Understanding the triad of host response, microbiome and disease status is potentially informative for disease prediction, prevention, early intervention and treatment. Using longitudinal assessment of saliva and disease status, we demonstrated that partial least squares modelling of microbial, immunological and clinical measures, grouped children according to future dental disease status. Saliva was collected and dental health assessed in 33 children aged 4 years, and again 1-year later. The composition of the salivary microbiome was assessed and host defence peptides in saliva were quantified. Principal component analysis of the salivary microbiome indicated that children clustered by age and not disease status. Similarly, changes in salivary host defence peptides occurred with age and not in response to, or preceding dental caries. Partial least squares modelling of microbial, immunological and clinical baseline measures clustered children according to future dental disease status. These data demonstrate that isolated evaluation of the salivary microbiome or host response failed to predict dental disease. In contrast, combined assessment of both host response together with the microbiome revealed clusters of health and disease. This type of approach is potentially relevant to myriad diseases that are modified by host-microbiome interactions.