Research & Medicine

When a member asks you tomorrow whether there is anything to the entourage effect, the honest answer until recently was a shrug. Four original studies looked at the same question, and two of them found nothing while two found something. Since 2026 the contradiction can be resolved cleanly, and the resolution teaches you more than the result itself. The question at stake Not whether terpenes do anything at all. That is uncontested, and it runs through TRP channels, GABA, adenosine and other targets. The narrow question is this: do cannabis terpenes activate CB1 and CB2 directly, the receptors THC also acts on? Only if that holds is the popular picture of the terpene turning the same dial pharmacologically correct. Four papers address it head-on. Two from the negative camp, two from the positive camp. The negative findings of 2019 and 2020 Santiago and colleagues used AtT-20 cells expressing human CB1 or CB2. The readout was the change in membrane potential, that is, a single endpoint. The terpenes tested showed no effect, neither on their own nor together with THC (Santiago et al., 2019). Finlay and colleagues cast a wider net: five terpenes, radioligand binding with labelled CP55,940 plus a cAMP assay, up to a maximum of 10 micromolar (µM). The conclusion was worded unambiguously: “no data were produced to support the hypothesis that any of the five terpenes tested have direct interactions with CB1 or CB2”. Only beta-caryophyllene showed weak CB2 binding. What is remarkable is the caveat the authors put in the same paper: “this study cannot rule out the existence of an entourage effect for terpenoids” (Finlay et al., 2020). That caveat was almost entirely lost in how the work was received. Between 2019 and 2021, much of the field treated the entourage effect as settled. The positive finding of 2021 LaVigne and colleagues tested four terpenes in CHO cells with cAMP and ERK readouts and added the mouse tetrad test, that is, a behavioural model. Alpha-humulene, geraniol, linalool and beta-pinene activated CB1, some of them with additive effects alongside the synthetic agonist WIN55,212. The authors discussed the discrepancy with Finlay openly and named three points: lower concentrations, an assay with only one output, no in vivo model (LaVigne et al., 2021). The weak point in LaVigne is the concentration. Up to 500 µM in vitro is far beyond what is physiologically plausible. What Raz does differently The papers by Raz and colleagues are the methodologically most solid work in this set, for four reasons. Breadth. 16 terpenes plus mixtures, tested at both receptors with matched concentration ranges. None of the other three papers tests anything like that breadth. Sensitivity of the readout system. Xenopus oocytes with GIRK current recording. That picks up subtle partial agonists far more reliably than membrane potential or cAMP. Antagonist verification. Rimonabant for CB1R and SR-144,528 for CB2R abolish the effects completely. Receptor specificity is therefore not inferred, it is shown. That is what makes the finding hard to argue away on methodological grounds. Note that the same series carries a declared conflict of interest, which the section on caution below spells out. Realistic concentrations. The group stayed at the water-solubility limit rather than pushing beyond it. The core findings: dose dependent activation at both receptors for most terpenes, with a maximum response at 10 to 60 percent of the THC response, that is, partial agonists. In the same assay, the EC50 values came out comparable to THC or lower (Raz et al., 2026). Important for context: this is a series of three publications by the same group, not three independent confirmations. 2023 brought the co-application with THC at the CB1 receptor, 2026 the terpenes on their own at both receptors, and also in 2026 the systematic analysis of the interaction with THC. Same working group, same measurement method, same funding. If you count the three papers as three pieces of evidence, you count one too many. The discrepancy has a methodological explanation Three factors carry the resolution. First, concentration. Finlay capped at 10 µM. Several terpenes only become measurably active above that. Second, assay sensitivity. A membrane potential readout or a cAMP assay simply sees a partial agonist with 20 percent maximum response less well than a GIRK current recording does. Third, and this is the most interesting point, competition within mixtures. Terpenes given together without THC do less than the sum of their parts, because they compete for the same binding site. Finlay also worked with mixtures in which no single terpene made up more than half. That makes Finlay’s null result for mixtures pharmacologically consistent with the positive result for single substances. These were never contradictory answers, they were different questions. Together with THC it looks different This is where the decisive refinement sits, the one added in 2026. The competition applies to terpenes among themselves. Towards THC, many terpenes behave not as competitors but as amplifiers. The group co-applied terpenes with THC at a ratio of one to ten, so with the terpene as the minority component, the way it occurs in the plant. An isobolographic analysis, which compares the observed effect with the effect you would predict from simple addition, produced genuine enhancement for a subset of the terpenes: borneol, limonene, sabinene, terpineol, alpha-pinene and ocimene at the CB1 receptor, beta-caryophyllene and linalool at the CB2 receptor. The authors’ reasoning holds up. If terpenes were pure competitors at the same docking site, a weak compound alongside a strong one would have to lower the overall effect. What was observed was mostly the opposite. From that they conclude that a second mechanism is probably also in play, an effect via a different site on the receptor. Three limitations come with that, and all of them are in the same paper. One terpene ran the other way. Eucalyptol abolished the THC effect at the CB2 receptor. Humulene and nerolidol showed a trend in the same direction. Not one of these examples appears in the paper’s abstract. Of eight mixtures,

Few arguments turn up in the cannabis debate as reliably as this one: “Cannabis causes psychosis, so obviously the stuff is harmful.” The sentence sounds final. It isn’t. The research on cannabis and psychosis paints a far more precise picture, one that neither plays down the risk nor lumps every kind of use together. We went through the current evidence and worked out what it actually says. Cannabis and psychosis: a link nobody seriously disputes Let’s start with what holds up. There is a connection between cannabis use and psychosis, and it is well documented. People who use regularly have, on average, a higher risk of developing a psychotic disorder or schizophrenia than people who never use. That is not a fringe position; it runs through dozens of studies and reviews. But the slogan “cannabis causes psychosis” glosses over the crucial point: how much, how potent and who account for almost the whole difference. That is where it gets interesting, and that is exactly where the blanket harm argument falls apart. How high the risk actually is Before talking about risk factors, it pays to look at the baseline, because without it, talk of “five times the risk” means nothing. The lifetime risk of developing schizophrenia in the general population sits at roughly 0.6 to 1%. Counting all psychotic disorders together (which includes delusional, schizoaffective and substance-induced forms) it rises to around 3%. Isolated, temporary psychosis-like experiences, without any illness being present, are something an estimated 5 to 7% of people have had at some point. That single percentage point is the anchor for everything that follows. When a study reports a “doubled” or “fivefold” risk, it almost always means relative risk, measured against this baseline. A fivefold risk sounds like 50%, but starting from 1% it works out to roughly 5%. Keeping that difference between relative and absolute risk in mind gets you halfway to an honest debate. The risk map: where cannabis actually sits With that baseline in hand, you can put the whole list of known risk factors in order, and you see immediately that cannabis is not at the top of it. The overview below pulls together what meta-analyses report in terms of relative risk increase, along with a rough absolute reference point. Factor Risk (relative) Absolute lifetime risk (approx.) General population (baseline) n/a ~1% (schizophrenia), ~3% (all psychoses) Identical twin affected n/a ~40–50% Both parents affected n/a ~40% Parent or sibling affected ~6x ~6–10% Racism and discrimination ~4x ~4% Childhood trauma (multiple, 5+ adversities) ~3x (up to ~6x) ~3–6% Urban birth or urban upbringing ~2–3x ~2–3% Cannabis: daily and high-potency ~5x ~3–5% Cannabis: frequent use (weekly or more) ~1.3–2x ~2% Migration or minority status ~2x ~2% The absolute figures are rough reference points based on a baseline risk of around 1%. The factors do not simply add up; some of them interact or overlap. Relative risks for discrimination and childhood trauma follow the 2021 Frontiers in Psychiatry umbrella review and Varese et al. (2012). Two things stand out. First, predisposition dominates everything else. Heritability estimates run from 60 to 80%, with a Danish twin cohort landing at around 79%, and an affected identical twin lifts the risk into an order of magnitude that no single environmental condition reaches (Hilker et al., 2018). Second, cannabis lands in the middle of the environmental factors: the daily high-potency extreme case is on a par with discrimination and childhood trauma, while ordinary use sits below that. Then there is the population-level view, which flips the public weighting. For society, what counts is not only how strongly a factor acts but how widespread it is. Adverse childhood experiences are common, which is why an estimated third of all psychosis cases can be attributed to them statistically, more than to any other single factor (Varese et al., 2012). That is precisely why social adversity, trauma and discrimination often sit at the top of the list in prevention research, even though genetics carries the most weight in any individual case. On this map cannabis is one factor among several: real, moderate in size, and unlike genes, background or childhood, one of the few you can influence directly through your own behaviour. Where the risk starts to rise In 2022, a Canadian meta-analysis pulled together the evidence on how the frequency of use affects psychosis risk, across ten studies covering around 7,400 people (Robinson et al., 2022). The result is remarkably clear and rarely quoted, because it is hard to turn into a scandal. People who use only occasionally, yearly or monthly, show no statistically significant increase in risk. Only from weekly use onwards does it climb measurably, by around a third. At daily use it then sits at roughly double that of non-users. So there is a threshold below which the data do not show an elevated risk. That is the first nuance the slogan loses: the relevant question is not “use, yes or no” but “how often”. Potency is the real crux The second nuance matters at least as much, and it explains why the debate looks different today than it did thirty years ago. That THC potency has risen steadily is well documented (Hoch et al., 2024). The figures usually cited: in the 1980s, THC content in cannabis was typically 2 to 4%, while today flower in legal markets often reaches 20% and more, and concentrates 80 to 90%. It is no longer the same substance. A British longitudinal study tackled exactly this question (Hines et al., 2024). It followed adolescents over years and asked whether high-potency cannabis in adolescence is linked to psychotic experiences later on. The result: those who used high-potency cannabis as teenagers had roughly double the risk of new-onset psychotic experiences. For “any” cannabis use, by contrast, the association was weak and statistically uncertain. That is the decisive finding. The effect hangs on potency, not on use as such. An even more recent analysis from the European research network EU-GEI marks the upper end of