ORIGINAL www.jpeds.

com • THE JOURNAL OF

PEDIATRICS
ARTICLES

A Systematic Review and Meta-Analysis of Intensive

Multidisciplinary Intervention for Pediatric Feeding Disorders: How
Standard Is the Standard of Care?
William G. Sharp, PhD1,2, Valerie M. Volkert, PhD1,2, Lawrence Scahill, MSN, PhD1,2, Courtney E. McCracken, PhD1,
and Barbara McElhanon, MD1,2

Objective To assess models of care and conduct a meta-analysis of program outcomes for children receiving
intensive, multidisciplinary intervention for pediatric feeding disorders.
Study design We searched Medline, PsycINFO, and PubMed databases (2000-2015) in peer-reviewed jour-
nals for studies that examined the treatment of children with chronic food refusal receiving intervention at day
treat- ment or inpatient hospital programs. Inclusion criteria required the presentation of quantitative data on food
consumption, feeding behavior, and/or growth status before and after intervention. Effect size estimates were
calculated based on a meta-analysis of proportions.
Results The systematic search yielded 11 studies involving 593 patients. Nine articles presented outcomes based
on retrospective (nonrandomized) chart reviews; 2 studies involved randomized controlled trials. All samples in-
volved children with complex medical and/or developmental histories who displayed persistent feeding concerns
requiring formula supplementation. Behavioral intervention and tube weaning represented the most common
treat- ment approaches. Core disciplines overseeing care included psychology, nutrition, medicine, and
speech- language pathology/occupational therapy. The overall effect size for percentage of patients successfully
weaned from tube feeding was 71% (95% CI 54%-83%). Treatment gains endured following discharge, with 80%
of pa- tients (95% CI 66%-89%) weaned from tube feeding at last follow-up. Treatment also was associated with
in- creased oral intake, improved mealtime behaviors, and reduced parenting stress.
Conclusions Results indicate intensive, multidisciplinary treatment holds benefits for children with severe feeding
difficulties. Future research must address key methodological limitations to the extant literature, including im-
proved measurement, more comprehensive case definitions, and standardization/examination of treatment approach.
(J Pediatr 2017;181:116-24).

See editorial, p 7

P
ediatric feeding disorders involve severe disruptions in nutritional and caloric intake exceeding ordinary variations in
hunger, food preference, and/or interest in eating.1 Feeding problems of this magnitude affect as many as 5% of chil-
dren and represent one of the most frequent concerns in pediatric settings.2,3 Avoidant/restrictive intake disorder (ARFID),
the broader psychiatric diagnosis for feeding disorders, requires failure to meet nutrition and/or energy needs as the result of
avoidance or restriction of oral intake of food. 4 Clinical manifestations of ARFID include faltering growth, significant nutri-
tional deficiencies, and/or reliance on enteral feeding or oral nutritional supplementation to meet energy needs. Infants and
children with feeding disorders also may have impaired cognitive and emotional development, 5 compromised immune func-
tioning, and may require recurrent hospitalizations. 6 Severe feeding difficulties also contribute to parental stress, anxiety, and
depression, as well as fear of social stigmatization due to unconventional feeding practices. 6,7
Estimates suggest 40%-70% of children with chronic medical concerns (eg, congenital or acquired respiratory, cardiac, and
gastrointestinal [GI] problems) experience feeding difficulties. 3 These medical problems may promote conditioned food aver-
sion by pairing unpleasant consequences, such as pain, nausea, and/or fatigue, with eating.1,8 Resolution of underlying
medical concerns, however, may not improve oral intake because of persistent, disruptive mealtime behaviors (eg, intense
tantrums, tearful protests) aimed at avoiding contact with food.9 In response to these behaviors, caregivers may coax,
comfort, and/or

ARFID Avoidant/restrictive food intake

disorder BMI Body mass index From the 1Department of Pediatrics, Emory University School of Medicine, Atlanta, GA; and 2Marcus Autism Center, Atlanta, GA
ES Effect size The authors declare no conflicts of interest.
GI Gastrointestinal
NRS Nonrandomized studies 0022-3476/$ - see front matter. © 2016 Elsevier Inc. All rights reserved.

RCTs Randomized controlled

trials
116
 Volume 181 • February 2017

reprimand and then understandably remove food and end the possible indicators of the treatment approach/setting (eg, “mul-
meal. Consequently, the child learns to avoid food by tidisciplinary treatment,” “tube weaning”). In addition, we re-
engag- ing in disruptive behaviors. 1 As a result, meals
increasingly involve little or no consumption, and a vicious
cycle takes hold. Limited exposure to food circumvents key
sensory, develop- mental, physiological, and social
processes associated with eating, which further erodes an
already fragile parent-child mealtime relationship. Without
intervention, this cycle leads to continued refusal,
inadequate nutrition, and the need for artificial supports (eg,
tube feeding) to support growth.1
Expert consensus increasingly recognizes intensive multi-
disciplinary intervention at day hospital programs and inpa-
tient settings as the standard of care for children with
complex feeding problems.1,3 This level of support allows
monitoring for potential complications (eg, aspiration,
severe weight loss, and/or allergic reactions) associated with
the introduction of new food types and textures,
advancement of oral volumes, and reduction of enteral
nutrition among children with little or no experience eating.
Previous reviews consistently report positive effects
associated with multidisciplinary intervention.1,3,6 The evidence
base, however, primarily involves single-case re- search and
nonrandomized studies (NRS) with few random- ized
controlled trials (RCTs). Lukens and Silverman3 identified 13
studies (11 NRS and 2 RCTs) published during a 15-year
period (1998-2013). Ten of the 13 studies involved multidis-
ciplinary treatment at day treatment or inpatient hospital
pro- grams; all reported positive outcomes associated
with intervention. Support for intensive intervention,
however, was derived solely from NRS.
Despite provisional support for the treatment of feeding
dis- orders at inpatient and day treatment programs,
important questions remain regarding this method of
treatment deliv- ery. Notably, previous reviews focus on
behavioral1 and/or psy- chological intervention3 for ARFID
implemented in a range of settings (eg, outpatient, inpatient)
spanning various thera- peutic approaches (eg, parent
education groups, therapist- directed protocols). Research,
however, has yet to exclusively examine intensive
multidisciplinary intervention. The current review sought to
survey the medical literature regarding treat- ment of
pediatric feeding disorders at inpatient and day treat- ment
programs, summarize treatment models and outcome
measures, and evaluate the evidence with the use of both de-
scriptive and meta-analytic procedures.

Methods
Following the Preferred Reporting Items for Systematic
Reviews and Meta-Analyses statement, we searched Medline,
PsychINFO, and PubMed (January 2000 and December
2015) and con- ducted ancestral and online searches in peer-
reviewed, English- language journals for eligible studies,
cross-checking search results and removing overlapping
citations. The search pa- rameters (Table I; available at
www.jpeds.com) included com- binations of key words
regarding the patient population (eg, “feeding
disorder,”“failure to thrive”) paired systematically with
viewed references from identified articles and previous
systematic reviews.1,3
Selection criteria required articles meet the following
cri- teria: the sample involved a pediatric population (birth
to 18 years of age) with ARFID, as evidenced by
dependence on enteral feeding or oral nutritional formula
supplementation; the study evaluated multidisciplinary
intervention at a day treat- ment or inpatient hospital setting
on a group level (vs case report); intervention primarily
targeted improving the volume of solid food intake vs
concerns regarding dietary variety (ie, food selectivity);
and the study presented pre/postintervention data on food
consumption (eg, grams consumed, use of feeding tube),
feeding behavior (eg, acceptance of food), and/or growth
status either descriptively (eg, frequencies, percentages) or
sta- tistically (eg, P values, t scores). Given the recognized
lack of RCTs in the field,1,3 we included NRS and RCTs to
examine group effects over time. Finally, this review
excluded studies that investigated the treatment of eating
disorders (eg, an- orexia nervosa, bulimia nervosa), which
involve a different eti- ology and treatment approach. 4 Two
authors (W.S., V.V.) independently searched the literature,
reviewed and screened potential articles, and reached
consensus on final inclusion (Figure 1; available at
www.jpeds.com).
Data Extraction, Variables Coded, and Reliability
Data extraction involved a standardized protocol to code eli-
gible studies (available on request). Variables captured
during this process included study descriptors (eg, experimental
design, treatment setting), patient demographic variables (eg,
sample size, age, sex), treatment approach (eg, duration,
disciplines involved), outcomes measures, and summary of
findings. The research team independently double-coded all
data extracted during the review process. The double-entered
data allowed for the calculation of percent agreement. Coder
agreement was 89% (range 80%-99%), exceeding the 80%
acceptable stan- dard of agreement recommended during
quantitative synthe- sis of research.1 To further ensure the
accuracy, we reached consensus on all areas of discrepancy
highlighted during the inter-rater analysis.
Data Analyses
We first analyzed extracted data on a descriptive level, sum-
marizing patient characteristics, treatment techniques, con-
tributing disciplines, and reported outcomes. This level of data
analysis involved summary statistics (percentages, means)
to identify commonalities and differences in approach to care
and outcomes. We then calculated effect sizes (ES) for
outcome vari- ables reported in at least 6 studies according to
standards for systematic reviews.10,11 For these calculations,
we used means (SDs) or frequency (percentages), as
appropriate. When summary statistics were not available,
we attempted to contact the corresponding authors via e-
mail. The primary goal of the meta-analysis was to
determine the omnibus impact of inter- vention. Given the
preponderance of NRS, the analysis focused on the
magnitude of pre/postchange associated with inter- vention.
As a result, we only analyzed reported changes for chil- dren
exposed to intervention for RCTs. The small number of
117
THE JOURNAL OF PEDIATRICS • www.jpeds.com Volume 181

studies precluded analysis of potential moderators (eg, age, appropriate mealtime behaviors, bite persistence (aka,
sex, diagnosis, number of treatment elements). contingency contacting, escape
We entered and analyzed data using Comprehensive
Meta- Analysis v3 (Biostat, Englewood, New Jersey).
Because of the heterogeneity in the reported outcomes
(Cochran Q = 52.1 and 32.8; P < .001 for discharge and
follow-up, respectively), we used random effects models to
estimate the overall ES. We con- ducted a sensitivity analysis
by repeatedly calculating the overall ES with one study
omitted per iteration and comparing the results with the
overall study effect. The threat of publication bias used
Duval and Tweedie’s trim-and-fill method. This non-
parametric method estimates the number of “missing”
studies in a meta-analysis and then determines the effect
these missing studies would have had on the outcome. If the
added studies significantly change the result, then
publication bias is pos- sible. We also calculated the fail-safe
N, which estimates whether nonsignificant missing studies
would nullify the observed effect. The risk of bias and
methodological quality also considered the limitations of the
contributing studies. For NRS, we de- veloped a checklist
system based on established guidelines for conducting
retrospective chart reviews 12 informed by the
Strengthening the Reporting of Observational Studies in
Epi- demiology recommendations for cohort studies (Table II;
avail- able at www.jpeds.com).13-22 RCTs were evaluated
with the Cochrane risk of bias assessment tool.23

Results
The search yielded 11 articles from a pool of 2436 studies
(Figure 1). Of these 11 studies, 2 were RCTs and 9 involved
NRS (Table III).13-21,24,25 Eight studies occurred at treatment pro-
grams in the US, 2 from a program in the Netherlands, and 1
from Austria. Collectively, these studies included 593 chil-
dren (age range 15.7-48 months; 314 boys and 279 girls).
The majority of children (n = 535; 90.2%) were receiving
treat- ment for feeding tube dependence. Two studies included
a total of 22 children (3.7% of the sample) who relied on
liquid formula to meet nutritional needs.17,25 The remaining
36 sub- jects had various feeding problems but were not tube
or formula dependent.15,17 All studies reported a high
frequency of co- occurring medical problems such as
gastroesophageal reflux, failure to thrive, and other
unspecified GI concerns (eg, “disease of the gut,”20 “GI
abnormality”13,17).

Treatment Setting and Approach to Intervention

Eight studies delivered treatment during inpatient hospital-
ization, and 4 studies involved day treatment programs, with
Greer et al studying both inpatient and day treatment pro-
grams (Table IV).13-21,24,25 In both settings, treatment in-
volved multiple therapeutic meals per day overseen by a
multidisciplinary professional team. All programs included
psy- chologists, physicians (ie, general pediatrician or
pediatric gas- troenterologist), and nutritionists. Nine studies
also involved speech-language pathologists, 6 included
nurses, and 6 in- volved occupational therapists. Behavioral
intervention— which included positive reinforcement of
extinction), and/or stimulus fading—represented the most
common treatment approach. Three studies15-17 involved
be- havioral intervention and oral-motor therapy.
Descriptions of oral-motor exercises included activities
aimed at decreasing tactile hypersensitivity16 and/or
increasing the range, strength, and control of the lips,
cheeks, jaw, and tongue.13,15 Nutrition- ists most often
calculated energy needs,18 monitored hydra- tion, and
tracked advances in oral intake to adjust tube feeds.16
Physicians monitored vitals14 and/or managed chronic (eg,
reflux, constipation) and acute illness.13
Six studies reported on tube weaning, also called
“hunger provocation” or “appetite manipulation.” In this
approach, tube feeds are reduced below recommended daily
needs to promote hunger. Of the 6 studies, 2 studies14,19
combined behavioral intervention with tube weaning.
Brown et al13 combined be- havioral intervention, tube
weaning, nutrition education, and oral-motor exercises.
Three studies18,20,24 identified tube weaning as the primary
mechanism of action. Trabi et al20 combined tube weaning
with psychoanalytically oriented play therapy during
meals, nutritional counseling, and oral-motor therapy (eg,
tactile stimulation). The authors also described meals as
unstructured, prohibiting the use of behavioral approaches
(eg, bite persistence). The tube-weaning model described
by Kindermann et al18 and Hartdorff et al24 cited use of
some be- havioral techniques (ie, positive reinforcement)
while concur- rently emphasizing oral feeding was “not
forced.”
The 6 descriptions of tube weaning all involved an
initial restriction at admission followed by subsequent
118
reductions. In Trabi et al,20 tube feeding was reduced by 40%
on the first day of admission, 60% on day 2, and
discontinued entirely at the end of the first week. Brown et
al13 reported average re- duction of 73% on day 1, with
further reductions reported during the hospital stay.
Kinderman et al18 and Hartdorff et al24 reduced tube feeding
to 50% by day 2 and eliminated tube feeding by day 8.
Byars et al14 restricted tube feeding by more than 50% at the
onset of treatment with further reduction as oral intake
improved. Silverman et al19 noted supplemental feeding was
discontinued or significantly reduced during the admission.
Five of the 6 studies also cited precautionary medical
monitoring during tube weaning (eg, monitoring hydration
status, access to fluids as needed, daily assessment of
weight).13,14,18,19,24
All studies involved caregivers in treatment, although the
specificity of participation and training procedures varied. The
8 studies that used behavioral intervention reported formal care-
giver training procedures to promote generalization into the
home setting. This most often involved a sequential process
during which caregivers first observed intervention in an ad-
jacent observation room or by closed-circuit video, followed
by the transition of caregivers into meals under the guidance
of a feeding therapist.13-16,19,21 No study, however, provided spe-
cific data on caregivers’ acceptance, mastery, and adoption
of treatment strategies. The 3 descriptions of tube-weaning
in- terventions without behavioral intervention18,20,24 provided
less specificity regarding parent training procedures.
Kinderman et al18 and Hartdorff et al24 indicated parents
fed their children under the guidance of a nurse during the
second week
Sharp et al
A Fe
Syst br
emat ua
ic ry
Revi
ew 20
and 17
Table III. Summary of sample
Meta Study
-
characteristics
Clawson Cornwell Kindermann Hartdorff Silverman Williams
Anal et al15 et al16 Greer et et al18 et al24 Sharp et et al19 Trabi et et al21
ysis al17 al25 al20
of Brown
et al13 Byars et
Inten
al14
sive
Multi
disci Institution Children's
Cincinnati Children's Our Kennedy Emma Emma Marcus Children's Medical Penn State
plina Hospital of
Orange Children's Hospital Children's Krieger Children's Children's Autism Hospital of Univeristy Hershey
ry Hospital House at Institute Hospital Hospital Center Wisconsin of Graz Medical
Inter County
Medical Baylor Center
venti Center
on Location Orange, CA Cincinnati, Richmond, VA Dallas, TX Baltimore, Amsterdam, Amsterdam, Atlanta, GA Milwaukee, Graz, Austria Hershey, PA
for OH MD The The WI
Pedi Netherlands Netherlands
atric Design NRS NRS NRS NRS NRS NRS RCT RCT NRS NRS NRS
Feed Total (%)*
ing Sample size 30 9 8 40 121 10 21 10 77 221 46 593
Diso Sex, n (%)
Male 18 (60) 5 (55) 4 (50) 20 (50) 71 (58.7) 3 (30) 10 (48) 5 (50) 40 (52) 118 (53) 23 (50) 317 (53)
rders
Female 12 (40) 4 (45) 4 (50) 20 (50) 50 (41.3) 7 (70) 11 (52) 5 (50) 37 (48) 103 (47) 23 (50) 276 (47)
: Age, mo
How Median – – – – – – – – – – 37
Stan Mean 48 37.2 32 47.88 45.62 15.7 19.7 44.9 54 26.4 –
dard SD 16.8 14.4 13.92 16.29 29.70 – 5.4 19.2 26.4 18 –
Is Range 23-84 21.6-66 18-55 22-84 10-162 9-21 – – – 4.5-93 16-133
the Primary feeding concern Studies
Stan Tube dependence (n) X (30) X (9) X (4) X (40) X (72) X (10) X (21) X (5) X (77) X (221) X (46) 11 (82%)
dard Formula dependence (n) X (17) X (5) 2 (18%)
of Other/not specified X (4) X (32) 1 (9%)
Mean age of onset, mo 3 11.6 – – – – – – 10.8 – –
Care
Duration problem, mo 30 26.4 – – – 13.5 17.5 – 44.4 21 –
?
Previous intervention reported X X – – – X X – X X X 7 (64%)
Medical concerns, n (%) Participants
Cardio/pulmonary 9 (30) 4 (44) 5 (63) 3 (8) – 2 (20) 8 (38) 7 (70) 39 (51) 41 (19) 10 (22) 128 (27)
Failure to thrive – – 6 (75) – – – – 4 (40) – – 19 (41) 29 (47)
Food allergies – – – – – 3 (30) 1 (5) 1 (10) – – 7 (15) 12 (14) OR
Gastroesophageal reflux 23 (77) 9 (100) 5 (63) 10 (25) – 1 (20) 3 (14) 6 (60) – – 39 (85) 96 (55)
General GI problem 9 (30) 6 (66) 1 (13) – 84 (69) 1 (20) – – 71 (92) 46 (21) 11 (24) 229 (44) IGI
Prematurity 17 (57) – 7 (88) 24 (55) 24 (20) 3 (30) 7 (33) – – 78 (36) 6 (13) 142 (31) NA
DD/autism/neurologic 10 (33%) 3 (33) 8 (100) – 21 (17) – 4 (19) 3 (30) 52 (77) 18 (8.2) 20 (43) 136 (25)
L
DD, developmental delay; X, study characteristic or variable reported in study; –, study characteristic or variable not available/omitted in article.
*Percentage of studies or sample calculated based on contributing studies providing data. AR
TIC
LE
11
9
12
0
T
HE
JO
UR
NA
L
Table IV. Treatment setting and intervention characteristics
OF
Study
PE
Brown Byars Clawson Cornwell Greer Kindermann Hartdorff Sharp Silverman Trabi Williams
et al13 et al14 et al15 et al16 et al17 et al18 et al24 et al25 et al19 et al20 et al21 Total (%) DI
Setting AT
Inpatient X X X X X X X X 8 (73)
Day treatment
Treatment duration, d
X X X X 4 (36)
•RI
Mean (SD) 19 11.4 29 46.43 46.8 17 14.4 5 10.9 21.6 24 22.3 (13.7) w.j
Range
Contributing disciplines
5-16 15-80 9-26 2-52 8-45 pe
Gastroenterologist/physician X X X X X X X X X X X 11 (100) ds.
Nursing/nurse practitioner X X X X X X 6 (55)
Nutrition/dietician X X X X X X X X X X X 11 (100) co
Occupational therapist
Psychologist
X
X X X
X
X
X
X X X
X
X X
X
X
X
X
6 (55)
11 (100)
m
Speech-language pathologist X X X X X X X X X 9 (82)
Social worker X X 2 (18)
Intervention mechanism(s)
Behavioral intervention X X X X X X X X 8 (73)
Nutrition education X X 2 (18)
Oral-motor exercises X X X X X 5 (45)
Tube weaning X X X X X X 6 (55)
Behavioral elements
Contingency contacting/extinction X X X X X X X 7 (64)
Differential attention X X 2 (18)
Negative reinforcement X X X 3 (27)
Positive reinforcement X X X X X† X† X X X 9 (82)
Response cost X X X 3 (27)
Shaping/fading X X X X X 5 (45)
Not specified/used X‡ X§ 2 (18)
Caregiver training X X X X X X X X X X X 11 (100)
†Describes positive reinforcement of acceptance, but no formal recognition of behavioral intervention.
‡Describes general use of antecedent and consequent based procedures, but no additional detail provided.
§Indicates no formal structure to meals.

Sh Vo
ar lu
p m
et e
al 18
THE JOURNAL OF PEDIATRICS • www.jpeds.com Volume 181

of intervention. Trabi et al20 indicated that parents were reported improvements in mealtime interactions based on
“strictly told not to feed their child” but also noted that the parent-rated questionnaires.
team sup- ported caregivers in their feeding activities and
attempts.

Approach to Measurement
Across studies, the number and definition of outcome mea-
sures varied (Table V; available at www.jpeds.com).13-21,24,25
All 11 studies presented data reflecting changes in oral
consump- tion. Of these, 8 reported the percentage of
patients weaned from enteral feeds, 6 reported the
percentage of calories re- ceived by oral vs enteral feeds,
and 4 reported on the volume of food consumed (eg, grams,
kcal). All but one study 21 as- sessed changes in
anthropometric parameters, with out- comes including
weight (kg), percentage of ideal body weight, and body mass
index (BMI). Three studies involved direct ob- servation of
mealtime performance, which included tracking bite
acceptance, mouth clean (ie, a proxy measure for swal-
lowing), and disruptive behaviors (eg, pushing away the
spoon). Greer et al17 and Silverman et al19 assessed mealtime
difficul- ties and caregiver stress using parent-rated
questionnaires. Sharp et al25 included a post-treatment
satisfaction questionnaire. Nine studies included post-
treatment follow-up data (range 1-24 months; mean 9
months).

Summary of Treatment Outcomes

The percentage of patients weaned from enteral feeds
ranged from 43% to 100% (69.8% [21.6%]) across the 8
studies that reported this outcome (Table VI; available at
www.jpeds.com).13-21,24,25 At follow-up, 4 studies14,19-21 re-
ported additional gains in patients weaned; 3 studies18,24,25 re-
ported resumption of tube feeding in some patients. Six
studies reported improvement in oral consumption during
meals, ranging from 38% to 100% (74.5 [21.5]) following
interven- tion. Three studies13,14,19 reported continued
improvements in oral intake at follow-up; Kinderman et al18
and Hartdorff et al24 reported a slight decline. The 4 studies
reporting on calories (kcal) or grams indicated significant
improvements in the volume of food consumed.
Four studies15,17,24,25 that included behavioral intervention
without tube weaning reported stabilization or improvement
in weight. The 6 studies that involved tube weaning as a
primary treatment component reported weight loss at
discharge. Of these, 4 reported on the percentage of weight
loss, which ranged from 4% to 9.2%; Brown et al13 reported
a slight decline in BMI z score, and Byars et al14 noted a
decline in percentage of ideal body weight. Byars et al,14
Kindermann et al,18 and Hartdorff et al24 reported an
increase in weight (kg) at follow-up. Byars et al,14 however,
also reported a continued decline in the per- centage of ideal
body weight. Kindermann et al18 and Hartdorff et al24 did not
report on growth trajectory over time. Brown et al 13 reported
a continued decline in BMI at follow-up.
The 3 studies15,17,25 that included direct observation re-
ported significant improvements in bite acceptance and
swal- lowing, as well as decreased disruptive behaviors.
Greer et al17 and Silverman et al19 reported reduced
caregiver stress follow- ing intervention. Both studies also
February 2017
Overall Estimate of ES
The percentage of patients weaned from enteral feeds
was the only variable that met the 6-study threshold. Each
of the 9 quali- fying studies reported that all participants
were 100% tube de- pendent before treatment. We,
therefore, calculated ES based on the proportion of
patients weaned successfully from tube feeding at
discharge and last follow-up. Eight of the 9 pos- sible
contributing articles provided sufficient data to contrib- ute
to ES estimates at discharge; 7 permitted ES estimation
to post-treatment follow-up. The overall ES for percentage
of pa- tients weaned from tube feeding was 71% (95%
CI 54%- 83%), with 6 of the 8 studies reporting a
success rate of 50% or more patients weaned from tube
feeding (Figure 2, A). On the basis of sensitivity
analysis, no one study was overly in- fluential to the ES
(range 68%-75%). At last follow-up, all 7 studies
reported 50% or more patients were weaned from tube
feeding. The overall ES for percentage of patients weaned
from tube feeding at last follow-up was 80% (95% CI
66%-89%; Figure 2, B). Here again, sensitivity analysis
showed that no one study was overly influential in
calculation of the ES (range 74%-83%).
Analyses also suggested a low threat of publication bias
for the proportion of patients weaned at discharge. Duval
and Tweedie’s trim-and-fill method showed that one study
needed to be imputed to the left of the mean ES (ie, a tube
weaning rate < 0.71). After we imputed this one study, the
ES slightly decreased to 0.68 (95% CI 0.52-0.81). The fail-
safe N analy- sis suggested that 79 nonsignificant studies
would be needed to nullify the ES at discharge. Duval and
A Systematic Review and Meta-Analysis of Intensive Multidisciplinary Intervention for Pediatric Feeding Disorders:
How Standard Is the Standard of Care?
ORIGINAL
ARTICLES
Tweedie’s method sug- gested that 0 studies needed to be
imputed, and the fail-safe N showed that 146 nonsignificant
studies would be needed to nullify the ES at follow-up.
Quality Assessment
The rigor of the 9 NRS studies was variable. All involved
ret- rospective chart reviews to collect data (Table II). Data
ab- straction represented the greatest potential source of bias
for NRS. No studies used blinded data abstractors,
standardized data abstraction protocols, or presented inter-
rater reliability for the abstracted data. The management of
missing data also was not reported. The 2 RCTs provided
greater protection against bias through blinding of outcome
assessors and clear management strategies for missing data
(Table VII; available at www.jpeds.com).24,25 Together,
conclusions from the current review should be viewed as
conditional based on method- ological limitations of
available studies.
Discussion
Our findings corroborate conclusions from previous
single-subject1 and qualitative reviews3 indicating positive
out- comes associated with day treatment and inpatient
pro- grams. All identified studies reported improvement
in consumption following intervention. On average,
depen- dence on enteral feeds was eliminated in 71% of
children at discharge. When documented, these benefits
appear to persist, with 80% of patients tube-free at follow-
up. Treatment also
121
Figure 2. Meta-analysis of patients tube weaned. Outcomes at A, discharge and B, at follow-up.
promoted increased oral intake, improved mealtime behav-
iors, and reduced parenting stress if reported. Evidence was
largely derived from NRS; however, 2 recent RCTs lend
further credence to positive outcomes of these programs.
Together, the available evidence suggests intensive
multidisciplinary treat- ment likely holds benefits for
children with severe feeding dif- ficulties, particularly in
cases involving complex medical histories that cannot be
effectively managed in an outpatient setting.
Our study also provides guidance for standards of care
(Table VIII). For example, all programs included psychol-
ogy, nutrition, and medicine; most involved a speech-
language pathologist or occupational therapists. The
involvement of these specialty areas provides important
oversight and safeguards when designing intervention with
consideration to behav- ioral, dietary, medical, and oral-
motor concerns (respec- tively) that often contributed to
the development and maintenance of chronic food refusal.
Closer examination, however, of the relative contributions of
discipline-specific ap- proaches (eg, oral-motor exercises,
nutrition education) is war- ranted. Consistent with
previous reviews, 1,3,6 behavioral intervention remains the
most frequently documented and well- supported treatment
occurring at multidisciplinary pro- grams. Behavioral
packages often combined a formalized meal
122
structure with consequence-based procedures (eg, escape
ex- tinction; reinforcement) and antecedent manipulations
(eg, reduced bite volume, modified food texture) to promote
con- sumption. Most programs also involved caregiver
training pro- cedures to foster generalization to home and
community settings.
Our findings also raise important questions regarding current
practices and potential differences in approach to care. The
rela- tive contribution of aggressive tube weaning as a
standalone or adjunctive therapy to behavioral intervention
is uncer- tain. Although 3 studies reported improved oral
intake in the absence of behavioral intervention, systematic
evaluation of potential costs (eg, weight loss) and benefits
(eg, time to effect) is needed. Current practices involved
notable differences in the sequence, timing, and volume of
tube feed reduction. Greater specificity regarding the
target(s) of intervention and dis- charge criteria is
recommended. For example, tube-weaning programs appear
largely focused on complete weaning from enteral feeds
before discharge. Although this is an end point for
treatment, restructuring meals so children accept and
swallow bites with few concomitant problems behaviors
(eg, tantrums, gagging) also is critical. Ideally, a
comprehensive ap- proach to advancing oral intake balances
both of these im- portant end goals.
Sharp et al
 Table VIII. Summary of key findings for clinical and research activities
Recommendations for standard of care at intensive day and inpatient programs
1.Multidisciplinary intervention should involve, at a minimum, a professional team that includes psychology, medicine, nutrition, and speech-language
pathology/ occupational therapy.
Rationale: Offers the necessary oversight and clinical guidance needed to address the behavioral, organic, dietary, and oral-motor concerns ubiquitous to severe
feeding disorders.
2.Behavioral intervention is a central treatment element for increasing oral intake while concurrently addressing the mealtime difficulties that prohibit
consumption.
Rationale: Promotes individualized mealtime structure and data-driven decision making to address the operant function of food refusal with consideration to the possible
side effects associated with extinction procedures.1
3.Treatment includes the active participation and involvement of caregivers, ideally with systematic training to promote generalization into the home and
community settings.
Rationale: Programmed caregiver training provides the structure and support necessary to maintain and expand upon treatment gains following discharge.
4.Discharge planning involves a transition plan for outpatient follow-up as a step-down process from daily
intervention. Rationale: Follow-up services provide additional support and assurances that treatment gains will endure
following discharge.

Recommendations to enhance the evidence base

1.Tube weaning: Work toward standardization of approach, determine the relative contribution to treatment outcomes, and analyze the cost/benefits for use as
an adjunct or standalone intervention.
Justification: There is high variability in the implementation of tube weaning, and weight loss was reported by all programs that used this approach.
2.Outcomes measures: Increase uniformity of assessment methods, ideally involving a multimethod battery that assesses changes in mealtime behaviors,
consumption, growth, and social validity of treatment.
Justification: There is inconsistent reporting and use of definitions for program outcomes, unclear methods for statistical analysis, and unspecified plans for data
abstraction and management.
3.Case definitions: Better describe patient population, including patient characteristics, inclusion/exclusion criteria, symptom severity, and/or barriers to achieving
oral intake.
Justification: The similarity across patient populations remains unclear, which may impact outcomes in terms of overall response to intervention, level of pre/post
improvements, and selection of treatment elements.
4.Better document durability of treatment: Include detailed follow-up data on key outcome measures, preferably providing multiple points in time to assess both
short- term and long-term progress following discharge.
Justification: Studies varied in the length of time (ranging from 1 to 24 mo) and number of time points (between 1 and 4 data points) when
documenting treatment durability, with follow-up absent in 2 studies.
5.Development of intervention manuals: Develop and evaluate treatment manuals or guidelines, ideally involving uniform procedures with built-in flexibility to tailor
the intervention to individual children.
Justification: This is a necessary prerequisite to replicating and evaluating treatment across settings, increasing access to care, and determining means to reduce the
overall length and cost of intervention.

The evidence base also would benefit from increased uni-
formity in the reporting of outcomes, ideally involving a
multimethod assessment battery capturing changes in meal-
time behaviors, oral intake, and anthropometric variables, as
well as caregiver satisfaction. More consistent reporting of
follow-up data also is needed to assess the durability of
treat- ment over time. Improved measurement also should
entail better characterization of patients at baseline,
including clarity regarding medical and/or behavioral
barriers to achieving oral intake. It is unclear, for example,
why samples differed in the level of oral intake at admission
—with estimates ranging from 0% to 30%. Without
additional data regarding actual meal- time performance, it
is difficult to determine whether admis- sion to programs
was driven by medical complexity (eg, risk of aspiration),
recent medical clearance to introduce food (eg,
passing a swallow study), and/or significant behavioral barri-
ers (eg, active and persistent refusal behaviors). Given the
need for better patient characterization, more uniformity in
outcome measurement and unanswered questions on the
necessary com- ponents of treatment, these 11 studies
prohibit definitive con- clusions regarding optimal models of
care.
Encouragingly, there appears to be some movement in the
field to address previous limitations highlighted by the
extant literature.1,3 The RCTs by Sharp et al25 and Hartdorff
velopment of standardized treatment approaches and
manuals represents a necessary prerequisite to replicating
and evalu- ating treatment across settings.25 With
converging evidence highlighting the benefits of intensive
multidisciplinary inter- vention for children with chronic
and severe food refusal, more systematic evaluation of
different treatment approaches and adjuncts to behavioral
intervention and/or tube weaning is war- ranted. In
addition to a continued call for growth of RCTs in this
field, NRS will continue to represent a valuable tool for
documenting treatment outcomes and providing
important insights into current models of care.
Researchers are encour- aged to enhance the
methodological rigor of both NRS and RCTs, taking
advantage of existing guidelines when planning and
conducting studies focusing on this treatment
environment.10,22,23 ■
et al24 reflect movement towards standardizing treatments at
day hos- pital programs and inpatient settings (respectively).
The de-
Submitted for publication Jun 15, 2016; last revision received Sep 22, 2016;
accepted Oct 3, 2016
Reprint requests: William G. Sharp, PhD, Pediatric Feeding Disorders
Program, The Marcus Autism Center, 1920 Briarcliff Rd, Atlanta, GA
30329. E-mail: wgsharp@emory.edu
A Systematic Review and Meta-Analysis of Intensive Multidisciplinary Intervention for Pediatric Feeding Disorders:
How Standard Is the Standard of Care?
2. Satter E. The feeding relationship: problems and interventions. J
Pediatr 1990;117:S181-9.
3. Lukens CT, Silverman AH. Systematic review of psychological
interven- tions for pediatric feeding problems. J Pediatr Psychol
2014;38:903-17.
4. American Psychiatric Association. Diagnostic and statistical manual of
mental disorders. 5th ed. Washington (DC): American Psychiatric Pub-
lishing; 2013.
5. Silverman AH, Tarbell S. Feeding and vomiting problems in pediatric
popu- lations. In: Roberts MC, Steele RG, eds. Handbook of pediatric
psychol- ogy. 4th ed. New York (NY): Guilford Press; 2009, p. 429-55.
6. Volkert VM, Piazza CC. Pediatric feeding disorders. In: Sturmey P,
Hersen M, eds. The handbook of evidence based practice in clinical
psychology, volume 1, child and adolescent disorders. New York (NY):
John Wiley & Sons, Ltd; 2012, p. 456-81.
7. Garro A, Thurman SK, Kerwin ME, Ducette JP. Parent/caregiver stress
during pediatric hospitalization for chronic feeding problems. J Pediatr
Nurs 2005;20:268-75.
8. Hyman P. Gastroesophageal reflux: one reason why baby won’t eat. J
Pediatr 1994;125:S103-9.
9. Piazza CC, Fisher WW, Brown KA, Shore BA, Patel MR, Katz RM, et
al. Functional analysis of inappropriate mealtime behaviors. J Appl
Behav Anal 2003;36:187-204.
10. Fu R, Gartlehner G, Grant M, Shamliyan T, Sedrakyan A, Wilt TJ, et
al. Conducting quantitative synthesis when comparing medical
interven- tions: AHRQ and the Effective HealthCare Program. J Clin
Epidemiol 2011;64:1187-97.
11. Higgins J. Cochrane Handbook for Systematic Reviews of Interven-
tions. www.cochrane.org/resources/handbook/. Accessed February 8,
2016.
12. Gearing RE, Mian IA, Barber J, Ickowicz A. A methodology for
conduct- ing retrospective chart review research in child and
adolescent psychia- try. J Can Acad Child Adolesc Psychiatry
2006;15:126-34.
13. Brown J, Kim C, Lim A, Brown S, Desai H, Volker L, et al. Successful
gas- trostomy tube weaning program using an intensive multidisciplinary
team approach. J Pediatr Gastroenterol Nutr 2014;58:743-9.
14. Byars KC, Burklow KA, Ferguson K, O’Flaherty TO, Santoro K, Kaul
A. A multicomponent behavioral program for oral aversion in children
de- pendent on gastrostomy feedings. J Pediatr Gastroenterol Nutr
2003;37:437- 80.
15.
References
1. Sharp WG, Jaquess DL, Morton JS, Herzinger C. Pediatric feeding dis-
orders: a quantitative synthesis of treatment outcomes. Clin Child Fam
Psychol Rev 2011;13:348-65.
123
Clawson EP, Kuchinski KS, Bach R. Use of behavioral interventions and
parent education to address feeding difficulties in young children with
spastic diplegic cerebral palsy. Neurorehabilitation 2007;22:397- 406.
16. Cornwell SL, Kelly K, Austin L. Pediatric feeding disorders:
effectiveness of multidisciplinary inpatient treatment of gastrostomy-
tube depen- dent children. Child Health Care 2010;39:214-31.
17. Greer AJ, Gulotta CS, Masler EA, Laud RB. Caregiver stress and out-
comes of children with pediatric feeding disorders treated in an inten-
sive interdisciplinary program. J Pediatr Psychol 2009;33:520-36.
18. Kindermann A, Kneepkens CMF, Stok A, van Dijk EM, Engels M,
Douwes AC. Discontinuation of tube feeding in young children by
hunger provo- cation. J Pediatr Gastroenterol Nutr 2008;47:87-91.
19. Silverman AH, Kirby M, Clifford LM, Fischer E, Berlin KS, Rudolph
CD, et al. Nutritional and psychosocial outcomes of gastrostomy
tube- dependent children completing an intensive inpatient behavioral
treat- ment program. J Pediatr Gastroenterol Nutr 2013;57:668-72.
20. Trabi T, Dunitz-Scheer M, Kratky E, Beckenbach H, Scheer P. Inpatient
tube weaning in children with long-term feeding tube dependency: a
ret- rospective analysis. Inf Mental Hlth J 2010;31:664-81.
21. Williams KE, Riegel K, Gibbons B, Field DG. Intensive behavioral
treat- ment for severe feeding problems: a cost-effective alternative to
tube feedings? J Dev Phys Disabil 2007;19:227-35.
22. von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC,
Vandenbroucke JP. Strengthening the Reporting of Observational
Studies in Epidemiol- ogy (STROBE) statement: guidelines for
reporting observational studies. J Clin Epidemiol 2008;61:344-9.
23. Higgins JP, Altman DG, Gotzsche PC, Juni P, Moher D, Oxman AD.
The Cochrane Collaboration’s tool for assessing risk of bias in
randomized trials. BMJ 2011;343:d5928.
24. Hartdorff CM, Kneepkens CM, Stok-Akerboom AM, van Dijk-Lokkart
EM, Engels MA, Kindermann A. Clinical tube weaning supported by
hunger provocation in fully-tube-fed children. J Pediatr Gastroenterol Nutr
2015;60:538-43.
25. Sharp WG, Stubbs KH, Adams H, Wells BM, Lesack RS, Criado KK,
et al. Intensive manual-based intervention for pediatric feeding disorders:
results from a randomized pilot trial. J Pediatr Gastroenterol Nutr
2016;62:658- 63.
124 Sharp et al
 Table I. Feeding and intervention key words used in da- tabase search
Feeding disorder search terms
Avoidant restrictive food intake disorder
Failure to thrive
Feeding disorder
Food phobia
Food refusal
Pediatric feeding disorders
Tube-dependent children
Intervention search terms
Behavioral intervention
Behavioral treatment
Intervention
Multidisciplinary treatment
Treatment
Tube weaning

Figure 1. Flow diagram of included and excluded studies.

 124.e1
A Systematic Review and Meta-Analysis of Intensive Multidisciplinary Intervention for Pediatric Feeding Disorders:
How Standard Is the Standard of Care?
 Table II. Quality assessment of NRS studies in the systematic review
BrownByarsClawsonCornwellGreerKindermannSilvermanTrabiWilliams et alet alet alet alet alet alet alet ale

Background and objectives

Clearly articulated research hypothesis Case selection X X X
Defined inclusion/exclusion criteria Described sampling time frame
Data abstraction XX XX XX XX XX X XX
Blinded data collection procedures Standardized data abstraction protocol Reliability check for abstracted data
Data management
Established rules for missing data Defined outcome measures Described data analysis
Confidentiality
Local ethic board approval

XX X XX XX X
X

X X X X

Table V. Summary of measurement approach

Study
Brown Byars Clawson Cornwell Greer Kindermann Hartdorff Sharp Silverman Trabi Williams Total
et al13 et al14 et al15 et al16 et al17 et al18 et al24 et al25 et al19 et al20 et al21 (%)
Consumption X X X X X X X X X X X 11 (100)
Calories/grams consumed x x x x 4 (36)
% oral intake x x x x x x 6 (54)
% patients weaned x x x x x x x x 8 (73)
Anthropometrics X X X X X X X X X X 10 (91)
BMI (z score; percentile) x x 2 (18)
Percent ideal body weight x x x 3 (27)
Weight, kg x x x x x x 6 (54)
Weight for height percentile x 1 (9)
Weight percentile x 1 (9)
Percent weight loss x x x x 4 (36)
Direct mealtime observation X X X 3 (27)
Bites accepted x x x 3 (27)
Mouth clean x x 2 (18)
Inappropriate behaviors x x 3 (27)
Negative vocalizations x 1 (9)
Caregiver satisfaction X 1 (9)
Parenting Stress Index X X 2 (18)
Standardized feeding X X 2 (18)
questionnaire
Discharge follow-up X X X X X X X X X 9 (82)
Time frame, mo 4, 12 3 1, 4, 7, 12 3, 6 1, 3, 6 1 12 7.5 12, 24

124. e2
Sharp et al
A Fe
Syst br
emat Table VI. Pre/post outcomes by study ua
ic Cornwell Trabi Williams
Brown et al16 Greer et al17 Kindermann et al18 Hartdorff et al24 Sharp et al25 Silverman et et al20 et al21 ry
Revi et al13 Byars et al14 Clawson et al19 20
ew al15
and 17
Meta Consumption
- % patients weaned*
Anal Discharge 90% 44% – 43% – 100% 86% – 51% 81.4% 67%
ysis Follow-up 83% 67% – – – 80% 86% – 63% 91.8% 74%
of Percent oral intake
Admission 22% 14.6% 10% – – 0%† 0%† – 30% – –
Inten Discharge 82% 63.4% 37.6% – – 100% 88% – 82% – –
sive Follow-up 92% 88.1% – – – 92.5% – – 85% – –
Multi Calories/grams
disci Admission – – a. 48.74 542.21 353.79 – – 4 – – –
plina b. 246.54
ry Discharge – – a. 89.51 897.83 898.12 – – 34 – – –
b. 486.95
Inter Follow-up – – – – – – 71 – – –
venti Metric a.grams kcal kcal grams
on b.kcal
for Anthropometrics
Pedi Admission −0.96 a. 96.4% a. 10.5 13.5 14.52 a. – a. 10.5 31.7% a. 98% – –
atric b. 13.1 b. 86.33% b. 9.19
c. 0.68%
Feed Discharge −0.97 a. 92.7% a. 11.0 13.56 15.24 a. 8.39 a. 9.7 57.5% a. 96% 4.32% –
ing b. 12.7 b. 89.59% b. – b. 7.4% b. 4%
Diso c. 1.76% c. 9.2%
rders Follow-up −1.01 a. 91.2% a. 13.1 – – a. – a. 10.1 – – –
: b. 13.3 b. 92.15% b. 13.6
c. 10.28%
How Metric BMI z score a. % ideal body a. weight - kg weight - weight - kg a. weight – kg a. weight – kg BMI/age %tile a. % idea body weight % weight
Stan weight kg
dard b. weight - kg b.% ideal body weight b. weight/height %tile b. % weight b. % weight loss loss
c.weight - %tile c.% loss
Is weight
the loss
Stan Other
dard Admission – – a. 51.88 – a. 27.49 – – a. 11.1% a. 75.36 – –
of b. 57.63 b. 31.55 b. 81.3% b. 80.59
c. 69.13 c. 24.57 c. 38.82
Care d. 81.1
? e. 106.68
Discharge – – a. 92.00 – a. 88.40 – – a. a. 71.82 – –
b. 91.75 b. 84.46 100% b. 59.27
c. 30.88 c. 6.17 b. 30% c. 28.56
d. 76.3
e. 101.81 OR
Follow-up – – a. –
b.–
– a. –
b.–
– – – – – – IGI
c.– c.–
d.PSIBites
total accepted Eating
NA
Metric a.% Bites accepted a.% a.% Bites a.PSI total
b.%behavior
Mouth clean b.%e.CEBI
Mouth clean
total
accepted b.MBQ: Behavior problems L
AR
AYCE, About Your Child's Eating; CEBI, Children's Eating Behavior Inventory; MBQ, Mealtime Behavior Questionnaire; PSI, Parenting Stress
TIC
12
Index. Follow-up based on last reported point of contact; –, study characteristic or variable not available/omitted in article.
*All patients dependent on feeding tubes at admission.
LE
4. †Patients described as “fully dependent” on tube feeding.
e3
THE JOURNAL OF PEDIATRICS • www.jpeds.com Volume 181

Table VII. Cochrane assessment of risk of bias of RCTs by study

Hartdorf et alSharp et al
Random sequence generation?+
Allocation concealment?+
Blinding++
Incomplete outcome data++
Selective reporting?+
Other bias+?

+, low-risk bias; −, high-risk bias; ?, unclear.

124.e4 Sharp et al