Author: Ani Murtskhvaladze
Gulisa Turashvili, a TSMU alumna and Associate Professor of Pathology at Harvard Medical School and Massachusetts General Hospital, delivered a public lecture at Tbilisi State Medical University on the diagnostic challenges in endometrial and cervical tumors on October 16th, 2025. During the event, she discussed modern diagnostic approaches, international best practices, and key clinical
challenges in contemporary pathology. The interactive session provided students, faculty, and clinicians with valuable theoretical and practical insights. To further explore the complexities of endometrial cancer diagnostics, treatment guidelines, and pathologic differential diagnosis, we invited Dr. Turashvili for an in-depth interview.
How have recent updates in WHO classification or NCCN guidelines influenced the reporting standards of diagnostic criteria in your work?
Thank you for the question. The WHO classification mainly provides histologic subtypes for all cancers, including gynecologic cancers. The most recent edition was published about five years ago, and the sixth edition is currently in development. I’ve been fortunate to be involved in that work. It is scheduled for release in early 2026. The planned WHO edition will primarily refine diagnostic criteria and prognostic factors rather than introducing major changes. Most key criteria will remain the same, but the upcoming updates will help clarify subtle morphologic features that prevent misclassification. We follow the WHO classification closely and will reassess our reporting once the new edition is released, though I expect only minor adjustments.
Regarding practice guidelines, since we are in the United States, we primarily follow NCCN rather than ESMO. One notable update is the recommendation for HER2 testing in endometrial cancer. Our institution follows this closely. NCCN recommends HER2 testing for serous carcinomas and carcinosarcomas, and encourages testing for all endometrial cancers with a TP53 mutation, regardless of histologic type. In my own practice, if I diagnose serous carcinoma, carcinosarcoma, clear cell carcinoma, or even certain endometrioid carcinomas with TP53 mutations, I perform HER2 immunohistochemistry without waiting for a clinician’s request, as this practice is supported by the guidelines. Overall, we stay current with WHO and NCCN updates and integrate them into our practice, and most clinicians in our institutions do the same.
How do current HER2 scoring systems in endometrial carcinoma differ from those used in breast and gastric cancer, since I know that the HER2 scoring system is not fully established in endometrial carcinoma?
HER2 scoring in endometrial carcinoma is somewhat established now, primarily through clinical trials enrolling patients with endometrial serous carcinoma. Initially, HER2 assessment used modified breast criteria. The main difference is the cutoff for HER2 positivity: in breast cancer, a 3+ result requires at least 10% of tumor cells with strong circumferential membrane staining, whereas in endometrial cancer, a 3+ result requires 30% of tumor cells due to the more heterogeneous HER2 expression in these tumors. Gastric criteria are also used, specifically when patients are being treated with trastuzumab deruxtecan (T-DXd). These criteria differ slightly between biopsies and surgical specimens, and have been applied to not only endometrial but also other gynecologic cancers to determine eligibility for T-DXd.
The key point is that which criteria are used depends on the intended therapy: trastuzumab uses endometrial-specific criteria with the 30% cutoff, while T-DXd uses gastric criteria which differs between biopsies and surgical specimens. For ovarian and cervical cancers, it is usually safe to assume gastric criteria will be applied, but for endometrial cancers, it’s essential to confirm with the clinician. One of the major challenges in standardizing HER2 assessment is that many pathologists and clinicians get confused by the different criteria. Reports sometimes simply state “HER2 2+” or “3+” without specifying which criteria were used, which can have major implications for treatment, because a 1+ score in one system might be a 2+ score in the other, etc. It’s therefore critical that the pathologist knows which scoring system to apply based on the intended therapy, and that clinicians understand the nuances of HER2 evaluation. A careful clinician should look for these details in the pathology report, because the treatment plan depends on it, and a report that simply says “HER2 positive” without context is simply not sufficient.
And in HER2 2+ cases, do you proceed with FISH without a clinician’s request?
For HER2 2+ cases, we only perform FISH if we are using endometrial-specific criteria, to determine eligibility for trastuzumab therapy. This is an important distinction because the clinical trial that established these endometrial-specific criteria used modified breast criteria, where 2+ cases are always reflexed to in situ hybridization, just like in breast cancer. So when we assess HER2 using endometrial-specific criteria for trastuzumab eligibility, we do perform FISH in tumors with 2+ scores.
However, when using gastric criteria for T-DXd eligibility, it is important to remember that the DESTINY Pan-Tumor trial did not require FISH for 2+ cases; patients were treated based on HER2 protein expression alone—tumors with both 3+ and 2+ were included, and while tumors with 3+ scores responded better, even tumors with 2+ scores showed some response which indicates that the response is independent of HER2 gene amplification. So when applying gastric criteria for determining T-DXd eligibility we do not perform FISH in tumors with 2+ scores. It really comes down to the intended anti-HER2 targeted therapy and the scoring criteria used.
How have molecular classifications such as the TCGA system improved diagnostic accuracy and prognostication in endometrial cancer?
The main contribution of the TCGA system is prognostic rather than diagnostic. It does not change how we make the traditional morphologic histotype diagnoses; we still use the same morphologic criteria. TCGA defines four molecular subtypes: POLE-mutated, MSI-high, Copy-number low, and
Copy-number high (often p53-abnormal). The molecular classification is clinically significant because these subtypes have distinct progression-free survival outcomes. It has also been shown that different histologic types of endometrial cancer—serous, endometrioid, and other rare subtypes—can fall into any of the four molecular groups. So the TCGA molecular classification provides an additional prognostic layer that complements, rather than replaces, the traditional morphologic diagnosis. The ideal approach is to report both, for example: endometrioid carcinoma, grade 3; molecular subtype: POLE-mutated or MSI-high, etc. Both elements together give the most accurate picture for optimal patient management.
Are you also doing POLE mutation testing along with HER2, or does it need to be requested separately by the physician?
That’s a great question. Unfortunately, at this time we do not perform POLE testing in every endometrial cancer, mainly because of reimbursement issues. In the U.S., insurance companies generally won’t cover sequencing for all patients without a clear clinical indication. Some also argue that if knowing a POLE mutation status wouldn’t change management, routine testing isn’t justified. Although we don’t test universally, if we see morphologic features suggestive of a POLE mutation, we mention this in the report. Clinicians can then request sequencing for those selected patients. So at our institution, pathologists help guide POLE testing in a targeted way. Other centers may have broader access to sequencing depending on funding or state-specific resources, but there is still a significant variability. I agree that this may change in the coming years, especially as more clinicians are requesting POLE testing, even though insurance coverage remains inconsistent.
- What’s the selection criteria when deciding to perform POLE mutation testing? Several features, while not pathognomonic, suggest a POLE mutation:
- Serous-like areas within an otherwise endometrioid carcinoma, a morphology that doesn’t fully fit the endometrioid morphology.
- Abundant tumor-infiltrating or peritumoral lymphocytes.
- Scattered tumor giant cells with markedly enlarged, pleomorphic nuclei that are not typical of serous carcinoma but stand out within an endometrioid background.
If I see all three features, I document them and state that they are suggestive of a POLE mutation and recommend sequencing. I may also contact the clinician, so they can arrange testing. One major barrier to routine sequencing for all endometrial cancers is the need for patient consent. We don’t just test POLE, we use broader next-generation sequencing panels that can detect potential germline variants, which requires patient consent. Insurance approval is usually required as well. This is different from fusion testing in sarcomas, which is purely diagnostic and does not involve germline findings, so consent and preauthorization are not needed.
This leads me to the next question about the diagnostic pitfalls in distinguishing between endometrioid and serous endometrial carcinoma in histopathology.
Yes, this is a well-known diagnostic challenge because endometrioid and serous carcinomas can mimic each other. Traditionally, serous carcinomas show characteristic architectural patterns—papillary, solid, or glandular. Papillary growth was once considered defining, but we now know that not all serous carcinomas are papillary, and not all papillary tumors are serous. A serous carcinoma with a predominant glandular pattern can closely resemble an endometrioid carcinoma. The key is cytologic atypia. Even in gland-forming serous carcinomas, the tumor cells show much higher grade atypia than we would accept for endometrioid carcinoma. This is why I always teach residents and fellows to first assess architecture at low power, then evaluate cytology at high power, and ask whether the two are concordant. Glandular architecture with disproportionately high grade atypia should raise suspicion for serous carcinoma mimicking endometrioid carcinoma. Immunohistochemistry can help: loss of MMR proteins, PTEN, and/or ARID1A favoring endometrioid carcinoma, and abnormal p53 expression with retained MMR, PTEN, and ARID1A favoring serous carcinoma.
However, we must remember that some endometrioid carcinomas are MMR-proficient, retain PTEN/ARID1A, yet may still have TP53 mutations. So morphology remains essential, and immunostains are supportive, not definitional. Diagnosis always relies on integrating architecture, cytology, and immunohistochemistry.
In case of ambiguous morphology and discordant immunostaining how do you approach the final diagnosis, like do you lean more on molecular findings or morphologic impression?
Ambiguous morphology often occurs in high-grade endometrial carcinomas, where a solid growth pattern can mimic serous,clear cell or grade 3 endometrioid carcinoma In such cases, we typically perform a panel of immunostains—MMR proteins, p53, PTEN, or ARID1A, and sometimes Napsin A for clear cell carcinoma. However, these stains are not always definitive. For example, MMR deficiency as well as ARID1A loss may occur in both endometrioid and clear cell carcinomas. So if morphology and immunostains remain inconclusive, it is acceptable to report the tumor descriptively as “high-grade endometrial carcinoma” and recommend sequencing and definitive subtyping in a hysterectomy. Even sequencing may not always clarify the subtype; for example, a copy-number low tumor could still be an endometrioid or clear cell carcinoma. Pragmatically, as long
as the endometrial carcinoma is recognized as high-grade, clinical management will be similar. The patient will undergo appropriate surgery—hysterectomy with bilateral salpingo-oophorectomy, lymph node assessment, and possibly omentectomy—so it is not harmful to report descriptively when a definitive subtype cannot be established. The key is careful morphologic examination and immunohistochemical studies; molecular findings can help but do not override morphologic impression when the data are inconclusive.
Are current staging systems adequate for reflecting the biological differences among endometrial cancer histologic types?
FIGO 2023 is the most recent staging system, replacing FIGO 2009. FIGO 2023 is drastically different as it incorporated aggressive vs. non-aggressive histologic subtypes, lymphovascular invasion, and molecular features—including TCGA subtypes. Therefore, application of FIGO 2023 staging requires the knowledge of extensive information, especially molecular data, which many institutions cannot routinely obtain. There was substantial backlash after FIGO 2023 was released. Only one pathologist was involved in its development, and many institutions in the U.S., including mine, do not use FIGO 2023 for staging endometrial carcinomas. When FIGO 2023 system first came out, the College of American Pathologists (CAP) immediately updated the endometrial cancer synoptic protocol to include only FIGO 2023, removing FIGO 2009. After widespread criticism—and once I joined the CAP Cancer Committee, we reinstated FIGO 2009 staging as an optional reporting element, while also keeping FIGO 2023 staging as an optional element. The essential system the pathologists are required to apply is AJCC pTNM staging. FIGO staging, whether 2009 or 2023, is optional and primarily the clinician’s responsibility. Do these staging systems reflect biological differences? FIGO 2009 is based on traditional, well-validated prognostic factors such as depth of myometrial invasion, cervical stromal or uterine serosal involvement, and adnexal metastases. These criteria remain biologically meaningful and are supported by robust data.
FIGO 2023, while attempting to incorporate molecular findings, cannot accurately reflect true biology when many institutions lack the complete molecular data. Applying FIGO 2023 staging using only partial data risks misclassification. FIGO 2023 “downstages” POLE-mutated cancers because of their excellent prognosis, creating an opportunity for de-escalating chemotherapy. However, many clinicians are not yet comfortable withholding chemotherapy from patients who have historically benefited from it, especially without stronger prospective data. So while the intent is biologically sound, the evidence is not mature enough for widespread practice change. In summary, FIGO 2009 remains reliable and biologically robust. FIGO 2023 aims to incorporate molecular findings but remains difficult to apply and may not accurately reflect tumor biology without comprehensive data. More evidence is needed before molecularly driven staging can confidently guide major treatment decisions.
I would also like to address the following topic: What are the main barriers to implementing universal Lynch syndrome screening among women with endometrial cancer? And could you make a small parallel between the United States and Georgia?
I have to say upfront that I don’t want to present inaccurate information about Lynch syndrome screening in Georgia, because I’m really not sure how universally it’s done there. I hope that at least
mismatch repair (MMR) immunochemistry is being performed reflexively on every endometrial cancer but I honestly don’t know. There are so many different clinics and pathology laboratories, and I’m not closely familiar with how many patients are actually getting screened. I can, however, speak about the status of Lynch syndrome screening in the US. I think we’ve made really good progress over the past few years. So in 2025, it’s hard to imagine a pathology department here that doesn’t perform mismatch repair immunochemistry either in-house or have a workflow to send it out.. In consultation cases, we sometimes see that the specimen is being sent out for MMR immunochemistry, with results to be reported in an addendum, indicating that the pathology department is aware of the significance of performing MMR immunohistochemistry even if they don’t have the tests in-house or don’t have enough cases per year to justify setting up the assay themselves. I always smile when I see that workflow in place—it’s a sign that universal screening is being taken seriously.
I also think there should be no age cutoff for testing. Some countries such as Canada used 70 years as a cutoff for MMR testing a few years ago; the yield of diagnosing Lynch syndrome in older women is lower, not zero. If we’re talking about universal Lynch syndrome screening, everyone with newly diagnosed endometrial cancer should be tested, regardless of age.
I really don’t know how things work in Georgia. Inter-laboratory practices most likely vary, and it would be impossible for me to say whether every pathology laboratory is performing universal Lynch screening. I’m genuinely curious to learn more.
Regarding somatic versus germline testing, most laboratories start with mismatch repair immunochemistry as a screening tool. You can’t predict germline mutations perfectly, but the pattern of protein loss gives you a clue. For example, loss of MLH1 and PMS2 is usually sporadic, not Lynch syndrome related, whereas isolated loss of PMS2, or loss of MSH6 and/or MSH2, would be more suggestive of a germline mutation. That’s why we screen this way: not every patient needs germline testing, only selected patients need to undergo germline evaluation. Otherwise, we’d be doing unnecessary germline testing for a lot of patients who don’t need it.
And in your experience, what practical steps could help smaller or resource-limited hospitals to adopt routine Lynch syndrome screening more efficiently?
I think the first step is really just knowing that it’s important. As long as a hospital has the ability to perform immunohistochemistry, whether it’s automated or manual, they could implement Lynch syndrome screening. Most laboratories already run other immunostains, so adding MLH1, PMS2, MSH2, and MSH6 stains would just be a matter of bringing in the antibodies and, optimizing them. It’s not necessarily simple, because they need the capacity to run immunohistochemistry and the knowledge to interpret the stains correctly. However, fundamentally, if a laboratory can perform immunostains, it can adopt routine Lynch syndrome screening. Education and awareness are key—pathologists need to know that Lynch syndrome screening is required for every endometrial cancer patient, and should be familiar with classic and uncommon staining patterns for accurate interpretation. Once those pieces are in place, it becomes much more feasible, even in smaller or resource-limited hospitals.
What do you think about the artificial intelligence influencing histopathological analysis in your field in this century, since it really gained popularity in the last two years?
Yes, I think that’s the million-dollar question. Artificial intelligence (AI) is already influencing our field, and it’s only going to increase. Some people feel threatened and worry that pathologists might lose their jobs, but I don’t really believe that. I see AI as a tool we can use to become more efficient and maybe even more accurate in certain tasks—such as screening lymph nodes and detecting isolated tumor cells or micrometastases. The problem is relying entirely on AI without human oversight. You don’t really know what the algorithm is doing on its own, and that can be dangerous. Right now, in 2025, I can’t see AI completely replacing pathologists. But as the saying goes, AI won’t replace pathologists—it will replace pathologists who don’t use AI. I think that’s true. It’s important for us to learn how to use these tools to improve accuracy and efficiency. Avoiding AI altogether isn’t the answer. When properly trained, these AI algorithms can be extremely helpful—for pathologists and ultimately for patient care.
What exact aspects of pathology do you think AI will be less likely to replace?
AI could work well for screening common, straightforward things, but when it comes to complex neoplastic cases, that’s a different story. Cases where tumors are unusual, involve multiple sites, or the subtype is unclear, especially when molecular testing is inconclusive or doesn’t fit the morphology, require experience and expertise. You need someone who can think critically about the final classification of the tumor. Depending on your practice, the number of these challenging cases can be quite high. For those situations, I think we will still need human pathologists to make the final call.
