Do We Need Sleep?

The researcher at University of Chicago did something cruel in 1999. He took eleven healthy young men and let them sleep only four hours a night for six nights. Not a month, not a year. Six nights. Then he gave them an oral glucose tolerance test, the same one we use to diagnose diabetes.

All eleven became temporarily diabetic. Their insulin sensitivity dropped by 40%. Their glucose disposal slowed to levels typically seen in sixty-year-olds. One week of bad sleep had aged their metabolisms by three decades.

Then he let them sleep twelve hours a night for a week. Everything reversed.

I think about this study constantly because nobody believes it applies to them. The executive who emails me at 3 AM. The resident who brags about thirty-hour shifts. The patient who tells me she “functions fine” on five hours while her hemoglobin A1c climbs month after month. They all think they’re different. Their bodies are doing exactly what those eleven men’s bodies did, they just can’t feel it happening.

Here’s what we missed about sleep for most of medical history. We thought of it as passive, a shutdown state, the biological equivalent of turning off your computer. Metabolically inert. The evidence now suggests the opposite. Sleep is when your body does some of its most critical metabolic work. Glucose regulation recalibrates. Insulin sensitivity resets. Growth hormone pulses. Cortisol drops to baseline so it can spike appropriately the next day.

Miss that window and your metabolism doesn’t just pause. It degrades.

A woman came to see me six months ago. Thirty-four, tech worker, ate impeccably, trained five days a week. Her fasting glucose had crept to 112 over two years despite doing everything the internet told her to do for metabolic health. Cut carbs, tried intermittent fasting, added more cardio. Nothing worked.

I asked about sleep. She was averaging 5.5 hours, had been for three years since her startup raised Series B funding. Her Oura ring gave her decent sleep scores. She thought she was managing fine.

We didn’t change her diet. Didn’t adjust her training. Just prioritized eight hours in bed with actual sleep hygiene. No screens after 9 PM, cool dark room, consistent schedule. Three months later her fasting glucose was 89.

This is the thing about sleep and glucose that makes it so insidious. The relationship runs both ways and creates a trap. Sleep deprivation impairs your glucose metabolism. Poor glucose control disrupts your sleep architecture. You end up in a metabolic doom loop where each problem makes the other worse, and because you can’t feel your insulin sensitivity declining in real-time, you don’t connect the dots until you’re pre-diabetic.

The mechanism is straightforward once you see it. During deep sleep, your sympathetic nervous system finally shuts up. Cortisol drops. Growth hormone peaks. This hormonal shift is when your cells restore insulin sensitivity. It’s not incidental. It’s scheduled maintenance that only happens during specific sleep stages.

Cut that maintenance window short and your cells start ignoring insulin’s signals. Your pancreas compensates by making more insulin. That works until it doesn’t. The path from sleep deprivation to type 2 diabetes isn’t metaphorical. It’s mechanical.

A 2015 study in the American Journal of Respiratory and Critical Care Medicine found that sleep serves as the foundation for cellular toxin removal in the brain. The glymphatic system, discovered only in 2012, operates primarily during sleep to clear metabolic waste including amyloid-beta, the protein that accumulates in Alzheimer’s disease. Your brain literally takes out the trash while you sleep. Stay awake and the trash accumulates.

We’ve known this matters for women’s health specifically. The 2022 Lancet review on menopause as a cardiometabolic transition pointed out that sleep disruption during perimenopause accelerates metabolic dysfunction independent of hormonal changes. Women who sleep poorly through menopause show faster deterioration in glucose regulation, lipid profiles, and inflammatory markers than women who maintain sleep quality even with the same hormonal fluctuations.

But here’s where it gets strange. We have better measurement tools than ever. Wearables track sleep stages with reasonable accuracy. They measure heart rate variability, respiratory rate, body temperature fluctuations. They generate sleep scores that correlate with actual polysomnography. And somehow we’re using all this precision measurement to convince ourselves we need less of what we’re measuring.

I see this pattern constantly. Someone shows me immaculate sleep tracking data. Their device says they got ninety minutes of deep sleep, perfect REM cycles, optimal sleep efficiency. Then I look at their metabolic labs over six months and see steady degradation. Fasting glucose climbing. Hemoglobin A1c creeping up. Inflammatory markers elevated. The wearable said they were fine. Their endocrine system tells a different story.

The problem is we’re measuring the wrong thing. Or rather, we’re measuring the right things but ignoring what the measurements mean. A sleep score tells you about sleep architecture. It doesn’t tell you whether that architecture was sufficient for your metabolic load, your training stress, your cortisol burden, your inflammatory state. Those interactions are invisible to the wearable but visible in your bloodwork if anyone bothers to look at both simultaneously.

This is where artificial intelligence becomes actually useful instead of just hyped. Not because AI can tell you to sleep more, that’s trivial. Because AI can find patterns in multivariate physiological data that humans miss.

Your sleep quality affects your glucose regulation, which affects your inflammatory response, which affects your recovery capacity, which determines how your body responds to training, which influences your sleep architecture. That’s a five-way interaction playing out across weeks or months with individual variation that makes population studies nearly useless for predicting your specific trajectory.

A human doctor looking at your labs every three months might notice your glucose trending up. They probably won’t connect it to the sleep deterioration that started six weeks earlier, or notice that the relationship is non-linear in you specifically because of how your cortisol responds to poor sleep differently than the average patient. We don’t have the bandwidth to track these multivariate relationships across time in every patient. The system barely has bandwidth for the quarterly lab review.

This is what systems like moccet are actually built to do. Not replace clinical judgment, but see connections across data streams that clinical judgment operating quarterly in fifteen-minute visits will miss. The platform ingests sleep architecture from wearables, metabolic markers from labs, training load, calendar stress, subjective recovery scores, and maps how these variables interact in you specifically over time.

It’s not telling you generic truths about sleep and glucose. It’s showing you that your glucose regulation degrades specifically when your deep sleep drops below seventy-five minutes for more than four consecutive nights, and that this happens predictably ten days before your HRV starts declining, giving you a warning before the metabolic damage becomes measurable in standard labs.

That kind of personalized pattern recognition is how you’d practice medicine if you had infinite time and perfect recall. Since neither exists, you build systems that approximate it.

The longevity medicine crowd finds this inconvenient. They’re optimizing everything else. Supplements stacked six deep. Peptides injected. Metformin off-label for anti-aging. Ice baths and sauna protocols. Blood panels quarterly tracking fifty biomarkers. Enormous effort and expense trying to extend healthspan by manipulating every pathway except the one that’s free and we already know works.

Sleep is probably the most powerful longevity intervention available and we treat it like an obstacle to overcome. The same people taking NMN and resveratrol for their theoretical effects on NAD+ metabolism will short their sleep to get more done, directly impairing the cellular repair processes those supplements supposedly enhance.

Some researchers have looked at whether we could engineer our way past needing sleep. Genetic studies in flies and mice show you can reduce sleep need through specific mutations. The problem appears immediately. Shortened lifespan. Impaired learning. Reduced stress resistance. Memory deficits. The mutations that let you sleep less break other things because sleep isn’t one function, it’s the time when dozens of essential maintenance functions happen.

Evolution had four billion years to optimize this. If skipping sleep were free, something would have evolved to skip it. Instead every animal ever studied has some form of sleep or sleep-like state. Even jellyfish, which don’t have brains, have periods of reduced activity and responsiveness that look like sleep. This is conserved across everything with a nervous system because it’s not optional.

The military spent decades trying to crack this. If you could create soldiers who don’t need sleep, the tactical advantage would be decisive. They tested every stimulant, every pharmacological intervention, every training protocol. The conclusion was always identical. You can defer sleep temporarily. You cannot eliminate the need for it. And the cognitive and physical performance degradation during extended wakefulness happens regardless of how alert the person believes they are.

That last part is crucial. Subjective alertness and objective performance separate dramatically under sleep deprivation. People think they’re functioning normally while their reaction times, decision-making, and metabolic regulation have degraded to dangerous levels. This is why sleep-deprived surgeons make more errors while being confident in their performance. It’s why drowsy driving kills more people than drunk driving in some states. The subjective feeling of capability persists long after the capability itself has vanished.

Which brings us back to wearables and measurement. The devices are measuring real things. The question is whether we use that information to optimize our biology or override it.

My patient with the climbing glucose now sleeps eight hours consistently. She joined the moccet waitlist because she wanted to see how sleep changes affected all her markers simultaneously, not just glucose. Within six weeks the platform flagged something her quarterly doctor visits had missed. Her inflammatory markers dropped, her HRV stabilized, and her training performance improved, but the improvements happened with a specific lag pattern. Her inflammation responded within two weeks of better sleep. Her HRV took four weeks. Her glucose took six weeks but then stabilized at a better level than she’d seen in three years.

That temporal pattern matters for behavior change. If you fix your sleep and check your glucose a week later and see no improvement, you might conclude sleep wasn’t the problem. The AI pattern recognition showed her that her body needed six weeks to rebuild insulin sensitivity, but the process was happening predictably even when it wasn’t immediately visible.

This is what the future of healthcare looks like, assuming we’re smart enough to build it. Not replacing doctors with algorithms, but giving doctors and patients both the ability to see physiological patterns that emerge over weeks and months across multiple systems. Your doctor can’t track how your sleep architecture correlates with your inflammatory markers with your training recovery with your glucose control in real-time. The system can. Then you have that conversation with your doctor informed by actual data about your physiology instead of population averages.

The answer to whether we need sleep is boringly straightforward. Yes. Completely, non-negotiably yes. Every controlled study on sleep deprivation shows progressive physiological dysfunction. Every attempt to eliminate sleep in mammals has failed or killed the animal. Total sleep deprivation in humans would be fatal if we could keep someone awake long enough, we know this from animal models.

The interesting question isn’t whether we need it. It’s whether we can finally build systems good enough to show each person specifically how much they need and what happens when they don’t get it. Because the fundamental problem with sleep is the same as the fundamental problem with most preventive medicine. The damage is invisible while it’s happening and obvious only after it’s done.

You can feel tired from bad sleep. You cannot feel your insulin sensitivity declining, your inflammatory markers rising, your amyloid-beta accumulating, your cortisol dysregulating. Those processes happen silently until they manifest as disease. By then you’ve spent years ignoring signals your body couldn’t make loud enough for you to notice.

Measurement alone doesn’t solve this. We’ve been measuring things for decades and people still wreck their health through sleep deprivation. What might solve it is measurement sophisticated enough to show you the specific, individual, quantified consequences of your sleep patterns before those consequences become irreversible.

The technology exists. The platform at moccet integrates it, sleep tracking from wearables, metabolic labs, inflammatory markers, training data, recovery metrics, and maps how they interact in you personally. Not generic advice about sleep hygiene, though that matters too. Specific data showing how your sleep patterns over the last six weeks affected your metabolic markers and what trajectory you’re on if nothing changes.

You can join the waitlist at moccet and get early acess. What you get isn’t motivation to sleep more. Most people know they should sleep more. What you get is evidence that your specific biology requires it, visible in your own data, connected across the systems you didn’t realize were connected.

The woman with the climbing glucose didn’t need me to tell her sleep was important. She needed to see that her particular metabolic dysfunction was mechanistically linked to her particular sleep pattern in a way that was reversible if she acted and progressive if she didn’t. Once she could see that in her own numbers, the behavior change wasn’t about willpower anymore. It was about not lying to herself about what the data showed.

Maybe that’s the actual revolution in healthcare. Not finding ways to transcend our biology, but building systems precise enough that we stop pretending we can.