Edible Tolerance Explained: Your Receptors Adapt, the Scans Are of Chronic Users, and Nobody Has Timed a Gummy

The same gummy stops working, everybody agrees on why, and almost nobody has checked. Tolerance to THC is real, it happens at the CB1 receptor, and four PET studies we can find have photographed it. All four recruited chronic or dependent users. The oral THC studies mostly gave capsules at 40 to 240mg a day to people who already smoked. The number of published studies that followed a patient on a 5mg bedtime gummy for months and measured what happened, as far as PubMed can tell us on November 7, 2026, is zero.

Our position: the mechanism transfers and the timetable does not. The same receptor takes the hit however THC arrives, so a nightly patient has every reason to expect some version of the same adaptation. How much, how fast and how long it takes to reverse were all measured in somebody else, and the tidy sentences built on those numbers ("two days resets you," "low doses don't cause it") go further than the papers do. The practical half of this lives in our tolerance-break guide. This page is the evidence, including the parts that embarrass the folklore.

Why does the same gummy stop working?

Because the brain adapts. Regular THC lowers the number of available CB1 receptors and, in animals, weakens the signal from the ones left. The liver gets a little faster too, but the classic human study found that change too small to explain the lost effect, and blood levels can climb while the high falls.

Hunt and Jones tested the liver theory directly in 1980. Six people took 30mg of oral THC every four hours for 10 to 12 days, and an injected tracer dose before and after showed clearance rising from 605 to 977 ml a minute. The authors concluded that changes of that kind could not account for the tolerance they saw. (That regimen is 180mg a day, nearly two 100mg Colorado tins, about $31 at the Star Buds price on our 11-hydroxy piece. Keep it in mind whenever someone cites "the tolerance literature" at you.)

Gorelick's NIDA group ran the cleanest modern version on 13 male daily smokers living on a research unit: 20mg capsules around the clock, 40mg on day one, 100mg on days two to four, 120mg on days five and six. Plasma THC and 11-hydroxy-THC climbed significantly over the first five days. The morning dose lifted intoxication ratings on day two and did much less on days four and six. More drug in the blood, less on the rating scale, which is what tolerance looks like once you take the liver out of the argument.

Some of the fade is the product. Vandrey's 2015 JAMA letter found 17% of 75 edibles accurately labeled, a problem our dose chart guide follows into the 2026 Colorado numbers, and THC oxidizes on the shelf at a rate our shelf-life guide covers. A tin from a new batch is a new variable.

What happens to CB1 receptors with daily use?

Two things, on two clocks. Receptors first respond less well (desensitization), which rodents show within days, and then the brain removes some of them (downregulation). PET scans can see only the second: CB1 availability ran about 12 to 20 percent lower in chronic users than in controls across four human studies.

The receptor work starts in rodents. Breivogel's group gave rats daily THC for 3, 7, 14 and 21 days and watched both receptor numbers and receptor-driven G protein activation fall with time, at different speeds in different brain regions. Piscura's 2023 review collects the signaling steps since, and makes the point that matters here: desensitization has only been demonstrated in animals, because nobody has an imaging probe for it in a living person. The newest piece is Álvaro-Blázquez 2026 in PNAS, from the Guzmán and Marsicano labs. Activated receptors recruit an enzyme called NEDD4L, which tags four specific sites for disposal. Block the tagging in mice and receptor levels hold and behavioral tolerance disappears, with the acute drug response intact. That is a mouse result and it has reached no human. It is also the most specific answer yet to how the receptor gets removed.

The table shows who each of the four human PET studies scanned, what it found, and what happened next.

Read the D'Souza row twice. Those men averaged 59 joint equivalents in the previous month, about a fifth of what Hirvonen's men smoked, and their deficit came out about the same size. All three male-dominated studies looked for a link between the deficit and recent use and found none; Hirvonen's did track years of smoking. None of this proves a 5mg gummy does the same thing, and Hirvonen's authors say plainly that they don't know about moderate or occasional use. What it removes is the comfortable assumption that a small habit means a small adaptation.

The other thing worth reading is what these papers never claim. Hirvonen's discussion says the team did not objectively measure tolerance. Spindle's team tested the link in their 10 women and found receptor availability unrelated to the acute effects of the cannabis they smoked. D'Souza's correlation was with withdrawal on day two. So the human link between fewer receptors and a weaker gummy is an inference from the animal work, a strong one, and nobody has measured it in the same person on the same day.

How fast does tolerance build?

Within days at high oral doses. Intoxication ratings were falling by day four on around-the-clock capsules in Gorelick 2013, and appetite and sleep effects were gone after day eight at 40mg a day in Bedi 2010. At gummy doses nobody has timed it. The closest number is 0.5 percent less relief per session in app data.

The oral studies agree on the direction and share a limitation. Haney's 12 marijuana smokers built tolerance to the subjective effects across four-day blocks at 20mg and then 30mg, four times a day. Hart's 12 volunteers at 10 or 20mg four times a day rated a sample joint lower by day three. Every one of those doses sits in the very high tier on our dosing guide once you add up the day, and a 5mg gummy sits at the top of the low tier.

Stith 2025 is the largest dataset we found from people treating something. The authors (three of them employed by the app's owner) took the first ten sessions logged by 16,395 patients in the Releaf App, 120,691 symptom ratings in all, and found relief shrinking 0.5 percent a session, or 5 percent by session ten. Eating or drinking made up 15 percent of the observations; the rest was smoke and vape. The fade came mainly from people who called themselves experienced before they downloaded the app, sessions were about four days apart on average, and the authors warn against extrapolating to months or years. Counting calendar days instead of sessions gave the same answer, so this paper cannot tell you whether nights off slow anything.

The 2020 review from Ramaekers's lab ends where this section does: tolerance is likeliest in people using high doses continuously for a long time, and knowledge of the frequency, dose and duration needed to build, hold or lessen it is very limited.

How long does recovery take?

Receptor availability starts climbing within two days of stopping and was not fully back at four weeks in the two studies that rescanned. Nobody has measured how long a fixed dose takes to feel the way it used to. Every break length quoted online is an extrapolation from scans or a forum consensus, including the short ones.

The two-day figure is the one that travels, so here is what it is. D'Souza's group rescanned 10 men after two days and the 15 percent gap was no longer statistically significant. Availability was still lower than in controls, and the rise inside the same men only trended toward significance (p=0.09). At 28 days, in eight men, the authors conclude that it had still not reached control levels, while allowing that the sample was small or that some of the gap predated cannabis. Spindle's women were still 16 to 18 percent low at three days, and Ceccarini's users were still down within the first week. Hirvonen's men came back in most affected regions at about four weeks, and not in the hippocampus.

The mouse data puts a shape on that. After 15 days of escalating THC, Sim-Selley's group found receptor signaling back to normal in the striatum by day 3 and in the hippocampus only by day 14, with receptor numbers back at 7 and 14 days. The hippocampus came back last, and it is the same region that had not recovered in Hirvonen's men at the second scan.

What a patient wants to know is when 5mg works again, and that study does not exist. Jones and colleagues reported in 1981 that tolerance is largely lost rapidly under the right conditions, with no figure in the abstract. D'Souza's team raise the possibility of a dose adjustment for anyone returning to a cannabinoid medicine after more than a few weeks off, which is an inference from the scans. The two papers we found on tolerance breaks as such measured other things. Ansell's 170 young recreational users linked a break to higher hazardous-use scores six months later, with break length making no difference, in an observational design. Fontana's measured whether a guide helped people finish 21 days.

Meanwhile the forum record keeps its own books. In September 2020 a Growers Network member posting as katmwall, using cannabis daily for chronic pain and starting from edibles only, reported needing 250mg at minimum for a slight buzz, twenty-five 10mg gummies, and said a previous break had barely moved the number and made them miserable throughout. The replies offered weekly strain changes and a three-day detox. Neither has a trial behind it, and the guide reports what patients do in that gap as what it is: observed practice.

Is tolerance to oral THC different from tolerance to smoked THC?

Nobody has run the head-to-head, with matched people building tolerance by each route and then compared. Swallowed THC plainly produces tolerance on its own. The route changes the chemistry, with two active molecules circulating after an edible, and no study has tested whether that changes the tolerance.

The nearest thing is Newmeyer's 2017 NIDA study, which gave frequent and occasional smokers cannabis by joint, vaporizer and brownie (about 54mg of THC). Frequent smokers' subjective ratings rose above placebo after inhaling and did not after the brownie. Occasional smokers' ratings rose on all three routes, with no difference in the peak. The companion paper has the awkward number: after that brownie, frequent smokers' peak blood THC averaged 17.7 micrograms per liter against 8.2 for occasional ones, some of it likely residue from the habit. Twice the THC in the blood and no significant lift on the rating scale. So much for the folk rule that a heavy smoker gets flattened by an edible, at least at this dose and in this sample. The tolerance showed up more clearly on the oral route than on either inhaled one.

Stith's app data found less relief from eating and drinking than from vaping, until THC and dose entered the model and the difference stopped being significant. The paper reports no tolerance slope split by route. So the question of whether a liver-made metabolite at parity with THC changes how tolerance builds stays open.

Does tolerance from smoking carry over to edibles, and the other way?

Yes, at least partly, in both directions where anyone has looked. Frequent users felt less than occasional users from oral doses of 15 and 54mg in two studies, and high-dose capsules blunted the response to a joint in two of three. The untested cases are the ones that matter here: low-dose edible patients, and edible-only users handed flower.

Kirk and de Wit's 1999 study has the detail worth keeping. At 15mg of oral THC, frequent users (the bar was 100 lifetime uses) felt less than infrequent ones and were less sedated. At 7.5mg it reversed: frequent users registered the capsule and infrequent users felt nothing different from placebo. Carryover here depends on the dose, and experience seems to do something besides blunt.

Going the other way, Benowitz and Jones reported in 1975 that hospitalized volunteers on prolonged oral THC had nearly complete tolerance to both the racing heart and the psychological effects of smoked marijuana. Hart's volunteers rated a joint lower by day three at 10 or 20mg four times a day, though they kept choosing to smoke it. Vandrey's 13 daily smokers, after five days on up to 120mg a day, rated five puffs no differently than on placebo, which the authors found surprising, while the heart rate rise did shrink at 60 and 120mg. Those men smoked daily to begin with, so there was less left to blunt.

Won's 2026 Colorado study is the one piece of this that looks like the real world. Daily and occasional users, all with some regular edible use, brought their own licensed edible to the lab, nearly all of them gummies, and took their usual dose under observation. The daily group's median was 20mg against 7.5mg for the occasional group, the high tier against the moderate one on our dosing guide. That ratio is a measurement of what regular users actually reach for, with the caveat that frequency counted every route, so nobody here is confirmed edible-only.

Which effects build tolerance, and which do not?

Thinking and memory show the most tolerance, and the high and anxiety partial tolerance, according to the one systematic review of human studies we found. Heart rate and blood pressure effects outlast the high in the oral studies. Appetite and sleep results conflict from study to study, and only appetite has been followed at anything like a gummy dose.

The table sets each effect against what the studies found and the conditions they found it under.

Hirvonen's paper offers a neat explanation: memory impairment tolerates because the cortex loses receptors, while the high and motor effects hold because they live in regions that don't. Its authors call the idea simplistic themselves, and the oral data undercuts it, with the high fading in most capsule studies. Ramaekers 2016 is the main dissenter on impairment generally, and its authors note that users in some earlier studies used about twice as often, so tolerance to impairment can arrive, if it arrives, well beyond the daily use they studied. For anyone asking about the wheel, the guide takes that up, and Won's authors are explicit that their finding does not make daily use safer. If the heart is your concern, our cardiovascular piece covers the vascular data, and the short version from the table is that the pulse and the pressure do not fade with the high.

Does tolerance develop to the medical effect, or only to the high?

Sometimes to both, and the popular belief that only the high fades is not what the evidence shows. A 16-day oral study watched appetite and sleep benefits vanish while the high held. Long-term spray trials found pain and spasticity relief holding for years in patients who stayed, and app data shows a slow fade.

Jones, Benowitz and Herning said it in 1981, after studying 120 subjects: "Tolerance to therapeutic effects or target symptoms poses problems." They listed intraocular pressure and sleep among the effects that tolerate. Flach's nine California glaucoma patients, on capsules or inhaled marijuana, are the sharpest example: pressure fell at first in all nine, the drops did not last, and all nine stopped within one to nine months, which is why our glaucoma page reads as it does.

Then there is Bedi 2010, which should be better known. Seven HIV-positive marijuana smokers lived in a lab for two 16-day stays on 10mg of dronabinol four times a day. Calories and sleep improved for eight days and then stopped differing from placebo. Ratings of positive mood, the high and wanting another dose held for all 16. That is selective tolerance to the medical target, and the authors point out that nobody knows whether it happens in people who don't smoke. Haney 1999, at higher doses, found the exact reverse for appetite. Both are small and both are in smokers, and they cannot both be the rule.

The table lists the long-term studies, with the catch in each.

Notcutt's withdrawal trial is the best evidence on the page that a medical effect can survive years of use, because a placebo arm is what the open-label extensions lack. Its enrollment is also the catch for the whole genre: people whose relief faded had already left. In Wade's 2006 MS extension, 24 of the 58 withdrawals were for lack of efficacy, and the patients who remained held steady on the same dose. "No evidence of tolerance" in these papers means no evidence among the people still in the study. All of the spray trials also test one company's product through the mouth lining, at about 20mg of THC a day for the ones reporting sprays, and only Zajicek, Beal and Schimrigk tested THC that was swallowed. (Beal's 2.5mg once or twice a day is the only study in the table anywhere near a gummy dose, and its measure was hunger.)

The app data doesn't settle it either. Stith's fade was similar across pain, depression and anxiety. Cuttler's 2018 analysis of 11,953 Strainprint sessions, on the other hand, found no change in perceived efficacy for depression, anxiety or stress over time, while baseline depression worsened. Cuttler's 2020 paper on inhaled cannabis for headache and migraine found effectiveness shrinking as doses grew, which our migraine page treats as an overuse warning. Two apps, three answers, all self-selected users who chose to keep logging.

Is raising the dose the fix?

It buys something, costs something, and has a ceiling. In app data, higher doses gave more relief and more side effects. In a 360-patient cancer pain trial, the two lower dose ranges beat placebo on a secondary analysis, and only the highest compared unfavorably on adverse events. Doubling a very high oral dose barely moved the rating scales.

Stith's arithmetic is the one to remember. Offsetting the fade from one more session took about 6 percent more dose, and patients actually added 0.6 percent. Nobody tested whether that held past session ten and the authors decline to claim it. Run it anyway and a 5mg dose passes 10mg around session 12 and 20mg around session 24, and the shelf has a product waiting at every stop. What the climb costs is a table on the guide.

The trial data is less encouraging than the app. Portenoy's nabiximols study randomized 360 opioid-treated cancer patients to placebo or 1 to 4, 6 to 10, or 11 to 16 sprays a day for five weeks. The primary 30 percent responder analysis failed for all doses together (p=0.59). A secondary analysis favored the low and medium groups (p=0.008 and p=0.039). Side effects climbed with dose, and the high-dose group fared worse than placebo on them. That was a dose-finding trial and it did not test tolerance. It does show that more THC is not reliably more relief, which the tolerance argument tends to assume.

At the top of the range the curve goes flat. Gorelick's men climbed from 40 to 120mg a day with ratings falling. Schlienz's 13 daily users went from 120mg a day to as much as 240mg a day of dronabinol for 12 days, and few subjective ratings differed between the two doses; the authors reach for dronabinol's partial agonism to explain the lack of an orderly dose response, the same pharmacology our endocannabinoid piece walks through. MacCallum and Russo's dosing review recommends generally keeping THC at 30mg a day or less, partly to avoid tolerance, which is expert judgment and no measured threshold.

None of this makes raising the dose unreasonable for a patient with a real symptom and a physician in the room; D'Souza's and Bedi's authors both say tolerant patients can need more. The more interesting finding sits in the extension trials. Johnson's and Rog's patients titrated their own dose and did not climb, Hoggart's did not seek more, and Serpell's averaged 7.3 sprays a day, about 20mg of THC, against a permitted maximum of 130mg. Patients with a symptom to track settled. Survivorship explains some of that, as above, and it still means the climb is no law of pharmacology.

What has nobody measured about edible tolerance?

Almost everything specific to the way patients take edibles. Beyond one appetite study, we found no human data on a low oral dose over months, no receptor scan in anyone recruited for eating cannabis, and no measurement of when an effect returns after a break. These are the questions patients ask first.

• What a low nightly oral dose does to CB1 availability. Hirvonen's and Spindle's authors both list lighter users as future work.
• How long a fixed dose takes to regain its original effect after a pause.
• Whether nights off, weekends off or every-other-day schedules slow tolerance at all.
• Whether stepping down to a microdose lets receptors recover. Every recovery scan above used abstinence.
• Whether edible-only users carry tolerance into flower, and whether the slope differs by route in the same people.
• What happens in women beyond a single scan at three days, given Spindle's finding that women without cannabis use had higher availability than men to begin with.

The next tolerance experiment we can find on ClinicalTrials.gov, NCT07471347 at King's College London, plans around 30 heavy users and a vaporizer. The gummy patient remains, scientifically speaking, an assumption.