Iron-deficiency anaemia: a finding, not a diagnosis

Iron is what haemoglobin uses to carry oxygen. Run short of it and the bone marrow makes red cells that are too few and too small, and the body runs short of oxygen delivery: fatigue, breathlessness on exertion, pallor, and a set of odder signs, including hair loss ^[1] and pica, the craving to chew ice or other non-foods.

It is the commonest nutritional deficiency in the world. The WHO estimated global anaemia prevalence at 24.8 percent, affecting 1.62 billion people across 1993–2005 — 47.4 percent of preschool children, 41.8 percent of pregnant women and 30.2 percent of non-pregnant women ^[2], with the global burden tracked over time in a separate systematic analysis ^[3]. Iron deficiency is the largest single contributor. Even in a wealthy country it is common: NHANES III found 9 percent of toddlers and 9–11 percent of adolescent girls and women of childbearing age iron deficient ^[4], and anaemia prevalence rises steeply after age 50, exceeding 20 percent past 85 ^[5].

And here is the principle the whole review turns on: iron-deficiency anaemia is a finding, not a diagnosis. Iron is not lost in any quantity except through bleeding. So a person who is iron deficient is either losing blood, failing to absorb iron, or has demand outstripping intake — and identifying which is not an optional extra after treatment. It is the treatment. Replacing iron without finding the cause corrects the number and leaves the reason in place, which in a subset of patients means leaving a cancer in place.

Start here: iron is conserved, not excreted

The body has no regulated route for getting rid of iron. Losses are tiny — shed cells, a little sweat — so total body iron is set almost entirely by how much is absorbed, and absorption is what the body controls ^[6].

The controller is hepcidin, a peptide hormone made in the liver. Hepcidin binds ferroportin — the only known cellular iron exporter ^{[7] [8]} — and causes it to be internalised and degraded ^[9]. Ferroportin is what lets iron out of the gut lining into the blood and out of macrophages recycling old red cells, so degrading it does two things at once: it blocks intestinal absorption and it locks iron inside the storage cells ^{[10] [11]}.

Hepcidin is regulated exactly as a controller should be: up by iron loading, down by anaemia and hypoxia ^{[12] [13]}, with the feedback loop closed by iron stores and plasma iron on one side and erythroid demand on the other ^[6]. Mice lacking hepcidin develop severe tissue iron overload ^[14]; mice overexpressing it in the liver are born pale and severely microcytic and anaemic ^[15]. Two mouse genotypes, opposite phenotypes, one hormone.

And it is also an acute-phase protein — induced by inflammation, principally through interleukin-6 ^{[16] [17] [18]}. That single fact generates the most common diagnostic confusion in this field, discussed below.

Iron in the blood travels on transferrin and is taken up through transferrin receptor 1 ^{[19] [20]}; what is not used is stored in ferritin, an iron-storage protein whose molecular properties and regulation are well characterised ^[21], and released from it by a dedicated autophagy route ^[22].

Pillar 1: measurement and diagnosis

The blood count first

Iron deficiency produces a microcytic anaemia — low haemoglobin with a low mean cell volume — and red cell indices give a sensitive indication of iron deficiency in the absence of chronic disease or haemoglobinopathy ^[23]. That caveat matters: thalassaemia trait also produces microcytosis and is common in some populations, so haemoglobin electrophoresis is recommended when microcytosis and hypochromia appear in patients of appropriate ethnic background, specifically to prevent unnecessary gastrointestinal investigation ^{[23] [24]}.

Ferritin is the test

Serum ferritin is the most powerful test for iron deficiency ^[23]. A low ferritin means depleted stores and essentially nothing else does.

Its one great weakness is inflammation. Ferritin is an acute-phase reactant, so it rises with inflammatory disease, and it correlates with both disease activity and body iron stores at once ^[25]. A normal or high ferritin therefore does not exclude iron deficiency in someone with infection, cancer, autoimmune disease, chronic kidney disease or heart failure — which is precisely the population in which the question arises most often. This is where anaemia of inflammation overlaps: inflammation raises hepcidin, hepcidin blocks absorption and traps iron in macrophages, and the result is iron-restricted erythropoiesis with iron-replete stores ^{[26] [27]}. Distinguishing that from true iron deficiency, or from the two coexisting, is the hard case.

Transferrin saturation and total iron-binding capacity add to the picture, and soluble transferrin receptor is the marker that is not confounded by inflammation, which is why it earns its place in exactly the situations where ferritin fails ^[28].

The order in which things fail

The three markers do not move together, and the sequence is not a textbook convention — it was established experimentally.

Fourteen volunteers were bled repeatedly to take them stepwise into iron deficiency, with serum iron, TIBC, MCV, erythrocyte protoporphyrin, ferritin and transferrin receptor measured serially throughout. Transferrin receptor did not change at all during the storage-depletion phase. It began rising only once ferritin had already reached subnormal values, and continued rising thereafter. The authors' own summary is the staging model: ferritin measures iron stores, transferrin receptor measures mild tissue deficiency, and haemoglobin measures advanced deficiency ^[28].

The check the arithmetic was never given. NHANES III reports its findings twice over — once as percentages and once as absolute numbers of people — and never divides either pair. Both divisions answer the same question: what share of iron deficiency has progressed to anaemia? From the percentages, 3 percent of toddlers with anaemia against 9 percent iron deficient gives 33.3 percent. From the headcounts, approximately 240,000 against approximately 700,000 gives 34.3 percent. Two independent routes through one paper's own reporting, agreeing to within one percentage point. For women of childbearing age the counts give 3.3 million of 7.8 million, or 42.3 percent, which sits inside the range the reported percentages allow ^[4].

The consequence is the clinical point. In both groups the majority of iron deficiency has not yet produced anaemia. A haemoglobin alone therefore misses roughly two-thirds of iron deficiency in toddlers and three-fifths in women — which is the argument for measuring ferritin rather than waiting for the blood count to fall, in anyone with a reason to be deficient.

A second check, on a different paper. WHO reports a global anaemia prevalence and an absolute number affected in separate sentences. Dividing 1.62 billion by 24.8 percent recovers 6.5 billion — the world population at the midpoint of the 1993–2005 data window ^[2]. Neither figure is presented as a population estimate; they are internally consistent anyway.

Three honest limits. NHANES defined iron deficiency as two abnormal results out of three tests, so the denominator here is that operational definition rather than a gold standard. The percentage ranges for women are wide enough that the count-derived 42.3 percent is confirmed only loosely. And the right-hand panel shows the order and direction the phlebotomy experiment established, not magnitudes — the substrate does not carry the numeric curves, and the panel says so.

One further gap: the figure the brief preferred was the hepcidin dosing one — fractional absorption against dose and dosing frequency. That study is not in this substrate; the crawl returned hepcidin biology densely but no absorption-versus-schedule dataset. The alternate-day dosing insight is discussed below on mechanism, and this review quotes no absorption percentages for it.

Then find the cause

This is the part that gets skipped. The British Society of Gastroenterology guideline is unusually direct about it: iron-deficiency anaemia occurs in 2–5 percent of adult men and postmenopausal women in the developed world and is a common reason for gastroenterology referral, with gastrointestinal blood loss from colonic or gastric cancer, and malabsorption in coeliac disease, the most important causes that need to be sought ^[23].

Its specific recommendations are worth stating because they are frequently not followed. Upper and lower GI investigation should be considered in all postmenopausal women and all men with confirmed iron-deficiency anaemia, unless there is a history of significant overt non-GI blood loss. All patients should be screened for coeliac disease. Any level of anaemia should be investigated in the presence of iron deficiency, and the lower the haemoglobin, the more likely serious underlying pathology and the more urgent the investigation. Notably, lower GI investigation should still be considered even if coeliac disease is found — one cause does not exclude another ^[23].

How often is coeliac disease the answer? A prospective study endoscoped and colonoscoped 105 consecutive patients referred with iron-deficiency anaemia — defined as ferritin below 25 ng/mL with haemoglobin below 12 g/dL — taking duodenal biopsies in everyone rather than only in selected patients, against a background in which occult coeliac disease had been reported in 0 to 6 percent of such adults ^[29]. Testing everybody rather than the ones who look coeliac is what makes a prevalence figure mean anything.

For occult bleeding, faecal immunochemical testing for haemoglobin has largely replaced guaiac-based tests, being specific for human haemoglobin ^[30], with standardised reporting units ^[31].

The other major causes need no investigation to find, only asking: heavy menstrual bleeding and pregnancy, in which demand rises sharply. Iron deficiency is also near-universal in some chronic diseases — half of a large pooled heart-failure cohort met the definition ^[32], and it is common in inflammatory bowel disease ^[33]. Obesity has emerged as a risk factor in its own right, through inflammation-associated impairment of duodenal iron absorption ^[34].

Pillar 2: treatment

The treatment is two things, and the first is not iron: treat the cause, and replace the iron.

Oral iron, and a mechanism-based dosing lesson

Ferrous salts are the standard, cheap and effective. The traditional prescription is a ferrous salt two or three times a day ^[35].

That schedule is probably wrong, and hepcidin explains why. Each dose of iron raises serum hepcidin; hepcidin degrades ferroportin; and ferroportin is the door through which the next dose would have to pass. A second dose given while hepcidin is still elevated is therefore absorbed less well than the first, and the unabsorbed iron stays in the gut lumen, and gastrointestinal side effects are the usual reason people stop taking it ^{[36] [37]}. Spacing doses out — a single daily dose, or alternate-day dosing — lets hepcidin fall back before the next dose arrives. The consequence is counterintuitive and practically important: taking iron less often can absorb as much or more of it, with fewer side effects.

Three grounded strands support that. The mechanism is directly demonstrated — synthetic hepcidin causes rapid, dose-dependent hypoferraemia ^[38], so raising hepcidin really does drop circulating iron rather than merely correlating with it. Reviews of iron biology in the hepcidin era note explicitly that this understanding has an impact on the established schedules of oral iron treatment and on the choice between oral and intravenous iron ^{[11] [10] [6]}. And the clinical end of it has been seen in the setting where iron is prescribed most: a Cochrane review of treatments for iron-deficiency anaemia in pregnancy concluded that daily low-dose iron supplements may be effective with fewer gastrointestinal side effects than higher doses ^[36].

As flagged above, the quantitative absorption study behind this is not in this substrate, so no absorption percentages are quoted. The mechanism is grounded; the effect size is not stated here.

Intravenous iron

Intravenous iron bypasses the gut entirely, and therefore bypasses hepcidin's control of absorption. It is used for intolerance of oral iron, malabsorption, continuing losses that outrun oral replacement, and when repletion needs to be fast. Modern preparations allow large doses in very few visits: a randomised comparison gave iron isomaltoside 1,000 mg as a single day-0 infusion against ferric carboxymaltose 750 mg on days 0 and 7 ^[39]. Indications, advantages and side effects across oral and intravenous formulations have been reviewed together ^[37].

The strongest randomised evidence for intravenous iron comes from populations where the deficiency is inseparable from the underlying disease. In chronic heart failure, ferric carboxymaltose has been tested against placebo in patients with iron deficiency defined as ferritin below 100 µg/L, or 100–300 with transferrin saturation below 20 percent ^{[40] [41] [42]}, and iron sucrose against exercise tolerance in anaemic and non-anaemic patients ^[43]. Note that last detail: benefit in patients who were not anaemic, which is the clearest demonstration that iron deficiency matters before it reaches the blood count — the same point the figure makes epidemiologically. It is not the only such demonstration: iron supplementation was tested against placebo in a double-blind randomised trial for unexplained fatigue in non-anaemic women ^[44].

In chronic kidney disease, intravenous iron optimises the response to erythropoiesis-stimulating agents ^[45], and the CKD anaemia story has its own hard-won lesson: correcting haemoglobin fully with epoetin did not help and in trials caused harm ^{[46] [47] [48]}. In inflammatory bowel disease, iron sucrose combined with erythropoietin has been used successfully ^[33].

Transfusion, and the response to expect

Transfusion treats the haemoglobin, not the deficiency, and the threshold for it has been settled by trial: a landmark randomised study of transfusion requirements in critical care established that a restrictive strategy is at least as good as a liberal one ^[49]. Transfusion is for severe or symptomatic anaemia, not for a number.

Two things about the expected response. A failure of haemoglobin to rise means one of three things — the diagnosis is wrong, the patient is not taking the iron, or the losses exceed replacement — each of which sends you back to the cause. And treatment must continue after the haemoglobin normalises, because normalising the blood count only refills the last compartment to empty; the stores that ferritin measures are still empty at that point, which is exactly what the staging figure shows.

Pillar 3: what is unresolved

Hepcidin-informed dosing and hepcidin-targeted drugs. The regulatory axis is now well enough understood to be a therapeutic target in both directions: agents that antagonise hepcidin to redistribute endogenous iron in anaemia of inflammation, and agents that mimic it in overload ^{[26] [11]}. A validated serum hepcidin immunoassay exists ^[50], but hepcidin is not yet a routine clinical measurement.

Diagnosing iron deficiency inside inflammation. Ferritin fails exactly where it is needed most ^{[25] [26]}. Soluble transferrin receptor is the principled answer ^[28] and is still not in routine use in many systems.

Rare genetic forms delineate the pathway: mutations in the hepcidin inhibitor TMPRSS6 cause iron-refractory iron-deficiency anaemia by upregulating hepcidin signalling — a natural experiment showing the axis is sufficient to cause the disease on its own ^[11].

Screening. Given that most iron deficiency has not yet become anaemia ^[4] and that WHO puts the global burden at over a billion people ^{[2] [3]}, who should be screened with ferritin rather than a blood count remains a live question, particularly in pregnancy and in at-risk groups.

And the persistent implementation failure, which is not scientific: the guideline says investigate men and postmenopausal women, and screen everyone for coeliac disease ^[23]. Iron replacement without that search is the commonest avoidable error in this condition.