I Compared 443 Nights of Sleep With My Ketone Log

Four attempts, three breaks, one imperfect watch — and the thing I was surest about turned out to be the one thing the watch does not see

I have never been a naturally deep sleeper.

When I am not in ketosis, my sleep is thin. Not absent. Not catastrophic. Just thin — like floating on top of a pool instead of sinking to the quiet bottom. I wake easily. I rarely remember dreams. I can spend enough hours in bed and still wake up feeling like my brain's battery stopped charging at 63 percent.

In sustained ketosis, I sink. Dreams return. Mornings feel cleaner. And waking up comes with something I almost never get otherwise: a heavy disorientation. What day is it? Where am I? It takes real time to surface — the way you come up slowly from deep water. I read that grogginess as a clue that I had actually been down there — in genuinely deep sleep.

So for years I have told people, confidently: I sleep better in ketosis.

That sentence may be true. But a sentence repeated often enough starts to sound more proven than it is — and that is dangerous when the person saying it is a physician. So this month I finally audited the claim. I lined up 443 nights of watch data against my blood-ketone log.

The broad pattern survived. Some of my favorite explanations did not. Finishing the audit forced two confessions I almost left out of the original report. And then — after the first draft of this letter was already written — the deep-sleep and REM data finally arrived, all of it, and the thing I was most sure about turned out to be the one thing the watch does not see.

My baseline, honestly

I have rarely remembered dreams. Since high school, other people talked about their dreams and I rarely had one to offer. That does not mean I wasn't dreaming — almost everyone dreams. Think of dream recall as a notebook left closed: the pen may have been moving all night, but you don't get to read the pages in the morning. Still, the difference was noticeable. Research says I'm not unusual: dream recall varies enormously between people, and those differences are stable traits [2,3].

I needed more sleep than my peers. By college and medical school, I was more tired than my siblings and friends. Nine hours was when I felt right; anything less and I was dragging. I never asked anyone whether that was normal — the culture says seven to eight, so I quietly assumed I was just bad at the rule. I was also, and remain, a natural night owl.

Residency broke my sleep on purpose. On call every third night. Home at 3 a.m., sometimes later. Back at the hospital by 6 a.m. And here is the cruel part: the more urgently I needed to sleep, the more impossible sleep became. Trying to sleep under a deadline is like trying to force a locked door — the harder you push, the more it sticks. This is real physiology, not weakness: stress and "sleep effort" genuinely worsen sleep, and some people's sleep systems are more stress-reactive than others [4]. I started taking Ambien during residency. That is how real it was.

My fifties got worse. Longer time to fall asleep. Extremely light sleep — I woke at everything. Even fewer remembered dreams.

Then came the bladder. In my mid-fifties I started waking to urinate — sometimes three or four times a night, especially when I am not in ketosis, when I notice a bladder urgency that pushes me out of bed. Each trip is an alarm clock with no snooze button: you are up, you are walking, and then you have to fall asleep again — or lie there hoping. Nocturia (the medical term for waking at night to urinate) is strongly linked to fragmented, poor-quality sleep, and the damage grows with each awakening [5,6]. I see the same pattern constantly in my patients.

And now the confession. My wife told me years ago: when I am heavier — and not in ketosis — I snore. At my worst, my BMI was over 30. It wasn't just snoring. I tested myself, and the test showed significant obstructive sleep apnea — a condition where the airway repeatedly collapses during sleep, cutting off breathing and shattering sleep architecture. When I lose the weight, the snoring stops. The apnea resolves. This has happened more than once, in both directions, and it is not subtle. I have watched the same thing in my own patients, including patients on CPAP machines (the bedside air-pressure devices that hold the airway open) whose apnea disappeared after major weight loss. That matches the trial evidence: weight loss reduces apnea severity, and remission happens often enough to be a realistic goal — though never a guarantee [9,10].

One rule outranks everything in this paragraph: if you use CPAP, do not stop it because you lost weight or feel better. Get re-tested and let the data decide.

What 443 nights showed

The spreadsheet covered January 2025 through late September 2026: 446 recorded dates, 443 nights with a usable sleep score, four labeled ketosis attempts, and three off-periods [1]. Quick glossary so the numbers mean something: the sleep score is my Polar watch's single 0–100 grade for the night — a report-card number that blends several ingredients into one mark. Continuity rates how broken the night was. Sleeping heart rate is my average overnight pulse. BHB (beta-hydroxybutyrate) is the blood ketone I measure — the liver's receipt confirming how ketotic I am.

The headline:

And it repeated, period after period: 60.2 vs 55.0 in early 2025; 59.2 vs 53.1 in spring; 60.4 vs 54.8 in the fall; then a long 2026 run averaging 62.6 [1]. A single good week is weather. A pattern that returns after you stop and start again is climate.

The rough beginning also mostly survived: in three of four attempts, weeks three through six beat the first two weeks (January 2025: 52.6 rising to 63.3). A recurring personal pattern — not a universal law, but enough to warn against judging any metabolic transition by its noisiest chapter.

What did not survive the audit:

1. Higher ketones did not clearly mean better sleep. Inside ketosis periods, bigger BHB numbers did not predict better scores [1]. Being in ketosis seemed to matter more than pushing the number higher once I was there. A light switch doesn't make the room brighter when you press it harder after it's already on.
2. HRV was not a ketosis story. My heart-rate variability (the small beat-to-beat timing variations used as a rough recovery gauge) rose over time — but it tracked the calendar, not ketosis [1]. Science is also about losing the explanations you wanted to keep.

And the standing caveat: a watch is not a sleep lab. Consumer wearables estimate sleep from movement and pulse; they cannot diagnose anything. The American Academy of Sleep Medicine says so explicitly [15]. When I say the watch shows a pattern, I mean a pattern — not a verdict.

The data I was missing finally arrived

The first version of this letter had an honest hole in it. My workbook had no nightly REM or deep-sleep columns. So my strongest claims — deeper sleep, more REM, dreams almost every night — had to live in the "strong impression" bucket, with a promise that I was trying to get the numbers.

I got them the slow way. The watch app will not export sleep stages, so I opened it and read out every night it ever recorded, one night at a time, into a spreadsheet: 431 nights, from January 2025 to last week. That covers all four of my ketosis attempts and all three of the breaks in between. And I set aside the first two weeks of every attempt, because I have always said — and the scores agree — that the start of ketosis is the roughest sleep of the whole process. Judging ketosis by its first two weeks is like judging a marriage by the wedding-planning.

Quick reminder of what the two numbers mean. Deep sleep is the slow-wave stage: the body's repair shift, the part you cannot be shaken out of easily, the part a watch estimates from how still you are and how slow and steady your pulse is. REM is the dreaming stage. Most adults get roughly an hour to an hour and a half of each. I expected my ketosis nights to show much more of both.

Here is what the watch actually showed.

Six more minutes of deep sleep in ketosis. Five more minutes of REM. That is the difference between a 54-minute nap and a one-hour nap. And there is a simpler explanation for even that small gap: in ketosis I sleep about half an hour longer, and a longer night carries a little more of every stage with it. Measured as a share of the night, deep sleep was 15 percent whether I was in ketosis or out of it. Same slice of the pie; the pie was just slightly bigger.

I did not trust one pooled number, so I checked attempt by attempt, each one against the break that came right after it. If ketosis deepens sleep, the stage minutes should drop every time I stop. After my January 2025 attempt, they did: deep sleep went from 62 minutes a night down to 49 when I quit. After my April 2025 attempt, nothing moved — 51 minutes in ketosis, 51 minutes out. After my October 2025 attempt, it went the wrong way: 54 minutes in ketosis, 59 in the break. And the long, carbohydrate-eating winter of 2025–26, which I would have called my worst sleep in years, gave me 59 minutes of deep sleep and 79 of REM a night — the same as this year's run, at 61 and 80. One for, one neutral, one against. That is not a pattern. That is a coin.

Then I asked the dimmer question again: within ketosis, do higher ketones mean deeper sleep? Across 141 nights, lining up my rolling week of ketone readings against that night's deep sleep, the correlation was 0.02. Zero would mean no relationship at all. A nightly reading of 5 bought me no more deep sleep than a reading of 2.

So I had to stop and ask the obvious question. If the stages do not change, why does the watch score my ketosis nights higher, every single attempt? I went back to the composite data to see which ingredients of the score were actually moving. Three were, and they moved the same way in all four attempts: my sleep was less broken (continuity 2.6 out of 5 in ketosis, 2.1 out of it — roughly one fewer stretch of lying awake per night); my sleeping heart rate was four beats a minute slower; and I slept about half an hour longer. Not deeper sleep. Less interrupted sleep, a quieter heart, and more of it.

That changed how I understand my own experience, and I think it changed it correctly. The heavy, disoriented wake-ups; the feeling of surfacing from somewhere far down; the dreams I actually remember — those are real, and they come back every time I am in ketosis. But the explanation may not be "more deep sleep." It may be "unbroken sleep." Think about what a night out of ketosis looks like for me: a bladder alarm at 2 a.m., then another at 4, each one followed by the lying-there-hoping, each one a chance to snore and gasp myself half awake. Now take those away. A night with one awakening instead of four feels deeper — you go down and stay down — whatever the stage minutes say. The switch I kept describing may be in my bladder and my airway, not in my brain.

Two cautions, in fairness to the reading I expected. First, a wrist watch has no brain-wave sensor. It guesses stages from movement and pulse, and it is weakest at exactly the distinction that matters here — deep versus light. It is possible ketosis deepens my sleep in a way this device simply cannot see. Second, the whole record drifted: my nights in 2026 were deeper than my nights in 2025 whether I was in ketosis or not, the same slow calendar effect I found with HRV. I cannot tell you what caused that either.

What I can tell you is this. I went looking for proof of the sentence I had repeated for years, and the proof I found was for a different sentence.

Why your version of this story matters

My history, piece by piece, is not unusual. Each piece has a literature behind it — and describes people who read health articles and think, that's me.

"I never remember my dreams." Recall varies hugely and stably between people, and low recall says nothing about how much you actually dreamed [2,3]. If that's you, that's a clue, not a quirk.

"The harder I try to sleep, the worse it gets." That's sleep reactivity — a measurable trait and a known vulnerability for insomnia [4]. A clue, not a quirk.

"I wake up to pee, and then the night is over." Nocturia fragments sleep directly [5,6] — and it travels with sleep apnea more often than most people realize [7,8]. Two or more trips a night, most nights, is worth saying out loud to a doctor. A clue, not a quirk.

"When I gain weight, I snore. When I lose it, I sleep." The weight–apnea relationship is one of the most established facts in sleep medicine. Weight-loss interventions reduce apnea severity; remission is realistic but never guaranteed [9,10]. A clue — and one of the few that comes with a treatment.

"Does keto itself do anything to sleep?" Honestly: possibly. A small 2008 sleep-lab study found more slow-wave (deep) sleep within 48 hours on a very-low-carb diet — and less REM [11]. A 2021 meta-analysis echoed that trade-off: lower carbohydrate, more deep sleep; higher carbohydrate, more REM [12] — a useful warning against any "ketosis raises everything good" story. A 2024 review of 20 studies found encouraging signals, but the studies were mostly in people with epilepsy and other neurological conditions, with messy methods [13]. Mechanistically, ketones may influence adenosine — the brain's sleep-pressure signal — which is plausible, not proven [14].

The literature's verdict and my spreadsheet's verdict are the same three words: interesting, plausible, unsettled.

Seven questions to ask about your own sleep

This is the part I most want my patients to read. You don't need a watch. You need honest answers.

1. Falling asleep. How long does it take — fifteen minutes, or an hour of lying there wired? Does pressure make it worse?
2. Staying asleep. How many times do you wake? When you do, are you half-awake — or were you truly out? Do you fall back asleep, or does a 2 a.m. waking end the night?
3. Depth. When you wake, are you instantly alert — or briefly disoriented, unsure of the day, needing time to surface? That heavy grogginess (doctors call it sleep inertia) is compatible with waking from deep sleep. But be honest about its limits: it's a clue, not a measurement.
4. Dreams. Do you remember dreams rarely, sometimes, or almost nightly? Has that changed with weight, stress, alcohol, medications, or diet?
5. Bathroom trips. Zero to one, or two to four? Do they change with weight, blood sugar, evening fluids, alcohol, or diet?
6. Breathing. Has anyone heard you snore loudly, gasp, choke, or stop breathing? Is it worse when your weight is up? Have you ever actually been tested?
7. Mornings. Do you wake restored, or already tired? Do you doze in meetings, on the couch — or, the red flag, while driving?

How sure am I?

• Established: my history as I've reported it; the 443-night dataset; the repeated on/off association in the watch numbers [1]; wearables don't diagnose [15]; weight loss improves apnea [9,10].
• Established (new): across 431 nights with stage data, my deep sleep and REM are essentially the same in ketosis and out of it — about an hour of deep and an hour and a quarter of REM either way, with a six-minute edge to ketosis that longer nights explain. What does change in ketosis, every attempt, is how unbroken the night is, how slow my sleeping heart rate is, and how long I sleep [1,16]. Within ketosis, higher ketones buy no extra deep sleep (correlation 0.02) [16].
• Likely: sustained ketosis genuinely coincides with better sleep for me — faster onset, fewer awakenings, remembered dreams, a calmer heart. It has repeated too many times, in both directions, to be noise.
• Plausible, not proven: much of the benefit may flow through fewer bathroom trips and a resolved apnea burden when leaner [5–10]. Ketosis may also act directly on sleep pressure and slow-wave sleep [11,13,14]. My data cannot separate five changes that moved together — and anyone who claims their N-of-1 can is selling something.
• Still not supported: that higher ketone numbers beat moderate ones — and now, that ketosis adds deep sleep or REM at all, at least as a wrist device measures them.
• Still not established: causation. Ketosis, weight, breathing, and bladder all changed together, and my file cannot pull them apart. And certainly not that ketosis treats insomnia or apnea in adults generally.

The sentence I trust now

Before the audit, I wanted to write:

Therapeutic ketosis makes me sleep deeply because high ketones raise my REM and deep sleep.

After the audit, the confessions, and the stage data, I trust a longer, less elegant sentence:

A lifetime of light sleep, rare remembered dreams, midlife nocturia, and weight-linked sleep apnea set my baseline. Across four recorded attempts — and now 431 nights of stage data — ketosis repeatedly lined up with falling asleep faster, waking less, a slower sleeping heart, remembered dreams, and better watch scores. It did not add deep sleep or REM. The depth of my sleep did not track the height of my ketones. And because ketosis also brought weight loss, quieter breathing, and fewer bathroom trips, my file still cannot prove that ketones themselves changed my sleep at all. The state looks like the switch. The number is not the dimmer. And the switch may be in my bladder and my airway, not in my brain.

Fewer fireworks. Better odds of being true. And a lesson I did not go looking for: the one claim I would have bet money on — that ketosis gives me deeper sleep — is the one the data would not give me. What it gave me instead is a better question, and a better place to look. Ask me after the next break.

And if any part of my story sounded like your nights — the thin sleep, the closed notebook of dreams, the alarm clock with no snooze button — then the most important sentence in this letter is yours to say to your doctor:

"I'd like to talk about my sleep — and whether I should be tested."

The stage data, attempt by attempt (blog only)

The letter gives the pooled numbers. Here is every stretch on its own, in order, so you can see what I saw. “Missing” is the number of calendar nights in that stretch with no recording at all — the watch failed to score them, and in my experience those were almost always the worst nights, so they are a signal too. Hours asleep and score come from the watch; deep and REM are the minutes I read out of the app by hand.

Read down the deep-sleep column and the pattern is the one in the letter: the stage minutes drift upward with the calendar — 2025 nights are shallower than 2026 nights — and the ketosis rows do not stand out from the off rows next to them. Now read the score column: every ketosis row beats every off row. Whatever ketosis is doing to my nights, the watch sees it in the score, the continuity and the heart rate, not in the stages. (The watch stopped saving sleep from mid-March to late June 2026 — a software fault, not a break in the experiment — which is why that stretch is absent.)

Two honest caveats that belong in the blog version. First, I set the first fourteen days of each attempt aside for the pooled numbers in the letter, because the start of ketosis is reliably my roughest sleep; this table includes them, which is why the Jan–Feb 2025 row shows 59 minutes of deep sleep while the letter’s day-15-onward figure for the same attempt is 62. Second, every stage minute here is a wrist estimate transcribed by hand from the app, one night at a time, and I have flagged a handful of nights where the app’s own minutes and percentages do not agree. None of them move the averages.

What this may mean for patients

A few lessons survive every caveat in this piece.

If your weight is up and your sleep is poor, think about apnea and nocturia first. They are common, they travel together, they fragment sleep night after night, and both are treatable.[5–10]

If anyone has heard you snore, gasp, or stop breathing — get tested. Not eventually. A home sleep test is simple, and untreated apnea taxes the heart, the brain, and blood pressure every single night.

If you lose weight and your sleep transforms, let a re-test make the big decisions. Apnea can retreat with weight loss — sometimes dramatically — and it can return quietly. Feelings are not a sleep study.

Ketosis is one tool, not a universal cure. My data says it coincides with less broken, longer sleep and a calmer heart for me, repeatedly, across four attempts. It does not say it will do that for you, it does not show deeper sleep on a wrist device, and it cannot say whether the ketones, the weight, the quieter airway, or the calmer bladder deserves the credit.

And if you track, track long enough to see climate. One good week is weather. The pattern that returns after you stop and start again is the one worth discussing with your doctor.

— Barry

I post my blood ketone number every day — including the bad ones — in a 90-second video: youtube.com/@neuraliftmd. Day one is there. So is every failure since.

If you want to do something like this with a physician watching, I’m taking on a small founding cohort. It starts with a conversation: calendly.com/hello-neuralift/30min

Educational only, not medical advice. Consumer wearables estimate sleep and cannot diagnose a sleep disorder. Do not stop CPAP or any prescribed sleep treatment without medical reassessment.

Where all this comes from

1. Dublin B. Personal Polar Unite sleep records (sleep score, continuity, hours, heart rate, HRV — no sleep-stage columns) and blood beta-hydroxybutyrate log, Jan 6, 2025 – Sep 28, 2026. Unpublished N-of-1 dataset. — The primary dataset: every night the watch recorded on the Polar Flow website, Jan 2025 to Sep 2026 — score, continuity, hours, heart rate, HRV. It holds no sleep-stage columns; those came from ref 16.
2. Schredl M, Remedios A, Marin-Dragu S, et al. Dream recall frequency, lucid dream frequency, and personality during the COVID-19 pandemic. Imagination, Cognition and Personality. 2022;42(2):113–133. — A single online survey of 1,537 people during the pandemic. It documents the large person-to-person spread in dream recall and its link to personality (openness). It does not test stability over time — that is ref 3.
3. Watson D. To dream, perchance to remember: individual differences in dream recall. Personality and Individual Differences. 2003. — 193 students kept daily dream diaries for 14 weeks. Dream recall was strongly stable across two months and tracked openness, absorption and fantasy-proneness. The classic evidence that recall is a trait.
4. Kalmbach DA, Anderson JR, Drake CL. The impact of stress on sleep: pathogenic sleep reactivity as a vulnerability to insomnia and circadian disorders. J Sleep Res. 2018;27(6):e12710. — A review from the Henry Ford sleep group. Sleep reactivity — how much a person’s sleep falls apart under stress — behaves like a trait and predicts who develops insomnia. “Sleep effort” is my phrase, not theirs.
5. Ancoli-Israel S, Bliwise DL, Nørgaard JP. The effect of nocturia on sleep. Sleep Med Rev. 2011;15(2):91–97. — Review. Two or more trips a night fragments sleep, lowers daytime function and quality of life. Published online 2010, in print 2011.
6. Schneider T, et al. Impact of nocturia on sleep and energy. 2007. — A narrative review in a urology supplement: frequent night-time awakenings disturb normal sleep architecture and daytime energy.
7. Kemmer H. Schlafapnoe in der Urologie [Sleep apnea in urology]. Der Urologe. 2009;48(10):1199–1202. — A German urology review listing nocturia among the urological consequences of obstructive sleep apnea. Cited for the direction of the link; ref 8 carries the weight.
8. Di Bello F, Napolitano L, Abate M, et al. Nocturia and obstructive sleep apnea syndrome: a systematic review. Sleep Med Rev. 2023;69:101787. — Systematic review. Nocturia rises with apnea severity; the mechanism is not settled, and whether treating apnea fixes the nocturia still needs trials.
9. Foster GD, et al. A randomized study on the effect of weight loss on obstructive sleep apnea among obese patients with type 2 diabetes: the Sleep AHEAD study. Arch Intern Med. 2009. — The Sleep AHEAD trial: 264 adults with obesity and type 2 diabetes. At one year the intensive weight-loss group’s apnea index fell about 9.7 events an hour more than controls; remission 13.6% vs 3.5%. Realistic, not guaranteed.
10. Carneiro-Barrera A, Amaro-Gahete FJ, Guillén-Riquelme A, et al. Effect of an interdisciplinary weight loss and lifestyle intervention on obstructive sleep apnea severity: the INTERAPNEA randomized clinical trial. JAMA Network Open. 2022;5(4):e228212. — The INTERAPNEA trial: 89 men with moderate-to-severe apnea. An 8-week programme (weight loss plus exercise, sleep hygiene and alcohol/tobacco advice) cut the apnea index 51% at 8 weeks and 57% at 6 months; complete remission 15% then 29%. The intervention was broader than weight loss alone.
11. Afaghi A, O'Connor H, Chow CM. Acute effects of the very low carbohydrate diet on sleep indices. Nutritional Neuroscience. 2008. — 14 healthy non-obese men in a sleep lab. After a very-low-carbohydrate meal, slow-wave (deep) sleep rose and REM fell, both acutely and at 48 hours. Tiny, acute, and the opposite REM direction to what I expected — which is partly why I went and checked my own stages.
12. Vlahoyiannis A, et al. A systematic review, meta-analysis and meta-regression on the effects of carbohydrates on sleep. Nutrients. 2021. — Meta-analysis of 11 studies. Lower carbohydrate intake modestly raised deep sleep (roughly 8–9 minutes); higher intake modestly raised REM. Small effects, small evidence base.
13. Pasca L, Quaranta CA, Grumi S, et al. The effects of ketogenic dietary therapies on sleep: a scoping review. J Sleep Res. 2024;33(4):e14073. — Scoping review of 20 studies, mostly in people with epilepsy and other neurological conditions. Signals of better sleep quality and more REM; methods were heterogeneous. Not a study of ordinary adults.
14. O'Hearn LA. The therapeutic properties of ketogenic diets, slow-wave sleep, and circadian synchrony. Curr Opin Endocrinol Diabetes Obes. 2021. — A narrative opinion review proposing mechanisms — adenosine, NAD, slow-wave sleep, circadian timing — by which ketogenic metabolism might shape sleep. Hypothesis-generating, not a trial.
15. Khosla S, et al. Consumer sleep technology: an American Academy of Sleep Medicine position statement. J Clin Sleep Med. 2018. — The American Academy of Sleep Medicine’s position: consumer sleep technology is not a substitute for validated diagnostic tools and should not be used alone to diagnose or treat sleep disorders.
16. Dublin B. Polar Sleep Plus Stages nightly readings (deep sleep, REM) read from the Polar Flow app for every recorded night, Jan 8, 2025 – Mar 12, 2026 and Jun 25 – Sep 25, 2026; 431 nights with stage data across four ketosis attempts and three breaks, matched to the rolling 7-day blood BHB average. Unpublished N-of-1 dataset. — My own record, and the backbone of the new section. The watch app does not export stages, so I read every recorded night out of the app by hand: 431 nights with deep and REM minutes, all four attempts and all three breaks, matched to my 7-day ketone average. Missing nights were counted, not ignored — a night the watch failed to record was almost always a bad one.