Foetal, neonatal and child vitamin D status and enamel hypomineralization

Abstract Objectives Recent literature suggested that higher vitamin D concentrations in childhood are associated with a lower prevalence of molar incisor hypomineralization (MIH). As tooth development already starts in utero, we aimed to study whether vitamin D status during foetal, postnatal and childhood periods is associated with the presence of hypomineralized second primary molars (HSPMs) and/or MIH at the age of six. Methods Our study was embedded in the Generation R Study, a population‐based, prospective cohort from foetal life onwards in Rotterdam, the Netherlands. HSPMs and MIH were scored from intraoral photographs of the children at their age of six. Serum 25(OH)D concentrations were measured at three points in time, which resulted in three different samples; mid‐gestational in mothers’ blood (n = 4750), in umbilical cord blood (n = 3406) and in children's blood at the age of 6 years (n = 3983). Results The children had a mean (±SD) age of 6.2 (±0.5) years at the moment of taking the intraoral photographs. After adjustment for confounders, no association was found between foetal 25(OH)D concentrations and the presence of HSPMs (OR 1.02 per 10 nmol/L higher 25(OH)D, 95% CI: 0.98‐1.07) or MIH (OR 1.05 per 10 nmol/L increase, 95% CI: 0.98‐1.12) in 6‐year‐olds. A higher 25(OH)D concentration in umbilical cord blood resulted in neither lower odds of having HSPM (OR 1.05, 95% CI: 0.98‐1.13) nor lower odds of having MIH (OR 0.95, 95% CI: 0.84‐1.07) by the age of six. Finally, we did not find higher 25(OH)D concentrations at the age of six to be associated with a significant change in the odds of having HSPM (OR 0.97, 95% CI: 0.92‐1.02) or MIH (OR 1.07, 95% CI: 0.98‐1.16). Conclusions 25(OH)D concentrations in prenatal, early postnatal and later postnatal life are not associated with the presence of HPSMs or with MIH at the age of six. Future observational research is required to replicate our findings. Furthermore, it is encouraged to focus on identifying other modifiable risk factors, because prevention of hypomineralization is possible only if the causes are known.


| INTRODUCTION
Dental enamel hypomineralization is an anomaly of dental enamel in which the affected enamel contains less mineral than sound enamel and is more susceptible to caries. [1][2][3] This anomaly can be divided into hypomineralization of second primary molars, called hypomineralized second primary molars (HSPMs), and hypomineralization of permanent first molars, called molar incisor hypomineralization (MIH). [3][4][5] In patients with MIH incisors of the upper jaw can also be involved and in rare cases incisors of the lower jaw. 3 Although hypomineralization is not restricted to those few index teeth and can be diagnosed in any tooth of both dentitions, a patient can only be diagnosed with HSPM/MIH if he or she has at least one affected second primary molar or first permanent molar, respectively. 6 The prevalence of HSPMs is about 4.9% in 6-year-old Dutch children. 7 For MIH, the prevalence ranges between 8% and 19% among Dutch and Scandinavian children aged six to thirteen years. 3,5,7,8 Children with HSPM have a higher chance of developing MIH. 9,10 Identifying modifiable risk factors is important to prevent development of dental enamel hypomineralization in children.
Several early life risk factors for HSPM and MIH have been identified. For HSPM, maternal alcohol consumption during pregnancy, low birth weight and fever during the first year of life are mentioned. 11 Other illnesses in early life and the use of antibiotics were proposed as risk factors for MIH. 12,13 The exact aetiology of dental enamel hypomineralization, however, remains unclear. 4,[11][12][13][14] In the search to unravel the aetiology of dental hypomineralization, a recent study of K€ uhnisch et al 15 showed that higher serum 25-hydroxyvitamin D (25 (OH)D) concentrations were correlated with less MIH and dental caries in 1048 German children at age ten. To our knowledge, this is the only study to have examined 25(OH)D and dental enamel hypomineralization. Several other studies examined vitamin D in relation to caries and generally observed that vitamin D supplementation in early life may be preventative for dental caries, as reviewed by Hujoel et al 16 The main function of vitamin D is to maintain plasma calcium concentrations at a constant level, which is important for healthy bone development and increasing evidence suggests also for healthy tooth development. 17,18 Vitamin D stimulates mineralization of dental enamel and bone by binding to receptors that are expressed in both dental cells and bone cells. 19,20 Because vitamin D is important in the mineralization of these tissues, it is noteworthy that we recently discovered that lower bone mass is associated with the presence of HSPM but not with MIH in 6-year-old children. 21 Our hypothesis is that this association could be explained by differences in 25(OH)D status between children, affecting mineralization of dental enamel and bone.
A limitation of the previous study of K€ uhnisch et al 15

| Study design and population
The analysis was embedded in the Generation R Study, a population-based, prospective cohort from foetal life onwards in Rotterdam, the Netherlands. 23 Figure 1).

| HSPM and MIH diagnoses
To visualize HSPM and MIH, an intraoral camera was used (Poscam USB intraoral [Digital Leader PointNix] or Sopro 717 [Acteon] autofocus camera, 640 9 480 pixels). The minimal scene illumination of both cameras was 3.0 lx (F1.4). During the data collection period, pictures of the teeth were taken by six trained nurses, twelve dental students and six PhD students. A paediatric dentist (ME) gave them a presentation about the how and why of taking the dental photographs and repeated that each half a year. Before the employees/students were allowed to make photographs themselves, they had to accompany an experienced employee/student for a day and learned how to make high-quality photographs. Afterwards, a paediatric dentist (ME) evaluated all photographs within 2 or 4 weeks. If she found the quality to be too low, she further instructed the respective employee/student on how to improve their quality or she trained them individually. Before taking the photographs, the children had to brush their teeth and excess saliva was removed with a cotton roll. Photographs were scored by a paediatric dentist (ME) on the presence of HSPM and MIH using the European Academy of Pediatric Dentistry (EAPD)criteria. 24 After completion of the data collection period, the same paediatric dentist (ME) re-evaluated the photographs of 649 children (10%) with a minimal time gap of 6 weeks. This resulted in a kappa for the intraobserver agreement of 0.82 for HSPM and 0.85 for MIH. 21 A second paediatric dentist (JV) re-evaluated the photographs of 648 children (10%). The kappa's for the interobserver agreement for ME and JV were 0.60 for HSPM and 0.69 for MIH. 10,21 JV evaluated the photographs only once. Hence, we were not able to calculate a kappa value for the intraobserver agreement of this examiner. ME and JV had a calibration session each 2 or 3 months. Before this session, ME randomly chose a couple photographs and discussed them together with JV. As photographs were taken at the age of six, not all children had their permanent first molars yet, resulting in a smaller number of children with data on MIH than HSPM. Children without data on MIH were on average younger (mean age 6.00 vs 6.41 years), were more often male (52.7% vs 44.7%) and more often had a Dutch or other Western background (68.3% vs 59.8%) than children with data on MIH (Table S1).

| 25(OH)D measurement
Maternal venous blood samples were collected during mid-preg- Samples were quantified using isotope dilution liquid chromatography/tandem mass spectrometry (LC-MS/MS). The method limit of quantification was 6 nmol/L and interassay imprecision was <11%. 25 Vitamin D status of children's blood samples was measured at the Endocrine Laboratory of the VU University Medical Center, Amsterdam, the Netherlands as described in detail previously. 26 Briefly, 25(OH)D was measured using isotope dilution online solid phase extraction LC-MS/MS, a similar method as used for the foetal sample. The limit of quantitation was 4.0 nmol/L; intra-assay coefficient of variation was <6%, and interassay coefficient of variation was <8% for concentrations between 25 and 180 nmol/L. 26 This method was perfectly aligned with the reference methods. 27 A cross-validation in 31 umbilical cord and pregnancy blood samples that were analysed in both laboratories showed an excellent correlation between both methods (r = .99). The Passing & Bablok   (Table S2).

| Covariates
Maternal age, educational level (low, mid-low, mid-high or high), parity, folic acid supplement use before/during pregnancy (start 1st 10 weeks, start periconceptional or never) and household income (<2000, 2000-3300, >3300 Euros/month) were assessed at enrolment in the study (ie, during pregnancy) using questionnaires. Maternal smoking and alcohol consumption during pregnancy were assessed in each trimester of pregnancy and categorized into never, until pregnancy was known, or continued. Information on child's birth weight was acquired from medical records and hospital registries. Low birth weight was defined as a birth weight below 2500 grams. Children's ethnicity was defined based on birth country of both parents 29  year follow-up, we re-assessed household income (<2000, 2000-3200 or >3200 Euros/month) and maternal educational level. 30 For all blood sample analyses, we kept a record of the month and season of the year in which blood was drawn.

| Statistical analyses
First, we constructed three binary logistic regression models in which having HSPMs at the age of six (yes/no) was defined as the outcome (dependent variable) and the foetal serum 25(OH)D concentrations were included as a predictor (independent variable). Foetal serum 25 (OH)D concentrations were included as both a categorical variable and as a continuous variable per 10 nmol/L. The categories were compared to an optimal serum concentration of ≥75 nmol/L (reference category). Model 1 adjusted only for the child's sex, gestational age at blood withdrawal, mother's age and BMI before pregnancy.
Model 2 additionally adjusted for variables that were associated with HSPMs in the Generation R Study population. 11 In model 3, we added variables that were associated with serum 25(OH)D concentrations in our study population. 26 We followed the same approach for studying the association between MIH (outcome) and foetal 25 (OH)D serum concentrations (predictor). Moreover, we made use of the same models to study the association between HSPM and MIH as outcomes and cord blood serum 25(OH)D concentrations as predictor. For the approach in which the child's serum 25(OH)D concentrations at the age of six was used as a predictor, minor modifications in the model were made as follows: Model 1 was adjusted for child's sex, age, weight and length, model 2 did not change, and model 3 was adjusted for household income and maternal educational level at the child's age of six instead of at enrolment, and child's watching television and playing outside were added because these factors have been shown to be important for children's vitamin D status. 27 To be able to compare results of foetal, birth and childhood 25 (OH)D, we repeated the analyses in a subgroup with data available on 25(OH)D at all three time points (n = 1840, Figure 1). We tested for statistical interaction between vitamin D status and children's age, sex and ethnicity separately in model 3. Multicollinearity was evaluated but was found not to be a problem in our models, because the tolerance statistic exceeded 0.20 for all variables. Moreover, we examined whether we could assume 25(OH)D levels to be linear to the logit using natural cubic splines (degrees of freedom = 3). Missing data of covariates were handled by applying multiple imputation (n = 10 imputations). 31 The pooled odds ratios (ORs) and 95% confidence intervals (95% CIs) were derived from pooling the results of the ten imputed datasets. Effect estimates were similar to the results of analyses of the original data, therefore, we only report pooled results after the imputation procedure. SPSS version 22.0 for Mac (IBM Corp, Armonk, NY, USA) was used for all analyses and a twosided P-value of <.05 was considered to be statistically significant.
The STROBE Guidelines were used to ensure adequate reporting of this observational study. 32

| RESULTS
Children in our sample had a mean (AESD) age of 6.2 (AE0.5) years at assessment ( Table 1)    We tried to minimalize the chance of having this bias using a reliable method, but some bias may still be present. 34   and dental enamel hypomineralization is the same for included and excluded participants. Another limitation of our study is that we did not consider different distribution patterns of HSPMs and MIH. This limited the possibility to associate vitamin D status with numerical data. Still, the number of children with HSPMs and MIH would have been the same, as the diagnosis was based on the index teeth as stated in the EAPD-criteria. 24 Furthermore, to prevent possible attrition bias, we implemented a multiple imputation method for missing data of covariates. Sensitivity analyses of the imputed data, however, did not result in significant differences in outcome compared to     15 because the development of teeth already starts in utero. 22 We were also able to examine 25(OH)D in a prenatal and early postnatal period. This was a major strength of our study. However, nei- Ideally, during the developmental period of teeth. Furthermore, it is important to keep on searching for different preventive possibilities and aetiological factors for dental hypomineralization in children, which still are unknown. 14 Moreover, despite null findings with hypomineralization, it would be interesting to study the association between 25 (OH)D status and dental caries in our population. The pathway in which vitamin D affects the risk of developing dental caries may involve pathways other than enamel mineralization. 40

ACKNOWLEDG EMENTS
First of all, we gratefully acknowledge the contribution of the participants, general practitioners, hospitals, midwives and pharmacies in

CONFLI CT OF INTEREST
The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. The authors declare no potential conflicts of interest with respect to the authorship and/or publication of this article.