Doing Everything Right and Still Anxious

What the science of habit formation can add to lifestyle medicine

A patient I’ll call Margaret comes to see me. She’s in her fifties, a recently retired pharmacist, and on paper she is the dream patient for any lifestyle medicine clinic. She doesn’t smoke. She hasn’t had a drink in eight years. She walks four miles a day, eats the kind of Mediterranean diet her cardiologist son helped her dial in. But that’s not all: she has a book club, a hiking group, and a husband she still likes. Her labs are fine. Her A1C is fine. Her blood pressure is fine. And as a pharmacist, she knows more about medications than most doctors.

She is also so anxious that she hasn’t been able to drive without raising her blood pressure and hasn’t gotten a good night’s sleep for several years.

Margaret has done everything right by the lifestyle medicine playbook, and she still spends a fair amount of her day rehearsing catastrophes. She echoes what a lot of patients have said over the years: “I don’t get it. I’m doing all the things you’re supposed to do.”

Yes, and... The playbook may be incomplete.

As a psychiatrist who has spent the past two decades studying how people get hooked on things, lately I’ve been thinking about some related subtleties and derivatives of that question. Not “why does anxiety persist,” which gets a lot of attention, but “why do the things we tell anxious people to do so often fail to stick?” Sleep more. Drink less. Move your body. Practice stress management. The advice is fine (and evidence-based). So why is Margaret, who follows it all, still white-knuckling her way through life?

I think part of the answer may be that lifestyle medicine may have a mechanism problem. We’ve built a generation of evidence around six pillars (nutrition, physical activity, restorative sleep, stress management, social connection, and avoiding risky substances) as a way to prevent and reverse chronic disease. I’m not here to argue with the pillars. And I’m a big fan of lifestyle medicine. I would go as far as saying that these pillars likely keep people healthier (in theory) than any combination of medications. Instead, I want to argue with the model we use to deliver them, which boils down to prescription + education + willpower. That’s roughly the same model that has failed addiction medicine for fifty years, and it fails for the same reason. It doesn’t account for how the brain actually learns.

Reward-based learning is older than humans. Eric Kandel won the Nobel Prize in 2000 for showing that this learning process operates in the sea slug, an organism with twenty thousand neurons in its entire nervous system. The basic structure is trigger, behavior, reward, and then repeat. Your brain encodes the reward value of each behavior in a region called the orbitofrontal cortex, where it gets compared against the reward value of other available behaviors and used to drive your future choices. This is how you learned to prefer dark chocolate to milk chocolate. It’s also how a forty-year patient of mine reinforced his smoking habit roughly 293,000 times before he walked into my clinic asking for help to quit.

Anxiety lives inside this same machinery. A trigger shows up. Maybe it’s an email, maybe it’s a tightness in your chest, maybe it’s a thought about tomorrow. Your brain reaches for whatever behavior has, in the past, made the discomfort go away even temporarily. Worry. A drink. An innocent scroll through your phone. A second helping of dessert. The behavior gets reinforced.

By the time someone like Margaret arrives at a clinic, her worry-and-avoid loop has been grooved thousands of times (or more), and her brain has stored that worrying has, paradoxically, some level of reward. Not because worry feels good. Because at some point it seemed to help her prepare for something bad, and her prefrontal cortex hasn’t been able to update that reward value since then.

There’s one more thing about this: the prefrontal cortex, which is where willpower-based interventions live, is the first brain region to go offline under stress. Amy Arnsten’s research at Yale has shown this in elegant detail. The very brain system we ask patients to use to implement our recommendations is the system that goes dark precisely when implementation is hardest. We hand people a flashlight and turn the lights out at the same time.

Substance use

The pillar I know best, and the one where the mechanism story is hardest to miss, is the avoidance of risky substances. Most clinicians treat substance use and anxiety as two conditions that travel together. The patient has anxiety and drinks, and we treat each in parallel. But for a big slice of patients, the substance isn’t the comorbidity. It’s the anxiety, in disguise.

I see this in my clinic pretty regularly. One patient stands out, a woman in her fifties whose anxiety was so severe in her twenties that she couldn’t make it to the grocery store without “pre-partying” first. A few decades of that, and by the time she walked into my office she had a textbook alcohol use disorder. The loop she described was almost too clean: drink to take the edge off the anxiety, drink past the edge because by then the off-switch was already underwater, wake up the next morning with anxiety worse than the one she had been trying to medicate. The substance wasn’t sitting next to her anxiety. It was the only tool she had ever found for it, and it was making the anxiety worse on a delay long enough that her brain never quite connected the two.

Alcohol is anxiety-relieving for an hour or two and anxiety-producing on the back end. The colloquial word “hangxiety” turns out to point at something real, with the day-after anxiety often hitting hardest in people who were already anxious to begin with. Nicotine is a stimulant that somehow reduces self-reported anxiety, because the withdrawal between cigarettes is itself anxiogenic, so each cigarette relieves a discomfort that the previous cigarette created. And when my patients and I examine this carefully, nicotine (a stimulant) isn’t the anxiety reliever; it’s the smoke BREAK that they take (which can be substituted by simply taking a break to stretch or go for a short walk, or to do a short mindfulness exercise).

Cannabis is more complicated, but for a lot of habitual users it follows the same arc. Benzodiazepines are the cleanest example, in a sad way: prescribed to treat anxiety, they generate, over time, a worse anxiety than the one they were prescribed for. And getting off of them is a bitch.

Telling these patients to “avoid risky substances” isn’t wrong. It’s just not a treatment. The substance is doing a job inside an anxiety habit loop, and unless we address the loop itself, the slot doesn’t stay empty. It gets filled by overeating, or doomscrolling, or, often enough, a new prescription.

Stress management

If risky substances is the pillar where the link to anxiety is most concrete, stress management is the pillar where I think the field has the thinnest mechanistic account. From my admittedly not complete read of the literature, most stress management content in lifestyle medicine boils down to a list. Diaphragmatic breathing. Progressive muscle relaxation. Yoga. Journaling. Time in nature. Good things, all of them, and not enough research on why they work or for whom.

Perhaps more importantly, anxiety and stress aren’t the same animal. Techniques that work for acute stress, like downregulating the sympathetic nervous system in the moment, often do nothing for the rumination loops that define generalized anxiety. I’ve talked to patients who diligently practiced box breathing for months and reported that it didn’t touch their 3 a.m. worry spirals.

The mechanism that does seem to address rumination is awareness applied to the loop itself. Noticing the trigger. Watching the behavior of worrying as it unfolds. Feeling clearly what the actual reward, or lack of reward, of worry is in the body. In trials from my lab and others, this kind of awareness training has produced clinically meaningful reductions in anxiety symptoms, and the proposed mechanism is exactly the one I’ve been describing: updating the reward value of worry in the orbitofrontal cortex so that your brain stops reaching for it. That’s a different approach that may get at the underlying driver of anxiety itself, and is quite different than being told to take a few deep breaths when you’re stressed.

Sleep

Sleep is the pillar where the two-way relationship with anxiety is best established. Anxious people sleep worse, and people who sleep worse become more anxious. A single night of sleep deprivation makes us (and our brains) more reactive (e.g. work that Yoo and colleagues laid out back in 2007 and that Goldstein and Walker have since built into a broader picture).

The lifestyle medicine teaching for patients is sleep hygiene. Consistent bedtime. Cool dark room. No screens before bed. Watch the caffeine. All sensible, and with quite a bit of evidence behind them. Unfortunately, none of what to do addresses the patient who already knows this stuff and still lies awake from 2:47 to 4:30 a.m. running through a list of things she can’t fix.

Insomnia, in a lot of these cases, is itself a habit loop. The bed becomes a conditioned anxiety cue. “Trying to sleep” becomes the failed control behavior, and lying awake feeling frustrated about lying awake gets reinforced as the actual behavior the brain practices, night after night. Cognitive Behavioral Therapy for Insomnia works partly by breaking this conditioned association, through techniques like stimulus control and sleep restriction.

In my own lab’s randomized trial of app-based mindfulness training for people whose worry was interfering with their sleep, we saw something pretty remarkable (yes I’m biased, but the data speak for themselves). In a randomized controlled trial, after two months, participants reported a 27% reduction in worry-related sleep disturbances, compared with 6% in the control group, and the mediation analysis pointed at a specific mechanism: their ability to not get hooked by their own anxious thoughts increased, which reduced their worrying, which reduced their perceived sleep problems. The part that complicates the story in an interesting and perhaps useful way is what didn’t change: their actual sleep duration, measured objectively, was about the same.

What changed was the loop. Their bedtime brains stopped being the same hyperaroused, ruminating brains they had brought into the bedroom for years, and the subjective experience of sleep improved even when the Fitbit data didn’t move. The thing that got better was the patient’s relationship with her own mind at 2 a.m. which is likely much better than your wearable telling you that you did/did not get a good night of sleep.

Exercise

I’m leaving exercise out of the main argument, but it deserves a paragraph, because it’s the pillar with arguably the strongest evidence for anxiety reduction (e.g. Stubbs and colleagues’ meta-analyses are reasonably solid), and it’s also the pillar that most clearly illustrates the limits of the prescription model. We can demonstrate in a trial that exercise reduces anxiety symptoms (and many have). We can also fail completely at getting an anxious patient to exercise, because the anxiety itself is what makes initiating and sustaining exercise hard. Even the pillar with the best evidence doesn’t implement itself. The implementation problem is the learning problem.

I’m not arguing that lifestyle medicine has the wrong pillars. The pillars are well-chosen and the evidence supporting them as targets is real. I’m suggesting that the model of how a patient gets from where they are to where the pillars say they should be is borrowed largely from a willpower-and-education tradition that hasn’t aged well in the addiction literature, and that anxiety is the cleanest place to see this, because anxiety renders willpower-based intervention nearly inert.

What might it look like to take learning seriously? Probably something like this. Clinical encounters that start not with “here’s what to do” but with “here’s the loop you’re already running.” Treatment plans that treat the OFC’s reward value as a thing to be updated, not a thing to be overridden. Time spent with patients in the actual texture of their behavior. What does that cigarette taste like? What does the worry feel like? What is the real reward of the second or third glass of wine? Sit with it long enough that your nervous system gets a chance to learn what it hasn’t yet had the chance to learn. That’s how reinforcement learning works (with a notable absence of the need for willpower).

I don’t think any of this is a departure from lifestyle medicine. Instead, I think it’s what lifestyle medicine gets up to date on the latest neuroscience about why so many of its prescriptions fail to stick.

A question worth sitting with, whether you’re a clinician or someone who has been on the receiving end of all this advice: when the change doesn’t happen, is the first instinct to ask for more education, more motivation, more discipline (or read another self-help book)? Or is it to ask what the loop is, and how to make it visible?

Those lead down completely different roads, and toward very different results.

Judson Brewer MD PhD is a psychiatrist and neuroscientist and professor at Brown University. He is the author of Unwinding Anxiety (NYTimes bestseller), The Craving Mind, The Hunger Habit and The Unwinding Anxiety Workbook. He co-founded MindshiftRecovery.org which provides free support for people with any type of addiction.

If you are struggling with anxiety, Dr. Brewer’s Going Beyond Anxiety program brings together his research and clinical experience to help people build effective skills to reduce anxiety and cultivate calm (www.goingbeyondanxiety.com).

Copyright © 2025, Judson Brewer, MD, PhD. All rights reserved.

REFERENCES

Arnsten, A. F. T. (2009). Stress signalling pathways that impair prefrontal cortex structure and function. *Nature Reviews Neuroscience, 10*(6), 410–422.

Arnsten, A. F. T. (2015). Stress weakens prefrontal networks: Molecular insults to higher cognition. *Nature Neuroscience, 18*(10), 1376–1385.

Brewer, J. A. (2019). Mindfulness training for addictions: Has neuroscience revealed a brain hack by which awareness subverts the addictive process? *Current Opinion in Psychology, 28*, 198–203.

Goldstein, A. N., & Walker, M. P. (2014). The role of sleep in emotional brain function. *Annual Review of Clinical Psychology, 10*, 679–708.

Kandel, E. R. (2001). The molecular biology of memory storage: A dialogue between genes and synapses. *Science, 294*(5544), 1030–1038.

Ong, J. C., Manber, R., Segal, Z., Xia, Y., Shapiro, S., & Wyatt, J. K. (2014). A randomized controlled trial of mindfulness meditation for chronic insomnia. *Sleep, 37*(9), 1553–1563.

Stubbs, B., Vancampfort, D., Rosenbaum, S., Firth, J., Cosco, T., Veronese, N., Salum, G. A., & Schuch, F. B. (2017). An examination of the anxiolytic effects of exercise for people with anxiety and stress-related disorders: A meta-analysis. *Psychiatry Research, 249*, 102–108.

Yoo, S. S., Gujar, N., Hu, P., Jolesz, F. A., & Walker, M. P. (2007). The human emotional brain without sleep: A prefrontal amygdala disconnect. *Current Biology, 17*(20), R877–R878.