Atopic dermatitis — eczema — is a chronic, relapsing inflammatory skin disease whose defining symptom is itch. Not rash first and itch second: the itch is the disease, and the visible eczema is substantially what scratching does to skin that was already inflamed.
It is one of the commonest chronic diseases of childhood, mapped across dozens of countries by the International Study of Asthma and Allergies in Childhood [1] [2], with prevalence and incidence systematically reviewed [3] and time trends examined internationally [4]. It is not only a childhood disease — adult prevalence has been surveyed internationally [5] [6] — but many children do improve as they grow [7].
The burden is dominated by the itch and by what it does to sleep. One review of childhood atopic dermatitis is titled "the misery of living with childhood eczema" and reports a UK lifetime prevalence of 16–20 percent by age 20 [8]. A disease that wakes a child every night, and therefore the parents, is not adequately described as a rash.
And it usually comes first. Atopic dermatitis is the opening move of the atopic march — the tendency for eczema in infancy to be followed by food allergy, then asthma, then allergic rhinitis [9] [10] [11] [7]. The march is not a fixed sequence; a large analysis reframed it as "many trajectories, many pathways" [12]. But the ordering is common enough, and the mechanism below suggests why.
Start here: two hits, and why the order matters
The barrier
The outermost skin layer is a brick-and-mortar arrangement of dead, flattened keratinocytes in a lipid matrix, and one of the proteins that builds it is filaggrin. Loss-of-function variants in the filaggrin gene predispose to atopic dermatitis [13] [14], and the mechanism is direct: filaggrin breaks down into the components of natural moisturising factor, so loss-of-function mutations reduce it and leave the skin dry and permeable [15].
The genetics did something more interesting than identify a risk gene. Filaggrin variants are also a significant risk factor for peanut allergy [16] and for allergic sensitisation and allergic disease generally [17]. A gene expressed in skin, not gut and not lung, raising the risk of a food allergy is hard to explain unless sensitisation is happening through the skin — which is the strongest available argument that the barrier defect comes first and the allergy follows, rather than both being expressions of a general atopic tendency.
And the barrier defect is not purely genetic. Type-2 cytokines themselves suppress filaggrin expression [18] [19]. So inflammation degrades the barrier that was letting the inflammation in — the loop closes.
The immune response
What comes through a leaky barrier is met by a type-2 immune response: IL-4 and IL-13 above all, with IL-25 and IL-33 driving type-2 innate lymphoid cells [20], and progressive activation of Th2 and Th22 cytokines characterising the lesions [21] [22] [23]. The disease is now routinely described as one of altered barrier plus immune dysregulation rather than either alone [24] [25] [26].
IL-31 deserves separate mention because it explains the symptom. A sensory-neuron-expressed IL-31 receptor mediates T-helper-cell-dependent itch [27] — a cytokine released by immune cells that acts directly on nerves. That is the molecular link between inflammation and scratching, and it later became a drug target.
The microbes
Atopic skin is colonised by Staphylococcus aureus, and the relationship is more than incidental [28]. Sequencing serial skin samples from children showed microbial communities at sites of disease predilection are dramatically different from controls, and — the striking part — treatment-associated increases in bacterial diversity preceded improvement in disease activity [29]. Diversity recovering before the skin does suggests the dysbiosis is participating rather than merely reflecting.
Pillar 1: measurement and diagnosis
The diagnosis is clinical
Chronic, relapsing, itchy eczematous rash in an age-typical distribution, with personal or family atopy, and no confirmatory test; studies operationalise it with the UK Diagnostic Criteria [30]. Age matters to the picture in another way too: the global-burden analysis found prevalence follows a bimodal curve, peaking in early childhood, falling in young adults, and rising again later [31]. Patch testing is used to identify contact allergy where that is suspected [32], and skin-prick testing has European standards [33], but neither diagnoses atopic dermatitis.
The scores, and which one to believe
Three instruments dominate: EASI (Eczema Area and Severity Index, clinician-assessed extent and intensity, 0–72), SCORAD (which mixes clinician signs with patient-reported itch and sleep loss) and POEM (Patient-Oriented Eczema Measure, symptoms only). Their responsiveness and minimal clinically important differences have been compared head-to-head [34], severity strata defined [35], POEM translated for clinical use [36], and the field went through a deliberate exercise — Harmonising Outcome Measures for Eczema — to stop every trial choosing its own [37] [38] [39].
Itch gets measured separately, because it does not track the visible disease closely enough to be inferred from it. The Peak Pruritus Numerical Rating Scale has been psychometrically validated with a defined responder threshold [40], and chronic pruritus has its own European guideline [41].
Centerpiece: what happened when the cytokines were blocked individually
For decades the options were topical steroids and, for severe disease, broad immunosuppression. Then the specific cytokines were targeted, and the results are best read from a single trial that put two mechanisms side by side.
JADE COMPARE randomised 838 patients 2:2:2:1 to oral abrocitinib 200 mg, oral abrocitinib 100 mg, subcutaneous dupilumab 300 mg every other week, or placebo — all on background topical therapy. At week 12, EASI-75 response was 70.3, 58.7, 58.1 and 27.1 percent; IGA response (clear or almost clear, improving at least 2 points) was 48.4, 36.6, 36.5 and 14.0 percent [42].
The check the arithmetic was never given. Look at abrocitinib 100 mg against dupilumab. These are as different as two drugs in this field get: an oral small molecule inhibiting JAK1 inside the cell against an injected antibody blocking the IL-4 receptor on its surface. They land 0.6 percentage points apart on EASI-75 (58.7 versus 58.1) and 0.1 apart on IGA (36.6 versus 36.5). Two independent endpoints, the same coincidence, in the same trial with the same patients and the same background therapy. The paper reports all four numbers and does not remark on it.
That convergence is not an artefact of a flat dose-response curve — abrocitinib 200 mg beats its own lower dose on both endpoints and beats dupilumab too. What it suggests is that at that level of response the ceiling is being set by the disease rather than by which node in the pathway you block, and that pushing past it takes more suppression rather than a different target.
A second check, and a warning about how these numbers get quoted. Tralokinumab, an anti-IL-13 antibody, reported EASI-75 of 25.0 and 33.2 percent in its two phase 3 trials [43] — against dupilumab's 58.1 percent above. Read side by side that looks like a large difference between drugs. It is mostly a difference between trial designs. The tralokinumab trials tested monotherapy, with control arms getting no topicals and achieving EASI-75 of 12.7 and 11.4 percent. The trials with background topicals had control arms achieving 27.1 and 29.6 percent [42] [44].
Within each design the control arms agree to under 2.5 percentage points; between designs the control response more than doubles (×2.35). Four trials, two designs, and the thing that moves the number most is what the control group was allowed to put on their skin. Which is why a monotherapy headline cannot be set beside a with-topicals headline, and why the only comparison a design licenses is against its own control. Tralokinumab has also been tested with topical corticosteroids, which is the comparable design [45].
The teaching point. Blocking IL-4/IL-13 signalling, or the JAK pathway they signal through, clears eczema far better than anything that preceded it — dupilumab's earliest trials already showed 85 percent achieving EASI-50 against 35 percent on placebo [46], and the phase 3 SOLO trials established the effect at scale [47] [48]. This is the same story as psoriasis: a pathway was mapped, individual nodes were drugged, and the response rates moved by a large multiple. The difference is that atopic dermatitis got there about a decade later.
Three honest limits. JADE COMPARE reports EASI-75 at week 12 while several other trials report week 16, and response is still rising over that interval. The convergence between abrocitinib 100 mg and dupilumab is a single observation at one dose in one trial, not a general property of the two classes. And EASI-75 is a threshold on a clinician-scored index — it says nothing directly about itch, which is what patients mind most and which is measured separately [40].
Pillar 2: treatment
The foundation is the barrier
Emollients, liberally and continuously, are the base of every treatment plan, and moisturisers have been reviewed specifically for their role across dermatitis [49]. This is not adjunctive comfort — it is treating half the mechanism.
Whether emollients from birth can prevent the disease is the most interesting open question in the field. A randomised controlled trial in 124 high-risk neonates in the US and UK tested full-body emollient from within three weeks of birth against no emollient, explicitly framed as testing whether barrier enhancement is a feasible prevention strategy [50]. The logic is compelling — if a leaky barrier initiates both the eczema and the sensitisation that follows it, sealing the barrier early should prevent both. Larger trials since have complicated the picture, and the definitive later trials are not in this substrate, so this review does not state a verdict.
Anti-inflammatory topicals
Topical corticosteroids remain the mainstay. Topical calcineurin inhibitors — tacrolimus and pimecrolimus — were the first steroid-sparing option, with tacrolimus ointment established in early trials [51], and they matter most for the face and skin folds, because a Cochrane review of topical tacrolimus found no evidence that it causes skin atrophy while finding 0.1% tacrolimus better than low-potency corticosteroids [52] — atrophy being the constraint that limits steroid use at those sites. Their own safety debate, over a theoretical cancer risk, has been reviewed and the authors note the risk of replacing "corticosteroid phobia" with "calcineurin phobia" [53]. Crisaborole, a PDE4 inhibitor, and topical ruxolitinib, a JAK1/2 inhibitor, extended the topical options — both FDA-approved for atopic dermatitis [54] [55] — with delgocitinib, tapinarof and others in trials [56] [57]. Cost has been noted as a barrier to the newer topicals [54].
Phototherapy and the old systemics
Narrowband UVB reverses atopic dermatitis with measurable changes in disease biomarkers [58], and phototherapy has been systematically reviewed [59] [60] [61]. Ciclosporin works and modulates the activated inflammatory pathways [62], but it is a potent broad immunosuppressant whose side effects limit its use [44] — which is exactly the constraint the targeted agents removed. Notably, the dupilumab-plus-topical-steroid trial was designed in patients who had failed, could not tolerate, or should not receive ciclosporin [44].
The targeted systemics
Dupilumab blocks the IL-4 receptor alpha subunit and therefore both IL-4 and IL-13 signalling [46] [47] [48]. It works with concomitant topical steroids [44] [63], in adolescents [64] and in children [65]. The same drug is effective in asthma [66] [67] and in chronic rhinosinusitis with nasal polyps [68] — which is the atopic march showing up in a pharmacology dossier.
Anti-IL-13 antibodies — tralokinumab [43] [45] and lebrikizumab [69] — narrow the target to one cytokine.
Nemolizumab targets the IL-31 receptor and is aimed at the itch specifically, with its phase 3 trial's primary endpoint being percent change in a pruritus visual-analogue scale rather than a skin score [70]. That choice of endpoint follows straight from the IL-31 biology [27], and it is the clearest example in this review of a mechanism dictating what gets measured.
Oral JAK inhibitors — abrocitinib [42] [71] [72] and upadacitinib — block signalling downstream of several cytokines at once. Upadacitinib was tested head-to-head against dupilumab in Heads Up, a 692-patient double-dummy trial with EASI-75 at week 16 as the primary endpoint [73]. They are oral, they act fast, and Heads Up reported upadacitinib superior to dupilumab on efficacy at 16 weeks with no new safety signals [73], with longer-term data to 52 weeks [74] and withdrawal-and-retreatment strategies studied [75]. The class carries specific cautions: the European Medicines Agency has issued recommendations restricting JAK inhibitor use in patients at risk — those with cardiovascular or thromboembolic disease or malignancy, current or former smokers, and people aged 65 and over [57]. That is the trade against the antibodies, which do not carry it.
Managing the itch-scratch cycle and infection
Scratching maintains the disease, so breaking the cycle matters independently of the inflammation. And because S. aureus colonisation tracks disease activity [28] [29], overt infection needs treating — though the microbiome data suggest the relationship is bidirectional rather than simply infective.
Pillar 3: what is unresolved
Prevention. The barrier hypothesis makes a clean prediction and early emollient trials tested it [50]. The evidence has become more contested rather than less, and this substrate does not contain the later definitive trials.
Endotypes. Atopic dermatitis varies by age, ethnicity and cytokine profile, and the Th2/Th22 balance differs between groups [21] [22]. Whether that predicts who responds to which of the now-numerous targeted options is the practical question.
The microbiome as a target. Diversity recovering before disease improves [29] is suggestive; a gut–skin axis has been proposed [76] [77] [78]; nothing has yet become a treatment.
Whether treating the eczema interrupts the march. If skin sensitisation drives food allergy and asthma [16] [17], then controlling eczema early ought to reduce them. The filaggrin–peanut-allergy link makes the hypothesis biologically serious, and the peanut-introduction trial showed the allergy end of it is modifiable [79], but whether eczema treatment does it is unproven.
Itch as its own endpoint. Nemolizumab's design [70] and the validation of dedicated itch scales [40] [41] reflect a shift from scoring skin to scoring symptoms — and, as the figure notes, EASI-75 says nothing directly about the thing patients most want stopped.
Dig deeper in lmmol
Psoriasis is the closest structural parallel anywhere in this collection: another immune-mediated skin disease where a cytokine pathway was mapped and then drugged node by node, with the same large jump in response rates — and the contrast is instructive, since psoriasis runs on IL-23/IL-17 while atopic dermatitis runs on IL-4/IL-13, so the two diseases sit on opposite arms of the immune system and their biologics are not interchangeable [80]. Asthma is the other end of the atopic march, and the link is pharmacological as well as epidemiological: dupilumab treats both [66] [67], which is unusual evidence that the two conditions share a driver rather than merely co-occurring [9] [10]. Celiac disease offers the barrier parallel from a different organ — a surface that should exclude an antigen, failing to. Rheumatoid arthritis shares the JAK-inhibitor class and its safety questions. And inflammatory bowel disease shares the pattern of broad immunosuppression giving way to targeted biologics, with the same trade of efficacy against specificity. The full collection is at health.