Homogeneity: Even Mixing as a Safety Specification

Three states, three different guarantees human data

Mixing a potent active into a carrier can end in one of three physical states, and they do not offer the same protection. In a true solution the active is molecularly dispersed in the solvent, and the concentration is the same in every aliquot by definition; take a millilitre from the top or the bottom and you have the same dose. In a suspension the active is present as solid particles dispersed through a liquid; the concentration is uniform only while the suspension is agitated, and Stokes settling begins the moment it stops, so the first dose out of an unshaken bottle and the last dose out of it differ, sometimes by a lot. In a dry blend the active is a minor solid component among larger masses of other solids, and the blend is uniform only to the extent that the blending process was designed to make it so and verified to have done so; dry blends also de-mix, segregating by particle size and density under vibration and handling, which is why the bottom of a shipped drum is not the same material as the top. The practical ranking follows directly: for a potent active, get it into solution and keep it there. A solution is the only one of the three states whose uniformity is a property of physics rather than a property of your process control.

• Solution: uniform by definition; the safest state for a potent minor component.
• Suspension: uniform only under agitation; settles by particle size and density difference; shake-before-use is a dose instruction, not a courtesy.
• Dry blend: uniform only if engineered and verified; actively de-mixes during handling and transport.

Sources: Johnson KC 2008 · United States Pharmacopeia 2023

Solubility is the constraint that silently converts a solution into a suspension human data

Every diluent has a saturation concentration for every solute, and it is a hard ceiling. Dissolve up to it and you have a solution; attempt to exceed it and the excess remains or comes out as solid, and what you have is a suspension that looks like a solution until it separates. This failure is quiet, which is what makes it dangerous: a stock that was prepared above saturation may appear clear while warm and then deposit crystal on the walls and the bottom as it cools, at which point the liquid phase is depleted and every volumetric dose drawn from it is lower than the label, while the material that eventually gets scraped or shaken back into suspension is a concentrated slug. Temperature, co-solvent ratio and the presence of other dissolved matter all move the saturation point, so a stock that is fine on the bench in summer can precipitate in a cold room. The checks are simple. Hold the prepared stock at the coldest temperature it will see and look for haze, crystal on the glass, or a film at the meniscus. Filter or discard anything that has deposited rather than trying to redissolve it in place. Prepare stocks comfortably below the saturation limit rather than at it, and accept a larger volume as the price of a solution that stays one. Lipophilic cannabinoids in particular are poorly soluble in water and readily soluble in oils, ethanol and glycols, which is why aqueous edible matrices are formulated as emulsions rather than solutions and why an aqueous beverage with an unemulsified cannabinoid will separate.

Sources: United States Pharmacopeia 2023 · Meehan-Atrash J 2021 · Lucas CJ 2018

Wetting, penetration and why a solvent has to do more than arrive human data

When a solution is applied to a solid carrier, the outcome depends on whether the liquid wets and penetrates the carrier or merely sits on its surface. Wetting is a matter of surface energy and viscosity: a liquid that beads on a waxy leaf cuticle deposits its solute in the footprint of the droplet, while a liquid that spreads and wicks distributes it along the capillary paths of the material. The competition is with evaporation. A volatile carrier solvent applied to a large surface area can be gone before capillary transport has moved it any distance, in which case the solute dries approximately where it landed and the distribution is a map of the spray pattern rather than of the material. Slower evaporation, gentler application, and mechanical redistribution during drying all push toward evenness; a fast-flashing solvent applied under a fan pushes hard the other way. The same logic governs edible bases: an active dissolved in a fat phase will distribute through the fat wherever the fat goes, so incorporation has to be complete before the matrix sets, and anything that sets or gels quickly locks in whatever distribution existed at that instant. This is also where the residual-solvent question begins, because the solvent that carried the compound in has to come back out, and a solvent chosen for good wetting is often one that is retained in an oily or viscous matrix far longer than intuition suggests.

Sources: Meehan-Atrash J 2021 · United States Pharmacopeia 2023 · International Council for Harmonisation 2021

Geometric dilution: the technique that actually works for a potent minor component human data

The intuitive way to mix a small mass into a large one — put the small mass in and stir — is close to the worst available method, because the small mass has to travel the whole distance and nothing in the process makes it do so. Geometric dilution inverts the problem. Combine the potent component with an approximately equal volume of diluent and mix that pair thoroughly. Then add another quantity of diluent approximately equal to the current total and mix again. Repeat, roughly doubling the mass at each step, until the whole batch is incorporated. Each step is a mix of two comparable masses, which is the case mixing handles well, and the number of steps grows only logarithmically with the dilution factor: taking 100 mg of active into 100 g of carrier is a thousandfold dilution and takes about ten doubling steps. Two supporting practices matter as much as the sequence. Particle size should be reduced and matched before blending, because content-uniformity variance in a low-dose blend scales with the particle size of the active — a coarse crystal is a hot spot with a delivery vehicle, and sieving or gentle milling of the active before the first step does more for uniformity than any amount of subsequent mixing. And the mixing action should fold and tumble rather than stir: rotational tumbling with a variety of particle paths distributes a minor component, while stirring in a fixed geometry can circulate material in stable paths that never intermix. Over-mixing a dry blend is also a real failure, because prolonged agitation of a mixture with dissimilar particle sizes eventually segregates it again.

• Double the mass at each step rather than adding a small mass to a large one; a thousandfold dilution is about ten steps.
• Reduce and match particle size before blending — content-uniformity variance scales with the particle size of the low-dose component.
• Tumble and fold rather than stir; vary the particle paths.
• Mix long enough to distribute and not so long that segregation restarts; determine the time empirically with assays, not by feel.
• A dry blend that will be shipped or stored will de-mix in transit. If the dose matters, unit-dose it before it travels.

Sources: Johnson KC 2008 · United States Pharmacopeia 2023 · United States Pharmacopeia 2023

Edibles: full incorporation and the fat phase human data

Edibles concentrate every difficulty on this page. The active is lipophilic, the matrix is usually a multi-phase system of sugar, water, hydrocolloid and fat, and the finished article is a discrete unit whose individual dose is what the consumer experiences — which means the relevant specification is unit-to-unit variance, not batch average. Because cannabinoids partition into fat, the fat phase is where the active lives, and the distribution of the fat therefore determines the distribution of the dose. An active dissolved in an oil that is then emulsified into a batter is distributed as well as the emulsion is; an active added directly to a sugar syrup will not dissolve, will form a separate phase, and will migrate. Anything that separates the phases after mixing re-sorts the dose: fat rising in a warm mould, a poorly stabilised emulsion breaking, or a syrup skinning over. Deposition timing matters as well, since a mould poured over several minutes while the mass is settling produces a dose gradient from the first cavity to the last. There is a documented reason to treat all of this seriously rather than as artisanal detail. Independent assays of retail edible products have repeatedly found labelled and measured cannabinoid content diverging in both directions, including substantial under- and over-labelling, and reviewers have noted regional variability in labelled unit strength on top of measurement disagreement. Meanwhile emergency-department data show that edibles account for a share of cannabis-attributable acute visits far out of proportion to their share of total cannabinoid sold, which is the clinical shape of a product class in which the delivered dose is poorly controlled and the pharmacokinetics punish error.

Sources: Vandrey R 2015 · Johnson-Arbor K 2023 · Monte AA 2019 · Lucas CJ 2018

How uniformity is verified rather than assumed human data

The distinguishing practice of a competent producer is not better mixing; it is measurement. Uniformity is verified by sampling the batch at multiple locations that are chosen to be the places most likely to differ — top, middle and bottom of a vessel, the centre and the walls, the first, middle and last units off a depositor, the first and last scoop out of a drum — assaying each sample independently, and then computing the spread rather than the average. Ten units is the conventional starting sample size, and the statistic that matters is the relative standard deviation between them, together with the range: a mean at target with a 25 percent RSD is a dangerous batch, and the mean alone will not say so. A composite sample, in which several increments are combined and homogenised before a single assay, measures the batch mean accurately and says nothing whatever about unit-to-unit variance; this distinction is important because a great many cannabis certificates of analysis report exactly one composite result for an edible batch, and that result is not a uniformity test. The pharmacopoeial analogue is worth knowing even where it does not legally apply: USP <905> Uniformity of Dosage Units assays ten individual units, combines the deviation of the mean from target with the observed standard deviation into a single acceptance value, and additionally requires that no individual unit fall outside 75 to 125 percent of the mean at the second stage. That structure exists precisely because pharmaceutical regulators concluded that a correct batch average is not evidence of a correct dose. Cannabis-industry practice varies enormously against that benchmark, from processors running in-process uniformity assays on every lot to processors who have never measured variance at all, and the certificate rarely distinguishes the two.

• Sample where the batch is most likely to differ, not where it is convenient.
• Assay increments individually. Report the RSD and the range, not only the mean.
• A composite assay measures the mean and is not a uniformity test.
• USP <905>: 10 units, an acceptance value combining mean deviation and standard deviation, individuals within 75-125 % of the mean at stage 2.
• Retain samples from each lot so that a later complaint can be investigated against the material that caused it.

Sources: United States Pharmacopeia 2023 · Johnson KC 2008 · International Organization for Standardization / International Electrotechnical Commission 2017 · Vandrey R 2015