Related Experiment Video
Updated: Sep 16, 2025
07:32
Author Spotlight: 3D Movement Assessment of Maxillary Posterior Teeth in Clear Aligner Treatment
Published on: February 23, 2024
1.3K
Spin and Hacking in Machine Learning Prediction Model Studies in Dentistry: A Systematic Review
Liandi Cheng1, Yixuan Pan1, Po-Kam Wo1
1State Key Laboratory of Oral Diseases & National Center for Stomatology & National Clinical Research Center for Oral Diseases, West China Hospital of Stomatology, Sichuan University, Chengdu, Sichuan, China.
Oral Diseases
|July 8, 2025
Summary
Researchers may exaggerate machine learning (ML) model value in dentistry through spin or hacking. This study found prevalent spin and some evidence of hacking, recommending TRIAL principles for transparent ML research.
Area of Science:
- Biostatistics
- Dental Research
- Machine Learning
Background:
- Machine learning (ML) and prediction models are increasingly popular in dentistry.
- Concerns exist regarding misleading reporting (spin) and statistical manipulation (hacking) in ML studies.
Purpose of the Study:
- To assess the prevalence of spin and potential hacking in ML-based prediction model studies within dentistry.
- To identify common spin practices and facilitators in dental ML research.
Main Methods:
- A systematic search identified eligible studies up to October 12, 2024.
- 1206 Area Under the Curve (AUC) values were analyzed for potential hacking.
- 209 studies were evaluated for spin using the SPIN-PM framework.
Main Results:
- Evidence of potential AUC-hacking was suggested by fluctuations near common thresholds.
- Spin practices were identified in 37.3% of studies, often involving optimistic language or unsubstantiated clinical applicability claims.
- Facilitators of spin, such as reporting performance measures without confidence intervals, were present in 39.2% of studies.
Conclusions:
- Spin and its facilitators are common in dental ML prediction model studies.
- Some evidence of hacking was detected.
- Adoption of TRIAL (Transparency, Reporting, Integrity, Adjustment, and Learning) principles is recommended to enhance research integrity.
