Terminal Delirium: Antipsychotic, Benzodiazepine, or Neither
A hospice patient's terminal delirium is frightening her family more than it seems to be distressing her, and the trial that supposedly settled how to treat it found the standard drugs made people worse — a finding the team has to decide whether it actually applies to this patient's more severe presentation.
M.T. moved into her daughter's house eleven months ago, when her kidneys finally failed past the point either of them was willing to fight with dialysis, and the two of them turned the downstairs sewing room into a hospice bedroom together, curtains and all. For ten days that arrangement worked exactly as intended — M.T. lucid, comfortable, visited daily by grandchildren she still recognized. On day eleven she stopped recognizing the room. She is picking at the bedsheets, calling out names of people not present, and twice tonight has tried to climb out of bed convinced she needs to catch a bus that stopped running decades ago.
Her daughter, listening from the kitchen, wants this stopped, and it's a reasonable thing to want — but the actual evidence on how to stop it runs against the reflexive answer. Agar and colleagues' 2017 randomized trial in JAMA Internal Medicine tested risperidone and haloperidol against placebo for delirium symptoms in palliative care inpatients and found that both antipsychotics performed worse than placebo rather than merely no better: delirium symptom scores at 72 hours were higher in the risperidone and haloperidol arms than in the placebo arm, both active arms had more extrapyramidal effects, and patients in the haloperidol arm had worse overall survival than placebo, at a hazard ratio of 1.73. That trial has functionally reversed the standing assumption that antipsychotics are simply the correct first move here. But its population skewed toward mild-to-moderate distress, and M.T.'s presentation — genuinely frightened, physically trying to leave the bed, not just restless — sits further along that severity range than most of the patients that trial actually enrolled.
The hospice nurse who screened her earlier tonight already ruled out the reversible causes worth checking before reaching for any sedating agent at all — no fecal impaction on exam, a bladder scan negative for retention, no obvious undertreated pain given her morphine was adjusted only three days ago and has otherwise held her comfortable. That negative workup is itself informative: it means tonight's agitation isn't a fixable trigger sitting underneath the delirium waiting to be found, which is exactly the situation the Agar trial's own patients were mostly in too, and exactly why its finding is hard for this team to simply set aside.
In the living room, with her daughter listening from the kitchen
I don't think we should start an antipsychotic. The Agar trial is the best randomized evidence we have on exactly this question — risperidone and haloperidol against placebo in palliative-care patients with delirium — and it found both drugs made delirium symptom scores worse than placebo rather than better, both increased extrapyramidal symptoms, and the haloperidol arm as a whole had worse overall survival. That last finding wasn't a high-dose effect, either — those were conservative, age-adjusted, titrated doses, which is the version of haloperidol we'd actually be giving her. That's not a weak signal to talk ourselves past because her daughter is frightened.
I'd agree with you completely if M.T. looked like the patients in that trial. But Agar's population skewed mild-to-moderate — restless, confused, not typically trying to physically climb out of bed the way she is tonight. A null-to-harmful finding in a milder population is real caution, but it isn't the same claim as 'antipsychotics don't work in severe hyperactive delirium,' because that trial didn't characterize enough patients at her severity to actually test that claim.
I want to be honest that this is an extrapolation, not a direct citation — I'm reasoning from the shape of the trial's population, not from a subgroup result the paper itself reports.
What's actually driving her to try to leave the bed twice tonight looks like fear to me — she thinks she's late for a bus, not that she's simply confused. Neither of your two positions is really arguing about that fear directly. I'd try low-dose lorazepam first. I want to be straight about what backs that: Hui and colleagues' 2017 randomized trial is the one clearly positive result in this space, and it found lorazepam added to haloperidol reduced agitation in persistent terminal delirium compared with haloperidol alone. That was lorazepam as an addition, not as monotherapy, so I'm outside what was actually tested and I know it. Which is why I'd watch for the paradoxical agitation it can cause in delirium, and treat that watch itself as the test: if she settles, the fear component was real and benzodiazepine-responsive; if she gets more agitated, we've learned something and the antipsychotic question comes back on the table with better information than we have right now.
Agreed: lorazepam 0.5mg given once, with the hospice nurse staying an extra hour to watch her response directly rather than by phone report. Her daughter was told plainly that the standard antipsychotic approach has real evidence against it in trials like hers, and that tonight's plan is a genuine trial-and-observe step, not a settled answer.
Not agreed: whether M.T.'s severity actually takes her outside the Agar trial's findings, or whether that's a comforting story the team is telling itself to justify treating a family's fear. The Clinical Pharmacologist said directly he still thinks starting an antipsychotic tonight would have been the wrong call regardless of severity; the Geriatric Psychiatrist's position was accepted as reasonable to test, not as proven.