Lactoferrin test formats, units and published cutoffs

A stool assay has to survive its own specimen. Feces carries bacterial proteases, the sample waits at ambient temperature between collection and the laboratory, and most proteins do not come through that intact. The reason this test exists at all is a 1996 Osaka study that set out to find which neutrophil protein does.

Sugi and colleagues measured four of them by ELISA in 196 stool samples from 75 patients with inflammatory bowel disease, alongside 25 control subjects: lactoferrin, neutrophil elastase, myeloperoxidase and lysozyme. All four rose in active disease. The authors then stimulated blood neutrophils with phorbol myristate acetate and with latex beads to see which protein the cells actually let go of, and stored homogenized stool at 4, 25 and 37 degrees C to see which one lasted. Lactoferrin was released most readily and held up best in feces. Their abstract gives no figures for either experiment, so the result is a ranking rather than a measurement, and that ranking is why the protein was taken forward instead of the other three.

Thirty years later the thresholds for that assay still do not transfer between laboratories. The reason is visible in the validation literature itself.

What the published studies actually used

One row in that table carries a threshold with its sensitivity and specificity attached. The other five either declined to set one, fitted one to their own patients, or returned no number at all.

The units are not the same unit

Kane and colleagues report micrograms of lactoferrin per gram of stool. Walker and colleagues report micrograms per milliliter. Neither abstract states a conversion between the two, and a stool extract's dilution is a property of the laboratory protocol rather than of the patient, so the two sets of figures cannot be placed on one axis.

The consequence shows up as soon as the numbers are read side by side. Kane's healthy controls averaged 1.45 µg/g and Walker's averaged 1.17 µg/mL, which look like the same finding. Kane's ulcerative colitis group averaged 1,125 µg/g and Walker's averaged 1,880 µg/mL. Borkowska's optimal threshold in children sits at 13 µg/g, two orders of magnitude below either disease mean and one order above both control means. A figure quoted from this literature without its unit, its assay and the population it was fitted to is not a result that can be checked.

The standard errors deserve the same attention. Kane reported 440 ± 128 µg/g in Crohn's disease and 1,125 ± 498 in ulcerative colitis; Walker reported 1,701 ± 382 µg/mL and 1,880 ± 565. Those are standard errors on group means, and they are large enough that the distance between an individual result and a group average says less than the distance between the diseased and undiseased groups does.

What a cutoff is fitted to

A threshold is only ever optimal for the comparison it was drawn against, and the Borkowska study makes that concrete inside a single cohort. Its 13 µg/g cutoff separates children with inflammatory bowel disease from children without it at 80.7% sensitivity and 92.7% specificity. In the same 52 patients, concentrations tracked clinical severity grade, with moderate Crohn's disease higher than mild or inactive disease and moderate ulcerative colitis higher than mild. They found no significant relationship between lactoferrin concentration and the severity of the endoscopic lesions.

A threshold that separates disease from no disease therefore does a different job from one that grades how inflamed a known-inflamed bowel currently is. The pooled figures on the fecal lactoferrin hub reach that from a different direction: specificity against endoscopic activity in known disease falls to 0.79 while specificity against irritable bowel syndrome reaches 0.94 or better. Langhorst's approach was to tune the cutoffs to his own scoped cohort, which lifted accuracy for every marker he measured and is a demonstration of the problem rather than a solution to it.

Walker's cohort adds the one prospective signal in this set. Patients whose lactoferrin was measured and who then flared clinically within two months averaged 845 ± 452 µg/mL against 190 ± 90 in those who stayed in remission. The gap is real at P = 0.003 and the standard error on the higher group is more than half its mean, so this supports a population statement about risk and not much about any one specimen.

Quantitative against qualitative

The formats answer different questions. A quantitative ELISA returns a concentration that can be compared against a previous result from the same laboratory. A rapid lateral-flow test returns a band, read by eye, placing the sample above or below a fixed threshold built into the device.

Puolanne and colleagues in Helsinki validated one rapid lactoferrin test and two rapid calprotectin tests against ELISA calprotectin in 72 patients with colonic inflammatory bowel disease, all of whom had an ileocolonoscopy, with clinical, endoscopic and histologic scoring. The rapid lactoferrin test correlated with those scores about as well as the better of the two rapid calprotectin tests. Their conclusion was about the format: semiquantitative and qualitative tests identify patients with inactive or mildly active disease, and cannot replace a quantitative calprotectin ELISA where the result will drive a treatment decision.

Two limits on reading that further. The study's comparator was calprotectin, so it measures how a rapid lactoferrin device performs beside the established rapid alternative rather than against a lactoferrin ELISA. And 72 patients split across three devices leaves small groups, so the ranking between the devices rests on less than the correlation with endoscopy does.

What is still missing

Across these six studies no two share a combination of assay, unit, population and reference standard. There is no common reference material, the extraction step is left to the laboratory, and the one ROC-derived threshold published with its full performance figures came from 52 children in Gdansk. A clinician comparing a result against a published cutoff is comparing it against whichever study happened to use a similar platform.

That is a reporting problem in the literature rather than a failure of the molecule. The biology holds up well: lactoferrin survives the specimen, which the biology of lactoferrin and the granule chemistry behind it explain, and the protein being measured is the same one described in the entry on what lactoferrin is. Where lactoferrin is being measured in stool, the platform and the threshold belong beside the number.

This page describes assays and the thresholds published for them. It does not interpret a result. A measured value belongs to the clinician who ordered the test and who holds the history it has to be read against.