CBD And Caffeine: Can I Mix Them?

Eight notes worth writing down

1. A can of cold brew prints its caffeine in milligrams. A bottle of CBD oil prints its cannabidiol in milligrams. Put both on the same kitchen counter and there is no third figure to read off, because no controlled human study has measured the subjective effects of caffeine and cannabidiol taken together.
2. Caffeine is broken down in the liver. The enzyme leading that work is cytochrome P450 1A2, one member of the cytochrome P450 family. An organ and a family of enzymes with names: that is the overlap, stated plainly.
3. Jiang and colleagues, in 2011, worked with human liver microsomes, which are fragments of liver tissue kept in glassware, and identified the specific cytochrome P450 enzymes responsible for metabolising cannabidiol, together with the family members involved [1]. Tissue and instruments, not volunteers with a cup in hand.
4. So the two molecules meet the same enzyme family in the same organ. That is an address, not a measurement. A shared route tells you the two questions belong on one page. It does not tell you what happens in one person on one morning.
5. Millar and colleagues, in their 2018 systematic review of cannabidiol in humans, found that how much reaches circulation, how long it persists and how widely the published figures differ all depend on processing in the liver and on the route of administration [2].
6. Type the question into a search box and the results answer in confident, complete sentences. Behind most of them sits no trial to cite: no arms, no blinding, no agreed outcome measures. Confidence of tone and weight of evidence are separate things, and here they have come apart.
7. Colloca and Barsky, in their 2020 clinical review, describe placebo and nocebo effects as measurable and mechanistically investigated, able to alter what people report after a dose, including inside blinded trials [3]. One good reason a single morning's impression, however vivid, is not data.
8. Since 2014 we have written these pages the same way: name what has been measured, name what has not, and leave the reader holding a label rather than a promise. On this particular pairing, the unmeasured column is the longer one.

Caffeine's route through the liver

Caffeine does not pass through untouched. It gets broken down, and the site of that work is the liver. The enzyme leading it carries a name that reads like a filing reference: cytochrome P450 1A2.

Worth unpacking, because the code carries the whole argument. Cytochrome P450 is a family, not a single protein. The 1 marks the family group, the A the subfamily, the final 2 the individual enzyme. Plenty of relatives sit under the same surname.

That distinction is where most online answers go loose. Saying caffeine and cannabidiol both use P450 is true at the level of the surname. It is the same sort of statement as saying two people are both Swiss. Useful for knowing where to look. Silent on everything else: how much of each compound is processed, how fast, whether one shifts the other's pace, and by how much.

The word chief is doing quiet work here too. Chief means principal, not sole. Naming one enzyme as the main route for caffeine is a statement about which door most of the traffic uses, not a claim that the others are shut.

The other half of the caffeine picture is ordinary and easy to skip past. Coffee arrives in a known quantity, by a known route, swallowed as a liquid. Route turns out to matter on the cannabidiol side as well, where Millar's 2018 review ties the published human figures to how the compound was taken [2].

So the honest version of the overlap is short: one organ, one enzyme family, named members [1]. Caffeine's principal enzyme is identified by number. Cannabidiol's enzymes were identified in liver tissue examined outside the body [1]. Past that point, the record stops handing out numbers for the pair.

People expect a mechanism to come with a magnitude attached. It usually does not. A map showing two roads meeting at a junction tells you nothing about the traffic there on a Tuesday morning, and none of the cited work put a cup of coffee alongside a pipette to find out.

Microsomes, routes and wide ranges

Two publications carry most of the weight on the cannabidiol side, and they answer different questions.

Jiang and colleagues, in 2011, used human liver microsomes [1]. Microsomes are what you get when liver tissue is broken up and spun down: the enzyme machinery, isolated, working in a controlled setting with no person attached to it. The output of that study was a set of names. Which cytochrome P450 enzymes handle cannabidiol, and which members of that family take part [1]. Mapped, in other words, rather than measured in the way a coffee drinker would like.

Millar and colleagues, in 2018, did something else entirely. A systematic review means gathering the published human studies against fixed criteria and reading them side by side, rather than running a new experiment [2]. Their picture of cannabidiol in humans holds three things together: how much of it turns up in circulation, how long it persists there, and how far apart the published numbers sit. All three track back to processing in the liver and to the route of administration [2].

That last point is the practical one. The spread across the published figures is wide, and the variables behind the spread are the format and the route [2]. An oil held under the tongue, a capsule swallowed with breakfast, something inhaled: the same molecule, different numbers on the page.

Which is why the arithmetic people want here cannot be done. If the human figures for cannabidiol on its own already span a wide range depending on how it was taken [2], and the caffeine side has never been run against it under controlled conditions, then adding the two is not a calculation. It is a guess with citations stapled to the back.

There is a narrower question hiding inside the popular one, and that narrower question does have an answer. Not what a coffee plus a few drops feels like. Rather: where in the body do these two molecules pass through the same machinery. The answer is the liver, and the cytochrome P450 family, with specific enzymes named for cannabidiol in 2011 by Jiang and colleagues [1].

What a real trial would need

1. Four arms, not two. Caffeine on its own, cannabidiol on its own, both together, and neither. That fourth arm looks redundant until you remember it is the only one that shows what a plain morning looks like in the same group of people, on the same scoring sheet, at the same hour.
2. Blinded conditions. Participants and assessors kept unaware of which arm is which, because Colloca and Barsky's 2020 review documents reported experience shifting after a dose even inside blinded trials [3]. Take the blinding away and you are no longer measuring the compounds. You are measuring expectation.
3. Outcomes agreed before the first dose is given, not chosen afterwards from whatever moved. Which things get recorded, in what words, on what scale. A trial that decides its questions at the end can find something in almost any dataset, and that finding will not replicate.
4. Fixed intervals. Readings taken at set points after dosing rather than whenever someone thinks of it, since the 2018 review by Millar and colleagues shows persistence in circulation is itself one of the moving parts, tied to processing in the liver and to route [2].
5. One format, one route, held constant across the arms. The published human figures for cannabidiol vary widely with format and route [2], so a study mixing an oil, a capsule and an inhaled product would fold that variation into its own results and never be able to separate it out again.
6. A publication at the end that anyone can look up and check. For caffeine and cannabidiol taken together, that publication does not currently exist. What exists is a mapped metabolic route [1], a human pharmacokinetic picture with wide margins [2], and a well described reason to be careful with self-report [3].

What the bottle can tell you

None of this makes the question pointless. It moves it. From what the mix will do, to what can actually be checked, and the checkable part is printed on the packaging.

A bottle states how many milligrams of cannabidiol it contains and, for an oil, how much sits in a single drop. A batch analysis states what an independent lab measured in that specific batch, not in the product line generally. The format tells you the route, and route is precisely the variable the 2018 review ties the published human figures to [2]. Those are numbers with a source behind them.

Time of day is yours to note down. So is quantity, and whether coffee was in the picture. Keep that as a record of what you did rather than evidence of what followed, and Colloca and Barsky's 2020 review is a solid reason to keep those two categories apart [3].

If you take prescription medication, the enzyme question stops being an academic one and becomes a conversation for your doctor or pharmacist, who can look at the specific medicines on your list. The liver and the cytochrome P450 family are where these molecules meet on paper [1], and that is exactly the kind of detail a clinician is equipped to read.

Named, unnamed, and the label

Two columns. In the first: the liver as the site of caffeine metabolism and cytochrome P450 1A2 as its chief enzyme; the cytochrome P450 enzymes identified for cannabidiol in human liver microsomes [1]; the human pharmacokinetic picture as it stood in 2018, route and format included [2]; placebo and nocebo as measurable, investigated phenomena [3]. In the second: any controlled human study of the two together, any figure for the pair, any trial with four arms and fixed outcomes.

We have kept those columns apart since 2014. It reads less dramatically than a headline. It also stays true when the evidence changes, and it leaves you holding a label you can verify instead of a promise nobody has tested.