Pharmacotherapy

TL;DR — No medication is established as a stand-alone treatment for AN, and the US Food and Drug Administration has approved no psychoactive treatment for the disorder at all (Cassioli 2020, PMID 32448045). Olanzapine has the most replicated weight-gain signal — a 152-patient RCT found BMI rising 0.259 kg/m² per month on drug versus 0.095 on placebo — but no corresponding benefit on obsessionality, and a meta-analysis pooling the five BMI-reporting trials found no significant olanzapine–placebo difference (Attia 2019, PMID 30654643; Cassioli 2020, PMID 32448045). Antidepressants may be used for comorbidity but have not reliably restored weight or treated core acute AN. Two later attempts to widen the pharmacological repertoire — intranasal oxytocin and olanzapine in young people — returned a null result and a failed feasibility study respectively (Maguire 2024, PMID 38520886; Filiz 2026, PMID 42116676). Medication evidence is adjunctive to nutrition, psychotherapy and medical care.

Evidence map

Class	Target/rationale	Evidence conclusion	Key risk/context
Olanzapine	Weight gain, anxiety/rumination	Modest weight signal; core-symptom benefit inconsistent	Sedation, metabolic effects, orthostasis, QT/drug context
Other antipsychotics	Rigidity/anxiety/weight	Sparse and inconsistent	Class adverse effects; very small trials
SSRIs/other antidepressants	Depression/anxiety/relapse	No established acute core-AN or weight benefit	Malnutrition may affect response/tolerability
Mood stabilizers	Comorbidity/impulsivity	Insufficient AN-specific evidence	Organ/electrolyte and pregnancy considerations
Intranasal oxytocin	Anxiolysis around eating; oxytocinergic perturbation in AN	Negative: 61-inpatient phase II RCT found no drug–placebo difference on ED psychopathology or BMI at any timepoint	Pilot findings did not replicate (Maguire 2024, PMID 38520886)
Dronabinol and other appetite agents	Weight/metabolic target	Described as a promising but unapproved study result alongside olanzapine; no approved pharmacological option exists for AN (Himmerich 2021, PMID 32858054)	Weight change is not complete recovery

Olanzapine trials

The 2008 double-blind, 10-week flexible-dose trial randomized 34 adult women to olanzapine or placebo alongside day-hospital treatment and reported a greater rate of weight increase, earlier attainment of target BMI and a greater rate of decrease in obsessive symptoms, with no difference in adverse effects; its size and co-intervention made replication essential, as the authors said explicitly (Bissada 2008, PMID 18558642). A two-site outpatient trial then randomized 23 participants to 8 weeks of olanzapine or placebo with medication-management sessions: end-of-treatment BMI adjusted for baseline was significantly greater on olanzapine (F(1,20)=6.64, p=0.018), psychological symptoms improved in both groups without between-group differences, 17 of 23 (74%) completed, and sedation was the only frequent side effect (Attia 2011, PMID 21426603).

The definitive test was a five-site North American RCT of 152 adult outpatients (96% women, mean BMI 16.7) randomized 1:1 to olanzapine or placebo and seen weekly for 16 weeks. The treatment-by-time interaction was significant for weight — BMI rose 0.259 kg/m² (SD 0.051) per month on olanzapine versus 0.095 (SD 0.053) on placebo — but the Yale-Brown Obsessive Compulsive Scale obsessions subscale did not differ (−0.325 vs −0.017 points per month), and there were no between-group differences in metabolic blood abnormalities (Attia 2019, PMID 30654643). Olanzapine is accordingly described as the first medication to safely promote weight gain in AN, a narrower claim than "effective for AN" (Muratore 2022, PMID 35576089).

In young people the evidence is limited and negative. A 20-participant placebo-controlled pilot included females aged 12.3–21.8 years and found no olanzapine advantage for weight or psychological outcomes (Kafantaris 2011, PMID 21663423). The later OPEN study set out to test feasibility for a future RCT of olanzapine in 55 patients aged 12–24 with AN or atypical AN who had gained <2 kg after at least a month of treatment as usual. It pre-screened 52 people, found 35 eligible, recruited 20, and retained 15 on olanzapine for ≥16 weeks; weight and BMI rose and eating-disorder psychopathology fell on average, but the study missed its recruitment target and reported low adherence (Filiz 2026, PMID 42116676). A one-armed open-label study that under-recruits cannot estimate effect.

Antidepressants: a clear negative and a possible subgroup

The definitive antidepressant trial in AN was a relapse-prevention design, not an acute one. Ninety-three women with AN were weight-restored to a BMI of at least 19.0 in intensive inpatient or day-programme care at two sites, then randomized double-blind to fluoxetine or placebo with individual CBT for up to one year. The proportion maintaining BMI ≥ 18.5 and remaining in the study for 52 weeks was 26.5% on fluoxetine versus 31.5% on placebo (p = .57), and time to relapse did not differ (HR 1.12, 95% CI 0.65–2.01, p = .64) after adjustment for pre-randomization BMI, site and subtype (Walsh 2006, PMID 16772623; NCT00288574). The authors' conclusion was that the study "failed to demonstrate any benefit from fluoxetine".

A 2025 secondary analysis reopens the question narrowly. Dichotomizing the 92 randomized participants by Beck Depression Inventory score at randomization (BDI > 20 indicating moderate/severe depression, n = 26), there was a significant medication × depression-severity interaction for time to relapse (HR 0.46, 95% CI 0.25–0.85, p = .01), and depression severity modified fluoxetine's effect on the trajectory of depressive (β = −0.27, 95% CI −0.42 to −0.12, p = .001) and bulimic (β = −0.15, 95% CI −0.25 to −0.05, p = .004) symptoms over 12 months. Fluoxetine was more effective than placebo at preventing relapse among the more depressed weight-restored patients (LLoyd 2025, PMID 41022691). This is a post-hoc subgroup analysis of a negative trial with 26 patients in the high-depression cell; the authors state it requires replication. It is hypothesis-generating, not a basis for an AN-specific prescribing recommendation.

The same cohort produced a finding that shapes when any relapse-prevention agent should be given. Modelling the daily hazard of relapse after full weight restoration showed risk rising immediately, peaking at approximately 60 days, then declining gradually with no inflection point at which it fell sharply (Walsh 2021, PMID 34154394). The first two months after acute care are the window in which relapse-prevention treatment has to work.

Earlier antidepressant work is mostly of historical interest but contains one durable observation. In 72 inpatients randomized to cyproheptadine, amitriptyline or placebo, cyproheptadine had only a marginal overall effect on days to normal weight — but it significantly improved treatment efficiency in non-bulimic patients and significantly impaired it in bulimic patients relative to both comparators, which the authors described as the first pharmacological evidence for the validity of the AN subtypes (Halmi 1986, PMID 3511877). Subtype-by-drug interaction has not been systematically pursued since.

Antipsychotics beyond olanzapine, and olanzapine in adolescents

Trial	Population	Result
Kafantaris 2011, PMID 21663423	20 underweight adolescent/young-adult females, restricting AN, 10 weeks, adjunctive olanzapine vs placebo in a comprehensive programme (15 completers; NCT00592930)	No difference in change in %median body weight at week 5 or 10; both groups gained similarly and improved similarly on eating attitudes, psychological functioning and resting energy expenditure. A trend of rising fasting glucose and insulin appeared only in the olanzapine group
Hagman 2011, PMID 21871373	40 females aged 12–21, risperidone (n=18, mean 2.5 mg) vs placebo (n=22), mean 9 weeks, during weight restoration (NCT00140426)	No benefit for weight or for most scales. EDI-2 Drive for Thinness fell more on risperidone over the first 7 weeks (effect size 0.88, p = .002) but the difference was not sustained to study end (p = .13); Interpersonal Distrust improved more on risperidone (effect size 0.60, p = .03); prolactin rose (p = .001)
Andries 2014, PMID 24105610	25 women with AN of ≥5 years' duration, randomized double-blind crossover, dronabinol 2.5 mg twice daily for 4 weeks vs placebo with 4-week washout (24 completers)	Weight gain of 0.73 kg above placebo (t = 2.86, df = 22, p < 0.01) with no significant psychotropic adverse events; EDI-2 subscale scores unchanged

Two things follow. First, olanzapine's weight signal in adults (Attia 2019, PMID 30654643) has never been reproduced in a randomized trial in young people — the adolescent pilot was null and the only subsequent attempt was a feasibility study that under-recruited (Kafantaris 2011, PMID 21663423; Filiz 2026, PMID 42116676). Second, dronabinol's 0.73 kg over four weeks in long-duration illness is a real randomized result and also an illustration of the ceiling: no agent in this table changed eating-disorder psychopathology.

What the drug evidence has in common

Every positive pharmacological finding in AN is a weight or physiological finding. No trial has shown a medication improving the cognitive core of the disorder — weight/shape overvaluation, fear of gain, drive for thinness — beyond transient or unsustained signals (Hagman 2011, PMID 21871373; Attia 2019, PMID 30654643; Andries 2014, PMID 24105610; Maguire 2024, PMID 38520886). That pattern is consistent with drugs acting downstream of nutrition rather than on the illness, and it is why "no approved agent for AN" (Muratore 2022, PMID 35576089) is a description of the science rather than a regulatory accident. Leptin replacement is the one candidate whose rationale targets the starvation-driven mental state directly, and its human evidence is a three-patient open case series (Milos 2020, PMID 32855384); see neurobiology and cognition.

Meta-analytic context

Review	Corpus	Conclusion
Cassioli 2020	19 acute-phase RCTs; 8 meta-analyzed	Five BMI-reporting studies showed no significant olanzapine–placebo difference; no significant psychopathology, depression or anxiety result across molecules (PMID 32448045)
Fornaro 2023	5,122 records screened, 203 full texts reviewed, 62 studies in qualitative synthesis (AN=22), 22 meta-analysed (AN=9)	Heterogeneous pharmacotherapies and eating-disorder diagnoses; the AN evidence within it is nine trials, which is why interpretation requires AN-specific subgroup reading (PMID 37393954)
Himmerich 2021	Pharmacological review	No established medication replaces nutritional/psychological treatment (PMID 32858054)
Muratore 2022	Narrative review across AN, BN, BED	Olanzapine is the first medication to safely promote weight gain in AN; by contrast fluoxetine is FDA-approved for BN and lisdexamfetamine for BED — AN remains the eating disorder with no approved agent (PMID 35576089)

Different meta-analytic conclusions can reflect inclusion dates, populations, outcomes and models. A positive weight slope in one adequately powered trial and a null pooled estimate from sparse heterogeneous trials are not logically incompatible.

Oxytocin: a well-powered negative

Perturbation of the oxytocinergic system in AN, and pilot findings that intranasal oxytocin reduced anxiety and eating-related concern, made it one of the more mechanistically motivated candidates. A multi-site phase II RCT gave 61 female inpatients (mean age 24.4, SD 7.9; mean BMI 16.24, range 11.43–18.55) 18 IU of intranasal oxytocin or placebo twice daily for four weeks during nutritional rehabilitation, with EDE psychopathology as the primary outcome and BMI secondary, assessed pre-treatment, post-treatment and at 6 months. There were no significant differences between oxytocin and placebo at any timepoint on any outcome of interest, and oxytocin did not reduce anxiety ratings or salivary cortisol in response to a high-energy snack; both groups improved on almost all psychological parameters over time (Maguire 2024, PMID 38520886). The authors attribute non-replication of the pilot to variance, higher baseline psychopathology and possible ceiling effects. This is the largest randomized placebo-controlled trial of repeated-dose intranasal oxytocin in AN, and it is negative.

What is actually prescribed

The gap between the evidence and practice is measurable at national scale, and it is large. A Danish register study identified every medication prescription from two years before to five years after a first AN diagnosis, for all patients diagnosed 1998–2011 against age- and gender-matched controls (1:10). The odds of receiving medication were increased for every class except a small residual category. The highest were alimentary-tract medication (OR 2.8, p < 0.001), peaking one year before first diagnosis, and psychopharmacological medication (OR 5.5, p < 0.001), peaking one year after. Older patients had higher odds across almost all classes, with a fivefold OR for cardiovascular medication; patients with psychiatric comorbidity had a threefold OR for psychotropics against those without. Calendar year made little difference (Clausen 2023, PMID 36379077). The pre-diagnosis alimentary peak may reflect gastrointestinal symptoms or diagnostic delay, but prescription timing alone cannot distinguish those explanations.

A follow-up study in 7,654 Danish patients stratified by the register-based severity index found that individuals with severe AN were more likely than those with less-severe AN to be prescribed alimentary-tract drugs (OR 1.4, 95% CI 1.3–1.6), cardiovascular drugs (OR 1.1, 1.0–1.3), analgesics (OR 1.2, 1.1–1.3) and psychotropics (OR 2.4, 2.1–2.7) — and the pattern persisted among individuals with no diagnosed comorbidity at all. Latent class analysis found five distinct comorbidity profiles within the severe group, each prescribed more than the comorbidity-free cluster (Zhang 2026, PMID 41688100). The authors' conclusion is the one this page has been building toward: widespread, prolonged prescribing continues in a disorder with no approved agent, and evidence-based pharmacotherapy guidelines are urgently needed.

A systematic scoping review of second-generation antipsychotics in AN quantifies how uneven the underlying evidence is. Across 55 articles published 2000–2022 (48 on effectiveness, 7 on tolerability alone), olanzapine had 7 randomized double-blind controlled trials; quetiapine had 2, risperidone had 1, and aripiprazole had none. The drugs were well tolerated with mild, transient adverse effects in this somatically vulnerable population, but the review concluded that the studies do not permit conclusive, reliable or reproducible findings on efficacy — while noting that international guidance nonetheless suggests olanzapine and aripiprazole may be of interest in severe or first-line-resistant situations (Thorey 2023, PMID 36928656). Aripiprazole is therefore recommended in some guidance on zero randomized evidence.

Comorbidity and safety

Treating depression, anxiety or OCD symptoms may be appropriate on their own merits, but symptom timing relative to starvation matters. Medication selection should account for bradycardia, orthostasis, electrolyte disturbance, QT-prolonging combinations, hepatic/renal function, seizure risk and pregnancy potential. Abruptly stopping established psychotropics can also cause harm; this page does not provide individual prescribing advice.

Open questions

• Which phenotype, if any, predicts clinically meaningful olanzapine response? The 152-patient trial reported an average trajectory difference, not responder identification (Attia 2019, PMID 30654643).
• Does medication-assisted weight gain translate into durable remission or quality of life? No olanzapine trial in AN has followed participants beyond the acute treatment window with remission as an endpoint (Attia 2019, PMID 30654643; Cassioli 2020, PMID 32448045).
• Can a feasible recruitment and adherence model be built for drug trials in young people with AN, given that a study designed for 55 participants recruited 20 and retained 15 (Filiz 2026, PMID 42116676)?
• Can metabolic-genetic findings yield a mechanism-based drug without pathologizing weight alone (Watson 2019, PMID 31308545)?
• Does the depression-moderated fluoxetine effect replicate? It rests on 26 patients in the high-depression cell of a trial that was negative overall (Walsh 2006, PMID 16772623; LLoyd 2025, PMID 41022691).
• Should relapse-prevention pharmacotherapy be concentrated in the first two months after weight restoration, where the modelled hazard of relapse peaks (Walsh 2021, PMID 34154394)?
• Does the cyproheptadine subtype interaction — benefit in non-bulimic and harm in bulimic AN — reflect a real biological division, and why has no trial since 1986 tested drug × subtype (Halmi 1986, PMID 3511877)?
• Why does olanzapine's adult weight signal not appear in adolescents, and is the difference developmental or a matter of sample size and concurrent programme intensity (Attia 2019, PMID 30654643; Kafantaris 2011, PMID 21663423)?

Related pages

• Treatment in adults — primary specialist therapies.
• Medical complications — medication-risk context.
• Clinical trials landscape — registered drug studies.

References

1. Cassioli E, et al. Pharmacological treatment of acute-phase anorexia nervosa: evidence from randomized controlled trials. J Psychopharmacol. 2020. PMID 32448045.
2. Bissada H, et al. Olanzapine in low body weight and obsessive thinking in women with anorexia nervosa. Am J Psychiatry. 2008. PMID 18558642.
3. Attia E, et al. Olanzapine versus placebo for out-patients with anorexia nervosa. Psychol Med. 2011. PMID 21426603.
4. Attia E, et al. Olanzapine versus placebo in adult outpatients with anorexia nervosa. Am J Psychiatry. 2019. PMID 30654643.
5. Fornaro M, et al. Psychopharmacology of eating disorders: systematic review and meta-analysis of RCTs. J Affect Disord. 2023. PMID 37393954.
6. Himmerich H, et al. Pharmacological treatment of eating disorders and comorbid problems. Pharmacol Ther. 2021. PMID 32858054.
7. Muratore AF, Attia E. Psychopharmacologic management of eating disorders. Curr Psychiatry Rep. 2022. PMID 35576089.
8. Watson HJ, et al. GWAS identifies eight risk loci and implicates metabo-psychiatric origins. Nat Genet. 2019. PMID 31308545.
9. Maguire S, et al. A phase II randomised controlled trial of intranasal oxytocin in anorexia nervosa. Psychoneuroendocrinology. 2024;164:107032. PMID 38520886.
10. Filiz ES, et al. Olanzapine for young people with anorexia nervosa: a synopsis of the main results from the OPEN open-label feasibility study. Health Technol Assess. 2026;30:1-17. PMID 42116676.
11. Walsh BT, et al. Fluoxetine after weight restoration in anorexia nervosa: a randomized controlled trial. JAMA. 2006;295:2605-2612. PMID 16772623. (NCT00288574)
12. LLoyd EC, et al. Fluoxetine for anorexia nervosa after weight restoration: moderation of effect by depression. Psychol Med. 2025;55:e290. PMID 41022691.
13. Walsh BT, et al. Time course of relapse following acute treatment for anorexia nervosa. Am J Psychiatry. 2021;178:848-853. PMID 34154394.
14. Halmi KA, et al. Anorexia nervosa: treatment efficacy of cyproheptadine and amitriptyline. Arch Gen Psychiatry. 1986;43:177-181. PMID 3511877.
15. Kafantaris V, et al. A placebo-controlled pilot study of adjunctive olanzapine for adolescents with anorexia nervosa. J Child Adolesc Psychopharmacol. 2011;21:207-212. PMID 21663423. (NCT00592930)
16. Hagman J, et al. A double-blind, placebo-controlled study of risperidone for the treatment of adolescents and young adults with anorexia nervosa: a pilot study. J Am Acad Child Adolesc Psychiatry. 2011;50:915-924. PMID 21871373. (NCT00140426)
17. Andries A, et al. Dronabinol in severe, enduring anorexia nervosa: a randomized controlled trial. Int J Eat Disord. 2014;47:18-23. PMID 24105610.
18. Milos G, et al. Short-term metreleptin treatment of patients with anorexia nervosa. Transl Psychiatry. 2020;10:303. PMID 32855384.
19. Clausen L, et al. Pharmacotherapy in anorexia nervosa: a Danish nation-wide register-based study. J Psychosom Res. 2023;164:111077. PMID 36379077.
20. Zhang ZP, et al. Medication use in severe anorexia nervosa: a Danish register-based study. BMJ Ment Health. 2026;29:e301894. PMID 41688100.
21. Thorey S, et al. Efficacy and tolerance of second-generation antipsychotics in anorexia nervosa: a systematic scoping review. PLoS One. 2023;18:e0278189. PMID 36928656.