[Construction of a prediction model for peri-implantitis based on logistic regression and analysis of influencing factors].

PURPOSE: To explore the influencing factors of peri-implantitis in patients with dentition defects after oral implant surgery, and to construct and validate a personalized prediction model.

METHODS

Patients who underwent oral implant surgery in the Affiliated Suzhou Hospital of Nanjing University Medical School from September 2021 to March 2025 were included retrospectively. Through 1∶1 propensity score matching, 100 cases in the infected group (diagnosed with peri-implantitis) and 100 cases in the non-infected group were finally included. The baseline data of patients and serum interleukin-17A (IL-17A) levels were collected. Binary logistic regression was used to analyze the influencing factors. A nomogram prediction model was constructed based on the screening results, and the prediction performance was evaluated by Bootstrap validation, receiver operating characteristic(ROC) curve and decision tree model.

RESULTS

The proportions of diabetes, smoking history, chronic periodontitis history, poor alveolar bone around implants, and IL-17A levels in the infected group were significantly higher than those in the non-infected group(P＜0.05). Logistic regression showed that diabetes, smoking history, chronic periodontitis history, poor alveolar bone around implants, and elevated IL-17A were independent risk factors for peri-implantitis(OR＞1, P＜0.05). The C-index of the nomogram model was 0.905, the area under the ROC curve(AUC) was 0.905 (95%CI: 0.865-0.946, P＜0.001), and the optimal cut-off value was 48.80 points (specificity was 0.880, sensitivity was 0.820, Youden index was 0.700). The decision tree model showed that IL-17A was the primary predictor; when IL-17A ＞ 14.380 ng/L, the incidence of peri-implantitis reached 87.50%.

CONCLUSIONS

Diabetes, smoking history, chronic periodontitis history, alveolar bone condition around implants and IL-17A level are key influencing factors of peri-implantitis. The constructed nomogram model has excellent prediction performance and can be used for preoperative individualized risk assessment.