys (← →) to move between questions
• Use keyboard shortcuts: press 1 = correct answer yes, 2 = distractors yes, 3 = MCQ OK
• Use filters (All / Open / OK / Revise / Reject) to focus on unreviewed questions
• Switch to the Overview tab to see your progress across all sections

5. Take Breaks

Your progress is saved automatically in your browser. You can close the tab, shut down your laptop, and come back later — everything will still be there. Just make sure to always use the same browser (and, if you use the online version, the same device).

6. Export Your Results

When you are done (or at any point for a backup):

• Click Export CSV in the top right
• A CSV file with all your ratings and comments will download automatically
• Send this CSV file back to us

---

What to Watch For

Please pay special attention to these clinical nuances:

• HR-low (1–10%): These tumors may behave more like TNBC. Check if the question and treatment reflect this
• HER2-low (IHC 1+ or IHC 2+/ISH−): Now a targetable category (e.g., T-DXd). Verify classification and treatment options
• HER2-ultralow (IHC >0 but <1+): Emerging category — check if questions account for this
• gBRCA-mutated tumors: Affects treatment (PARPi eligibility). Ensure this is reflected where relevant
• Guideline alignment: Does the correct answer match current guidelines (NCCN, ESMO, AGO, S3)?
• Outdated treatments: Has the standard of care changed since the question was written?

---

Verdict Definitions

Verdict	Meaning
OK	Clinically accurate and ready for the expert panel
Revise	Has fixable issues — please describe what needs to change in the comment field
Reject	Fundamentally flawed or misleading — please explain why in the comment field

---

Questions?

If you are unsure about a question or need clarification about the validation process, feel free to reach out.

Thank you for your time — your clinical expertise is essential to making this benchmark reliable!