ORIGINAL ARTICLE

Long-term effects of enteral feeding on growth and mental healthin adolescents with anorexia nervosa—results of a retrospectiveGerman cohort studyI Nehring1, K Kewitz2, R von Kries1 and U Thyen2

BACKGROUND/OBJECTIVE: Anorexia nervosa (AN) is a severe eating disorder with a high mortality rate. Treatment regimes showregional and global variation and are sometimes supported by enteral feeding (EF) via nasogastric tube, although risks and benefitsare still unclear. We aimed to find out whether EF improves growth and AN recovery and prevents psychiatric comorbidities.SUBJECTS/METHODS: Data were retrospectively collected from medical records and follow-up data were collected viaquestionnaires. Two hundred and eight female AN patients who were hospitalized below the age of 18 years with a mean follow-upof 6 years were analyzed. We calculated relative risks for the association between EF and suboptimal growth, remission of AN andthe occurrence of psychiatric comorbidities, adjusting for potential confounders.RESULTS: A third of the analyzed girls received EF at any time. In the adjusted analyses, we found no significant associationsbetween EF and suboptimal growth, the persistence of AN and the occurrence of psychiatric comorbidities, respectively.CONCLUSION: Our data suggest EF to be neither a risk factor nor beneficial for growth, recovery or persistence of ANand the occurrence of psychiatric comorbidities.

European Journal of Clinical Nutrition (2014) 68, 171–177; doi:10.1038/ejcn.2013.244; published online 4 December 2013

Keywords: anorexia nervosa; enteral feeding; treatment; adolescence

INTRODUCTIONAnorexia nervosa (AN) is an eating disorder related with severeweight loss and disturbance of body function affecting mostlyfemale children and adolescents.1 Though AN is a rare diseasewith a prevalence of some 0.3%, it is the mental disorder with thehighest mortality.2 A mortality rate of at least 4% per decade offollow-up was reported to be related to AN.3 Early onset of AN canbe a risk factor for impaired growth of body length.4,5

Furthermore, about 20% of the AN patients remain chronically illand up to 70% develop psychiatric comorbidities.6

There are different approaches for treating AN, such as familytherapy, individual psychological therapies, inpatient care andpsychopharmacologic interventions.7,8 Parenteral nutrition andenteral feeding (EF) via nasogastric tube are suggested to besuccessful opportunities to support AN therapy.9–11 This kind ofcompulsory treatment is increasingly challenged in young patientsunwilling or unable to consent to the procedure.12 EF seems to bea tightrope walk: preventing starvation in the malnourishedpatient confronts to leaving out his individuality and freedom.12

EF is not only a psychological but also a physiological challengefor the patient and, thus, it should be planned and managedcarefully.13 As a result, EF is claimed to be the last resort in ANtherapy.13

The literature does not provide sufficient evidence to identify anoptimal AN therapy.8 The current German guidelines state thatforced EF ‘should only take place when all other measures havebeen exhausted’.14 There are only few studies with very small

sample sizes on the efficiency of EN,10,15 which did not provideevidence for recommendations.

We therefore analyzed in a retrospective study the associationbetween EF and the three outcomes: suboptimal growth,persistence of AN and occurrence of other mental disorders atfollow-up.

MATERIALS AND METHODSParticipantsPatients were recruited in four hospital units for child and adolescentpsychiatry and psychosomatic paediatrics in Northern Germany. Tworesearchers reviewed medical records of AN patients, as defined byICD-10, in the respective hospital. The head of the service invited formerpatients by mail to participate in a follow-up study. To reach as manypatients as possible, not responding persons were contacted bytelephone.

Patients were included eligible if they met the following inclusioncriteria:

� at least one inpatient visit in one of the four included hospitalsbetween 1996–2006

� AN was the main reason for admission� hospitalization for more than 14 days� last in-patient visit at least one year ago at time point of

reviewing medical records� age at first admission was below 18 years� female sex

1Division of Epidemiology, Institute for Social Paediatrics and Adolescent Medicine, Ludwig-Maximilians-Universitat Munchen, Munich, Germany and 2Department of Paediatricand Adolescent Medicine, University of Lubeck, Lubeck, Germany. Correspondence: I Nehring, Division of Epidemiology, Institute for Social Paediatrics and Adolescent Medicine,Ludwig-Maximilians-University of Munich, Heiglhofstrasse 63, 81377 Munich, Germany.E-mail: ina.nehring@med.uni-muenchen.deReceived 20 June 2013; revised 10 September 2013; accepted 18 October 2013; published online 4 December 2013

European Journal of Clinical Nutrition (2014) 68, 171–177& 2014 Macmillan Publishers Limited All rights reserved 0954-3007/14

www.nature.com/ejcn

Treatment regimes across the four units were similar and included atleast individual psychological therapy and family therapy during the in-patient stay.

We included only females, as males are not comparable pertaining thecharacteristics of AN.16 Patients with incomplete data on EF, age andheight (n¼ 57) were excluded.

Ethics approval was obtained from the Ethical Review Board at theUniversity of Lubeck and written informed consent was obtained fromparticipants.

Data collectionData collection was conducted in two stages: first, data were abstractedfrom medical records. These included: information on sex, date of birth,dates of hospital admission and discharge, repeated hospitalizations,outpatient therapies, psychiatric comorbidities, anthropometric data, typeof AN, comorbidities, occurrence of menarche and EF.

Second, a follow-up was conducted by contacting the patients via mailand asking them to answer questionnaires to assess the persistence orrelapse of eating disorders and occurrence of any other mental disorders.Furthermore, a self-created questionnaire was used to obtain data aboutthe current life situation, physical health, age and anthropometric data.

The Structured Inventory for Anorexic and Bulimic Eating Disorders(SIAB-S)17 was used to determine the presence of AN at follow-up. Thisstandardized validated questionnaire on eating disorders collects self-reported information on AN, bulimia nervosa (BN), binge eating disordersand eating disorders not otherwise specified (EDNOS) as defined by DSM-IV18 (EDNOS) and ICD-101 (AN, BN). Presence of AN was defined as meetingat least three ICD-10 criteria for AN. In the following text we use the word‘recovery’ for no longer diagnosis of AN.

The occurrence of other mental disorders (except eating disorders) wasidentified with the Hopkins Symptom Checklist (SCL-90-R),19 which enablesa quantification of the general psychopathology based on self-reportedphysiological and psychological symptoms. A mental disorder wassuspected by a General Symptomatic Index of 60 or more.19

Definitions of variablesThe explanatory variable EF was dichotomized and defined as any EF atany hospitalization versus no EF. Height percentiles were calculated usingthe current German age- and sex-specific height references.20 Body massindex (BMI) was defined as body weight divided by squared body length(kg/m2). BMI s.d. scores are based on German age- and sex-specificreference percentiles for BMI.21 As the patients had different follow-upperiods due to varying starting points we report the time since lastdischarge as a proxy for follow-up time. The three outcomes optimalgrowth, persistence of AN and occurrence of other psychiatriccomorbitities were dichotomized (yes/no). We used body lengthpercentiles in increments of 10 to define optimal and suboptimal growthof height; girls who maintained their initial height decile until follow-up orreached a higher decile had optimal growth. Girls who fell on a lowerdecile had suboptimal growth. As in girls with a body height below thethird percentile both initially and at follow-up, suboptimal growth couldnot be defined, these girls were included in the category ‘optimal growth’,although the girls might not realize their full growth potential.

Statistical analysesFor descriptive analyses, we compared characteristics of responders withnon-responders and tube-fed with non-tube-fed girls. We furthercompared patients with and without optimal growth of height, patientswith and without persistent AN at follow-up and patients with and withoutoccurrence of any other mental disorders at follow-up. Data are shown asmean and s.d. or frequency (%) as appropriate. To test for statisticalsignificant differences between the two respective groups we used t-testor Pearson’s w2-test, respectively (a¼ 0.05).

We calculated Poisson Regression models to quantify the association ofEF with growth, remission of AN and the occurrence of other mentaldisorders. Relative risks (RRs) and corresponding 95% confidence intervals(95% CIs) were calculated for crude and adjusted models. All variables thatwere significantly associated with the exposure and the respectiveoutcome, as suggested in the descriptive analyses, were considered aspotential confounders. These variables were added stepwise to themodels. We additionally performed some stratified analyses to detectpotential effect modification (data not shown).

We calculated the detectable RR assuming a study power of 80% and a5% level of significance by using the software Power and Sample SizeCalculations version 2.1.31.

Descriptive and regression analyses were performed using SAS 9.2 forWindows.

In addition, we calculated a propensity score22 considering age atbeginning of disease, BMI at first hospitalization and time betweendiagnosis and hospitalization as covariates. On the basis of the calculatedpropensity score subclasses (quintiles) were formed and RRs between tubefeeding and the specific outcome were calculated for each quintile. TheRRs were pooled using Mantel–Haenzel method.

RESULTSFive hundred and twenty-eight patients met the inclusion criteriaand data were abstracted from the records, 265 of whom agreedto participate in our study. Two hundred and sixty-three patientsdid not participate because they could not be contacted, refusedbecause they did not wish to talk about the disease again, had notime, had unsuccessful therapy or did not want to give confidentinformation. After excluding subjects with incomplete data(n¼ 57) 208 remained for our analyses (Figure 1). The contributionof the four sites ranged from 16–32% of the entire sample. Themedian follow-up period was 5.5 years (mean: 6.3 years) andranged between 1.3 and 12.7 years. Sixty-eight cases (32.7 %) hadmore than one hospitalization.

There were no statistically significant differences betweenparticipants and non-responders pertaining age at beginning ofthe disease, sex, EF, beginning of disease before first menarche,BMI at first hospitalization and BMI at discharge (data not shown).

We found a recovery rate of 69.7% (95% CI: 63.0–75.9), whichwas similar in patients with (65%) and without EF (72%) (P¼ 0.27).

Thirty-four percent of the analyzed girls received EF at anyhospitalization. Table 1 shows the characteristics of girls with andwithout EF. Patients with EF were significantly younger, shorterand had lower BMI at the time of hospitalization. Time betweenonset of disease and hospital admission was shorter and theduration of hospitalization was significantly longer in girls with EF.Girls who experienced EF had more hospitalizations. At follow-up,16.9% (n¼ 12) of the EF girls and 14.6% (n¼ 20) of the non-EF girlsreported to have BN.

Descriptive data on the outcomes optimal growth, persistenceof AN and psychiatric comorbidity are shown in Tables 2–4. Onehundred and fifty-five (75%) girls showed normal growth in heightby our definition; three girls had initially and at follow-up heightsbelow the third percentile. Girls with optimal growth were olderand had higher BMI s.d. scores at their first hospitalization.

Patients with AN persistence had more frequent outpatienttreatments and hospitalizations than patients with recovery atfollow-up (Table 3). Age at beginning of disease did not differbetween patients with and without persistent AN.

Patients with persistent AN have an increased risk to reportsymptoms of psychiatric comorbidities at follow-up (RR¼ 1.39,95% CI: 1.04–1.84). Forty-five percent (n¼ 94) of the girls have hadpsychiatric comorbidities at time of hospitalization. Of these, 45%(n¼ 42) had high scores in the Symptom Checklist indicatingpersisting psychiatric comorbidity, although the others recoveredfrom their comorbid conditions. Forty girls without comorbidity atthe time of hospitalization reported symptoms indicating psy-chiatric comorbidities at follow-up resulting in 39.6% (n¼ 82) ofthe study population with suspected psychiatric comorbidities atfollow-up (Table 4). These participants were also smaller and hadhigher BMI s.d. scores at onset. The follow-up period wassignificantly shorter in patients with persistent AN and in patientswith other mental disorders compared with their more healthycounterparts.

The regression for the association between EF and suboptimalgrowth yielded a crude RR of 1.64 (95% CI: 0.94–2.87) (Table 5).Stepwise adjustment for potential confounders reduced the RR

Long-term effects of enteral feedingI Nehring et al

172

European Journal of Clinical Nutrition (2014) 171 – 177 & 2014 Macmillan Publishers Limited

towards 1.0. Likewise, we did not find significant associations inthe crude and adjusted analyses between EF and the remission ofAN or occurrence of psychiatric comorbidity, respectively.

DISCUSSIONIn a catamnestic retrospective cohort study on 208 adolescentgirls, who were hospitalized for AN with a follow-up of at least one

year, we found no evidence for an association between EF vianasogastric tube (EF) and growth, persistence or remission of ANor the occurrence of psychiatric comorbidities at follow-up,respectively.

A number of outcomes need to be addressed to balance thepotential benefits and drawbacks of EF. Previous studies focusedon safety of EF,9 weight restoration,23 psychological outcomessuch as anxiety, compulsive disorders and depression,9,10,15

Table 1. Characteristics of study subjects stratified by EF

EF(n¼ 71)

No EF(n¼ 137)

Total samplen¼ 208

n Mean (s.d.) n Mean (s.d.) P-valuea n Mean (s.d.)

Age at first hospitalization (years) 71 14.3 (1.6) 137 15.3 (1.6) o0.0001 208 14.9 (1.7)Height at onset (m) 71 1.62 (0.08) 137 1.65 (0.08) 0.01 208 1.64 (0.08)Height percentile at onset 71 51.8 (26.1) 137 53.3 (29.2) 0.72 208 52.8 (28.1)Menarche was before onset of disease (%) 70 N¼ 49

70.0%131 N¼ 104

79.4%0.14 201 N¼ 153

76.1%BMI at first hospitalization (kg/m2) 70 14.3 (1.3) 134 15.1 (1.4) o0.001 204 14.8 (1.4)BMI s.d. scores at first hospitalization 69 � 2.7 (1.1) 134 � 2.9 (1.1) 0.10 203 � 2.75 (1.1)BMI percentile at first hospitalization 70 1.7 (4.7) 136 2.7 (5.5) 0.2 206 2.3 (5.3)Outpatient therapy before first hospitalization (%) 71 N¼ 25

35.2%137 N¼ 48

35.0%0.98 208 N¼ 73

35.1%Time between onset of disease and first hospitalization (weeks) 66 34.5 (23) 122 53.3 (49) 0.004 188 46.7 (42.7)Duration of first hospitalization (days) 71 136 (98) 137 82 (48) o0.0001 208 100.5 (73.6)BMI at first discharge (kg/m2) 68 17.1 (1.6) 132 17.4 (1.4) 0.24 200 17.3 (1.5)BMI s.d. scores at first discharge 66 � 1.28

(0.8)128 � 1.33 (0.8) 0.69 194 � 1.31

(0.82)Total number of hospitalizations 71 2.0 (1.4) 137 1.4 (0.9) 0.0002 208 1.63 (1.1)Mental disorders at hospitalization 71 N¼ 33

46.5%137 N¼ 62

45.3%0.87 208 N¼ 95

45.7%

Follow-upTime since last discharge (years) 71 7.2 (3.1) 137 5.8 (2.8) 0.002 208 6.3 (3.0)BMI at follow-up 67 19.7 (3.3) 131 19.6 (2.5) 0.8 198 19.6 (2.8)

Abbreviations: BMI, body mass index; EF, enteral feeding. at-Test or Pearson’s w2. Bold entries indicate Po0.05.

Figure 1. Flow of participants through study.

Long-term effects of enteral feedingI Nehring et al

173

& 2014 Macmillan Publishers Limited European Journal of Clinical Nutrition (2014) 171 – 177

medical complications15 and recovery rate9,10 after AN therapysupported by EF. In adolescent girls persistence or remission ofAN, the occurrence of psychiatric comorbidities at follow-up andthe growth of body length appear to be of particular relevancewith respect to potential harms and benefits of EF.

In his comprehensive review, Steinhausen reported an on-average recovery rate of only 46.9% (range: 0–92) for AN patients,whereas 20% (range: 0–79) of the patients had persistent chronicAN and a third improved only partly (range: 0–75).6 In our studywe found a higher recovery rate of 69.7%, which was similar inpatients with (65%) and without EF (72%). There are surprisinglyfew studies comparing artificial feeding, either by tube orparenterally, to oral feeding. A previous cohort study with amean follow-up of 3 years showed that patients who received oraland parenteral refeeding had nearly the same recovery rate (61%)as patients who received oral refeeding alone (64%).9 A recentrandomised controlled trial (RCT) on adult AN and BN patients byRigaud and colleagues,10 however, concluded that tube fedpatients had significantly less binge eating and purging episodes

one year after hospitalization than non-tube-fed patients. Thistrial, however, had patients with an unusual high compliance, andthe authors may have had a conflict of interest as it was supportedby Nestle Home Care. Our fully adjusted regression model yieldedno evidence for an association between EF and AN remission,which is in accordance with two previous smaller observationalstudies.9,15 Because of comparatively high sample size our studysubstantially supports the concept of no clinically relevantassociation between EF and persistence or recovery of AN: thestudy was sufficient to detect a RR below 0.56 (risk related to EF)or above 1.66 (beneficial effect of EF) with a power of 80%.

Psychiatric comorbidities, such as depression and compulsivebehaviors, are frequent phenomena during and after recovery ofeating disorders.24,25 More than half of the AN patients sufferedfrom psychiatric comorbidities in a 12-year follow-up study byFichter et al.26 We found nearly 40% of the patients reportedsymptoms indicating another mental disorder at follow-up. Half ofthese (n¼ 42) have already had any psychiatric comorbidity athospitalization, as reported in medical records.

Table 2. Characteristics of subject stratified by growth

Suboptimal growth (height)(n¼ 53)

Optimal growth (height)(n¼ 155)

P-Valuea

n Mean (s.d.) n Mean (s.d.)

Age at first hospitalization (years) 50 13.6 (1.2) 158 15.5 (1.5) o0.0001Height at onset (m) 50 1.61 (0.06) 158 1.65 (0.08) 0.002Height percentile at onset 50 55.1 (25.7) 158 52.2 (28.6) 0.52Menarche was before onset of disease (%) 49 N¼ 32 65.3% 152 N¼ 121 79.6% 0.04BMI at first hospitalization (kg/m2) 50 14.8 (1.3) 154 14.9 (1.5) 0.8BMI s.d. scores at first hospitalization 50 � 2.25 (0.9) 153 � 2.91 (1.1) 0.0002Outpatient therapy prior to first hospitalization (%) 50 N¼ 13 26.0% 158 N¼ 60 38.0% 0.12Time between beginning of disease and first hospitalization (weeks) 47 34.4 (25.0) 141 50.9 (46.5) o0.0001Total number of hospitalizations 50 1.8 (1.1) 158 1.6 (1.1) 0.13BMI at first discharge 48 17.07 (1.4) 152 17.4 (1.5) 0.17

Follow-upTime since last discharge (years) 50 5.8 (3.2) 158 6.4 (2.9) 0.18BMI at follow-up 48 18.9 (2.7) 150 19.9 (2.8) 0.03

Abbreviation: BMI, body mass index. Optimal growth: stayed on height percentile or reached higher percentile; suboptimal growth: fell to lower percentile.at-Test or Pearson’s w2. Bold entries indicate Po0.05.

Table 3. Characteristics of subjects stratified by persistence of AN at follow-up

Persistence of AN at follow-up(n¼ 63)

Remission of AN at follow-up(n¼ 145)

P-valuea

n Mean (s.d.) n Mean (s.d.)

Age at first hospitalization (years) 63 14.9 (1.8) 145 15.1 (1.5) 0.39Height at onset (mean±s.d.) 63 1.63 (0.09) 145 1.64 (0.07) 0.26Height percentile at onset 61 50.4 (27.8) 143 52.7 (28.0) 0.59Menarche was before onset of disease (%) 60 N¼ 42 70.0% 141 N¼ 111 78.7% 0.18BMI at first hospitalization (kg/m2±s.d.) 61 14.9 (1.4) 143 14.8 (1.5) 0.81BMI s.d. scores at first hospitalization 61 � 2.65 (1.1) 142 � 2.80 (1.1) 0.37Outpatient therapy before first hospitalization (%) 63 N¼ 29 46.0% 145 N¼ 44 30.3% 0.03Time between onset of disease and first hospitalization (weeks) 61 43.5 (44.2) 127 48.3 (42.1) 0.48BMI at first discharge (kg/m2) 63 17.0 (1.3) 137 17.5 (1.5) 0.06Total number of hospitalizations 63 1.84 (1.1) 145 1.54 (1.1) 0.08

Follow-upTime since last discharge (years) 61 6.5 (3.2) 142 7.5 (3.1) 0.02BMI at follow-up 62 18.0 (3.0) 136 20.4 (2.4) o0.0001

Abbreviations: AN, anorexia nervosa; BMI, body mass index. at-Test or Pearson’s w2. Bold entries indicate Po0.05.

Long-term effects of enteral feedingI Nehring et al

174

European Journal of Clinical Nutrition (2014) 171 – 177 & 2014 Macmillan Publishers Limited

There are only few studies examining the effect of EF onpsychiatric outcomes.10,15 The above mentioned RCT by Rigaud10

suggested that EF combined with cognitive behavioral therapysignificantly improved anxiety and depression levels comparedwith cognitive behavioral therapy only in patients with AN and BN.Our regression model indicated neither an increased nor areduced risk for the occurrence of other mental disorders atfollow-up after EF. The sample size was sufficient to detect a RRbelow 0.60 (beneficial effect of EF on the occurrence of mentaldisorders) or above 1.62 (increased risk for mental disorders by EF)with a power of 80%.

We additionally performed an analysis stratified by absence ofdocumented mental disorders at hospitalization to compareincidence of and remission from mental disorders; we could notobserve any significant differences of the effect of EF between thestrata (data not shown).

This study was designed as a part of health services researchand used a catamnestic approach, which allows for retrospectivecohort analyses. Owing to the catamestic approach diagnosticcriteria for psychiatric disorders were different at hospitalization(as documented in medical records) and follow-up (based onvalidated questionnaires). We therefore based our analyses onsuspected diagnosis of mental disorders at follow-up indicated.

An additional analysis corroborated that patients with persistentAN had an increased risk for later psychiatric comorbidities.Stratification by persistence of AN did not suggest effectmodification of EF by AN.

We are not aware of a study examining the effect of EF onyoung patients’ growth. It is known that AN patients havediminished growth hormone activity, which might account forgrowth retardation.4 Nutritional rehabilitation may increasegrowth hormone status allowing for ‘normal’ growth.4 We

Table 4. Characteristics of subjects stratified by the occurrence of psychiatric comorbidities, other than eating disorders, at follow-up

Psychiatric comorbiditiesat follow-up (n¼ 82)

No psychiatric comorbiditiesat follow-up (n¼ 125)

P-valuea

Mean (s.d.) n Mean (s.d.)

Age at first hospitalization (years±s.d.) 82 15.0 (1.8) 125 15.2 (1.5) 0.42Height at onset (mean±s.d.) 82 1.62 (0.08) 125 1.65 (0.08) 0.01Height percentile at onset 81 47.0 (27.6) 122 55.2 (27.7) 0.04Menarche before onset of disease (%) 81 N¼ 65 80.3% 119 N¼ 88 74.0% 0.30BMI at first hospitalization (kg/m2±s.d.) 80 15.1 (1.6) 123 14.7 (1.3) 0.09BMI s.d. scores at first hospitalization 79 � 2.56 (1.1) 123 � 2.89 (1.1) 0.03Outpatient therapy before first hospitalization (%) 82 N¼ 34 41.5% 125 N¼ 39 31.2% 0.13Time between beginning of disease and first hospitalization (weeks, s.d.) 74 44.2 (37.4) 113 48.7 (46.0) 0.49BMI at first discharge (kg/m2) 81 17.31 (1.3) 118 17.34 (1.6) 0.91Total number of hospitalizations 82 1.7 (1.0) 125 1.6 (1.2) 0.48Mental disorders at first assessment 82 N¼ 42 51.2% 125 N¼ 52 41.6% 0.17

Follow-upTime since last discharge (years) 82 5.8 (3.0) 125 6.6 (3.0) 0.05BMI at follow-up 78 19.4 (3.0) 119 19.8 (2.7) 0.37

Abbreviation: BMI, body mass index. at-Test or Pearson’s w2. Bold entries indicate Po0.05.

Table 5. RRs (95% CI) for the association between tube feeding and suboptimal growth (I), remission of AN (II) and occurrence of other mentaldisorders at follow-up (III)

Outcome N Relative Risk (95% CI)

(I) Suboptimal growthCrude 208 1.644 (0.943–2.867)

Adjusted for:(A) Age at first hospitalization 208 1.016 (0.574–1.799)(B) Aþheight at first hospitalization 208 1.139 (0.567–1.810)(C) Bþ time between beginning of disease and first hospitalization 188 1.030 (0.558–1.900)

(II) Remission of AN at follow-upCrude 208 0.897 (0.632–1.272)

Adjusted for:Time since last discharge 208 0.835 (0.584–1.196)

(III) Psychiatric comorbidities at follow-upCrude 207 1.129 (0.720–1.770)

Adjusted for:(A) Height at beginning 207 1.048 (0.664–1.653)(B) Aþ time since last discharge 207 1.148 (0.720–1.830)

Abbreviations: AN, anorexia nervosa; CI, confidence interval, RRs, Relative Risks.

Long-term effects of enteral feedingI Nehring et al

175

& 2014 Macmillan Publishers Limited European Journal of Clinical Nutrition (2014) 171 – 177

therefore examined whether EF—as a sort of nutritionalrehabilitation—prevents growth retardation. A quarter of theanalyzed patients did not grow normally by our definition. Ourdata, however, did not suggest EF to prevent suboptimal growthin the adjusted analyses. Treated patients might feel worse withtheir AN and be more prone to surreptitious compensatorybehaviors. Thus, a significant difference in growth might not havebeen detected. The sample size, however, was sufficient to detecta RR below 0.52 (beneficial effect of EF on growth) or above 1.72(risk for suboptimal growth after EF). As the study was aretrospective cohort study, we are unable to provide evidencethat the outcome for participants with EF would have been similarhad they not received EF. We cannot exclude that they wouldhave shown less optimal growth below a RR of 1.72. An RCTcomparing adolescent patients with severe nutritional deprivationrandomly allocated to EF could provide data on the efficacy ofsuch treatment, however is unlikely to be feasible.

We therefore additionally applied a supplemental analysisbased on a propensity score, which has been suggested to mimican RCT with observational data, as it reduces the bias in covariatessimultaneously.22 The analyses yielded similar results for alloutcomes (see Supplementary Information).

Strength and limitationsThis retrospective cohort study examined a selected sample of ANpatients of considerable size with a mean follow-up time of 6.3years. Therefore, clinically relevant effects of EF on AN remission,mental disorders and growth could have been detected.

We used validated self-reporting questionnaires to assess theoutcomes of AN persistence and mental disorder occurrence. Thedefinition of suboptimal growth was not based on height at lastmeasurement but considered growth velocity between admissionto hospital and follow-up in adherence to up-to-date Germanpercentiles.20 Anthropometric follow-up data were self-reportedand thus might have been biased.27

As anorectic women tend to slightly overestimate their trueweight,28 reporting bias cannot be excluded. Information onexposure to EF and data pertaining hospitalization wereretrospectively abstracted from medical records by researcherswho were not aware of the outcome. Therefore, information biasis unlikely.

Unfortunately, we do not have sufficient information onfrequency and duration of EF, which might have had an impacton patients’ growth.

As this was no RCT, other risk factors for the outcomesconsidered might not be balanced between the observationgroups. We therefore compared potentially relevant variables ofthe patients with and without EF and the patient groups with andwithout the respective outcome and considered significant factorsas potential confounders. Nevertheless, residual confoundingcannot be excluded.

Furthermore, there might be some risk of attrition bias becausethe response rate was only about 50%. The non-responderanalysis, however, did not show any significant differencesbetween participants and non-responders. Particularly the fre-quency of EF was almost identical in the two groups suggesting alow risk of attrition bias.

The supplemental analyses are based on subclassification onthe propensity score. This approach allows for a better compar-ability in treatment and control groups without randomization.22

These analyses yielded similar results as detected by Poissonregression. This further strengthens the results of our study.

CONCLUSIONEF does not appear to account for a clinically relevant improve-ment of growth in young AN patients. Regarding potential harms,

there was no indication of a clinically relevant risk for persistenceof AN or occurrence of mental disorders related to EF. Decisionsfor EF in adolescent girls can neither be justified by potentialimproved length growth nor by reduced AN persistence oroccurrence of mental comorbidities. However, it may remain atreatment option as a last resort in very young nutritionallydeprived patients if other treatments fail, as EF is oftenexperienced as extremely stressful and should, therefore, beavoided as much as possible through intensive motivationalsupport and treatment. Should it still be necessary, a stepwiseperformance (arrangement of guardianship, compulsory admis-sion, EF)14 is recommended.

CONFLICT OF INTERESTThe authors declare no conflict of interest.

ACKNOWLEDGEMENTSWe are grateful to the support of Dr med. Gunther Hinrichs (ZIP Kiel), Dr med.Joachim Walter (Katholisches Kinderkrankenhaus Wilhelmstift, Hamburg), Dr med.Jan Gerrit Behrens (MediClin Seepark, Bad Bodenteich) and Dr phil. Dipl. Psych.Dorothe Verbeek (UKSH Lubeck) who gave clinical advice and made the dataavailable. We thank all patients who took part in this study. We thank Kristiane Krugerfor her help in recruiting the sample and collecting the data and to Sabine Brehm, themedical documentalist. We furthermore thank Lucia Albers and Dr Otmar Bayer forstatistical support. Parts of the manuscript arose from the PhD Thesis of KK (at theUniversity of Lubeck) and from the Master Thesis of IN (at the University of Munich).

AUTHOR CONTRIBUTIONSThe author’s responsibilities were as follows: IN: analyses and principleauthorship of the manuscript; KK: recruitment and data collection, descriptiveanalyses; RvK: interpretation of the data and contribution to final draft;UT: conception of research question, project management, contribution to finaldraft of the manuscript.

REFERENCES1 The World Health Organization. ICD-10 Classification of Mental and Behavioural

Disorders Geneva1993. Available from http://www.who.int/classifications/icd/en/GRNBOOK.pdf.

2 Harris EC, Barraclough B. Excess mortality of mental disorder. Br J Psychiatry 1998;173: 11–53.

3 Signorini A, De Filippo E, Panico S, De Caprio C, Pasanisi F, Contaldo F. Long-termmortality in anorexia nervosa: a report after an 8-year follow-up and a review ofthe most recent literature. Eur J Clin Nutr 2007; 61: 119–122.

4 Golden NH, Kreitzer P, Jacobson MS, Chasalow FI, Schebendach J, Freedman SM et al.Disturbances in growth hormone secretion and action in adolescents with anorexianervosa. J Pediatr 1994; 125: 655–660.

5 Swenne I, Thurfjell B. Clinical onset and diagnosis of eating disorders inpremenarcheal girls is preceded by inadequate weight gain and growthretardation. Acta Paediatr 2003; 92: 1133–1137.

6 Steinhausen HC. The outcome of anorexia nervosa in the 20th century.Am J Psychiatry 2002; 159: 1284–1293.

7 Eisler I, Simic M, Russell GF, Dare C. A randomised controlled treatment trial of twoforms of family therapy in adolescent anorexia nervosa: a five-year follow-up.J Child Psychol Psychiatry 2007; 48: 552–560.

8 Lock J, Agras WS, Bryson S, Kraemer HC. A comparison of short- and long-termfamily therapy for adolescent anorexia nervosa. J Am Acad Child Adolesc Psychiatry2005; 44: 632–639.

9 Diamanti A, Basso MS, Castro M, Bianco G, Ciacco E, Calce A et al. Clinical efficacyand safety of parenteral nutrition in adolescent girls with anorexia nervosa.J Adolesc Health 2008; 42: 111–118.

10 Rigaud D, Brondel L, Poupard AT, Talonneau I, Brun JM. A randomized trial on the efficacy of a 2-month tube feeding regimen in anorexia nervosa: A 1-yearfollow-up study. Clin Nutr 2007; 26: 421–429.

11 Serfaty M, McCluskey S. Compulsory treatment of anorexia nervosa and themoribund patient. Eur Eat Disord Rev 1998; 6: 27–37.

12 Thiels C. Forced treatment of patients with anorexia. Curr Opin Psychiatry 2008; 21:495–498.

Long-term effects of enteral feedingI Nehring et al

176

European Journal of Clinical Nutrition (2014) 171 – 177 & 2014 Macmillan Publishers Limited

13 Cockfield A, Philpot U. Feeding size 0: the challenges of anorexia nervosa.Managing anorexia from a dietitian’s perspective. Proc Nutr Soc 2009; 68:281–288.

14 Herpertz S, Hagenah U, Vocks S, von Wietersheim J, Cuntz U, Zeeck A.The diagnosis and treatment of eating disorders. Dtsch Arztebl Int 2011; 108:678–685.

15 Zuercher JN, Cumella EJ, Woods BK, Eberly M, Carr JK. Efficacy of voluntarynasogastric tube feeding in female inpatients with anorexia nervosa.JPEN J Parenter Enteral Nutr 2003; 27: 268–276.

16 Gueguen J, Godart N, Chambry J, Brun-Eberentz A, Foulon C, Divac Phd SM et al.Severe anorexia nervosa in men: comparison with severe AN in women andanalysis of mortality. Int J Eat Disord 2012; 45: 537–545.

17 Fichter MM, Quadflieg N. Comparing self- and expert rating: a self-reportscreening version (SIAB-S) of the structured interview for anorexic and bulimicsyndromes for DSM-IV and ICD-10 (SIAB-EX). Eur Arch Psychiatry Clin Neurosci2000; 250: 175–185.

18 American Psychiatric Association. Diagnostic and Statistical Manual of MentalDisorders DSM IV. 4th edn (American Psychiatric Association, 2000.

19 Derogatis LR, Lipman RS, Rickels K, Uhlenhuth EH, Covi L. The HopkinsSymptom Checklist (HSCL): a self-report symptom inventory. Behav Sci 1974; 19:1–15.

20 Rosario AS, Schienkiewitz A, Neuhauser H. German height references for childrenaged 0 to under 18 years compared to WHO and CDC growth charts. Ann HumBiol 2011; 38: 121–130.

21 Kromeyer-Hauschild K, Wabitsch M, Kunze D, Geller F, Gei� HC, Hesse V et al.Perzentile fur den Body-mass-Index fur das Kindes- und Jugendalter unterHeranziehung verschiedener deutscher Stichproben. Monatsschrift Kinderheilkunde2001; 149: 807–818.

22 Rosenbaum PR, Rubin DB. Reducing bias in observational studies usingsubclassification on the propensity score. J Am Stat Assoc 1984; 79: 516–524.

23 Robb AS, Silber TJ, Orrell-Valente JK, Valadez-Meltzer A, Ellis N, Dadson MJ et al.Supplemental nocturnal nasogastric refeeding for better short-term outcome inhospitalized adolescent girls with anorexia nervosa. Am J Psychiatry 2002; 159:1347–1353.

24 Herpertz-Dahlmann B, Muller B, Herpertz S, Heussen N, Hebebrand J, RemschmidtH. Prospective 10-year follow-up in adolescent anorexia nervosa--course,outcome, psychiatric comorbidity, and psychosocial adaptation. J Child PsycholPsychiatry 2001; 42: 603–612.

25 Rastam M, Gillberg C, Wentz E. Outcome of teenage-onset anorexia nervosa in aSwedish community-based sample. Eur Child Adolesc Psychiatry 2003; 12: I78–I90.

26 Fichter MM, Quadflieg N, Hedlund S. Twelve-year course and outcome predictorsof anorexia nervosa. Int J Eat Disord 2006; 39: 87–100.

27 Merrill RM, Richardson JS. Validity of self-reported height, weight, and body massindex: findings from the National Health and Nutrition Examination Survey,2001-2006. Prev Chronic Dis 2009; 6: A121.

28 Meyer C, Arcelus J, Wright S. Accuracy of self-reported weight and height amongwomen with eating disorders: a replication and extension study. Eur Eat DisordRev 2009; 17: 366–370.

Supplementary Information accompanies this paper on European Journal of Clinical Nutrition website (http://www.nature.com/ejcn)

Long-term effects of enteral feedingI Nehring et al

177

& 2014 Macmillan Publishers Limited European Journal of Clinical Nutrition (2014) 171 – 177