Prazosin for PTSD Nightmares After a Large Negative Trial
A large VA-sponsored trial found prazosin no better than placebo for PTSD nightmares, reversing years of smaller positive studies. The drug remains in widespread use anyway.
R.O. is a 47-year-old man who has driven long-haul freight routes since leaving the Army fifteen years ago, a job he specifically chose, he says, because the cab is the only place he has ever felt like his own hypervigilance was useful rather than exhausting. He has been on prazosin for trauma-related nightmares for six years, titrated slowly by a prior VA prescriber to 6 mg at bedtime, and reports the nightmares dropped from nearly every night to perhaps twice a month somewhere in his second year on the drug. He has no cardiac history and his blood pressure has run on the low-normal side throughout treatment, requiring one dose reduction early on for lightheadedness.
He arrives today having read about a large VA-sponsored trial — the PACT trial, published in 2018 — that found prazosin no better than placebo for nightmares or overall sleep quality in a population larger and more rigorously assessed than any of the smaller studies that had established the drug's reputation over the preceding two decades. He wants to know whether he has spent six years on a drug that a real trial has now shown does not work. The honest answer sits uncomfortably between two true things: PACT is genuinely the best-powered evidence available and its negative result cannot be waved away, and R.O.'s own six-year trajectory — a real, sustained, dose-dependent reduction in a symptom he tracked carefully himself — is not the kind of thing a population-level null result erases for an individual patient who was never enrolled in it.
At the follow-up visit
PACT is not a study to explain away. It was larger, longer, and more methodologically rigorous than any trial that built prazosin's original reputation, and it found no separation from placebo on nightmares or global sleep quality. The mechanism itself — alpha-1 blockade reducing central noradrenergic tone during REM — is plausible, but plausibility is exactly what smaller positive trials also had, and they did not hold up at scale.
One real limit of PACT worth naming: it was designed to test whether starting prazosin helps veterans with chronic PTSD and frequent nightmares as a group, not whether an established, dose-titrated individual responder should stop. Those are different questions, and a null result on the first does not directly answer the second.
That distinction matters more than a straight reading of PACT allows. R.O. is not a treatment-naive enrollee in a randomized population — he is a known, dose-titrated responder with six years of self-tracked data showing sustained benefit. A negative population-level trial tells you prazosin does not reliably work on average; it does not tell you it isn't working in this specific patient, and stopping a drug that is demonstrably working for someone because a larger study found no average effect would be substituting statistics for his actual clinical course.
The honest position sits between both of you, and I'd rather say that to him directly than pretend PACT resolves it. We are not stopping a drug that is working; we are also not going to pretend the evidence base underneath it is what it looked like before PACT. Plan: continue at his current dose, but build in a real discontinuation trial — a slow taper over several weeks, watching his own nightmare frequency — sometime in the next year, so that his continued use rests on his own demonstrated need rather than on inertia plus a reputation the drug earned from the smaller trials that preceded PACT.
Prazosin continued unchanged at 6 mg at bedtime, with an explicit plan for a supervised taper trial later in the year rather than either an immediate stop or indefinite continuation without re-testing.
Not agreed, and left as an open tension rather than a false resolution: whether PACT should change the threshold for STARTING prazosin in a new patient going forward, independent of what it means for a patient already six years into treatment. That question was set aside for a separate conversation, not answered by today's visit.