Predicting Dental Caries Outcomes in Children: A "Risky" Concept

Abstract

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.

Keywords: disease susceptibility; evidence-based dentistry; oral health; pediatric dentistry; risk assessment; systems biology.

Figure 1.

Schematic illustration of individual variations in early childhood caries (ECC) susceptibility within ECC risk groups. Risk has dimensions of a probability often expressed as a fraction over time or categorically (e.g., low, moderate, and high) at the population level; however, children’s individual diagnoses can vary between only 2 statuses, ECC case or healthy. Importantly, individual children within the same “risk group” have varying causal risk sets (“pies”), preventing the personalization or privatization of risk (as discussed by Rockhill 2001) and the predictive ability of traditional, population-level risk factors at the person level. Upstream determinants (e.g., socioeconomic factors) are the major influences on the population incidence of ECC, with more proximal factors (e.g., simple sugars, as discussed by Sheiham and James 2015) being important risk factors. As illustrated, children stop being part of the at-risk population when they develop active disease because the disease is defined at the person level; however, they can return to the at-risk group if they do not have active disease. In a similar fashion, they are not at risk when they do not have susceptible tooth surfaces and can return to the risk pool when they acquire new susceptible surfaces (e.g., permanent teeth). It is also depicted that children can change risk category at any time.

Figure 2.

Schematic illustration of varying tooth surface–level susceptibilities within a population group estimated to have relatively homogeneous (e.g., moderate) early childhood caries (ECC) risk, within the context of proximal causes and distal determinants of ECC development. Susceptibilities of tooth surfaces vary individually and may be grouped in biologically informative clusters (as discussed by Batchelor and Sheiham 2004; Shaffer et al. 2013): case A depicts the dentition of a child with highly susceptible upper anterior smooth surfaces, B reflects high pit and fissure susceptibility, and C an overall uniform moderate susceptibility. Panels D and E illustrate the potential of using proximal measures of disease activity and a systems biology approach (discussed by Nyvad et al. 2013) to ascertain precise estimates of disease propensity, which, for example, are likely to vary between upper anterior facial (D) and lower proximal [E] surfaces. Achieving “precision dentistry” warrants a comprehensive understanding of genome influences on various proximal and distal factors (e.g., enamel properties and dental anatomy, saliva quality and quantity, oral microbiome, interactions with fluoride) and an ability to integrate multiple-level ‘omics data (Ritchie et al. 2015) including the environmental influences on the structure and function of the genome (depicted as epigenetic effects). For simplicity, a host of major influences is included in the “Environment” category, including social determinants of health, diet, and other oral health behaviors.