Glasgow-Imrie Criteria

Formula

PANCREAS

P = PaO₂ < 60 mmHg; A = Age > 55; N = Neutrophils (WCC) > 15 ×10⁹/L; C = Calcium (corrected) < 2.0 mmol/L; R = Renal, urea > 16 mmol/L; E = Enzymes, LDH > 600 IU/L; A = Albumin < 32 g/L; S = Sugar, glucose > 10 mmol/L.

Threshold of 3

Derived, not arbitrary: in the Glasgow validation series, 31% of the 131 episodes with three or more criteria were severe against 8% of the 274 with fewer.

• Score the worst value for each variable across the first 48 hours, not the admission value. Scoring the admission panel only will under-score, because several of these variables peak on day two.
• There is one criteria set for all aetiologies. Unlike Ranson's criteria, the thresholds do not change for gallstone pancreatitis.
• Calcium must be albumin-corrected. Failing to correct it in a patient with a low albumin counts the same physiological abnormality twice.
• The original 1984 system had nine factors; one did not predict severity in prospective testing and was removed. The eight-factor version is the one to use, and it is what 'modified Glasgow' refers to.
• AST above 200 IU/L substitutes for LDH in some published versions of the criteria.
• The glucose criterion should be disregarded as prognostically meaningful in known diabetes, though the score offers no formal adjustment.

Interpreting the result

Three or more criteria predicts severe disease and should change where the patient is managed. But read the validation figures carefully, because they are frequently mis-stated: in the Glasgow series 31% of episodes with three or more criteria were severe, not 90-something percent. That is a positive predictive value of roughly one in three. The score earns its place on the other side of the threshold — 8% of episodes with fewer than three criteria were severe, so a low score is the more informative result, and it is the number to lean on when deciding a patient does not need high-dependency care. Treat two criteria as a caution rather than as a negative result; the threshold is a convention drawn through a continuous gradient of risk, and a patient with two criteria and a rising urea is not the same as a patient with two criteria who is improving. Finally, this is a 48-hour statement about physiological severity, not about anatomy. A patient can meet one criterion and still be developing extensive necrosis, which is why deterioration prompts a contrast CT at around 72 hours regardless of what the score said.

What the Glasgow-Imrie Criteria needs (8 inputs)

P — PaO₂ below 60 mmHg (8 kPa)

On room air. Hypoxaemia here reflects early acute lung injury from circulating inflammatory mediators rather than aspiration or fluid overload. Some published versions state 7.9 kPa, which is the same threshold after unit conversion rather than a different criterion — and this is the one variable requiring an arterial sample, which is why it is occasionally skipped and the score reported out of seven.

A — Age over 55 years

The same threshold as the non-gallstone Ranson criteria. Unlike Ranson, Glasgow-Imrie does not raise it for gallstone disease — there is a single criteria set here regardless of aetiology, which is a genuine simplification.

N — White cell count above 15 ×10⁹/L

The 'N' stands for neutrophils in the mnemonic, but the criterion is the total white cell count. A marker of the systemic inflammatory response, not of infection — this is not a reason to start antibiotics in the first 48 hours.

C — Corrected calcium below 2.0 mmol/L (8 mg/dL)

Corrected for albumin, which matters here more than usual because low albumin is itself one of the eight criteria, so an uncorrected calcium in a hypoalbuminaemic patient will double-count the same abnormality. The hypocalcaemia reflects calcium being consumed in the saponification of necrotic retroperitoneal fat.

R — Urea above 16 mmol/L (BUN above approximately 44.8 mg/dL)

Despite adequate fluid resuscitation. Renal impairment is among the strongest single predictors of a bad outcome in acute pancreatitis, and unlike Ranson's version this criterion is an absolute value rather than a rise, which makes it far easier to score.

E — LDH above 600 IU/L

Reflecting tissue necrosis. Some versions of the criteria accept AST above 200 IU/L in place of LDH, which is a documented variant rather than an error — worth noting if you are comparing a score across two hospitals with different panels.

A — Albumin below 32 g/L (3.2 g/dL)

A criterion Ranson's set does not contain. It captures capillary leak and third-space loss, doing something similar to Ranson's fluid-sequestration criterion but from a single blood result rather than 48 hours of fluid charting.

S — Glucose above 10 mmol/L (180 mg/dL)

In the absence of a known diabetes diagnosis. Reflects loss of islet function in an inflamed gland, so it carries no prognostic information in a patient whose glucose was already high before the admission.

Units. The criteria are conventionally written in SI units because of their British origin, which is the mirror image of Ranson's US conventional units and the usual source of confusion when the two are compared. Conversions: calcium 2.0 mmol/L is 8.0 mg/dL; glucose 10 mmol/L is 180 mg/dL; albumin 32 g/L is 3.2 g/dL; PaO₂ 8 kPa is 60 mmHg, and the 7.9 kPa quoted in some versions is the same threshold rather than a different one. Urea and BUN are distinct measurements — urea 16 mmol/L corresponds to a BUN of approximately 44.8 mg/dL, since BUN in mg/dL is roughly urea in mmol/L multiplied by 2.8. The calcium must be albumin-corrected before comparison with the threshold. LDH and AST cut-offs are assay-dependent and should be read against your local reference range rather than assumed to transfer from the 1984 paper.

What it returns

Number of criteria met (0–8)

Unweighted — each of the eight criteria counts one point regardless of how strongly it predicts on its own.

Predicted mild or predicted severe

A binary output at a threshold of three, which is how the score was validated and how the guidelines use it. Two criteria sits close enough to the threshold to deserve treating as a caution rather than as reassurance.

How it is calculated

The Glasgow criteria are what happens when a score is pruned by prospective testing rather than assembled by expert opinion. Imrie's group in Glasgow had been running trials in acute pancreatitis and needed a way to stratify entrants; they began from a nine-factor system, then assessed it prospectively across 405 episodes accumulated over seven years — deliberately excluding the patients whose data had built the system in the first place, which is why the 1984 paper reads as a validation rather than a derivation. Testing each factor individually showed that one of the nine carried no prognostic signal at all. Removing it improved overall classification from 72% to 79% of episodes correctly predicted, and the eight that remained are the criteria in use today. The mechanistic spread is deliberate: respiratory failure, age, inflammatory load, necrosis markers, renal function, capillary leak and endocrine failure of the gland, each represented once. Nothing about the arithmetic is clever — it is an unweighted count — and the score's durability comes from the selection of variables, not from the model.

Facts & figures

Evidence

How it compares

Pearls & pitfalls

• Use the worst value in the 48-hour window for each variable. Scoring the admission bloods alone is the commonest error and it under-scores, because urea, calcium and albumin typically move over day two.
• Correct the calcium for albumin. Low albumin is separately one of the eight criteria, so an uncorrected calcium scores the same abnormality twice.
• Three or more criteria gave a 31% rate of severe disease in the validation series, not 90%. The score's real strength is the 8% rate below the threshold — it excludes better than it confirms.
• The 'N' in PANCREAS is neutrophils but the criterion is the total white cell count.
• There is only one criteria set. If you find yourself looking for gallstone-specific thresholds, you are thinking of Ranson.
• The urea criterion is an absolute value, not a rise despite fluids. This is a real difference from Ranson and it makes Glasgow much easier to score retrospectively.
• AST above 200 IU/L is an accepted substitute for LDH in some published versions — check which your local guideline uses before comparing scores between hospitals.
• Missing the arterial gas turns this into a seven-criterion score with an unknown threshold. Note the omission explicitly rather than reporting the count as if complete.
• In known diabetes the glucose criterion carries no prognostic weight, and the score has no adjustment for that.
• 'Modified Glasgow' means the eight-factor version. Nine-factor references are to the superseded original.

Critical actions

• Start treatment on the clinical picture, not on the score. Fluid resuscitation, analgesia and early enteral feeding belong in the first hours, well before any 48-hour score exists.
• Use BISAP or a SIRS assessment for the day-one decision about level of care, and let Glasgow-Imrie confirm or revise it at 48 hours.
• Take the arterial gas. It is the criterion most often skipped and it represents early lung injury, which is the organ failure that most often declares first.
• At three or more criteria, escalate to high-dependency or intensive care before organ dysfunction is established.
• Establish the aetiology in parallel — biliary ultrasound, triglycerides, calcium, a drug and alcohol history. The score says how sick, never why.
• In gallstone pancreatitis, proceed to urgent ERCP if there is coexisting cholangitis, and plan cholecystectomy on the same admission in mild disease to prevent recurrence.
• Image with contrast CT at around 72 hours, or sooner if the patient deteriorates, and grade necrosis with CTSI. Earlier imaging under-stages it.
• Do not drain sterile necrosis on imaging appearance alone — intervention follows clinical deterioration, and a delayed, minimally invasive approach outperforms early necrosectomy.

Why this score exists

The Glasgow criteria exist because of a clinical trial, not a prognostic ambition. Imrie's unit was testing Trasylol in acute pancreatitis in the mid-1970s, and a trial in a disease whose severity varies this widely is uninterpretable without some way of showing the arms were comparable — so a prognostic factor system was built to do that job. The 1984 paper is explicit about the three purposes the authors saw for such a system, and they are worth reading in order: alerting the clinician to potentially severe disease, allowing severity to be compared within and between patient series, and permitting rational selection of patients for trials of new treatments. Two of those three are research purposes. That framing explains why the score is a binary classifier at a threshold rather than a graded mortality estimate — it was designed to sort patients into comparable groups, and the bedside triage use grew on top of a tool built for a different question. It also explains the intellectual honesty of deleting a factor that did not work: a research instrument has to justify each variable, where a clinical mnemonic tends to accumulate them.

About the creator

• Clement W. Imrie

  Senior author of the prognostic factor system; Royal Infirmary, Glasgow

  Developed the original Glasgow prognostic factor system during trial work in acute pancreatitis, and co-authored its prospective validation.

• Scott L. Blamey

  First author, Gut 1984 validation paper

  Led the prospective assessment across 405 episodes that reduced the system from nine factors to eight.

• David C. Carter

  Senior co-author, Gut 1984

  Co-authored the validation study and much of the associated Glasgow work on acute pancreatitis.

Limitations

• Positive predictive value is modest — only 31% of episodes with three or more criteria were severe in the validation series — so a positive result raises concern rather than establishing severity.
• Not complete until 48 hours, so it cannot inform the admission decision about level of care.
• Requires an arterial blood gas, the one component not on a routine panel, and the criterion most often quietly omitted.
• Unweighted, so age over 55 and a urea above 16 mmol/L count equally despite carrying very different prognostic force.
• Derived and validated in a single Scottish centre in a population of the late 1970s and early 1980s, with the aetiology mix and baseline outcomes of that setting.
• The glucose criterion is confounded by pre-existing diabetes and the albumin criterion by chronic liver disease or malnutrition, with no adjustment for either.
• LDH thresholds are assay-dependent, and the accepted AST substitution introduces variation between hospitals using different versions.
• Predicts physiological severity only — silent on aetiology, on necrosis, and on whether intervention will be needed.
• Not validated for serial re-scoring, in children, or in pregnancy.
• Shares the ceiling of every score in this field: the 2012 nine-score comparison concluded these systems have reached their maximal efficacy and that better prediction will require a different approach rather than a refined checklist.

If you are the patient

The Glasgow-Imrie criteria are a list of eight things doctors check in the first two days after someone is admitted with pancreatitis, to work out how serious the attack is likely to be. Most come from ordinary blood tests — the amount of calcium, protein, sugar, urea and certain enzymes in your blood, and the number of white cells — plus your age and a test of the oxygen level in your blood taken from an artery in the wrist. Each one that is abnormal counts as one point. Doctors wait the full two days because pancreatitis often reveals how severe it will be over time rather than immediately, and they use the worst reading from that period. Fewer than three points suggests a milder attack that can usually be looked after on a normal ward. Three or more means you will be monitored much more closely, often in a high-dependency or intensive care unit, and the team will look sooner for complications with a CT scan. It is important to understand what a higher score does and does not mean: most people with three or more points still recover. The score is a signal to watch you carefully, not a prediction that things will go badly. You may hear it called the PANCREAS score, because the first letters of the eight items spell out the word.

Frequently asked questions

Related calculators

• HAPS — Harmless acute pancreatitis score
• Tokyo Guidelines — Cholangitis — TG18 diagnosis and severity grade for acute cholangitis
• Ranson's Criteria — Acute pancreatitis severity at 48 hours
• BISAP Score — Bedside index for severity of pancreatitis
• CTSI — CT severity index — pancreatitis
• Revised Atlanta Classification — Acute pancreatitis severity — mild, moderately severe, severe

References