On June 8, at the American Diabetes Association meeting, Adaptyx Biosciences put up data the field has chased for ten years and never produced: the
first continuous, multi-day readings of free cortisol
pulled from a sensor sitting in the fluid just under the skin. The Stanford spinout walked off with the
2026 ADA Innovation Challenge
.
Investors had already placed their bets.
Level Zero Health
is putting hormones on a microneedle patch for $6.9 million.
London's Sava Technologies raised €16.6 million on a line that tells you exactly how this market is being pitched:
"glucose is only the beginning."
The promise is that cortisol will do for stress and sleep what continuous glucose monitoring did for diabetes. It skips the part that matters, which is that cortisol is a far nastier thing to measure than glucose, in almost every way that breaks a sensor.
Why cortisol monitoring is one million times harder than glucose tracking
Let's start with how much of it there is.
Glucose floods the body, running in the millimolar range and barely thinning out between blood and the interstitial fluid a sensor sits in, which is why a glucometer can be slightly sloppy and still land right.
Free cortisol, the fraction that actually crosses into cells and does anything,
moves in the low nanomolar range
, somewhere between ten thousand and a million times more dilute.
If glucose is sugar you can taste in your coffee, free cortisol is a few grains of something dropped in a bathtub, and the sensor has to read it correctly while sitting in that bathtub.
Glucose also came with a free gift. A cheap, durable enzyme, glucose oxidase, eats glucose and hands back an electrical signal, and it has been the engine inside every glucometer since 1967. Cortisol has no such enzyme.
Nobody sells cortisol oxidase, and nobody will invent one for a patch. So instead of letting chemistry do the counting, a cortisol sensor has to physically catch the whole intact molecule and recognize its shape, and that is where the engineering gets expensive.
Catching it is the part everyone can already do - a hospital ELISA grabs cortisol with an antibody and reads it cleanly and has for decades. The trouble is that those tests fire once, like a mousetrap.
They bind the target, give one number, and the surface is spent. A device meant to follow cortisol across a night of sleep needs the opposite, something that grabs the molecule, reports it, lets it go, resets, and repeats that thousands of times over several days without anyone refilling a reagent. That reset is the whole problem, and where most of the real science of the last few years has gone.
How DNA aptamer sensors solved the continuous cortisol problem
The frontrunner solution is a strand of DNA. An
electrochemical aptamer
is a short piece of DNA anchored to a gold electrode and tipped with a marker that shuttles electrons, usually methylene blue.
Cortisol lands on it, and the strand folds, swinging the marker toward the electrode so the current climbs. Cortisol drifts off, the strand relaxes, the current settles back. Nothing is consumed, so the same electrode answers over and over.
Adaptyx calls its version a molecular switch,
licensed from the Stanford labs
of H. Tom Soh and Joseph DeSimone and built on a $14 million seed round last October, and it runs sixteen at once, eight reading cortisol against eight controls. Then the body fights back.
Drop one of these electrodes into living tissue, and proteins start settling onto the gold within minutes, like grease creeping across a camera lens, and the baseline wanders ten to thirty percent in an hour until the signal is buried under the gunk.
The repair that made continuous sensing usable came from
Kevin Plaxco's lab
: add a second marker at a different voltage that feels only the fouling, then subtract it out. Drift drops under 2%. It works, and it is worth being clear about what it is, a correction propping up a measurement that is not stable on its own.
A second camp throws out the biology entirely.
Molecularly imprinted polymers
cast a hard synthetic cavity in the exact shape of cortisol, a lock with no key returned, which fouls less and lasts longer but is slow to reset.
A Zhejiang University group built a sweat version in 2025 that hit a 0.36 nanomolar detection limit across eight reuse cycles, real numbers that arrive with a tax: it needs iontophoresis, a small current that pushes a drug into the skin to force sweat out, which means more electronics, more power, and more to break on the patch.
Interstitial fluid vs. sweat: Which cortisol sensor approach will win?
The fluid a company chooses decides its future.
Adaptyx
and Level Zero push microneedles through the skin into interstitial fluid, which shadows free blood cortisol closely and lands them squarely in regulated-device territory.
The sweat crowd stays on top of the skin, noninvasive, easier to sell, and pays in noise. John Rogers's group at Northwestern published a genuinely clever design in
Nature Sensors
this year that turns paper test strips into time-stamped sweat readings and tracks serum at 0.94, but each zone fires once.
Lined up, that is a string of snapshots, and calling it continuous is a stretch. Down at the consumer end, the language gets looser than the data. EnLiSense sells Corti, a passive-sweat cortisol and melatonin band, and to its credit, calls it
a general wellness device
with no FDA clearance, backed by an
October study
showing better than 0.92 agreement with saliva across 43 people.
Clair Health
, fresh off an $11.6 million round in June, does not measure a hormone at all. It infers cortisol from heart rate, skin temperature, and similar signals through a machine-learning model, the same move Oura uses to guess menstrual-cycle phase.
That can be a perfectly good product. It is not cortisol sensing, and the gap stops being a technicality the instant anyone makes a medical claim. That medical claim is the entire ballgame, and FDA has already tipped its hand.
A cortisol test is Class II under
21 CFR 862.1205
, but every cleared predicate is a bench immunoassay, so a continuous on-body device has nothing to call itself equivalent to and almost certainly enters through De Novo, the
same door Dexcom's G6 used in 2018
to create the integrated-CGM category.
FDA approval: The real barrier to continuous cortisol devices
That route is the map worth copying, special controls and all, down to the
over-the-counter clearances
Stelo and Abbott's Lingo collected in 2024. What will not work is hiding behind the word wellness.
Last July, FDA sent WHOOP a
warning letter
over a blood-pressure feature and threw out the wellness framing flatly, ruling that estimating someone's blood pressure cannot be walked back from diagnosing hypertension, whatever the marketing says. A band that murmurs about your stress has room to operate. That warning letter was recently resolved.
The moment it mentions Addison's disease, Cushing's, or when to take a hydrocortisone dose, it is a medical device and gets treated like one. Cushing's syndrome is less about a single high reading than a rhythm that has broken, the overnight low that should arrive and does not, which a one-shot saliva or urine test tends to miss, as a March 2026
review in JCEM
spells out.
People on hydrocortisone for Addison's or congenital adrenal hyperplasia are largely blind, and
published pharmacology
shows some of them sitting uncovered for nearly half a day while the labs their doctor orders come back normal. Intensive care is the same gap from the other side, where
sepsis guidance
tells clinicians not to trust a single spot cortisol in a critically ill patient, which is precisely the hole a continuous line could fill.
So, the need is real, and the proof is not. Adaptyx says its readings showed concordance with mass-spec blood draws, and concordance is carrying a lot of unpaid weight there, because the company has not shown a correlation coefficient, a sample size, or a Bland-Altman plot behind it.
Sweat sensors still agree with blood only loosely at rest, and worse the moment someone moves or panics. Not one continuous cortisol device holds any FDA clearance, which makes this pre-commercial, and the billion-dollar market figures already circulating are worth nothing until a product exists to pin them to.
The only honest benchmark sitting next to it is CGM, a US market one, a 2025 report puts
on track for $8.4 billion by 2033
, built over roughly twenty years on the backs of tens of millions of diabetic patients.
Here is the part Adaptyx and the labs around it have earned- they cleared the one bar that demanded new science, proving a reusable sensor can track a hormone present at a few billionths of a mole through an ordinary day on a real body.
That was the hard problem, and it is solved. Everything left is the unglamorous work that buries most device companies, the long haul from a sensor that works to a cleared product that a hospital will pay for. Cortisol may well become the next thing worth watching around the clock.