This Week in Pathology — Sep 12, 2026

Welcome to This Week in Pathology. This week we're covering 10 notable papers spanning classification and molecular taxonomy, new immunohistochemical markers that solve everyday diagnostic problems, and the recurring challenge of cutaneous and soft tissue mimics. Let's dive in.

We start with how we classify tumours, and the first paper comes from Virchows Archiv, where Kushima lays out the framework of the sixth edition of the World Health Organization Classification of Digestive System Tumours for gastric carcinoma [1]. Morphology remains the foundation, but the refinements matter for daily sign-out. Crawling-type adenocarcinoma is now recognised as a distinctive variant of tubular adenocarcinoma, poorly cohesive carcinoma is subclassified into signet-ring cell and non-signet-ring cell subtypes, and the concept of pure signet-ring cell carcinoma is introduced. The uncommon subtypes have been expanded too — carcinoma with lymphoid stroma, alpha-fetoprotein-producing carcinoma, micropapillary adenocarcinoma, fundic-gland type adenocarcinoma and sarcomatoid carcinoma. The larger shift is that the molecular subgroups originally proposed by The Cancer Genome Atlas, along with HER2, Claudin 18.2, mismatch repair deficiency and PD-L1, have moved from research categories into essential reporting for precision oncology. If your gastric reports don't yet reflect that integrated histomolecular framework, they will need to. Also in Virchows Archiv, Švajdler and colleagues report a real-world central European experience with genome-wide DNA methylation profiling in central nervous system tumours — a prospective, consecutively accrued three-year cohort of 291 cases across adults and children [2]. Profiling succeeded in about 96 percent of cases, and roughly seven in ten analysable samples reached a high-confidence classifier match, with only about 3 percent completely unclassifiable. Integrated into a full diagnostic workup, methylation gave clinically useful results in about eight of every ten cases, resolving uncertainty or prompting major revision in 149 histologically challenging tumours. Among the genuinely ambiguous lesions, methylome data changed the tumour grade in close to four in ten cases, mostly upgrading. The practical message is about the low-confidence band: more than half of those lower-confidence cases still yielded meaningful integration when combined with morphology and ancillary markers, so a score below the cutoff should not be read as assay failure. Discrepant or frankly misleading classifications occurred in about 2 percent, and simply updating the bioinformatic pipeline version rescued several ambiguous entries. Computational epigenomics is powerful, but it is an adjunct to neuropathology, not a replacement for it.

A second theme is new immunohistochemical markers that address problems we have all been stuck with. In The American Journal of Surgical Pathology, Wang and colleagues address the absence of a reliable marker for gastric and oesophageal adenocarcinoma, where CDX2 has always underperformed [5]. Mining The Cancer Genome Atlas, they identified friend of GATA1, or FOG1, as showing high messenger RNA expression essentially only in gastro-oesophageal adenocarcinoma across 29 solid tumour types, then tested it by immunohistochemistry in 187 gastro-oesophageal and oesophageal squamous cases. About nine in ten gastro-oesophageal adenocarcinomas showed moderate to high FOG1, whereas only about half showed meaningful CDX2 — and the two markers were close to reciprocal in colonic adenocarcinoma, where most tumours were CDX2 positive and over nine in ten were FOG1 negative or low. Neither marker stained oesophageal squamous cell carcinoma. There is some cross-reactivity to be aware of: about one in six pancreatic and about one in nine ampullary adenocarcinomas showed moderate to high FOG1, though staining was rare in breast, lung and ovary and absent in a long list of other carcinomas. For carcinoma of unknown primary work-ups above the ligament of Treitz, this is a marker worth validating. Staying with marker validation, a multi-institutional series in Virchows Archiv from Kalomeris and colleagues examined cribriform tumour of the skin, the rare adnexal neoplasm of uncertain malignant potential [6]. Across 11 morphologically classic tumours, CD38 was expressed in about eight in ten, while all 28 morphologic mimics — adenoid cystic carcinoma, digital papillary adenocarcinoma, eccrine and apocrine adenomas, hidradenoma and endocrine mucin-producing sweat gland carcinoma — were uniformly negative. Recurrent deletions of 6q and 9q were found in seven of eight successfully tested cases. So CD38 is highly specific among those mimics but imperfectly sensitive: two classic tumours were negative, including one with confirmed co-deletion. A negative stain does not exclude the diagnosis.

That brings us to the broader problem of mimicry, which dominates several papers this week. In Human Pathology, Gilbert and colleagues studied 22 dedifferentiated or undifferentiated melanomas that had initially been called sarcoma [4]. These were mostly older men, most without any known melanoma history, nearly all with metastatic disease and a strong preference for lymph node basins, especially the axilla. Just over half were negative for melanocytic markers altogether. The discriminator was PD-L1: every one of the 22 tumours showed strong expression, with a mean combined positive score of 93, whereas a comparison cohort of 87 sarcomas averaged a combined positive score under 6 with a median of zero, and only three exceeded 50. High tumour mutational burden and an ultraviolet mutational signature were present in every case tested. Of the ten patients treated with PD-L1-targeted immunotherapy, eight had a partial or complete response. The practical implication is direct — a spindled or rhabdoid tumour in an older man with a brisk inflammatory infiltrate and diffuse strong PD-L1 should raise dedifferentiated melanoma before you settle on sarcoma, and the answer changes treatment. Two Modern Pathology papers extend the cutaneous theme. Moran and colleagues characterised 27 primary cutaneous apocrine carcinomas, most in men and most in the axilla [9]. Regional nodal metastasis was present in over half the patients, and about a third were deceased at last follow-up, so this is not an indolent entity. Every case tested showed diffuse strong androgen receptor and gross cystic disease fluid protein-15, with oestrogen and progesterone receptor positive only in a minority; six cases showed 3+ HER2 and three showed amplification. Recurrent alterations included ERBB2, KMT2C and PIK3CA — a profile that both separates these from mimics and opens targeted options. And Ebbelaar and colleagues examined 16 cutaneous melanocytic tumours carrying both an NRAS Q61 and an IDH1 R132C mutation, showing that this genotype spans a spectrum [10]. Seven were melanocytomas with reproducible biphasic architecture, low Ki-67, PRAME negativity, retained p16 and minimal copy number change; nine were melanomas, distinguished by higher-grade cytology, TERT promoter mutation in all nine, 9p21 loss in several, and higher copy number burden. Reassuringly, over a median melanoma follow-up of about four years there were no distant metastases and no melanoma-related deaths. Moderate-to-severe atypia carried the strongest association with progression, raising the risk roughly six-fold, with a Ki-67 of 10 percent or above, lymphocytic infiltrate, loss of the typical biphasic pattern and complete p16 loss also flagged — though these were exploratory analyses in a small series.

Finally, a set of papers on reproducibility and rare entities. In Modern Pathology, Li and colleagues ran the first systematic interobserver study of the core criteria used to classify fibroepithelial lesions, with 19 subspecialised academic breast pathologists scoring stromal cellularity, stromal cell atypia and tumour border [3]. Agreement was moderate for stromal cellularity and for tumour border, and poor for stromal cell atypia on a three-tier scale — improving only to moderate when collapsed to a two-tier system. Structured virtual training with a washout period barely moved the needle; what it did do was make readers more conservative, with most participants assigning lower cellularity and lower atypia afterwards. That is a sobering result for phyllodes grading, and it argues for two-tier reporting and for the reference images this group has now established. Also in Modern Pathology, Roark and colleagues review C-cell lesions and medullary thyroid carcinoma, emphasising the grading criteria based on mitotic activity, Ki-67 and necrosis that now drive risk stratification, alongside germline and somatic RET testing and selective RET inhibitors [7]. And in The American Journal of Surgical Pathology, Dundr and colleagues add three uterine sarcomas with KDM2B fusions, with partners EPC1, EPC2 and CREBBP [8]. Morphology ranged from pure spindle cell to mixed spindle and round cell with prominent hyalinisation, overlapping with low-grade endometrial stromal sarcoma — but behaviour was aggressive, with two patients dying of disease and the third recurring in the pelvis. These belong with the BCOR-altered high-grade endometrial stromal sarcomas, and fusion testing is what separates them from a low-grade diagnosis.

If you only have time for one paper this week, make it the Human Pathology series on dedifferentiated and undifferentiated melanoma masquerading as sarcoma [4]. It identifies a misdiagnosis most of us will encounter, gives a simple, widely available immunohistochemical discriminator, and the correction leads directly to an effective therapy.

Here are the key takeaways from this week in Pathology. First, the sixth edition World Health Organization gastric framework moves predictive biomarkers — HER2, Claudin 18.2, mismatch repair and PD-L1 — from optional to essential, alongside new morphologic categories including pure signet-ring cell carcinoma. Second, methylation profiling in central nervous system tumours delivers clinical utility in roughly eight of ten cases, and low-confidence scores are still often informative when read with morphology, so do not discard them. Third, strong diffuse PD-L1 in a sarcoma-like tumour should prompt serious consideration of dedifferentiated melanoma. Fourth, FOG1 is a promising sensitive and specific marker for gastro-oesophageal adenocarcinoma where CDX2 fails, and CD38 is a specific but imperfectly sensitive adjunct for cutaneous cribriform tumours. And fifth, interobserver agreement on phyllodes stromal atypia remains poor even among subspecialists, and training alone does not fix it — favour two-tier reporting.

That's your roundup for This Week in Pathology. The full transcript and references are available on the episode page in your AudioScholar library. This is an AI-curated summary — for clinical decisions, always consult primary sources and current guidelines. See you next week.

If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And if you have a paper you have been meaning to read, upload the PDF, or paste any link, at audioscholar dot C C. We will turn it into audio like this one, in any of thirty-one languages.

This is an automated summary generated by artificial intelligence, which can make mistakes. Always review the original source materials.