Lorazepam, PTSD, and Why the Alcohol Paradigm is Bullshit

February 25, 2026. Gate B22 at Berlin Tegel. My Lufthansa flight is 40 minutes late. I have a coffee that tastes like hot disappointment and a phone full of fresh PubMed alerts.

So I'm reading.

There's a paper from this guy Norstrom — Swedish researcher, published this month in Addiction — and he wants to apply the total consumption model to cannabis use. The ALCOHOL model. The same framework Ledermann cooked up in 1956 to predict liver disease from population drinking rates.

I genuinely cannot.

My wife would tell you: this is the face I make when I'm about to send a very long email that I probably shouldn't send. She's seen this face a lot. She's a patient woman.

OK here's the problem. The total consumption model works for alcohol because ethanol is a simple toxin. Dose equals damage, linearly. Your liver cells don't have CB1 receptors. They don't adapt. They don't compensate. Pour more ethanol through them, more cells die. A high school student could graph it.

Your endocannabinoid system? Adapts. CONSTANTLY. CB1 receptors disappear from synapses when flooded, reappear when the signal drops. It's homeostatic. It's designed to maintain balance. Applying Ledermann's alcohol model to the ECS is like diagnosing a fever with a bathroom scale — wrong instrument, wrong disease, wrong century.

But Norstrom isn't even the thing that really got me.

It's the Rodas paper.

Rodas and colleagues took veterans with PTSD — real people, people having nightmares about things that actually happened — and forced them off cannabis for thirty days. Monitored the effects. Called it "extended abstinence research."

OK so here's the biology that should make this obviously cruel. PTSD is not "bad memories." It's fear extinction failure. The brain keeps treating a past threat as a present one because the circuit that would update the threat assessment — the circuit that says "that was then, you're safe now" — isn't firing right. That circuit lives in the basolateral amygdala. Runs on CB1 receptor activation. Marsicano's lab in Bordeaux showed this definitively in 2002. Lutz's group has replicated it six different ways. Without endocannabinoid signaling, fear extinction cannot happen. CANNOT. Not "happens more slowly." Cannot.

So these researchers took people whose fear extinction circuit was already broken and cut off the molecular fuel that circuit requires.

For a month.

And they wrote it up like they'd done something useful.

Look. I know this is how academic medicine works. IRBs, protocols, consented participants. I know. But sometimes the design of a study reveals something about the underlying assumptions — about how the researchers fundamentally conceptualize the relationship between cannabis and the brain — and those assumptions were wrong before the study started.

Bob would've had the phone out of my hand and been dictating a letter to the editor before I finished explaining it. He had OPINIONS about this kind of work.

One paper in the batch was actually worth something. Kamoltham out of Thailand did a real Phase II RCT — whole plant cannabis preparations versus lorazepam for insomnia. Properly randomized. Blinded outcomes assessment. Cannabis won on sleep quality, next-day function, patient satisfaction, basically everything measured.

Which tracks. Benzos flatten your REM architecture. That's just what they do. Your brain cycles through sleep stages — N1, N2, N3, REM — and the REM phase is where emotional memory consolidation happens, where the brain does its nightly maintenance. Lorazepam suppresses that. Cannabis leaves the architecture mostly intact while reducing sleep onset time and nighttime awakenings.

Bob Melamede used to say benzos were pharmacological sledgehammers. He would have loved the Kamoltham trial. He would have printed it out and taped it to his office door.

I miss him.

Flight's boarding. More papers tomorrow.