Elisabetta Xue

The clinical significance of human herpesvirus 6 (HHV-6) after allogeneic hematopoietic cell transplant (HCT) is debated. We prospectively collected weekly whole-blood HHV-6 in 217 posttransplantation cyclophosphamide (PTCy)–based HCT recipients. Overall, 149 patients (69%) had HHV-6 detected, at a median of 21 days after HCT; in 63 (29%) patients, HHV-6 DNAemia spiked to ≥4 × log10 copies per mL (high-level HHV-6 DNAemia) at median day +27. Hematologic malignancy, myeloablative conditioning, HLA-partially mismatched donor, marrow graft, PTCy dose, and sirolimus use were univariates associated with high-level HHV-6 DNAemia. All spikes precipitously declined spontaneously, but after initial decline, 27% persisted at ≥3 × log10 copies per mL for >1 month. Those with HHV-6 spikes had lower day +28 CD4+ T-cell counts. During initial spikes, there was a strong correlation between plasma and whole-blood HHV-6 DNAemia. By contrast, in patients with HHV-6 persisting for >1 month, HHV-6 was predominantly found within CD4+ T cells but not in plasma. Only 1 patient developed HHV-6–associated encephalitis. High-level HHV-6 DNAemia was associated with significantly increased rates of otherwise unexplained fever (35% vs 21% in patients without high-level HHV-6 DNAemia), rash (27% vs 12%), transaminitis (52% vs 27%), and gastrointestinal symptoms (23% vs 7%), as well as increased risk of acute (36% vs 18%) and moderate-to-severe chronic graft-versus-host disease (GVHD; 22% vs 8%). HHV-6 detection occurs frequently in PTCy recipients but infrequently leads to classic HHV-6–related disease (encephalitis), although it associates with various self-limited clinical symptoms, may play a role in shaping immune reconstitution, and may serve as a marker for increased GVHD risk.

Purpose: Olfactory neuroblastoma (ONB) is a rare sinonasal malignancy with limited therapeutic options in the recurrent/metastatic setting; little is known regarding its responsiveness to immunotherapy. Inhibition of enhancer of zeste homolog 2 (EZH2) has been shown to improve T‐cell‐mediated killing and susceptibility to immune checkpoint inhibitors in a variety of cancers. We aimed to evaluate the expression and activity of EZH2 in ONB and its association with immune characteristics. Materials and Methods: We studied a cohort of 36 ONB real‐world patient samples that underwent molecular profiling at a centralized lab (Caris Life Science). To infer EZH2 methyltransferase activity, we adopted an EZH2 gene repression signature (ERS) score: ONB samples were stratified into ERS‐low and ERS‐high subgroups, corresponding to high and low inferred EZH2 methyltransferase activity, respectively. Transcriptomic data were utilized to calculate the T‐cell‐inflamed (TCI) score and mitogen‐activated protein kinase (MAPK) pathway activation score (MPAS). Tumor immune microenvironment composition was inferred from tumor‐derived bulk RNA sequencing data. We analyzed immunologic differences between ERS‐low and ERS‐high ONB. Results: In ERS‐high ONB, we observed a higher expression of immune‐related genes, a higher proportion of TCI tumors, and an enrichment in inflammatory pathways. ERS‐high ONB also displayed increased macrophages, and to a lesser extent, B cells and CD8 ⁺ T cell infiltration in the tumor microenvironment. Also, ERS‐high was associated with increased MPAS, potentially identifying ONB with increased susceptibility to MAPK inhibitors. These data were confirmed in an independent validation cohort using a publicly available dataset. Conclusions: Taken together, our data suggest that low EZH2 activity is associated with a more immunogenic microenvironment, paving the way for potential combinations of EZH2 inhibitors with checkpoint blockade in ONB.

Cytomegalovirus (CMV) infection post‐hematopoietic cell transplantation (HCT) remains a significant cause of morbidity and mortality. While letermovir prophylaxis is available for CMV‐seropositive recipients, optimal donor selection for CMV‐seronegative recipients remains unclear, with donor age often prioritized over CMV serostatus. We investigated the relative impact of donor age and CMV serostatus in CMV‐seronegative recipients (n = 1013) with either CMV‐seropositive (n = 318) or CMV‐seronegative donors (n = 695), who underwent HCT with HLA‐matched sibling donors with calcineurin inhibitor‐based or post‐transplant cyclophosphamide (PTCy)‐based graft‐versus‐host disease prophylaxis, or haploidentical donors with PTCy. Multiple conventional and machine‐learning approaches were employed to account for confounding factors. Across all analyses, CMV‐seropositive donor status was consistently associated with significantly inferior overall survival (OS) and disease‐free survival, primarily driven by increased non‐relapse mortality (NRM). Older donor age showed a statistically significant association with OS and DFS in some but not all models, and its effect size was small (about 1% increased hazard per year in Cox proportional hazard model) compared to the substantial impact of CMV‐seropositive donors (about 27%–33% increased hazards for worse OS, approximately 50%–60% higher NRM in Cox proportional hazard model). However, our machine learning model revealed a more nuanced and complex non‐linear effect of donor age, further demonstrating the adverse impact of donor CMV serostatus. Our findings suggest that prioritizing a CMV‐seronegative donor may be associated with improved outcomes in CMV‐seronegative recipients. Further research is needed to validate these findings and explore underlying mechanisms.

Olfactory neuroblastoma (ONB), sinonasal undifferentiated carcinoma (SNUC), and sinonasal neuroendocrine carcinoma (SNEC) are rare malignancies arising from the sinonasal tract with limited therapeutic options. The expression of the somatostatin receptor 2 gene (SSTR2), which is expressed in other neuroendocrine neoplasms and is therapeutically actionable, has been reported in these tumors. Here, we analyzed SSTR2 gene expression and its associations with genomic features, established biomarkers predicting of immune response, and the tumor immune microenvironment in a cohort of ONB, SNUC, and SNEC tumor samples (26, 13, and 8 samples, respectively) from a real-world database. SSTR2 gene expression was high in neural-type ONB and low in basal-type ONB and in most of the SNUC and SNEC cases; there was no difference in expression between primary and metastatic tumors. The T cell-inflamed (TCI) score analysis classified 38.5% of SNUC cases as T cell-inflamed compared to only 3.9% of ONB and 0% of SNEC cases; 26.9% of ONB cases were classified as intermediate TCI; and SNEC had the lowest relative immune cell infiltration by deconvolution. In high SSTR2-expressing ONB, there was a higher proportion of infiltrating of Natural Killer cells and dendritic cells by deconvolution. Additionally, high SSTR2-expressing ONB was enriched for proliferation pathways, including E2F and Myc targets and G2M checkpoints. In conclusion, our findings delineate significant differences between these three types of sinonasal malignancies that were examined. In ONB, relative to SNUC and SNEC, the SSTR2 expression profile, combined with its immune profiles, indicates potential novel therapeutic strategies and combinations for this unmet clinical need. Conversely, the inflammatory microenvironment of SNUC may be targetable using immuno-oncologic therapies.

Background: Artificial intelligence is increasingly being applied to many workflows. Large language models (LLMs) are publicly accessible platforms trained to understand, interact with, and produce human-readable text; their ability to deliver relevant and reliable information is also of particular interest for the health care providers and the patients. Hematopoietic stem cell transplantation (HSCT) is a complex medical field requiring extensive knowledge, background, and training to practice successfully and can be challenging for the nonspecialist audience to comprehend. Objective: We aimed to test the applicability of 3 prominent LLMs, namely ChatGPT-3.5 (OpenAI), ChatGPT-4 (OpenAI), and Bard (Google AI), in guiding nonspecialist health care professionals and advising patients seeking information regarding HSCT. Methods: We submitted 72 open-ended HSCT–related questions of variable difficulty to the LLMs and rated their responses based on consistency—defined as replicability of the response—response veracity, language comprehensibility, specificity to the topic, and the presence of hallucinations. We then rechallenged the 2 best performing chatbots by resubmitting the most difficult questions and prompting to respond as if communicating with either a health care professional or a patient and to provide verifiable sources of information. Responses were then rerated with the additional criterion of language appropriateness, defined as language adaptation for the intended audience. Results: ChatGPT-4 outperformed both ChatGPT-3.5 and Bard in terms of response consistency (66/72, 92%; 54/72, 75%; and 63/69, 91%, respectively; P=.007), response veracity (58/66, 88%; 40/54, 74%; and 16/63, 25%, respectively; P<.001), and specificity to the topic (60/66, 91%; 43/54, 80%; and 27/63, 43%, respectively; P<.001). Both ChatGPT-4 and ChatGPT-3.5 outperformed Bard in terms of language comprehensibility (64/66, 97%; 53/54, 98%; and 52/63, 83%, respectively; P=.002). All displayed episodes of hallucinations. ChatGPT-3.5 and ChatGPT-4 were then rechallenged with a prompt to adapt their language to the audience and to provide source of information, and responses were rated. ChatGPT-3.5 showed better ability to adapt its language to nonmedical audience than ChatGPT-4 (17/21, 81% and 10/22, 46%, respectively; P=.03); however, both failed to consistently provide correct and up-to-date information resources, reporting either out-of-date materials, incorrect URLs, or unfocused references, making their output not verifiable by the reader. Conclusions: In conclusion, despite LLMs’ potential capability in confronting challenging medical topics such as HSCT, the presence of mistakes and lack of clear references make them not yet appropriate for routine, unsupervised clinical use, or patient counseling. Implementation of LLMs’ ability to access and to reference current and updated websites and research papers, as well as development of LLMs trained in specialized domain knowledge data sets, may offer potential solutions for their future clinical application.

Invasive fungal diseases (IFDs) still represent a relevant cause of mortality in patients affected by hematological malignancies, especially acute myeloid leukaemia (AML) and myelodysplastic syndrome (MDS) undergoing remission induction chemotherapy, and in allogeneic hematopoietic stem cell transplantation (allo-HSCT) recipients. Mold-active antifungal prophylaxis (MAP) has been established as a standard of care. However, breakthrough IFDs (b-IFDs) have emerged as a significant issue, particularly invasive aspergillosis and non-Aspergillus invasive mold diseases. Here, we perform a narrative review, discussing the major advances of the last decade on prophylaxis, the diagnosis of and the treatment of IFDs in patients with high-risk neutropenic fever undergoing remission induction chemotherapy for AML/MDS and allo-HSCT. Then, we present our single-center retrospective experience on b-IFDs in 184 AML/MDS patients undergoing high-dose chemotherapy while receiving posaconazole (n = 153 induction treatments, n = 126 consolidation treatments, n = 60 salvage treatments). Six cases of probable/proven b-IFDs were recorded in six patients, with an overall incidence rate of 1.7% (6/339), which is in line with the literature focused on MAP with azoles. The incidence rates (IRs) of b-IFDs (95% confidence interval (95% CI), per 100 person years follow-up (PYFU)) were 5.04 (0.47, 14.45) in induction (n = 2), 3.25 (0.0013, 12.76) in consolidation (n = 1) and 18.38 (3.46, 45.06) in salvage chemotherapy (n = 3). Finally, we highlight the current challenges in the field of b-IFDs; these include the improvement of diagnoses, the expanding treatment landscape of AML with molecular targeted drugs (and related drug–drug interactions with azoles), evolving transplantation techniques (and their related impacts on IFDs’ risk stratification), and new antifungals and their features (rezafungin and olorofim).

BACKGROUND Artificial intelligence is increasingly being applied to many workflows. Large language models (LLMs) are publicly accessible platforms trained to understand, interact with, and produce human-readable text; their ability to deliver relevant and reliable information is also of particular interest for the health care providers and the patients. Hematopoietic stem cell transplantation (HSCT) is a complex medical field requiring extensive knowledge, background, and training to practice successfully and can be challenging for the nonspecialist audience to comprehend. OBJECTIVE We aimed to test the applicability of 3 prominent LLMs, namely ChatGPT-3.5 (OpenAI), ChatGPT-4 (OpenAI), and Bard (Google AI), in guiding nonspecialist health care professionals and advising patients seeking information regarding HSCT. METHODS We submitted 72 open-ended HSCT–related questions of variable difficulty to the LLMs and rated their responses based on consistency—defined as replicability of the response—response veracity, language comprehensibility, specificity to the topic, and the presence of hallucinations. We then rechallenged the 2 best performing chatbots by resubmitting the most difficult questions and prompting to respond as if communicating with either a health care professional or a patient and to provide verifiable sources of information. Responses were then rerated with the additional criterion of language appropriateness, defined as language adaptation for the intended audience. RESULTS ChatGPT-4 outperformed both ChatGPT-3.5 and Bard in terms of response consistency (66/72, 92%; 54/72, 75%; and 63/69, 91%, respectively; P=.007), response veracity (58/66, 88%; 40/54, 74%; and 16/63, 25%, respectively; P<.001), and specificity to the topic (60/66, 91%; 43/54, 80%; and 27/63, 43%, respectively; P<.001). Both ChatGPT-4 and ChatGPT-3.5 outperformed Bard in terms of language comprehensibility (64/66, 97%; 53/54, 98%; and 52/63, 83%, respectively; P=.002). All displayed episodes of hallucinations. ChatGPT-3.5 and ChatGPT-4 were then rechallenged with a prompt to adapt their language to the audience and to provide source of information, and responses were rated. ChatGPT-3.5 showed better ability to adapt its language to nonmedical audience than ChatGPT-4 (17/21, 81% and 10/22, 46%, respectively; P=.03); however, both failed to consistently provide correct and up-to-date information resources, reporting either out-of-date materials, incorrect URLs, or unfocused references, making their output not verifiable by the reader. CONCLUSIONS In conclusion, despite LLMs’ potential capability in confronting challenging medical topics such as HSCT, the presence of mistakes and lack of clear references make them not yet appropriate for routine, unsupervised clinical use, or patient counseling. Implementation of LLMs’ ability to access and to reference current and updated websites and research papers, as well as development of LLMs trained in specialized domain knowledge data sets, may offer potential solutions for their future clinical application.