Bone Lesions and a Rash: Langerhans Cell Histiocytosis

This post walks the disease as it is understood now: a clonal neoplasm rather than an immune reaction, the older names that still show up in reports, what imaging suggests and what only a biopsy settles, why one site and many are treated differently, and the consequence that keeps follow-up lifelong.

Three findings that only mean something together

LCH most commonly presents as a painful bone lesion, with skin the second most commonly involved organ, per the NCI PDQ health professional summary. Bone is affected in an estimated 80% of patients. It can occur in any bone, though the hands and feet are often spared, and the most frequent site in children is a lytic lesion of the skull vault, painful or completely silent. After the skull come the femur, ribs, humerus, pelvis and vertebrae.

Spine lesions can involve any vertebra, most often in the cervical spine, and may collapse the vertebral body into the flattened shape radiologists call vertebra plana. A vertebral lesion with soft tissue extension tends to present with pain and can cause significant neurological deficits, which is what prompts an MRI to assess cord compression (NCI PDQ).

The skin half of the picture is where the delay usually lives. In infants, seborrhoeic involvement of the scalp may be mistaken for prolonged cradle cap unless the classic purpuric component is present, per PDQ, and the next most common skin site is the body creases such as the antecubital fossa and the perineum. Infants can also present with a generalised rash that mimics other skin disorders, and lymph nodes are a third site, most often in the neck.

None of this is common: two to ten cases per million children aged 15 or younger, a median age at presentation of 30 months, and 1.06 per million adults a year in an England study, a figure that counts everyone from age 15 upward as an adult (NCI PDQ).

Why the name on the report may not match the textbook

For most of the twentieth century LCH was filed as an immune or inflammatory reaction. That has changed. The NCI PDQ health professional summary states that LCH results from the clonal proliferation of immature dendritic cells and is now considered a myeloid neoplasm, while adding that some controversy remains over whether it is a true malignancy. A 2025 Histiocyte Society review in Haematologica says LCH was historically considered a reactive inflammatory condition and is now best described as a clonal myeloid neoplastic disorder with inflammatory features.

That reclassification is not a labelling exercise. A reactive process gets watched. A clonal neoplasm gets staged, gets treated by extent, and carries a driver mutation that can be aimed at.

The old names did not retire with the old theory. Paperwork may say histiocytosis X, where the X stood for the unknown cell, or eosinophilic granuloma, or Hand-Schuller-Christian disease, or Letterer-Siwe disease. The US National Library of Medicine still indexes histiocytosis X, Hand-Schuller-Christian disease and Letterer-Siwe disease as entry terms for Langerhans-cell histiocytosis in MeSH, and a 2020 skull vault imaging review in Insights into Imaging still labels the classic lesion eosinophilic granuloma. Two reports can carry two names for one condition, and neither is wrong.

What the X-ray suggests and what only the biopsy settles

Skull lesions are typically well defined and lytic. That Insights into Imaging review describes beveled edges and a hole within a hole appearance, produced by endosteal scalloping and slight expansion beyond the inner and outer tables, with a central button sequestrum classically described. It also notes margins may be sclerotic or non-sclerotic, which matters because a sharply outlined hole is suggestive, not diagnostic.

The differential is real. The same review places fibrous dysplasia, recognised by its ground-glass matrix, beside eosinophilic granuloma for a lytic skull lesion in a young patient, and lists epidermal inclusion and dermoid cysts among the lookalikes. In an older patient the same shape raises different candidates. Our pieces on telling a bone metastasis from a benign lesion and on the vague symptoms of multiple myeloma cover how the same hole on a film gets worked up in adults.

So the tissue decides. PDQ is explicit that a pathological diagnosis is always required, with narrow exceptions where the risk outweighs the benefit, such as isolated pituitary stalk disease or vertebra plana with no soft tissue mass. Lytic bone lesions, skin and lymph nodes are the sites most frequently biopsied. LCH cells stain positively with antibodies to S100, CD1a and anti-langerin (CD207), and PDQ states that staining with CD1a or langerin confirms the diagnosis.

Molecular testing sits alongside the immunostains. An activating BRAF V600E variant was found in 35 of 61 cases (57%) in the 2010 report that opened this work, and later reports confirmed it in 50% or more of childhood cases (NCI PDQ). PDQ notes pathway testing is not a required part of the workup, while the variant can be detected in tissue or peripheral blood, and that trials of BRAF and MEK inhibitors such as vemurafenib and dabrafenib are open for adults and children. The molecular line at the bottom of a pathology report can matter more than its placement suggests.

One site or many: the line that sets the treatment

The organising question after diagnosis is extent. PDQ separates single-system disease, one system involved, from multisystem disease, two or more. Inside multisystem disease the further split is risk-organ involvement: the liver, spleen or haematopoietic system including bone marrow. Survival is closely linked to the extent of disease at presentation when those high-risk organs are involved, and to the response to initial treatment (NCI PDQ).

A second map runs across the skull. PDQ defines a CNS-risk group: lesions of the facial bones or the anterior or middle cranial fossae, such as the temporal, orbital, sphenoid, ethmoid and zygomatic bones, with intracranial tumour extension. Those patients carry a threefold increased risk of other central nervous system disease and of diabetes insipidus, and systemic treatment is recommended for that reason even at a single bony site.

Treatment therefore ranges widely, and the range is the point. For a single skull lesion of the frontal, parietal or occipital region, or a single lesion of any other bone, PDQ gives curettage only as the recommended therapy when possible, with curettage plus an injection of methylprednisolone as an alternative. Observation is described for a single vertebral body lesion without extradural extension. At the other end, the most commonly used systemic regimen is vinblastine with prednisone for 12 months, per the HISTSOC-LCH-III trial, for multifocal bone disease, low-risk multisystem disease and CNS-risk bones. Reactivation risk tracks the same axis, from 10% in single-system unifocal bone disease to close to 50% in multisystem LCH.

The thirst that can arrive before the diagnosis or years after it

Diabetes insipidus is the consequence that shapes the rest of the story. It comes from LCH damage to the antidiuretic hormone-secreting cells of the posterior pituitary and is the most frequent endocrine manifestation of the disease, per PDQ. Among patients who survive LCH, central diabetes insipidus is the most common permanent condition, and CNS neurodegenerative LCH is the most severe.

The timing is what catches people out. Roughly 4% of patients present with an apparently idiopathic form of diabetes insipidus before any other LCH lesion is identified, and in a prospective study of children presenting with idiopathic central diabetes insipidus, 19% eventually developed signs of LCH (NCI PDQ). Running the other way, 75% of LCH patients with diabetes insipidus had CNS-risk bone lesions, and 56% of those who developed it went on to develop anterior pituitary hormone deficiencies within ten years.

Reported long-term consequences range from 20% to 70%: about 20% with low-risk-organ involvement such as skin, bone, lymph nodes or pituitary, and about 70% for multisystem disease when treatment lasted only six months (NCI PDQ). The Haematologica review adds that neurodegeneration may arise decades after the original presentation, with a cumulative incidence of 6% in the largest published paediatric cohort. A disease whose worst outcome can surface twenty years later is a disease whose record has to survive twenty years.

How Healz holds a case like this

The skull X-ray report, the scalp rash, the biopsy immunostains and the hormone results arrive from four departments on four dates, and the treatment depends on all of them, because single-site and multi-site disease are not managed the same way. Healz keeps them in one chat, not ten apps and four portals, with the date attached to each.

Healz is equipped with memory. It pulls the medical facts out of every upload, keeps the abnormal values up front, and lines the findings up in order, so last spring's rash sits beside this month's CD1a result instead of in a folder nobody reopens, and a value that drifts over years reads as a line, not a number somebody has to recall. Frontier AI works the whole case, so the lesion, the skin and the endocrine results are read against each other.

Healz is also equipped with root-cause technology. It does not stop at the first label, asks the questions a rushed visit skips, and cross-checks the case against more than a million rare cases, which is what a disease this uncommon needs. As a cancer ai and pathology report reader, it is equipped to read a pathology report and say in plain words what CD1a, langerin and a BRAF result mean, search current studies and trials, and produce a PDF of questions for the next appointment. You can bring a board-certified doctor into the same chat for a second opinion.