Artificial Intelligence and non-/Micro-Invasive was Cost-Effective in a Randomized Trial

Objectives: We assessed the cost-effectiveness of AI-supported detection of proximal caries in a randomized controlled clustered cross-over superiority trial.
Methods: Twenty-three dentists assessed 20 bitewings; 10 of which were randomly evaluated supported by an AI-based software and the other 10 without AI. We evaluated the proportion of true and false positive and negative detections and the treatment decisions taken for each detected lesion (non-invasive, micro-invasive, invasive). Cost-effectiveness was assessed from a public-private-payer perspective in Germany. A Markov simulation model and Monte Carlo microsimulations of posterior permanent teeth were used to analyze cost-effectiveness over patients’ lifetime. Costs were estimated using the German public and private fee catalogues, and our outcome was the time teeth were retained.
Results: For detecting early (E2 or D1) lesions (i.e., those usually recommended for non- or micro-invasive management), dentists were significantly more sensitive when using AI. However, treatment decisions determined the lifetime cost-effectiveness: Managing early lesions restoratively (as occurred in many cases in trial) resulted in AI and no AI showing similar effectiveness (tooth retention for a mean (2.5-97.5%) 49 (48-51)) and costs (AI: 330 (250-409) Euro, no AI: 330 (248-410) Euro). If, however, all detected early lesions had been treated non- or micro-invasively, AI was far less costly (266; (200-352) Euro) than no AI.
Conclusions: AI applications should not only support caries detection, but also subsequent evidence-based management of caries lesions.