Sleep architecture after 45: the deep-stage collapse.

Total sleep hours stay roughly flat from 35 to 65. What collapses after 45 is the architecture — and that's the part that matters for cognition, mood, and immune function.

If you wear a sleep tracker, here is what you have probably noticed: your total sleep time at 50 looks similar to what it was at 35. Six and a half, maybe seven hours. What's actually different — and what the tracker shows once you look at the right screen — is that deep sleep has dropped from 90 minutes a night to 30 or 40. REM is shorter. Wake-after-sleep-onset is up. The same total clock time produces a fraction of the restorative output.

The 2024 Walker lab follow-up data quantified it: by 50, women have lost roughly 60% of the deep-sleep duration they had at 30. Men lose slightly less but start their decline earlier. The story is not "older adults need less sleep." It's "older adults are getting less restorative sleep inside the same window."

What's actually happening in the brain

Deep sleep is when the glymphatic system clears metabolic waste from neural tissue — including beta-amyloid, the protein implicated in Alzheimer's pathology. REM sleep consolidates emotional memory and supports the prefrontal regulation that lets you not snap at people the next day. Lose 60% of either and the cognitive, mood, and immune outputs follow.

The estrogen-progesterone story matters here too. Progesterone has GABA-ergic effects that promote deep sleep. As progesterone production drops through perimenopause, deep sleep drops with it. The bidirectional relationship is part of why HRT conversations now include sleep architecture, not just hot flashes.

The four interventions that work, in order

1. Temperature drop. Body temperature has to fall to enter and maintain deep sleep. The most reliable way is environmental: a room at 65-68°F. Cooling sheets, fans, mattress chillers all help. A warm bath 90 minutes before bed paradoxically helps because vasodilation accelerates the subsequent drop. This is the highest-leverage intervention by a wide margin.

2. Alcohol cut-off. Alcohol shreds REM sleep at exactly this age. The 2025 BMJ Open paper put the threshold at any consumption within 4 hours of sleep, even one drink, with measurable REM degradation. The half-glass-of-wine-with-dinner crowd does not have to stop drinking — they have to move it earlier.

3. Light timing. Bright outdoor light within the first hour of waking is one of the few free interventions with strong randomized data for sleep consolidation that same night. Inside-light at 100 lux is not enough; outdoor light at 2,000-10,000 lux is. Ten minutes outside before coffee, every morning, for six weeks.

4. Hormonal-metabolic support. Magnesium glycinate (200-400mg) has the cleanest data for women. Apigenin (chamomile extract) at 50mg has emerging data. Progesterone — for women in the right window with a clinician — is the bigger lever where appropriate.

What doesn't work as well as you'd think

Sleep trackers themselves do not improve sleep. The data is unambiguous: people who wear them score the same as people who don't, after the initial 2-week behavior-change window. Use the tracker for information, not for the intervention.

Melatonin at typical OTC doses (3-10mg) is 10-30x physiological. The actual sleep-onset dose with the best data is 0.3mg. Most products are dosed for marketing, not biology. CBD has surprisingly thin sleep data despite the marketing. Tart cherry juice has decent data but the effect size is small.