Adherence to a healthy and potentially sustainable Nordic diet is associated with child development in The Norwegian Mother, Father and Child Cohort Study (MoBa)

Study population

This study was conducted within MoBa, which is a population-based pregnancy cohort study conducted by the Norwegian Institute of Public Health [25]. Pregnant women were recruited from all over Norway from 1999 to 2008, and 41% of those invited consented to participate. The cohort now includes 114.500 children, 95.200 mothers and 75.200 fathers. The study also includes data from the Medical Birth Registry Norway (MBRN), a national health registry containing information about all births in Norway [26]. The two datasets are linked by using the Norwegian National security number which is available in all Norwegian National health registries. The linkage is performed by the MoBa data team and the research file contains an anonymous serial number. The MoBa cohort study is an ongoing longitudinal health study where data still is collected from participants. The current study is based on MoBa version 12 of the quality-assured data files released for research in January 2019. Response rate for the questionnaires answered during pregnancy (Q1-Q3) was between 91 and 95% with decreasing participation rate over time. When the child was 3 years and 5 years old the response rate was at 59 and 54% respectively [25].

We included participants who had responded to the baseline questionnaire (Q1) around gestational week 17, covering general health and sociodemographic information, the food frequency questionnaire (FFQ) (Q2) answered around gestational week 22 and participants who were registered in the MBRN with singleton births. We excluded women with calculated energy intakes outside the range 4.5–20 MJ/day. The final study population consisted of a baseline 83,800 mother-child pairs (Fig. 1).

Fig. 1figure 1

Flow chart of inclusion of participants in the study

Ethics

The establishment of MoBa and initial data collection was based on a license from the Norwegian Data Protection Agency and approval from The Regional Committees for Medical and Health Research Ethics. All MoBa participants provided written informed consent before enrolment into the study. The MoBa cohort is now based on regulations related to the Norwegian Health Registry Act. The current study was approved by The Regional Committees for Medical and Health Research Ethics (2019/339).

Main exposure

NND scores for maternal pregnancy diet and child diet at 6 and 18 months and 3 years have been developed and are described in detail in previously published papers [18, 24, 27]. A brief summary from these papers is presented here.

The child NND scores were developed under the rationale of being as similar as possible to a previously developed maternal NND score in MoBa [18]. Despite referring to the child diet scores as ‘child NND scores’ in this paper for simplicity reasons, it should be noted that they may not reflect the NND to the same extent as the maternal NND score. The NND scores and their corresponding subscales are presented in Table 1.

Table 1 Description of the New Nordic Diet scores

For each child score, dietary variables from the child questionnaires were selected to construct subscales based on the maternal NND subscales score. The questions assessing child diet were far less extensive and specific compared to the maternal FFQ, hence, the included dietary components differ to some degree from the maternal score and additionally differ between age-specific scores.

In the child dietary assessment, the mothers were mainly asked to respond to “How often does your child usually eat/drink the following” with response alternatives varying slightly between questionnaires. All responses were subsequently recoded to reflect a weekly consumption. We defined missing as having incomplete data on all food items that were included in the construction of each child diet score. If information was missing for some food items only, an assumption of null intake was made in accordance with recommendations by Cade et al. [28]. These items were recoded to 0 (never/seldom) to avoid losing all dietary information for respondents with incomplete data for a given item. For the maternal score, all missing items were null-imputed.

Further, the included food or drink items were added together in the respective subscales, either to yield a subscale measuring frequency of weekly consumption or to generate a relative measure of consumption of one food group compared to another. The subscales were mostly dichotomized by the median and coded to give either 0 or 1 point, where receiving 1 point acknowledged a healthier food choice or a frequency of consumption above the median of a healthy food item. Some subscales were scored according to responding “yes” or “no” to a question, where “yes” indicated the favourable health behaviour.

The sum of the subscales was further computed into continuous age-specific NND scores. Finally, each score was divided into low, medium, and high adherence groups with the intention to create as equally sized groups as possible. Where this was not possible, cut-offs were chosen to yield the low and high adherence groups as equally proportioned as possible. The rationale for categorizing participants into low, medium, and high adherence was to be able to compare high vs. lower adherence to the described dietary pattern. This was further used to explore associations between maternal and child diet and developmental outcomes in participating children.

Child developmental outcomes

Child communication skills and motor development at 6 and 18 months, 3 and 5 years, respectively, were assessed with short forms of Ages and Stages Questionnaires (ASQ) and Child Development Inventory (CDI) [29]. ASQ is a parent-completed questionnaire tool that is used to identify potential developmental delay compared to age-peers, in need for further assessment. It is a widely used developmental screening-tool validated for use in Norwegian populations [30, 31] and has been applied in previous MoBa studies of prenatal exposure through maternal diet and child development [32, 33]. In MoBa, the ASQ at 5 years only covers language development. At this age, we used the CDI, completed by the parents in the 5-year questionnaire to assess motor skills and determine the child’s developmental level based on skill-assessment at given ages throughout the first 2 years in life and upwards to six and half years [34].

Outcome dimensions were defined according to the respective instrument manuals [29, 30, 35]. The outcome data were calculated as sum scores as the basis for the analysis, with a lower score indicating fewer milestones achieved by the child at the time of measure. We used simple imputation for participants with less than 50% of items missing on total scores. The missing items were recoded to mean of total score. Participants missing more than 50% of items in a score were excluded at that timepoint of measure. Sensitivity analysis between excluding missing values and imputed values were conducted and there was no significant change in results.

The items used to measure developmental skills changed across age as the child grew older, thus, the measures are not directly longitudinally comparable. The range of the outcome measures across the five timepoints of data collections also differ.

Covariates

Covariates considered for inclusion in the models were baseline variables from questionnaires answered during pregnancy and at birth regarding maternal health and socioeconomic status identified as adjustment factors in previous studies investigating the relationship between diet and child development [32, 36]. The covariates included were parity, maternal age at delivery in four categories, maternal education, maternal pre-pregnancy body mass index (BMI) and marital status. Maternal symptoms of depression were additionally included as a covariate and was measured by a five-item short version of the Hopkins Symptom Checklist, psychometrically derived from the 25-item version [37]. Also included were child sex, gestational age (included in analysis with child NND scores), and age of the child when the questionnaire was answered.

Statistical analysis

Linear regression and logistic regression analysis with robust standard errors were employed to compute crude and adjusted estimates of associations of maternal and age-specific child NND scores with measures of child development from 6 months to 5 years. Both methods were applied to establish a potential positive linear association between the NND scores and child development scores, and to examine whether scoring in the low NND score-categories was associated with developmental delay as expressed by scoring 2 standard deviations (SD) below or lower than the mean developmental score. The given cut-off allows for identification of the lowest scoring individuals within the sample, although the cut-off is not clinically validated for assessment of a specific developmental delay. The distribution of child development scores was highly left skewed with more than 90% of the children scoring within the 90th percentile range at each timepoint. After carefully considering the consequences of comparing exposure and outcome measurements which both had a different number of items and showed different range and SD at the different timepoints, we concluded that longitudinal analysis methods would not be suitable for this study. Instead, we performed cross-sectional analyses on the relationship between exposure and outcome at each timepoint and investigated potential trends and patterns in the results.

In the linear regression model, the percentage of maximum scoring for the developmental outcome and NND scores were computed to give more comparable results across timepoints. For the logistic regression models, the child development scores were dichotomized with 2SD below the mean as cut-off. Developmental score values at -2SD of the mean score or lower were assigned the value 1 (poor outcome), and the rest given the value 0. For these analyses, the values of the NND score categories (high/middle/low) were reversed with the low NND adherence group being assigned the value 2 and the high adherence was given the value 0 (reference group). More than 10% of the mothers participated more than once in the study and to correct for a possible impact of sibling covariance, we used robust cluster variance estimation in all analyses. The statistical programs, SPSS version 22 (SPSS, Inc., Chicago, IL, USA) and STATA/SE 16.1 were used for the analyses.

留言 (0)

沒有登入
gif