A Personalized, Biomarker-Based Risk Assessment Model for Peri-Implantitis: Integration of Clinical, Molecular, and Microbial Predictors

  1. Ioannis Fragkioudakis
  2. Leonidas Batas
  3. Ioannis Vouros
  4. Dimitra Sakellari

Reviewed by

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

Log in to save this article

Abstract

Peri-implantitis is a common and biologically complex complication of dental implants, often diagnosed only after irreversible bone loss has occurred. This study aimed to develop a personalized risk assessment tool by integrating clinical, molecular, and microbial data. A total of 124 participants were stratified into peri-implantitis and healthy/mucositis groups. Clinical parameters, including clinical attachment loss (CAL), were measured, and biological samples were collected for active-matrix metalloproteinase-8 (aMMP-8) quantification, quantitative microbial profiling, and MMP-8 polymorphism genotyping. Multivariable logistic regression identified CAL, elevated aMMP-8 (>20 ng/mL), and Staphylococcus epidermidis load as independent predictors of peri-implantitis, yielding a highly accurate model (Area Under the Curve = 0.982, accuracy = 94.2%). Additionally, MMP-8 mRNA gene expression and aMMP-8 levels were significant predictors of bleeding on probing. Based on the regression model, a clinically applicable, point-based risk assessment tool was developed that integrates three key predictors—CAL, aMMP-8, and MMP-8 SNP (T allele, rs11532004). Weighted scoring enabled stratification into low-, moderate-, and high-risk categories, with scores ranging from 0 to 7. This system demonstrated strong alignment with clinical diagnosis and offers a personalized, screening approach to support early intervention in peri-implantitis. The findings highlight the diagnostic value of combining host-response biomarkers and microbial profiling in assessing peri-implant disease risk. This model advances precision dentistry by enabling early detection and personalized monitoring of peri-implantitis, supporting more targeted prevention and intervention strategies.

Article activity feed

  1. PREreview

    This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/17280462.

    The authors develop a personalized risk‐assessment model for peri-implantitis by integrating clinical parameters (notably clinical attachment loss, CAL), host-response biomarkers (active MMP-8 in peri-implant crevicular fluid), microbial profiling (including Staphylococcus epidermidis load), and genetic susceptibility (MMP-8 promoter SNP rs11532004). In a cross-sectional cohort of 124 participants stratified into peri-implantitis vs. healthy/mucositis, multivariable modeling retained CAL, elevated aMMP-8 and S. epidermidis as independent predictors, achieving very high discrimination (reported AUC up to ~0.98) and yielding a point-based score (0–7) that stratifies patients into low/moderate/high risk. This integrated approach advances precision dentistry by moving beyond retrospective clinical signs to a biologically informed screening tool that could enable earlier detection and tailored monitoring, pending prospective validation

    Major issues

    • Study design & causality. The cross-sectional design precludes temporal inference; the model's ability to predict future peri-implantitis onset/progression remains untested. A prospective, multi-center validation with repeated measures is needed to establish temporal predictive value.

    • External validation & calibration. Reported discrimination is very high (AUC ~0.98 in abstract) but there is limited detail on optimism correction (e.g., bootstrap), calibration (plots, calibration slope), decision-curve analysis, or clinical utility (NRI/IDI). Without these, real-world performance and net benefit are uncertain.

    • Single-center cohort & spectrum bias. Recruitment from one academic center may limit generalizability across implant systems, maintenance protocols, and population microbiomes; spectrum/selection bias is possible (case/control balance, referral patterns). Multi-center sampling is recommended.

    • Predictor handling & thresholds. There is an apparent inconsistency in aMMP-8 cut-offs (e.g., >20 ng/mL in abstract vs. >24 ng/mL in scoring section), and limited detail on how cut-points were derived (Youden, clinical rationale) and whether continuous predictors were checked for nonlinear effects. Please harmonize thresholds and justify binarization.

    • Microbial signal specificity. The inclusion of S. epidermidis as a predictor warrants careful discussion of site contamination risks and ecological plausibility across centers; clarify sampling controls and lab QA to support its robustness as a risk marker.

    • Genetic analysis details. Provide Hardy–Weinberg equilibrium checks, call rates, genotype QC, and population structure control. Explicitly report genetic model (additive/dominant) used for rs11532004 and whether multiple-testing was addressed across molecular/microbial markers.

    • Model reporting (TRIPOD). Ensure full TRIPOD adherence: missing-data handling, candidate predictor pre-specification, events-per-variable, internal validation method, final model coefficients (with CIs), and a worked example of the risk score with predicted probabilities.

    • Clinical integration & workflow. The manuscript would benefit from a clear algorithm for chairside use (e.g., sequence: POC aMMP-8 → CAL → genotype/microbiome), expected retest intervals, and cost/feasibility considerations (turnaround time, device/assay availability).

    Minor issues

    • Terminology consistency. Standardize the aMMP-8 threshold across abstract, methods, and scoring; use a single name and ID for the SNP (MMP-8 −799C>T, rs11532004).

    • Risk categories. Provide the numerical score-to-probability mapping, with calibration plot, and clarify clinical actions tied to low/moderate/high categories (e.g., recall intervals, adjunctive imaging/therapy).

    • Confounders. Expand on adjustment for known risk factors (history of periodontitis, smoking, plaque control, implant depth/time in function), and present a sensitivity analysis including these covariates.

    • Data & code availability. Consider depositing analysis code and de-identified dataset (where ethical) to facilitate reproducibility and external validation.

    • Figures/tables. Add a decision-tree graphic of the scoring system and a calibration plot; improve table footnotes with units and assay details (e.g., POC brand, LOD/LOQ).

    • Editing. Minor editorial issues (commas, spacing) and a few long sentences in Introduction/Discussion could be tightened for clarity.

    Competing interests

    The author declares that they have no competing interests.

    Use of Artificial Intelligence (AI)

    The author declares that they did not use generative AI to come up with new ideas for their review.

  2. Version published to 10.20944/preprints202505.1629.v1