Longitudinal untargeted salivary metabolomic profiles following fluoride varnish application in high-caries and no-caries preschool children

Background Short-term salivary metabolomic changes following fluoride varnish application remain poorly characterized. This study examined longitudinal molecular-feature profiles in preschool children with contrasting caries experience. Methods Twenty children with high caries experience (dmft ≥ 8) and 20 with no caries (dmft = 0) provided saliva at baseline and at 1 and 14 days after varnish application. Samples were analyzed using LC–MS/MS-based untargeted metabolomics. Longitudinal changes were assessed using mixed-effects models and paired contrasts with Benjamini–Hochberg correction. Upstream gap filling was enabled; no additional imputation was applied to the complete post-QC matrix. Supervised models were evaluated using seven-fold cross-validation and 200 label-permutation tests to assess model stability and potential overfitting; nested cross-validation and bootstrap validation were used to examine ROC optimism. Results Of 3,893 initially detected molecular features, 3,196 passed quality-control filtering, including 420 with Level 1/2 annotations. Mixed-effects analysis identified 602 features with an FDR-supported overall time effect, whereas no feature showed an FDR-supported overall caries-group effect or group-by-time interaction. Using the same robustness criterion in both groups, 32 features in the high-caries group and 32 in the no-caries group changed consistently from baseline at both follow-up assessments. Twenty-three persistent feature records were shared by both groups; 11 carried Level 1/2 annotations, of which ten decreased and 18-HETE increased. Eight of these 11 records were excluded from biomarker and mechanistic interpretation; only crotonic acid, DL-stachydrine and 18-HETE were unflagged. The shared day-14 set produced a sparse, nominal sphingolipid-metabolism signal supported by two compounds with opposite directions ( p = 0.00237; q = 0.332). Acetophenone had an apparent AUC of 0.815; baseline classification with feature reselection yielded mean AUCs of 0.558 in nested cross-validation and 0.511 in out-of-bag bootstrap validation. Conclusion Shared short-term salivary molecular-feature changes followed fluoride varnish application. Only three persistent shared Level 1/2 records—crotonic acid, DL-stachydrine and 18-HETE—remain unflagged provisional targets for analytical confirmation. Their identities and quantitative reliability require verification, including clarification of the DL-stachydrine/proline betaine relationship. The sparse sphingolipid-related signal remains hypothesis-generating. Without an untreated comparator, these results describe post-application temporal associations rather than a fluoride-specific mechanism.