This week in Pathology

This AudioScholar briefing is generated by artificial intelligence for healthcare professionals and trainees. It is not medical advice.

This week: a computational TROP2 score that tracks benefit from datopotamab deruxtecan in lung cancer, endometrial cells on Pap tests that carry more weight after sixty, and methylation profiling that splits a rare ovarian cancer into two subtypes.

Welcome to This Week in Pathology. We're covering 10 papers this week: 3 in depth, 4 in brief, 3 quick signals.

In Nature Medicine, Garassino and colleagues report a biomarker analysis built on two trials of datopotamab deruxtecan, a TROP2-directed antibody-drug conjugate, in advanced non-small-cell lung cancer [1]. Conventional visual TROP2 scoring by pathologists does not predict response to these drugs. The group therefore built a supervised computational pathology tool that measures TROP2 in individual cells on standard immunohistochemistry slides. From it they derived a normalized membrane ratio, which is membrane staining relative to membrane plus cytoplasmic staining. The score was developed in 115 patients from TROPION-PanTumor01. Its cut-point was then optimized and validated in 352 biomarker-evaluable patients from TROPION-Lung01, with 221 patients in the main intended population of nonsquamous disease without actionable alterations. With datopotamab deruxtecan, median progression-free survival was 7.2 months in ratio-positive tumours against 4.0 months in ratio-negative tumours, roughly a sixty percent lower risk of progression. With docetaxel, the score made no difference. That treatment-by-biomarker interaction was statistically significant, which supports a predictive role rather than a purely prognostic one. The main limitation is that the cut-point was tuned and validated within the same trial programme, and the abstract reports progression-free survival only, with no overall survival result. This is the first evidence that a quantitative subcellular readout succeeds where eyeball scoring has failed.

Antibody-drug conjugate selection is also the subject of a Human Pathology study from Ramineni and colleagues [2]. They reviewed 241 non-breast, non-gastroesophageal solid tumours that underwent HER2 immunohistochemistry at one large academic institution in the United States between 2024 and 2025. Twenty-two tumour types were tested, most often endometrial, colorectal and urothelial carcinomas. About one in ten tumours was HER2 positive. Urothelial carcinoma had the highest positivity rate among the more common types, at roughly a third of cases. The positive predictive value of immunohistochemistry for ERBB2 amplification using gastric criteria was 50 percent overall. It reached 100 percent outside urothelial carcinoma but was only about 14 percent within it. Nine of nineteen eligible patients received trastuzumab deruxtecan, and seven of those nine had stable disease or no detectable tumour over a median of 21 months. On the practice side, fourteen cases were scored with criteria that did not match recommendations. A quarter of eligible reports lacked gastric-criteria results needed for trastuzumab deruxtecan eligibility. Numbers per tumour type are small and the treated group is tiny, so the outcome data are descriptive. The study documents real gaps in how HER2 is scored outside breast and stomach, and it flags urothelial carcinoma as the place where staining and amplification diverge most.

In the American Journal of Clinical Pathology, Burke and colleagues examined benign-appearing endometrial cells on Pap tests in women aged 45 or older, combining a retrospective cohort with a meta-analysis [3]. Among 133,280 Pap tests from 2021 to 2023 at one institution, these cells were reported in 2,030 women. Of the 604 women with histologic follow-up, six percent had atypical hyperplasia or endometrial carcinoma. Carcinoma risk climbed with age, from under one percent at 45 to 49 years to about seven percent at 60 and over. Carcinoma risk was also higher in postmenopausal than in premenopausal women, about five percent against under one percent. Clinically significant pathology stayed at or below five percent through age 59, then rose to just under thirteen percent from age 60. Pooling seven cohorts and 4,989 women gave an overall risk of about four and a half percent, with substantial heterogeneity between series. The main limitation is that fewer than a third of the women had tissue follow-up, so selection toward symptomatic or higher-risk patients is likely. The authors conclude that combining age with menopausal status may stratify risk better than age alone, a position current guidelines do not yet reflect.

Now, four in brief.

Three Modern Pathology papers test DNA methylation against morphology in rare tumours, and they do not all point the same way. Malik and colleagues studied 14 superficial small round cell tumours with FET::ETS fusions or EWSR1 rearrangements and S100 or SOX10 expression [4]. Methylation clustering agreed with histology in nine of ten confidently classified tumours. However, it placed every SOX10-expressing tumour with superficial neurocristic tumour, regardless of morphology. Two such cases with high-grade cytology metastasized, and one patient died. Pooled survival did not differ significantly from superficial Ewing sarcoma. The authors favour treating ambiguous or high-grade lesions as Ewing sarcoma, a conclusion resting on very few cases.

Patton and colleagues profiled 63 perineurial and related tumours [5]. Conventional and sclerosing perineuriomas formed one distinct methylation group, separate from their soft tissue mimics. A quarter of benign perineuriomas carried mutually exclusive NF1 or NF2 deletions. Hybrid schwannoma-perineuriomas grouped with neurofibromas. Ten of eighteen malignant perineuriomas sat closer to undifferentiated sarcoma, with heavy copy number change and alterations such as TP53 and RB1. Here methylation backs up the benign diagnosis but questions whether malignant perineurioma is a single entity, with follow-up of under two years.

Diwan and colleagues profiled 63 small cell carcinomas of the ovary, hypercalcemic type [6]. Methylation split them into two clusters. The first showed classic small cells, diffuse WT1 and hypercalcemia. The second showed large rhabdoid cells, absent WT1, and keratin and cyclin D1 expression. Both clusters had low copy number burden, consistent with SMARCA4 loss as the shared driver, and both subtypes were aggressive. The therapeutic implications remain untested.

Outside soft tissue, Khutanthong and colleagues, also in the American Journal of Clinical Pathology, tested apolipoprotein E staining on 39 kidney biopsies [7]. Staining was absent in all seven C3 glomerulonephritis cases and present in eleven of twelve dense deposit disease cases. Five of twenty intermediate cases stained, often segmentally. Electron microscopy deposit density did not differ significantly between groups. The marker looks useful as an adjunct, but it is not specific, and the sample is small.

Three quick signals.

In Cancer Cytopathology, Gao and colleagues report a rapid onsite molecular adequacy protocol for ultrasound-guided sampling of pancreatic cancer in 61 patients [8]. DNA sequencing succeeded in 95 percent of sequenced specimens, but RNA fusion testing succeeded in only about half. Overall adequacy did not substantially improve on prior experience. This is single-centre quality data.

Kasajima and Klöppel, in Virchows Archiv, review biopsy diagnosis of digestive neuroendocrine neoplasms [9]. They focus on separating tumours from carcinomas, Ki-67 assessment, heterogeneity, mixed neoplasms and primary-site assignment in metastases. It is an expert narrative review rather than new data.

Jen and Loghavi, in the American Journal of Clinical Pathology, set out a diagnostic framework for myeloid or lymphoid neoplasms with tyrosine kinase gene fusions [10]. They anchor diagnosis in molecular confirmation rather than eosinophilia, with tiered testing escalating to RNA sequencing or optical genome mapping. It is expert opinion integrating classification updates.

Stepping back across this week's papers, most of the evidence comes from small, retrospective, often single-institution series of rare tumours, where molecular classifiers are advancing faster than the outcome data needed to judge them. Methylation profiling refines categories, yet in one study it conflicted with morphology on behaviour. Only the antibody-drug conjugate biomarker work rests on randomized trial material, and even there validation stayed within the same programme.

If you only have time for one paper this week, make it the TROP2 normalized membrane ratio study [1]. It reopens the question of whether TROP2 can be a usable predictive biomarker, and it places computational scoring of immunohistochemistry at the centre of that answer.

That's your roundup for This Week in Pathology. The full transcript and references are available on the episode page. See you next week.

If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And a quick rating on Apple Podcasts or Spotify helps other physicians discover the show.

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