Childhood feeding difficulties have an impact on behavior, dietary variety, and parental stress (1–4). The prevalence of feeding difficulties ranges from 1% to 2% in typically developing children (5,6), to 46% to 89% in children with autism spectrum disorder (ASD) (7). Feeding difficulties have been associated with a number of risk factors, including altered growth patterns (8), digestive problems (9), and impaired parent-child relationships (9,10).
Restricted dietary variety can adversely affect children's health in the long term. Where the child's preference is for energy-scarce foods (ie, a low energy/low nutrient balance), limited dietary variety may lead to growth faltering (11), cognitive impairments (12), and an altered metabolic state (13). Where the child has a preference for energy-dense foods (ie, a high energy/low nutrient balance), restricted dietary variety may lead to overweight/obesity and the resulting concomitant disease states, including cardiovascular disease and type 2 diabetes mellitus (14). Overweight/obesity in childhood has also been associated with a number of psychological sequelae, including low self-esteem and depression (15).
Despite the high prevalence and significant short- and long-term risks associated with feeding difficulties in childhood, well-designed intervention studies are scarce. The primary goals of feeding intervention typically include increasing dietary variety/volume, and reducing undesirable mealtime behaviors, but the styles of therapy by which these goals are achieved vary considerably, and lack a strong evidence base. A recent survey of practice identified that clinicians predominantly provided child-focused treatment guided by the principles of systematic desensitization (SysD, bottom-up, play-based, modeling style of intervention) or operant conditioning (OC, top-down, prompt-and-reward style of intervention), and/or provided parent-focused training (written educational material, education sessions) (16). Most clinicians were, however, underconfident in their application of these approaches (16). Within the present literature, interventions involving techniques based on OC are the most well represented (17–19), but studies are limited to single cases or case series, which makes generalization of results difficult. A recent systematic review of feeding difficulties in children with ASD indicated that interventions appeared to result in medium-large improvements in food volume intake, but only small-negligible improvements in decreasing undesirable mealtime behaviors (20). Dietary variety was seldom included as an outcome measure (20).
The primary aim of this study was to determine whether intervention across 2 therapy arms (OC vs SysD) had an impact on increasing dietary variety (number of foods consumed) and quality (number of macro-/micronutrients in which recommended daily intake was met), and decreasing the frequency of undesirable mealtime behaviors in children with feeding difficulties, to ascertain whether one approach was superior to the other. A secondary aim was to compare outcomes across 2 cohorts with feeding difficulties (children with ASD and nonmedically complex children, NMC), and 2 intensity streams (intensive vs weekly).
This prospective study for children with feeding difficulties took place at a tertiary pediatric hospital in Brisbane, Australia, between October 2011 and July 2013. Children with ASD and NMC children between the ages of 2 and 6 years were recruited via referral from parents or medical professionals. Children with ASD had a documented diagnosis by a pediatrician, psychologist, or psychiatrist. Children defined as NMC had never received treatment by a specialist physician for a medical condition. Children were eligible to participate if they had a diagnosed feeding difficulty. Diagnosis of feeding difficulties was confirmed via clinical assessment. For the diagnosis of feeding difficulties, participants were required to present with either food selectivity by type (<10 foods across each food group: fruits/vegetables, proteins, carbohydrates) (21) or food selectivity by texture (eg, only consuming purees) (21). Participants may also have presented with mealtimes averaging >30 minutes (22), and/or clinically significant difficult mealtime behaviors (1) that were having an impact on parental stress.
Children were excluded from the study if they were acutely medically unwell; clinically underweight (<5th percentile body mass index [BMI]); fed via tube; diagnosed with dysphagia for which modified fluids and textures had been prescribed or for whom severe dysphagia was limiting dietary capacity (eg, children on a liquid-only diet); unable to eat or drink orally (eg, because of severe aspiration risk; because of gastrointestinal obstruction); or had >2 allergies or intolerances, or had a risk of anaphylaxis.
In this parallel-group randomized clinical trial, individuals were randomly assigned to the OC or SysD intervention arm following baseline assessment. The study statistician (R.W.) allocated participants to either SysD or OC using a sequence of computer-generated random numbers. Treatment allocations were stored in opaque envelopes consecutively numbered by a research assistant independent of the study. These were opened sequentially as each child was eligible for intervention. Each participant was offered 10 treatment sessions in their allocated arm, and posttreatment outcomes were collected at a review session 3 months after the completion of intervention. Blinding of the therapist and the parents was not possible owing to the nature of the intervention provided. The same therapist provided all child-focused intervention sessions (J.M.).
Table 1 provides detailed information regarding the intervention arms. The OC arm was based on a prompt-and-reward style of therapy, in which the child was prompted (using verbal and/or visual prompts) to try foods outside their comfort level, and received rewards for doing so (verbal and object). Object reinforcements that were motivating to the child were used on a 1:1 schedule at the commencement of intervention, and “thinned” on a schedule as success was observed. The SysD arm was delivered as a play-based intervention that provided repeated exposure to goal foods through modeling and play, with no specific requirements for consumption. Task chaining and social reinforcement were used across both arms. All of the parents were asked to refrain from providing their child with food and drink (aside from water) in the 2 hours before each intervention session.
Parents had the option of requesting the intervention be provided in a weekly (10 sessions for 10 weeks) or intensive (10 sessions in 1 week) manner. It was not feasible to randomize intensity of treatment, owing to potential issues with family structure, parent work arrangements, and/or child schooling arrangements.
A tailored parent training program focused on feeding skills, behavior, and nutrition was delivered across both arms by a second therapist (parent educator) (P.D.). Parent training comprised 3 major features. First, parents were provided with and guided through written educational materials developed by the multidisciplinary team (speech-language pathology, occupational therapy, psychology, and nutrition) from national guidelines and existing best evidence. Second, the parent educator provided guided commentary for the parent while they observed the treating therapist completing intervention sessions with their child via linked cameras. Finally, every second session, the parent was immersed in therapy sessions with his/her child and encouraged to direct an increasing proportion of the intervention (eg, during session 2, they were present in the room providing verbal reinforcement; during session 4, they provided object reinforcement). Following this form of immersive training, the parent educator discussed session outcomes and provided troubleshooting advice in preparation for the next immersive session.
This work is presented as part of a larger study investigating different subgroups of children with feeding difficulties, the Healthy Eating Learning Program (HELP) Study. This trial had ethical approval from the Children's Health Services Queensland Human Research Ethics Committee (HREC/10/QREH/30), and The University of Queensland Medical Research Ethics Committee (Ref#2010000677).
Parents of participants completed a range of questionnaires and attended a face-to-face clinical assessment in an outpatient clinic. Measures collected at baseline only included medical history and demographic information, oral motor skills using a modified checklist (24), sensory processing using the Sensory Profile (25), and developmental level using the Parent Evaluation of Developmental Milestones—Assessment Version (PEDS-DM) (26). Primary outcome measures included a 3-day weighed food diary to measure dietary intake, food lists across food group categories to measure dietary variety, and the Behavioral Pediatrics Feeding Assessment Scale (BPFAS) (2) to measure mealtime behaviors. Secondary outcome measures included anthropometrical measures (weight, height, BMI), the Eyberg Child Behavior Inventory (ECBI) (27) to measure behaviors outside of mealtimes, and the Parenting Stress Index—Short Form (3rd edition) (PSI-SF) (28) to measure parent stress. Assessment measures were scored by the multidisciplinary team. Dietary analysis and review of Sensory Profiles and PSI-SF were completed by independent assessors, who were masked to treatment allocation and stage of analysis in the case of the food diary (ie, whether the food diary was completed pre- or posttreatment).
Fidelity to Treatment
A fidelity measure was developed and used to monitor compliance with the key features of the intervention protocols (Appendix 1, http://links.lww.com/MPG/A415). All sessions for all of the participants were rated to capture fidelity to treatment.
Features monitored for each participant included adherence to session dose recommendations, environmental considerations, and variety of foods offered (≥30 foods total). Features monitored that were specific to the OC arm included offering 3 foods per session, presenting an antecedent (verbal and/or visual prompt), providing a consequence for desired behavior (verbal or object reinforcement), “thinning” object reinforcement over the course of intervention, and providing a consequence for undesired behavior (withdrawal of attention, verbal redirection). Features modeled that were specific to the SysD arm included offering 10 foods per session, offering a range of textures during each session, linking foods within a session by sensory-motor properties, providing modeling of food play using a hierarchy of increasing exposure, and providing verbal reinforcement for desired behaviors. In the parent training, features monitored included the content of sessions (education regarding feeding skills, behavior, and nutrition) and the transition of the parent into intervention sessions.
To detect a clinically important between-group difference of 0.75 standard deviations, power calculations indicated that a sample size of ≥32 was required (with power = 0.8 and α = 0.05). Descriptive statistics are presented as a mean (±standard deviation) for continuous variables and frequency (percentage) for categorical variables. To calculate the between-group differences at follow-up, linear regression models were used with the intervention arm (ie, OC/SysD) included as the main effect, and the baseline score of the outcome of interest included as a model covariable. Including baseline scores in the model adjusts for the possibility of pregroup imbalance. The results are presented as adjusted mean differences (±95% confidence intervals [CIs]). A P value of <0.05 was considered to be statistically significant. Analyses were conducted at the individual patient level on an intention-to-treat basis.
In addition, the overall difference between the pre- and postintervention scores was calculated using a univariable linear regression model, with time (baseline/follow-up) entered as the main effect. The effect sizes were calculated for the pre–post test comparisons, where d ≥ 0.2 was considered to be a small effect size, d ≥ 0.5 was considered to be a medium effect size, and d ≥ 0.8 was considered to be a large effect size. To investigate the effect of etiological group (ASD vs NMC), and intensity (weekly vs intensive), separate linear regression models were constructed. Both models adjusted for the baseline score of the outcome of interest.
In total, 86 participants prospectively enrolled in the study, of which 78 were eligible to participate. Data are presented for 68 participants who had full datasets available for analysis (ASD = 33, NMC = 35; OC = 36, SysD = 32; intensive = 25, weekly = 43). Two children from the NMC group received an ASD diagnosis during the course of the study, and were reallotted to the ASD group before data analysis. In addition, in 2 patients, there were significant issues with therapy suitability, and these participants were treated according to protocols for the opposing intervention arm (Fig. 1).
Table 2 provides demographic and baseline characteristics of the groups. Between-group results for baseline characteristics indicated that the intervention groups were not different across most characteristics. The OC group, however, presented with significantly fewer foods consumed at baseline in the overall foods category (OC = 18.6 foods, SysD = 23.9 foods, P < 0.05), as well as in the carbohydrates (OC = 4.1 carbohydrate-rich foods, SysD = 6.5 carbohydrate-rich foods, p < 0.01) and proteins (OC = 3.3 protein-rich foods, SysD = 4.9 protein-rich foods, p < 0.05) categories. As expected, the ASD group presented with significantly more global developmental delays than the NMC group (ASD = 67%, NMC = 44%, p < 0.01). Of note, a significantly greater proportion of children in the intensive arm were receiving oral supplements, either in formula or in vitamin form, compared with the weekly arm (intensive = 64%, weekly = 28%, P < 0.01) (data not shown). Most children in the study presented with some degree of oral motor impairment (n = 56, 82%), or oral sensory sensitivity (n = 39/61, 64%), or both (n = 45/61, 74%).
With regard to differences in outcomes between the 2 intervention arms, after adjusting for differences at baseline, there was a trend toward greater increases in the total number of foods consumed (adjusted mean difference −3.3 foods, 95% CI −6.8 to 0.1, P = 0.06), and the total number of unprocessed fruits and vegetables consumed (adjusted mean difference −1.3 unprocessed fruits and vegetables, 95% CI −2.7 to 0, P = 0.05) in the OC arm compared with the SysD arm. There was also a trend toward greater reduction of difficult mealtime behaviors (Total Frequency Score [TFS] Child–adjusted mean difference 3.5, 95% CI −1.4 to 8.4, P = 0.15) in the OC arm when compared with the SysD arm. The only statistically significant difference between groups was a decrease in height z score in the OC arm compared with the SysD arm (adjusted mean difference 0.2, 95% CI 0–0.4, P = 0.01), but this was not considered clinically significant (Table 3). Decrease in height z score was reflected in trends toward greater changes to BMI z score with OC intervention.
ASD Versus NMC
Comparison of outcomes across the ASD and NMC groups revealed only modest differences in performance, none of which were statistically significant (Appendix 2, http://links.lww.com/MPG/A416). It was considered clinically relevant that the ASD group demonstrated more improvement in overall dietary quality (total areas where recommended daily intake/adequate intake was met) (adjusted mean difference 1.1 areas met, 95% CI −0.6 to 2.7), relative to the NMC group. In contrast, the NMC group made slightly more improvement to dietary variety (total proteins adjusted mean difference −1.1 protein-rich foods, 95% CI −2.7 to 0.4) compared with the ASD group.
Intensive Versus Weekly
There were no statistically significant differences in outcomes observed between weekly and intensive intervention models (Appendix 3, http://links.lww.com/MPG/A417). It was observed that greater reductions in difficult parent mealtime behaviors (BPFAS TFS-Parent) occurred in the weekly arm (adjusted mean difference −2.0, 95% CI −1.0 to 0.7) compared with the intensive arm. This trend was also observed in the number of behaviors reported to be a problem outside of mealtimes, wherein greater change was made in the weekly arm compared with the intensive arm (Eyberg Child Behavior Inventory Total Problem Score) (adjusted mean difference −1.9, 95% CI −4.6 to 0.7). Although not statistically significant, this was considered to be clinically relevant.
Overall Change From Baseline to 3-Month Follow-Up
Given the limited significant differences observed between intervention arms, etiological groups, and intensity levels, comparison of pre–post scores across all outcome measures was completed, with all of the participants included in the same group (Table 4). Favorable results were observed for all primary outcomes (all P < 0.05). Difficult mealtime behaviors and dietary variety were the variables wherein the largest effect sizes were demonstrated, indicating that most change had been achieved across these outcomes.
Fidelity to Treatment
Total fidelity to treatment was 94% across all of the participants. The median number of treatment sessions provided was 9 (of a possible 10). In some cases, in which therapy sessions were cancelled as a result of illness, it was not possible to provide the full 30 foods across therapy sessions, but 80% of participants received 30 foods or more, and the mean number of foods provided was 29.6 (±3.4). In addition, cancellations because of illness meant that some parents were unable to be fully transitioned into the role of the therapist. In all of these cases, however, the parent was providing some aspect of treatment (eg, reinforcement, offering food, modeling). Compliance was noted with all other features of intervention.
Favorable outcomes following a structured intervention block were achieved in this study, regardless of intervention type, etiological group, or intensity. Despite significant changes being observed across both intervention arms, OC appeared to have a somewhat greater effect than SysD on increases in dietary variety and reduction in difficult mealtime behaviors at 3 months postintervention. This may be because OC therapy demands adult-determined changes to child behavior and intake, which may occur more quickly (because they do not require the child to make internal motivational changes, as is required in SysD). It is well understood that improvements made in OC intervention may be followed by a relapse in behavior once the intervention has ceased (29). Therefore, review of participants >3 months after intervention may have provided further information about long-term intervention relative effects in the OC and SysD arms.
There were limited differences observed in response to intervention between participants with ASD and those with an NMC history. It is possible that underlying diagnosis is not as important in predicting intervention outcomes as other features of feeding difficulty, including oral motor impairment and oral sensory sensitivity, which were similar across both groups. Different features of feeding difficulty may, in fact, be more responsive to particular intervention approaches. Further research is required in refining subgroups for feeding difficulties in developing interventions.
Parent training was important in this study, but because of the study design it was impossible to know whether it was integral to intervention success, because every parent received the same training. In a structured, postintervention survey, many parents reported the opportunity to observe their child in a neutral environment outside of the therapy situation and receive guided commentary was invaluable to the experience (data not shown). In addition, the opportunity to participate in immersive therapy situations was observed to enhance generalizability of the intervention. We have found parents of children with clinically significant oral hypersensitivity also reported high frequencies of difficult behaviors at mealtimes. It is suggested that perhaps through education the parents’ perceptions of difficult mealtime behaviors were reshaped into understanding regarding oral sensory and oral motor problems, which led to increased patience and support for the child at mealtimes, rather than a reaction to “naughty” behavior. Altering parents’ perceptions of difficult mealtime behavior may have also alleviated some stress associated with mealtimes. It is suggested, therefore, that immersive parent training contributed to successful intervention outcomes, but its relative contribution to child-focused intervention needs to be further tested.
Although dietary analyses were completed without vitamin and mineral supplements to measure the participants’ intake from real food and drinks only, we did not require children to stop taking either vitamin/mineral or oral calorie supplements to participate in intervention. Our study protocol requested that parents refrain from offering food or drinks in the 2 hours before sessions, but this was difficult to monitor, and intake may have impacted appetite and/or internal regulation of dietary needs. This may, therefore, have had an effect on motivation to try new foods during the intervention block. Overreliance on milk/formula for energy supplementation has been found to be an appetite suppressant in typically developing children with picky eating behaviors (30). Many children in our study were supplementing energy intake with high-energy drinks such as milk or toddler formula. This was reflected in the data demonstrating differences in total energy intake with/without drinks (Table 2). The limited difference in performance between the intensive and weekly groups, in which significantly more participants were receiving supplements in the intensive group, may suggest that the provision of vitamin/mineral and/or oral calorie supplements had a minimal impact on outcomes. It may also suggest that supplementation limited further progress in the intensive group owing to suppressed appetite (30) or that participants in the intensive group had a more severe presentation. It was encouraging to note that improvements in the total percentage of energy intake met occurred with and without drinks included in analysis (Table 4). This suggests that the children in this study began eating more foods in response to intervention, rather than supplementing their intake further with drinks.
Another factor that was not possible to control without significantly affecting recruitment was prior access to intervention. Some children in the study had received clinical assessment and/or intervention before participating in the study. In addition, although it was requested that no other intervention be accessed during involvement with the study, 3 participants received some form of input between finishing intervention and attending their 3-month review. In 2 cases, this involved general classroom input at school, and in 1 case, this involved clinical assessment only. Exposure to external input may have biased responses to the questionnaires (because the parents may have had prior exposure to these), or falsely inflated outcomes, but in reality the numbers were small.
Given that participants were all in a healthy BMI range at the start of intervention, increases in BMI after treatment were of concern, given that one of the aims of feeding intervention is to improve long-term health outcomes. In terms of clinical relevance, although the change was not great, a small reduction in height z score was observed, which likely inflated BMI outcomes. It may be that some of the children in the study experienced a plateau in growth during the period of the study, which affected their height z scores. An alternate hypothesis, given that change to BMI was greatest in the ASD cohort, is that children with ASD added new foods to their diet, but continued consuming the same volumes of the old ones because of the often observed rigidity in routine in this group (31). Overgeneralization owing to rigidity is of urgent consideration in providing feeding intervention for children with ASD.
Although this study was a prospective randomized trial, with good compliance and limited attrition, as with any study, there are some limitations to be acknowledged. As a randomized clinical trial, this study lacked a control group, which would have provided information about whether improvement in dietary variety and mealtime behavior would have occurred without intervention. Many of the participants in this study, however, had feeding difficulties for some time before commencement of the study, so it is unlikely that these difficulties would have self-resolved for all of the children. In addition, given the high level of parental stress and long-term health impacts associated with feeding difficulties, it was unethical not to provide intervention. High parental concern also suggested that being randomized to a control group would have resulted in high levels of attrition, and many families may have sought assistance via other means, which would have potentially biased the results because of differential dropout. Another limitation of this study is the lack of objective measurement possible in a study that measures behavior and diet, predominantly through parent-reported outcomes. Wherever possible, bias was minimized through the use of masked assessors, but the influence of parent bias should not be discounted.
This study is the first known prospective, randomized clinical trial to compare dietary and behavioral outcomes for OC and SysD intervention in children with feeding difficulties. Comparisons were also made between children with ASD and children with an NMC history, and interventions offered in an intensive or weekly manner. The results demonstrated positive outcomes across all primary outcomes measured, regardless of randomization, intensity, or group. This suggests that, overall, intervention delivered by experienced therapists to a standardized protocol is effective in increasing dietary quality and variety and decreasing difficult behaviors at mealtimes. Further research is required in examining other cohorts of children with feeding difficulties and exploring outcomes after longer periods postintervention.
1. Dovey TM, Jordan C, Aldridge VK, et al. Screening for feeding disorders. Creating critical values using the Behavioral Pediatrics Feeding Assessment Scale. Appetite
2. Crist W, Napier-Phillips A. Mealtime behaviors of young children: a comparison of normative and clinical data. J Dev Behav Pediatr
3. Nicholls D, Bryant-Waugh R. Eating disorders of infancy and childhood: definition, symptomatology, epidemiology, and comorbidity. Child Adolesc Psychiatr N Am
4. Greer AJ, Gulotta CS, Masler EA, et al. Caregiver stress and outcomes of children with pediatric feeding disorders treated in an intensive interdisciplinary program. J Pediatr Psychol
5. Chatoor I. Feeding disorders in infants and toddlers: diagnosis and treatment. Child Adolesc Psychiatr N Am
6. Mascola AJ, Bryson SW, Agras WS. Picky eating during childhood: a longitudinal study to age 11 years. Eat Behav
7. Ledford JR, Gast DL. Feeding problems in children with autism spectrum disorders
: a review. Focus Autism Other Dev Disabl
8. Ramsay M, Gisel EG, Boutry M. Non-organic failure to thrive: growth failure secondary to feeding-skills disorder. Dev Med Child Neurol
9. Dovey TM, Farrow CV, Martin CI, et al. When does food refusal require professional intervention
? Curr Nutr Food Sci
10. Goh DY, Jacob A. Perception of picky eating among children in Singapore and its impact on caregivers: a questionnaire survey. Asia Pac Fam Med
11. Shields B, Wacogne I, Wright CM. Weight
faltering and failure to thrive in infancy and early childhood. BMJ
12. Dowdney L, Skuse D, Heptinstall E, et al. Growth retardation and developmental delay amongst inner-city children. J Child Psychol Psychiatry
13. Barker DJP. Early growth and cardiovascular disease. Arch Dis Child
14. Bjorge T, Engeland A, Tverdal A, et al. Body mass index in adolescence in relation to cause-specific mortality: a follow-up of 230 000 Norweigan adolescents. Am J Epidemiol
15. Goodell LS, Pierce MB, Bravo CM, et al. Parental perceptions of overweight during early childhood. Qual Health Res
16. Marshall J, Hill RJ, Dodrill P. A survey of practice for clinicians working with children with autism spectrum disorders
and feeding difficulties
. Int J Speech Lang Pathol
17. Sharp W, Jaquess D, Morton J, et al. Pediatric feeding disorders: a quantitative synthesis of treatment outcomes. Clin Child Fam Psychol Rev
18. Linscheid TR. Behavioral treatments for pediatric feeding disorders. Behav Modif
19. Kerwin ME. Empirically supported treatments in pediatric psychology: severe feeding problems. J Pediatr Psychol
20. Marshall J, Ware RS, Ziviani J, et al. Efficacy of interventions to improve feeding difficulties
in children with autism spectrum disorders
: a systematic review and meta-analysis. Child Care Health Dev
2014; [Epub ahead of print].
21. Field D, Garland M, Williams K. Correlates of childhood feeding problems. J Paediatr Child Health
22. Reau NR, Senturia YD, Lebailly SA, et al. Infant and toddler feeding patterns and problems: normative data and a new direction. J Dev Behav Pediatr
24. Therapy Skill Builders, Morris SE, Klein MD. Pre-Feeding Skills: A Comprehensive Resource for Mealtime Development. 2nd ed.2000.
25. Dunn W. Sensory Profile: Caregiver Questionnaire. San Antonio, TX: PsychCorp; 1999.
26. Glascoe FP, Robertshaw NS. Parents’ Evaluation of Developmental Status: Developmental Milestones. Victoria: Hawker Brownlow Education; 2009.
27. Eyberg S. ECBI: Eyberg Child Behavior Inventory. Lutz, FL: Psychological Assessment Resources; 1999.
28. Abidin R. Parenting Stress Index (PSI). Lutz, FL: Psychological Assessment Resources; 1995.
Sydney, Australia, Department of Health and Ageing, National Health and Medical Research Council. Australian Dietary Guidelines. 2006.
29. Bouton ME, Winterbauer NE, Todd TP. Relapse processes after the extinction of instrumental learning: renewal, resurgence, and reacquisition. Behav Processes
30. Wright CM, Parkinson KN, Shipton D, et al. How do toddler eating problems relate to their eating behavior, food preferences, and growth? Pediatrics
31. Hsu W-S, Ho M-H. Ritual behaviours of children with autism spectrum disorders
in Taiwan. J Intellect Dev Disabil