Abstract Peri-implant diseases remain a major cause of late implant failure, and current risk assessment tools show limited capacity to integrate prosthetic factors, salivary biomarkers and artificial intelligence-based prediction. This study aimed to develop the Implant Success Prediction Tool (ISPT), a multifactorial peri-implant risk stratification system structurally designed for modular integration with artificial neural networks and salivary omics data. ISPT development followed three main pillars: (1) incorporation of Implant Disease Risk Assessment (IDRA)-validated clinical vectors, including bleeding on probing percentage, number of sites with probing depth ≥ 5 mm, bone loss in relation to age, periodontitis susceptibility, supportive periodontal therapy and hygiene/compliance parameters; (2) qualitative usability testing of IDRA by implantologists, who identified elements to maintain, clarify or expand; and (3) alignment with a precision medicine framework, establishing collaboration with a salivary diagnostics laboratory SalivaTec ( https://ciis.ucp.pt/salivatec ) to enable systematic saliva collection and future deep phenotyping. The final ISPT structure comprises ten standardized risk vectors displayed in a colour-coded radial traffic-light diagram, integrating six adapted IDRA-derived vectors and four novel vectors: abutment height/angulation; saliva collection/deep phenotyping vector (“ salivaomics ”); foreign bodies, titanium particles and tribocorrosion; and other for occlusal loading and functional risk. The tool is conceptually prepared to function as a structured input matrix for artificial neural networks, supporting longitudinal training with combined clinical and salivary data to predict implant outcomes (peri-implant health, mucositis, peri-implantitis) over a minimum 5-year monitoring period. ISPT represents the first peri-implant risk assessment tool explicitly designed for modular integration of artificial intelligence and salivary omics data within a precision dentistry framework. Its standardised vectors, traffic-light visualisation and longitudinal validation methodology provide a scalable structure for future externally validated predictive models of implant success and failure.
Bornes et al. (Mon,) studied this question.