Two dangers sit on either side of the drug chart in AKI: giving too much, so a renally cleared drug accumulates to toxicity, and giving too little, so an antibiotic never reaches the level that treats the sepsis. AKI makes both easy, because its falling and then recovering function is a moving target that standard equations cannot track. The same kidney that cannot clear a drug can, on dialysis, be used to remove a poison — so this chapter is about subtracting drug where the kidney has failed, and adding clearance where a toxin must go.
Loading versus maintenance: the first distinction
The single most useful dosing principle is that loading and maintenance doses obey different rules. The loading dose fills the volume of distribution to reach a therapeutic concentration, and that volume is not much changed by AKI, so the loading dose is usually given in full. This matters most for antibiotics in sepsis, where the instinct to 'go easy on the kidneys' by reducing the first dose is exactly wrong — it delays therapeutic levels when the infection is most dangerous. The maintenance dose, by contrast, replaces what the body clears, and since AKI reduces clearance, the maintenance dose is reduced or its interval extended. Full load, reduced maintenance: the rule covers most situations.
The non-steady-state trap
The equations clinicians reach for to estimate kidney function — the eGFR formulae — are built on an assumption that fails in AKI: steady state. They take a single creatinine and assume production and clearance are balanced, which is true in stable CKD but false in an acute injury where the creatinine is still rising or falling. In a patient whose creatinine is climbing, the true GFR is lower than the equation reports, because the creatinine has not yet caught up — so the equation flatters the kidney and risks over-dosing. In recovery, with creatinine falling, the true GFR is higher than reported, risking under-dosing. The practical response is to distrust the calculated number in non-steady-state AKI, to dose with caution in the direction the creatinine is moving, and to lean on therapeutic drug monitoring wherever it exists, since measured levels sidestep the estimation problem entirely.
Counting the dialysis clearance
Once a patient is on renal replacement therapy, the circuit adds clearance that the dosing must account for, and the commonest error — met repeatedly in this volume — is to dose as if the patient were anuric and so under-dose. CRRT in particular provides continuous, moderate clearance, and antibiotics dosed for anuria reach sub-therapeutic levels, causing treatment failure in exactly the septic patients who can least afford it. Intermittent haemodialysis removes drugs in bursts, so dialysable drugs are timed around the sessions — typically dosed after dialysis to avoid removing the dose just given. SLED sits in between. Across all modalities, the drugs with narrow therapeutic windows are monitored directly: vancomycin and the aminoglycosides are dosed by level, not by formula, precisely because the kinetics on dialysis are too unpredictable to guess.
The drugs that accumulate
A practical list of culprits earns its place because these drugs cause avoidable harm when their accumulation is forgotten. Gabapentin and pregabalin are renally cleared and accumulate to cause sedation and confusion. Morphine's active metabolites accumulate and prolong its effect, so safer opioids are preferred and all are used cautiously. Low-molecular-weight heparin accumulates and raises bleeding risk, so it is monitored by anti-Xa or replaced with unfractionated heparin in significant AKI. Digoxin, the sulfonylureas with their hypoglycaemia, and many others join the list. And metformin deserves singling out: it is held in significant AKI because its accumulation drives a metabolic lactic acidosis. The discipline is to scan the chart in any AKI for renally cleared drugs and to adjust or stop them before they accumulate.
When the kidney becomes the treatment: dialysable poisons
The same dialysis that supports a failed kidney can be turned to a different purpose — removing a poison from the blood. Whether a toxin is dialysable is predictable from its physicochemistry: small molecules that are water-soluble, minimally bound to protein, and confined to a small volume of distribution are removed well, because dialysis can only clear what is free in the plasma water. The classic dialysable poisons follow from these properties — methanol and ethylene glycol, salicylates, lithium, metformin, valproate in massive overdose, theophylline, phenobarbital, and dabigatran. The mirror-image rule is just as useful: drugs with a large volume of distribution or high protein binding — the tricyclic antidepressants, digoxin, most benzodiazepines — are barely touched by dialysis, because most of the drug is hidden in tissues or bound, not free to be cleared. For these, dialysis is futile and other measures, such as digoxin-specific antibody fragments, are used instead.
Delivering extracorporeal treatment
When a poison is dialysable and the poisoning is severe, extracorporeal treatment is delivered with the modality that fits. Intermittent haemodialysis is usually the most efficient for rapid removal, because its high clearance clears a small-volume toxin quickly. CRRT is chosen for the haemodynamically unstable patient and for managing rebound — lithium is the archetype, redistributing out of cells back into the plasma after an intermittent session, so that levels climb again and a repeat or continuous treatment is needed. The toxin-specific approaches pair removal with an antidote where one exists. Toxic alcohols are treated by blocking alcohol dehydrogenase with fomepizole to stop the formation of toxic metabolites, alongside haemodialysis to remove the parent alcohol and correct the acidosis, with dialysis indicated for severe acidosis, high levels, or end-organ effects. Salicylate poisoning is treated by alkalinising the urine to enhance elimination, with haemodialysis added for severe features. Lithium and metformin-associated lactic acidosis are both managed with haemodialysis, the latter clearing the drug and correcting the acidosis at once. Consensus guidance from the extracorporeal-treatments-in-poisoning literature defines which poisonings warrant these interventions, and the poison centre is an early call.
After removal, and a closing principle
Two final points tie the chapter together. First, removal is not always one-and-done: rebound from tissue redistribution, classically with lithium, means levels and the patient must be rechecked after a session and treatment repeated or continued as needed. Second, the dosing story does not end when the AKI does — as the kidney recovers, clearance rises, and a maintenance dose set for anuria becomes an under-dose, so every drug must be re-dosed for the function the patient actually recovers to, the same reconciliation discipline of the recovery chapter. The unifying idea is simple: in AKI, subtract drug where the kidney can no longer clear it, add clearance where a toxin must be removed, and keep adjusting as the moving target moves.
