1 INTRODUCTION

Childhood obesity is considered a major global health challenge1 associated with long-term morbidity and mortality.2 Most prevention and intervention strategies targeting childhood obesity focus on altering the obesogenic lifestyle by increasing physical activities or promoting a healthy diet, or a combination of the two.3, 4 As these interventions show only limited effectiveness in randomized controlled trials,3 and even less when implemented as scale-up interventions under real world conditions,5 there remains a need to identify further modifiable risk factors for increased body mass index (BMI) in children, which could inform effective preventive and interventive measures in the future. While various biological, environmental and sociopsychological factors contribute to the development and maintenance of overweight and obesity in children,2 several studies also indicate the family to be a factor of central influence.6 Thus, it is unsurprising that previous research has not only suggested links of children's risk for overweight or obesity with specific parenting behaviours related to food consumption or physical activity,7, 8 but also has indicated links with the observed general quality of the early mother-child interaction, specifically maternal sensitivity.9 However, previous studies have not disentangled the effect of maternal sensitivity from the potential effect of maternal intrusiveness, as these studies have often used instruments capturing aspects of both constructs in one measure. Therefore, this study aims to replicate and extend previous research by exploring both the role of maternal sensitivity and intrusiveness in mother-child interaction on children's weight in a sample from Germany.

Ever since the introduction of the maternal sensitivity construct to the field, developmental research has stressed the relevance of sensitive caregiving for a broad range of child outcomes.10 Maternal sensitivity refers to the mother's capacity to recognize and correctly interpret the infant's cues and to respond appropriately, consistently and promptly.11 When facing arousal, infants highly depend on their caregivers' sensitive behaviour in order to stay well-regulated. According to attachment theory, infants internalize these early caregiver relationship experiences. Consequently, these experiences serve as prototypes for regulatory strategies.12 Empirical studies show that caregiver sensitivity positively affects self-regulatory abilities in children, such as executive functioning, emotion regulation or inhibitory control.13-15 Moreover, such sensitive interactions support the child in developing the ability to correctly perceive inner affective states or signals and to differentiate them from hunger and satiation.16 There is evidence that impairments in both self-regulation and the perception or understanding of inner (affective) states are associated with obesity17-21 and that individuals with overweight compared to individuals with normal weight show higher levels of “emotional eating”, i.e., eating due to distress or uncomfortable affective states in the absence of hunger.22 It is not completely understood whether such impairments in individuals with obesity are rooted in non-sensitive experiences with their caregivers during early childhood. However, Wendland et al23 showed that lower maternal sensitivity towards a child at the age of 6 months predicted increased BMI standard deviation scores (BMI-SDS) in girls—but not boys—at the age of 48 months. Likewise, low maternal sensitivity compared to high sensitivity towards a child at the age of 6 and 54 months, respectively, predicted a higher risk of the child having overweight or obesity at school age.24, 25 In a large cohort study, Anderson et al9 examined the effect of maternal sensitivity in combination with child attachment security at the ages of 15, 24 and 36 months on the development of obesity in adolescence. Low relationship quality (combined score of attachment and sensitivity) of the mother-child interaction compared to high relationship quality was associated with a greater risk of developing obesity at the age of 12 to 15 years. Interestingly, lower maternal sensitivity was associated with adolescent obesity to a greater extent than an insecure attachment of the child. In contrast, in another cohort study, a significant link between the mother-infant relationship (including maternal sensitivity, responsiveness, and fostering of cognitive and socio-emotional growth) at 9 months and obesity at 5.5 years was no longer observed when maternal sociodemographic factors such as maternal education were taken into account.26 One study even suggested more positive mother-child interactions in 2-to 3-years-old children with obesity compared to children with normal weight.27

While the majority of these studies seem to emphasize the importance of sensitive behaviour towards the child during early childhood for healthy weight development, the role of intrusive behaviour remains unclear. Maternal intrusiveness refers to behaviour characterized by overprotection or over-directiveness, and to behaviour interfering with or dominating the child's activities and restricting the child's autonomy.28, 29 It has been associated with a variety of negative child outcomes, such as internalizing and externalizing problems, poor academic performance and deficits in socioemotional development.30, 31 While maternal sensitivity in interaction with the child promotes the development of children's self-regulation, maternal intrusiveness appears to compromise self-regulatory capacities.32-34 However, little is known about the effect of maternal intrusiveness on the development of overweight and obesity or increased BMI-SDS in children. Evidence from research focusing on feeding situations found observed intrusive maternal feeding behaviours to be positively associated with children's BMI-SDS.35 In addition, a closer look at the studies that reported an effect of maternal sensitivity on children's weight revealed that a majority of these studies9, 24, 25 used instruments to assess sensitivity which—besides others—did include aspects of (non-)intrusiveness (e.g., “respect for autonomy”, “intrusiveness”). Hence, it is not clear whether the effect of the quality of the mother-child interaction found in these studies is attributable to sensitive behaviour or intrusive behaviour on the part of the mother. As positive and negative parenting behaviours exert unique effects on child outcomes,36, 37 gaining more knowledge about the specific associations of maternal sensitivity and intrusiveness with children's increased BMI-SDS could be helpful for tailoring prevention or intervention programmes.

Taking all this into consideration, this study aims to help fill the research gaps outlined above by analysing the effect of specific aspects of the mother-child interaction during early childhood on children's BMI-SDS at school age. First, we want to examine the effect of maternal sensitivity on children's prospective weight. In line with previous research,9, 23 we hypothesize that lower maternal sensitivity in early childhood predicts higher BMI-SDS in children at school age. Second, we want to expand the current research on the quality of mother-child interaction and its influence on children's weight development by focusing on maternal intrusiveness as a potential risk factor. We hypothesize that higher maternal intrusiveness (i.e., lower non-intrusiveness) in the mother-child interaction predicts higher BMI-SDS in children at school age. Since high maternal BMI38 and low maternal education in higher economic status countries,39 as well as high children's weight at birth,40 are considered to be determinants of increased BMI-SDS in children, we take these aspects into account as potentially confounding covariates. This specifically extends previous research that did not include maternal BMI/weight status in the analyses,24, 25 or only collected data on maternal obesity when youth were adolescent rather than in early childhood.9 In addition, we control for demographic characteristics of the child (age, gender) and explore whether child age and child gender moderate the effects of maternal sensitivity and maternal intrusiveness on child BMI-SDS at school age.

2 METHODS 2.1 Sample

The current study was conducted at the IFB Adiposity Diseases Centre, University of Leipzig, investigating risk and protective factors for children's weight development in a risk group (mother and/or father with obesity, i.e., BMI ≥ 30) and a control group (both parents with normal weight).41 Families were recruited using flyers and posters in kindergartens and at the practices of health care professionals in and around Leipzig, Germany,42 and screened by telephone for parental weight status. Only families who met the inclusion criteria regarding parental weight status were invited to participate in the study. A total sample of N = 209 young children aged 5–47 months (M = 24.87, SD = 11.35) and their parents were recruited at the first point of assessment (t1) and were assessed repeatedly over a longer period of time. This study uses data from the first point of assessment (t1) and the latest assessment at school age (t2), which took place about 6 years (M = 6.00, SD = 0.50, range 4.92–7.17 years) after t1. At t2, 77 families had dropped out of the study (typically due to moving away, lack of time, or no reaction to our contact attempts), but n = 132 children aged 5 to 10 years (M = 7.58, SD = 1.08) and their mothers participated again. This subsample did not differ from the whole sample in terms of maternal education, BMI or children's age or gender and BMI-SDS at birth. For n = 129 children out of these 132 children, anthropometric data were obtained either by assessment in the laboratory (n = 95) or via parental reports (n = 34). Following the recommendation by Pinquart43 regarding the investigation of associations between parenting and continuous weight data, we removed underweight children from our data, yielding a final sample of n = 116. Table 1 shows the descriptive characteristics of the sample. In line with the inclusion criteria mentioned above, we oversampled for mothers with obesity at t1. Hence, our sample included 45 (38.8%) mothers with obesity (BMI ≥ 30) and 10 (8.6%) mothers with overweight (25 ≤ BMI < 30).*

TABLE 1. Descriptive characteristics of the sample (N = 116) t1 t2 Variables M (SD) or N (%) M (SD) or N (%) Mothers' characteristics Age in years, M (SD) 31.48 (4.59) 37.49 (4.69) BMI, M (SD) 28.27 (8.59) - Education Certificate of general or secondary educationa 47 (40.5) - General qualification for university entranceb 36 (31.0) - University degree 33 (28.4) - Children's characteristics Age in months, M (SD) 24.00 (11.36) 96.15 (12.69) Gender Male 60 (51.7) 60 (51.7) Female 56 (48.3) 56 (48.3) BMI-SDS At birth 0.50 (1.34) 0.37 (0.90) BMI At birth 13.22 (1.69) 17.22 (2.55) Weight status Obesity 7 (6.0%) Overweight 14 (12.1%) Normal weight 95 (81.9%) Note: t1 is the baseline (average child age = 24 months) and t2 is the school age (average child age = 96 months). a The certificate of general education is an elementary school diploma, which is obtained on successful graduation from grade 9; the certificate of secondary education is obtained on successful graduation from grade 10. b This group also includes mothers with the entrance qualification for a university of applied sciences. 2.2 Procedures

The ethics committee of the Medical Faculty, University of Leipzig approved this study. Caregivers gave their informed consent and mothers completed sociodemographic information forms at t1. We also assessed the anthropometric data of mothers and videotaped mother-child interactions in the laboratory at t1 during a free-play situation. Data on children's birth weight and birth height were taken from official booklets that included routine medical check-ups for children (“U-Heft”). At t2, we obtained data on children's height and weight.

2.3 Measures 2.3.1 Anthropometric data

Trained personnel assessed the body weight and height of mothers at t1 using calibrated body scales (Kern, model MPT 300K100M) with a reading accuracy of ±0.1 kg. Mothers stood freely, without shoes. Maternal height was measured in a free-standing position with a calibrated stadiometer (Soehnle) with a reading accuracy of ±0.5 cm. None of the mothers reported being pregnant at the point of data collection. We then calculated maternal BMI at t1. We also collected information on children's birth weight and length from an official booklet issued by the Ministry of Health which mothers brought to the laboratory. This booklet (“U-Heft”) includes data on physical development collected by medical staff during routine medical check-ups for children throughout childhood and is for the parents to take home for their records. Immediately after birth, the baby receives a first examination by medical staff (doctors or midwives) during which birth weight and length are assessed and documented accordingly in the official booklet. Based on the anthropometric data from the routine medical check-up at birth we also calculated children's BMI-SDS at birth. BMI-SDS (or BMI z scores) are standardized for age and gender with respect to a reference standard44 and indicate the extent to which the BMI of an individual lies above or below the median BMI value, considering the individual's age and gender. According to the German reference standard including children aged 0–18 years, a BMI-SDS of 1.28 and above indicates overweight and a BMI-SDS of 1.88 and above indicates obesity. At t2 we obtained parental reports regarding current weight and height for all children. Ninety of 116 children were also assessed by trained personnel using the same body scales and stadiometer as used for mothers at t1. Whenever anthropometric data collected in the laboratory were available, we used these for the calculation of BMI-SDS at t2. When these data were not available—which was the case for 26 children—we calculated BMI-SDS based on parental reports of children's weight and height.† As the majority of children in this study did not show overweight or obesity at school age, we used child BMI-SDS as a continuous outcome measure (see statistical analyses).

2.3.2 Maternal sensitivity and intrusiveness in mother-child interactions

We used the dimensions sensitivity and non-intrusiveness of the emotional availability scales (EAS; 4th edition28) to assess maternal sensitivity and (non-)intrusiveness at t1 based on videotaped mother-child interactions in a 16-minute free-play situation using standard and age-appropriate toys provided in the laboratory. Sensitivity refers to the extent to which the caregiver's affect is well regulated and authentic, and to which the caregiver is able to perceive and interpret children's cues correctly and to react to them appropriately, including adequate timing and handling of conflicts, flexibility of behaviour, creativity, and an accepting attitude towards the child. Non-intrusiveness refers to the absence of overly protective, suggestive, stimulating or controlling behaviour in the caregiver as well as the ability to follow the child's lead without interfering in the child's activities or impairing the child's autonomy. Both dimensions were rated on 7-point scales (1 = non-optimal, 7 = optimal) by two female coders (the senior author and another researcher) who were blind to further information about the families. Before coding the study data, the two coders had been successfully trained and accredited to use the EAS by Biringen. For 16% of the study videos inter-rater reliability between the two coders was assessed. The ICCs were ICC = 0.75 for sensitivity and ICC = 0.77 for non-intrusiveness, respectively.

2.4 Statistical analyses

We used SPSS statistical software, version 25.0 (SPSS Inc.) for our analyses. We used correlation analyses (Pearson's r) to investigate associations between aspects of mother-child interaction (i.e., sensitivity and non-intrusiveness), child BMI-SDS at t2 and control variables. To test our hypotheses that sensitivity and non-intrusiveness would negatively predict child BMI-SDS 6 years later, we conducted separate linear regression analyses with child BMI-SDS at t2 as the dependent variable (as we expected multicollinearity between sensitivity and non-intrusiveness). The predictors were maternal sensitivity and non-intrusiveness. In the second step, we also included maternal BMI, and maternal education at t1 as well as child age at t1 and child gender as covariates. In the third step, we added interaction terms sensitivity × child age as well as sensitivity × child gender and non-intrusiveness × child age as well as non-intrusiveness × child gender, respectively. In the fourth step, we included child BMI-SDS at birth as an additional covariate reflecting a proxy for genetic, biological or intrauterine influences. Whenever our hypothesized effects reached significance, we also calculated effect sizes (Cohen's f2) for these effects.45

2.5 Results 2.5.1 Descriptive analyses

The mean scale scores for sensitivity (M = 5.01, SD = 0.87) and non-intrusiveness (M = 5.00, SD = 0.98) indicate that on average the quality of the mother-child interactions in this sample tended to be moderate. Both scales covered a range from 3 to 7, suggesting that there were no mothers in this sample who showed extremely insensitive behaviour or immensely (physically) intrusive behaviours as these mothers would have received even lower scores (1 or 2).

Intercorrelations of sensitivity, non-intrusiveness and child BMI-SDS at t2 as well as child BMI-SDS at birth, maternal BMI, maternal education as well as child age and child gender are presented in Table 2. There was a significant strong positive cross-sectional correlation between maternal sensitivity and non-intrusiveness in early childhood (t1), a small positive correlation between maternal sensitivity and maternal education, and a significant small longitudinal negative correlation between maternal non-intrusiveness in early childhood (t1) and child BMI-SDS at school age (t2). The greater the mother's intrusiveness in interactions with her child was in early childhood, the greater the BMI-SDS of her child was at school age. Moreover, child BMI-SDS at t2 showed a small positive correlation with child BMI-SDS at birth, a moderate positive correlation with maternal BMI, a small positive correlation with child gender and a small negative correlation with maternal education. In addition, child gender was moderately negatively associated with maternal education.

TABLE 2. Correlation analyses (Pearson's r) between aspects of mother-child interaction, demographic variables and child BMI-SDS at school age and at birth 1 2 3 4 5 6 7 8 1 Sensitivity at t1 1 2 Non-intrusiveness at t1 0.75** 1 3 Child BMI-SDS at birth 0.13 0.02 1 4 Maternal BMI at t1 −0.04 −0.07 0.09 1 5 Maternal education at t1 0.20* 0.08 −0.10 −0.27** 1 6 Child age at t1 0.04 0.01 0.03 −0.06 −0.05 1 7 Child gender 0.02 0.01 0.16 0.12 −0.30** 0.07 1 8 Child BMI-SDS at t2 −0.12 −0.23* 0.26** 0.44** −0.27** 0.18† 0.21* 1 Note: Child gender (1 = male, 2 = female). t1 is the baseline (average child age = 24 months) and t2 is the school age (average child age = 96 months). * p < 0.05. ** p < 0.01. † p < 0.06. 2.5.2 Prediction of children's BMI-SDS at t2

By applying a regression analysis with child BMI-SDS at t2 as the dependent variable and maternal sensitivity as predictor, we found that maternal sensitivity did not predict children's BMI-SDS at t2—with or without the covariates (see Table 3). Of the covariates, maternal BMI as well as child age positively predicted children's BMI-SDS at t2, while there were no significant effects of maternal education and child gender (see Table 3, model 2). All variables explained 27% of the variance. When exploring whether child age or gender moderated the effect of maternal sensitivity, we only found a small significant effect for the interaction sensitivity × child gender (f2 = 0.04) with all variables explaining 30% of the variance (see Table 3, model 3). Figure 1 illustrates that there was a negative association between maternal sensitivity and child BMI-SDS at t2 in girls but not in boys. Multiple linear regressions conducted separately for girls and boys revealed a small negative effect of maternal sensitivity on child BMI-SDS in girls (β = −0.31, p = 0.022, f2 = 0.10) but not in boys (β = 0.10, p = 0.438). Hence, the lower the mother's sensitivity in the interaction with her young daughter, the higher her daughter's BMI-SDS was at school age. The effect of maternal sensitivity in girls remained significant after covariates were added (β = −0.21, p = 0.048, f2 = 0.08).

TABLE 3. Regression analyses summary for maternal sensitivity, maternal BMI, and demographic variables predicting children's BMI-SDS at school age B SE β 1 (Constant) 0.37 0.08 Sensitivity at t1 −0.12 0.10 −0.12 2 (Constant) 0.37 0.07 Sensitivity at t1 −0.09 0.09 −0.09 Maternal BMI at t1 0.04 0.01 0.40*** Maternal education at t1 −0.10 0.10 −0.09 Child age at t1 0.02 0.01 0.19* Child gender 0.21 0.15 0.12 3 (Constant) 0.38 0.07 Sensitivity at t1 −0.08 0.09 −0.08 Maternal BMI at t1 0.04 0.01 0.39*** Maternal education at t1 −0.10 0.10 −0.09 Child age at t1 0.02 0.01 0.19* Child gender 0.21 0.15 0.12 Sensitivity × child age −0.00 0.01 −0.01 Sensitivity × child gender −0.35 0.17 −0.17* 4 (Constant) 0.37 0.07 Sensitivity at t1 −0.11 0.09 −0.11 Maternal BMI at t1 0.04 0.01 0.38*** Maternal education at t1 −0.08 0.10 −0.08 Child age at t1 0.02 0.01 0.19* Child gender 0.17 0.15 0.10 Sensitivity × child age 0.00 0.01 0.00 Sensitivity × child gender −0.29 0.17