Journal of Dental Research

Purpose: To compare the performance of three large language models (LLMs), ChatGPT-o1, Grok 3, and Gemini 2.0 Pro, on the American College of Prosthodontists' National Prosthodontic Resident Examination (NPRE). Materials and Methods: Official NPRE examinations from years 2015-2024 were retrieved from the American College of Prosthodontists website and used to prompt each LLM. Paid access tiers and advanced reasoning were enabled. Responses from LLMs were recorded and compared against the official answer keys for accuracy, variability across years, and consistency of responses. Duplicate questions were retained for the primary analysis and removed in a sensitivity analysis. Accuracy (percent correct) was compared using generalized estimating equations binary logistic regression (primary); Cochran's Q with Bonferroni-adjusted McNemar tests (sensitivity). Year-to-year variability (SD of annual accuracy) was compared using weighted Levene tests. Consistency across duplicate items was quantified as the percentage of perfect matches and compared using Cochran's Q with Bonferroni-adjusted McNemar tests. Results: Across 1,487 questions, overall accuracy differed among models (global P < 0.001). ChatGPT-o1 achieved 78.4% correct, Grok 3 76.4%, and Gemini 2.0 Pro 73.2%; ChatGPT-o1 and Grok 3's accuracy significantly exceeded Gemini 2.0 Pro (both P [&le;] 0.016), with no significant difference between ChatGPT-o1 and Grok 3 (P = 0.099). After removing duplicates (n = 1,200), results were similar: ChatGPT-o1, Grok 3, and Gemini 2.0 Pro were 77.1%, 76.6%, and 71.5% correct, respectively; ChatGPT-o1 and Grok 3's accuracy was significantly higher than Gemini 2.0 Pro (both P < 0.001). Variability across years did not differ significantly (P [&ge;] 0.097). Grok 3 showed significantly higher perfect-match consistency (79.2%) than ChatGPT-o1 (69.4%) and Gemini 2.0 Pro (64.5%) (P [&le;] 0.010). Conclusions: Contemporary LLMs showed 70-80% accuracy on postgraduate prosthodontic multiple-choice content. While promising for educational support, residual error and inconsistency warrant human oversight and transparent methods reporting before curricular adoption.

Objective. A deep learning (DL) model was used to convert smartphone videos of a complete arch implant cast into 3D scans. The aim of current study was to determine if a custom scan body (SB) with geometric features and coating would outperform regular PEEK stock SB in this DL scenario. The DL-derived scan outcomes were compared with those obtained from a conventional splinted open-tray impression and from photogrammetry. Materials and Methods. A maxillary edentulous model with six implants and multi-unit abutment analogs was scanned using four protocols: conventional splinted open-tray impression (CO), photogrammetry (PG; Icam4D), DL using stock SBs (DLS) and DL using custom SBs (DLC). Each protocol was repeated for 10 times. The DL scans were produced from smartphone videos with a high-fidelity, multi-view 3D construction AI model (Neuralangelo). The custom designed SB incorporated geometric features and was fabricated via 3D printing followed by a spray coating. Accuracy (trueness and precision) was assessed using three measurements: Root Mean Square (RMS), linear deviation, and angular deviation. Results. DLC outperformed DLS in both trueness and precision regarding RMS and linear measurements (p<0.001). CO and PG demonstrated the highest RMS and linear trueness, with no significant difference between them (RMS: p=0.93; linear: p=0.663). PG achieved the best precision across RMS, linear and angular measurements. Conclusion. The optimised SB significantly improves the accuracy of DL-based approach for full-arch implant scan comparing to regular PEEK stock scan bodies. While early stage, neural surface reconstruction has potential as a viable option for full-arch implant rehabilitation.

Objectives: To systematically compare the outcomes of zirconia-based versus metal framework veneered implant-supported fixed complete dental prostheses (ISFCDPs). We integrate cost effectiveness analysis comparing monolithic zirconia versus titanium PMMA ISFCDPs to model the 20 year costs and benefits associated with each arm. Data sources: PubMed/MEDLINE, Cochrane CENTRAL, Scopus, ClinicalTrials.gov and WHO ICTRP were searched on March 6 2026 with no date restrictions applied. Study selection: Two arm comparative studies reporting clinical outcomes for both zirconia-based and metal framework veneered ISFCDPs with a minimum of one year follow up after prosthesis delivery. Data extraction and synthesis: Two independent reviewers extracted data and assessed the risk of bias using the Newcastle Ottawa Scale. Narrative synthesis followed SWiM guidelines. A tiered material classification system was employed to handle framework alloy heterogeneity. Meta-analysis was not possible due to heterogeneity in outcomes reported and methodological differences. A Markov cohort model estimated cost-effectiveness from a patient perspective over a 20 year horizon. Deterministic and probabilistic sensitivity analysis (PSA) tested results robustness. Results: Eight retrospective cohort studies (396 patients; 551 prostheses) were included in final analysis. All were classified as Tier 1 due to inconsistent alloy reporting. Prosthesis and implant survival rates were high across both arms but with no statistical difference for any primary outcome. Cost-effectiveness analysis estimated total discounted per-patient costs of {pound}23,295 for titanium-PMMA and {pound}29,170 for monolithic zirconia over 20 years. This results in an ICER of {pound}315,417 per QALY gained. PSA confirmed zirconia as cost-effective in only 0.50% of 10,000 iterations at the {pound}20,000/QALY threshold. Net monetary benefit was consistently negative for zirconia. Conclusions: Both types of ISFCDPs provided high prosthesis and implant survival with no statistically significant difference in primary clinical outcomes, though all evidence was graded at very low certainty. We estimated a per-patient premium of {pound}5,875 for monolithic zirconia over a 20 year horizon with marginal QALY gains (0.019). Adequately powered RCTs with standardised material reporting are needed to strengthen our clinical and economic conclusions. Systematic review registration: OSF: 10.17605/OSFJO/CF796

Abstract Aim: To test whether neighbouring-tooth restorations affect a focal tooths periodontal status (within-mouth spillover), and whether accounting for them alters the established association at the restored surface. Materials and methods: Prospective tooth-surface data from the Study of Health in Pomerania (SHIP): SHIP-START (N=2,715/1,982/1,397, follow-ups 1-3) and SHIP-TREND (N=2,123). Own and neighbouring-tooth restoration were defined at baseline; probing depth (PD), clinical attachment level (CAL), bleeding on probing (BOP) and deep pockets (PD >=4 mm) at follow-up. A neighbourhood-exposure mixed model with generalized propensity-score adjustment estimated both effects, with two corroborating estimators. Results: Own restoration was associated with worse periodontal status at that surface (crown PD exp(beta) up to 1.10). Neighbouring-tooth restorations raised focal-tooth PD (spillover exp(beta) 1.03-1.05) and deep-pocket risk (risk ratio up to 1.25); CAL and BOP showed none consistently. It attenuated with tooth-lag, nearing the null by two to three teeth in most estimators. Ignoring interference modestly inflated the direct effect; an opposing-tooth negative control showed none; untreated caries reproduced it. Conclusions: Dental restorations exert a within-mouth spillover on neighbouring-tooth PD: their periodontal footprint extends beyond the restored tooth. The direct association persists, modestly attenuated, after accounting for the neighbourhood.

Background: Subepithelial connective tissue grafting remains a reference treatment for predictable root coverage. Although short- and medium-term outcomes of tunnel-based procedures are well documented, evidence regarding stability beyond 10 years remains limited. This study evaluated the long-term clinical performance of a minimally invasive tunnel technique combined with subepithelial connective tissue grafting (SCTG) under routine clinical conditions. Methods: This retrospective longitudinal cohort study included 74 patients (57 women and 17 men) contributing 710 gingival recession sites treated between 2009 and 2025. All sites were treated with a tunnel approach and SCTG, with enamel matrix derivative (EMD) used in selected cases. The mean follow-up was 6.0 for 4.0 years, with a maximum observation period of 16 years. The primary outcome was recession depth reduction. Secondary outcomes included complete root coverage (CRC), mean root coverage, and long-term marginal stability. Clinically relevant relapse was defined as a 1 mm increase in recession after initial healing. Results: Mean recession reduction was 2.72 mm. Complete root coverage was achieved at 83.4% of treated sites. At the final available follow-up, no treated site showed a clinically relevant relapse of 1 mm after initial healing, and no site deteriorated beyond its baseline recession level. Treatment effects were observed across anterior and posterior regions. Conclusions: Within the limitations of a retrospective cohort design, tunnel surgery combined with SCTG was associated with high root-coverage predictability and durable marginal soft-tissue stability for observation periods extending to 16 years. These real-world data support phenotype-enhancing, minimally invasive soft-tissue augmentation as a durable therapeutic strategy for localized and multiple gingival recessions. Keywords: gingival recession; tunnel technique; subepithelial connective tissue graft; root coverage; periodontal plastic surgery; long-term stability

Automated dental implant planning using artificial intelligence (AI) faces challenges in interpretability, safety assurance, and applicability beyond single-tooth cases. This paper presents a unified explainable framework for cone-beam computed tomography (CBCT) that handles both single-tooth and consecutive multiple missing teeth. The method combines nnU-Net-based anatomical segmentation with a transparent geometric optimization pipeline. Implant placement is formulated as a lexicographic multi-objective constrained optimization problem that enforces clinical safety margins while balancing bone engagement, mesiodistal spacing, and buccolingual position. For multi-tooth cases, a global coordination strategy ensures inter-implant parallelism, minimum spacing, and a common insertion path. The framework was validated on 20 single-tooth and 9 consecutive multi-tooth cases. The AI plans achieved a centroid distance of 1.60+/-1.05 mm for single-tooth cases and 1.22+/-0.32 mm for multi-tooth cases, with approximately 7 degrees axial angle difference. All plans achieved 100% compliance with predefined safety thresholds. In a Turing-style test, experts were unable to distinguish AI from human intelligence (HI) plans under blinded conditions. By decoupling perception from reasoning and embedding clinical knowledge as explicit constraints, the framework offers an interpretable and clinically trustworthy solution for computer-aided implant dentistry.

Objective: To critically evaluate the influence of extracellular matrix stiffness on tumor progression, prognosis, and clinical implications in patients with oral squamous cell carcinoma. Design: This systematic review followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, and the protocol was registered in PROSPERO (CRD420251156429). Five databases (PubMed, Scopus, Embase, Web of Science, and OATD) were searched without restrictions on publication year or language. Human studies that evaluated extracellular matrix stiffness in oral squamous cell carcinoma, either as an assessment method or as a prognostic factor, were included; in vitro and animal studies were excluded. Results: Eight human studies involving 436 patients were included. Elastography was the most frequent measurement method, employed in four studies. Mean tumor thickness was reported in three studies and ranged from 3.0 to 4.4 mm. Extracellular matrix stiffness was expressed in various units (e.g., 37.7 gf, 23.5 kPa) and was consistently higher in tumor tissue than in normal mucosa in all studies. Three studies demonstrated a significant positive correlation between stiffness and tumor thickness. One study identified predictive cut-off values for lymph node metastasis and reported that high matrix stiffness was an independent predictor of worse overall survival (HR = 4.022; 95% CI: 1.294-12.495; p = 0.016) and recurrence-free survival (HR = 5.680; 95% CI: 1.146-28.143; p = 0.033). A meta-analysis of five studies (307 patients) showed a significantly higher standardized stiffness in oral squamous cell carcinoma compared with normal oral mucosa (standardized mean difference = 5.28; 95% CI: 2.61-7.94; p < 0.001), with substantial heterogeneity (I2= 98.6%). Conclusions: Extracellular matrix stiffness is elevated in oral squamous cell carcinoma and is associated with greater tumor thickness, lymph node metastasis and, in one study, worse survival. These findings support extracellular matrix stiffness as a potential biomechanical marker, pending further validation in standardized prospective studies.

Background: Temporomandibular disorders (TMDs) and orofacial pain (OFP) conditions affect approximately one-third of the global population, yet diagnosis often relies on subjective clinical assessment rather than standardized, evidence-based criteria. This diagnostic uncertainty contributes to misdiagnosis, misdirected or delayed treatment and significant healthcare costs. Case description: We prospectively collected structured clinical data from 1,584 patients using a custom, structured, note-documentation system designed for machine learning compatibility (Smart Medical Note [SmartNote]). From the 77 possible TMD-OFP diagnoses, we identified 17 conditions with sufficient case volumes ([&ge;]24 exemplars) for analysis. The clinical features which were present at or above >50% in these cases underwent Monte Carlo statistical analysis to determine which ones are most useful for creating an objective diagnostic profile. The 17 evidence-based diagnostic profiles demonstrated strong discriminative performance (median AUC values 0.71-0.99). Practical Implications: These diagnostic profiles represent the first large-scale, statistically derived criteria for TMD and orofacial pain conditions. By providing objective, quantifiable diagnostic standards, this statistical analysis of structured electronic records can support algorithmic diagnostic-assist software, reduce the time to appropriate treatment, and enhance clinical decision-making and documentation of diagnoses for dental practitioners, regardless of their level of specialized training in orofacial pain. Diagnostic alignment analysis confirms complete concordance with DC-TMD criteria for seven major diagnostic categories, validating the clinical applicability of our data-driven approach.

Hard-tissue surgery requires precise spatial control, while conventional computer-assisted navigation presents guidance separately from the operative field. Augmented reality can bring navigation information closer to the surgeon, but how best to translate this information into intuitive intraoperative guidance remains underexplored. We therefore developed an augmented reality navigation system (ANS) that translates real-time deviations from the planned trajectory into co-located holographic action cues. Evaluation comprised typodont and ex vivo porcine experiments, a randomized crossover study in 40 dental trainees, and eight clinical cases. In porcine mandibles, ANS showed lower angular deviation than static and dynamic navigation (1.78 vs 6.42 and 4.02 degrees). Among trainees, accuracy was comparable with dynamic navigation, while workload was lower (29.18 vs 33.15), usability higher (73.13 vs 68.88), and visual-focus ratings improved (3.55 vs 2.28). Clinical cases supported feasibility. These findings support co-located action guidance as an alternative interface paradigm for surgical navigation.

Objective: To examine whether Medicaid-enrolled preschoolers with untreated decayed teeth received dental treatment within one year of enrollment and identify the factors associated with a treatment gap. Methods: A retrospective cohort analysis of data from a cluster-randomized trial conducted in 18 community-based pediatric primary care practices in Northeastern Ohio (2017-2022). Treatment receipt was determined using Medicaid claims, with treatment gap defined as fewer teeth with treatment claims than teeth found on baseline exam with decay. Multivariable logistic regression assessed the association of treatment gap with child age, sex, race/ethnicity, caregiver education, and number and location of baseline decayed teeth. Results: Of 766 eligible children, 487 (63.6%) attended the dentist within one year. Among 155/487 (31.8%) with baseline untreated decay, 90/155 (58.1%) had a treatment gap. Odontograms visually showed that decay was concentrated on upper anterior and posterior teeth. A treatment gap was associated with a greater number of decayed posterior teeth (OR = 1.90, 95% CI: 1.60-2.30) and decayed anterior teeth (OR = 2.19, 95% CI: 1.51-3.39), both p < 0.001. Other socio-demographic variables were not significantly associated with a treatment gap. Conclusion: More than half of Medicaid-enrolled children attending well-child visits had a dental treatment gap after 1 year. This pattern may reflect dentists' hesitancy to restore primary teeth nearing exfoliation and needing multiple dental visits to complete needed restorative treatment. To address this gap, non-surgical interventions such as silver diamine fluoride can be applied by pediatric primary care providers to control the bacteria and prevent disease progression.

Background: Nicotine pouches (NPs) are increasingly popular tobacco-free nicotine products, particularly among young people. Despite growing use, evidence regarding their oral health effects remains limited. Methods: We employed a mixed-methods approach comprising: (1) a national cross-sectional survey of NP users, (2) a survey of UK dental professionals, and (3) in-person clinical assessments of NP users. Survey data explored patterns of NP use, oral and dental side effects, and perceptions of risk. Clinical assessments included comprehensive oral examinations, periodontal assessments, and histopathological investigation of lesions where indicated. Results: A total of 3,659 NP users completed the public survey. Participants were predominantly male (89%) and aged 16-22 years (78%). Overall, 45% reported at least one oral or dental side effect. The most commonly reported effects were gingival recession (60%), burning sensations (41%), white patches (30%), and dry mouth (29%). Higher nicotine exposure was associated with increasing numbers of reported side effects. Among 126 dental professionals, oral lesions associated with NP use were commonly observed clinically, while only 17% reported seeing no oral or dental effects. Thirty-three NP users underwent in-person assessment. Ten lesions were biopsied; six demonstrated hyperkeratosis and four showed oral epithelial dysplasia (three mild, one moderate). All dysplastic lesions occurred at pouch placement sites. In two cases, clinical follow-up demonstrated partial or complete resolution of lesions following cessation of NP use. Periodontal assessments demonstrated generally low levels of periodontal disease and gingival recession across the cohort. Conclusions: Oral and dental side effects associated with NP use were commonly reported by users and observed by dental professionals. Histopathological assessment identified biopsy-confirmed oral epithelial dysplasia in lesions occurring at pouch placement sites, representing, to our knowledge, the first such report associated with NP use. Evidence of lesion improvement following discontinuation of use suggests that at least some NP-associated oral changes may be reversible. Although causality cannot be established, these findings highlight the need for vigilance among dental professionals and urgent longitudinal research to determine the incidence, progression, and reversibility of NP-associated oral lesions.

Importance: Clinicians often look for a dental source in Staphylococcus aureus bacteremia (SAB), but how often dental disease is found, compared with other bacteremias, is unclear. Objective: To compare documented dental disease in adults with SAB, Escherichia coli bacteremia, and a general population cohort. Design, Setting, and Participants: Retrospective cohort study in the Protege database of 48 975 adults with bacteremia from January 1, 2015, through December 31, 2023, assigned by earliest in-window organism, and 81 945 general population adults. Exposures: SAB, with E coli bacteremia as the primary comparator and the general population as the secondary comparator. Main Outcomes and Measures: Documented dental disease from 7 days before through 30 days after index, defined by a qualifying dental procedure or disease-specific note text and adjusted for age, sex, race, and documentation intensity. Results: Dental disease occurred in 528 of 16 435 patients with SAB (3.21%) and 179 of 24 942 patients with E coli bacteremia (0.72%). The adjusted odds ratio was 3.86 (95% CI, 3.23 to 4.63). By the pathology-only definition, disease occurred in 517 patients with SAB (3.15%) and 399 general population patients (0.49%). The adjusted odds ratio was 3.71 (95% CI, 3.08 to 4.46). Within SAB, rates were highest with prior infective endocarditis (9.8%), opioid use disorder (6.2%), injection drug use (5.3%), and prosthetic valve (5.0%). Prevalence ranged from 0.49% in the general population to 7.98% in other streptococcal bacteremia. Of 437 patients with SAB and an assignable first finding date, 333 (76.2%) had the finding after the index date. Conclusions and Relevance: Dental disease was documented more often in SAB than in E coli bacteremia after adjustment. Most findings were first recorded after bacteremia was recognized, during SAB evaluation. These data do not show that dental disease caused the bacteremia.

Objectives: A previously published conceptual mapping review integrated fragmented evidence on dental anxiety (DA) into a broader conceptual model of potential causes and consequences. This study focused on consequences, extending from oral health-related behaviours through oral disease to systemic health, and empirically evaluated these hypothesized relationships on the population level. Methods: Cross-sectional data from the 46-year follow-up of the Northern Finland Birth Cohort 1966 (n = 1,859) were analysed using structural equation modelling. DA was modelled as a latent construct; delayed/problem-oriented dental attendance and home oral care as behavioural proxies; periodontal disease burden as the percentage of sites with probing pocket depth >=4 mm; and cardiovascular risk as a Framingham-based risk score. Generalized anxiety, sex, and education were covariates. Smoking-stratified multigroup analyses assessed model robustness. Results: The prespecified associations were observed across behavioural, periodontal, and systemic-risk domains. Higher DA was associated with more delayed/problem-oriented attendance and poorer home care; both were associated with greater periodontal disease burden, which in turn was associated with higher cardiovascular risk. Model fit was good (chi-square(119) = 529.98, CFI = 0.962, TLI = 0.951, RMSEA = 0.043, SRMR = 0.051). The overall association between DA and cardiovascular risk was small (beta = 0.066, p = .004), with a small decomposed indirect association (beta = 0.013, p = .006) and a remaining direct association (beta = 0.053, p = .022). Smoking-stratified multigroup comparisons found no differences in the structural associations by smoking status. Conclusions: This population-based evaluation provides preliminary empirical support for the conceptual model. Findings extend the relevance of DA beyond dental attendance and oral health outcomes to a broader oral-systemic public-health context. Cross-sectional data cannot establish temporal or causal ordering; longitudinal replication is needed.

Objectives: Dentists prescribe approximately one in ten antibiotics worldwide, yet antimicrobial stewardship (AMS) remains underemphasized in dental education. Large language models (LLMs) may support AMS training, but their proficiency and clinical reasoning in this context remain unclear. We evaluated GPT-4o's accuracy and clinical reasoning on dental antibiotic prescribing questions, stratified by question difficulty. Methods: We assembled 125 multiple-choice questions on dental antibiotic prescribing from eight peer-reviewed studies (2017-2023). GPT-4o answered each question and generated a clinical justification. Accuracy was assessed against source-study answer keys and examined across difficulty quartiles. Justifications were evaluated using an adapted 12-axis human-evaluation framework assessing scientific consensus, extent and likelihood of harm, inappropriate and missing content, bias, and both correct and incorrect comprehension, retrieval, and reasoning. Prophylaxis-specific questions were analysed separately. Results: GPT-4o correctly answered 72% of questions. Accuracy remained relatively stable across difficulty quartiles (78%, 78%, 65%, 70%). Experts rated 95.4% of justifications positively across the 12 axes. Comprehension, retrieval, and reasoning each exceeded 96.2% positive ratings. Missing content was the main weakness (7.8%), and 7.1% of justifications showed a moderate-to-severe potential for harm. Performance on prophylaxis-specific questions (98.1%) exceeded non-prophylaxis questions (93.0%). Conclusions: GPT-4o demonstrated moderate-to-high proficiency and clinically defensible reasoning in dental antibiotic prescribing questions. However, residual risks indicate that it is not suitable for unsupervised clinical use but shows potential as a supervised AMS educational tool.

Since model dose-response is critical when assessing caries lesion development over time, this study evaluated the influence of fluoride dose and treatment duration on caries progression in a rat caries model. Streptococcus mutans-infected Sprague-Dawley rats were treated with deionized water, 226 ppm F-, or 2,260 ppm F- twice daily for 3, 4, or 5 weeks. Caries lesions were assessed using Larson's modification of the Keyes scoring system and complemented by micro-computed tomography (microCT). Intraoral fluoride availability, serum and bone fluoride concentrations and microbial counts were also determined. Fluoride reduced caries severity in a dose- and time-dependent manner. While early enamel lesions were detected in all groups, extensive dentine lesions increased over time, in a dose-dependent manner, in the control and 226 ppm F- groups, and were not observed in the 2,260 ppm F- group after 5 weeks. Intraoral and bone fluoride availability increased significantly with fluoride concentration and treatment duration, whereas serum fluoride levels reflected fluoride dose instead of treatment duration. MicroCT-derived enamel volume correlated negatively with both total and extensive caries scores, supporting its utility as an objective measure of lesion severity. In conclusion, extending model length from 3 to 5 weeks increased the severity of caries lesions in a dose-dependent manner. Fluoride intraoral availability and bone fluoride also demonstrated a dose and time-dependent response.

Background: Cone-beam computed tomography (CBCT) frequently captures the maxillary sinuses incidentally, and reliable automated detection of sinus abnormality is clinically relevant. Unlike most vision-language benchmarks in medical imaging, which pair images with pre-existing, human-authored clinical reports, findings text can also be generated directly by a large language model from the image itself--raising the question of how much diagnostic value such AI-derived text carries, and whether that value depends on independent verification. Multimodal artificial intelligence (AI) benchmarks risk overstating performance if the provenance of each input--image, raw AI-generated text, or radiologist-verified text--is not clearly separated and reported. Methods: We used 300 mid-sagittal CBCT slices from the MMDental dataset. ChatGPT generated findings text and a provisional normal/abnormal label for every slice (majority vote, three independent readings from the image alone); primary classification performance was assessed on this full, unfiltered set (n=300). A radiologist then independently reviewed each case's image together with ChatGPT's description, producing their own diagnosis; three cases were excluded as insufficient, yielding 297 confirmed cases. On this subset, every model was retrained and re-evaluated under identical 10-fold cross-validation on both the provisional ChatGPT-only labels ("pre") and the radiologist-confirmed labels ("post"), isolating the effect of label provenance from image or architecture. Eight vision architectures, seven language classifiers, and five VLMs were evaluated throughout; three generative models performed exploratory note-drafting. Findings: Raw ChatGPT-generated text produced the highest performance of any modality or condition: language models reached near-ceiling AUC (0.992 to 1.000, n=300), exceeding every vision model (AUC 0.799 to 0.880) and every VLM image-only probe (AUC 0.63 to 0.69). On the 297-case pre/post analysis, this advantage depended heavily on label source: language and text-derived VLM performance fell substantially from ChatGPT-only to radiologist-confirmed labels (e.g. BERT-base AUC 0.999 to 0.837), while vision-model performance was stable or modestly improved (e.g. DenseNet-121 0.867 to 0.891). The radiologist reclassified 62 of 297 cases (21%) relative to ChatGPT's provisional read, and a meaningful proportion of raw ChatGPT text was clinically uninterpretable or unsupported by the imaging. Interpretation: As shown here for the first time, raw, image-derived AI-generated text yields the highest apparent classification performance in this benchmark, but this reflects the text's alignment with its own self-generated labels rather than verified diagnostic content, and a substantial share of that text is not clinically explainable. Radiologist-confirmed text and labels give a lower but trustworthy estimate of true performance, on which convolutional neural network (CNN) vision models remain a stable, comparatively inexpensive baseline. Multimodal dental AI should report performance separately by modality and label provenance rather than pooling headline metrics.

Multimodal large language models (MLLMs) are increasingly applied to image-based clinical reasoning, yet their diagnostic reliability in periodontal image interpretation, and the underlying source of their errors, remain poorly characterized. This study evaluated six architecturally distinct MLLMs (Claude Sonnet 4.5, GPT-5.0, Gemini 2.5, GLM-4.6, Sonar, and Grok 4.1) using 50 image-based multiple-choice questions drawn from the American Academy of Periodontology In-Service Examination, spanning clinical photographs, histopathology, radiographs, cardiac rhythm strips, and anatomical illustrations. A sequential two-phase experimental design was used: in Phase 1, each model independently described each image, selected an answer, and provided a supporting citation; in Phase 2, applied only to questions answered incorrectly, models were given an expert-validated visual description and asked to re-answer, allowing diagnostic improvement through visual correction to be measured directly. Expert ground truth for image content was established by a board-certified periodontist and independently validated by a second board-certified periodontist. Model outputs were classified using a dual-process error taxonomy adapted from Norman's model of diagnostic reasoning, distinguishing perceptual errors, arising from inaccurate visual feature extraction, from cognitive errors, arising from flawed reasoning despite accurate perception, with cognitive errors further subdivided into correctable and persistent subtypes, and additional categories capturing compound perceptual-cognitive failures and compensatory reasoning that overcame inaccurate perception. Diagnostic accuracy and error type distribution varied significantly across models and image modality. Correcting inaccurate visual descriptions in Phase 2 improved diagnostic accuracy for a subset of previously incorrect responses, indicating that a meaningful share of errors originated at the level of visual perception rather than clinical reasoning; conversely, a distinct subset of errors persisted despite accurate corrected visual input, indicating reasoning-level failures independent of perceptual accuracy. Some models also reached correct answers despite generating inaccurate image descriptions, reflecting compensatory reasoning resilient to perceptual error. These findings show that aggregate accuracy scores conflate mechanistically distinct failure modes, and that perceptual and cognitive errors carry different implications for how MLLMs might be safely deployed or improved for diagnostic image interpretation. The expert-guided visual correction framework introduced here provides a generalizable, mechanism-based approach to benchmarking multimodal AI diagnostic performance that extends beyond periodontics to other visually driven diagnostic domains in medicine. As MLLMs become increasingly accessible to clinicians, residents, and dental educators, distinguishing perceptual from cognitive failure is essential for guiding responsible clinical use, targeting model refinement, and informing AI-augmented dental education and competency assessment.

Objective Describe the profile of consumers utilising Emergency Departments (ED) across Victorian hospitals for dental conditions between financial year 2018-19 to 2022-23. Methods Dental conditions were grouped according to the diagnosis categories using the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM). Presentations for non-dental conditions were classified as 'Other'. Dependent variables included age, sex, Aboriginal and Torres Strait Islander (Aboriginal) background, language other than English, interpreter required, remoteness, and socio-economic disadvantage according to the Index of Relative Socio-economic Disadvantage (IRSD). The primary outcome evaluated is the relative risk ratio (RRR) of having a hospital ED presentation for a dental condition compared to a non-dental condition and age-specific rates per 100,000 population. Multinomial logistic regression was performed following the backwards elimination method. Results Over the years 2018-19 to 2022-23, 0.55% (n=44,382) of ED presentations across Victorian hospitals by Victorian residents were due to dental conditions. Patients aged 20-44 years have the highest RRR compared to other age groups for dental-related ED presentations. Patients experiencing social inequity in terms of demographic factors such as Aboriginal background, requiring an interpreter, not living in a major city and living in areas of greater disadvantage (IRSD) have statistically significant increased RRR for dental-related ED presentations. Conclusion Patient characteristics were useful for identifying the relative risk of dental-related ED presentations between socio-economic and demographic groups to inform relevant public policy design focus on preventing dental-related ED.

Introduction: Early Childhood Caries (ECC) remains one of the most common chronic conditions in childhood, affecting over 600 million children globally. When untreated, it can lead to severe pain, abscesses, difficulty in chewing that contributes to malnutrition, impaired speech, disrupted sleep, low self-esteem, social withdrawal, and poor academic performance, all of which diminish a childs overall quality of life. Although largely preventable, ECC remains highly prevalent because of delayed preventive care and low parental awareness. Effective health education needs to be consistent, accessible, and available whenever parents need it. Although current evidence indicates that anticipatory guidance is a proven approach, its long-term reinforcement through a mobile app has not yet been tested. This study aims to develop and evaluate the effectiveness of a mobile health (mHealth) application delivering Sustained Anticipatory Guidance (SAG) for the prevention of ECC. Methods and analysis: A two-arm parallel hospital-based randomized controlled trial will be conducted at Sri Ramachandra Institute of Higher Education & Research, Chennai, India. The study will enrol 540 mother-infant dyads from the departments of Pediatric Medicine and the Centre for Early Childhood Caries Research (CECCRe). Eligible dyads with infants aged 6-12 months in the predentate stage will be randomized to the "Zero-Cavity" mHealth app (intervention) or standard care (control). Baseline and semi-annual data on oral hygiene, fluoride exposure, and feeding/diet will be collected. Calibrated dentists will perform quarterly clinical exams using ICDAS II criteria, recording cavitated and non-cavitated lesions. For children at high risk or with enamel defects/incipient lesions, clinical photographs will be taken to support individualized preventive guidance. The primary outcome will be the effectiveness of the app in preventing ECC, while the secondary outcomes will include ECC incidence and prevalence. Discussion: Despite being largely preventable, ECC remains a major public health challenge, sustained by limited preventive access and inconsistent counselling, especially in LMICs. This trial evaluates the Zero Cavity app, believed to bridge this gap by providing sustained, age-appropriate anticipatory guidance beyond clinic settings on oral development, hygiene, diet/feeding, and fluoride use. The behavioural theory-driven approach is expected to enhance cultural relevance, usability, and behaviour change, with optimism that process evaluation will illuminate engagement and implementation. It is believed that this large, adequately powered, parallel-group RCT with intention-to-treat analysis will offer, with optimism, a scalable, low-cost model that will expand the reach and consistency of continuous preventive care in diverse urban and peri-urban populations.

Increasing evidence highlights the important role the oral ecosystem plays in systemic health. Despite the fact that oral diseases affect nearly 3.5 billion people globally, our understanding of the precise drivers of these diseases remains limited. To bridge this gap and provide broader understanding of oral public health, a clinical study involving 595 participants 20-78 years of age was conducted over 8 weeks. Characterization of the oral ecosystem using metabolomics, metagenomics, and targeted immune markers were analyzed in conjunction with demographic, lifestyle, and medical history metadata. This comprehensive approach allowed for the identification of specific links between microbial and metabolic signatures and overall wellness, providing a contemporary baseline to enable the development of personalized prevention and treatment strategies to help safeguard oral and systemic public health.