The HPA axis & cortisol

The sequence is a hormonal relay with a built-in brake. Neurons in the paraventricular nucleus of the hypothalamus release CRH (corticotropin-releasing hormone) into the small portal blood supply feeding the pituitary. The pituitary answers with ACTH into the general circulation, which reaches the adrenal cortex and triggers release of cortisol ^[1]. Cortisol then feeds back onto pituitary and hypothalamus and shuts the signal off, which is what returns the system to baseline once the threat has passed.

What gets the hypothalamus going depends on the kind of stressor, and the distinction is more useful than it sounds. A reactive stressor is one the body can already feel: blood loss, low blood sugar, infection, pain. Those drive the paraventricular nucleus fairly directly through brainstem relays. An anticipatory stressor is a situation the brain has judged threatening before anything physical has happened: a deadline, a confrontation, a scan result. Those run through limbic structures and work largely by disinhibition, releasing a tonic GABAergic brake that normally holds the hypothalamus quiet ^[3]. That is why a purely psychological threat produces a full endocrine response, and why the amygdala and prefrontal cortex are as much part of this axis as any gland.

Cortisol acts through two receptors that are not interchangeable. The mineralocorticoid receptor (MR) binds it with roughly ten times the affinity of the glucocorticoid receptor (GR). At the daily trough MR is largely occupied and GR largely empty; only a pulse or a stress response pushes concentrations high enough to recruit GR in quantity ^[6]. That is the mechanism behind an otherwise puzzling fact: the same hormone supports normal function at low levels and causes damage at high ones, because it is engaging different machinery at each.

The pulsing deserves its own paragraph, because it explains why a single blood draw is a poor measurement. Cortisol is released in discrete ultradian pulses about once an hour, and gene transcription in target tissues tracks those pulses rather than the average concentration ^[5]. A constant infusion delivering the same total dose does not produce the same downstream effect. The axis encodes information in a pattern, not a level, so one number tells you very little about whether that pattern is intact.

The cascade is also not the whole stress system. The sympathetic branch runs alongside it and faster, putting out adrenaline within seconds while the HPA axis is still assembling its response over minutes ^[7]. Much of what feels like stress in the moment is that fast branch, not this one.

In an acute burst cortisol is a mobiliser. It raises blood glucose through hepatic gluconeogenesis and reduced uptake elsewhere, makes stored fat and protein available, and dials down building projects that can wait, including parts of digestion, reproduction and growth ^[1]. A sensible trade when the challenge is real and brief.

The more interesting half of the story is that cortisol often acts as a brake on the stress response itself. The classic synthesis sorts glucocorticoid actions into four categories rather than one ^[4]. A permissive action is one where baseline cortisol is needed for another system to work at all. A suppressive action is where the stress-level rise stops another response overshooting, the immune system being the clearest case: without glucocorticoid restraint an inflammatory response can do more damage than the pathogen. A stimulatory action amplifies the ongoing response. A preparative action changes how the body will handle the next stressor.

Sorting an effect into the right box changes the conclusion. If cortisol is mostly restraining inflammation, blunting it indiscriminately is not obviously a good idea, and "lower cortisol" is not a health goal on its own. This is the same error as treating every reactive oxygen species as damage; see antioxidants and oxidative stress.

Layered over the pulses is a large circadian rhythm. Cortisol falls through the evening to a trough around the middle of the night, climbs in the last hours of sleep, and peaks roughly 30 to 45 minutes after waking. That peak is a distinct event, the cortisol awakening response, driven by the act of waking rather than only by the clock, and it involves a change in adrenal sensitivity to ACTH as well as a rise in ACTH itself ^[8].

The rhythm is set by the same master clock that sets sleep, the suprachiasmatic nucleus, and glucocorticoids in turn are one of the main internal signals keeping peripheral clocks aligned with it ^[6]. That two way relationship is why sleep and the circadian system cannot be separated from stress physiology.

Sleep loss moves this measurably. Restricting eleven healthy young men to four hours in bed for six nights raised evening cortisol concentrations against a fully rested condition, alongside lower glucose tolerance, lower thyrotropin and increased sympathetic activity ^[11]. The authors noted the pattern resembled ageing: chronic short sleep does not just make a person tired, it makes the endocrine profile look older.

Measurement is where most consumer-facing claims fall apart. A single morning blood cortisol says almost nothing, because it depends on how long ago the person woke, whether the draw was stressful, and where in an ultradian pulse it landed. The measures that carry information are repeated salivary samples across a day, giving the awakening response and the diurnal slope, and hair cortisol, which integrates over months. The standard laboratory provocation is the Trier Social Stress Test: an unprepared speech and mental arithmetic before an impassive panel, reliably producing a two to fourfold rise in salivary cortisol ^[9]. Read any cortisol claim against which of these it used.

Sustained activation is the clearest worked example of allostatic load: the cumulative wear of a response that keeps switching on and never fully off ^[2]. The costs follow from what cortisol does acutely. Prolonged gluconeogenesis and reduced glucose uptake push toward insulin resistance and visceral fat. Prolonged suppression of anabolic processes impairs repair. Prolonged evening elevation fragments sleep, and fragmented sleep further dysregulates the axis, which is the loop that makes burnout self-sustaining.

The brain is a target as much as a controller. The hippocampus is dense in both MR and GR and normally applies the brake; sustained glucocorticoid exposure remodels its dendrites and impairs the memory it supports, so chronic stress degrades part of its own off switch ^[2]. The amygdala moves the other way, growing more reactive, while the prefrontal cortex loses the top-down control that would end the response.

Timing matters as much as amount. The same exposure produces different outcomes depending on when in a life it lands, because different structures develop at different rates, and one model of stress damage does not fit the prenatal period, childhood, adolescence and old age alike ^[10].

It is worth stating what is not established. There is no validated cortisol number that identifies a person as chronically stressed. Chronic activation can present as sustained high output, as an exaggerated response to each new stressor, or eventually as a blunted flattened profile ^[3], and those look nothing alike on a test. The shape of the rhythm carries more information than any single value, and even the shape is read against a wide normal range.

The supplement market sells a story in which chronic stress eventually exhausts the adrenal glands so that they can no longer make enough cortisol, producing fatigue a support formula can correct. This is not a recognised condition and the evidence does not support it. A systematic review screened 3,470 papers, included 58 that had measured adrenal axis function alongside fatigue, and found conflicting results throughout, with much of the literature resting on assessment methods endocrinology does not endorse ^[12]. Its conclusion was that adrenal fatigue remains a myth.

That is not the same as saying the symptoms are imaginary. People who present with this complaint are genuinely exhausted. The point is that a wrong mechanism sends them toward the wrong interventions and away from the right diagnosis. Adrenal insufficiency is a real and dangerous condition with a real diagnostic test; so are hypothyroidism, anaemia, sleep apnoea and depression, all of which present with fatigue and all of which are treatable. A store-bought "adrenal support" blend does not distinguish between them.

There is a related trap on the other side. Because a chronically activated axis can end up blunted ^[3], a low reading is sometimes offered as proof of the exhaustion story. It is not; low output has several causes and the direction of travel cannot be read off one measurement. Persistent unexplained fatigue is a reason to see a clinician, not a reason to buy a formula.

The heavy lifting is behavioural. Consistent sleep timing is the largest lever, because sleep debt raises evening cortisol directly ^[11] and because the axis is anchored to the circadian clock ^[6]. Morning light anchors that clock. Training loads need genuine recovery: exercise is a real physiological stressor, and the point of a training block is to spend that stress on adaptation rather than accumulate it. Physical activity and social support both track with a lower chronic stress burden ^[2].

On the compound side, the picture is thinner than the marketing suggests but not empty. Ashwagandha is the best supported of the adaptogens here. In a 60 day double-blind trial in 64 chronically stressed adults, a high-concentration root extract at 600 mg per day cut serum cortisol significantly against placebo alongside improvement on all three stress scales ^[13]. A later eight week study compared doses, finding reductions in perceived stress and serum cortisol at both 250 mg and 600 mg per day, larger at the higher dose ^[14]. A meta-analysis of 23 randomised placebo-controlled trials across 1,706 participants found a significant cortisol reduction, cautioning that heterogeneity between trials was substantial ^[15].

Phosphatidylserine has a narrower, older evidence base and targets the exercise-induced spike rather than baseline. In nine healthy men, 800 mg per day for ten days blunted both the ACTH and the cortisol response to cycling without touching growth hormone or prolactin ^[16]. A crossover in ten men at 600 mg per day found peak cortisol about 39 percent lower and area under the curve about 35 percent lower than placebo ^[17]. Both trials are small and old, and neither ran long enough to say whether the blunting improves training outcomes.

Two cautions. The goal is never to abolish cortisol; a suppressed axis is worse than a briefly elevated one, and the suppressive actions above are necessary work ^[4]. And a compound that lowers a number in a trial has not been shown to fix the underlying pattern, which is a rhythm rather than a level ^[5]. Not medical advice.