Titration: Start Low and Go Slow as a Protocol

Why it is a protocol and not a platitude human data

Titration is the practice of approaching an unknown effective dose from below in increments small enough that an overshoot is recoverable, waiting long enough between increments to observe the full effect of the last one. Stated that way it is obviously a measurement procedure rather than a piece of caution, and its two parameters are the increment size and the waiting interval. Both are set by pharmacology. The increment should be a fraction of the smallest dose expected to do anything, because the point is to find the threshold rather than to land on the target in one step. The interval is the parameter people get wrong, and it is set by the time to peak effect rather than by the time to onset. Redosing after onset but before peak is the single most reliable way to produce an overdose with a substance that is otherwise forgiving, because the second dose lands while the first is still rising and the two peaks sum. The reason the protocol is worth writing out formally is that the correct interval differs between routes by two orders of magnitude, and intuitions calibrated on one route are dangerous on the other.

Sources: Huestis MA 2007 · Lucas CJ 2018

Route pharmacokinetics, with numbers human data

Inhalation. Absorption across the alveolar surface is fast and the drug reaches the brain in one circulation time, so onset is within minutes; plasma concentrations of Δ9-THC peak within roughly the first ten minutes of the end of smoking and subjective effect commonly peaks somewhere around ten to thirty minutes, declining substantially over two to four hours. The feedback loop is therefore short: a person who waits half an hour has seen most of what a dose is going to do. The bioavailability of the inhaled route is variable but relatively high, and — importantly — it is dominated by technique, device and puff topography, which is why a labelled cartridge concentration does not translate into a delivered dose. Oral. Absorption is slower and irregular, onset is typically thirty to a hundred and twenty minutes, and peak plasma concentration commonly falls two to four hours after ingestion, with wide inter-individual variation and a strong food effect: a high-fat meal increases absorption of a lipophilic cannabinoid substantially and shifts the curve. Oral bioavailability is low and variable, in part because of extensive first-pass metabolism. That first pass is also the mechanism that makes the oral route different in kind and not only in timing: hepatic metabolism converts Δ9-THC to 11-hydroxy-Δ9-THC, a metabolite that is at least as pharmacologically active as the parent and readily enters the central nervous system, so an oral dose delivers a different mix of active species than an inhaled one. The combination of a long, variable delay and a metabolite with strong central activity is the mechanistic explanation for why edible overconsumption is the classic acute presentation: the person waits, feels little, takes more, and then receives the sum of the doses at a peak they cannot shorten. Emergency-department data bear this out, with edibles accounting for a share of cannabis-attributable acute visits far larger than their share of total cannabinoid sold.

Sources: Huestis MA 2007 · Lucas CJ 2018 · Monte AA 2019

Why 11-OH-THC matters to the protocol contested human data

The oral route is not simply the inhaled route delayed. Substantial first-pass hepatic metabolism of Δ9-THC produces 11-hydroxy-Δ9-THC, and this metabolite is itself a cannabinoid receptor agonist that is reported to be equipotent with or more potent than the parent compound, with ready brain penetration. An oral dose therefore delivers a pharmacologically different exposure, weighted toward a metabolite, arriving over a longer window. The practical consequences for titration are three. Milligram figures do not transfer between routes, so an oral dose cannot be inferred from an inhaled dose that felt right. The oral peak may be both later and disproportionately stronger than the plasma curve of the parent compound alone would suggest. And inter-individual differences in hepatic metabolism, including genetic variation in the enzymes involved and competing substrates or inhibitors, act on the oral route far more than on the inhaled one — which means the same product produces a wider spread of experiences orally than it does by inhalation.

Contested — caveat. The relative potency of 11-OH-THC against Δ9-THC comes from a small body of older controlled human work and animal data, and the figures quoted in the literature vary. That it is active, brain-penetrant and formed in quantity by the oral route is not in dispute; a precise potency ratio is not well established and should not be treated as a number to calculate with.

Sources: Huestis MA 2007 · Lucas CJ 2018

Tolerance to a partial agonist does not transfer to a full agonist contested in vitro

This is the point on this page that has actually killed people, and it deserves to be stated without softening. Δ9-THC is a partial agonist at CB1, with reported binding affinity in the tens of nanomolar. The synthetic cannabinoid receptor agonists sold as herbal-incense products are, as a class, high-affinity full agonists: JWH-018, the first-generation compound in wide circulation, has reported CB1 affinity around an order of magnitude tighter than Δ9-THC and behaves as a full agonist, and the later indazole- and indole-carboxamide generations are more potent again. A partial agonist has a ceiling: beyond a certain receptor occupancy, additional drug produces no additional maximal effect, which is a large part of why acute cannabis overconsumption is usually a bad few hours rather than a medical emergency. A full agonist has no such ceiling, so the dose-response relationship keeps climbing into effects that have no counterpart in cannabis intoxication — seizures, tachyarrhythmia, extreme hypertension or hypotension, hyperthermia and agitated delirium. Tolerance acquired through heavy cannabis use is largely CB1 receptor downregulation and desensitisation. Against a full agonist that can drive a maximal response from a reduced receptor population, that adaptation offers much less protection than it feels like it should — and it is worse than useless in one specific respect: it licenses a larger dose. The heavy user is precisely the person whose experience tells them that a small amount will do little, and who therefore takes an amount that a full agonist will turn into a toxicological event. The same logic applies whenever a product of unknown composition is approached with tolerance built on a known one, which is a common situation in a market where novel cannabinoids appear faster than they can be characterised.

Contested — caveat. The non-transfer of tolerance is a mechanistic inference from receptor pharmacology — partial versus full agonism at CB1, and the affinity difference — combined with clinical case series in which experienced cannabis users presented with severe synthetic-cannabinoid toxicity. There is no controlled human cross-tolerance study comparing Δ9-THC tolerance against a synthetic full agonist, and there could not ethically be one. The reported affinity figures are single-laboratory values from different assay systems and are not directly comparable to two significant figures.

Sources: Showalter VM 1996 · Huffman JW 2005 · Banister SD 2018 · Castaneto MS 2014 · Hermanns-Clausen M 2013

Batch variance, and the compound you cannot titrate human data

Titration works within a stable dose-response relationship. Two things break it. The first is batch-to-batch variance: a dose established on one batch is information about that batch, and a new batch is a new unknown that has to be approached from below again. Where distribution was uneven the situation is worse still, because even one batch does not have a single dose-response relationship — the variance is within it, so a previously safe portion is not evidence about the next portion. This is the same argument the hot-spots page makes, arriving from the other direction: titration, which is the standard defence against an unknown potency, is defeated by a product whose potency varies inside the package. The second and absolute limit is identity. Titration presupposes that there is a dose-response curve for the substance being taken, even if the person does not know where they sit on it. An unidentified compound has no such curve available: the increment size cannot be chosen, because the scale is unknown; the waiting interval cannot be chosen, because the time to peak is unknown and for many novel compounds is longer than intuition suggests; and the shape of the curve is unknown, including whether there is a ceiling at all. A substance that is only known by what it was sold as is not being titrated; it is being sampled. The honest statement of the limit is that harm reduction can make consuming an identified compound at an unknown personal dose much safer, and can do relatively little about consuming an unidentified compound at an unknown dose. That is why identity, not dose, is the first question.

• A dose established on one batch is information about that batch only.
• Within-batch variance defeats titration itself, because the next portion is a new draw.
• Without identity there is no dose-response curve, so no increment and no interval can be chosen rationally.
• Sold-as is not identity. Only analysis against an authentic reference standard is identity.

Sources: Frinculescu A 2016 · Luzio A 2019 · Castaneto MS 2015 · Trecki J 2015

What the protocol logic looks like, stated generally human data

Written out as logic rather than as a recommendation for any substance: establish identity before anything else, because every subsequent step depends on it. Establish concentration, by assay or by a documented dilution you performed and labelled. Choose an increment that is a fraction of the smallest dose expected to produce any effect for that compound by that route. Take one increment. Wait past the expected peak for the route, not past the expected onset — which means minutes to tens of minutes for inhalation and hours for ingestion, with the oral interval extended if food was involved. Record what happened, because the next increment is a decision that needs data. Then, if a further increment is taken at all, take one increment and not two. Do not change route and dose in the same step, and do not change product and dose in the same step, because a single experiment with two variables answers nothing. Treat any new batch as a new unknown. None of this is specific to a substance and none of it is a dosing recommendation; it is the general structure of approaching an unknown dose-response relationship from below, which is the same structure a pharmacologist would use and the same one a clinical dose-escalation study is built on.

Sources: Huestis MA 2007 · Lucas CJ 2018 · Castaneto MS 2015