Large Language Models to Aid Gynecological Oncology Treatment

Not Applicable

Not yet recruiting

• Conditions: Breast Cancer

• Registration Number: NCT06865534
• Lead Sponsor: Philipps University Marburg

Brief Summary

This trial aims to assess the impact of providing medical students with access to large language models, in comparison to treatment guideline pdfs, on treatment concordance with a conventional multidisciplinary tumor board

Detailed Description

Advanced artificial intelligence (AI) technologies, particularly large language models such as OpenAI's ChatGPT, hold significant potential for enhancing medical decision-making. While ChatGPT was not specifically designed for medical applications, it has shown utility in various healthcare scenarios, including answering patient inquiries, drafting medical documentation, and aiding clinical consultations. Despite these advancements, its role in supporting treatment decision-making-particularly in complex oncological cases-remains underexplored.

Treatment decision-making in gynecological oncology is a multifaceted process that integrates evidence-based guidelines, tumor biology, patient-specific factors, and clinical expertise. AI tools like ChatGPT could potentially assist in synthesizing relevant guideline-based recommendations, improving decision accuracy, and facilitating more efficient clinical workflows. However, ChatGPT is not specifically tailored for oncological treatment decisions and lacks comprehensive validation in this domain. Additionally, it may generate misinformation or plausible-sounding but inaccurate recommendations, which could impact clinical judgment. Therefore, understanding how medical professionals, including students and early-career physicians, interact with such AI tools is essential before broader integration into clinical practice. Locally deployable models, such as Llama, enable secure, on-premise usage while retrieval-augmented generation ensures guideline-compliant recommendations.

This study will investigate the impact of language models on treatment decision support for medical students managing gynecological oncology cases. This is a crossover study, where participants will be randomized into two groups. All participants begin with access to ChatGPT for two vignettes. They then proceed with two cases using either a locally deployed language model, followed by two cases relying on guideline PDFs, or vice versa.

Each participant will analyze clinical cases, propose treatment plans, and rate their confidence in their decisions and decision support system usability. This study aims to provide insights into the potential benefits and limitations of integrating AI tools like ChatGPT into oncological treatment decision-making.

Study Timeline

• Study First Submitted: 2025-02-11
• Study Start: 2025-02-24
• Status Verified: 2025-03
• Study First Submitted QC: 2025-03-03
• Last Update Submitted: 2025-03-03
• Study First Posted: 2025-03-10
• Last Update Posted: 2025-03-10
• Primary Completion: 2025-06-01
• Study Completion: 2025-06-01

Recruitment & Eligibility

• Status: NOT_YET_RECRUITING
• Sex: All
• Target Recruitment: 68

Inclusion Criteria

• Medical students having started with clinical subjects

Exclusion Criteria

• Not being a medical student

Study & Design

• Study Type: INTERVENTIONAL
• Study Design: CROSSOVER

Primary Outcome Measures

Name	Time	Method
Treatment concordance with tumor board decisions	directly (within 10 minutes) after Intervention	Participants in each group select treatment modalities for case vignettes

Secondary Outcome Measures

Name	Time	Method
Treatment confidence	directly (within 10 minutes) after Intervention	For each case participants will be asked for their treatment confidence (VAS 0-10). The mean score will be compared between decision support groups.
Time spent for treatment decision	directly (within 10 minutes) after Intervention	Time (in seconds) participants spend per case between the decision support groups will be compared.