Discussion
The present study findings demonstrated that children with moderate-to-severe AD, but not mild AD, to be more susceptible to ECC development, compared to those without the disease. Although complex pathogenic mechanisms of AD are still evolving, current concepts suggest that defective skin barrier may be a potential driving factor in AD pathogenesis and not a consequence of the disease 17. As skin/hair and dental enamel originate from ectoderm, we proposed a sub-clinical structural hypothesis to explain this link based on presence of hair keratin proteins in the mature enamel organic matrix 18 and association of its mutations with increased caries risk 19. This was substantiated in a prospective twin study reporting children with infantile eczema to be at 2 times higher risk of developmental enamel defects 20, which is a well-established risk factor for ECC susceptibility 21.
Degree/type of skin barrier dysfunction may be one of the major differences between mild and severe forms of the disease, as evidenced by, a) a two-fold increased risk of genetic defects in skin proteins (filaggrin, FLG ) in severe AD individuals compared to mild AD22, and b) an increased skin permeability dysfunction (transepidermal water loss, TEWL) in severe AD individuals, compared to milder forms of the disease 23. Furthermore, polymorphisms in genes regulating skin barrier function/homeostasis have been linked with tooth developmental anomalies. First, defects in desmosomal proteins such as desmoplakin have been associated with severe AD 24 and enamel dysplasia 25, respectively. Second, defects in genes encoding for laminin-332 and type XVII collagen have not only been associated with AD26, but also play a critical role in differentiation of enamel-forming (ameloblasts) cells 27. Hence, it may be plausible that severity of AD (with corresponding skin barrier defects) may parallel degree of enamel defects and hence risk of caries. Further characterization of enamel proteins in exfoliated deciduous teeth from children with moderate-to-severe AD and non-AD are ongoing and may elucidate or even confirm the pathogenic pathway proposed in the study.
It can be argued that children with exacerbated AD symptoms, especially in moderate-to-severe AD, may exhibit compromised compliance and/or attention towards oral hygiene 4 and thus lead to increased caries susceptibility. Hence, the association between AD severity and ECC was further tested controlling for post-natal dietary (such as duration of breastfeeding, child’s daily frequency of sweet snacks) and oral hygiene factors (such as child’s brushing frequency) (Table 3). Consistently, similar findings were revealed, with moderate-to-severe AD cases to be at 2-fold higher ECC risk compared to those without AD, substantiating an underlying biological link independent of the dietary-behavioural effect, between moderate-to-severe AD and ECC.
Diagnosis of AD, especially in epidemiological studies, may be subject to potential errors especially in in young children. Conventionally, caregiver-reported history of doctor-diagnosed AD is employed in epidemiological studies 9, with a reported sensitivity of 70% and a positive predictive value (PPV) of 87% in children/adolescents (<18 years) in a referral dermatological care setting with high prevalence rates (∼29.5%) 10. However, in the present study when the true disease outcome was low (7.3%), our results showed lower sensitivity (51.9%) and PPV (24.8%) in identifying true AD cases (Table 4). These findings indicate that although caregiver-reported criteria may be useful in estimating AD burden in populations with high prevalence rates, its utility in very young children and/or cohorts with low prevalence rates may be compromised. Furthermore, although more than two-thirds of moderate-to-severe AD children in the first 2 years of life demonstrate sensitization to food allergens 11, low PPV (22.6%) and sensitivity (40.3%) of a positive skin prick test suggested that allergic sensitization alone may not be sufficient for AD diagnosis in early-life. In contrast, combining a positive SPT with a caregiver-reported history of doctor-diagnosed AD, as employed in our previous work 8, resulted in a ~2-fold higher PPV (45.4%) compared to single use of SPT or caregiver-report alone. However, since overall PPV values still remained low (inflated Type I error), its utility in epidemiological studies needs to weighed against the more established diagnostic systems such as Hanifin and Rajka criteria and UK Working Party criteria 28.
The strengths of this study include the longitudinal study design, use of physician diagnosis in AD estimation, and enumeration of caries at the surface-level (number of surfaces affected), instead of individual-level outcome (yes/no) reported previously8. Study limitations include few moderate (N=12) and severe AD (N=2) cases available in the cohort. Second, missing data seen due to multiple time-points of data collection reduced the sample size (N=581). Nevertheless, post-imputation results using multiple imputation by chained equation (creating 40 imputed datasets), demonstrated that exclusion of few missing cases did not affect the results of the study (Appendix Table 2). Third, it was logistically impossible to obtain inter- and intra-examiner reliability among physicians for AD assessment, although both Hanifin-Rajka criteria 29and SCORAD scores 30 have been shown to exhibit high inter-rater agreement (ICC>0.80) among physicians/dermatologists. Fourth, severity of AD cases may be underestimated in the current study owing to possible regression of skin lesions before the time of clinic visits and the logistic inability to obtain SCORAD in the first year when the prevalence of the disease is at its peak 1. Lastly, there may be potential residual confounding as not all confounding factors could be accounted in the analysis due to constraints of sample size.
In conclusion, children with moderate-to-severe atopic dermatitis demonstrated a two-fold increased caries risk at 3 years compared to those without the disease, and early dental referral may be beneficial. Furthermore, diagnostic accuracy of caregiver-reported history of doctor-diagnosed AD, skin prick test, and the combined criteria may be compromised in identifying true AD cases in certain populations with low AD prevalence rates.