Longitudinal associations between early child weight gain, parent feeding, child self-regulation, and later child body mass index

Summary

Background:

Few longitudinal studies have examined associations of child weight trajectories, maternal demandingness and responsiveness during feeding, and child self-regulation.

Objective:

We examined if child weight-for-length trajectories from 6 weeks to 2 years of age were associated with maternal demandingness and responsiveness at child age 3 years old, and if maternal feeding dimensions predicted child BMI trajectories from 4.5 to 7.5 years among Mexican American children from low-income families. Child self-regulation was evaluated as a potential mechanism linking maternal feeding with child BMI.

Method:

Child (N = 322) weight and length/height were assessed at 10 timepoints from 6 weeks through 7.5 years. Mothers completed the Caregiver Feeding Style Questionnaire when the child was 3 years of age.

Results:

A steeper slope of weight-for-length z scores from 6 weeks to 2 years (indicating more rapid weight gain) was associated with less maternal demands during feeding at 3 years. More maternal demandingness at child age 3 years predicted lower child BMI at 4.5 years, but not trajectories from 4.5 to 7 years. Child self-regulation was not associated with child BMI from 4.5 to 7.5 years.

Conclusion:

The findings highlight how the relationship between mothers and children during feeding can be bidirectional and potentially influenced by the developmental stage.

1 |. INTRODUCTION

Low-income Hispanic children have higher rates of childhood obesity than low-income non-Hispanic children.¹ Mexican Americans, who comprise approximately 80% of the Hispanic population in the western United States, are at higher risk of obesity than Hispanic Americans from other countries of origin.² Parents can influence child food consumption and weight in diverse ways. Feeding styles capture a parent’s general approach to feeding across situations and domains, providing the socio-emotional context in which parents socialize the child towards food and eating.^3–5 A significant portion of research on feeding styles uses the Caregiver Feeding Style Questionnaire (CFSQ).⁵ The CFSQ can be used to classify parents into four feeding style categories (authoritarian, authoritative, indulgent, and uninvolved) based on median splits on responsiveness and demandingness dimensions.^5,6 Responsiveness refers to how sensitive parents are to the child’s needs during eating.^5,6 Demandingness refers to the demands parents place on children and how much parents encourage children to eat.^5,6

While a number of cross-sectional studies have found relationships between parental feeding dimensions and child weight,^7–10 few have looked at this association longitudinally.^11–13 Parental feeding dimensions have been shown to concurrently predict child weight⁷; however, few longitudinal studies exist. Several cross-sectional studies document a positive association between a combination of low demandingness and high responsiveness and child weight across ethnic minority⁸ and low-income families,^9,10 accounting for 26% of the variance in child body mass index (BMI), and suggesting demandingness and responsiveness are strong predictors of child BMI.⁸ This relation remains when controlling for demographic characteristics, child temperament, and parental affect.⁸ Limited longitudinal work offers preliminary evidence that demandingness and responsiveness are stable over a 2-year period¹¹ and prospectively predict child BMI across 1.5 year,¹² 2 year,¹¹ and 3 year¹³ time periods. However, not all studies have found this association,^14,15 while others have found responsiveness but not demandingness associated with child BMI.¹⁶ Collectively, research suggests a combination of low demandingness and high responsiveness feeding dimensions is associated with child BMI, however, the existing longitudinal studies assess the impact of feeding on child BMI at baseline and then at a later timepoint,,^11,12,13,17 and cannot address associations of early life weight gain trajectories with later parent feeding dimensions, and in turn, child weight gain trajectories across later years.

Child self-regulatory ability is one potential mechanism commonly theorized to play an important role in child weight gain.^13,17 General self-regulation ability in childhood encompasses varied attentional, emotional, biological, cognitive, and behavioural processes that support goal-directed action.¹⁸ Research has tended to focus on impulsivity, effortful control, and the ability to delay gratification as characteristics of self-regulation that may directly impact child eating behaviour and weight management.¹⁹ Poor child self-regulation predicts higher BMI, obesity, and more rapid weight gain across childhood.^20,21 For children at high risk for obesity, poor child self-regulation has been shown to mediate rapid increases in weight gain.²⁰

The current longitudinal study of low-income, Mexican American mothers and children addressed limitations in prior research by prospectively examining whether a child’s weight-for-length trajectory from 6 weeks to 2 years of age significantly predicted maternal demandingness and responsiveness at 3 years (see Figure 1). Further, we examined if demandingness and responsiveness at 3 years predict growth in child BMI from 4.5 to 7.5 years. Self-regulation was evaluated as a potential mechanism linking maternal feeding dimensions with later child BMI. Specifically, we hypothesized that a more positive slope of weight-for-length trajectory through age 2 years would predict higher demandingness and lower responsiveness. Additionally, the present study hypothesized that: (1) higher demandingness and lower responsiveness would predict poorer child self-regulation; and (2) poor child self-regulation would predict higher child BMI at 4.5 years and an increasing slope of BMI from 4.5 to 7.5 years.

FIGURE 1.

Full model with all tested paths. Note: Solid lines represent statistically significant paths shown at p < 0.05. Dashed lines represent non-significant tested paths. Standardized estimates shown for significant paths. Error terms not shown. Latent variable sz-score percentile; ‘BMI’ = body mass index; ‘self-reg’ = self-regulation; ‘wks.’ = weeks; ‘yrs.’ = years. Model fit: χ2(N = 322; df = 120) = 256.098, p < 0.0001; RMSEA= 0.059 (95% CI: [0.049, 0.069]); CFI = 0.834; SRMR = 0.070. The complete model included an interaction term between responsiveness and demandingness at 3 years, however, it was not significantly related to any of the other variables and was removed for a more parsimonious model

2 |. METHOD

2.1 |. Participants

A total of 322 Mexican-origin women were included in the present longitudinal study. Women were recruited from prenatal clinics that serve low-income or uninsured families. Eligibility criteria included (1) low-income status (e.g., family income <$25,000), (2) age 18 or older, (3) Spanish or English fluency, (4) self-identification as Mexican or Mexican American, (5) singleton delivery, and (6) no prenatal evidence of an infant health or developmental problem. Additional demographic data are shown in Table 1. Women received financial compensation for each visit ($75 for the prenatal visit, $50 for each postpartum home visit, $100 for each lab visit). Travel or reimbursement for travel costs were provided for lab visits.

TABLE 1.

Descriptive statistics for all study variables

Variable name	Min	Max	Mean	SD	%	n
Mother’s age	18	42	27.79	6.48		322
Mother’s education	0	18	10.14	3.21		322
Mother’s country of birth
United States					14	44
Mexico					86	278
Child’s gender
Male					45.7	145
Female					54.3	172
Marital status
Married					30	96
Living with partner					45	147
Never married					15	49
Number of other biological children	0	9	1.98	1.68		320
Estimated total income
≤$5,000					13.7	44
$5,001–10,000					18.9	61
$10,001–15,000					27.0	87
$15,001–20,000					11.5	37
$20,001–25,000					12.4	40
≥$25,001					13.9	45

Note: N = 322; Living with partner are mothers who are not married but living with a partner.

2.2 |. Design and procedure

The current analyses include data from 5 home visits (prenatal, 6 weeks, 12 weeks, 18 weeks, 24 weeks) and 7 lab visits (1 year and 1.5, 2, 3, 4.5, 6, and 7.5 years). Informed consent for participation was obtained at the prenatal home visit (26–38 weeks gestation, M = 35.4, SD = 2.8). Consent forms were read aloud to address variations in literacy. Survey questions were read aloud in the mother’s choice of Spanish (82%) or English (18%). Child length/height and weight were assessed at each time point. At the 4.5-year lab visit, mother-child dyads participated in three interaction tasks, which were video-recorded for later coding. All aspects of the study complied with ethical standards and were approved by the university’s Institutional Review Board.

2.3 |. Attrition

The study followed a ‘planned missingness’ design²² for the postpartum home visits to minimize participant burden. All participants were assigned to complete the 6-week visit, but dyads were randomly assigned to complete two of the three remaining home visits. Of the 322 women who consented and participated in the prenatal visit, 312 mother-infant dyads (97%) completed the 6 weeks visit. Of those randomly assigned, 99% (203 dyads) attended the 12-week visit, 96% (209 dyads) attended the 18-week visit, and 93% (209) attended the 24-week visit. For later time points, participation rates were 82.2% (265 dyads) at 1-year, 73.6% (237 dyads) at 1.5-years, 75.4% (243 dyads) at 2-years, 66.7% (215 dyads) at 3-years, 71.4% (230 dyads) at 4.5-years, 65.6% (212 dyads) at 6-years, and 74.5% (240 dyads) at 7.5-years.

2.4 |. Measures

2.4.1 |. Infant weight-for-length and child BMI

Weight, infant length, and child height measurements were assessed once by trained staff at each time point. Length was measured with the infant laying prone, while height was assessed with the child standing against the wall. Weight was measured with portable digital scales. Weight-for-length percentile z-scores (WLZ%) at 6, 12, 18, and 24 weeks and 1, 1.5, and 2 years of age were calculated using World Health Organization (WHO) software (available at http://www.who.int/childgrowth/software/en/). These z-scores were used to calculate growth parameters. BMI-for-age percentile z-scores (BMIz%) were calculated using the same program for the 4.5-, 6-, and 7.5-year time points. The program flags and sets to missing any WLZ% and BMIz% scores considered ‘biologically implausible’, based on a z-score <−5 or >5.0.

2.4.2 |. Maternal feeding demandingness and responsiveness during childhood

Mothers completed the Caregiver’s Feeding Styles Questionnaire (CFSQ) at the 3-year visit, which was developed to measure parental feeding dimensions in two low-income ethnic groups.^5,23 It contained 12 parent-centred feeding directives and 7 child-centred feeding directives with items scored on a 5-point Likert scale ranging from 1 = never to 5 = always. The current analyses use the responsiveness and demandingness feeding dimensions. Although a growing body of literature has examined the categorical approach to feeding styles as it relates to childhood obesity,⁷ this scoring process is less than ideal. Categorizing continuous variables by median splits can yield arbitrary categorization of individuals with a lack of consistency across studies.²⁴ Further, individuals slightly above and below the cut point are treated as very different rather than very similar.²⁴ Importantly, categorization of data can conceal non-linear relations between variables, and seriously underestimate the extent of variation in outcomes between the groups.²⁴ For these reasons, the current study analyses responsiveness and demandingness from the CFSQ as continuous variables.⁵ This approach has been used in previous studies.^14–16,25

The demandingness dimension constitutes both the child-centred and parent-centred feeding directives to assess the degree to which parents make demands or promote child eating (α =0.90). The responsiveness dimension assesses sensitivity to the child’s needs (e.g., reasoning with the child) while considering overall feeding directives (α = 0.79). Each dimension has shown sound psychometric properties, validity, and cultural invariance across low-income Hispanic and African–American families.²⁵ This factor structure was replicated in a second Latino sample.¹¹

2.4.3 |. Child self-regulation

Child behaviour questionnaire.

At the 4.5-year visit, mothers completed the 94-item Child Behaviour Questionnaire (CBQ).²⁶ Respondents indicate how representative each item is for their child using 1 = extremely untrue to 7 = extremely true. The current study replicated the hierarchical second-order factor structure of effortful control validated by Backer-Grøndahl and colleagues.²⁷ Using confirmatory factor analysis in the structural equation framework, four first-order factors representing attention focusing (3 items), inhibitory control (3 items), low-intensity pleasure (3 items), and perceptual sensitivity (3 items) were specified and all four loaded onto the effortful control second-order factor. The factor score was saved and used as a continuous variable in all subsequent analyses, with higher scores indicating greater effortful control.

Observed child dysregulation.

At the 4.5-year visit, mothers and children participated in three interaction tasks for five minutes each: free play, during which mothers play with their child as they normally would if they were at home; clean up, during which mothers have their child clean up the toys; and teaching task, wherein mothers teach their child an activity slightly above the child’s developmental ability. Interaction tasks were video-recorded and children’s behaviour was rated based on the extent to which they exhibited signs of globally dysregulated affect or behaviour.²⁸ Global ratings were indexed by levels of appropriateness, lability, intensity, duration, frequency, and recovery time that children displayed during the individual episodes. Scores ranged from 1 to 5, with 5 signifying a very high degree of emotional, behavioural, and attentional dysregulation. Undergraduate research assistants were highly trained with weekly reliability meetings held between undergraduate coders and graduate student master coder to prevent observer drift. Reliability was done on 13% of the videos (ICC = 0.82). An average dysregulation score was calculated for the child across the three tasks. This behavioural coding system by Crnic and colleagues,^29,30 has been previously used to connect independently observed behaviour with maternal reports of her own and her offspring’s emotional and behavioural functioning³¹ and infant psychobiological functioning.³²

Head-toes-knees-shoulders (HTKS).

Children completed the head-toes-knees-shoulders task (HTKS)³³ at the 4.5-year visit. The HTKS is a measure of children’s inhibitory control, working memory, and cognitive flexibility. During the HTKS, children first follow the interviewer’s instructions (e.g., ‘touch your head’) and then are asked to do the opposite of what the interviewer instructs (e.g., ‘when I say touch your head, you touch your toes’). A total accuracy score is calculated by summing accurate responses for each of the 30 trials (2 = correct; 1 = self-correct; 0 = incorrect). HTKS tasks were double coded from video recordings. Because very few children performed accurately beyond the first 10 items, only accuracy scores for the first 10 trials were used.

Kiddie continuous performance test (K-CPT).

The kiddie continuous performance test (K-CPT)³⁴ was administered at the 4.5-year visit. The child viewed a series of pictures that appeared on a computer screen and were asked to press the spacebar after each target stimuli (e.g., fish), which were interspersed with non-target stimuli as distractors. The task lasted an average of 7 min and consisted of three blocks of 120 stimuli each, with each block containing 24 targets and 12 distractors. CPT performance was measured in terms of inhibition, indexed by a ratio of the number of correct target responses (i.e., hits) divided by the number of total response (i.e., hits plus commission errors).

Marshmallow test.

At the 4.5-year visit, children completed the Marshmallow Test,³⁵ as a measure of food-specific delay of gratification. After being asked if they preferred one marshmallow or two, the child was told that they would receive two marshmallows if they waited to eat the first and did not get out of their seat until the interviewer returned. The number of seconds a child waited (15-min maximum) was recorded and included in analyses.

2.4.4 |. Covariates

Prenatal maternal acculturation.

Previous research has found relations between acculturation and parental feeding,^36,37 and maternal acculturation and infant weight³⁸ thus mother’s acculturation was a distal variable in the model. Acculturation was assessed at the prenatal visit with maternal country of birth (Mexico or United States), preferred language use (Spanish or English), and the Anglo (α = 0.93) and Mexican (α = 0.90) Orientation subscales of the Acculturation Rating Scale for Mexican Americans-II.³⁹ A latent variable was formed³⁸ and factor scores were saved and used as a continuous variable in statistical models. Higher factor scores indicate higher acculturation.

Infant temperamental regulation.

To control for earlier temperamental regulation in the prediction of later effortful control, mothers completed the short form of the Infant Behaviour Questionnaire-Revised (IBQ-R)⁴⁰ at the 6-week visit. Mothers responded to 91 items using 1 = Never to 7 = Always. The Orienting/Regulatory Capacity (ORC) factor, comprised of Duration of Orienting, Low Intensity Pleasure, Cuddliness, and Soothability subscales, has been shown to predict emerging effortful control.⁴¹ For the current analyses, Cuddliness and Soothability were dropped from the ORC factor to improve scale reliability (α = 0.64). Higher scores indicate higher child regulation.

3 |. ANALYTIC STRATEGY

3.1 |. Preliminary analyses

Descriptive statistics and correlations among study variables were examined using SPSS version 26. A latent growth model was estimated to characterize the change in WLZ% across 7 time points from child age 6 weeks through 2 years. Latent growth models with and without a latent quadratic growth factor were tested to account for possible non-linearity in growth rates. A latent variable approach was used to identify shared variance across observed, objective, and maternal report indicators of child self-regulation at 4.5 years (see Figure 2). The loadings of each indicator and the overall fit of the confirmatory factor analysis were examined. A second linear latent growth model was estimated to characterize change in BMIz% across 3 time points from child age 4.5 years through 7.5 years.

FIGURE 2.

Measurement model of child self-regulation at 4.5 years. Note: Standardized factor loadings shown. Error terms not shown. ‘Dysreg’ = observed child dysregulation; ‘HTKS’ = Head-Toes-Knees-Shoulders task; ‘K-CPT’ = Kiddie Continuous Performance Test; ‘Marsh. Task’ = Marshmallow Task. Model fit was acceptable: χ2(N = 219; df = 2) = 1.324, p = 0.5158; RMSEA= 0.000 (90% CI: [0.000, 0.119]); CFI = 1.000; SRMR = 0.021

3.2 |. Primary analyses

Mplus 8.4 was used for primary analyses. Missing data were accounted for using full information maximum likelihood estimation. The previously fit latent growth models and self-regulation latent variable were merged into a larger structural equation model (see Figure 2). Using grand mean centring, an interaction term (responsiveness × demandingness) was included as a predictor of self-regulation, effortful control, and BMIz% intercept and linear growth in the larger model. The full structural equation models were evaluated using including chi-square (χ²), comparative fit index (CFI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR). A non-significant chi-square statistic, CFI greater than 0.95, RMSEA less than 0.05, and SRMR less than 0.08 suggest good model fit.⁴²

4 |. RESULTS

4.1 |. Preliminary analyses

Descriptive statistics.

See Table 2 for the prevalence of children classified as underweight, overweight, and obese at each time point. Table 3 presents descriptive statistics and zero-order correlations for the indicators of self-regulation at 4.5 years. Table 4 presents descriptive statistics and zero-order correlations for the primary study variables. None of the primary study variables were skewed or kurtotic (>2 or >7, respectively). Maternal prenatal acculturation was positively correlated with the WLZ% intercept, r = 0.144, p = 0.01, and BMI% intercept at 4.5 years, r = 0.145, p = 0.025. Infant 6-week temperamental regulation was positively correlated with maternal use of responsive feeding at 3 years, r = 0.179, p = 0.01, and child effortful control at 4.5 years, r = 0.224, p = 0.001.

TABLE 2.

Prevalence of children who meet criteria for underweight, overweight, and obesity at each timepoint

Child age	Underweight	Overweight	Obese
6 wks.	22.7%	7.4%	32.5%
12 wks.	22.8%	7.3%	18.9%
18 wks.	15.2%	10.0%	22.5%
24 wks.	9.7%	16.9%	14.4%
1 year	1.1%	19.5%	22.3%
1.5 years	2.4%	20.0%	22.3%
2 years	1.5%	14.2%	23.5%
4.5 years	2.0%	15.3%	19.2%
6 years	5.1%	7.3%	19.1%
7.5 years	3.4%	13.8%	29.1%

Note: Percentile derived from weight-for-length (6 weeks–2 years) and weight-for-height (4.5–7.5 years) z-scores calculated from WHO growth charts. Underweight is defined as a BMI that is below the 5th percentile, overweight as BMI in the 85 –94th percentile, and obesity as BMI ≥ 95th percentile (in accordance with CDC standards; U.S. Preventive Services Task Force, 2010). ‘Wks.’ = weeks.

TABLE 3.

Means, SDs, and zero order correlations among self-regulation variables at 4.5 years

Variables	1	2	3	4
1. Child Dysregulation composite	–
2. CPT: Inhibition ratio	0.3	–
3. Marshmallow Task	−0.11	0.3	–
4. Head-Toes-Knees-Shoulders Accuracy	−0.13	0.26	0.19	–
Mean	4.36	0.50	3.72	0.19
SD	1.76	0.27	4.60	0.31

Note: All ages refer to child age. Correlation coefficients presented in bold are statistically significant, p < 0.05. CPT, Kiddie Continuous Performance Test; Marshmallow Task, time to completion in minutes.

TABLE 4.

Descriptive statistics and zero-order correlations among primary study variables

Variables	1	2	3	4	5	6	7	8	9
1. WLZ intercept	–
2. WLZ slope	−0.88	–
3. WLZ quadratic	0.85	−0.92	–
4. Responsiveness 3 years	0.05	−0.03	0.01	–
5. Demandingness 3 years	0.07	−0.15	0.07	−0.28	–
6. Effortful control 4.5 years	0.06	−0.03	0.02	0.19	0.12	–
7. Self-regulation 4.5 years	−0.03	0.02	−0.08	0.02	0.05	0.18	–
8. BMI percentile intercept	0.22	0.03	0.16	0.14	−0.30	<0.01	<0.01	–
9. BMI percentile slope	−0.07	0.01	−0.09	−0.04	0.15	0.03	−0.03	−0.16	–
Mean	50.05	41.78	−20.71	1.16	2.86	0.00	0.00	65.15	6.49
SD	23.66	45.86	21.71	0.18	0.78	0.79	0.08	26.08	77.73

Note: All ages refer to child age. Correlation coefficients presented in bold are statistically significant, p < 0.05. WLZ, weight-to-length z-score; Self-regulation, WLZ intercept, WLZ slope, WLZ quadratic, BMI percentile intercept, and BMI percentile slope are latent variables and thus factor scores from the measurement model are used in this table. Main analyses did not use factor scores. BMI percentile is the weight-to-height ratio based on age and gender percentile score.

Growth model for WLZ% from 6 weeks through 2 years.

The latent growth model with a quadratic growth factor fit the data better than the model without the quadratic, indicating non-linearity in growth rates. Traditional model fit indices (e.g., CFI, SRMR, and RMSEA) are not appropriate for growth models due to definitional violations, incorrect baseline models, and mean structure mis-specifications.⁴³ The residual variance of the 2-year variable was constrained to zero due to lack of unexplained variance not accounted for by the other predictors.⁴⁴

The mean WLZ% for the first 7 time points were as follows: 6-week M = 56.64 (SD = 40.96), 12-week M = 46.20 (SD = 38.35), 18-week M = 55.58 (SD = 37.12), 24-week M = 57.26 (SD = 34.70), 1-year M = 75.73 (SD = 27.00), 1.5-year M = 69.45 (SD = 29.52), and 2-year M = 67.91 (SD = 28.50). From 6 weeks to 2 years, children showed significant linear gains in WLZ% relative to the reference sample, β = 4.178, SE = 0.614, p < 0.001. The results also indicated a significant non-linear growth factor, β = −2.071, SE = 0.345, p < 0.001. There was sufficient variability among infants for the intercept (β = 9.003, SE = 1.226, p < 0.001), slope (β = 43.966, SE = 9.155, p < 0.001), and quadratic growth factors (β = 11.518, SE = 2.883, p < 0.001).

The mean BMIz% for the last three time points were as follows: 4.5-year M = 65.65 (SD = 29.64), 6-year M = 56.04 (SD = 32.20), and 7.5-year M = 68.36 (SD = 31.14). Results of the linear latent growth model revealed significant variance in both intercept (β = 7.884, SE = 1.275, p < 0.001) and slope (β = 0.926, SE = 0.208, p < 0.001).

Self-regulation measurement model.

Indicators for the 4.5-year model of self-regulation included the dysregulation composite score (negative loading), the HTKS accuracy score across the first ten trials, the number of seconds lasted in the Marshmallow Test, and a measure of inhibitory control and sustained attention from the CPT. The measurement model indicated significant factor loadings for all variables and fit the data well: χ²(N = 219; df = 2) = 1.32, p = 0.52; RMSEA= 0.00, 90% CI [0.00, 0.12]; CFI = 1.00; SRMR = 0.02.

4.2 |. Primary analyses

The fit of the final structural equation model was acceptable, χ² [120, N = 322] = 256.098, p < 0.001; CFI = 0.83; RMSEA = 0.06, 90% CI [0.05, 0.07]; SRMR = 0.07. Standardized results are reported on Figure 2. Higher prenatal maternal acculturation predicted a higher infant WLZ% intercept, β = 0.191, SE = 0.072, p = 0.008. Prenatal maternal acculturation was not significantly related to the linear and quadratic WLZ% growth rates or the two feeding dimensions. Children with linearly increasing rates of change in WLZ% across time were more likely to have mothers who used more demanding feeding practices, β = −0.852, SE = 0.241, p < 0.001 at 3 years. The direct paths from the intercept, slope, and quadratic factors to child self-regulation and effortful control at 4.5 months did not reach statistical significance. The WLZ% intercept, linear slope, and quadratic factors from 6 weeks to 2 years significantly and positively predicted the intercept of child BMIz% from 4.5 to 7.5 years (all p’s < 0.001). The WLZ% linear rate of change negatively predicted the BMIz% linear rate of change from 4.5 to 7.5 years, β = [C0]0.672, SE = 0.321, p = 0.036.

Higher infant temperamental regulation predicted more responsive feeding at 3 years, β = 0.179, SE = 0.065, p = .006, and higher effortful control at 4.5 years, β = 0.179, SE = 0.065, p = 0.006. Higher effortful control at 4.5 years was also predicted by greater maternal responsiveness, β = 0.183, SE = 0.079, p = 0.021, and demandingness, β = 0.167, SE = 0.08, p = 0.037, at 3 years. Higher effortful control was associated with higher self-regulation at 4.5 years, β = 0.284, SE = 0.107, p = 0.008. Self-regulation and effortful control at 4.5 years were not significantly related to BMIz% intercept or linear slope from 4.5 to 7.5 years. Children of mothers who reported more demandingness during feeding at 3 years had a lower BMIz% intercept, β = [C0]0.172, SE = 0.073, p = 0.018. The interaction of responsiveness and demandingness at 3 years was not significantly related to any of the other variables. Removal of the interaction from the structural equation model yielded a more parsimonious model without compromising model fit.

4.3 |. Post-hoc analyses

To determine whether lower demandingness predicted clinically significant overweight classification (85–94th percentile), a multiple regression was run that accounted for child WLZ% at 2 years. One SD above the mean of demandingness was associated with a child BMI at the 61st percentile, while 1 SD below the mean of demandingness was associated with a child BMI at the 70th percentile.

5 |. DISCUSSION

The existing literature on parent feeding styles and child weight has used a categorical approach in both cross-sectional and longitudinal studies, finding that an indulgent feeding style is associated with weight gain and obesity risk.⁷ However, some research has not found this association.^14,15,45 Previous researchers have noted that although a person-centred (i.e., categorical) approach can be informative, a variable-centred approach (i.e., dimensions as continuous variables) can be superior and more powerful in examining maternal feeding with child eating and weight.²⁵ Further, there can be costs to dichotomizing data. Several studies have used the dimensions of demandingness and responsiveness in analyses instead of categorical approaches, similar to the current study.^14–16,25 Jansen, Domènech Rodriguez, and colleagues have raised concerns about the conceptual and assessment equivalency of the labels ‘authoritarian’, ‘authoritative’, ‘uninvolved’, and ‘indulgent’ when applied to ethnic minority populations in the parenting styles and the feeding literature.^8,46,47 Thus, future research should consider alternative approaches to assessing feeding styles, and careful language use with ethnic minority populations.

The present study significantly contributes to the literature in several ways. Few studies have employed a longitudinal design, with 12 assessments from pregnancy to 7.5 years child age among a low-income Mexican American population, who are at elevated risk for obesity.² Extending unidirectional models whereby parents influence child weight, we assessed how early child weight-for-length trajectories may influence later maternal demandingness and responsiveness during meal times, and conversely, how maternal feeding dimensions may influence subsequent child BMI after accounting for early child weight-for-length trajectories. However, the relationship and interaction between parents and children during the feeding context is bidirectional and can be influenced by a myriad of factors such as culture, income, parent temperament, child temperament, just to name a few.

We found that more rapid weight-for-length gain in infancy was linked with more maternal demandingness during feeding at 3 years. In turn, more demandingness was associated with lower child BMI at age 4.5 years after adjusting for child weight-for-length at 2 years. This result is consistent with previous research among low-income families demonstrating an association between demandingness and lower child BMI.²⁵ Contrary to research on toddlers and preschoolers,^8,12 the interaction term between responsiveness and demandingness was not significant in predicting child BMI. However, these findings are consistent with another study where the interaction term was not significantly associated with child BMI among Black and Latino preschoolers and their mothers.²⁵ At age 3, it is normative for children to continuously change the foods they like or do not like to eat (sometimes even daily), refuse to try new foods, self-feed more regularly, and exhibit fussiness. However, between 4.5 to 7.5 years of age, children start school. It may be that maternal feeding at age 3 is not predictive across longer time spans of childhood and/or the introduction of school diminishes the influence of maternal feeding. Therefore, future research is needed to assess whether our findings replicate in other studies and developmental periods.

Contrary to predictions, child self-regulation did not explain the effect of maternal feeding on later child BMI. Our measures of child self-regulation assessed general self-regulatory ability with a focus on child’s effortful control ability. There are several possibilities for the lack of a relation. First, we evaluated the impact of child self-regulation at 4.5 years on weight gain from 4.5 to 7.5 years. Prior research found that child self-regulation early in life (e.g., age 2–3) predicted change in child BMI from age 2 to 5.5.^20,21 The uniqueness of our sample may also help explain the null findings. With a similar sample of Hispanic preschoolers, Hughes and colleagues¹⁹ found that only poorer self-regulation during a food-related task was associated with child BMI; general self-regulation was not related to child BMI. Other factors may play a stronger role in child weight gain among this population, such as cultural traditions, expectations, and norms surrounding child weight.

Maternal report of infant temperamental regulation was linked to more responsive feeding at 3 years and higher child effortful control at 4.5 years. Consistent with previous research, infant temperamental regulation may have an evocative effect on maternal behaviours, including feeding.⁴⁸ Further, our results suggest that responsiveness is positively related to later effortful control, even after accounting for continuity in the maternal report of child temperamental regulation. Mothers with well-regulated infants may be able to use more responsive feeding than those feeding a difficult infant. Over time, more responsiveness and less control over child eating and food choices may promote broader aspects of child self-regulation. Our results add support to the body of literature establishing the link between maternal support and responsiveness and the emergence of effortful control during the preschool years.

The results must be considered in the context of study limitations. Although we examined the influence of child weight trajectories on maternal feeding at 3 years, we were unable to evaluate concurrent influences of general maternal parenting style on child weight from 6 weeks to 2 years of age. In addition, as is typical with low SES, largely immigrant ethnic minority samples, we experienced attrition across 7.5 years. Also, whereas prior work has suggested child general self-regulatory ability influences child weight and BMI,^20,21 other work among low-income and ethnic minority children demonstrated only eating-specific self-regulation was associated with child weight.¹⁹ Although our assessment of child self-regulation benefited from the use of multiple standardized assessments, including a marshmallow task, we did not evaluate eating-specific self-regulation independently. Next, while the mothers in our sample report being the primary caregiver responsible for child feeding, assessment of others involved in child feeding may provide a more complete picture of the feeding environment (e.g., grandmothers, aunts, fathers). In addition, this study did not find a significant interaction between demandingness and responsiveness, which needs to be further explored in future research. Finally, our results may not be generalized to higher-SES children or those from other ethnic minority groups.

In a sample of low-income Mexican American children, the current study examined the relation of child weight trajectories in infancy with maternal demandingness and responsiveness in toddlerhood, and in turn, the association of maternal feeding dimensions with later child BMI.

Larger longitudinal studies are needed to disentangle the bidirectional relationship between child weight and maternal feeding across child development. The finding that infant temperamental regulation was associated with more responsive feeding and effortful control suggests that maternal feeding interventions should be approached through a dyadic framework. Rather than having standard prescriptions for all mothers, our findings suggest that exploring the dyadic relationship between the specific mother and child will offer most promise for positive outcomes. Future research may explore how other aspects of a child (e.g., temperament, personality, self-regulation) combined with mother’s perception of child weight over time longitudinally impact maternal feeding.