What a calprotectin test actually measures

Somewhere between the first appointment and the colonoscopy there is usually a small pot, a stool sample, and a number that comes back a week later. That number decides a good deal: how urgent the referral is, whether a camera goes in, and whether the working diagnosis is inflammatory bowel disease or irritable bowel syndrome. It is worth knowing what it can and cannot tell the person reading it.

What the number is

Calprotectin is a protein carried by neutrophils, the white cells that arrive wherever there is inflammation. When the lining of the bowel is inflamed those cells turn up in the bowel wall, and some of what they carry ends up in stool. Measuring it is therefore an indirect way of asking whether there are neutrophils in the gut — which is a different question from asking whether someone has symptoms. Irritable bowel syndrome can produce severe symptoms with no neutrophils behind them, and that gap is exactly the gap this test is built to exploit.

Better at ruling out than ruling in

A systematic review by Freeman and colleagues in BMJ Open pooled the accuracy studies at the commonly used threshold of 50 micrograms per gram. Comparing inflammatory bowel disease against irritable bowel syndrome, it found a sensitivity of 0.97, with a confidence interval of 0.91 to 0.99, and a specificity of 0.76, with an interval of 0.66 to 0.84. Against a broader group without inflammatory bowel disease, the figures were 0.95 for sensitivity and 0.67 for specificity.

Those two numbers do different jobs, and the asymmetry between them is the point. High sensitivity means very few people who have the disease are missed, so a low result is genuinely reassuring. The lower specificity means a raised result is much weaker evidence in the other direction: plenty of people above the threshold turn out not to have the disease at all. A negative test closes a door. A positive one only opens one.

Moving the threshold trades one property for the other. The same review reports that raising the cut-off from 50 to 100 reduces sensitivity and increases specificity across every clinical question it examined — fewer false alarms, at the cost of missing more.

Two reviews, two different answers

The earlier Health Technology Assessment review by Waugh and colleagues, which is the work behind the test being used this way in the United Kingdom, put the pooled figures at the same threshold of 50 at 93 per cent sensitivity and 94 per cent specificity in adults separating inflammatory bowel disease from irritable bowel syndrome. That specificity is far higher than the 0.76 the later review arrived at. Both are published pooled estimates, and they do not agree. Neither has been averaged into the other here, because the average of two disagreeing estimates is a third figure nobody measured.

The borderline band

The Health Technology Assessment review names the awkward zone explicitly: results between 50 and 150, most of which belong to people who do not have inflammatory bowel disease. The later review also found accuracy lower in primary care than in specialist clinics. A borderline number is neither a diagnosis nor a clean negative, and what happens next — repeat it, or investigate further — is a decision for the team holding the result.

Other things lift it

Because the test measures neutrophils rather than a diagnosis, anything that brings neutrophils to the bowel can raise it. The Freeman review names several:

• diverticular disease
• coeliac disease
• rectal adenocarcinoma
• non-specific inflammation

That is less a flaw in the test than a description of what it is: a sensitive detector of inflammation in the bowel that says nothing about what caused the inflammation. Naming the cause is what the colonoscopy is for, which is why a raised result leads to one rather than replacing it.