How to Evaluate CBD Oil Strength and Quality: A Physician’s Guide

In a study of 105 hemp topical products purchased online and at national retailers, gas chromatography and mass spectrometry found that only 24 percent of the products listing a CBD amount were within ten percent of that figure. Measurable THC was present in 35 percent of products, including four that were labeled THC free.

In a separate analysis of 97 hemp haircare, cosmetic, and food products, ten contained no detectable cannabidiol at all despite label claims. Label accuracy did not differ between products that advertised third-party testing and those that did not.

Dose only means something when the amount is known. A patient carefully titrating from ten milligrams to twenty is performing a meaningless exercise if the bottle contains half or double what it claims, and that patient will reasonably conclude that cannabidiol does not work for them.

The stakes rise with dose. Cannabidiol interacts with liver enzyme pathways shared by a substantial number of prescription medications, and the likelihood of a clinically relevant interaction increases as dose increases. Chasing the highest milligram figure on the shelf is where an unverified product becomes a clinical problem rather than a wasted purchase.

Patients ask some version of which cannabidiol oil is strongest nearly every week. It is the wrong first question. The right one is whether the bottle in front of them contains anything close to what the label says, and the published testing on that point is not reassuring.

A case series published in JAMA Network Open in 2022 purchased 105 hemp topical products, 45 from retail stores and 60 online, and measured cannabinoid content by gas chromatography and mass spectrometry. Of the 89 products that listed a total CBD amount, 18 percent contained more than ten percent less cannabidiol than advertised, 58 percent contained more than ten percent more, and 24 percent were accurate within that margin. Measurable THC appeared in 35 percent of the products, all below the 0.3 percent hemp threshold, and four of those were labeled THC free.

A 2024 analysis in Cannabis and Cannabinoid Research examined 97 hemp haircare, cosmetic, and food or drink products. Of the 71 with a specific labeled amount, 49 percent were underlabeled and 38 percent overlabeled, leaving about 13 percent accurate. Ten products contained no detectable cannabidiol. Notably, accuracy was no better among products claiming external laboratory testing.

The broader hemp market shows the same pattern. An analysis of 53 hemp-derived delta-9 THC products found that 66 percent differed from their stated dose by more than ten percent, and that 71 percent of the laboratory reports supplied to customers did not test for impurities at all.

Once a product’s content is verified, milligrams of cannabidiol per milliliter, roughly one dropper, is a reasonable way to think about intensity. The tiers below are practical categories used in clinical conversation rather than thresholds established by trial evidence, and they should be read that way.

Entry level, around 5 to 10 mg per mL, suits someone who has not used a cannabinoid product before and wants to observe their own response before adjusting anything.

Moderate strength, around 25 to 50 mg per mL, is where many patients managing everyday discomfort, mild sleep difficulty, or low-level anxiety settle after titrating upward from an entry dose.

Higher strength, around 60 to 100 mg per mL, is the range where more substantial symptom management is attempted, and it is the point at which I want a patient working with a clinician or a knowledgeable pharmacist rather than adjusting alone, chiefly because of drug interaction risk.

Very high strength, above roughly 150 mg per mL, is a tool for specific supervised situations rather than a default upgrade. A product at that concentration should be approached by taking a fraction of a dropper, not a full one.

Cannabidiol does not follow a simple linear dose-response curve. The clearest human demonstration comes from a double-blind trial in which 57 healthy men received 150 mg, 300 mg, or 600 mg of oral cannabidiol or placebo before a simulated public speaking test. Only the 300 mg dose significantly reduced anxiety. Neither 150 mg nor 600 mg differed from placebo.

That is an inverted U-shaped curve, and it means a middle dose can outperform a higher one for a given effect. It is one of the strongest arguments against treating milligram count as a proxy for benefit.

Exposure also varies substantially between people and between routes. A systematic review of human cannabidiol pharmacokinetics found reported half-lives ranging from 1.4 to 10.9 hours after oromucosal spray, two to five days after chronic oral dosing, and about 31 hours after smoking. Absolute oral bioavailability in humans has not been established. Peak concentration rises when cannabidiol is taken with food and in lipid-based formulations. Two people taking the same labeled dose can end up with meaningfully different exposure.

The practical consequence is that the same number on two different bottles, taken by two different people at two different times of day, describes two different pharmacologic events.

Cannabidiol is metabolized through cytochrome P450 pathways that it also inhibits, and those pathways are shared with a large number of prescription medications. A 2022 review in Expert Review of Clinical Pharmacology assembled the available evidence into guidance on the dose ranges at which clinically significant interactions become likely and which drug classes are most affected.

The classes that come up most often in practice are antiepileptics, benzodiazepines and other sedatives, certain anticoagulants, and immunosuppressants with narrow therapeutic windows. The review’s own conclusion is worth quoting in spirit rather than overselling: individual prediction is limited by gaps in dose-response data and by variation in comorbidity and concurrent medication, so clinician awareness plus clinical and therapeutic drug monitoring remains the best available approach.

Translated into a rule a patient can use: if you take any prescription medication regularly, review your full list with a physician or pharmacist before moving into the higher strength tiers, not after something feels wrong.

A Certificate of Analysis is the only practical way to verify a strength claim. Four checks separate a useful one from decoration.

Confirm the testing date is recent, and confirm that the batch or lot number on the certificate matches the number on the bottle exactly. A certificate for a different batch tells you nothing about the product in your hand.

Confirm the testing was performed by an independent accredited laboratory rather than the manufacturer’s own. Note that in the published product analyses, accuracy was no better among products that advertised external testing, so the presence of a testing claim on the package is not itself evidence of anything.

Confirm the panel covers more than cannabinoid content. A trustworthy certificate reports heavy metals, pesticide residue, microbial contamination, and, for concentrates, residual solvents. In the hemp delta-9 analysis cited above, 71 percent of the laboratory reports provided to customers omitted impurity testing entirely.

If a company cannot produce a current, batch-matched certificate on request, that alone is reason enough to look elsewhere, regardless of what the label promises.

The label accuracy evidence is strong for what it measures. These are laboratory analyses using gas chromatography and mass spectrometry on products purchased the way consumers purchase them, reported with percentage deviations rather than impressions. For the question of whether commercial cannabidiol labels can be trusted, the answer is well established.

The dosing evidence is weaker. The inverted U-shaped finding comes from a single trial of 57 healthy men using one outcome in one experimental setting, and it should be read as a demonstration that the curve is non-linear rather than as a map of optimal doses for any clinical condition. The pharmacokinetic review documents wide variation and, importantly, documents gaps.

Product testing studies sample a market at a point in time and in specific places. The topical and cosmetic analyses drew from the Baltimore area and from online retailers, with purchases in 2020 and 2022. The hemp market moves quickly, and the specific numbers should be treated as characteristic of an unregulated category rather than as current figures for any particular shelf.

The strength tiers described above are clinical convention. No trial has established that 25 to 50 mg per mL is the right range for anxiety or that 60 to 100 mg per mL is where symptom control begins. They organize a conversation; they do not carry evidentiary weight.

The dose-response trial that showed a middle dose outperforming a higher one used healthy male volunteers and an acute stress paradigm. Extending that finding to chronic dosing for a clinical condition is an inference, not a result.

None of this evidence shows what dose of cannabidiol treats any particular condition. The label accuracy studies say nothing about efficacy, and the dose-response trial addressed acute experimental anxiety in healthy volunteers.

It does not show that full-spectrum products outperform isolates. The entourage hypothesis, which holds that minor cannabinoids and terpenes work together with cannabidiol, remains a hypothesis with limited controlled human support.

It also does not establish long-term safety at higher doses. Most human cannabidiol safety data at therapeutic doses come from epilepsy trials of purified pharmaceutical cannabidiol under medical supervision, which is a different exposure from an unverified retail product taken indefinitely.

The 2018 Agriculture Improvement Act removed hemp from the controlled substances list and created a large consumer market without a corresponding testing requirement. The label accuracy findings are the predictable result of that arrangement rather than evidence of unusual bad faith among manufacturers.

For clinicians, the practical consequence is that a patient’s report of what they took frequently cannot be taken at face value. When a patient says a cannabidiol product did nothing, the differential includes an ineffective compound, an inadequate dose, and a bottle that did not contain what it claimed. Only the third is verifiable, and only with a certificate.

I have had patients bring in bottles they were certain had failed them, and the most likely explanation in several cases was that the product never contained a meaningful amount of cannabidiol to begin with. That is a frustrating thing to discover after months.

So I have stopped starting these conversations with dose. I start with the certificate. If a patient cannot produce a current batch-matched certificate from an accredited lab, the dose discussion is premature, because we would be titrating an unknown.

The other thing I say often is that reaching for the highest milligram figure on the shelf is not a strategy. The human data we have on cannabidiol dose-response show a curve that can bend back down, and the interaction risk with prescription medications rises with dose. More is a decision, not an upgrade.

None of this is an argument against cannabidiol. It is an argument for knowing what you are taking, which is a lower bar than the market currently clears.

Verify before you titrate. Ask for a current Certificate of Analysis with a batch number matching your bottle, from an independent accredited laboratory, covering contaminants as well as cannabinoid content. Then start low, hold each change for at least a week, and increase gradually rather than jumping tiers. If you take prescription medication regularly, review the full list with a physician or pharmacist before moving into the higher strength ranges.

Take away two things. Label potency in this market is unverified by default, and the published testing suggests that most products deviate from their stated amount by more than ten percent in one direction or the other. And higher strength is a different product, not a better one, with a dose-response curve that does not reward escalation and an interaction profile that punishes it.