Salivary Porphyromonas gingivalis was associated with clinically confirmed intra-oral halitosis independently of periodontal status in a cross-sectional study. Investigators found that models incorporating additional bacterial species offered limited improvement over the single-species approach.
The study included 102 adults recruited at Kyung Hee University Dental Hospital in Seoul, South Korea, between May 2021 and October 2022. Of these, 30 had halitosis confirmed by agreement between self-reported symptoms and clinician-rated breath odor (organoleptic score of 2 or higher), while 72 had neither self-reported nor clinically detected halitosis. Participants with discordant assessments were excluded. The authors analyzed unstimulated whole-saliva samples using 16S ribosomal ribonucleic acid gene sequencing and quantitative polymerase chain reaction (qPCR) to characterize bacterial communities and measure microbial concentrations.
Investigators compared logistic regression models using individual bacterial species with models incorporating co-occurring species to classify halitosis status. Model performance was evaluated using repeated stratified 5-fold internal cross-validation across 50 repetitions, yielding 250 training and testing splits. Additional analyses examined differences in bacterial abundance, associations with volatile sulfur compound concentrations, and the influence of periodontal status and demographic characteristics on classification performance.
Among 202 bacterial taxa evaluated, 10 showed statistically significant differences between groups after correction for multiple comparisons, with all 10 enriched among patients with halitosis. P gingivalis showed the strongest difference and was detected in all 30 patients with halitosis compared with 27 of 72 patients without halitosis (100% vs 37.5%). Median sequencing-derived abundance was 72.5 in the halitosis group compared with 0 in the nonhalitosis group.
Targeted qPCR testing confirmed higher P gingivalis concentrations among patients with halitosis, with median bacterial loads of 77,800 vs 10,000 copies/mL. Total bacterial load was also higher in the halitosis group, at 211 million vs 119 million copies/mL. Other species enriched in the halitosis group included Peptostreptococcus stomatis, Parvimonas micra, Solobacterium moorei, andEubacterium nodatum.
A classification model based solely on P gingivalis achieved a cross-validated area under the receiver operating characteristic curve (AUROC) of 0.913, with balanced accuracy of 83.3%, sensitivity of 83.1%, and specificity of 83.5%. Models combining P gingivalis with Tannerella forsythia, and either Prevotella intermedia or Fusobacterium nucleatum achieved AUROCs of 0.961 and 0.962, respectively.
Although the combined models demonstrated numerically higher discrimination, the improvements were not statistically significant after adjustment for multiple comparisons. Comparisons with 66 alternative models that combined P gingivalis with pairs of other bacterial taxa also indicated that the observed gains were not unique to the correlation-derived groups. Models excluding P gingivalis performed approximately at chance level, with AUROCs ranging from 0.410 to 0.574.
The association between P gingivalis and halitosis persisted after accounting for periodontal status. In models adjusted for periodontitis and number of teeth, each 1-standard-deviation increase in bacterial abundance was associated with approximately 9 times the odds of halitosis. Adding periodontal characteristics changed classification performance by less than 0.01 AUROC.
Subgroup analyses showed that P gingivalis was detected in all patients with halitosis across the periodontally healthy, gingivitis, and periodontitis groups. Among patients without halitosis, detection rates were 25%, 35.7%, and 62.5%, respectively. Periodontal status alone demonstrated limited discriminatory ability, with an AUROC of 0.589.
Additional analyses examined associations between salivary bacteria and volatile sulfur compounds associated with oral malodor. P gingivalis abundance correlated with hydrogen sulfide, methyl mercaptan, and total volatile sulfur compound concentrations across the study population. Models incorporating the bacterium explained approximately 30% of the variation in log-transformed volatile sulfur compound concentrations. However, the associations were not observed within the halitosis group alone, suggesting that they primarily reflected differences between groups rather than variation in odor severity among patients with halitosis.
Investigators also evaluated whether demographic and behavioral characteristics improved classification. Models incorporating age, sex, smoking status, and oral hygiene without microbial predictors performed poorly, achieving an AUROC of 0.542. Adding these characteristics to bacterial models did not improve discriminatory performance.
Several limitations warrant consideration. The single-center, cross-sectional design and modest sample size limit generalizability and prevent causal conclusions. Classification models underwent repeated internal cross-validation but lacked external validation, and bacterial co-occurrence groups were identified within the same cohort used for classification. Excluding patients whose self-perceived halitosis differed from clinician assessments may have increased classification performance while limiting applicability to broader clinical populations. Species-level identification from sequencing was also provisional, and the study did not directly assess microbial functional activity. Additionally, the periodontally healthy subgroup included only 5 patients with halitosis. The correlation-based analyses also could not establish biological interactions between co-occurring bacterial species.
The findings support further investigation of salivary P gingivalis as a potential marker of intra-oral halitosis, although its clinical utility remains unestablished.
“Co-occurring periodontal taxa added little discriminatory information beyond P. gingivalis,” wrote lead study author Yeon-Hee Lee, DDS, PhD of the School of Dentistry at Seoul National University, Seoul, South Korea, and colleagues.
The authors declared no competing interests. The study received institutional and research grant funding; the funders reported no role in study design, analysis, or publication decisions.
Source: Scientific Reports
