Dental caries risk assessment remains challenging, particularly among adolescents, where biological and behavioural factors influence caries risk. The aim of this thesis was to evaluate existing multivariable prediction models for caries development and to investigate microbial and host-related characteristics in dental plaque and saliva associated with caries experience in adolescents. Paper I systematically reviewed and critically appraised development and validation studies of multivariable prediction models for caries development. Twenty-one studies describing 66 prediction models were included. Model performance was generally low to moderate, and only a few models demonstrated clinically useful predictive ability. All studies were assessed as having a high risk of bias, most commonly related to the analysis domain. Previous caries experience was the most frequently retained predictor, whereas biological variables were included less consistently. Paper II investigated bacterial composition, acid tolerance, and metabolomic profiles of dental plaque in adolescents with and without caries experience. No significant difference in overall acid tolerance was observed between groups, although acid tolerance tended to be higher among adolescents with caries experience. Caries experience was associated with differences in microbial community composition and metabolic activity, including higher relative abundances of Abiotrophia, Prevotella, and Veillonella, whereas Rothia and Corynebacterium were associated with caries-free individuals. Community- level microbial and metabolic patterns discriminated between groups more clearly than individual taxa, supporting an ecological understanding of caries. Paper III examined salivary proteins and metabolites in the same adolescent cohort as in paper II. Individual salivary variables generally showed limited discriminatory ability. In contrast, multivariate models combining a small number of salivary compounds demonstrated good discrimination between adolescents with and without caries experience, with area under the curve values ranging from 0.87 to 0.93 across morning and afternoon samples. In conclusion, current caries prediction models show important methodological limitations and limited predictive performance. The findings support the ecological plaque hypothesis and suggest that microbial community characteristics, metabolic activity, and salivary profiles may complement traditional caries risk assessment. Further longitudinal studies are required to determine their value as predictors of future caries development.