Sleep

The sleep field in 2026 is splitting along a clean line: interventions with a measurable mechanism and replicated polysomnography evidence, and interventions sold on sleep-tracker vibes. The four articles cited below sit on the first side of that line. Each one is built on a mechanism the body actually runs — thermoregulation, magnesium-mediated GABA-A modulation, suprachiasmatic light entrainment, hypothalamic GH pulse circuitry — not on a hopeful nutrient panel or a tracker score.

The strongest current finding is mechanistic, not pharmacologic. The 2024–2025 Cell paper mapping the hypothalamic circuit behind sleep-driven growth hormone release closed a 50-year gap in our understanding of why deep sleep is the dose-response variable for tissue repair. The implication is practical: chasing total time in bed is the wrong target. Chasing the slow-wave architecture — temperature drop at sleep onset, no alcohol in the last three hours, no late-evening blue light pulse — is the one that moves the recovery dial.

The supplement story has narrowed. Magnesium bisglycinate has one decent placebo-controlled trial behind it (n=155, 2025) and a plausible mechanism via NMDA antagonism and GABA-A potentiation. The effect size is real but modest — about 17 minutes faster sleep onset, roughly half a point on the Pittsburgh Sleep Quality Index. Glycine sits in a similar zone but works through a different lever entirely: it pulls core body temperature down at night by signaling peripheral vasodilation, which is the same lever a cool bedroom pulls. That is not a coincidence; it is the same physiological pathway.

DORAs (dual orexin receptor antagonists like suvorexant and lemborexant) are the only pharmacologic class with a clean mechanistic story for chronic insomnia — they block the wakefulness signal instead of forcing sedation. They are also expensive, prescription-only, and not yet covered by most North American plans. For most readers, they remain a referral, not a buy.

Everything else — magnesium threonate marketing claims, melatonin doses above 1 mg, the tracker-as-protocol industry — is still where we said it was a year ago: tolerated, unsupported, sometimes worse than nothing. The pieces below explain why.

Three questions are likely to move the sleep field in the next 12 months. First, the DORA price wall: lemborexant goes off-patent in several markets between now and 2027, which would put a real pharmacologic option in the under-$30/month range for the first time. Second, the GLP-1 cross-talk: semaglutide and tirzepatide both alter sleep architecture in ways the trial data are only now starting to characterize — some users report dramatically improved deep sleep, others report fragmented REM. The mechanism is probably appetite-driven autonomic stabilization, but the picture isn't clean yet.

Third, and most quietly: the consumer-tracker accuracy problem. Independent validation studies in 2025 showed that the major rings and watches still misclassify roughly 30% of REM episodes when compared with polysomnography. That matters because protocol decisions are increasingly being made off tracker data. Our position remains the same as it was 18 months ago: trackers are good for trend (did this week look like last week?) and bad for absolute scoring (is this stage really REM?). Use them accordingly.

There is also a quiet access conversation happening around CBT-I. The trial evidence for cognitive behavioral therapy for insomnia is the strongest of any intervention in this category, but most regions have wait lists measured in months. Digital CBT-I apps with FDA clearance have closed part of that gap; whether they retain users past the friction point of stimulus-control compliance is the open question, and the early dropout curves are not encouraging.