Benzodiazepines in Delirium: Almost Always Wrong, Except When They're the Only Right Answer
Two delirious patients, two very different verdicts on the same drug class. One case shows why benzodiazepines usually make hospital delirium worse; the other shows the specific pharmacology that makes them the correct, guideline-mandated first-line choice.
Helen G., 88, was admitted four days ago for a hip fracture repair and has been slow to recover mentally since surgery — not confused on admission, but now intermittently disoriented to place, anxious, and reporting she can hear her late husband's voice in the hallway at night. Her daughter, at bedside most of the day, says this is nothing like her mother, who lives independently and still drives during daylight hours.
The overnight nurse, worried about her anxiety and poor sleep, asks whether a small dose of lorazepam would help her settle. It is, in a hospital, an understandable and common request — but Helen's delirium has an identified, more specific driver: three days of accumulating oxycodone for post-surgical pain, on top of an opioid-naive baseline, plus persistent sleep fragmentation from overnight vitals. Nothing about her presentation suggests alcohol withdrawal or any GABA-related etiology; benzodiazepines here would be treating none of her actual contributing causes.
She's anxious, she's not sleeping, and the overnight team is stretched thin. I don't think a low dose of lorazepam to help her settle is an unreasonable ask tonight — it's what gets requested in this exact situation more often than not.
I'd hold off on the lorazepam. Nothing about her picture points to a GABA-related cause — there's no alcohol history, no autonomic instability, no tremor. What we do have is three days of accumulating oxycodone in an opioid-naive 88-year-old, plus overnight sleep fragmentation from vitals and labs, both of which are well-established, specific delirium drivers in older post-surgical patients.
Adding a benzodiazepine on top of that wouldn't address either cause — it would add a second sedating drug to a delirium picture that benzodiazepines are consistently associated with worsening, not resolving, in patients like her.
That's fair, and I hadn't actually worked through what's driving it specifically — I'd defer to the opioid and sleep-protocol plan instead and hold the lorazepam in reserve rather than reach for it tonight.
Agreed: no benzodiazepine given. Opioid dosing adjusted and scheduled around documented pain, and overnight vitals consolidated to reduce sleep disruption, with reassessment the following morning.
Helen's disorientation and nighttime hallucinations improved over the following two nights as the opioid regimen was adjusted and sleep disruption reduced, without any benzodiazepine having been used.
Dennis R., 52, was brought to the emergency department by his brother two days after what the brother describes as a sudden, unplanned stop to drinking following a decade of daily heavy alcohol use — a stop forced by a hospitalization for an unrelated injury that left him without access to alcohol. He is now profoundly disoriented, tremulous, sweating through his gown, and intermittently seeing insects crawling on the wall that aren't there. His heart rate is 128 and his blood pressure is 178/104.
His CIWA-Ar score is 27, and the clinical picture — autonomic hyperactivity, tactile and visual hallucinations, profound disorientation, onset 48 hours after his last drink — is textbook delirium tremens, the most severe and dangerous stage of alcohol withdrawal, with a real risk of seizure and death if untreated. Unlike Helen's case, the underlying pathophysiology here is a specific, well-characterized GABA-A receptor state: years of heavy alcohol use downregulate GABA-A receptor sensitivity and upregulate excitatory NMDA signaling as compensation, and abrupt cessation unmasks that adaptation as a genuine, life-threatening hyperexcitable state.
This is the one situation where a benzodiazepine is exactly the right first-line drug, and not treating him with one is what would be dangerous. Delirium tremens reflects a specific, well-characterized adaptation: chronic alcohol exposure downregulates GABA-A receptor sensitivity and upregulates NMDA-mediated excitatory tone as compensation. Abrupt cessation removes the GABAergic input that adaptation was built around, unmasking a genuine hyperexcitable state — the tremor, the autonomic surge, the hallucinations, and the real risk of a withdrawal seizure.
Lorazepam potentiates GABA-A receptors directly, functioning as a cross-tolerant substitute for the alcohol his brain has adapted around. This isn't a sedative masking an unrelated problem the way it would be in Helen's case — it's the specific, guideline-recommended correction for the actual receptor state driving his symptoms.
I agree completely on the drug — I'd just push on how we dose it once it's running. Saitz and colleagues, JAMA, 1994, randomized patients with alcohol withdrawal to symptom-triggered dosing, guided by a structured score like his CIWA-Ar, against fixed-schedule dosing given regardless of symptom severity. The symptom-triggered group needed substantially less total medication and finished treatment in a shorter time, with no difference in safety.
I'm not disputing that he needs lorazepam right now — I'm saying that once we're past this acute moment, dosing him against his actual CIWA-Ar trajectory rather than a fixed clock will get him through this with less total drug on board, not more.
Agreed: symptom-triggered lorazepam per CIWA-Ar protocol, IV thiamine given before any dextrose-containing fluids, and close monitoring for escalating autonomic instability or seizure.
Not agreed, though narrowly: the addiction medicine specialist would have been comfortable with a more cautious, front-loaded initial dosing strategy given how severe his presentation already was at arrival; the emergency medicine physician held that symptom-triggered dosing should govern from the very first dose, not just once he's past the acute window. Both agreed this did not change tonight's immediate plan.