Pasquale Santamaria

Aim: To compare the prognostic performance of an artificial intelligence (AI) model with that of experienced clinicians in predicting tooth loss over a 10‐year period. Materials and Methods: An AI model trained on structured clinical and radiographic data was compared with 12 periodontists and 11 general dentists (GDs), who independently assigned prognostic scores (0–10 scale) to 300 teeth with known 10‐year outcomes. AI and clinician performance were evaluated under a fixed threshold and group‐specific optimal thresholds. Accuracy, sensitivity, specificity, predictive values, area under the receiver operating characteristic curve and calibration were used for comparison. Results: Using the fixed threshold (score > 5 = survival), clinicians achieved higher overall accuracy than AI (75.6% periodontists, 74.9% GDs, 69.2% AI; p < 0.05), with sensitivity low and comparable across groups (14.7%–22.7%). Using group‐specific optimal thresholds, AI achieved the highest accuracy for all‐cause tooth loss (67.5%) and performed comparably to GDs for periodontitis‐related tooth loss (76.2% vs. 76.3%). Calibration was substantially worse for AI (Brier score 0.247) than for periodontists and GDs (0.171, 0.170). Individually, AI performance overlapped with that of several clinicians. Conclusion: Clinicians outperformed AI under a fixed threshold, but AI performance was comparable to clinicians using optimal thresholds, situating AI within the range of experienced clinicians despite modest overall predictive ability.

Background: The aim of this study was to study the effect of non‐surgical periodontal therapy (NSPT) on gingival crevicular fluid (GCF) and serum progranulin (PGRN) levels in the early healing phases. Methods: The study included periodontitis (test) ( n = 24) and periodontal health (control) ( n = 24) groups. PGRN, vascular endothelial growth factor (VEGF), interleukin (IL)‐1β, tumor necrosis factor alpha (TNF‐α), and IL‐10 levels were assessed at baseline, at the 1st, 2nd, and 14th day, and 1st and 3rd month after NSPT in serum and GCF samples by Luminex bead‐based multiplex immunoassay method. Results: GCF PGRN, IL‐1β, TNF‐α, VEGF, and IL‐10 levels were higher in the test group than in the control group at baseline ( p < 0.05). GCF PGRN and VEGF levels decreased from day 14 after NSPT, while IL‐1β levels decreased gradually from day 2 ( p < 0.001). TNF‐α levels rapidly increased on day 1 after NSPT and gradually decreased from day 14 ( p < 0.001). GCF PGRN/ TNF‐α molar ratio levels dramatically decreased from baseline day 1 after treatment and then increased gradually from day 14 to the 1st month ( p < 0.001). There were no differences in serum parameters between groups and among time points ( p ≥ 0.05), while a strong positive correlation was detected between GCF PGRN and IL‐1β, and TNF‐α levels ( p < 0.001) at baseline. Conclusions: GCF PGRN total amount levels decreased gradually at each time point during the early healing period after NSPT, in parallel with IL‐1β. Changes in GCF PGRN and PGRN/TNF‐α molar ratio may be associated with periodontal disease and post‐treatment outcomes (ClinicalTrials.gov ID: NCT05535049). Plain language summary: Progranulin (PGRN) is a protein with complex physiological functions, producing granulin peptides that promote inflammatory and anti‐inflammatory activity. This study aimed to evaluate PGRN levels in the presence of periodontal disease and the local and systemic changes after following non‐surgical periodontal therapy (NSPT). The study included periodontitis (test) ( n = 24) and periodontal health (control) ( n = 24) groups. PGRN, vascular endothelial growth factor (VEGF), interleukin (IL) ‐1β, tumor necrosis factor alpha (TNF‐α), and IL‐10 levels were assessed in serum and gingival crevicular fluid (GCF) samples at baseline and at various time points after NSPT. GCF PGRN total amount levels were higher in the presence of periodontitis. Their levels decreased after NSPT from the 14th day in patients with periodontitis in parallel with inflammatory and regenerative mediators. However, PGRN/TNF‐α molar ratio levels increased after treatment at the end of the early healing phase. Molecular mediators hold promise as a diagnostic and therapeutic tool in periodontal therapy. Monitoring the level of PGRN in GCF before and after periodontal therapy could in the future be useful for personalized care.

Aim: To assess the feasibility of applying sequential, multiple assignment, randomized trial (SMART) design in periodontology and to assess the response of grade III furcation‐involved molars to various treatments.Methods: A member of the public served as a co‐investigator. Twenty participants with at least one grade III furcation‐involved molar tooth were randomized to either non‐surgical periodontal treatment (NSPT) or open‐flap debridement (OFD). They were reassessed 6 months later, where ‘successful response’ to therapy was defined with a ‘combined outcome’ incorporating patient‐reported and clinical parameters. Based on SMART principles, non‐responders underwent further treatment and were followed up to 12 months.Results: All study feasibility criteria were met, except ‘planned recruitment rate’, which was slower than expected. At the 6‐month review, those who underwent OFD showed a greater reduction in probing pocket depth (PPD) compared to those who received NSPT (p = < 0.01). ‘Success’ was reached by 6 out of 10 NSPT participants and by 7 out of 10 OFD participants. Non‐responders were further randomized and followed up to 12 months.Conclusion: This SMART design is feasible in periodontology trials. This study demonstrates its advantages and limitations along with clinical and patient‐reported data of grade III furcations treated surgically or non‐surgically.

Aims: To externally validate the potential applicability of the 2018 classification of periodontal diseases as a risk assessment tool, through the use of a nomogram built on a multivariate predictor model.Materials and Methods: Data from 459 patients with periodontitis, across four cohorts (three in the United States, and one in the United Kingdom), were retrospectively analysed. After staging and grading periodontitis before active periodontal therapy (APT), patients were categorised by the model as having ‘low tooth loss’ (≤ 1 teeth lost due periodontal reasons [TLP]) or ‘high tooth loss’ (≥ 2 TLP) at a 10‐year follow‐up. Model discrimination was evaluated using the area under the receiver operating characteristic (AUC‐ROC) curve. Calibration was assessed through calibration plots, calibration‐in‐the‐large (CITL), calibration slope and the expected:observed (E:O) ratio. Recalibration methods, including Temperature Scaling, Isotonic Regression and Beta Calibration, were also tested.Results: The original nomogram yielded an aggregate AUC‐ROC 0.72 but showed poor calibration. Isotonic recalibration improved the AUC‐ROC to 0.77 and enhanced calibration metrics, achieving an E‐statistic of 1.00, CITL of 0.00 and a calibration slope of 1.00.Conclusion: A nomogram based on the components of the 2018 periodontal disease classification can serve as a prognostic tool with cross‐site applicability across clinical settings in industrialised countries, accurately predicting the 10‐year risk of tooth loss due to periodontitis from the initial assessment conducted before the start of APT.

Background: Artificial intelligence (AI) has the potential to enhance healthcare practices, including periodontology, by improving diagnostics, treatment planning and patient care. This study introduces ‘PerioGPT’, a specialized AI model designed to provide up‐to‐date periodontal knowledge using GPT‐4o and a novel retrieval‐augmented generation (RAG) system. Methods: PerioGPT was evaluated in two phases. First, its performance was compared against those of five other chatbots using 50 periodontal questions from specialists, followed by a validation with 71 questions from the 2023–2024 ‘In‐Service Examination’ of the American Academy of Periodontology (AAP). The second phase focused on assessing PerioGPT's generative capacity, specifically its ability to create complex and accurate periodontal questions. Results: PerioGPT outperformed other chatbots, achieving a higher accuracy rate (81.16%) and generating more complex and precise questions with a mean complexity score of 3.81 ± 0.965 and an accuracy score of 4.35 ± 0.898. These results demonstrate PerioGPT's potential as a leading tool for creating reliable clinical queries in periodontology. Conclusions: This study underscores the transformative potential of AI in periodontology, illustrating that specialized models can offer significant advantages over general language models for both educational and clinical applications. The findings highlight that tailoring AI technologies to specific medical fields may improve performance and relevance.

Aim: To describe the microbiological composition of subgingival dental plaque and molecular profile of gingival crevicular fluid (GCF) of periodontal furcation‐involved defects.Materials and Methods: Fifty‐seven participants with periodontitis contributed with a degree II–III furcation involvement (FI), a non‐furcation (NF) periodontal defect and a periodontally healthy site (HS). Subgingival plaque was analysed by sequencing the V3–V4 region of the 16S rRNA gene, and a multiplex bead immunoassay was carried out to estimate the GCF levels of 18 GCF biomarkers. Aiming to explore inherent patterns and the intrinsic structure of data, an AI‐clustering method was also applied.Results: In total, 171 subgingival plaque and 84 GCF samples were analysed. Four microbiome clusters were identified and associated with FI, NF and HS. A reduced aerobic microbiota (p = .01) was detected in FI compared with NF; IL‐6, MMP‐3, MMP‐8, BMP‐2, SOST, EGF and TIMP‐1 levels were increased in the GCF of FI compared with NF.Conclusions: This is the first study to profile periodontal furcation defects from a microbiological and inflammatory standpoint using conventional and AI‐based analyses. A reduced aerobic microbial biofilm and an increase of several inflammatory, connective tissue degradation and repair markers were detected compared with other periodontal defects.

Aim: The aim of this analysis was to compare a clinical periodontal prognostic system and a developed and externally validated artificial intelligence (AI)‐based model for the prediction of tooth loss in periodontitis patients under supportive periodontal care (SPC) for 10 years.Materials and Methods: Clinical and radiographic parameters were analysed to assign tooth prognosis with a tooth prognostic system (TPS) by two calibrated examiners from different clinical centres (London and Pittsburgh). The prediction model was developed on the London dataset. A logistic regression model (LR) and a neural network model (NN) were developed to analyse the data. These models were externally validated on the Pittsburgh dataset. The primary outcome was 10‐year tooth loss in teeth assigned with ‘unfavourable’ prognosis.Results: A total of 1626 teeth in 69 patients were included in the London cohort (development cohort), while 2792 teeth in 116 patients were included in the Pittsburgh cohort (external validated dataset). While the TPS in the validation cohort exhibited high specificity (99.96%), moderate positive predictive value (PPV = 50.0%) and very low sensitivity (0.85%), the AI‐based model showed moderate specificity (NN = 52.26%, LR = 67.59%), high sensitivity (NN = 98.29%, LR = 91.45%), and high PPV (NN = 89.1%, LR = 88.6%).Conclusions: AI‐based models showed comparable results with the clinical prediction model, with a better performance in specific prognostic risk categories, confirming AI prediction model as a promising tool for the prediction of tooth loss.