The children given 20 mL/kg died more often

In 2011, a report came out of six sites across Kenya, Uganda and Tanzania. It was a trial randomising 3141 children aged 2 months to 12 years with severe febrile illness accompanied by impaired consciousness or respiratory distress and impaired perfusion, to either a 20 mL/kg bolus (albumin or saline) or no bolus (maintenance fluids only).

48-hour mortality was 10.6% with the albumin bolus, 10.5% with the saline bolus, and 7.3% with no bolus. **More children died in the groups that received fluid** (absolute difference 3.3%, RR approximately 1.45, P=0.003).

Give rapid fluid to a child in shock. That was what resuscitation guidelines worldwide said at the time. FEAST broke that conventional wisdom — in reverse.

Fluid is something in short supply, and adding it saves lives — that was the belief for ten years. This piece traces how that decade began, and how it was folded up.

When you order fluids at the bedside, you are actually handling three questions at once: what to give, how much to give, and when to stop. Each of these three has followed its own twenty-year path.

Act I: Adding, because there wasn't enough

In 2001, Rivers and colleagues proposed a resuscitation protocol for the first six hours after arrival in the emergency department. Targeting central venous pressure, mean arterial pressure, and central venous oxygen saturation (ScvO2), it staged fluids, vasopressors, inotropes and transfusion accordingly. This was early goal-directed therapy (EGDT).

In a single-centre randomised trial of 263 patients, in-hospital mortality fell from 46.5% to 30.5% (P=0.009, absolute difference 16%). A gap of this size is rarely seen in critical care.

This one trial defined sepsis management for the following decade. "Early, aggressive, goal-directed." Fluid was something in short supply, and adding it saved lives — that era began.

The collapse of Act I: three countries arrived at the same answer

In 2014, ProCESS tested the same question across multiple centres. Three groups: EGDT, protocolised standard therapy, and usual care. 60-day mortality was 21.0% vs 18.2% vs 18.9% — no difference.

The same year, ARISE in Australia and New Zealand replicated the study across 51 sites and 1600 patients. 90-day all-cause mortality was 18.6% vs 18.8%, absolute difference −0.3 points (95% CI −4.1 to 3.6, P=0.90). What matters here is that the EGDT group was, in fact, treated properly and aggressively. Fluid volume in the first six hours was 1964±1415 vs 1713±1401 mL, vasopressors 66.6% vs 57.8%, red cell transfusion 13.6% vs 7.0%, dobutamine 15.4% vs 2.6% — all significantly higher, P<0.001. The protocol worked as intended, and mortality still didn't move.

The following year, 2015, the UK's ProMISe closed out the third trial. Across 56 hospitals and 1260 patients, 90-day all-cause mortality was 29.5% vs 29.2% (RR 1.01, 95% CI 0.85 to 1.20, P=0.90). This trial went further still: the EGDT group had significantly worse organ failure scores, and spent more days on advanced circulatory support and longer in the ICU. A cost-effectiveness analysis was built in as well, finding that the probability EGDT was cost-effective was under 20%.

Three countries, three trials, the same answer. EGDT bundles requiring a central venous catheter and an ScvO2 target fell out of the standard.

But a caveat is needed here. Over 13 years, the control group itself had changed. Mortality in usual care in Rivers' era was 46.5%; in ProCESS's usual care, it was 18.9%. It's probably more accurate to read this not as EGDT being refuted, but as the thinking EGDT popularised — recognise early, start antibiotics early, start fluids early — having spread across the board, erasing any incremental benefit.

And so "in what sequence do you add" reached a dead end. What was asked next was what to add.

Act II: What to give — the defeat of colloids

Fluid that stays in the vasculature is more efficient. Colloid osmotic pressure holds water inside the vessels. The logic is clean, and that's exactly why it was used for so long. It took twenty years and six trials for this logic to collapse.

Albumin was tested first. SAFE (2004, 6997 adult ICU patients) compared 4% albumin against saline. 28-day mortality was 20.9% vs 21.1% (RR 0.99, P=0.87). No difference in organ dysfunction, ICU stay, mechanical ventilation, or dialysis. In subgroup analysis, though, albumin was harmful in traumatic brain injury, and there was a trend toward benefit in sepsis.

ALBIOS (2014, 100 ICUs, 1818 patients) tested that sepsis subgroup specifically. 20% albumin was added to crystalloid, maintaining serum albumin at 30 g/L or above. Physiologically, it hit its mark — mean arterial pressure was significantly higher over the first seven days (P=0.03), and net fluid balance was significantly lower (P<0.001). Even so, 28-day mortality was 31.8% vs 32.0% (RR 1.00, P=0.94), and 90-day mortality 41.1% vs 43.6% (RR 0.94, P=0.29) — unchanged. Physiological improvement not translating into outcome is a pattern seen again and again in critical care.

Synthetic colloids (hydroxyethyl starch, HES) didn't get off with mere equivalence. The first to show harm was VISEP (2008). Comparing 10% pentastarch (HES 200/0.5) against Ringer's lactate in severe sepsis, the trial was stopped early on safety grounds. HES increased acute kidney failure and renal replacement therapy, and toxicity rose with cumulative dose. Not simple inefficacy — dose-dependent harm.

Four years later, two large-scale trials confirmed it. In 6S (2012, severe sepsis, 798 patients), 90-day mortality in the HES 130/0.42 group was 51% vs 43% (RR 1.17, P=0.03), renal replacement therapy also rose, 22% vs 16% (P=0.04), and there was more bleeding. The same year's CHEST tested 6% HES 130/0.4 in a broader ICU population of 7000 patients; 90-day mortality was 18.0% vs 17.0% (RR 1.06, P=0.26), no difference, but renal replacement therapy was significantly higher with HES, 7.0% vs 5.8% (RR 1.21, P=0.04).

One trial asked the question of colloids as a group. CRISTAL (2013, 57 ICUs, 2857 patients) compared a colloid group — including gelatin, dextran, HES and albumin — against crystalloid. The primary outcome, 28-day mortality, showed no difference: 25.4% vs 27.0% (RR 0.96, P=0.26). 90-day mortality favoured colloids, 30.7% vs 34.2% (RR 0.92, 95% CI 0.86 to 0.99, P=0.03), but the authors themselves frame this as an exploratory finding. Because the colloid arm spanned multiple agents, no per-agent evaluation is even possible.

Lay the six trials side by side and the outline emerges. Albumin, at best, equivalent. HES, harmful. Colloids as a class — nothing that won on a primary outcome. Nothing gained, a burden placed on the kidneys — that was the twenty-year verdict.