A Cannabinoid Agonist Raised Insulin Sensitivity and a CB1 Blocker Lowered It: Inside an NIH Clamp Study

Using sequential hyperglycemic and euglycemic-hyperinsulinemic clamps, the reference method for separating insulin secretion from insulin action in humans, the investigators found that nabilone significantly increased insulin sensitivity compared with placebo, while 45 mg of the CB1 antagonist CP-945,598 significantly decreased it.

Insulin secretion during the hyperglycemic clamp was unchanged by either compound. The endocannabinoid system in this experiment modulated how tissues respond to insulin, not how much insulin the pancreas releases.

Most of what is claimed about cannabis and metabolism rests on surveys and observational cohorts, where confounding is impossible to exclude. This study asks a narrower question with a far more precise instrument and gets a clean answer about receptor pharmacology in healthy physiology.

It also speaks to a specific clinical history. CB1 blockade was pursued as an obesity strategy, and the class collapsed over psychiatric harm. These data suggest that CB1 blockade carries a metabolic cost as well, which reframes what was lost and gained in that episode.

The study, Acute metabolic effects of cannabinoid receptor modulators during sequential hyperglycemic, euglycemic-hyperinsulinemic clamps in healthy individuals, was conducted by Chee W. Chia, Eric Y. Tang, Roy G. Cutler, Susan S. You, Qinghua Chen, Adeline Y. M. Choo, Denise L. Melvin, Christopher E. Ramsden, Dimitrios Kapogiannis, and Josephine M. Egan at the National Institute on Aging Intramural Research Program in Baltimore. It was published in the American Journal of Physiology: Endocrinology and Metabolism, volume 330, issue 5, pages E659 to E674.

Twenty-one healthy men each completed four visits separated by at least six weeks. At each visit, in randomized and blinded order, they received one of four treatments: nabilone 2 mg, a nonspecific cannabinoid receptor agonist; CP-945,598 15 mg, a selective CB1 receptor antagonist at low dose; CP-945,598 45 mg, the same antagonist at high dose; or placebo.

After a fast of more than ten hours, each participant underwent a sequential hyperglycemic clamp followed by a euglycemic-hyperinsulinemic clamp, with a deuterated glucose disposal test to quantify glucose utilization and production. Serum endocannabinoids and a panel of N-acylethanolamines were measured across the procedure.

First, insulin secretion was unaffected. Neither nabilone nor either dose of CP-945,598 changed glucose-induced insulin secretion during the hyperglycemic clamp compared with placebo. The beta cell response was intact and unmoved.

Second, insulin sensitivity moved in opposite directions depending on the compound. Nabilone significantly increased insulin sensitivity relative to placebo during the euglycemic-hyperinsulinemic clamp. The 45 mg dose of CP-945,598 significantly decreased it. The authors describe the antagonist effect as dose-dependent.

Third, the relationship runs both ways. During the clamps, elevated plasma insulin suppressed circulating non-esterified fatty acids, anandamide, and several other N-acylethanolamines in a saturable, insulin-dependent manner. 2-arachidonoylglycerol, the other major endocannabinoid, was not affected by insulin.

The reason this design carries more weight than the usual cannabis metabolism study is measurement. Fasting glucose, HOMA-IR, and questionnaire-based exposure estimates all carry substantial noise. A euglycemic-hyperinsulinemic clamp holds plasma glucose at a fixed level while insulin is infused, and the glucose infusion rate required to maintain that level is a direct readout of whole-body insulin sensitivity.

Pairing it with a preceding hyperglycemic clamp separates two things that often get conflated in metabolic reporting: how much insulin the pancreas makes in response to glucose, and how well tissues respond to the insulin that arrives. This study found a cannabinoid effect on the second and none on the first.

A within-subject crossover design adds further precision, because each participant serves as his own control across all four conditions. With 21 participants and four visits each, the comparison is tighter than the raw sample size suggests.

Rimonabant, a CB1 inverse agonist, was approved in Europe in 2006 as an anti-obesity agent and withdrawn in 2008 after psychiatric adverse effects including depression, anxiety, and suicidality. The psychiatric signal was the reason it was pulled, and it was a sufficient reason.

These clamp data add a second consideration that was harder to see at the time. If blocking CB1 reduces insulin sensitivity in healthy physiology, then a CB1 antagonist prescribed for weight loss would be working against insulin action at the same time it was reducing body weight. This study does not establish that this happened in rimonabant trials, and it was not designed to. It establishes the underlying receptor-level effect in healthy men.

The broader lesson is that the endocannabinoid system is not a dial that can be turned down in one tissue without consequences elsewhere. Interventions at CB1 reach appetite, mood, and insulin action together.

Nabilone is a synthetic cannabinoid receptor agonist approved for chemotherapy-induced nausea and vomiting. It is not cannabis. It was given orally at a fixed 2 mg dose to fasted healthy men under laboratory conditions, with no cannabidiol, no terpenes, and no variability in route or product.

Cannabis delivers a variable mixture of compounds by a variable route to a population that includes people with obesity, insulin resistance, and diabetes, in the fed state, on other medications. None of those conditions were tested here.

The honest translation is narrow and still worth having. Cannabinoid receptor signaling participates in the regulation of insulin action in humans. That is a mechanistic claim, it is now supported by a controlled measurement rather than an inference, and it is a reasonable basis for designing trials in patients who actually have metabolic disease.

For the mechanistic question it asks, this is strong evidence. A randomized crossover design, a reference-standard measurement of insulin action, an active agonist, two doses of an antagonist, and a placebo arm together make a coherent pharmacological argument rather than a single observation.

For any clinical question, it is preliminary. Twenty-one healthy, non-obese men under fasting laboratory conditions is not a patient population, and acute single-dose administration says nothing about repeated exposure. Treat this as physiology that motivates trials, not as evidence supporting a therapy.

The sample was 21 men. No women were studied, which matters given documented sex differences in both endocannabinoid signaling and insulin sensitivity. No participant had obesity or insulin resistance, so the study cannot say whether the same receptor manipulations produce the same effects in a dysregulated system, which is precisely the clinical situation of interest.

The comparison also depends on CP-945,598, an investigational compound that was never marketed. Extrapolating from it to other CB1 antagonists requires assuming comparable selectivity and receptor occupancy, and the single-dose acute design leaves open whether these effects persist, attenuate, or reverse with repeated administration.

This study does not show that cannabis improves insulin sensitivity, that cannabis prevents or treats type 2 diabetes, or that any cannabis product should be used for a metabolic indication. Nabilone is a single synthetic agonist given orally at a fixed dose, and its behavior does not transfer to inhaled flower, edibles, or CBD-dominant preparations.

It also does not demonstrate a clinical benefit of any size. A statistically significant change in clamp-derived insulin sensitivity in healthy men is a physiological signal. Whether it translates into a change in A1c, weight, or cardiovascular outcome in patients is untested.

Population data have repeatedly shown lower diabetes prevalence and lower body mass index among cannabis users, a pattern that has persisted across datasets without a satisfying explanation. Observational work can keep documenting that pattern indefinitely without resolving it. This study supplies part of the missing mechanistic link in the right species, under controlled conditions.

It also sharpens a caution. A separate line of controlled human work has found that high-THC cannabis can raise inflammatory markers without improving insulin sensitivity. A clean receptor-level effect in fasted healthy men and a messier picture with real products in real users are not contradictory. They describe different experiments, and clinicians need both in view.

What I appreciate about this paper is the restraint of the question. It does not ask whether cannabis is good for metabolism. It asks whether cannabinoid receptor signaling participates in insulin action, and it uses a method precise enough to answer that. The result is that it does, in both directions.

The finding that has stayed with me is the one about insulin suppressing anandamide. That means hyperinsulinemia, which is the defining state of early metabolic syndrome, is also a state of reduced circulating anandamide. Whether that is protective, compensatory, or part of the problem is unknown, and it is a genuinely interesting question that nobody was asking a decade ago.

None of this changes what I do on Monday morning. My patients with metabolic disease are not 21 healthy fasting men, and nabilone at 2 mg is not what anyone is buying. What it does is make me more confident that when a patient with insulin resistance asks whether cannabis is doing something to their metabolism, the answer is that it probably is, and that the direction depends on details we can still only partly specify.

In healthy men under controlled clamp conditions, a cannabinoid receptor agonist increased insulin sensitivity and a high dose of a CB1 antagonist decreased it, with no effect from either on insulin secretion. The endocannabinoid system regulates insulin action in humans. That is a mechanistic finding, not a treatment recommendation, and the gap between the two is large.

Carry forward the receptor pharmacology and the bidirectional signaling. Leave behind any sentence that begins with the word cannabis, because cannabis was not studied here. The compound was nabilone, the population was 21 healthy fasting men, and the outcome was a laboratory measure of insulin action over a few hours.