Olanzapine for Anorexia Nervosa: Weight Effect, Not Anxiety Relief
An outpatient with anorexia nervosa is doing everything asked of her and still not gaining weight, because the obsessional anxiety around food consumes more of the session than the food itself. Olanzapine has real randomized evidence in her population — for the weight, not for the anxiety — which means the effect she would most want is the one the trial specifically failed to find, and the effect it does produce is the one she fears most.
C.M., a 24-year-old woman, freelances as a graphic designer out of a shared studio downtown and just signed her first solo client contract three weeks ago — a milestone she describes almost apologetically, as though it doesn't count next to what her treatment team is actually tracking. She was diagnosed with anorexia nervosa fourteen months ago after a college roommate noticed she'd stopped eating in front of anyone, and has been in structured outpatient treatment — weekly therapy, monthly dietitian visits, a written meal plan — for four months now. She follows the plan to the letter, by her own report and her dietitian's, and still her weight has not moved in six weeks.
The reason isn't defiance. It's that each meal now takes her upward of ninety minutes, most of it spent in a private, exhausting internal negotiation over portion size that she can describe in granular detail but cannot seem to talk herself out of, and by her own account the anxiety has gotten measurably worse, not better, since starting treatment — a pattern she finds confusing and demoralizing, since she expected structure to help. Her weight has been genuinely stable for six weeks despite adherence the dietitian independently corroborates, which is itself informative: this isn't a case of quiet non-adherence hiding behind a compliant story, it's a plateau happening in full view. What the evidence actually supports here is narrower than the reason the team wants to reach for it. Attia and colleagues' 2019 randomized trial — 152 adult outpatients with anorexia nervosa across five North American sites, the largest medication trial ever conducted in this illness — set two co-primary outcomes: rate of change in weight, and rate of change in obsessionality measured on the Yale-Brown Obsessive Compulsive Scale. Olanzapine beat placebo on the first, and the effect was real but modest: BMI rose about 0.26 points per month against 0.10 on placebo. On obsessionality it showed no significant benefit at all. So the drug has controlled evidence for moving the number C.M. is stuck on, and none for relieving the ninety-minute negotiation that is the reason she is stuck on it. That inversion is the whole difficulty: the team would be offering a weight-promoting drug to a patient whose illness is organized around fear of exactly that outcome, on the strength of an effect she has not asked for, while the effect she would actually want is the one the trial specifically failed to find.
Outpatient review, month four
I'd like to offer her olanzapine, and I want to be straight about what for. Attia and colleagues' 2019 trial randomized 152 adult outpatients with anorexia nervosa to olanzapine or placebo and found a real advantage in the rate of weight gain — modest, but it separated from placebo. It did not find a benefit on obsessionality, and I'm not going to pretend otherwise. My argument is that the weight trajectory is the thing that has actually stalled, and this is the one agent with randomized evidence of moving it in exactly her population. If weight starts moving, there's a reasonable chance the obsessional load comes down behind it, because some of that rigidity is itself a feature of being underweight — but that's my clinical reasoning, not something the trial demonstrated.
I want to be precise about what that trial actually showed, because it cuts against us more sharply than you've allowed. Obsessionality wasn't a secondary measure they happened to look at — it was a co-primary endpoint, measured on the YBOCS, chosen in advance precisely because this is the symptom everyone hopes olanzapine will touch. It didn't. The weight effect separated; the psychological effect didn't. So the one thing this drug is demonstrated to do to C.M. is the one thing she has told us directly, more than once, that she is afraid of. If we offer it and she experiences it as being medicated into compliance rather than treated, we could lose four months of built trust over a benefit she never asked for.
You said if weight moves, the obsessional load may come down behind it. That's a reasonable clinical intuition and I've seen it happen — but it's exactly the hypothesis this trial was built to test, and it came back negative. I don't think we get to lean on it as though the evidence were silent rather than unfavorable.
You're right that obsessionality was a co-primary endpoint and that it came back negative — I'm not going to pretend this is stronger evidence than it is, and I'd add that the version of this conversation I've heard most often in practice gets that backwards. But I don't think the conclusion is withholding it. I think it's refusing the comfortable framing. The tempting move here is to tell her this is an anxiety medication and that any weight change is an incidental side effect — that would be much easier to say to a frightened patient, and it would be false. What we can say honestly is narrower: her weight has been flat for six weeks, this is the only agent with randomized evidence of moving that in her population, the effect is modest, and we do not know whether the mealtime anxiety follows. That's a harder sentence to say. It's also the only one that survives her finding out later what the trial actually found.
Agreed: start olanzapine 2.5mg with planned uptitration, and describe it to C.M. as what the evidence actually supports — a modest effect on the rate of weight gain — rather than as a treatment for the mealtime anxiety, which the trial tested directly and did not improve. The team explicitly rejected the more comfortable framing of calling it an anxiety drug: telling a patient a medication targets the symptom she most wants relieved, when the best trial in her population found no such effect, was judged the larger alliance risk, not the smaller one. Metabolic baseline labs already normal; repeat fasting glucose and lipids at 3 months. Weekly therapy and monthly dietitian visits continue unchanged.
Not agreed: how directly to discuss the therapeutic-alliance risk with C.M. herself before starting, versus simply proceeding with the careful framing already planned. The Eating Disorder Therapist wanted to name the alliance concern to C.M. openly — telling her the team debated this specifically because of what she'd shared about her fear — as a way of demonstrating the concern was heard, not just managed around. The Attending Psychiatrist worried that surfacing the team's own internal disagreement might read as uncertainty and undermine confidence in the plan. Left for the Eating Disorder Therapist's next individual session to judge in the moment.