QUEST - Eating Disorders

• Fluoxetine in Anorexia Nervosa

• Overall risk of death w/diagnosis of anorexia nervosa

• Antidepressants for Bulimia Nervosa

• CBT and interpersonal therapy for Bulimia

---

AUTHOR:
Carolina Aponte Urdaneta

Fluoxetine In Anorexia Nervosa

Chairman’s Rounds, September 11, 2006

Citation: Walsh BT, et al. Fluoxetine after weight restoration in anorexia nervosa: a randomized controlled trial.  JAMA. 2006 Jun 14; (22):2605-12

Clinical Question: What is the efficacy of Fluoxetine at preventing relapse or deterioration after initial treatment (weight restoration) in anorexia nervosa?

Background Info: Anorexia Nervosa is a serious psychiatric illness with high morbidity and mortality. There is a high rate of relapse after initial treatment. Antidepressant medications (SSRIs) are often prescribed for these patients since they can exhibit symptoms of OCD and depression. Most controlled trials to date have shown no benefit from medication, despite this 60% of these patients are prescribed SSRIs.

Question Type: Therapy.

Validity Criteria for Appropriate Question Type:

Follow-up: follow-up was complete. All patients who entered the trial were properly accounted for and attributed at its conclusion. Randomization: Patients were appropriately randomized. Intention to treat:  patients were analyzed in the groups to which they were randomized. All randomized patient data was analyzed. Similar groups: groups were similar except for statistically significant difference in BMI for patients receiving placebo (20.16 vs. 20.45 P =.007). Blinding: patients, clinicians and study coordinators were blinded.  Equal treatment: There was equal treatment aside from the intervention. Non intervention treatment for all patients included one week inpatient and 12 months outpatients with 50 sessions of manualized CBT.

Study Design Type: Randomized Controlled Trial

Relevant Criteria for Appropriate Design:

Allocation: Randomized.

Blinding: Double blind. Upon study termination patients, physicians and therapists were asked to guess about assignments and physicians and patients were statistically more likely to guess correctly.

Follow up period: 1 week inpatient and 12 months outpatient.

Setting: New York State Psychiatric Institute and Toronto General Hospital.

Patients/Population: n= 93, women, ages between 16 and 45 year old, who met DSM-IV criteria for anorexia nervosa (except for amenorrhea), who had a BMI of 19 sustained for 2 weeks and had been in a weight restoration program at the study sites. Exclusion criteria: imminent risk of suicide, medical illness, medications. Number screened 150.

Intervention/exposure: Fluoxetine up to 80 mg.

Outcomes: Time to relapse and proportion of patients completing successfully 1 year of treatment. Secondary outcomes: proportion of patients in each group who failed to maintain BMI above 18.5.

Patient Follow Up: 40 completed entire course, 53 terminated prematurely.

Main Results: 

Cox proportional Hazard analysis with covariates for BMI, site and subtype were used. Time to relapse: no significant difference between placebo and fluoxetine groups. (Hazard ratio 1.12, 95% CI: 0.65-2.01) At 52 weeks 45% of the placebo and 43% of the fluoxetine group had not relapsed. Using more or less conservative analyses and omitting patients in the treatment group who did not have detectable levels of Fluoxetine in their blood did not affect results. In secondary analysis, the only significant difference was measured in BAI mean regression per month of -0.77 in fluoxetine vs. -0.22 in placebo group (P=0.07).

Conclusions:  This was well designed study which controlled for several possible biases and was analyzed in several different ways in order to find or exclude any possibility for Fluoxetine being effective in this patient population. The results are negative. In the secondary outcomes tested there was only one statistically significant difference that favored Fluoxetine over placebo which was a higher regression in BAI. The authors comment that this might be attributable to chance only.

Two Teaching Points: 

1. In negative studies confidence intervals can help interpret whether the results are truly negative or not, By looking at the upper end of the interval, if this excludes any important benefit of treatment, you can conclude that the trial is truly negative.

2.  Negative studies fail to exclude the null hypothesis (in this study no difference between placebo and fluoxetine). These studies are less likely to be published than studies that show apparent differences, positive studies may be as much as three times more likely to be published than negative studies, they also face an increased delay in publication.

Two Questions for Discussion:

1. What other pharmacological interventions should be studied in Anorexia at this point?
2. Would you prescribe SSRI to your anorectic patient after seeing this data?

Return to top of the page

---

Clinical Question: Is the overall risk of death increased in patient with a diagnosis of anorexia nervosa?

Korndorfer et al., Long-term survival of patients with anorexia nervosa: a population-based study in Rochester , Minn. Mayo Clin Proc. 2003;78:278-284.

Methods

-All hospitals in Rochester use the same record system
-208 residents identified who had lived in the community for at least 1 year when they presented with anorexia nervosa for the first time 1935 through 1989.
-Diagnostic indices included weight loss, amenorrhea, etc. (p279)
-Categorized into definite, probable or possible anorexia 
-Criteria
1.        self imposed wt loss of 15% from prior wt (or <15% if still growing)
2.        amenorrhea (2 cycles) or decreased testosterone and libido 
3.        psychological fear of fatness or loss of control of eating
4.         no physical illness to account for this
Definite met all 4 criteria
Probable met ³/ including 1 and 4
Possible met 1 and 4 (2 and 3 unknown)
One author also reviewed patient records and used clinical judgment
-Followed up until death or most recent clinical contact
-Vital status updated through commercial sources and patient letters - Termination date June 30,2000
-Looked at cause of death
-Kaplan-Meier curve for survival curve
-Used an age- and sex- matched cohort from the Minnesota white population 
-1-sample Gehan generalized Wilcoxon test to compare survival with cohort -Log rank test to compare survival
-Age and sex-specific death rates for the US population were used to calculate the expected numbers

Results
-208 met criteria (193 women 15 men) from 1935 to 1989
-Age 10 to 57, with median age of 19 and mean 21.5
-82 (39% definite), 92 (44% probable),34 (16%) possible
 -At last contact 17 patients died (23.7 expected from cohort)
-30 y after the study 93% (95% CI 88%-97%) compared to expected 94% SMR (standardized mortality ratio) 1.30

Strengths
Unique shared patient record
 No difference in survival between dx in first 30 years and last 25 96% followed 10 years
Study cohort followed for 92% of possible years (5646/6106) -Mean duration off/u 27.1 years, median 33, range 1 d to 63 y

Weaknesses
White patients only
Very small cohort
Community population
 Most had never seen a psychiatrist, few hospitalized. SMR is extremely low in this paper

Return to top of the page

Antidepressants vs. Psychological Treatments & their Combination for Bulimia Nervosa:  A Systematic Review

Ana Carla P. Smith, Chair's Rounds, April 26, 2004

Bottom Line: psychological treatment, is as effective as antidepressants for bulimia nervosa, but a combination is best.

Clinical Question: EP is a 19 y/o WF with a h/o bulimia, admitted to WW w/ sx of depression, s/p suicide gesture (narcotic OD). Ms. P. is treatment naYve, and needs tx for her eating disorder. In female patients w/ bulimia nervosa, is treatment with antidepressants, psychotherapy, or a combination thereof most effective for decreasing purging behavior?

Reference: Bacaltchuk, J. Hay, P. Trefiglio, R. Antidepressants versus psychological treatments and their combination for bulimia nervosa. Cochrane Depression, Anxiety and Neurosis Group. Cochrane Database of Systematic Reviews. 1, 2004.

Validity:
Did the overview explicitly address a sensible question? - yes
This was a systematic review of all RCTs comparing antidepressants with psychological approaches or comparing their combination with each single approach for the treatment of bulimia nervosa. All people w/ BN, diagnosed by Russell's, DSM, or ICD-10 criteria were included. They were mostly young adult females, with similar behavioral and demographic features, and a similar number of baseline bulimic episodes per week. Studies measured full remission (main outcome), reduction (either 100% or 50% in binge or purge episodes) or average difference in bulimic sx at endpoint (outcomes analyzed separately). 
Was the search for relevant studies detailed & exhaustive? - yes
Every RCT in which antidepressants were compared w/ psychological tx's or the combination was compared to each tx alone to reduce the sx of bulimia was included. Multiple electronic databases (MEDLINE, EMBASE, LILACS, PsycLIT, Cochrane database) were searched. Pharmaceutical companies were contacted, hand searches were performed, and various book chapters were reviewed.
Were the primary studies of high methodologic quality - yes
Included trials were those w/ low (adequate allocation concealment) or moderate (some doubt about the results) risk bias (category A or B in the Cochrane Handbook). ITT analysis was performed; dropouts were assigned to worst possible outcome (no remission or improvement in sx).
Were the assessments of the included studies reproducible? - yes
The quality of the eligible trials and data from subsequently included studies was extracted independently by two reviewers. Trials were rated independently, and any disagreements were resolved by discussion & consensus.

Results:
Comparison I - antidepressants vs psychological treatments­
5 trials included. Remission reported in all trials. Remission rates = 20% for single
antidepressants vs 39% for single psychotherapy (RR 1.28; 95% CI = 0.98 - 1.67). Dropout rates = 40% for antidepressants vs 18% for CBT (RR 2.18; 95% Cl 1.09 - 4.35. NNH = 4 over 17.5 weeks). Clinically relevant but not statistically significant difference favoring psychological treatments.

Comparison II - antidepressants vs combination -

5 trials included. Remission reported in 4 trials. Remission rates = 42% for combination vs 23% for antidepressants (RR = 1.40; 95% CI = 0.98 - 1.99; NNT = 5). Dropout rates HIGH = 41 % for antidepressants vs 34% for combination (RR 1.19; 95% Cl 0.69-2.05).

Comparison III - psychological treatments vs combination -

7 trials included. Remission rates = 36% for psychological approaches vs 49% for combination (RR = 1.21; 95% CI = 1.02­1.45; NNT = 8). Dropout rates = 16% for psychological vs 30% for combination (RR 0.57; 95% CI 0.38-0.88; NNH for 14 wks = 7; statistically significant).

Comments / Application:

How can I best interpret the overview's results to apply them to the care of my patients?
Combination tx's are superior to psychotherapy alone. This was the only statistically significant difference between treatments using the most conservative statistical approach (failsafe N calculation).
CBT = gold standard for tx of bulimia. Medication proposed as a supplement to education, nutritional rehab & psychotherapy. When adding meds, it may be a good idea to implement measures to prevent dropouts.
In light of the results of this trial, combination tx may be recommended as first-line tx for BN, especially when sx are severe and a single approach has only been partially successful.
CBT alone is generally considered superior to antidepressants alone; antidepressants may be considered first-line if a skilled psychotherapist is not available.
Reasons to not use meds as exclusive tx = overall effectiveness of psychological interventions is good; CBT seems more acceptable to patients (lower dropout rate); long term maintenance change is better w/ CBT (relapse after d/c of meds is high). 
Will results change my management? Yes.
In my patient's case, antidepressant therapy was indicated to treat sx of depression. Ms. P was scheduled for initial eval at the Duke eating disorders clinic, and will likely receive psychotherapy.
Will patients be better off as a result of the study? Yes.
Hopefully, patients with BN will receive the psychotherapy or combination therapy they need.

Return to top of the page

Sarah Parker, 3/7/05, Chairman’s Rounds

Bottom Line:  Limitations in this study specifically prevail which must be remembered as stated below. Yet, enough evidence exists to reveal one way bulimia could be treated in a residency setting with the use of guided self-help manuals. Perhaps the best combo is guided self-help along with individual CBT. Group CBT remains to be controversial as to specific types of bulimia if this is the sole treatment, but again mostly is better than placebo.

 Background: Individual CBT is the treatment of choice for bulimia along with possibly interpersonal therapy. However, controversy has prevailed about group therapy as it was initiated. Some studies have shown it can make a specific subtype worse if used alone. Most studies have shown improvement. A self-care manual would be useful for the treatment of eating disorders to cut down on costs and to use the limitations of insurance for treatment. Randomized trials have shown self-help manuals to be effective, and one English version of the self-help manual showed 8week CBT along with 8week self-help was equivalent to 16 weeks of individual CBT. This study is going to look at this theory within Austria .
Reference: Bailer, Ursula, et al. Guided Self-Help Versus Cognitive-Behavioral Group Therapy in the Treatment of Bulimia Nervosa. Int J Eat Disord 35:522-537, 2004.
Methods: Subjects: Family physicians, psychiatrists, counselors, and self-referrals could be referred to the study if age 17 and older. Held at Department of Genral Psychiatry at the University Hospital of Psychiatry in Vienna . Inclusion criteria: Met DSM-IV criteria for bulimia using SCID by 1 of 4 physicians (who did not end of treating the pt) Exclusion criteria: 1)medically unstable or 2)suicidal threat. Participants could be on SSRI only if unchanged for prior 3 months. 81 ended up signing consent and 41 assigned randomly but not at the same time to either group CBT or self-help with guidance. Treatment: Guidance was performed by 18 weekly sessions with three female residents in their first or second year in psychiatry. Group CBT was administered in three steps by two female therapists and one co-therapist. Participants completed study if went to 50% of sessions and group CBT also had 18 sessions but of 90 minute duration (8-12 in group)
Assessment:self-report, EDQ+EB-IV+BDI at weeks 1 10 and 18 and 1 year follow-up, along with SCID DSM-IV interview, BMI, ht and wt, vital signs Analysis: Ancova, intention to treat analysis, mixed-effect regression analysis to compare changes over time by condition controlling for a variety of variables
Validity: Strengths: randomized but not at the same time, groups apparently not treated differently significantly outside of study, study population is generarizable although it is in Vienna! It is a situation we often get presented with in real practice.
Weaknesses:   questionable selection bias vs real-world referral situation, Unsure on blinding, treatment groups not similar at baseline in lifetime major depression, current MD, medication, and hx suicide attempts (HOWEVER they did control for this in the analysis), low sample size, 48 of 81 total ended up completing study, is threshold of 50% completion adequate?
Results: (see back of sheet for compliance and attrition)  ITT analysis 7.5% is Low.

Completer Sample:
Self-Help                  Group CBT
End of treatment (Week 18)
Rec: 1 (4.3%)           3 (12%)     
Rem: 13 (56.5%)      9 (36%)    
1 Year Follow-Up
Rec: 7 (30.4%)           3 (12%)     
Rem: 17 (73.9%)       11 (44%)    

Comments: Many weaknesses in this study, but seems productive enough to replicate with a larger sample size to prove reliability.  The authors point all these out as well and recognize the weaknesses. The theory makes sense, has been proven in other studies, and has potential given the reduction in health care coverage for eating disorders. Due to low recovery rates overall these numbers are good and more to add to the literature.

 Return to top of the page

 

---