F. Perry Wilson, MD
@fperrywilson.bsky.social
Director, Clinical and Translational Research Accelerator @Yale. Columnist @medscape.
How Medicine Works and When It Doesn't in bookstores now!
How Medicine Works and When It Doesn't in bookstores now!
I break down the full paper, the physiology of the rebound, and what this means for the future of obesity medicine @medscape: buff.ly/4jvQhAO
/🧵
/🧵
November 24, 2025 at 11:27 PM
I break down the full paper, the physiology of the rebound, and what this means for the future of obesity medicine @medscape: buff.ly/4jvQhAO
/🧵
/🧵
But there were also weight independent effects. Here you see the increase in systolic blood pressure after stopping the drug. It goes up even among people who are able to keep the weight off.
November 24, 2025 at 11:27 PM
But there were also weight independent effects. Here you see the increase in systolic blood pressure after stopping the drug. It goes up even among people who are able to keep the weight off.
With the regain in weight came increases in weight-associated factors, like waist circumference.
November 24, 2025 at 11:27 PM
With the regain in weight came increases in weight-associated factors, like waist circumference.
People who lost more weight ON the drug were more likely to keep it off. That might reflect something like "commitment to weight loss" which is hard to measure. But you don't know how much people will lose before treatment starts.
November 24, 2025 at 11:27 PM
People who lost more weight ON the drug were more likely to keep it off. That might reflect something like "commitment to weight loss" which is hard to measure. But you don't know how much people will lose before treatment starts.
What predicts who will regain weight? Let's look at the baseline data.
Was it age? BMI? Insulin levels?
Answer: None of them. Nothing at baseline predicted who would be able to keep the weight off. We are essentially flying blind.
Was it age? BMI? Insulin levels?
Answer: None of them. Nothing at baseline predicted who would be able to keep the weight off. We are essentially flying blind.
November 24, 2025 at 11:27 PM
What predicts who will regain weight? Let's look at the baseline data.
Was it age? BMI? Insulin levels?
Answer: None of them. Nothing at baseline predicted who would be able to keep the weight off. We are essentially flying blind.
Was it age? BMI? Insulin levels?
Answer: None of them. Nothing at baseline predicted who would be able to keep the weight off. We are essentially flying blind.
We finally have a detailed look at the people who stopped.
This waterfall plot is dramatic. Each bar is one person.
82.5% of people regained significant weight. Almost NO ONE could maintain the loss without the drug.
This waterfall plot is dramatic. Each bar is one person.
82.5% of people regained significant weight. Almost NO ONE could maintain the loss without the drug.
November 24, 2025 at 11:27 PM
We finally have a detailed look at the people who stopped.
This waterfall plot is dramatic. Each bar is one person.
82.5% of people regained significant weight. Almost NO ONE could maintain the loss without the drug.
This waterfall plot is dramatic. Each bar is one person.
82.5% of people regained significant weight. Almost NO ONE could maintain the loss without the drug.
This comes from a new post-hoc analysis of the SURMOUNT-4 trial.
The setup: Participants took tirzepatide for 36 weeks. They lost ~20% of their body weight.
Then, half were switched to placebo. The other half stayed on the drug.
buff.ly/mCCtwHc
The setup: Participants took tirzepatide for 36 weeks. They lost ~20% of their body weight.
Then, half were switched to placebo. The other half stayed on the drug.
buff.ly/mCCtwHc
November 24, 2025 at 11:27 PM
This comes from a new post-hoc analysis of the SURMOUNT-4 trial.
The setup: Participants took tirzepatide for 36 weeks. They lost ~20% of their body weight.
Then, half were switched to placebo. The other half stayed on the drug.
buff.ly/mCCtwHc
The setup: Participants took tirzepatide for 36 weeks. They lost ~20% of their body weight.
Then, half were switched to placebo. The other half stayed on the drug.
buff.ly/mCCtwHc
I’ve called GLP-1 agents like Ozempic and Mounjaro "miracle drugs." But a new study forces me to re-evaluate that moniker.
The miracle has a problem: It ends when the prescription runs out.
Let's look at the data on the "Rebound Effect." 🧵
The miracle has a problem: It ends when the prescription runs out.
Let's look at the data on the "Rebound Effect." 🧵
November 24, 2025 at 11:27 PM
I’ve called GLP-1 agents like Ozempic and Mounjaro "miracle drugs." But a new study forces me to re-evaluate that moniker.
The miracle has a problem: It ends when the prescription runs out.
Let's look at the data on the "Rebound Effect." 🧵
The miracle has a problem: It ends when the prescription runs out.
Let's look at the data on the "Rebound Effect." 🧵
But consumption? Completely different story.
Average number of drinks:
• Placebo: ~3
• Low-dose THC: 2.4
• High-dose THC: 2.1
Nice dose-response. Statistically significant. And the direction is reversed from the authors’ original hypothesis.
Average number of drinks:
• Placebo: ~3
• Low-dose THC: 2.4
• High-dose THC: 2.1
Nice dose-response. Statistically significant. And the direction is reversed from the authors’ original hypothesis.
November 19, 2025 at 4:24 PM
But consumption? Completely different story.
Average number of drinks:
• Placebo: ~3
• Low-dose THC: 2.4
• High-dose THC: 2.1
Nice dose-response. Statistically significant. And the direction is reversed from the authors’ original hypothesis.
Average number of drinks:
• Placebo: ~3
• Low-dose THC: 2.4
• High-dose THC: 2.1
Nice dose-response. Statistically significant. And the direction is reversed from the authors’ original hypothesis.
Cravings went up slightly when participants were shown images of their preferred alcohol. But THC didn’t change cravings meaningfully. The craving curves looked the same whether people smoked placebo or high-dose THC.
November 19, 2025 at 4:24 PM
Cravings went up slightly when participants were shown images of their preferred alcohol. But THC didn’t change cravings meaningfully. The craving curves looked the same whether people smoked placebo or high-dose THC.
Then came the fun part: the bar lab.
For two hours, participants could drink up to eight mini drinks. But for every drink they didn’t consume, they earned $3. A tidy little behavioral economics setup.
For two hours, participants could drink up to eight mini drinks. But for every drink they didn’t consume, they earned $3. A tidy little behavioral economics setup.
November 19, 2025 at 4:24 PM
Then came the fun part: the bar lab.
For two hours, participants could drink up to eight mini drinks. But for every drink they didn’t consume, they earned $3. A tidy little behavioral economics setup.
For two hours, participants could drink up to eight mini drinks. But for every drink they didn’t consume, they earned $3. A tidy little behavioral economics setup.
And the THC definitely did something. Participants reported more happiness, more anxiety (classic), and had measurable increases in heart rate. So we know the doses weren’t trivial.
November 19, 2025 at 4:24 PM
And the THC definitely did something. Participants reported more happiness, more anxiety (classic), and had measurable increases in heart rate. So we know the doses weren’t trivial.
They recruited 157 people who regularly use both alcohol and marijuana. Each participant came into the lab on 3 occasions, smoking:
• a placebo joint
• a 3.1% THC joint
• a 7.2% THC joint
(in random order)
• a placebo joint
• a 3.1% THC joint
• a 7.2% THC joint
(in random order)
November 19, 2025 at 4:24 PM
They recruited 157 people who regularly use both alcohol and marijuana. Each participant came into the lab on 3 occasions, smoking:
• a placebo joint
• a 3.1% THC joint
• a 7.2% THC joint
(in random order)
• a placebo joint
• a 3.1% THC joint
• a 7.2% THC joint
(in random order)
The substitution model is especially interesting: even if both drugs show up together in the real world, one might suppress the other when conditions are controlled. But that’s a hard thing to study. Enter Dr. Jane Metrik’s group. buff.ly/nUuJPg5
November 19, 2025 at 4:24 PM
The substitution model is especially interesting: even if both drugs show up together in the real world, one might suppress the other when conditions are controlled. But that’s a hard thing to study. Enter Dr. Jane Metrik’s group. buff.ly/nUuJPg5
Substitution: people are just looking to “feel different,” and any drug that gets them there competes with the others.
November 19, 2025 at 4:24 PM
Substitution: people are just looking to “feel different,” and any drug that gets them there competes with the others.
Phenotype: the same “type” of person tends to use both; the drugs are correlated because people bring their traits with them.
November 19, 2025 at 4:24 PM
Phenotype: the same “type” of person tends to use both; the drugs are correlated because people bring their traits with them.
There are a few possible models in this space.
Complementarity: marijuana makes you more likely to drink; lower inhibitions, more risk-taking, maybe more craving.
Complementarity: marijuana makes you more likely to drink; lower inhibitions, more risk-taking, maybe more craving.
November 19, 2025 at 4:24 PM
There are a few possible models in this space.
Complementarity: marijuana makes you more likely to drink; lower inhibitions, more risk-taking, maybe more craving.
Complementarity: marijuana makes you more likely to drink; lower inhibitions, more risk-taking, maybe more craving.
Start with the obvious: observationally, cannabis use disorder and alcohol use disorder are extremely comorbid. Roughly 60% of people with cannabis use disorder also meet criteria for alcohol use disorder. Pot users drink more; drinkers smoke more. It's a tight pair.
November 19, 2025 at 4:24 PM
Start with the obvious: observationally, cannabis use disorder and alcohol use disorder are extremely comorbid. Roughly 60% of people with cannabis use disorder also meet criteria for alcohol use disorder. Pot users drink more; drinkers smoke more. It's a tight pair.
A new randomized trial in American Journal of Psychiatry just dropped, and the headline result is…unexpected: smoking marijuana REDUCED alcohol consumption. Yes, reduced. And before you say “that makes no sense", let’s unpack it. (🧵)
November 19, 2025 at 4:24 PM
A new randomized trial in American Journal of Psychiatry just dropped, and the headline result is…unexpected: smoking marijuana REDUCED alcohol consumption. Yes, reduced. And before you say “that makes no sense", let’s unpack it. (🧵)
PS: As I was reading this paper, one line kept running through my head:
"They're sharing a drink they call loneliness, but it's better than drinking alone." - Billy Joel, 1973
Turns out he understood this paradox 50 years before the research caught up.
"They're sharing a drink they call loneliness, but it's better than drinking alone." - Billy Joel, 1973
Turns out he understood this paradox 50 years before the research caught up.
November 12, 2025 at 7:54 PM
PS: As I was reading this paper, one line kept running through my head:
"They're sharing a drink they call loneliness, but it's better than drinking alone." - Billy Joel, 1973
Turns out he understood this paradox 50 years before the research caught up.
"They're sharing a drink they call loneliness, but it's better than drinking alone." - Billy Joel, 1973
Turns out he understood this paradox 50 years before the research caught up.
Turns out you can have high "social well-being" (lots of friends, companionship) AND high "social ill-being" (loneliness, disconnection) at the same time.
Most people in the study were in BOTH categories simultaneously.
Most people in the study were in BOTH categories simultaneously.
November 12, 2025 at 7:54 PM
Turns out you can have high "social well-being" (lots of friends, companionship) AND high "social ill-being" (loneliness, disconnection) at the same time.
Most people in the study were in BOTH categories simultaneously.
Most people in the study were in BOTH categories simultaneously.
I tend to think of loneliness and social connection as two poles on a spectrum, but this turns out to be overly simplistic.
November 12, 2025 at 7:53 PM
I tend to think of loneliness and social connection as two poles on a spectrum, but this turns out to be overly simplistic.
We've all heard about the "loneliness epidemic." And yes, loneliness is linked to heart disease, dementia, stroke, and higher mortality.
But the fascinating thing about this new PLOS One paper: loneliness and social connection aren't opposites.
buff.ly/xcNMxg1
But the fascinating thing about this new PLOS One paper: loneliness and social connection aren't opposites.
buff.ly/xcNMxg1
November 12, 2025 at 7:53 PM
We've all heard about the "loneliness epidemic." And yes, loneliness is linked to heart disease, dementia, stroke, and higher mortality.
But the fascinating thing about this new PLOS One paper: loneliness and social connection aren't opposites.
buff.ly/xcNMxg1
But the fascinating thing about this new PLOS One paper: loneliness and social connection aren't opposites.
buff.ly/xcNMxg1
Here's how that would mess things up:
Some kids who wouldn't normally need psychiatric care get pushed over that edge by doxycycline. Now they're in the study group, but they're actually the HEALTHIEST kids in that group.
Some kids who wouldn't normally need psychiatric care get pushed over that edge by doxycycline. Now they're in the study group, but they're actually the HEALTHIEST kids in that group.
November 6, 2025 at 6:42 PM
Here's how that would mess things up:
Some kids who wouldn't normally need psychiatric care get pushed over that edge by doxycycline. Now they're in the study group, but they're actually the HEALTHIEST kids in that group.
Some kids who wouldn't normally need psychiatric care get pushed over that edge by doxycycline. Now they're in the study group, but they're actually the HEALTHIEST kids in that group.
But I have doubts.
First, there's no clear dose-response effect. Higher doses should show stronger effects if this is causal. The medium dose looks best here, which is... odd.
First, there's no clear dose-response effect. Higher doses should show stronger effects if this is causal. The medium dose looks best here, which is... odd.
November 6, 2025 at 6:41 PM
But I have doubts.
First, there's no clear dose-response effect. Higher doses should show stronger effects if this is causal. The medium dose looks best here, which is... odd.
First, there's no clear dose-response effect. Higher doses should show stronger effects if this is causal. The medium dose looks best here, which is... odd.