Investigation of Diagnostic and Prognostic Testing for Peri-implantitis Using Quantitative Metabolomics

Abstract

Background: Research has primarily focused on finding diagnostic and prognostic tests that can distinguish between peri-implant health and disease. Current diagnostic modalities for peri-implant disease include radiographic and clinical measures that measure damage from previous episodes of peri-implant breakdown and are unable to predict susceptibility to future peri-implant disease. To date, biomarkers found in peri-implant sulcular fluid (PISF) have demonstrated low accuracy and predictability at diagnosing peri-implantitis. Recent advances in metabolomics have been of interest in the field of periodontics owing to its potential ability of providing more in-depth information on disease processes.Aim: Determine the spectra of metabolites found in PISF that can discriminate between peri-implant health and disease, and be used to accurately diagnose peri-implantitis. Methods: In a cross-sectional study, the PISF from 33 peri-implantitis and 26 healthy control subjects was collected around healthy (probing depth ≤3mm and radiographic bone loss <2mm) and diseased implants (probing depth ≥6mm and radiographic bone loss ≥3mm). PISF samples were analyzed using proton nuclear magnetic resonance (H-NMR) spectroscopy, to obtain 2D proton spectra profiles with water suppression pulse. Regions of interest (ROIs) were defined based on Total Correlation Spectroscopy (TOCSY) data from two public databases (MMCD and HMDB). Signal intensities for each ROI in PISF spectra were generated using rNMR software. A total of 35 PISF metabolites were assigned. The correlation of each individual metabolite with health or disease status was calculated with Spearman’s coefficient. The predictive ability of a metabolite and a combination of metabolites to diagnose peri-implantitis was determined via receiver operating curves. Results: Cadaverine/lysine, propionate, alanine/lysine, putrescine/lysine, valine, tyramine and threonine were significantly correlated with disease, whereas -ketoglutarate, isoleucine, proline and uracil were significantly correlated with a healthy state. AUROC values for individual metabolites correlated with disease were statistically significant (p<0.05) and ranged between 0.606 and 0.617. The combination of 2 to 4 metabolites slightly increased the AUROC values and ranged between 0.624 and 0.653 (p-value <0.05). Conclusions: Diseased peri-implant sites demonstrate a spectrum of metabolites that are statistically different than those from healthy peri-implant sites. Certain metabolites are positively and negatively correlated with disease. However, individual metabolites and the combination of 2 to 4 metabolites showed a low discriminatory ability (low sensitivity and specificity) to differentiate between peri-implant health and disease.