One arm to one brain
— and then the drug starts leaving.

• The definition is a performance specification. An intravenous anaesthetic induces loss of consciousness in one arm–brain circulation time. Ketamine does not, and that single failure explains most of what is unusual about it.
• Three molecular targets, not one. Four chemical classes converge on the GABA-A receptor; ketamine blocks the open NMDA channel; dexmedetomidine works through an α2 adrenoceptor and the brain’s own sleep switch.
• The bottle is pharmacology. An emulsion, an alkaline powder and a propylene-glycol solution each solve a solubility problem, and each charges for it in pain, in instability, or in what may share the cannula.
• A bolus wears off by redistribution. The patient wakes with almost the entire dose still in the body. Metabolism decides the hangover, not the emergence.
• Context-sensitive half-time is the infusion’s version of that story, and it is the reason propofol supports total intravenous anaesthesia and thiopentone does not.

Ketamine fails this definition, and it is worth saying so plainly at the outset rather than discovering it in lesson 4. Its central effects become evident at about 90 seconds, not at 15. That is not a defect of potency or of lipid solubility — ketamine is five to ten times as lipid soluble as thiopentone — but a consequence of where it acts and what it has to do there. Every other agent in this module meets the specification.

The history is short and it explains the shape of the current formulary. The barbiturates arrived in the 1930s, and thiopentone’s rapid onset and short duration made it different in kind from anything used before. Phencyclidine was withdrawn for psychotomimetic reactions but left behind the chemically related ketamine. The steroidal agents were abandoned: pregnanolone would not dissolve, and althesin’s solubilising agent, Cremophor EL, was associated with anaphylaxis. Etomidate survived on cardiovascular stability alone. Propofol displaced almost everything once a formulation existed that did not have to be reconstituted.

The ideal agent, and how each real one fails it

The list below is the standard one, and it is worth learning as a list because it is the spine the rest of this module hangs on: every agent is a different pattern of compliance and failure against the same sixteen properties. The right way to read the table is not “propofol is best” but “each of these drugs still exists because there is a patient for whom its particular failures do not matter”.

Almost every question on this topic opens with a classification, and the two available classifications answer different questions. Chemical class predicts formulation, metabolism and the pattern of adverse effects — it is why the barbiturate needs an alkaline vial and the alkylphenol needs an emulsion. Molecular target predicts the shape of the clinical effect: whether the patient looks asleep or dissociated, whether the drug is analgesic, whether it will stop them breathing.

Three agents are named in the classification and taught nowhere in this module, and they are named deliberately because a classification that omits them is incomplete. Methohexitone is the oxybarbiturate — no sulphur at C2 — cleared about three times faster than thiopentone and still the agent of choice for electroconvulsive therapy. Remimazolam is an ultra-short-acting benzodiazepine designed to be broken down by tissue esterases rather than by the liver, which gives it the organ-independent offset that midazolam lacks. Alfaxalone is the surviving neuroactive steroid; the class was abandoned in human practice when althesin’s solubilising vehicle proved to cause anaphylaxis, not because the pharmacology was wrong.

Read the two etomidate rows together. R-etomidate is among the most active agents at the GABA-A receptor in the table; S-etomidate is inactive at every receptor in it, with roughly a 30-fold difference in activity between them. Two molecules with identical atoms, identical bonds and identical lipid solubility, and only one of them is an anaesthetic. That is about as clean a demonstration as pharmacology offers that these drugs act on a protein with a shaped binding site, and not by dissolving into a membrane — which is the whole argument that replaced the Meyer–Overton lipid theories.

GABA is the principal inhibitory transmitter of the mammalian central nervous system, and the GABA-A receptor is the only site proven to be involved in barbiturate-induced anaesthesia. Structurally it belongs to the same pentameric ligand-gated family as the nicotinic acetylcholine receptor: five subunits arranged around a central pore, each contributing four transmembrane segments to the channel wall. There are at least thirty subunit combinations, and which combination a neurone expresses decides which drugs affect it and how.

The three sites, and why they are not interchangeable

The GABA sites. Two of them, at the interfaces the α subunits make with the β subunits. Both must be occupied for the channel to open. Occupying one does nothing — the same asymmetry that governs the nicotinic receptor, and for the same structural reason.

The benzodiazepine site sits at the α/γ interface, and this is where midazolam binds. What it does there is positive allosteric modulation and nothing else: it increases the frequency with which the channel opens in response to GABA that is already present. It cannot open the channel by itself at any concentration. That single mechanistic fact is the origin of the benzodiazepine ceiling effect — a benzodiazepine can only amplify the inhibition the brain is already generating, so there is a maximum beyond which more drug achieves nothing — and it is why a pure benzodiazepine overdose is survivable in a way a barbiturate overdose is not.

The general anaesthetic site is somewhere else entirely — in the transmembrane domain of the β subunit, not at an interface in the plane you would draw the receptor from above. Propofol, etomidate, the barbiturates and the halogenated volatiles all act here, and the β subunit subtype matters: β2 and β3 are more sensitive to etomidate than β1, and sites on all three β transmembrane domains are crucial to propofol’s hypnotic action.

Propofol is not confined to GABA-A. It also lengthens glycine channel opening, inhibits neuronal nicotinic and 5-HT3 receptors, and inhibits the NMDA receptor by modulating sodium channel gating. Etomidate, by contrast, is remarkably selective — which is one reason its adverse effects are endocrine rather than neurological.

The NMDA receptor is an excitatory glutamate-gated cation channel, and it is unusual among ligand-gated channels in requiring three conditions at once before it will pass current: glutamate bound, glycine bound as co-agonist, and the membrane depolarised enough to expel a magnesium ion that otherwise sits in the pore. That magnesium plug is a voltage-dependent block of a chemically gated channel, and it is what makes the receptor a coincidence detector — the property underlying long-term potentiation, learning and memory.

Ketamine binds the phencyclidine recognition site inside the pore. The consequences of that location, rather than of the binding itself, are what make this block behave unlike a competitive antagonism:

• The antagonism is non-competitive. Ketamine does not compete with glutamate for the glutamate site — glutamate can continue to bind normally. Raising glutamate does not overcome the block. Calling it competitive is the commonest mis-description of it, and it is wrong in a way that matters: a competitive antagonist is by definition surmountable by more agonist, and this one is not.
• The block is use-dependent. The drug can only reach a site inside the channel once the channel has opened, so the block develops fastest and deepest in the neurones that are firing most. A quiet neurone is relatively spared.
• The concentration changes the mechanism. At sub-anaesthetic doses ketamine also binds the closed receptor and reduces the frequency of opening by an allosteric mechanism; at higher concentration it blocks both open and closed channels, and that is the dissociative state. The analgesic dose and the anaesthetic dose are not the same drug action turned up.

Dexmedetomidine is a highly selective agonist at the α2 adrenoceptor, a G-protein-coupled receptor whose intracellular effects include inhibition of adenylate cyclase and modulation of calcium and potassium conductance. Its selectivity for α2 over α1 is about 1600:1, against roughly 220:1 for clonidine — which is the quantitative version of the statement that dexmedetomidine is the more specific drug.

The chain matters more than the receptor. Reducing noradrenergic output from the locus coeruleus lifts the inhibition that nucleus exerts on the ventrolateral preoptic nucleus; released, the preoptic nucleus increases GABA and galanin release onto the tuberomammillary nucleus; and histamine release into cortex and subcortex falls. That is the endogenous non-REM sleep pathway, entered at its own switch. The clinical signature follows directly: patients are rousable and cooperative, follow commands while intubated, and fall asleep again when left alone.

None of these drugs dissolves in water on its own, because the same lipid solubility that gets them into the brain keeps them out of the ampoule. What was done about that is not packaging trivia — it is the origin of pain on injection, of two drug incompatibilities, and of the only lethal complication in the module.

The lipid emulsion — propofol

Propofol 1% is an oil-in-water emulsion: 10% soybean oil as the lipid phase, 1.2% purified egg phospholipid as emulsifier, 2.25% glycerol to make it isotonic, and sodium hydroxide to adjust the pH to about 7. EDTA or another agent is added because the emulsion is otherwise an excellent bacterial growth medium. Four consequences follow, and all four are examinable:

• Pain on injection, from the aqueous-phase propofol in contact with venous endothelium — which is why raising the proportion held in the lipid phase, as the medium-chain triglyceride formulations do, reduces it.
• Microbial growth, hence the preservative and the rule about discarding drawn-up syringes.
• A lipid load. An infusion delivers a non-nutritional calorie load that is significant over days and must be counted in the nutritional prescription — though it is not a way of providing calories, and treating it as one is how a patient ends up overfed on lipid.
• An allergy question that turns out to have a clear answer, and lesson 2 gives it.

The alkaline vial — thiopentone

Thiopentone is supplied as a pale yellow powder of the sodium salt, mixed with 6% by weight anhydrous sodium carbonate, in a vial filled with nitrogen rather than air. Reconstituted to 2.5%, the solution has a pH of about 10.5. Both measures exist for the same reason. Sodium carbonate reacts with water to give a strongly alkaline solution, which drives the keto–enol tautomerism towards the water-soluble enol form. Nitrogen replaces air because atmospheric carbon dioxide would dissolve, form carbonic acid, lower the pH and push the equilibrium back towards the insoluble keto form.

Propylene glycol — etomidate

Etomidate is presented as a 0.2% solution at pH 4.1 containing 35% v/v propylene glycol, which improves stability and reduces the irritant properties of the injection. A lipid emulsion formulation also exists. The solvent is why etomidate hurts going in — about a quarter of patients — and the lipid formulation exists to address exactly that.

Read the pKa column against the pH column. Thiopentone’s pKa sits almost exactly at plasma pH, so its ionisation — and therefore the fraction free to act — moves with the patient’s acid–base state; lesson 3 makes that quantitative. Propofol’s pKa of 11 puts it essentially entirely unionised at any physiological pH, so nothing the patient does to their pH changes its availability. Etomidate’s pKa of 4.2 means the opposite arrangement: unionised and lipid soluble in the patient, and soluble enough to bottle only because the ampoule is acidic.

Peak plasma concentration after a bolus arrives within a few seconds of the injection finishing. Peak effect arrives one to four minutes later. The gap is real, it is measurable, and misjudging it is the most common reason an otherwise adequate induction produces coughing at laryngoscopy.

The reason is that the plasma is not the biophase. Drug has to leave the blood and equilibrate with the brain, and that transfer takes time. The standard way of describing it is to hang a notional effect-site compartment off the central compartment, small enough to contain no meaningful amount of drug and therefore not to perturb the kinetics, and to give it a single rate constant:

In words: the effect site chases the plasma at a rate proportional to the gap between them. C_p is the plasma concentration, C_e the effect-site concentration, and k_e0 the rate constant for elimination of drug from the effect site, in min⁻¹. A large k_e0 means the effect site fills and empties quickly, and the lag is short.

The peak of the effect-site curve is where it crosses the falling plasma curve — which has to be true, because that is where dC_e/dt is zero. At that moment the patient is at their deepest, and beyond it both curves fall together. That crossing point is what a pump means when it reports an effect-site concentration, and it is why an effect-site-targeted infusion pauses.

The same relationship explains a comparison that comes up repeatedly: alfentanil has far less hysteresis than fentanyl because its k_e0 is larger, not because it is more potent — it is considerably less potent. The general theory of hysteresis and effect-site modelling belongs to the pharmacokinetics module; what this lesson adds is the numbers for these particular drugs and what they do to a real induction.

Follow the mass, not the concentration. An induction dose is injected into a central compartment consisting of blood and the vessel-rich group — brain, heart, liver, kidney, endocrine tissue — which together are perhaps 10% of body mass and receive perhaps 75% of the cardiac output. The concentration there goes very high very fast, which is exactly what is required for an effect within one arm–brain circulation.

That high concentration is also a steep gradient into every tissue that has not yet received any. Muscle is about 50% of body mass and receives roughly a fifth of the cardiac output, so it fills over minutes, and while it fills it draws drug out of the central compartment. The plasma concentration therefore falls steeply — and the brain, in equilibrium with the plasma, empties with it. The patient wakes.

Two consequences follow, and both are more examinable than the statement itself.

1. A second bolus lasts longer than the first. The gradient into muscle is what terminated the first dose, and the first dose partly abolished it. Give the second dose into a partly loaded peripheral compartment and less drug leaves the centre, so the concentration stays higher for longer. This is why repeated boluses accumulate in a way that a single dose does not predict.
2. An infusion inverts the whole argument. Run the drug for hours and the peripheral compartments approach equilibrium with the plasma. Stop, and there is no longer a gradient out — in fact drug now returns from muscle and fat into the blood, opposing the fall. Recovery then depends on metabolism, and the agents diverge sharply. That is the next section.

Fat, the third compartment, is worth a separate sentence because it behaves unlike the other two. It is large and extremely poorly perfused, so its time constant is measured in hours; it plays almost no part in terminating a bolus, and it is the reason a long infusion of a lipid-soluble agent has an offset that a short one does not predict. Compartment models themselves — the derivation, the rate constants, the BET scheme — belong to kinetics and models.

It is also not the elimination half-life, and the reason is worth stating mechanically rather than as a slogan. The elimination half-life describes the terminal exponential alone — the slope measured after a single bolus, when the peripheral compartments were empty. After a long infusion they are not empty, and when the infusion stops they hand drug back. The terminal slope was measured in conditions that no longer apply.

A second wording point: half-lives are constant and half-times are not. After one context-sensitive half-time, the next halving takes longer still, because the faster distribution processes have already been used up and the slower ones increasingly dominate. A half-life is a property of an exponential; a half-time is an observed duration.

Two agents on that figure are worth reading against each other. Propofol stays under 40 minutes even at eight hours, because its clearance is so high that elimination keeps pace with return from the periphery — which is the entire pharmacological case for total intravenous anaesthesia, and lesson 2 takes it apart. Thiopentone leaves the axis, for two separate reasons that compound: a low clearance, and a metabolism that saturates at high dose so that handling moves from first-order towards zero-order. An infusion of thiopentone is a decision with a tail measured in days.

The textbook dose is a starting point for a 70 kg adult who is not bleeding. Everything below moves it, and each one moves it for a reason already established in this lesson — which is why this section is the synthesis rather than a separate list to learn.

The kinetic numbers, and a disagreement between sources

The last column is there because the two standard references do not agree, and pretending otherwise is worse than showing it. Peck reports volume of distribution, Miller reports volume of distribution at steady state, and the two draw on different source series — so propofol’s elimination half-life is 5 to 12 hours in one book and 4 to 7 in the other, and thiopentone’s is 6 to 15 against 7 to 17. Ketamine protein binding is 25% in Peck and 12% in Miller. In an answer, quote one source’s figure and say which — or give the range across both. What is not in dispute is the ordering: propofol’s clearance is an order of magnitude greater than thiopentone’s, and ketamine is by some distance the least protein-bound of the four.