Journal of Adolescent Health 53 (2013) 573e578

www.jahonline.org
Original article

Higher Caloric Intake in Hospitalized Adolescents With Anorexia Nervosa

Is Associated With Reduced Length of Stay and No Increased Rate of
Refeeding Syndrome
Neville H. Golden, M.D. a, *, Casey Keane-Miller, R.D. b, Kristin L. Sainani, Ph.D. c,
and Cynthia J. Kapphahn, M.D., M.P.H. a
a

Division of Adolescent Medicine, Stanford University School of Medicine, Stanford, California

Department of Clinical Nutrition, Lucile Packard Childrens Hospital at Stanford, Palo Alto, California
c
Department of Health Research and PolicyeEpidemiology, Stanford University School of Medicine, Stanford, California
b

Article history: Received January 11, 2013; Accepted May 25, 2013
Keywords: Anorexia nervosa; Refeeding syndrome; Hypophosphatemia; Hypomagnesemia; Hypokalemia

See Related Editorial p. 555

A B S T R A C T

Purpose: To determine the effect of higher caloric intake on weight gain, length of stay (LOS), and
incidence of hypophosphatemia, hypomagnesemia, and hypokalemia in adolescents hospitalized
with anorexia nervosa.
Methods: Electronic medical records of all subjects 10e21 years of age with anorexia nervosa,
rst admitted to a tertiary childrens hospital from Jan 2007 to Dec 2011, were retrospectively
reviewed. Demographic factors, anthropometric measures, incidence of hypophosphatemia
(
3.0 mg/dL), hypomagnesemia (
1.7 mg/dL), and hypokalemia (
3.5 mEq/L), and daily change
in percent median body mass index (BMI) (%mBMI) from baseline were recorded. Subjects
started on higher-calorie diets (1,400 kcal/d) were compared with those started on lowercalorie diets (<1,400 kcal/d).
Results: A total of 310 subjects met eligibility criteria (age, 16.1  2.3 years; 88.4% female, 78.5 
8.3 %mBMI), including 88 in the lower-calorie group (1,163  107 kcal/d; range, 720e1,320 kcal/d)
and 222 in the higher-calorie group (1,557  265 kcal/d; range, 1,400e2,800 kcal/d). Neither group
had initial weight loss. The %mBMI increased signicantly (p < .001) from baseline by day 1 in the
higher-calorie group and day 2 in the lower-calorie group. Compared with the lower-calorie group,
the higher-calorie group had reduced LOS (13.0  7.3 days versus 16.6  9.0 days; p < .0001), but
the groups did not differ in rate of change in %mBMI (p .50) or rates of hypophosphatemia
(p .49), hypomagnesemia (p 1.0), or hypokalemia (p .35). Hypophosphatemia was associated
with %mBMI on admission (p .004) but not caloric intake (p .14).
Conclusions: A higher caloric diet on admission is associated with reduced LOS, but not increased
rate of weight gain or rates of hypophosphatemia, hypomagnesemia, or hypokalemia. Refeeding
hypophosphatemia depends on the degree of malnutrition but not prescribed caloric intake,
within the range studied.
2013 Society for Adolescent Health and Medicine. All rights reserved.

* Address correspondence to: Neville H. Golden, M.D., Division of Adolescent

Medicine, Stanford University School of Medicine, 770 Welch Road, Suite 433,
Stanford, CA 94304.
E-mail address: ngolden@stanford.edu (N.H. Golden).
1054-139X/$ e see front matter 2013 Society for Adolescent Health and Medicine. All rights reserved.
http://dx.doi.org/10.1016/j.jadohealth.2013.05.014

IMPLICATIONS AND
CONTRIBUTION

In adolescents hospitalized with anorexia nervosa, commencing refeeding

with 1,400e2,000 kcal/d is
associated with reduced
length of stay and no
increased rates of hypophosphatemia, hypomagnesemia, or hypokalemia,
compared with starting on
<1,400 kcal/d. Refeeding
hypophosphatemia depends
on the degree of malnutrition but not prescribed
caloric intake. Concerns
about refeeding syndrome
should not limit amount of
calories prescribed, within
the range studied.

574

N.H. Golden et al. / Journal of Adolescent Health 53 (2013) 573e578

Hospitalized adolescents with anorexia nervosa (AN) are

frequently started on low-calorie diets, based on national and
international guidelines, because of concerns about refeeding
syndrome [1e3]. The refeeding syndrome is a life-threatening
constellation of clinical and metabolic changes induced by
refeeding a malnourished patient, and is most likely to occur
during the rst week of refeeding. Those <70% of expected body
weight are at particular risk [4,5]. The clinical features of the
refeeding syndrome reect cardiovascular, hematologic, respiratory, and neuromuscular compromise including arrhythmias,
cardiac failure, hemolytic anemia, acute respiratory failure,
seizures, coma, and sudden death [6,7]. Multiple case reports have
described refeeding syndrome in AN [4,8e15]. Hypophosphatemia
is the biochemical hallmark of the syndrome [4,6,7], but hypomagnesemia and hypokalemia may also contribute to the clinical
consequences. Early detection and correction of these electrolyte
abnormalities can help avert the full clinical syndrome [5,16,17].
Weight gain is an important early component of the treatment
of AN, and is necessary to reverse the medical complications and
enable effective psychological intervention. Hypocaloric diets can
be associated with initial weight loss and prolonged length of stay
(LOS). Garber et al. [18] recently found that starting patients on
a diet of 1,200 kcal/d was associated with initial weight loss in 83%
of subjects and signicant weight gain was achieved only after
8 days of hospitalization. They found that prescribing higher
calories at baseline was signicantly associated with faster weight
gain and shorter hospitalization. By starting low and going slow,
both the amount and the rapidity of weight gain may be
compromised.
Whereas initial weight loss during nutritional rehabilitation
of malnourished patients is well known [19,20], the safety of
more aggressive refeeding protocols has not been well studied.
The risk of developing the refeeding syndrome needs to be
balanced against the benets of more rapid weight gain. Before
national recommendations are revised, we need scientic
evidence to demonstrate the safety and efcacy of more
aggressive protocols. Such studies should include a large number
of participants, and should specically include subjects at highest risk for developing the refeeding syndrome: those <70%
expected body weight.
The aim of the present study was to determine the effect of
higher caloric intake on daily weight gain, LOS, and incidence of
hypophosphatemia, hypomagnesemia, and hypokalemia in a large
sample of hospitalized adolescents with AN, admitted to a tertiary
childrens hospital for medical instability. We hypothesized that
higher caloric intake would increase rate of weight gain and
reduce LOS, and would not be associated with increased incidence
of hypophosphatemia, hypomagnesemia, or hypokalemia.
Methods
Study population
Eligible subjects were adolescents aged 10e21 years with AN
admitted to Lucile Packard Childrens Hospital inpatient eating
disorders unit for medical stabilization, whose rst admission
occurred between January 2007 and December 2011. A child
psychiatrist made the diagnosis of AN on presentation by
according to the Diagnostic and Statistical Manual of Mental
Disorders, 4th edition [21]. Criteria for admission to the unit
include one or more of the following: severe malnutrition
(<75% expected body weight), vital sign instability, (heart rate

<50 beats/minute during the day or 45 beats/minute at night),

hypotension <90/45 mm Hg, hypothermia <36.3 C, orthostatic
changes in pulse and blood pressure, dehydration, electrocardiographic abnormalities, or electrolyte disturbances, according
to published national guidelines [1,22,23]. Patients with bulimia
nervosa or eating disorder not otherwise specied were
excluded. For subjects who were readmitted during the study
period, only the rst admission was included, to ensure that the
response to refeeding analyzed in this study would not reect
prior medical interventions. Subjects who were transferred
from another facility after nutritional rehabilitation had already
been initiated, those who signed out against medical advice
before they were medically stable, and those requiring nasogastric feeding were excluded.
Study design
Electronic medical records of eligible subjects with AN were
retrospectively reviewed. Duration of illness, rate of weight loss,
demographic factors (age, race, and sex), anthropometric measures
(height and daily weight), and the presence or absence of hypophosphatemia (
3.0 mg/dL), hypomagnesemia (
1.7 mg/dL), and
hypokalemia (
3.5mEq/L) were recorded. Duration of illness was
ascertained from the history and physical performed by the
admitting physician using a standardized template. One investigator (C.K.M.), a registered dietitian, reviewed the recorded
24-hour diet recall obtained by the unit dietitian in an interview
with the patient within 24 hours of admission. Prescribed caloric
intake on the rst day of hospitalization and on the day of
discharge were reviewed and recorded by C.K.M.
Body mass index (BMI) was calculated using the formula:
BMI weight in kilograms divided by the square of height in
meters. Median BMI (mBMI), the 50th percentile BMI for exact
age, was determined using the sex-specic 2000 Centers for
Disease Control and Prevention BMI-for-age growth charts for
children and adolescents aged 2e20 years (http://www.cdc.gov/
growthcharts). Percent median BMI (%mBMI) was calculated by
dividing daily BMI by median BMI  100.
Baseline weight was dened as the weight obtained at
0600 hours on the rst full day after admission. For each subject,
change in %mBMI was calculated for each day of hospitalization
from baseline. Length of stay was dened as the number of days
from admission to discharge. Rate of change of %mBMI was
calculated by dividing the total percent change in mBMI by the
number of days hospitalized. Subjects started on a higher-calorie
diet (1,400 kcal/d) were compared with those started on
lower-calorie diets (<1,400 kcal/d). The reason for selecting this
cutoff is that most protocols based on the belief that caloric
content predisposes to refeeding syndrome will start patients on
<1,400 kcal/d.
The Stanford University Human Subjects Research Committee
reviewed the protocol. A waiver of informed consent and a
Health Insurance Portability and Accountability Actecompliant
waiver of individual authorization were granted. The Stanford
University Institutional Review Board approved data collection
protocols.
Protocol for refeeding and monitoring of daily weight and
electrolytes on our unit
Weight is obtained daily at 0600 hours post-void in a hospital
gown only. Three meals and two snacks are prescribed. All meals

N.H. Golden et al. / Journal of Adolescent Health 53 (2013) 573e578

and snacks are supervised by a staff member and any missed

portions are replaced with a high-calorie liquid supplement.
Before 2008, because of concerns about refeeding syndrome,
patients were usually started on 1,000e1,200 kcal/d depending
on their 24-hour diet recall on the day before admission. Since
2008, initial caloric prescriptions have been increased to
1,400e2,000 kcal/d with no other major changes to the program.
Initial caloric prescription is decided by individual physicians on
admission, based on their assessment of the degree of malnutrition and caloric intake preceding admission. Prescribed calories are increased by approximately 200 kcal every 24e48 hours,
based on a daily weight gain goal of .2e.5 kg/d. Patients receive
a balanced diet composed of approximately 40%e50% carbohydrate, 25%e30% protein, and 25%e30% fat, supplemented by
a high-calorie liquid supplement consisting of 60% carbohydrate,
15% protein, and 25% fat. In a minority of cases, in which patients
either refuse to eat at all or are unable to achieve adequate
calories orally, nasogastric feeding is utilized in addition to oral
food. Serum chemistries including magnesium and phosphorus
are drawn on admission, and thereafter at 0600 hours every
24e48 hours for the rst week, and afterward as clinically
indicated. If phosphorus is
3.0 mg/dL, sodium potassium
phosphate is administered, 250 mg/packet (8 mmol phosphorus
and 7.1 mEq potassium), one to two packets three times a day by
mouth. Some providers supplement phosphorus when levels
approach 3.0 mg/dL (3.1e3.5 mg/dL) in severely malnourished
patients in whom phosphorus levels are rapidly dropping.
For hypomagnesemia (
1.7 mg/dL), magnesium is supplemented as magnesium amino acids chelate (133 mg elemental
magnesium per tablet), one to two tablets up to three times a day
by mouth. Patients are discharged when their vital signs are
stable for 24 hours, with weight > 75% mBMI.
Statistical analysis
The primary outcome variable was the difference in rate of
change of %mBMI between groups. Secondary outcome variables
were differences in LOS and incidence of hypophosphatemia,
hypomagnesemia, and hypokalemia between groups. Results
were analyzed using independent t test for continuous variables
and chi-square analysis for categorical data. For continuous
variables, Levenes test was used for equality of variance and the
Mann-Whitney U test was used when signicant difference in
variance was detected. The paired t test was used to compare
change in %mBMI from baseline on each day of hospitalization.
Multivariable linear regression was used to determine the effect
of initial prescribed calories (treated as a continuous variable) on
the rate of change of weight gain (controlling for potential
confounders). The dependent variable was the rate of change of
%mBMI during hospitalization. The predictor variables were:
initial prescribed calories (treated as a continuous variable); kcal
from 24-hour dietary recall from the day before admission, and
%mBMI on admission and year of admission. We also included an
interaction term between %mBMI on admission and initial
prescribed calories (which was the product of these two variables). Date of admission was included in this model because of
the change in our clinical practice that began in 2008. Logistic
regression analysis was used to determine the effect of initial
prescribed calories treated as a continuous variable on the
presence of hypophosphatemia, hypomagnesemia, or hypokalemia. The dependent variable for this model was the development of hypophosphatemia, hypomagnesemia, or hypokalemia.

575

The predictor variables were: initial prescribed calories

(continuous variable); kilocalories from 24-hour dietary recall
from the day before admission; rate of weight loss before
admission; year of admission; and initial %mBMI. Subgroup
analyses were conducted for patients who were severely
malnourished (<70% mBMI). Data are presented as mean 
standard deviation. Results were analyzed using SPSS v19.0
software (SPSS Inc, Chicago, IL).
Results
During the 5-year study period, 330 subjects met eligibility
criteria for inclusion into the study. Twenty subjects were
excluded for the following reasons: transferred from another
facility (n 10), required nasogastric feeding (n 8), and signed
out against medical advice before medically stable (n 2). The
study sample was therefore composed of 310 subjects. They were
predominantly female (88.4%) and Caucasian (82.9%). Mean age
was 16.1  2.3 years (range, 10.4e21 years). They had lost on
average 13.7  9.6 kg over a period of 1.1  1.0 years. Subjects
were moderately malnourished (78.5  8.3 %mBMI) with a mean
BMI of 16.0  1.8 kg/m2.
Table 1 shows that 88 subjects were in the lower-calorie
group (1,163  107 kcal/d; range, 720e1,320 kcal/d) and 222 in
the higher-calorie group (1,557  265 kcal/d; range, 1,400e2,800
kcal/d). In the higher-calorie group, 90% of subjects received
1,400e1,800 kcal/d and 95% received 1,400e2,000 kcal/d. At
baseline, there were no signicant differences in age, weight loss,
rate of weight loss, 24-hour dietary recall on the day before
admission, admit weight, admit BMI or %mBMI between groups.
However, those started on a lower-calorie diet had been ill longer
(p .002). At the end of hospitalization, absolute weight gain
was higher in the lower-calorie group (p .01) but after
correction for their longer LOS, rate of change of %mBMI did not
differ between groups. Neither group had initial weight loss. The
%mBMI increased signicantly from baseline by day 1 in the
higher-calorie group (p < .001) and by day 2 in the lower-calorie
group (p < .001). By multivariable regression analysis, rate of
change of %mBMI was signicantly associated with lower %mBMI

Table 1
Demographic and clinical features: higher-calorie versus lower-calorie groups
Lower-calorie
intake (n 88)
Age, year
Duration illness, year
Weight loss, kg
Rate of weight loss,
kg/mo
24-h dietary recall,
kcal/24 hours
Admit weight, kg
Admit BMI, kg/m2
Admit %mBMI
Initial prescribed calories
Discharge BMI, kg/m2
Discharge %mBMI
Discharge calories
Weight gain, kg
% Change mBMI
Rate of change %mBMI
Length of stay, days

16.2
1.39
14.4
1.6






2.4
1.3
11.0
1.4

886  631
41.8
15.9
77.9
1,163
17.2
84.3
2,531
3.6
9.3
.57
16.6













6.5
2.2
9.6
107
1.9
8.2
608
2.3
6.9
.30
9.0

Data are means  standard deviation.

BMI body mass index; mBMI mean BMI.

Higher-calorie
intake (n 222)
16.1
1.00
13.4
1.6






2.3
.9
9.0
1.6

973  558
42.9
16.1
78.7
1,557
17.1
83.7
2,560
2.9
7.2
.55
13.0

 7.5
 1.7
7.8
 265
 1.5
 6.9
 598
 1.9
 5.5
 .37
 7.3

p
.69
.002
.45
.97
.30
.21
.48
.44
<.001
.63
.54
.70
.01
.01
.50
<.0001

576

N.H. Golden et al. / Journal of Adolescent Health 53 (2013) 573e578

Table 2
Multivariable regression analysis of effects of initial prescribed calories, 24-hour
dietary recall, percent mean body mass index on admission (mBMI), and date of
admission on rate of change of percent mBMI
Model
Constant
%mBMI on admission
Initial prescribed calories
24-hour recall of calorie intake
Admit date, year

Standardized coefcient, b
.347
.026
.059
.013

p
.63
<.0001
.69
.33
.84

at baseline (p
.0001) but not with initial prescribed calories

(p .69), 24-hour dietary recall (p .33), or date of admission
(p .84) (Table 2). There was no evidence of interaction between
initial %mBMI and initial prescribed calories treated as a continuous variable (p .92).
Length of hospital stay was signicantly shorter in the highcalorie group 13.0  7.3 days versus 16.6  9.0 days; p
.0001.
A total of 49 subjects developed hypophosphatemia (15.8%), 47
developed hypomagnesemia (15.2 %), and 62 developed hypokalemia (20.0%). Figure 1 shows there was no signicant difference in the rates of hypophosphatemia (p .49),
hypomagnesemia (p 1.0), or hypokalemia (p .35) in the
higher-calorie group compared with the lower-calorie group. No
subjects in either group developed clinical refeeding syndrome.
By logistic regression, hypophosphatemia was associated with %
mBMI on admission (b .055; p .004) and rate of weight loss
prior to admission (b .2; p .03), but not caloric intake on
admission (b .001; p .14).
A total of 49 subjects were <70% mBMI on admission. In this
subgroup, mean %mBMI was 65.1%  3.8% (range, 51%e69.9%)
and mean BMI was 13.7  1.2 (range, 11.0e16.6). Of the 49
subjects in this severely malnourished group, 31 were in the

Figure 1. Prevalence of hypophosphatemia, hypomagnesemia and hypokalemia:

higher-calorie vs. lower-calorie intake.

higher-calorie group and 18 were in the lower-calorie group.

Those in the lower calorie group had lost more weight than those
in the higher calorie group, 22.5  13.2 kg versus 13.1  7.5 kg,
p .01, but there were no other signicant differences between
groups in demographic or clinical variables. Similar to the total
group, there was no signicant difference in rate of change of
%mBMI between groups. Length of stay was 20.2  9.0 days in the
higher-calorie group compared with 23.1  10.5 days in the
lower-calorie group. This difference did not reach statistical
signicance in this smaller sample, but the magnitude of the
difference (approximately 3 days) is similar to that seen in the
larger sample. In the higher-calorie group 8 of 31 (25.8%)
developed hypophosphatemia, compared with 7 of 18 subjects
(38.9%) in the low-calorie group, c2 .92, p .34. The incidence
of both hypomagnesemia and hypokalemia was similar in the
higher-calorie group and the lower-calorie group (5 of 31 [16.1%]
versus 3 of 18 [16.7%], c2 .002, p .96; and 8 of 31 [25.8%]
versus 4 of 18 [22.2%], c2 .08, p .78), respectively.
Discussion
The results of our study demonstrate that in adolescents
hospitalized with AN, starting patients on an average of
1,550 kcal/d (range, 1,400e2,000 kcal/d) is associated with
reduced LOS and is not associated with increased rates of hypophosphatemia, hypomagnesemia, or hypokalemia. No subjects in
either group developed clinical refeeding syndrome. Most
important, we found no evidence of an association between
higher prescribed calories and refeeding syndrome in those who
are severely malnourished (<70% mBMI). Although statistical
power was limited in this subgroup, the trend was toward lower
rates of hypophosphatemia and hypokalemia in the higher-calorie
group.
Our ndings are in support of those of Garber et al. [18], who
also found that higher caloric intake on admission predicted
shorter hospital stay. Reduction in LOS is important not only
because it reduces cost of treatment, but also because it enables
the adolescent to return home sooner, where family-based
treatment, the most effective treatment for adolescent with
AN [24], can be initiated or continued. Earlier studies, also
conducted in adolescents with AN, demonstrated initial weight
loss for 5e8 days when started on hypocaloric diets of
1,000e1,200 kcal/d [18,20]. In contrast to these studies, in our
present study, the higher-calorie group gained weight within
1 day and the lower-calorie group gained weight within 2 days
of admission. The most likely reason for these differences is the
higher amount of calories prescribed on admission. However,
despite a greater rapidity of onset of weight gain and reduced
LOS in the higher-calorie group, after correcting for LOS, we did
not nd an increase in the rate of weight gain in the highercalorie group. This nding is in contrast to the study of Garber
et al., in which the rate of weight gain was also higher in those
prescribed higher calories. The reasons for these differences in
rate of weight gain are not clear. One possible explanation is
that in the study of Garber et al., although patients were started
on fewer calories, they were discharged on more calories than
were our patients (2,668  387 kcal/d compared with 2,547 
603 kcal/d). Presumably, caloric intake was increased more
rapidly in that study.
Other investigators have documented the safety of more
aggressive refeeding protocols in hospitalized patients with AN. In
a small sample of 46 admissions of 29 adolescents with AN,

N.H. Golden et al. / Journal of Adolescent Health 53 (2013) 573e578

Whitelaw et al. [25] found that a protocol starting patients on

a minimum of 1,900 kcal/d was associated with mild hypophosphatemia (serum phosphorus <3.4 mg/dL) in 38% of subjects,
but was otherwise safe. In a population of severely malnourished
young adults with AN (mean BMI, 11.3  .7 kg/m2), Gentile et al.
[26] utilized higher caloric intakes than recommended by current
guidelines by using a combination of nasogastric and oral feeding,
and demonstrated that with prophylactic supplementation of
thiamine, phosphate, and potassium on a specialized unit, no
subjects developed refeeding syndrome. Caloric prescription was
based on indirect calorimetry measurements of resting energy
expenditure. They argued that even the most severely malnourished patients could safely be treated on specialized units with
careful monitoring. There is no consensus regarding prophylactic
administration of phosphorus supplementation during the rst
weeks of nutritional rehabilitation, or whether to supplement
phosphorus only when levels drop. A recent survey of adolescent
medicine practitioners in North America who treat patients with
eating disorders conrmed the wide variations in clinical practice
in this regard [27].
With good reason, concerns about the refeeding syndrome
have led to cautious recommendations regarding refeeding
malnourished patients. The results of our study demonstrate that
hypophosphatemia was associated with the degree of malnutrition, but not prescribed calorie intake. The relationship between
both the refeeding syndrome and hypophosphatemia and the
degree of malnutrition is well known [4,5,14,15,25,28], but there
is an assumption that the refeeding syndrome is associated with
the amount of calories prescribed during refeeding, without good
data to support such an assumption. Case reports have shown
that the refeeding syndrome can develop while consuming
a wide variety of caloric intakes, including intakes as low as
500e800 kcal/d [4,15]. Whereas some organizations have
deliberately avoided recommending a specic starting calorie
amount to avoid refeeding syndrome, based on lack of available
data [22,23], other organizations have made specic recommendations. The Academy of Nutrition and Dietetics recommends starting at 1,000e1,200 kcal/d, with incremental
advances to achieve a weight gain of .5e1.0 lb/week [2]. The
American Psychiatric Association recommends starting at
1,000e1,600 kcal/d [1]. The National Institute for Health and
Clinical Excellence recommends starting on a maximum intake of
10 kcal/kg/d (500 kcal/d for a patient weighing 50 kg) for adults
at high risk of developing refeeding problems (dened as those
with one or more of the following: BMI <16 kg/m, unintentional
weight loss > 15% within the prior 3e6 months, little or no
nutritional intake for > 10 days or low levels of phosphate,
magnesium, or potassium before feeding) [3]. These guidelines
pertain to adults and not specically to those with eating disorders. Caution is certainly warranted, especially in those who are
severely malnourished, but the caution may need to focus more
on careful monitoring of electrolytes and cardiac function than
on the amount of calories prescribed. Even though electrolytes
can be normal on admission, hypophosphatemia, hypomagnesemia, and hypokalemia may develop only after refeeding has
been initiated, so electrolyte abnormalities and cardiac arrhythmias need to be anticipated and promptly treated to prevent
clinical refeeding syndrome.
The predominant theory for the development of refeeding
syndrome is electrolyte shifts from the extracellular to intracellular spaces with refeeding, stimulated by insulin secretion in
response to reintroduction of carbohydrate. The intracellular

577

shifts of these minerals result in a drop of serum phosphorus,

magnesium, and potassium, which can lead to cardiac and
neuromuscular dysfunction. Whereas hypophosphatemia tends
to occur early in refeeding, hypomagnesemia tends to occur later
[28]. Because it is the carbohydrate component of the diet that
primarily initiates the insulin surge, some investigators have
advocated that development of hypophosphatemia depends
more on the carbohydrate content of the diet rather than the
amount of calories, and they have recommended restricting the
carbohydrate content of the diet to <40% of total daily energy
intake [29]. Our patients received balanced diets supplemented
by a high-calorie liquid supplement, but we did not specically
document the macronutrient composition of daily meals for each
patient.
The major strength of our study is the large sample size. We
believe this to be the largest study of nutritional rehabilitation in
adolescents with AN. Our sample also includes 49 subjects with
severe malnutrition, who therefore were at highest risk for
development of refeeding syndrome. In addition, our patients
were all admitted to a specialized unit in which data were
collected systematically. Data were collected on weight loss and
caloric intake before admission, in addition to %mBMI, which
provided a richer assessment of the degree of malnutrition. The
major limitation of this study is that it is a retrospective study
subject to systematic bias. Subjects were not randomized into
higher-calorie or lower-calorie groups, and the initial caloric
prescription was made by individual providers based on their
clinical assessments of the patient, which included the degree of
malnutrition as well as the severity of caloric restriction in the
days before admission. Furthermore, our reported prevalence of
hypophosphatemia is an underestimate because some providers
elected to supplement phosphorus prophylactically before levels
dropped below 3.0 mg/dL. Finally, there is not a large difference
in the amount of calories prescribed between the two groups,
which may explain the lack of difference in rate of weight gain.
We agree with the comments of Katzman [30] that there is
a need to challenge existing recommendations with scientic
evidence demonstrating safety and efcacy of different protocols. The results of our study add to that body of literature to
inform clinical practice. Our results demonstrate that in hospitalized adolescents with AN, compared with starting on a caloric
intake of approximately 1,160 kcal/d, a higher caloric diet of
approximately 1,550 kcal/d is associated with reduced LOS and
no increased rates of refeeding syndrome. We found that
refeeding hypophosphatemia depended on the degree of
malnutrition but not prescribed caloric intake, within the range
studied. Our study suggests that concerns about refeeding
hypophosphatemia have unnecessarily limited the amount of
calories prescribed to hospitalized adolescents with AN, and
more aggressive feeding protocols should be implemented.

References
[1] American Psychiatric Association. Treatment of patients with eating
disorders, third edition. Am J Psychiatry 2006;163(suppl):4e54.
[2] American Dietetic Association. Position of the American Dietetic Association:
Nutrition intervention in the treatment of anorexia nervosa, bulimia nervosa,
and other eating disorders. J Am Diet Assoc 2006;106:2073e82.
[3] National Institute for Clinical Excellence. Nutrition support in adults. Oral
nutrition support enteral tube feeding and parenteral nutrition. www.nice.
org.uk/CG032. Accessed January 11, 2013.
[4] Kohn MR, Golden NH, Shenker IR. Cardiac arrest and delirium: Presentations
of the refeeding syndrome in severely malnourished adolescents with
anorexia nervosa. J Adolesc Health 1998;22:239e43.

578

N.H. Golden et al. / Journal of Adolescent Health 53 (2013) 573e578

[5] Ornstein RM, Golden NH, Jacobson MS, et al. Hypophosphatemia during
nutritional rehabilitation in anorexia nervosa: Implications for refeeding
and monitoring. J Adolesc Health 2003;32:83e8.
[6] Solomon SM, Kirby DF. The refeeding syndrome: A review. JPEN J Parenter
Enteral Nutr 1990;14:90e7.
[7] Fuentebella J, Kerner JA. Refeeding syndrome. Pediatr Clin North Am 2009;
56:1201e10.
[8] Weinsier RL, Krumdieck CL. Death resulting from overzealous total parenteral nutrition: The refeeding syndrome revisited. Am J Clin Nutr 1981;34:
393e9.
[9] Sheridan PH, Collins M. Potentially life-threatening hypophosphatemia in
anorexia nervosa. J Adolesc Health Care 1983;4:44e6.
[10] Hall DE, Kahan B, Snitzer J. Delirium associated with hypophosphatemia in
a patient with anorexia nervosa. J Adolesc Health 1994;15:176e8.
[11] Kaysar N, Kronenberg J, Polliack M, et al. Severe hypophosphataemia
during binge eating in anorexia nervosa. Arch Dis Child 1991;66:
138e9.
[12] Beumont PJ, Large M. Hypophosphataemia, delirium and cardiac
arrhythmia in anorexia nervosa. Med J Aust 1991;155:519e22.
[13] Birmingham CL, Alothman AF, Goldner EM. Anorexia nervosa: Refeeding
and hypophosphatemia. Int J Eat Disord 1996;20:211e3.
[14] Fisher M, Simpser E, Schneider M. Hypophosphatemia secondary to oral
refeeding in anorexia nervosa. Int J Eat Disord 2000;28:181e7.
[15] Norris ML, Pinhas L, Nadeau PO, et al. Delirium and refeeding syndrome in
anorexia nervosa. Int J Eat Disord 2012;45:439e42.
[16] Golden NH, Meyer W. Nutritional rehabilitation of anorexia nervosa: Goals
and dangers. Int J Adolesc Med Health 2004;16:131e44.
[17] Tresley J, Sheean PM. Refeeding syndrome: Recognition is the
key to prevention and management. J Am Diet Assoc 2008;108:
2105e8.
[18] Garber AK, Michihata N, Hetnal K, et al. A prospective examination
of weight gain in hospitalized adolescents with anorexia nervosa
on a recommended refeeding protocol. J Adolesc Health 2012;50:
24e9.

[19] Keys A. The biology of human starvation. Minneapolis (MN): University of

Minnesota Press; 1950.
[20] Solanto MV, Jacobson MS, Heller L, et al. Rate of weight gain of inpatients
with anorexia nervosa under two behavioral contracts. Pediatrics 1994;
93(Pt 1):989e91.
[21] American Psychiatric Association. Diagnostic and statistical manual for
mental disorders. 4th ed. Arlington, VA: American Psychiatric Association,
1994.
[22] Golden NH, Katzman DK, Kreipe RE, et al. Eating disorders in adolescents:
Position paper of the Society for Adolescent Medicine. J Adolesc Health
2003;33:496.
[23] Rosen DS. Identication and management of eating disorders in children
and adolescents. AAP Policy 2010;126:1240.
[24] Lock J, Le Grange D, Agras WS, et al. Randomized clinical trial comparing
family-based treatment with adolescent-focused individual therapy for
adolescents with anorexia nervosa. Arch Gen Psychiatry 2010;67:1025e32.
[25] Whitelaw M, Gilbertson H, Lam PY, et al. Does aggressive refeeding in
hospitalized adolescents with anorexia nervosa result in increased hypophosphatemia? J Adolesc Health 2010;46:577e82.
[26] Gentile MG, Pastorelli P, Ciceri R, et al. Specialized refeeding treatment for
anorexia nervosa patients suffering from extreme undernutrition. Clin Nutr
2010;29:627e32.
[27] Schwartz BI, Mansbach JM, Marion JG, et al. Variations in admission
practices for adolescents with anorexia nervosa: A North American sample.
J Adolesc Health 2008;43:425e31.
[28] Raj KS, Keane-Miller C, Golden NH. Hypomagnesemia in adolescents with
eating disorders hospitalized for medical instability. Nutr Clin Pract 2012;
27:689e94.
[29] Kohn MR, Madden S, Clarke SD. Refeeding in anorexia nervosa: Increased
safety and efciency through understanding the pathophysiology of
protein calorie malnutrition. Curr Opin Pediatr 2011;23:390e4.
[30] Katzman DK. Refeeding hospitalized adolescents with anorexia nervosa: Is
start low, advance slow urban legend or evidence based? J Adolesc
Health 2012;50:1e2.