GLP-1 Sleep Repair: 3am Wakes Gone, Waking Cortisol Down 34% in 12 Weeks

A 51-year-old partner at an Amsterdam law firm was three months into semaglutide, down 6 kg, and quietly falling apart at 3am. Waking every night between 3:10 and 3:40, mind racing on billing matters. Waking cortisol on a home salivary test was 24 nmol/L — the top of the reference range and well above his own historical baseline. In 12 weeks: wakes gone, waking cortisol 15.8 nmol/L (−34%), and weight loss kept moving. Sleep was the accelerant nobody had wired in.

The presenting state

He had done everything the leaflet said. Injected on schedule. Ate less. Weight moved. But he was accumulating a specific fatigue — the kind that reads as "focused" until it doesn't, and then everything falls off at once. HRV was suppressed. Sleep tracker showed 5.8h effective, four wake events per night, and REM squeezed into the last 90 minutes before the alarm.

GLP-1s alter gastric emptying and can shift the timing of the evening glucose curve. Combined with post-loss metabolic adaptation and pre-existing high work-load cortisol, this reliably surfaces as a middle-of-the-night wake pattern — the 3am cortisol surge finding a light-sleep window it can walk through (Van Cauter et al., Sleep Med Rev 2007).

The protocol

Sleep and cortisol got prioritised over training volume. Counter-intuitive on a weight-loss protocol; correct on a fat-loss with sanity intact one.

• Consistent 22:30 sleep window — held even on court-prep nights. Sleep regularity matters more than duration for cortisol rhythm (Windred et al., Sleep 2024).
• Last meal 3h before bed — the GLP-1 slows gastric emptying already; late food was extending the digestive window into the sleep window.
• Protein floor 1.6 g/kg — front-loaded to breakfast and lunch so evening meal could be smaller without triggering 2am hunger.
• Zero caffeine after 12:00 — caffeine half-life is ~5–6h; the 3pm espresso was still 25% active at bedtime (Drake et al., J Clin Sleep Med 2013).
• Magnesium glycinate 300 mg at 21:30.
• Light exposure protocol — 10 min outdoor light before 09:00, dim indoor light after 20:00. Anchored the circadian phase.
• VILPA bursts moved to AM — three 60-second bouts before 11:00. Kept the cardiovascular signal, kept the evening cortisol curve clean.
• Journal (GLP-1 protocol) — wake events, waking cortisol proxy (subjective wired-tired 0–10), sleep window, caffeine, evening meal timing. The 3am wake correlated 0.71 with meals eaten after 20:30.

What changed

• Sleep continuity: 4 wakes → 0 by week 8.
• Effective sleep: 5.8h → 7.1h.
• Waking cortisol: 24 → 15.8 nmol/L on repeat salivary test at week 12 (−34%).
• HRV: 34 → 46 ms.
• Weight loss continued: −4 kg further across the 12 weeks — the myth that "you have to eat more to fix sleep" got quietly retired.
• Nausea complaints on injection days dropped — the earlier eating window meant the drug wasn't sitting on a full stomach.

Why this worked

GLP-1 protocols routinely miss sleep, and cortisol quietly re-loads the whole system. Fix the 3am wake, and the parasympathetic recovery window (deep sleep, HRV rebuild) that normally protects lean mass and appetite regulation gets to run. Ignore it, and you're paying interest on the medication every single night.

The drug can't do this piece. The pattern layer can.

Sources

• Van Cauter et al., 2007 — Cortisol rhythm and sleep, Sleep Med Rev.
• Windred et al., 2024 — Sleep regularity and cortisol/mortality, Sleep.
• Drake et al., 2013 — Caffeine and sleep, J Clin Sleep Med.
• Fothergill et al., 2016 — Post-loss metabolic adaptation, Obesity.