Feeding Problems as an Indicator of Developmental Delay in Early Childhood

Abstract

Objective:

To determine if feeding problems are an indicator of developmental delay.

Study design:

In this prospective longitudinal cohort study, mothers of 3,597 children (49% female, 35% multiples) reported on their children’s feeding problems and developmental delays (using the Ages and Stages Questionnaire, ASQ) when children were 18, 24, and 30 months of age. Average scores of feeding problems were computed at each age, as well as a categorical score indicating persistently high feeding problems ≥90^th percentile across time. The Battelle Developmental Inventory (BDI-2) assessed development in 5 domains for a subset of children at 4 years.

Results:

In adjusted analyses, feeding problems (per point increase) were increasingly associated with six ASQ domains from 18 months (Odds Ratios (ORs) ranged from 1.30 to 1.98) to 24 months (ORs = 2.07 – 2.69) to 30 months (ORs = 3.90 – 5.64). Compared with children who never experienced feeding problems, children who experienced high feeding problems at one or two time points were more than twice as likely to have a delay on all ASQ domains (ORs = 2.10 – 2.50), and children who experienced high feeding problems at all three time points were four or more times as likely to have a delay on all ASQ domains (ORs = 3.94 – 5.05). Children with one-point higher feeding problems at 30 months scored 3-4 points lower on all BDI-2 domains at 4 years.

Conclusions:

Frequent feeding problems, especially those that persist into the third year, could be used to identify children at risk for developmental delay for more targeted screening.

Children with severe developmental disabilities have high rates of feeding disorders¹ that result in insufficient weight gain or growth. Feeding problems (eg, crying during meals, pushing food away, and gagging)² are also common in young children without disabilities.^3-6 Few population-based studies of associations between feeding problems and developmental delays are available, and feeding problems are often operationalized in a limited way (e.g., only picky eating). Feeding problems may relate to developmental delays because they are indicative of underlying neurological problems,⁷ they contribute to undernutrition and poor physical growth which impact development,^{8, 9} or they relate to emotional temperament,^{10, 11} which may contribute to social and behavioral delays.¹²

Feeding is a complex activity that requires coordinating oral-motor, neurologic (e.g., hunger cues), gastrointestinal, cognitive (e.g., sensation and perception), and social cues and feedback.⁷ Feeding problems can be organic or nonorganic (e.g., behavioral/social).^{13, 14} Feeding is a major parental concern,¹⁵ with nutrition/feeding discussed extensively at pediatric visits.¹⁶ Feeding problems can be stressful for parents,^{3, 17} and may reinforce coercive interaction patterns between parents and children.^{10, 18} The severity of feeding problems may be an early indicator of developmental delays and may indicate failure of one or more of its integrated systems (e.g., neurodevelopment, behavioral regulation, cognition).¹

One longitudinal study noted stability in feeding problems from 5-20 months and 4.5 years, and that feeding problems at 5 months, and persistent problems over time, were associated with more problems in adaptive and social behavior at 4.5 years.¹⁹ In another prospective study, persistent feeding problems (refusing food and not eating well at 1.5, 3, and 6 years) were predictive of pervasive developmental problems at age 7 years.²⁰ Additional cross-sectional studies also suggest a link between early childhood feeding disorders and developmental disabilities.^21-23

The current study investigated associations between the frequency of feeding problems and developmental delays at three time points from 18 to 30 months. We explored feeding problems in the second and third years because this is when solid food feeding habits are established,²⁴ and studies have demonstrated stability in feeding problems across early childhood.^{19, 22} In a subgroup, we explored associations between feeding problems at 30 months and performance on a standardized developmental test at 4 years. We hypothesized that feeding problems and developmental delays would be associated at all time points with stronger associations later in development and for high-level feeding problems that persist across the three time points.

Method

Participants

The Upstate KIDS Study is a population-based birth cohort designed to evaluate the impact of infertility treatment on child health.²⁵ Infants conceived by infertility treatment were oversampled, but no association between treatment and early development was observed.²⁶ A total of 5,034 mothers and 6,171 children were recruited into the study when children were 4 months old. Main analyses were restricted to 3,597 children with feeding behaviors and developmental delay data for at least one timepoint from 18 to 30 months of age, including 2,338 singletons and 1,259 multiples.

A subgroup of children was invited to participate in a developmental assessment at age 4 years including children who had a probable developmental delay on the Ages and Stages Questionnaire²⁷ (ASQ) at age 30-36 months or the Modified Checklist for Autism in Toddlers²⁸ (MCHAT) at 18-24 months; children referred to the New York State Early Intervention Program for evaluation; twins and higher order multiples (regardless of their screening status); and a random selection of singleton children who passed both the ASQ and MCHAT. Of the 2,154 children invited, 601 (27.9%) participated and feeding problem data were available from 516 children, 408 (79%) of whom passed the ASQ and MCHAT, and 108 (21%) of whom failed one or both.

Human subjects research approval was obtained from participating institutions (NYSDOH IRB #07-097; UAlbany #08-179), and informed consent was obtained.

Procedures

Feeding Problems.

Mothers responded to questions, derived from previous research,^{8, 9} about how often their children engaged in behaviors indicating feeding problems (0=never, 1=rarely, 2=sometimes, 3=often) at 18, 24, and 30 months. Items focused on observable behaviors at mealtimes indicating distress, food refusal, and mechanical feeding difficulties. Additional items were asked at 24 and 30 months that reflect behaviors more common at older ages (Table I (available at www.jpeds.com) shows specific items). At each age, items were averaged to create scales indicating the severity of feeding problems (range = 0-3). Internal consistency (α) reliability ranged from .81 to .86 across age, and feeding problem scales were correlated across age, rs ranged from .37 to .45, ps < .001. A revised version of this scale has been validated in healthy and undernourished 6- to 24-month-old children in six countries.²⁹

Table 1.

Feeding problem items and administration schedule

How often does your child do the following when given food?	Subscale	18 months	24 months	30 months
Push food/spoon away	R	X	X	X
Turns head away repeatedly	R	X	X	X
Closes mouth when offered food	R	X	X	X
Can’t chew solid foods	M	X	X	X
Gags on food	M	X	X	X
Holds food in mouth	M	X	X	X
Spits food out	M	X	X	X
Throws food	M	X	X	X
Cries/screams during meals	M	X	X	X
Refuses to eat a specific meal	R		X	X
Refuses to eat a specific type of food	R		X	X
Spills food	M			X

Note. R= Item loaded on food refusal subscale. M = Item loaded on the mechanical/distress subscale. X = item was administered at this age. See Wright, Gurney ²⁹ for a revised and validated version of this questionnaire.

In addition to the continuous feeding problems scale score, we coded persistent feeding problems as 0=never, 0 time points, 1=sometime, 1-2 time points, or 2=always, all 3 time points for scoring ≥ 90^th percentile on the feeding problems score from 18-30 months. The 90^th percentile cutoffs were 1.66 at 18 months, 1.64 at 24 months, and 1.75 at 30 months (on a 0-3 scale).

Developmental Delay.

Mothers completed the ASQ, third edition²⁷ at 18, 24, and 30 months. The ASQ is a validated screening instrument designed to detect developmental impairments in five domains: fine motor, gross motor, communication, personal-social functioning, and problem-solving skills.^{30, 31} ASQ items were scored as “yes” (10 points), “sometimes” (5 points), and “not yet” (0 points). A probable developmental delay on a given domain of the ASQ is defined by a score that is two or more standard deviations below the United States national average for the age group. In addition to domain scores, a total ASQ delay score was computed as having a probable delay on any domain of the ASQ. Probable delay on the ASQ was somewhat stable across time, phi (ϕ)=.40–.55, ps<.001 for ASQ total, and ϕ=.33–.60, ps<.001 for subdomains.

At the 4-year clinic visit, a subset of children were administered the Battelle Developmental Inventory, Second Edition³² (BDI-2). The BDI-2 is a standardized (M=100, SD=15) assessment that is used to measure developmental skills. The BDI-2 provides a total score as well as 5 subdomain scores: motor, communication, cognitive, personal/social, and adaptive. The BDI-2 has demonstrated reliability, internal consistency, and content and criterion validity.³³

Congenital Malformations.

Participants were linked to the New York State Department of Health Congenital Malformations Registry. Malformations include permanent conditions like chromosomal and central nervous system abnormalities as well as correctable issues (e.g., cleft lip/palate, cataract, and hernia). Overall, 244 children (6.7%) had one or more congenital malformations, which is slightly higher than the percentage across New York in 2008 (5.0%),³⁴ probably explained by the oversampling of multiples, who have a higher rates of malformations than singletons.³⁵

Statistical Analyses

All statistical analyses were performed using SAS version 9.4. To be eligible for inclusion in the analysis, participants had at least one measurement of feeding problems and ASQ from 18 to 30 months. Missing data was handled with multiple imputation using fully conditional regression estimation for continuously distributed variables and discriminant function estimation for binary variables. To account for differential missingness (detailed in the Appendix [available at www.jpeds.com]), all independent and dependent variables and covariates were included in the imputation model. Fifty datasets were imputed and statistics aggregated over the different datasets using Rubin’s rules.³⁶

To assess associations between the feeding problems scale and developmental delay on the ASQ across age, generalized linear mixed models (GLMM) with restricted maximum likelihood estimation and the logit link were performed. Because multiple births were included in the sample, a random effect of family with unstructured covariance was included to account for potential within-family variance. In addition, child age was modeled with a random intercept with an auto-regressive covariance structure to account for correlations over time. An interaction between feeding problems and age group was modeled to capture age-varying associations between feeding problems and developmental delay, and if significant, analyses were stratified by age group. Following unadjusted analyses, models were adjusted for child sex, gestational age, sleep problems, and history of GERD, and maternal age, race/ethnicity, education, marital status, infertility treatment, parity, plurality, and depressive symptoms (see justification and measurement of covariates in the Appendix).

Next, persistence of high feeding problems (1 sometime and 2 always ≥ 90^th percentile versus 0 never) from 18-30 months was modeled in relation to ASQ at 30 months using the same measures previously described.

Four sensitivity analyses were conducted. First, children with registered congenital malformations were removed and analyses re-run.³⁷ This analysis was conducted to ensure that a small number of children with abnormalities that often impact feeding and development were not skewing the results.³⁸ Second and third, multiples and children born preterm (< 37 weeks gestation) were sequentially removed from the sample and analyses re-run. The study design oversampled twins, who may be at higher risk of feeding and developmental problems,³⁹ which could skew the results. Preterm children are also likely to be at risk for both feeding problems and developmental delay. Finding consistency in the results for only full-term children would support the generalizability of the findings. Fourth, exploratory principal components analyses with promax rotation were conducted to determine whether the feeding items separated into subscales. Two components were extracted at each time point representing food refusal and mechanical and distress problems (Table 1). Given the theoretical distinction between mechanical and distress items, we explored mechanical/distress as a single scale, and also separated these items into two scales. Analyses were explored with these subscales to determine if one or another type of feeding problem was more strongly associated with developmental delays.

In subgroup analyses of the Battelle, the analyses of the continuous feeding scale were repeated (substituting the logit link for an identity link), using the 30-month feeding problems scale as a predictor of developmental functioning on the Battelle total and subdomains.

Results

Table 2 displays the descriptive statistics. On average, children experienced feeding problems between rarely and sometimes. Most children (77%) did not experience high levels of feeding problems any time between 18 and 30 months, but 21% experienced high feeding problems at one or two time points, and 2% experienced persistently high feeding problems at all three time points. A small percentage of children has an ASQ delay at any time (8-11% had a delay in one or more domains of the ASQ, and 3-7% had a delay in an individual domain).

Table 2.

Demographic and perinatal characteristics of the sample, feeding problems, and developmental delay status across waves

	n	M (SD) or %	range
Feeding Problems Scale
18 months	2795	1.14 (0.46)	0.11-2.89
24 months	2462	1.00 (0.49)	0.09-2.82
30 months	2475	1.26 (0.39)	0.08-3.00
Persistent Feeding Problems 1	3597
Never ≥ 90th percentile		77
Sometime ≥ 90th percentile		21
Always ≥ 90th percentile		2
Developmental Delays
ASQ Total
18 months	2573	8
24 months	2180	11
30 months	1957	10
ASQ Fine Motor
18 months	2623	3
24 months	2221	4
30 months	2012	3
ASQ Gross Motor
18 months	2626	3
24 months	2229	4
30 months	2030	5
ASQ Communication
18 months	2616	3
24 months	2234	7
30 months	2011	5
ASQ Personal-Social
18 months	2626	3
24 months	2236	5
30 months	2029	3
ASQ Problem Solving
18 months	2605	3
24 months	2214	3
30 months	2006	4
4-year Development
Battelle Total	515	113.31(18.91)	46-145
Motor	516	107.79(17.97)	56-144
Communication	515	111.81(19.31)	55-145
Cognitive	514	109.43(18.67)	55-143
Personal-Social	516	115.15(16.13)	55-145
Adaptive	515	105.11(16.55)	55-142
Covariates
Child sex (female)	3597	49
Gestational age (weeks)	3597	37.57 (2.83)	23-42
Sleep problem
18 months	2788	11
24 months	2557	9
30 months	2475	12
History of GERD (4-24 months)	3251	17
Maternal age (years)	3597	31.27 (2.83)	23-42
Non-Hispanic white ethnicity/race	3597	84
College educated	3597	59
Married or civil union	3454	83
Infertility treatment	3597	37
Nulliparous	3565	50
Plurality	3597	35
Maternal depression (24 months)	2724	2.21 (2.40)	0-15

Note: ASQ = Ages and Stages Questionnaire©. GERD = gastroesophageal reflux disease.

¹Weighted frequencies based on multiply imputed data.

In unadjusted models of the feeding problem scale with ASQ delay, there was always a significant or marginally significant feeding problem by age interaction (24 month vs. 18 month ORs=1.12–1.36; 30 month vs. 18 month ORs=1.67–2.78) so analyses were stratified by child age. In stratified analyses (Table 3), feeding problems were associated with an increased risk of total ASQ delay and delay in nearly all subdomains at all ages. Reflecting the interactions, point estimates indicated stronger associations between feeding problems and ASQ delay as the children aged (ORs=1.50–2.45 at 18 months, ORs=2.36–3.12 at 24 months, and ORs=4.50–6.60 at 30 months).

Table 3.

GLMM results for the feeding problems scale with ASQ at 18, 24, and 30 months

	18 months			24 months			30 months
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
Unadjusted
ASQ Total	1.77	1.26, 2.47	<.001	2.41	1.85, 3.15	<.001	4.50	3.08, 6.58	<.001
Fine Motor	2.45	1.46, 4.12	<.001	3.11	2.03, 4.75	<.001	6.08	3.52, 10.52	<.001
Gross Motor	1.50	0.92, 2.44	.102	2.36	1.48, 3.77	<.001	5.38	3.37, 8.60	<.001
Communication	2.39	1.42, 4.03	.001	2.57	1.84, 3.58	<.001	5.40	3.33, 8.76	<.001
Personal-Social	1.99	1.20, 3.29	.008	2.92	1.99, 4.29	<.001	6.60	3.79, 11.49	<.001
Problem Solving	2.23	1.39, 3.58	.001	3.12	1.96, 4.96	<.001	4.78	2.83, 8.06	<.001
Adjusted
ASQ Total	1.49	1.05, 2.11	.027	2.10	1.58, 2.79	<.001	3.90	2.61, 5.82	<.001
Fine Motor	1.98	1.16, 3.38	.012	2.69	1.74, 4.15	<.001	5.32	3.01, 9.42	<.001
Gross Motor	1.30	0.78, 2.16	.304	2.07	1.25, 3.43	.005	4.92	2.98, 8.11	<.001
Communication	1.76	1.05, 2.97	.033	2.14	1.51, 3.03	<.001	4.42	2.65, 7.36	<.001
Personal-Social	1.55	0.92, 2.61	.101	2.38	1.60, 3.54	<.001	5.64	3.14, 10.12	<.001
Problem Solving	1.87	1.14, 3.07	.012	2.63	1.65, 4.20	<.001	4.08	2.37, 7.03	<.001

Note: GLMM = generalized linear mixed model. ASQ = Ages and Stages Questionnaire©. Adjusted models control for child sex, gestational age, sleep problems (time-varying), history of gastroesophageal reflux disease, maternal age, non-Hispanic white ethnicity/race, college education, marital status, infertility treatment, parity, plurality, and maternal depression. Models are stratified by child age because interactions between child age and feeding problems were significant or marginally significant.

In adjusted analyses (Table 3), children with a one-point increase in the feeding problems score were more than twice as likely to have a delay on the ASQ total and all subscales at 24 months and four or more times as likely to have a delay at 30 months. Associations were weaker at 18 months, but consistent with the directions observed at 24 and 30 months. Full model results are in Tables 4, 5, 6, 7, 8, and 9.

Table 4.

Adjusted GLMM results for the feeding problems scale with ASQ Total

	18 months			24 months			30 months
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
ASQ total
Feeding problems	1.49	1.05, 2.11	.027	2.10	1.58, 2.79	<.001	3.90	2.61, 5.82	<.001
Child sex	0.77	0.59, 1.00	.046	0.62	0.49, 0.78	<.001	0.60	0.46, 0.78	<.001
Gestational age	0.90	0.86, 0.95	<.001	0.87	0.83, 0.90	<.001	0.87	0.84, 0.91	<.001
Sleeping problem	1.07	0.72, 1.59	.724	1.17	0.81, 1.67	.403	1.30	0.88, 1.91	.186
History of GERD	1.62	1.15, 2.29	.006	1.68	1.23, 2.28	.001	1.40	0.99, 1.99	.055
Maternal age	0.98	0.96, 1.00	.115	1.00	0.97, 1.02	.808	0.98	0.96, 1.01	.148
Non-Hispanic white ethnicity/race	0.81	0.57, 1.03	.233	0.79	0.58, 1.08	.136	0.68	0.50, 0.93	.015
College educated	0.77	0.57, 1.03	.080	0.80	0.62, 1.04	.100	0.94	0.70, 1.26	.679
Married or civil union	0.78	0.54, 1.15	.210	0.90	0.64, 1.26	.539	0.62	0.43, 0.87	.006
Infertility treatment	1.25	0.92, 1.68	.149	1.06	0.80, 1.40	.685	1.01	0.75, 1.36	.942
Nulliparous	0.69	0.52, 0.91	.009	0.74	0.57, 0.97	.027	0.69	0.52, 0.91	.008
Plurality	1.35	0.99, 1.84	.056	1.04	0.79, 1.36	.789	1.43	1.06, 1.91	.017
Maternal depression	1.08	1.01, 1.14	.020	1.08	1.02, 1.14	.006	1.07	1.01, 1.14	.029

Note: GLMM = generalized linear mixed model. ASQ = Ages and Stages Questionnaire©. GERD = gastroesophageal reflux disease.

Table 5.

Adjusted GLMM results for the feeding problems scale with ASQ Fine Motor

	18 months			24 months			30 months
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
ASQ Fine Motor
Feeding problems	1.98	1.16, 3.38	.012	2.69	1.74, 4.15	>.001	5.32	3.01, 9.42	>.001
Child sex	0.65	0.45, 0.94	.023	0.87	0.63, 1.21	.406	0.65	0.45, 0.93	.020
Gestational age	0.94	0.88, 1.01	.075	0.88	0.83, 0.93	>.001	0.90	0.85, 0.96	>.001
Sleeping problem	1.18	0.67, 2.09	.565	1.41	0.86, 2.33	.177	1.20	0.71, 2.03	.487
History of GERD	1.47	0.89, 2.44	.132	1.47	0.95, 2.27	.082	1.28	0.76, 2.15	.352
Maternal age	0.97	0.94, 1.01	.121	0.98	0.95, 1.01	.167	0.99	0.96, 1.03	.727
Non-Hispanic white ethnicity/race	0.65	0.41, 1.04	.075	0.79	0.51, 1.22	.289	0.60	0.40, 0.89	.012
College educated	0.74	0.48, 1.14	.174	0.84	0.59, 1.21	.354	0.88	0.60, 1.30	.514
Married or civil union	0.78	0.45, 1.35	.374	0.85	0.53, 1.37	.512	0.56	0.35, 0.90	.017
Infertility treatment	1.18	0.76, 1.82	.467	1.15	0.77, 1.70	.497	1.04	0.70, 1.53	.860
Nulliparous	0.71	0.47, 1.08	.106	0.64	0.44, 0.93	.019	0.51	0.34, 0.75	.001
Plurality	1.33	0.85, 2.08	.205	1.22	0.85, 1.76	.274	1.57	1.07, 2.30	.020
Maternal depression	1.08	0.99, 1.18	.072	1.07	0.99, 1.17	.095	1.04	0.96, 1.14	.320

Note: GLMM = generalized linear mixed model. ASQ = Ages and Stages Questionnaire©. GERD = gastroesophageal reflux disease.

Table 6.

Adjusted GLMM results for the feeding problems scale with ASQ Gross Motor

	18 months			24 months			30 months
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
ASQ Gross Motor
Feeding problems	1.30	0.78, 2.16	.304	2.07	1.25, 3.43	.005	4.92	2.98, 8.11	>.001
Child sex	0.90	0.63, 1.29	.576	0.84	0.60, 1.18	.315	0.76	0.56, 1.04	.084
Gestational age	0.92	0.86, 0.98	.009	0.87	0.82, 0.92	>.001	0.86	0.82, 0.91	>.001
Sleeping problem	0.99	0.54, 1.80	.969	1.23	0.73, 2.06	.438	1.30	0.82, 2.06	.266
History of GERD	1.73	1.09, 2.75	.020	1.53	0.98, 2.37	.060	1.44	0.96, 2.17	.077
Maternal age	0.98	0.95, 1.02	.337	0.99	0.96, 1.02	.575	0.98	0.95, 1.01	.142
Non-Hispanic white ethnicity/race	0.82	0.50, 1.36	.444	0.76	0.49, 1.18	.223	0.82	0.56, 1.21	.322
College educated	0.83	0.55, 1.25	.365	0.97	0.67, 1.41	.878	1.12	0.79, 1.60	.528
Married or civil union	0.86	0.50, 1.50	.603	0.79	0.49, 1.28	.343	0.56	0.37, 0.85	.007
Infertility treatment	1.27	0.85, 1.90	.249	1.39	0.92, 2.11	.114	1.24	0.88, 1.75	.220
Nulliparous	0.74	0.51, 1.09	.132	0.62	0.42, 0.90	.013	0.52	0.37, 0.72	>.001
Plurality	1.22	0.80, 1.84	.353	1.33	0.90, 1.98	.155	1.55	1.10, 2.20	.012
Maternal depression	1.04	0.96, 1.14	.326	1.05	0.96, 1.14	.305	1.03	0.96, 1.11	.401

Note: GLMM = generalized linear mixed model. ASQ = Ages and Stages Questionnaire©. GERD = gastroesophageal reflux disease.

Table 7.

Adjusted GLMM results for the feeding problems scale with ASQ Communication

	18 months			24 months			30 months
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
ASQ Communication
Feeding problems	1.76	1.05, 2.97	.033	2.14	1.51, 3.03	<.001	4.42	2.65, 7.36	<.001
Child sex	0.57	0.37, 0.87	.010	0.52	0.38, 0.71	<.001	0.53	0.38, 0.75	<.001
Gestational age	0.94	0.87, 1.02	.144	0.89	0.85, 0.94	<.001	0.89	0.85, 0.94	<.001
Sleeping problem	1.03	0.56, 1.87	.933	1.40	0.91, 2.15	.130	1.19	0.74, 1.93	.470
History of GERD	2.41	1.47, 3.94	.001	1.79	1.22, 2.62	.003	1.54	1.01, 2.37	.047
Maternal age	0.98	0.94, 1.01	.182	1.01	0.98, 1.04	.620	0.99	0.95, 1.02	.356
Non-Hispanic white ethnicity/race	0.68	0.41, 1.12	.126	0.82	0.54, 1.23	.327	0.60	0.42, 0.86	.006
College educated	0.91	0.57, 1.45	.685	0.75	0.53, 1.06	.100	0.82	0.57, 1.18	.280
Married or civil union	0.61	0.36, 1.03	.065	0.73	0.49, 1.08	.114	0.59	0.39, 0.89	.012
Infertility treatment	1.06	0.68, 1.66	.789	0.75	0.52, 1.08	.122	0.96	0.66, 1.39	.821
Nulliparous	0.61	0.38, 0.96	.034	0.80	0.58, 1.10	.171	0.69	0.48, 0.98	.040
Plurality	1.70	1.09, 2.65	.020	1.32	0.94, 1.85	.112	1.37	0.95, 1.99	.091
Maternal depression	1.13	1.03, 1.24	.009	1.11	1.04, 1.18	.001	1.08	1.00, 1.16	.059

Note: GLMM = generalized linear mixed model. ASQ = Ages and Stages Questionnaire©. GERD = gastroesophageal reflux disease.

Table 8.

Adjusted GLMM results for the feeding problems scale with ASQ Personal-Social

	18 months			24 months			30 months
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
ASQ Personal-Social
Feeding problems	1.55	0.92, 2.61	.101	2.38	1.60, 3.54	<.001	5.64	3.14, 10.12	<.001
Child sex	0.62	0.43, 0.91	.014	0.58	0.42, 0.81	.001	0.61	0.43, 0.88	.007
Gestational age	0.91	0.85, 0.98	.018	0.89	0.84, 0.94	<.001	0.88	0.84, 0.93	<.001
Sleeping problem	1.26	0.72,2.21	.414	1.54	0.99, 2.42	.058	1.28	0.77, 2.13	.331
History of GERD	1.61	1.01, 2.59	.047	1.68	1.12, 2.53	.013	1.68	1.08, 2.61	.021
Maternal age	0.99	0.95, 1.02	.437	0.99	0.96, 1.02	.593	0.99	0.95, 1.02	.420
Non-Hispanic white ethnicity/race	0.85	0.51, 1.41	.519	0.58	0.39, 0.84	.005	0.74	0.50, 1.10	.137
College educated	0.76	0.49, 1.19	.229	0.90	0.63, 1.27	.533	0.91	0.62, 1.35	.651
Married or civil union	0.75	0.45, 1.25	.269	0.85	0.56, 1.29	.446	0.55	0.35, 0.88	.012
Infertility treatment	1.29	0.85, 1.96	.227	1.09	0.75, 1.57	.656	1.17	0.80, 1.72	.410
Nulliparous	0.67	0.44, 1.03	.066	0.73	0.52, 1.02	.068	0.60	0.41, 0.87	.008
Plurality	1.38	0.91, 2.11	.132	1.30	0.92, 1.85	.139	1.39	0.95, 2.04	.087
Maternal depression	1.11	1.02, 1.22	.018	1.11	1.02, 1.20	.011	1.03	0.95, 1.12	.436

Note: GLMM = generalized linear mixed model. ASQ = Ages and Stages Questionnaire©. GERD = gastroesophageal reflux disease.

Table 9.

Adjusted GLMM results for the feeding problems scale with ASQ Problem Solving

	18 months			24 months			30 months
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
ASQ Problem Solving
Feeding problems	1.87	1.14, 3.07	.013	2.63	1.65, 4.20	>.001	4.08	2.37, 7.03	>.001
Child sex	0.71	0.50, 1.01	.056	0.72	0.50, 1.03	.072	0.71	0.51, 0.99	.042
Gestational age	0.96	0.90, 1.03	.257	0.90	0.85, 0.95	>.001	0.90	0.85, 0.95	>.001
Sleeping problem	0.97	0.56, 1.70	.924	1.44	0.85, 2.42	.174	1.17	0.71, 1.93	.533
History of GERD	1.59	0.98, 2.57	.061	1.64	1.03, 2.61	.036	1.42	0.93, 2.18	.104
Maternal age	0.99	0.95, 1.02	.498	1.00	0.96, 1.03	.860	0.99	0.96, 1.02	.487
Non-Hispanic white ethnicity/race	0.85	0.52, 1.36	.490	0.85	0.53, 1.36	.487	0.75	0.51, 1.09	.130
College educated	0.68	0.45, 1.01	.055	0.81	0.55, 1.19	.285	1.04	0.72, 1.51	.846
Married or civil union	0.84	0.50, 1.39	.490	0.77	0.47, 1.25	.288	0.56	0.37, 0.87	.009
Infertility treatment	1.08	0.72, 1.62	.716	1.10	0.73, 1.65	.649	1.02	0.71, 1.47	.918
Nulliparous	0.90	0.62, 1.30	.576	0.72	0.49, 1.08	.112	0.61	0.43, 0.86	.005
Plurality	1.16	0.73, 1.83	.530	1.43	0.97, 2.10	.073	1.50	1.05, 2.16	.027
Maternal depression	1.10	1.01, 1.20	.027	1.10	1.00, 1.20	.047	1.07	0.99, 1.15	.089

Note: GLMM = generalized linear mixed model. ASQ = Ages and Stages Questionnaire©. GERD = gastroesophageal reflux disease.

In adjusted analyses of persistent feeding problems with ASQ at 30 months (Table 10), compared with children who never experienced high (≥ 90^th percentile) feeding problems, children who experienced high feeding problems at one or two time points were more than twice as likely to have an ASQ delay, and children who experienced high feeding problems at all three time points were four or more times as likely to have an ASQ delay.

Table 10.

Adjusted models of feeding problems >= 90^th percentile never (reference), sometime (1-2 time points), or always (all 3 time points) from 18 to 30 months predicting ASQ at 30 months

	OR	95% CI	p
ASQ Total
Sometime	2.10	1.53, 2.89	<.001
Always	3.94	1.89, 8.21	<.001
Fine Motor
Sometime	2.28	1.47, 3.54	<.001
Always	5.05	2.24, 11.41	<.001
Gross Motor
Sometime	2.17	1.49, 3.17	<.001
Always	3.99	1.77, 9.01	<.001
Communication
Sometime	2.28	1.55, 3.35	<.001
Always	4.22	1.93, 9.19	<.001
Personal-Social
Sometime	2.50	1.64, 3.80	<.001
Always	4.34	1.91, 9.86	<.001
Problem Solving
Sometime	2.10	1.40, 3.15	<.001
Always	4.23	1.93, 9.28	<.001

Note: ASQ = Ages and Stages Questionnaire©. Models control for child sex, gestational age, sleep problems (time-varying), history of gastroesophageal reflux disease, maternal age, non-Hispanic white ethnicity/race, college education, marital status, infertility treatment, parity, plurality, and maternal depression.

Sensitivity Analyses

Removing 244 children with congenital malformations, 1259 twins and higher order multiples, and 887 preterm (gestational age < 37 weeks) children, the adjusted findings were similar to those reported above (Tables 11 and 12).

Table 11.

Sensitivity analyses removing children with (A) congenital malformations, (B) multiple births, and (C) prematurity (gestational age < 37 weeks): adjusted GLMM results for the feeding problems scale with ASQ at 18, 24, and 30 months

	18 months			24 months			30 months
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
A. Removing congenital malformations
ASQ Total	1.48	1.02, 2.17	.041	2.15	1.60, 2.90	<.001	3.95	2.62, 5.97	<.001
Fine Motor	1.97	1.11, 3.50	<.001	2.66	1.69, 4.17	<.001	5.06	2.83, 9.04	<.001
Gross Motor	1.34	0.76, 2.35	.102	2.24	1.30, 3.85	.004	5.19	3.10, 8.69	<.001
Communication	1.82	1.03, 3.21	.001	2.05	1.43, 2.93	<.001	4.41	2.61, 7.46	<.001
Personal-Social	1.53	0.87, 2.70	.008	2.25	1.49, 3.40	<.001	5.65	3.13, 10.21	<.001
Problem Solving	1.84	1.08, 3.14	.001	2.61	1.61, 4.25	<.001	4.11	2.36, 7.16	<.001
B. Removing multiples
ASQ Total	1.70	1.05, 2.74	.030	2.14	1.51, 3.03	<.001	4.53	2.77, 7.40	<.001
Fine Motor	2.19	1.10, 4.39	.027	3.27	1.88, 5.66	<.001	4.84	2.47, 9.47	<.001
Gross Motor	1.87	0.96, 3.64	.065	2.58	1.42, 4.71	.002	5.74	3.04, 10.83	<.001
Communication	2.10	0.99, 4.45	.052	1.90	1.22, 2.97	.005	4.21	2.31, 7.70	<.001
Personal-Social	2.13	1.02, 4.45	.043	2.45	1.49, 4.02	<.001	5.91	3.01, 11.59	<.001
Problem Solving	1.99	1.05, 3.78	.035	2.99	1.62, 5.52	<.001	4.25	2.24, 8.07	<.001
C. Removing preterm
ASQ Total	1.69	1.07, 2.65	.023	2.33	1.64, 3.30	<.001	4.59	2.84, 7.43	<.001
Fine Motor	2.20	1.09, 4.43	.027	3.41	2.00, 5.82	<.001	5.34	2.74, 10.44	<.001
Gross Motor	1.63	0.86, 3.10	.132	2.66	1.47, 4.81	.001	6.44	3.46, 12.02	<.001
Communication	2.04	1.00, 4.15	.050	2.20	1.40, 3.44	<.001	4.97	2.74, 9.00	<.001
Personal-Social	1.82	0.91, 3.64	.092	2.66	1.62, 4.37	<.001	6.45	3.27, 12.71	<.001
Problem Solving	1.96	1.07, 3.59	.030	3.37	1.86, 6.11	<.001	4.58	2.46, 8.55	<.001

Note: GLMM = generalized linear mixed model. ASQ = Ages and Stages Questionnaire©. Adjusted models control for child sex, gestational age (not in preterm model), sleep problems (time-varying), history of gastroesophageal reflux disease, maternal age, non-Hispanic white ethnicity/race, college education, marital status, infertility treatment, parity, plurality (not in multiples model), and maternal depression.

Table 12.

Sensitivity analyses removing children with (A) congenital malformations, (B) multiple births, and (C) prematurity (gestational age < 37 weeks): Adjusted models offeeding problems >= 90^th percentile never (reference), sometime (1-2 time points), or always (all 3 time points) from 18 to 30 months predicting ASQ at 30 months

	A. Removing congenital malformations			B. Removing multiples			C. Removing preterm
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
ASQ Total
Sometime	2.15	1.55, 2.99	<.001	2.07	1.37, 3.11	<.001	2.29	1.56, 3.37	<.001
Always	4.22	1.96, 9.10	<.001	4.17	1.61, 10.84	.004	3.74	1.41, 9.90	.008
Fine Motor
Sometime	2.19	1.39, 3.43	<.001	2.26	1.28, 3.98	.005	2.23	1.30, 3.81	.004
Always	5.11	2.18, 12.02	<.001	4.40	1.44, 13.42	.010	3.26	0.97, 11.00	.057
Gross Motor
Sometime	2.30	1.56, 3.39	<.001	2.32	1.38, 3.88	.002	2.55	1.58, 4.10	<.001
Always	4.58	2.01, 10.45	<.001	4.41	1.44, 13.49	.009	4.19	1.33, 13.21	.015
Communication
Sometime	2.31	1.55, 3.45	<.001	1.97	1.17, 3.30	.010	2.43	1.49, 3.96	<.001
Always	4.44	1.95, 10.13	<.001	3.17	1.07, 9.39	.038	3.28	1.06, 10.14	.039
Personal-Social
Sometime	2.59	1.69, 3.99	<.001	2.57	1.49, 4.44	<.001	2.78	1.62, 4.76	<.001
Always	4.49	1.88, 10.72	<.001	3.23	0.91, 11.44	.070	2.83	0.71, 11.25	.139
Problem Solving
Sometime	2.15	1.41, 3.27	<.001	2.05	1.21, 3.48	.008	2.20	1.33, 3.64	.002
Always	4.46	1.94, 10.28	<.001	3.55	1.12, 11.27	.031	3.44	1.11, 10.61	.032

Note: ASQ = Ages and Stages Questionnaire©. Models control for child sex, gestational age (not in preterm model), sleep problems (time-varying), history of gastroesophageal reflux disease, maternal age, non-Hispanic white ethnicity/race, college education, marital status, infertility treatment, parity, plurality (not in multiples model), and maternal depression.

The food refusal and mechanical/distress feeding problem subscales identified with PCA were moderately correlated at each time point, rs=.47-.52, ps<.001. When mechanical and distress items were further delineated into mechanical (can't chew solid foods, gags on food, holds food in mouth) and distress/behavioral (spits food out, throws food, cries/screams), correlations were slightly larger between the two, rs=.51-61, ps<.001, than between food refusal and mechanical, rs=.34-.43, ps<.001, and food refusal and distress/behavioral, rs=.47-.50, ps<.001. The food refusal, mechanical, and distress/behavioral subscales (as well as the combined mechanical/distress subscale) were similarly associated with developmental delays and no clear pattern emerged of differential associations of the subscales with developmental delays. All effects were in the same direction and similar strength as the total scale (data not shown).

Subgroup Analyses

In analyses of 516 children who completed the assessment at age 4, a one-point increase in feeding problems at 30 months was associated with scoring 4-6 points lower on the Battelle total and subscales (Table 13). Adjusting for child and maternal factors, the reduction in Battelle scores was 3-4 points.

Table 13.

GLMM results for the 30 month feeding problems scale with Battelle Developmental Inventory at 4 years

	Unadjusted			Adjusted
	B	95% CI	p	B	95% CI	p
Battelle Total	−6.72	−10.87, −2.58	.002	−4.48	−8.40, −0.56	.025
Motor	−6.30	−10.39, −2.21	.003	−4.29	−8.32, −0.26	.037
Communication	−5.62	−9.97, −1.27	.011	−3.84	−8.08, 0.39	.075
Cognitive	−4.16	−8.11, −0.22	.038	−2.98	−6.81, 0.85	.128
Personal-Social	−5.98	−9.62, −2.33	.001	−3.89	−7.44, −0.34	.032
Adaptive	−6.58	−10.33, −2.84	<.001	−4.51	−8.09, −0.94	.013

Note: GLMM = generalized linear mixed model. Adjusted models control for child sex, gestational age, sleep problems at 30 months, history of GERD, maternal age, non-Hispanic white ethnicity/race, college education, marital status, infertility treatment, parity, plurality, and maternal depression.

Discussion

The results of this study suggest that feeding problems, broadly defined to include meal-time distress, food refusal, and mechanical issues, are increasingly associated with developmental delays from 18 to 30 months, and feeding problems at 30 months are associated with somewhat lower developmental scores at 4 years. Children who had high feeding problems at one or two time points were more than twice as likely to have a probable developmental delay on the ASQ, and children who had high feeding problems at all three time points were 4 or more times more likely to have a delay on the ASQ at 30 months. Removing children with congenital malformations, twins and higher-order multiples, and preterm children did not change these conclusions.

This study expands beyond previous studies that have found associations between feeding problems and development in children at risk^{1, 21, 23} as well as population studies that have found associations between picky eating and development.^{19, 20} This study used a large cohort of children at three time points, measured a range of feeding problems and developmental delays, and included a full in-person developmental assessment for a subsample of children (rather than relying exclusively on maternal reports of child development).

Much of the literature on feeding problems in normative samples focuses on food refusal or picky eating.^{4, 11, 17, 18, 20, 40} Picky eating is distressing for parents¹⁷ and may put children at risk for growth faltering.⁹ However, sensitivity analyses in this study did not indicate that a particular kind of feeding problem was more strongly associated with developmental delays. The risk for developmental delays was similar across food refusal and mechanical/distress subscales.

The associations between feeding problems and development were not limited to a particular developmental domain. In this study, we explored feeding problems as an early indicator of developmental delays because feeding problems may be easier to identify early, not because we believe feeding problems are the cause of later delays. Except in rare cases of extreme feeding problems that lead to growth faltering and failure to thrive,^{8, 9, 21} it is more likely that feeding problems in the general population are simply a symptom of an underlying developmental issue. Feeding problems may be indicative of broad delays, or may have associations with a particular developmental domain for different children. For example, children may experience feeding problems because of neurological problems that impact the ability to coordinate oral-motor, communication, and social feeding skills.⁴¹ Children may experience feeding problems because of the inability to communicate with the caregiver, leading the child to exhibit feeding problem behaviors instead.^{42, 43} Lacking fine motor skills may also lead to feeding problems by impacting the ability to self-feed and increasing frustration during meals.

There were somewhat weaker associations between feeding problems and developmental delays at 18 months compared with 30 months. Children experience improvements in oral-motor and self-feeding skills in the second year of life, but there is wide variation in individual rates of development.^{44, 45} The wide range of normative individual differences at 18 months may have attenuated associations between feeding problems and developmental delays. For example, some typically developing children may be late in acquiring a particular oral-motor skill (e.g., coordination of chewing juicy foods) at 18 months, but by 30 months, most children have strong oral-motor and self-feeding skills unless there is a lingering developmental problem.⁴⁵

The American Academy of Pediatrics (AAP) recommends screening for developmental disabilities at 9, 18, and 30 months.⁴⁶ If AAP screening guidelines are followed, most children at risk for developmental delays will be identified before the pattern of feeding problems is clear. However, an estimated 37% of pediatricians did not screen all children under 3 years in 2016,⁴⁷ and screening varies widely by state.⁴⁸ Two common perceived barriers to universal screening were the amount of time the screenings take and inadequate reimbursement.⁴⁷ Many other countries, including low- and middle-income countries, do not have routine developmental screening programs.⁴⁹ There may also be disparities in referral rates for children who screen positive for developmental concerns. In a large U.S. hospital system, over half of children who screened positive for a developmental problem were not referred to early intervention services, and referral rates differed by child sex, age, and race.⁵⁰ Screening toddlers whose mothers raise concerns about feeding behavior could help to identify children at risk for developmental delays. However, the weaker associations between feeding problems and delays earlier in development suggest that more weight should be given to feeding problems that persist into the third year. Pediatricians could also use information about feeding problems to help decide whether a child who screens positive for a developmental delay should be referred to early intervention services.

Both feeding problems and the ASQ were derived from maternal reports, which may inflate associations among constructs due to common source variance. However, results were largely supported by independent testing of a subset of children at age 4 years. Maternal reports of feeding problems have been validated against observational feeding data in other studies.⁴⁰ Mothers’ reports of developmental status have also been shown to relate to independent testing of the child.⁵¹ The 90^th percentile cut-point used to identify children with persistently high feeding problems was sample-specific, but these analyses were complimented by those using the full range of scores.

The sample in this study was largely European-American and well educated. These demographics were reflected in children’s high average standard scores on the Battelle, but there was large variation (Table 2). It is unclear to what extent the results of this study will generalize to other groups. Finally, there is always the possibility of confounding due to unmeasured variables such as detailed indicators of growth that might be associated with both feeding behavior and developmental delay.⁵² However, we did include a range of important child and family covariates (see Appendix for details).

The subsample that completed the BDI-2 was not random. Only 28% of those invited chose to participate, and invitations were stratified according to various characteristics including performance on developmental screeners. It is unclear what effect this may have had on the BDI-2 findings and they should be interpreted with caution.

Despite feeding problems being a common occurrence in early childhood,^{2, 3 15, 16} children who are reported to have high levels of feeding problems in early childhood are also more likely to have a delay on multiple domains of a developmental screening test and score lower in independent testing. High levels of feeding problems translated to mothers reporting between rarely and sometimes or more in occurrence of 9 to 12 different feeding issues. Should our findings be confirmed, taking notes of such reports in pediatric records may be a practical way to add information for targeted screening. The relatively stronger results as children age from 18 to 30 months combined with the finding that children with persistently high feeding problems from 18 to 30 months are 4 or more times as likely to have a delay on the ASQ suggest that children who experience feeding problems into their third year may be at the greatest risk and should be screened for developmental delays.

Abbreviations:

ASQ

Ages and Stages Questionnaire

BDI-2

Battelle Developmental Inventory, second edition

MCHAT

Modified Checklist for Autism in Toddlers

GERD

gastroesophageal reflux disease

GLMM

generalized linear mixed models

OR

odds ratio

PCA

principal component analysis.