Sepsis
A Socratic walk-through of sepsis — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why can a defence that handles most infections quietly end up killing the patient?
Your immune system deals with something most days and you never notice. Occasionally you get a sore throat, a hot swollen finger, a few days of fever. Then, in a small minority of cases, the same system takes a patient from mild illness to failing kidneys and death within hours.
The tempting reading is that the infection got worse — a nastier organism, a bigger dose. Hold that up to the light, though. Patients die of sepsis whose infection has been sterilised, and patients develop the identical syndrome with no infection at all. So what is doing the killing?
Reasoning it through
REASONING #Start with what inflammation is for, in the place it was designed to work: a square centimetre of tissue with bacteria in it. Local vessels dilate, so more blood arrives. Capillary walls become leaky, so plasma and antibody flood the tissue. Clotting is activated locally, laying fibrin that fences the invader in. Neutrophils squeeze out of the vessels to kill. Every one is exactly right, and together they give you a red, hot, swollen, painful finger that gets better.
Now perform the thought experiment that defines sepsis. Take the same four responses and apply each to the entire vasculature at once.
Dilate everything: the container the blood sits in suddenly holds more than the blood available, and pressure falls. Make all capillaries leaky: plasma leaves the circulation everywhere, so volume falls too, the patient swells, and fluid crossing into the lung's air spaces makes breathing fail. Activate clotting throughout: microscopic clots form in small vessels while the factors and platelets that made them are consumed, so the patient thromboses and bleeds at once. Recruit neutrophils everywhere: enzymes meant for a bacterium are released into the patient's own capillary linings.
Not one of those is a malfunction. Each is the correct local behaviour, delivered at the wrong scale — which already explains why sterilising the bacteria does not fix it.
The second half is why it runs away rather than settling. Low pressure and leaked plasma mean tissue does not get enough oxygen; cells starved of oxygen die; dying cells spill mitochondrial fragments and nuclear proteins, which the immune system reads as danger signals essentially indistinguishable from bacterial ones. So the injury caused by the response becomes a fresh trigger for it. Below some level of mediator release the loop damps out and the patient has a bad few days; above it the loop's gain exceeds one and sustains itself without further help from the pathogen. That is what makes sepsis a threshold phenomenon rather than a graded one, and why deterioration is so often described as sudden.
Now test the folk account — "sepsis is blood poisoning, bacteria multiplying in the blood". Blood cultures never grow an organism in a large share of septic patients; the fraction commonly quoted is roughly a third to a half (recalled, and it varies with how hard and how early the culturing is done). More decisively, severe pancreatitis, major burns and multiple trauma produce the same picture — the same hypotension, lung failure and organ dysfunction — with no organism anywhere. Bacteria are a trigger, not the agent of harm.
The analogy
THE ANALOGY #A building's fire suppression works because it acts locally: sprinklers in one room, that room's fire doors shut, that corridor vented. Trip every sprinkler on every floor at once and the fire is certainly out — along with the electrics, the servers, the lifts and the sealed stairwells that were meant to be the way out.
a flooded building can be pumped out and rewired, and its systems are independent, whereas the harm here creates the next round of the same signal — the water, in effect, sets off more alarms.
Clarifying the model
THE MODEL #Three refinements, then a serious problem with everything above.
The refinements first. Sepsis's definition was substantially rewritten in 2016 — from criteria based on inflammatory signs to criteria based on measured organ dysfunction. This is why I quote no incidence or mortality percentage: those figures are definition-dependent and not comparable across the change. I also decline the widely repeated claim that mortality rises by a fixed percentage for every hour antibiotics are delayed. That number comes from one retrospective cohort of patients already in septic shock, and later work has found much smaller effects, or none, in less severe sepsis; the direction is well supported, the slope is not.
The neighbouring piece Why the body runs a fever makes the complementary case — that a costly response can be the correct one, deliberately deployed. The fixed point of difference here is scale rather than cost: the same response, still correct, becomes lethal when the boundary containing it fails.
Now the failure in my own account. If sepsis were runaway inflammation, blocking the inflammatory mediators should save lives. It has not. Trials of anti-TNF antibodies, interleukin-1 receptor antagonists, anti-endotoxin agents and high-dose corticosteroids have run for decades and have overwhelmingly failed to improve survival. That is a substantial strike against the pure hyperinflammation story.
What covers the gap is that the two phases are simultaneous rather than sequential. Alongside the cytokine surge, septic patients show profound immune suppression — lymphocytes dying by apoptosis in large numbers, monocytes that no longer respond to a bacterial stimulus. Many deaths occur late, from second infections acquired in intensive care by a patient whose defences are exhausted. So blanket suppression fails because a large part of the problem was already suppression. The honest current model is not "too much immunity" but "immunity disordered in both directions at once", and the reason no single blocking drug works is that patients differ in which direction dominates and when.
A picture of it
THE PICTURE #How to readEach box is a condition the patient occupies at one time, and each arrow is labelled with what moves them. Two paths leave the initial infection: containment on the left, which ends the story, or spill into the circulation. From the systemic state the arrow back to containment is the recovery most patients take; the arrow down is not. The loop from shock to itself is the whole mechanism of the runaway — the damage feeds the signal that caused it — and the box below it is why the story does not end when the surge does.
What became clearer
WHAT CLEARED #The defence is not misfiring. It is doing precisely what it does in a swollen finger, to a body that has no room for it: dilation becomes hypotension, leak becomes oedema, fencing becomes disseminated clotting. And because the resulting damage generates the same danger signals that bacteria do, past a certain intensity the response no longer needs the infection to keep going — which is why killing the organism can arrive too late to matter.
The least comfortable turn is that the obvious remedy, turning the response down, does not work, because by then the patient is also profoundly immunosuppressed. Sepsis is not an immune system doing too much. It is one that has lost its sense of scale in both directions.
Where to go next
ONWARD #- Whether biomarkers can identify which septic patients are hyperinflamed and which are immunoparalysed, so treatment could be aimed rather than blanket.
- Why the endothelial glycocalyx — the gel layer lining blood vessels — is shed in sepsis, and what that does to capillary leak.
Key terms
TERMS #| Term | What it means |
|---|---|
| Sepsis | life-threatening organ dysfunction caused by a dysregulated host response to infection, as defined by the 2016 consensus. |
| DAMPs | damage-associated molecular patterns: molecules released by dying host cells that the immune system reads much as it reads bacterial signals. |
| Disseminated intravascular coagulation | widespread small-vessel clotting that consumes clotting factors and platelets, causing thrombosis and bleeding at once. |
| Immunoparalysis | the profound suppression of immune function seen alongside inflammation in sepsis, and a common route to late death from a second infection. |
Every term the collection defines is gathered in the glossary.