Mediation Effect of Dental Caries in the Relationship between Parental Locus of Control and Oral Health-Related Quality of Life

Oral health-related quality of life (OHRQoL) is a multidimensional factor which reflects the level of comfort, self-esteem, and satisfaction with oral health and is an important part of overall well-being [1]. Although OHRQoL is increasingly recognized as an important outcome measure that measures clinically meaningful change in patients [1], there remain gaps in the understanding of factors that influence the OHRQoL of children. While associations between parental socioeconomic status, dental anxiety, oral health behaviors, and OHRQoL have been assessed [2], the effect of parental locus of control (LoC) on OHRQoL has not been well studied. Parental LoC is defined as a parent’s belief in the level of control and efficacy they have over events in their child’s life. Internal LoC refers to an individual’s perception that events and outcomes are determined by one’s choices and behaviors, whereas external LoC is related to an individual’s perception that events and outcomes are determined by external forces, such as bad luck or chance [3]. External parental LoC has been associated with negative child behaviors [4] and adverse child psychological outcomes [5]. However, it is currently unclear if a pathway exists between parental LoC and OHRQoL, as well as the magnitude of this relationship. Few studies have explored the impact of parental LoC on reported child OHRQoL [6]. Parental LoC is potentially a modifiable risk factor of OHRQoL. If a direct effect of parental LoC exists, efforts can be directed to reduce external parental LoC and increase internal parental LoC to improve the OHRQoL of children. These include programs to increase parenting confidence and improve parent-child interactions [4, 7, 8]. One study in Brazil utilized the Multidimensional Health LoC Scale (MHLC) classified parents as having an internal or external LoC based on the relative scores of the internal, external, and chance subscales. No association was found between the binary LoC factor and the OHRQoL of 5-year-old children as measured by SOHO-5 [9, 10]. Another study among adults aged 21–89 years old noted that a chance LoC on the MHLC was associated with a poorer OHRQoL as measured by OHIP-14 but did not find an association between internal LoC and OHRQoL [11]. The impact of LoC, especially as measured on a continuous scale, on OHRQoL is inconclusive. The prior studies were also cross-sectional and do not reflect the longitudinal effect of LoC on OHRQoL. The study, thus, seeks to determine the effect of parental LoC on OHRQoL and to quantify this effect if it exists.

The potential pathways through which parental LoC may affect OHRQoL are unclear. The study hypothesizes that one of the potential pathways through which parental LoC affects OHRQoL is through dental caries. While there may also be other mediators through which parental LoC influences child OHRQoL besides dental caries, such as toothache and traumatic dental injury [9, 10], the study sought to evaluate the mediation effect of dental caries as it is a preventable condition that can be controlled through multiple pathways [12]. Parental LoC has been found to be associated with dental caries. A study reported fewer caries increments with internal LoC, by 1.33 surfaces per unit increase in internal LoC [13]. Other studies have also suggested lower odds of caries in those with stronger internal LoC [14, 15]. One of the key predictors of dental caries among children is oral health-related behaviors [16, 17]. This includes the intake of sugar-containing beverages and foods, use of fluoridated dentifrices, frequency of toothbrushing, and visits to a dental professional. As primary caregivers, parents play a critical role in the development of their children, especially in early childhood [18], and their LoC may influence the diet and oral hygiene practices of young children in early childhood and, in turn, the development of dental caries [9, 14].

On the other hand, dental caries can cause toothache, and when untreated, is associated with poorer oral health-related quality of life (OHRQoL) among children [19‒22]. The greater the severity of dental caries, or the higher the decayed, missing and filled tooth (dmft) index, the poorer the OHRQoL [9, 23, 24]. Dental caries is an endemic condition among children, which imposes a significant burden and carries a risk of morbidity and mortality [25]. Globally, the incidence of dental caries in primary dentition in 2016 was 1.76 billion (95% CI: 1.26 to 2.39 billion) [26, 27]. No studies to date have quantified the average causal mediation effect of dental caries on the relationship between parental LoC and child OHRQoL. In addition, it is unclear if the mediation effect is primarily through active dental caries only, or through caries experience. If dental caries is indeed a mediator in the causal pathway between parental LoC and OHRQoL, timely and cost-effective interventions for the prevention and management of dental caries may potentially mitigate the negative effect of external parental LoC on OHRQoL.

The direct effect of parental LoC on child OHRQoL, and the proportion of the total effect that is mediated through dental caries, including both active caries and caries experience, has not been quantified to date. While parental LoC has been associated with dental caries outcomes [13], and dental caries status has separately been shown to have an impact on child OHRQoL [19, 28], a thorough analysis of the direct effect of parental LoC and mediation effect of dental caries on child OHRQoL can provide insight into the relative effectiveness of strategies to address parental LoC and other strategies to prevent dental caries. The aim of the study was thus to assess the total and direct effect of parental LOC on child OHRQoL and the indirect effect of parental LoC on child OHRQoL mediated through dental caries.

This prospective cohort study was carried out using data collected in the Growing Up in Singapore towards Healthy Outcomes (GUSTO) study, an ongoing multicenter longitudinal cohort study based in Singapore (Fig. 1) [29]. Recruitment of Singaporean mothers aged 18 years and above, and of Chinese, Malay, or Indian ethnicity, in their first trimester took place between June 2009 and September 2010. Mothers and their offspring were followed-up closely through a series of examinations and collection of samples. Ethics approval for the study was granted by the SingHealth Centralized Institutional Review Board (Reference No. 2009/280/D) and the National Health Care Group Domain Specific Review Board (Reference No. D/09/021). Written informed consent was obtained from all participants during recruitment.

Sociodemographic characteristics were obtained at baseline recruitment at birth between 2009 and 2011. Oral health status and dental caries outcomes were assessed when each child was 5 years of age (5th year post-natal visit; between 2014 and 2016). The LoC questionnaire was administered alongside the intraoral examinations. Oral examinations were carried out by 3 trained and calibrated dental professionals (intraclass correlation coefficient [ICC]: >0.80) using the modified International Caries Detection and Assessment System (ICDAS-II) criteria [30]. Compared to histological classifications, ICDAS-II has been reported to have a high sensitivity and specificity, with summary receiver operating characteristics curves above 0.75 and above 0.90 at D1 (enamel and dentine) and D3 (dentine only) thresholds, respectively [31]. Inter-rater and intra-rater reliability scores were above or equal to 0.67 [31]. The participants were examined in the supine position, using a torch as a source of illumination [32]. The tooth status was assessed by visual inspection using a mouth mirror, as well as tactile inspection with a blunt probe, after cleaning and drying tooth surfaces with sterile gauze. No radiographs were taken [32]. Using the ICDAS second digit codes as the thresholds for defining the presence of tooth decay, scores were summated across all teeth to derive decayed, missing, and filled surfaces (dmfs) representing caries experience, and decayed surfaces (ds) scores representing active caries for each child. A second digit between 1 and 6 was used to score the presence of any tooth decay including incipient caries (ICDAS 1–6), and a second digit between 3 and 6 was used to score the presence of cavitated lesions only (ICDAS 3–6). The 13-item LoC questionnaire was developed based on the LoC theory [14]. Responses were coded on a 5-point Likert scale (strongly disagree [1] to strongly agree [5]), and items expressing external LoC and a belief in chance were reverse-coded. All scores were aggregated through a summation of responses to all questions, with the potential range of total scores from 13 to 65. A higher total score suggests a stronger internal LoC, while a lower total score suggests a stronger external LoC.

At the 6-year post-natal visit (between 2015 and 2017), the Early Childhood Oral Health Impact Scale (ECOHIS) was administered to mothers to assess the OHRQoL of their child [33]. ECOHIS is a tool measuring the impact of oral health issues and treatment experiences on the quality of life of children aged 3–5 years old, and their families. Studies have assessed that ECOHIS has adequate construct validity with no ceiling effects [34, 35]. In terms of internal consistency, reported Cronbach’s alpha and intraclass correlation coefficients were high above 0.7 [22, 34, 36‒42]. The 13-item survey was completed by a proxy or parent of the child, and it comprised two sections (child impact and family impact), and six domains (symptoms, function, psychological, self-image/social interaction, parent distress, and family function). Mothers were asked to consider the child’s entire lifespan for the ECOHIS questionnaire. Responses to each question were coded on a five-point Likert scale (0 = never; 1 = hardly ever; 2 = occasionally, 4 = very often; 5 = do not know), with scores tabulated as an unweighted sum of responses after excluding “do not know” and missing responses [35]. Child impact and family impact scores were calculated by summing the scores of the items under the respective sections. The total score, thus, ranged from 0 to 52 while the child impact and family impact scores ranged from 0 to 36 and 0 to 16, respectively. If there were up to two missing items in the child impact section or up to one missing item in the family section, the mean of remaining items in each section was used to impute the scores for missing items. Questionnaires with more than two missing items in the children impact section and more than one missing item in the family impact section were excluded [35]. Higher ECOHIS scores represented higher impacts and poorer OHRQoL. ECOHIS was assessed on a continuous scale to assess the effects of the exposure on the severity of the outcome.

A directed acyclic graph (DAG) was developed on DAGitty [43], based on existing literature (online suppl. material S1; for all online suppl. material, see https://doi.org/10.1159/000545620) to characterize the theorized causal relationship between exposure (parental LoC), mediator (caries status), and the outcome (ECOHIS score) (Fig. 2). Caries status was measured by dmfs or ds and further subdivided into ICDAS 1–6 and ICDAS 3–6. Potential confounding factors were included as covariates in the analyses. These included the mother’s ethnicity (Chinese, Malay, or Indian), the mother’s highest education level (up to secondary school education, post-secondary, and above), mother’s age at delivery, household income (≤SGD 3,999, ≥SGD 4,000), and housing type (1–3 room public housing flats, 4–5 room housing flats, private housing, or others).

In other words, the total causal effect is equal to the sum of the mediation effect under one exposure and the natural direct effect under the other exposure. The degree of violation of the sequential ignorability assumption was assessed using the medsens package (online suppl. S2) [46]. The rho (ρ) at which the indirect effect turned 0, measured by the correlation between the error terms in the mediator and outcome models, was used to quantify the violation of the sequential ignorability assumption [47].

As interactions may be present [48], exposure-mediator interaction was assessed in separate models, including the interaction term between parental LoC and caries status. To assess for interactions, besides the LoC group, caries status was also dichotomized into two categories, based on the caries experience status (dmfs [ICDAS 1–6 or 3–6] ≥1 or <1) or the presence of active caries (ds [ICDAS 1–6 or 3–6] ≥1 or <1).

As part of robustness checks, the analyses were also carried out using the mediate package [49], with total effect, natural indirect effect, pure natural indirect effect, natural direct effect, and total natural direct effect assessed based on the decomposition recommended by Nguyen et al. [50]. A sequential approach was also adopted with inverse probability weighting to account for potential exposure-mediator and exposure-outcome interactions, along with bootstrapping with 1,000 replications to obtain 95% confidence intervals (online suppl. S3) [51].

A total of 542 children completed visits in both the 5th and 6th years (shown in Fig. 1). Of these, 25 were missing LoC data, and 211 were missing caries data (of which 23 were also missing LoC data). A total of 312 children had complete sociodemographic data, completed the oral examination at the 5th year post-natal visit, and had data for the LoC questionnaire in the 5th year and the ECOHIS questionnaire in the 6th year. The sociodemographic characteristics of the participants are presented in Table 1. At the 5th year post-natal visit, the median dmfs was 2 (IQR 0–5), while the median LoC was 47 (IQR 43–50). The distribution of characteristics of individuals excluded due to missing data (n = 230) compared with those with complete data (n = 312) is presented in online supplementary Table S1.1.

There was a higher proportion of Chinese and lower proportion of Malays in the internal LoC group compared to the external LoC group (Table 1). Higher proportions of mothers with post-secondary and university education, as well as children from households with incomes of SGD 4,000 and above, were also observed in the internal LoC group compared to the external LoC group (Table 1). Compared to those living in smaller 1- to 3-room public flats, a higher proportion of children staying in larger 4- to 5-room public flats and living in private housing were classified under the internal LoC group (Table 1).

With parental LoC on a continuous scale and evaluated at the mean value of the lowest and highest quintile, the total effect of internal parental LoC on ECOHIS, relative to external parental LoC, was −2.05 (95% confidence interval [CI], −4.03 to −0.08), of which the direct effect was approximately −1.45 (95% CI: −3.41 to 0.41) (Table 1). With the mediator set as dmfs (ICDAS 1–6), the indirect effect was −0.61 (95% CI: −1.28 to −0.08). The percentage of the total effect of parental LoC on ECOHIS that was mediated by the presence of dental caries was 29.5%, 30.1%, 23.7%, and 28.3%, respectively, for dmfs (ICDAS 1–6), dmfs (ICDAS 3–6), ds (ICDAS 1–6), and ds (ICDAS 3–6) (Table 2).

When categorized into two LoC groups, the total effect of internal parental LoC was −0.88 (95% CI: −2.20 to 0.45), with a direct effect of −0.63 (95% CI: −1.94 to 0.64). The indirect effect of dmfs (ICDAS 1–6) was −0.25 (95% CI: −0.66 to 0.11). The percentage of total effect mediated was 25.2%, 13.6%, 24.1%, and 16.3% for dmfs (ICDAS 1–6), dmfs (ICDAS 3–6), ds (ICDAS 1–6), and ds (ICDAS 3–6), respectively (Table 3). The rho at which the average causal mediation effect equates 0 was 0.26, 0.29, 0.20, and 0.24, respectively.

The coefficient of the interaction term between LoC group and caries status (dmfs [ICDAS 1–6]) in the association between LoC group and ECOHIS score was −1.37 (95% CI: −4.12 to 1.39). Following the inclusion of the exposure-mediator interaction term, the average mediation effect via dental caries was −0.12 (95% CI: −0.37 to 0.05), −0.12 (95% CI: −0.46 to 0.20), −0.10 (95% CI: −0.34 to 0.06) and −0.09 (95% CI: −0.36 to −0.17) with the mediator set as dmfs (ICDAS 1–6), dmfs (ICDAS 3–6), ds (ICDAS 1–6), and ds (ICDAS 3–6), respectively. The total effect was −0.87 for dmfs/ds (ICDAS 1–6) and −0.89 for dmfs/ds (ICDAS 3–6), while the average direct effect ranged from −0.75 (95% CI: −2.14 to 0.59) to −0.80 (95% CI: −2.22 to 0.47), respectively (Table 4). The proportion mediated was 11.2%, 11.5%, 9.5%, and 8.2%, respectively (Table 4).

Results of the sensitivity analyses using the mediate package are presented in online supplementary Tables S2.1 and S2.2. The indirect effect was −0.62 (95% CI: −1.16 to −0.08) without interaction terms, while the natural indirect effect was −0.95 (95% CI: −1.91 to 0.02) and the pure natural indirect effect was −0.42 (95% CI: −1.05 to 0.22) after accounting for interaction effects. Smaller mediation effects were noted using the sequential approach (online suppl. Table S3.1).

This study found an effect of parental LoC on the OHRQoL of children. When assessed as a continuous variable, the exposure (internal parental LoC) had an overall negative total effect on child ECOHIS scores compared to external parental LoC, with lower ECOHIS scores suggesting a better OHRQoL. However, the study did not manage to detect an effect of parental LoC on child OHRQoL when parental LoC was dichotomized into internal and external LoC. This study found that 23.7% of the total effect of parental LoC on ECOHIS was mediated through active decayed tooth surfaces (ds [ICDAS 1–6]). A higher mediated effect was found if only cavitated lesions were considered (ds [ICDAS 3–6]), with incipient caries excluded. Similarly, a higher mediated effect was observed if previously treated carious lesions (missing and filled surfaces; dmfs) were included beyond active carious lesions (ds). The direct effect of parental LoC on ECOHIS was not statistically significant after accounting for the mediation effect through dental caries.

There may be a loss in power [52] when continuous exposures (LoC) were categorized and dichotomized in this study, resulting in the inability to detect an effect on OHRQoL. Using the Multidimensional Health Locus of Control survey which categorized parental LoC by whether the lowest score was in the internal, external or chance subscale, Gomes et al. [10] found no association between internal and external parental LoC and the child version of SOHO-5. Similarly, Granville-Garcia et al. [9] did not find an association between internal and external LoC measured using the Multidimensional Health Locus of Control Scale on OHRQoL measured by the 14-item SOHO-5 questionnaire that captured both child and parent responses. It is also unclear if the lack of association between parental LoC and OHRQoL in prior studies was due to the administration mode, with child responses collected for SOHO-5 as opposed to solely parent or caregiver (proxy) responses for ECOHIS, or other differences in psychometric properties between SOHO-5 and ECOHIS. Non-differential misclassification of OHRQoL obtained from children may also have biased the findings toward the null.

In line with other studies [7, 24], this study found that greater severity of dental caries (higher dmfs and ds scores) was associated with more external parental LoC (lower LoC scores) and poorer OHRQoL (higher ECOHIS scores). Past caries experience, beyond just active caries, may also explain the pathway between parental LoC and child OHRQoL given this study’s finding of a greater mediation effect of dmfs compared to ds alone. While there are few studies that assessed the mediation effect of dental caries on the effect of parental LoC on child OHRQoL, Gomes et al. [10] found that compared to children with no dental caries, there was a higher relative risk for poorer OHRQoL as measured by a higher SOHO-5 score for children with cavitated dental lesions after adjusting for other individual and contextual factors. Nonetheless, the adjusted relative risk for poorer OHRQoL among children with white-spot lesions was only similar to those with cavitated lesions [10].

In terms of exposure-mediator associations, this study’s findings are consistent with studies that found that a higher internal parental LoC was associated with lower odds of dental caries [7, 24]. Findings from a previous study that adopted the Basic Research Factors Questionnaire among American Indian caregivers of preschool children suggested that a unit increase in caregiver internal LoC was associated with a smaller increase in dmfs over 1 year by 1.33 surface [13]. However, external LoC and chance LoC were not associated with change in dmfs [13]. Using the same 13-item LoC questionnaire as that in this study, Lencová et al. [14] noted a linear trend with higher odds of being free from untreated dental caries (dt = 0) and intact teeth (dmft = 0) with higher quintiles of internal parental LoC. Dental caries has been shown to have a negative impact on OHRQoL in multiple studies [19‒21], with poorer OHRQoL with increasing caries activity, severity, and number of affected teeth [23, 53]. Exposure-mediator interaction effects on OHRQoL were not detected in this study.

This is the first study assessing the average causal mediation effect of dental caries among children on the relationship between parental LoC and child OHRQoL. The results of this study suggest that oral health education for caregivers should focus on increasing their internal parental LoC by highlighting the role and impact of caregivers in preventing dental caries in their children [7, 8]. Oral health education for caregivers has been found to be effective in preventing early childhood caries [54], and ECOHIS scores have been found to be associated with health literacy [55]. Among parents with external LoC who are resistant to change, preschool-based interventions such as topical fluoride applications [56] can potentially help mitigate the impact on the child’s OHRQoL.

Nonetheless, this study has a few limitations. While the study was able to account for most pretreatment covariates in the assessment of the mediation effect of dental caries, there may be other contextual factors that may confound the exposure-outcome relationship between parental LoC and child OHRQoL such as child sense of coherence [57]. However, some studies did not find an association between sense of coherence and OHRQoL [9]. Furthermore, as ECOHIS is a proxy-administered measure of child OHRQoL, it is less likely to be influenced by the child’s sense of coherence. Exposure-induced mediator-outcome confounding may also occur through factors such as dental care utilization, which could not be quantified in this study. Nonetheless, this study sought to assess the impact of violating the sequential ignorability assumption [44]. Across the LoC group models without interaction terms, at a rho of between 0.20 and 0.29, an omitted confounder must explain at least 20% of the variance (Table 3) for the average causal mediation effect to be zero. A rho of 0.3 was said to represent a modest violation of the sequential ignorability assumption [58]. Future studies may also collect data on these additional factors and account for their impact on the exposure-outcome relationship. In addition, visual inspection of children’s teeth using a torch instead of on a dental chair may result in misdiagnosis of dental caries, although any misclassification is likely to be non-differential by LoC group. There may be potential selection bias as a substantial number of children did not undergo dental examinations in the GUSTO study due to scheduling conflicts with dental examinations (n = 211). While multiple imputation could be considered, this was not carried out in view of the large amount of missing data on dental caries status which would impact the precision of estimates [59]. Given that there was a higher proportion of children from lower income families and whose mothers had a highest education level of secondary and below in this study compared to those excluded (online suppl. Table S1.1), and that these factors were independently associated with more external parental LoC and poorer OHRQoL, the true effect of parental LoC on OHRQoL may be higher than that estimated in this study. Finally, there may be multiple mediators in the pathway from parental LoC to child OHRQoL. The mediation effect of a history of toothache and dental pain on child oral health-related life was not evaluated, although toothache may be considered a sequela of dental caries [9, 10, 60]. Given that the proportion mediated by dental caries is 30% or lower across models, the effect of other mediators can be further explored in future studies.

In conclusion, this study found that parental LoC had an impact on child OHRQoL, and this effect was partially mediated through dental caries. There is thus room for interventions to ensure the OHRQoL of children through community-based measures to improve parental LoC and prevent dental caries in children. Future studies to design and assess the effectiveness of interventions to improve parental LoC may be conducted.

The study team would like to sincerely thank the late Dr. Rahul Nair for conceptualizing this study and providing his invaluable guidance and contributions to the study. We would also like to thank the GUSTO Study Group: Airu Chia, Andrea Cremaschi, Anna Magdalena Fogel, Anne Eng Neo Goh, Anne Rifkin-Graboi, Anqi Qiu, Arijit Biswas, Bee Wah Lee, Birit Froukje Philipp Broekman, Candida Vaz, Chai Kiat Chng, Chan Shi Yu, Choon Looi Bong, Daniel Yam Thiam Goh, Dawn Xin Ping Koh, Dennis Wang, Desiree Y. Phua, E. Shyong Tai, Elaine Kwang Hsia Tham, Elaine Phaik Ling Quah, Elizabeth Huiwen Tham, Evelyn Chung Ning Law, Evelyn Keet Wai Lau, Evelyn Xiu Ling Loo, Fabian Kok Peng Yap, Falk Müller-Riemenschneider, Franzolini Beatrice, George Seow Heong Yeo, Gerard Chung Siew Keong, Hannah Ee Juen Yong, Helen Yu Chen, Hong Pan, Huang Jian, Huang Pei, Hugo P.S. van Bever, Hui Min Tan, Iliana Magiati, Inez Bik Yun Wong, Ives Lim Yubin, Ivy Yee-Man Lau, Jacqueline Chin Siew Roong, Jadegoud Yaligar, Jerry Kok Yen Chan, Jia Xu, Johan Gunnar Eriksson, Jonathan Tze Liang Choo, Jonathan Y. Bernard, Jonathan Yinhao Huang, Joshua J. Gooley, Jun Shi Lai, Karen Mei Ling Tan, Keith M. Godfrey, Keri McCrickerd, Kok Hian Tan, Kothandaraman Narasimhan, Krishnamoorthy Naiduvaje, Kuan Jin Lee, Li Chen, Lieng Hsi Ling, Lin Lin Su, Ling-Wei Chen, Lourdes Mary Daniel, Lynette Pei-Chi Shek, Maria De Iorio, Marielle V. Fortier, Mary Foong-Fong Chong, Mary Wlodek, Mei Chien Chua, Melvin Khee-Shing Leow, Michael J. Meaney, Michelle Zhi Ling Kee, Min Gong, Mya Thway Tint, Navin Michael, Neerja Karnani, Ngee Lek, Noor Hidayatul Aini Bte Suaini, Oon Hoe Teoh, Peter David Gluckman, Priti Mishra, Queenie Ling Jun Li, Sambasivam Sendhil Velan, Seang Mei Saw, See Ling Loy, Seng Bin Ang, Shang Chee Chong, Shiao-Yng Chan, Shirong Cai, Shu-E Soh, Stephen Chin-Ying Hsu, Suresh Anand Sadananthan, Swee Chye Quek, Tan Ai Peng, Varsha Gupta, Victor Samuel Rajadurai, Wee Meng Han, Wei Wei Pang, Yap Seng Chong, Yin Bun Cheung, Yiong Huak Chan, Yung Seng Lee, and Zhang Han.

Ethics approval for the study was granted by the SingHealth Centralized Institutional Review Board (Reference No. 2009/280/D) and the National Health Care Group Domain Specific Review Board (Reference No. D/09/021). Written informed consent was obtained from all participants at recruitment.

The authors have no conflicts of interest to declare.

This research is supported by the Singapore National Research Foundation under its Translational and Clinical Research (TCR) Flagship Programme and administered by the Singapore Ministry of Health’s National Medical Research Council (NMRC), Singapore (NMRC/TCR/004-NUS/2008; NMRC/TCR/012-NUHS/2014). Additional funding is provided by the Singapore Institute for Clinical Sciences, Agency for Science Technology and Research (A*STAR), Singapore. This study was also funded by Singapore’s NMRC (grant NMRC/CIRG/1341/2012) and National University Health Systems (NUHS) Bridging Funds 02/FY16 (R-221-000-110-733). S.H.X. Tan is supported by the National University Health System Clinician Scientist Program (NCSP 2.0) and the NMRC Research Training Fellowship Award. C.S. Hsu is supported by the NMRC Grant: CIRG12may049 “Building Oral Microbiome to Identify Novel Biomarkers/Modulators for Early Childhood Caries and Oral-Systemic Link.”

C.S. Hsu contributed to the conception and design of the work. C.S. Hsu, J.G. Eriksson, and K.H. Tan contributed to the acquisition of the data. S.H.X. Tan contributed to the analysis of the work and drafting the work. A. Singh, J.G. Eriksson, K.H. Tan, and C.S. Hsu contributed to the interpretation of data and reviewed the work critically for important intellectual content. All authors give final approval of the version to be published and agree to be accountable for all aspects of the work.

The datasets generated and/or analyzed during the current study are not publicly available due to an ethical restriction (patient confidentiality) which was imposed by the Centralized Institutional Review Board of SingHealth. Interested researchers may request the data by contacting the data team leader of GUSTO at [email protected].