The Music Before the Rocks
The Sirens did not lure sailors because they sang badly. They were dangerous because they sang beautifully.
In Homer’s Odyssey, Odysseus knows that the Sirens’ song will be irresistible. He does not pretend he is immune to it. Instead, he orders his crew to fill their ears with wax and bind him to the mast, commanding them not to release him however urgently he begs. This is not a story about ignorance. It is a story about wisdom: the recognition that even the strongest mind may need protection from a beautiful false certainty.1
The myth matters because intensive care has its own Siren: physiology, singing with the dangerous beauty of things that are true in isolation, but untested in the whole patient.
Physiology sings in the language of pressure gradients, receptor pathways, mucus viscosity, alveolar recruitment, oxygen delivery, intravascular volume, oncotic pressure, renal blood flow, glucose toxicity, catecholamine signalling and inflammatory modulation. It sings to the clinician at 3am, standing beside a ventilator, a vasopressor pump and a worsening blood gas. It offers the thing we most crave in critical illness: a coherent explanation.
The mucus is thick, so thin it. The albumin is low, so replace it. The glucose is high, so normalise it. The kidney is underperfused, so dilate its vessels. The alveoli are oedematous, so stimulate fluid clearance. The oxygenation is poor, so recruit the lung. The circulation is empty, so fill it.
These statements are not stupid. That is exactly why they are dangerous.
The Seduction Is Not New
The idea that physiology can seduce us is not new. A JAMA editorial from 2009 gave the problem its perfect name: “the seduction of physiology”.2 The phrase is worth lingering over. Physiology does not usually bully us into error. It flatters us. It tells us that we understand. It takes the complexity of a critically ill human being and reduces it to a mechanism that fits on a slide, in a lecture, in a guideline box, or in the satisfying simplicity of a ward-round sentence.
But mechanism is not medicine. It is the beginning of medicine.
No serious biological intervention enters a randomised trial without a physiological rationale. It would be unethical, irrational and scientifically incoherent to randomise patients to an intervention for which there was no plausible reason to expect benefit. Every biological interventional trial that has ever been run has had some physiological basis. That is the entry ticket. It is not the verdict.
Physiology generates hypotheses. Trials test whether those hypotheses survive contact with the whole patient.
The Intervention Is Never Just the Desired Effect
The human failure is subtle. We begin with a desired physiological effect and then mistake that desired effect for the whole intervention.
We say, “albumin exerts oncotic pressure,” as though that were the same as saying “albumin improves survival.” We say, “hypertonic saline hydrates airway secretions,” as though that were the same as saying “the patient is liberated from the ventilator sooner.” We say, “tight glucose control restores normoglycaemia,” as though the number and the patient were interchangeable. We say, “beta-agonists clear alveolar oedema,” as though receptor signalling were the same thing as recovery from ARDS.
But an intervention is never only the thing we want it to be.
It has the intended physiological effect, if we are lucky. It also has unintended physiological effects. It has pharmacological effects, procedural effects, behavioural effects, opportunity costs, nursing effects, physiotherapy effects, prescribing effects, monitoring effects, and effects on the wider ecology of critical care. Some are known. Some are guessed. Some are never measured individually. Some are invisible until they appear as bleeding, bronchospasm, hypoxaemia, renal failure, hypoglycaemia, longer ventilation, or death.
A trial does not merely ask whether the mechanism is plausible. It asks whether the totality of consequences helps the patient.
MARCH: The Hero of the Story
This is why MARCH is such an important trial. It should not be read as a disappointment. It should be read as a success of clinical science.
MARCH did not test an absurd idea. It tested an extremely sensible one. Mechanically ventilated patients with acute respiratory failure often develop retained secretions because invasive ventilation disrupts mucociliary clearance and alters secretion properties. Usual airway clearance includes suctioning, humidification, isotonic saline and respiratory physiotherapy, and mucoactive agents are often added when secretions are thick or difficult to clear.3
The practice was common, variable and empirical. The MARCH protocol described mucoactive prescribing as being driven by local availability, clinician preference and prior experience, rather than by national evidence-based guidance. It also made clear why the question mattered: if mucoactives worked, they could be used more appropriately; if they were ineffective, unnecessary or harmful, they could be prevented.4
That is not therapeutic nihilism. That is the moral purpose of a trial.
The physiology was seductive. Carbocisteine is a mucoregulator, intended to redress the balance between glycoproteins that regulate mucus viscosity and restore more normal viscoelastic properties. Hypertonic saline is an expectorant, intended to thin mucus by increasing mucosal water volume, maintaining the periciliary layer, decreasing mucus adhesiveness and stimulating cough.5
One can almost hear the Sirens: thick secretions obstruct the airway; these drugs loosen secretions; airway clearance improves; ventilation shortens; the patient recovers sooner.
That is a beautiful song.
What MARCH Asked
MARCH asked whether the song was enough.
The trial randomised 1,956 critically ill, mechanically ventilated patients with acute respiratory failure and difficult-to-clear secretions in a multicentre, open-label, 2-by-2 factorial design. Patients received usual care alone, carbocisteine, nebulised 6% or 7% hypertonic saline, or both interventions. The primary outcome was not a change in secretion viscosity, sputum appearance or clinician satisfaction. It was duration of mechanical ventilation from randomisation to first successful unassisted breathing.3
That choice of outcome is crucial. MARCH did not ask whether the mucus could be modified. It asked whether the patient outcome could.
The answer was clear. Neither carbocisteine nor hypertonic saline reduced the duration of mechanical ventilation. Median duration of mechanical ventilation was 186.1 hours with carbocisteine and 172.7 hours without carbocisteine. It was 184.5 hours with hypertonic saline and 174.3 hours without hypertonic saline.3
More importantly, the trial heard notes in the music that physiology alone had not made loud enough. Clinically important upper gastrointestinal bleeding occurred more often with carbocisteine. Bronchoconstriction requiring bronchodilator treatment and hypoxaemia during nebulisation occurred more often with hypertonic saline.3
The mucus may have been the target. The patient was the outcome.
The Trial Succeeded
MARCH did exactly what a trial is meant to do.
It took a common, plausible, low-glamour intervention and forced it to answer the only question that matters: when used in real critically ill patients, does the sum of its effects improve patient-centred outcomes? Not the desired effect. Not the most elegant effect. Not the physiological effect we teach. The sum of all effects.
This is why MARCH is the hero in this story. It gives clinicians permission to stop being enchanted. It protects patients from an intervention that sounded useful but did not improve liberation from mechanical ventilation and was associated with harm. It turns uncertainty into knowledge. It turns habit into evidence. It turns plausible practice into accountable practice.
De-implementation is not failure. De-implementation is one of the most mature forms of progress. It is medicine learning to put something down.
The Graveyard of Beautiful Mechanisms
MARCH belongs to a much larger history in critical care. The shoreline is littered with interventions that were once carried by a lovely physiological tune.
Albumin is a classic example. The physiological story is almost irresistible: albumin contributes to plasma oncotic pressure, so albumin should retain fluid in the intravascular compartment, restore circulating volume more efficiently and improve haemodynamics. Yet the SAFE trial found similar 28-day outcomes with albumin and saline in a broad ICU population.6 In traumatic brain injury, albumin was associated with higher mortality than saline.7 In severe sepsis and septic shock, ALBIOS showed no survival benefit from albumin replacement compared with crystalloids alone.8 The oncotic pressure was real. It was simply not the whole truth.
Tight glycaemic control is the metabolic version of the same story. Hyperglycaemia in critical illness is associated with adverse outcomes. Normal glucose therefore looked like normal care. NICE-SUGAR showed the danger of confusing a normal number with a better patient: intensive glucose control increased mortality compared with conventional control.9
Renal-dose dopamine may be the purest example of the spell. The promise was simple: increase renal blood flow, increase urine output, protect the kidney. The effect on urine output could feel reassuring at the bedside. But a multicentre randomised trial in critically ill patients with early renal dysfunction found no clinically significant renal protection.10 A better-looking fluid balance chart was not kidney salvation.
Beta-agonists in ARDS had their own elegance. If beta-2 stimulation accelerates alveolar fluid clearance, then salbutamol should help dry the injured lung. BALTI-2 tested that idea and found that intravenous salbutamol was poorly tolerated and associated with worse outcomes.11 The receptor biology was plausible. The patient did not benefit.
High-frequency oscillatory ventilation promised open lungs, tiny tidal volumes, less cyclic recruitment and derecruitment, and better oxygenation. It was a ventilatory strategy with intellectual beauty. But OSCILLATE found higher mortality with high-frequency oscillation than conventional ventilation in early ARDS, while OSCAR found no mortality benefit.12, 13 Oxygenation is not survival. Sometimes the monitor smiles while the patient suffers.
Hydroxyethyl starch carried the same colloid logic as albumin, but with a different price. Colloids should expand plasma volume efficiently; starch should therefore be a useful resuscitation fluid. In CHEST, hydroxyethyl starch did not reduce 90-day mortality compared with saline and was associated with increased use of renal replacement therapy.14 The circulation was not the only organ listening.
Early goal-directed therapy for septic shock was a symphony of targets: central venous pressure, mean arterial pressure, urine output and central venous oxygen saturation. The logic was seductive: correct oxygen delivery and perfusion variables early, and survival should improve. But ProCESS, ARISE and ProMISe found no mortality benefit from protocolised early goal-directed therapy compared with contemporary usual care.15, 16, 17 This is not merely a story of a wrong idea. It is a story of time, context and changing usual care. A physiological package may be useful in one era and redundant in another.
Even fluid boluses, perhaps the most primitive reflex in acute care, cannot claim exemption. Shock suggests impaired perfusion; fluid should increase preload, stroke volume and cardiac output. Yet in FEAST, fluid boluses increased 48-hour mortality among African children with severe febrile illness and impaired perfusion.18 If even “give fluid to shock” can be context-dependent and harmful, then no physiological reflex deserves immunity from trial evidence.
Not Anti-Physiology, but Pro-Patient
The lesson is not that physiology is useless. That would be a childish conclusion. Without physiology we would have no hypotheses, no discipline, no explanatory framework, no way of deciding what to test next. Physiology is the map that allows us to begin.
But the map is not the voyage.
Physiology tells us what might happen if we pull one thread. A clinical trial tells us what happens when the whole tapestry moves. The patient is not a pathway. The patient is not a mucus plug, a glucose concentration, an oncotic gradient, a PaO2:FIO2 ratio, a urine output, or a cardiac output curve. The patient is a whole biological system embedded in a clinical system, and interventions ripple through both.
That is why randomisation matters. It does not deny mechanism. It disciplines it. It balances known and unknown prognostic factors. It protects us from choosing the patients most likely to do well and then crediting the intervention. It captures intended and unintended consequences. It allows the intervention to be judged not by the charm of its rationale, but by its net effect on outcomes that matter.
MARCH is the rope around the mast
Return, then, to the mast: the place where the song is heard clearly, but no longer allowed to steer the ship.
The wisdom of the story is not that Odysseus denied the Sirens’ beauty. It is that he planned for it. He knew that recognising the danger would not be enough once the song began, so he created a discipline stronger than desire: wax in the sailors’ ears, rope around the mast, and an instruction not to trust his future self when the music took hold.
MARCH is that rope around the mast. It allowed the physiological song to be heard without being obeyed. Carbocisteine and hypertonic saline were plausible enough to deserve a trial; randomisation determined whether that plausibility translated into benefit for patients. It did not: neither intervention shortened mechanical ventilation, and each was associated with harm.3
That is not a failure of science. That is science working.
The Sirens will always sing in critical care. They should. We need their song to generate ideas. But we should not steer by it. We should listen, take notes, build hypotheses, design trials, and then lash ourselves to the mast of patient-centred evidence before we change practice.
MARCH is not the wreckage on the rocks. MARCH is the ship passing safely by.
Physiology may tell us why an intervention should work; randomisation tells us whether it does.