Transparency and Representation in Clinical Research Utilizing Artificial Intelligence in Oncology: A Scoping Review
Anjali J D'Amiano1, Tia Cheunkarndee1, Chinenye Azoba1
1Johns Hopkins University School of Medicine, Baltimore, Maryland, USA.
Cancer Medicine
|March 10, 2025
Summary
Few oncology studies report racial demographics, and those that do often lack diversity. Improving data transparency and representation is crucial for equitable artificial intelligence (AI) in cancer care.
Area of Science:
- Oncology
- Artificial Intelligence
- Health Equity
Background:
- Artificial intelligence (AI) shows promise for improving cancer care outcomes.
- Ensuring AI models do not perpetuate racial and ethnic disparities is critical.
- This review assesses demographic data reporting in AI oncology studies.
Purpose of the Study:
- To evaluate the transparency of demographic data reporting in AI oncology research.
- To assess the diversity of participants in clinical studies using AI in oncology.
- To identify gaps in reporting that could lead to health disparities.
Main Methods:
- Searched PubMed for AI, machine learning, and deep learning studies in oncology (2016-2021).
- Included original research and clinical trials, excluding reviews and meta-analyses.
- Collected data on reported demographics (age, sex, race) for AI training/validation sets.
Main Results:
- 40% of eligible studies had publicly available data sets.
- Demographic reporting was inconsistent: 58% reported age, 51% sex, and only 5% reported race.
- Studies reporting race predominantly included White individuals (70.7%-93.4%).
- Only three studies reported race with more than two categories.
Conclusions:
- A significant lack of racial and ethnic demographic data reporting was observed in AI oncology studies.
- Studies reporting race had limited representation of non-White populations.
- Enhanced transparency and diverse data are essential for unbiased AI implementation in oncology.
