This Week in Pathology — Jul 26, 2026

Welcome to This Week in Pathology. This week we're covering 8 notable papers spanning genitourinary oncology diagnostics, molecular pathology of gastrointestinal and breast cancers, and artificial intelligence applications in anatomic pathology. Let's dive in.

Beginning with genitourinary oncology and tissue-based diagnostics, a landmark phase 3 open-label randomized trial published in The New England Journal of Medicine by Galsky and colleagues investigated perioperative enfortumab vedotin combined with pembrolizumab versus standard neoadjuvant cisplatin-gemcitabine in patients with muscle-invasive bladder cancer. At a median follow-up of 33.6 months, the estimated 2-year event-free survival was 79.4% in the combination arm compared to 66.2% in the chemotherapy arm, with a hazard ratio of 0.53. Estimated overall survival at 2 years reached 86.9% versus 81.3%, and the pathological complete response rate roughly doubled, jumping from 32.5% with chemotherapy to 55.8% with the antibody-drug conjugate and immunotherapy combination. Grade 3 or higher adverse events were more frequent with enfortumab vedotin-pembrolizumab at 75.7% compared to 67.2%. Despite the toxicity, these findings establish a superior pathological and survival benefit over traditional cisplatin regimens [1].

Moving to molecular diagnostics in gynecologic pathology, Warren and colleagues published a study in The American Journal of Surgical Pathology highlighting a major clinical bottleneck in precision medicine: the severely limited availability of short tandem repeat genotype testing for diagnosing partial hydatidiform mole in the United States. Because a partial hydatidiform mole is defined by diandric triploidy, definitive diagnosis in formalin-fixed products of conception requires short tandem repeat genotype testing. Through a survey of the Association for Molecular Pathology and direct queries to commercial reference laboratories, the authors found that only 13 academic and non-academic health system laboratories and a single commercial reference laboratory nationwide offered the test. Furthermore, many required internal consultative review. To navigate this reality, the authors propose a new reporting nomenclature system reflecting diagnostic certainty based on available methodologies, whether genotype, ploidy, or morphology alone [2]. In urological cytology, Barth and colleagues evaluated Keratin 17 immunocytochemistry as an adjunctive marker in urine cytology for urothelial carcinoma in a prospective cohort published in Cancer Cytopathology. Keratin 17 expression was significantly higher in urothelial carcinoma samples, and combining Keratin 17 with standard Paris System cytology numerically improved the area under the curve to 0.76, offering a reproducible adjunct for cytologically equivocal cases [7].

Exploring molecular heterogeneity and targeted therapy in gastrointestinal pathology, Larson and colleagues investigated colorectal carcinomas with multivergent morphology, defined as two discrete nonconventional morphologic patterns, in The American Journal of Surgical Pathology. Interrogating two distinct morphologic areas by next-generation sequencing in 11 cases revealed that multivergent colorectal carcinomas were larger, presented at higher T and N stages, and were more frequently mismatch repair-deficient with atypical patterns. Crucially, targeted alterations eligible for Food and Drug Administration-approved therapies or clinical trials were identified in 63.6% and 72.7% of cases, respectively, with actionable alterations frequently isolated to only one of the sequenced areas, underscoring the value of multivergent sampling [3]. Complementing this in gastric cancer, Hwang and colleagues assessed FGFR2 amplification, fusions, and FGFR2b overexpression in whole-slide surgical specimens using next-generation sequencing and immunohistochemistry in Virchows Archiv. Analyzing 113 advanced gastric cancer gastrectomy specimens, the authors found substantial concordance between immunohistochemistry and next-generation sequencing, with rare FGFR2 fusions defining an aggressive, poor-prognosis subtype associated with amplification and overexpression [4].

Turning to artificial intelligence and computational pathology, three key studies evaluated workflow integration across different subspecialties. In prostate pathology, Pascarella and colleagues published a multi-reader study in Virchows Archiv evaluating an artificial intelligence-assisted slide review tool for transurethral resection of the prostate specimens. Across five readers reviewing 102 slides, artificial intelligence assistance significantly reduced review time by a mean of 34.5 seconds per slide while maintaining high diagnostic accuracy and decreasing classification errors from 24 unaided to 9 aided, proving particularly efficient for less experienced readers [5]. In breast pathology, Yue and colleagues addressed the diagnostic challenge of identifying HER2-ultralow expression across primary and metastatic lesions in a multi-center validation study published in Virchows Archiv. Using a deep learning model validated on 89 consensus-reviewed cases from 20 medical centers, the artificial intelligence model achieved a superior detection rate for HER2-ultralow cases at 93.8% compared to the 84.6% pathologist consensus rate, while maintaining lower mean absolute error, demonstrating that computational tools can help pathologists achieve expert-level reproducibility in faint staining ranges [8]. Finally, in computational literature synthesis, Seifaddini and colleagues performed a systematic scoping review in Modern Pathology examining large language models in cancer pathology reports across a four-level capability framework ranging from text preparation to guideline-based clinical reasoning and interpretive synthesis. Reviewing 41 studies, the authors noted that while lower-level extraction tasks show high performance, success does not consistently translate to higher-level clinical reasoning, highlighting the need for multi-site validation and error analysis before safe integration into oncology workflows [6].

If you only have time for one paper this week, make it the phase 3 trial by Galsky and colleagues in The New England Journal of Medicine on perioperative enfortumab vedotin and pembrolizumab in cisplatin-eligible muscle-invasive bladder cancer. This study establishes a transformative standard of care by nearly doubling the pathological complete response rate and significantly improving event-free and overall survival compared to standard chemotherapy [1].

Here are the key takeaways from this week in Pathology. Perioperative enfortumab vedotin and pembrolizumab significantly improves event-free and overall survival and pathological complete response rates in muscle-invasive bladder cancer compared to cisplatin-gemcitabine, despite higher toxicity [1]. Short tandem repeat genotype testing for partial hydatidiform mole has critically limited availability across the United States, necessitating adjusted reporting nomenclature when genotyping is absent [2]. Next-generation sequencing of distinct morphologic areas in multivergent colorectal carcinomas identifies isolated targetable molecular alterations that may expand salvage therapy options [3]. Artificial intelligence assistance significantly streamlines transurethral resection of the prostate slide review and improves detection rates for challenging HER2-ultralow breast cancer cases [5, 8].

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 for audio briefings on your own clinical questions and papers, visit audioscholar dot C C.

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