Predicting Dental Caries Outcomes in Children

In recent years, unprecedented gains in the understanding of the biology and mechanisms underlying human health and disease have been made. In the domain of oral health, although much remains to be learned, the complex interactions between different systems in play have begun to unravel: host genome, oral microbiome with its transcriptome, proteome and metabolome, and more distal influences, including relevant behaviors and environmental exposures. A reasonable expectation is that this emerging body of knowledge can help improve the oral health and optimize care for individuals and populations. These goals are articulated by the National Institutes of Health as "precision medicine" and the elimination of health disparities. Key processes in these efforts are the discovery of causal factors or mechanistic pathways and the identification of individuals or population segments that are most likely to develop (any or severe forms of) oral disease. This article critically reviews the fundamental concepts of risk assessment and outcome prediction, as they relate to early childhood caries (ECC)-a common complex disease with significant negative impacts on children, their families, and the health system. The article highlights recent work and advances in methods available to estimate caries risk and derive person-level caries propensities. It further discusses the reasons for their limited utility in predicting individual ECC outcomes and informing clinical decision making. Critical issues identified include the misconception of defining dental caries as a tooth or surface-level condition versus a person-level disease; the fallacy of applying population-level parameters to individuals, termed privatization of risk; and the inadequacy of using frequentist versus Bayesian modeling approaches to derive individual disease propensity estimates. The article concludes with the notion that accurate caries risk assessment at the population level and "precision dentistry" at the person level are both desirable and achievable but must be based on high-quality longitudinal data and rigorous methodology.