Dental caries (carious oral microbiome) vs Caries-free / healthy oral microbiome: Discrimination of caries vs health (random forest ROC-AUC)
Microbiome-based random forest classifiers could not reliably discriminate caries from health across the combined cohort (best AUC approximately 0.75 in a disease-only KO subset; ungraded).
What each review found
Include earlier review versions| Finding | Effect | Studies | GRADE certainty |
|---|---|---|---|
Dental caries (carious oral microbiome) vs Caries-free / healthy oral microbiome Discrimination of caries vs health (random forest ROC-AUC) · Individuals providing oral (saliva/supragingival plaque) 16S rRNA microbiome samples; 2016 samples from 19 studies… · Oral cavity microbiome (saliva and supragingival plaque samples); pooled across 19 international studies Microbiome-based random forest classifiers could not reliably discriminate caries from health across the combined cohort (best AUC approximately 0.75 in a disease-only KO subset; ungraded). Meta-analysis of caries microbiome studies can improve upon disease prediction outcomes. (2022) All outcomes | AUC 0.75 | 19 studies · 2016 participants | GRADE not reported |
Findings are shown side by side, never pooled across reviews.
The reviews
| Review | Searched up to | Included studies | AMSTAR 2 |
|---|---|---|---|
| Meta-analysis of caries microbiome studies can improve upon disease prediction outcomes. 2022 All outcomes | Not specified (studies published from 2012 onward included; no explicit search date reported) | 16 | Not applicable |
Studies behind the findings
None of these reviews ties its included studies to this outcome in the data here. Cochrane reviews and reviews with headline outcomes only do not list their studies yet.
How reviews are matched to this question
Reviews word the same question differently. Four decision models (TypeSafe's Jev, Liquid AI's d1-ultra, SubQ Snap and Baseten's Elephant) judged each wording of the population, intervention, comparator and outcome against the others; two wordings count as the same when at least three of them agree and none clearly disagrees, and findings count as this question only when all four elements match. Timing, setting and subgroup never split a question: they are shown on each finding.