Starting an SSRI During Active Malnutrition in Anorexia Nervosa
A 17-year-old with restricting-type anorexia nervosa is acutely depressed and underweight at the same time. The question isn't whether her mood deserves treatment — it's whether an SSRI can even work on a brain this malnourished, or whether starting one now just adds a drug with no real chance of helping.
R.N., a 17-year-old girl, used to be the one in her friend group who organized study sessions and remembered everyone's birthdays; her mother says she has watched that version of her daughter disappear over the past school year, replaced by someone who cries in her room most evenings and has stopped answering texts from the same friends. She plays clarinet in the school orchestra and made it to regional auditions last spring, but hasn't picked up the instrument since September. The restriction started, as it often does, with a summer of "eating healthier" before cross-country season — six weeks in, that had become skipped lunches, then skipped breakfasts, then a rigid list of foods she would allow herself, none of it framed by her as a problem until her pediatrician's routine physical caught a resting heart rate in the 40s.
Six weeks into structured outpatient treatment — family-based meal support, weekly weigh-ins, a dietitian — her weight has moved in the right direction, but slowly, and her mood has not moved at all. She meets criteria for a major depressive episode that predates the restriction by her own timeline, not just a byproduct of hunger: anhedonia, sleep disruption, a stated wish some nights that she "didn't have to wake up," though she denies any current plan or intent. Her PHQ-9 today scored 19, essentially unchanged from intake. The question in front of the team is not whether that deserves treatment — it clearly does — but whether starting an SSRI in a brain this malnourished has any real chance of working, or whether it is a drug added to a chart that cannot yet respond to it. Starvation measurably alters central serotonergic function: reduced dietary intake lowers plasma tryptophan, the amino-acid precursor SSRIs depend on the brain having enough of to begin with, and the principal placebo-controlled trial of fluoxetine in this setting — Attia and colleagues, 1998, a seven-week randomized study of 31 hospitalized women all weighing under 80% of ideal body weight — found no advantage over placebo on weight gain or on measures of psychological state. It is a small trial, and the only one of its kind, which is precisely why it is both the best evidence available and thin evidence to hang a decision on. The trial that argues against prescribing measured group averages in weight and mood; it never enrolled a question about what to do for the one patient in it who says some nights she doesn't want to wake up. A null result and an absent answer are not the same finding, and the case for waiting rests on treating them as though they were.
Outpatient team meeting, six weeks into treatment
I want to start fluoxetine today. Her mother's timeline is clear — the low mood came first, the restriction followed months later — and six weeks of behavioral treatment alone hasn't moved her PHQ-9 by a single point. She's telling us some nights she doesn't want to wake up. That's not a finding I'm comfortable sitting on while we wait for a weight number that could be months away.
I don't disagree that her depression is real and independent of the eating disorder — I'm not contesting the diagnosis. What I'm not convinced of is that fluoxetine can actually do anything for it at 79% of median BMI. Attia and colleagues randomized actively malnourished inpatients with anorexia nervosa to fluoxetine or placebo and found no separation on weight or mood outcomes. The proposed reason isn't that the diagnosis was wrong — it's that starvation lowers circulating tryptophan, the amino acid the brain converts into serotonin in the first place, so an SSRI has less substrate to work with regardless of dose. Plainly: the drug can only boost the signal that's already there, and there's real evidence that signal is turned down by hunger itself. We'd be adding QTc and GI risk to a chart for a drug that may not have a mechanism to fire yet.
I'd also gently push back on treating "we're doing something" as itself a reason to prescribe — a medication that can't work isn't a safety net, it's a false one.
You're right that the tryptophan mechanism is real and that Attia's trial found nothing at the group level — I'm not disputing the pharmacology. But that trial studied a population as a whole; it wasn't designed to tell us what to do with one specific 17-year-old who is telling us, this week, that she doesn't want to wake up some nights. A null finding across a starved-inpatient average doesn't mean the drug definitely can't help her individually, and passive suicidal ideation is exactly the kind of severity marker that should move a decision independent of what the average non-suicidal malnourished patient's drug response looks like.
My actual proposal: start it, but frame it honestly with her and her family as a trial with real uncertainty about whether it will work at her current weight, layer in more frequent safety check-ins rather than waiting for the next scheduled visit, and don't treat the drug as a substitute for continuing to push weight restoration, which remains the strongest lever we have regardless of what fluoxetine does or doesn't do.
Agreed: start fluoxetine 10mg with a plan to uptitrate as tolerated, continue family-based treatment and weekly dietitian visits unchanged, and move safety check-ins to every two weeks rather than monthly given the passive suicidal ideation. The family was told directly, not softened, that the medication's chance of working at her current weight is genuinely uncertain per the existing evidence, and that weight restoration itself remains the primary treatment regardless of how the SSRI trial goes.
Not agreed: whether the plan should specify a weight or interval at which to reassess and potentially discontinue if there's no response, or whether that decision should be made openly later once more is known. The Clinical Pharmacologist wanted a stated stopping point now, to avoid the drug quietly becoming permanent by default if it never clearly helps; the Attending Psychiatrist preferred to leave that open, worried that pre-committing to a stop date could read to the family as less than full effort on the depression. Both agreed to revisit explicitly at the six-week mark.