Maher A. Sughayer

Acute leukemia comprises heterogeneous hematologic malignancies. CD117 (c-KIT), encoded by the KIT proto-oncogene, is a receptor tyrosine kinase involved in hematopoietic cell survival, proliferation, and differentiation. CD117 expression was assessed in 118 acute myeloid leukemia (AML) and 106 acute lymphoblastic leukemia (ALL) cases and correlated with clinical, immunophenotypic, molecular, and survival parameters. CD117 was expressed in 82.2% of AML and 17.9% of ALL cases. In AML, CD117 positivity was associated with AML-M2 (p < 0.001), core-binding factor (CBF) abnormalities (p = 0.010), and favorable European LeukemiaNet (ELN) 2022 risk (p = 0.041). CD117 independently predicted improved overall survival (OS) in CBF-adjusted (aHR = 0.450, 95% CI: 0.219–0.924, p = 0.030) and ELN-adjusted models (aHR = 0.295, 95% CI: 0.127–0.684, p = 0.004). CD117 was associated with improved event-free survival (EFS) only in the ELN-adjusted model (aHR = 0.374, 95% CI: 0.159–0.882, p = 0.025). In ALL, CD117 was associated with T-cell lineage (p < 0.001) and independently predicted inferior OS (aHR = 4.171, 95% CI: 1.375–12.656, p = 0.012) and EFS (aHR = 3.145, 95% CI: 1.153–8.576, p = 0.025). In T-ALL, it predicted poorer OS (aHR = 3.316, 95% CI: 1.043–10.537, p = 0.042). In conclusion, CD117 has disease-specific prognostic significance, predicting favorable outcomes in AML but adverse outcomes in ALL, particularly T-ALL.

Claudin 18, a tight junction protein, has emerged as a therapeutic target in various gastrointestinal cancers, notably gastric adenocarcinoma. However, its prevalence and clinicopathological significance in colorectal cancer (CRC) remain incompletely understood. In the current single-center, retrospective, immunohistochemical study, the prevalence and prognostic value of Claudin-18 was investigated in CRC. Claudin-18 expression was detected in 15.9% (17/107) of CRC cases, with negative/weak expression in 84.1% (90/107). No significant associations were observed between Claudin-18 expression and age, sex, tumor site, T stage, lymph node status, distant metastasis, TNM stage, histological subtype, or lymphovascular invasion (all p > 0.05). No significant associations were identified between Claudin-18 expression and age, sex, tumor location, histological subtype, or TNM stage. Furthermore, Claudin-18 expression independently predicted survival outcomes in CRC where Kaplan–Meier analysis demonstrated significantly worse overall survival (OS) in patients with strong/moderate Claudin-18 expression compared with Claudin-18 -negative/weak expression (log rank test, p = 0.024). Mean OS was markedly longer in Claudin-18 -negative/weak expression group (121.0 vs.76.6 months). Progression-free survival (PFS) analysis demonstrated a trend toward poorer outcomes among patients with increased Claudin-18 expression; however, the difference did not reach statistical significance (log-rank test, p = 0.066). Event-free survival (EFS) was also significantly reduced in patients with strong/moderate Claudin-18 expression (log-rank test, p = 0.022). Mean EFS was markedly longer in CLDN18 negative/weak expression (93.1 vs. 31.1 months). Multivariable binary logistic regression showed that the overall model was statistically significant (χ2(7) = 22.657, p = 0.002). However, none of the evaluated variables—including gender, age, lymphovascular invasion, tumor laterality, TNM stage, MSI status, histological type, and Claudin-18 expression—was independently associated with the outcome (all p > .05). Claudin-18 expression was not a significant predictor in the adjusted model (p = 0.931; OR = 0.63). In conclusion, despite its low prevalence in the current cohort, Claudin-18 expression was associated with poorer overall survival; however, its independent prognostic value was not established.

Background: General-purpose large language models (LLMs) are increasingly evaluated in diagnostic pathology, but prior studies have largely emphasized diagnostic accuracy rather than how models fail. We evaluated four LLMs for diagnostic performance, pathology-relevant errors and hallucinations, their burden, and potential clinical impact across multi-organ pathology cases. Design: In this retrospective multicenter study, 153 pathology cases from two institutions spanning 20 organs were evaluated using ChatGPT-5.3 (LLM1), Gemini 3 (LLM2), Grok 4.20 (LLM3), and Claude Opus 4.6 (LLM4). Each LLM received multi-magnification histologic images with clinical context and generated a microscopic description and diagnosis. No data splitting, training, or fine-tuning was performed. Twenty-one pathologists assessed 612 outputs for diagnostic correctness, error and hallucination type and burden, clinical impact (0–4), and overall performance (1–5). Results: Strict diagnostic accuracy was 48.9% overall (60.1% including partially correct diagnoses) and ranged from 36.6% to 58.2% across LLMs. Errors occurred in 82.0% of outputs and hallucinations in 76.6%; 90.2% contained at least one error or hallucination. Misinterpretation was the most frequent error (72.1%), while fabricated histologic features were the dominant hallucination type (74.5%). LLM2 had significantly lower misinterpretation rates than the other three LLMs, while LLM3 had significantly higher fabricated-feature hallucination rates than LLM1 and LLM2. Errors were more frequent in incorrect than correct diagnoses (97.5% vs 66.2%), as were hallucinations (95.1% vs 59.9%; both p < 0.001). Every strictly incorrect diagnosis contained an error and/or hallucination, while 79.9% of strictly correct diagnoses also contained at least one. In multivariable analysis, LLM3 was independently associated with higher odds of an incorrect diagnosis. Diagnostic correctness was the strongest determinant of high clinical impact; each one-point increase in hallucination burden increased the odds by 63% (adjusted OR, 1.63). Conclusion: Diagnostic accuracy alone substantially underestimates the safety limitations of general-purpose LLMs in histopathology. Errors and hallucinations were common even when the final diagnosis was correct, and their burden was independently associated with clinical impact. Evaluation frameworks should therefore assess both diagnostic correctness and the reliability and potential consequences of accompanying generated content.

Background: Immunohistochemistry (IHC) remains the gold standard for assessing key breast cancer biomarkers such as estrogen receptor (ER) and progesterone receptor (PR). However, the staining process is resource-intensive, time-consuming, and dependent on specialized laboratory infrastructure. While recent advances in computational pathology have demonstrated the feasibility of predicting biomarker status directly from hematoxylin and eosin (H&E) stained tissue, most models lack visual interpretability, limiting clinical trust and adoption. Methods: We propose a two-stage deep learning pipeline for virtual immunohistochemistry. In Stage 1, a regression-based convolutional neural network predicts continuous ER/PR positivity from H&E image patches trained using a mean-squared-error objective; an Edge-Focused Loss was explored as an alternative but was not adopted. In Stage 2, the predicted biomarker percentage is incorporated as a spatial condition into a conditional generative adversarial network (cGAN), enabling the synthesis of virtual IHC images. The framework was evaluated on the IHC4BC dataset using 5-fold patient-level cross-validation. Results: The predictive model achieved a mean absolute error (MAE) of 9.14% for ER and 6.10% for PR under 5-fold patient-level cross-validation; treated as a clinical positive/negative classifier it reached an area under the ROC curve of 0.92 for ER and 0.74 for PR. The generative stage produced visually plausible virtual IHC images with biologically meaningful nuclear staining patterns. Although pixel-based metrics (SSIM and PSNR) favored the unconditioned baseline, domain-specific analysis of the deconvolved DAB channel showed substantially improved biomarker-specific staining fidelity for the conditional framework. For PR, the correlation between the generated DAB signal and ground-truth positivity increased from 0.32 (unconditioned baseline) to 0.61 in the deployed pipeline, reaching 0.82 under oracle conditioning, illustrating both the benefit of conditioning and the residual error propagated from Stage 1. Conclusion: This proof-of-concept study presents a dual-output computational framework that integrates quantitative biomarker prediction with interpretable virtual staining. The proposed approach may support preliminary biomarker assessment and workflow prioritization, particularly in resource-limited settings. However, external validation and domain-specific evaluation metrics are required before clinical application.

Introduction: Parasitic infections are prevalent in developing countries, and cancer patients with weakened immune systems are particularly vulnerable. This study examined the prevalence of protozoan parasites among cancer patients. Methodology: A total of 300 stool and heparinized plasma samples were collected from cancer patients at the King Hussien Cancer Center (KHCC). Cryptosporidium, Giardia duodenalis, and Blastocystis were detected by microscopy. Immunodiagnostic tests included electrochemiluminescence immunoassay (ECLIA) for Toxoplasma gondii IgG and IgM antibodies in plasma samples, and the ImmunoCard STAT Cryptosporidium/Giardia Kit for G. duodenalis and Cryptosporidium detection in stool. DNA extraction from stool samples was followed by nested polymerase chain reaction (PCR) to confirm the presence of intestinal parasites. Results: Toxoplasma gondii, Cryptosporidium, G. duodenalis, and Blastocystis were detected. T. gondii was found in 22% of patients via IgG antibodies and in 2.7% via IgM, with the highest IgG seropositivity in multiple myeloma and uterine cancer patients (50%), and the highest IgM seropositivity in multiple myeloma patients (12.5%). The prevalence of Cryptosporidium varied depending on the detection method: 8.3% by microscopy, 11% by immunodiagnostic tests, and 12.3% by PCR. The highest infection rate was among colorectal cancer (CRC) patients. Giardia duodenalis was detected at rates of 1.7% by microscopy, 2.0% by immunodiagnostic testing, and 0.7% by PCR. Blastocystis was most prevalent in CRC patients, with detection rates of 31.0% by microscopy and 48.3% by PCR. Conclusions: This study highlights the significance of protozoan parasitic infections among cancer patients, emphasizing the need for screening and management to improve patient outcomes.

Mesothelin is a surface glycoprotein overexpressed in several solid tumors and a known immunotherapy target. While present on blast cells in acute myeloid leukemia (AML), its role in acute lymphoblastic leukemia (ALL) remains unclear. This study evaluated mesothelin expression in 181 newly diagnosed acute leukemia cases using immunohistochemistry (IHC) on bone marrow biopsies and ELISA on plasma samples. Mesothelin was expressed in 31% of pediatric and 15% of adult AML cases (n=65), with no significant differences among AML subtypes. It was absent or rare in ALL (n=51) and undetectable in normal bone marrow. In AML, mesothelin was detected in 49% of CD38⁺, 60% of CD64⁺, 62.5% of KMT2A-rearranged, and 69% of core binding factor AML [83% with inv(16)/ t _(16;16) , 57% with t _(8;21) ]. It was largely absent in cases with NPM1 (87.5%), GATA2, and DNMT3A mutations. Expression correlated significantly with age, CD38, CD64, and mutations in NPM1, GATA2, DNMT3A, and KDM6A. No significant association was found between mesothelin expression and 5-year overall or event-free survival, measurable residual disease, remission, or relapse. In ALL, 83.3% of T-ALL and 85% of B-ALL cases showed no expression. However, mesothelin correlated with CD11c, PHF6, and CBLC mutations in ALL. Soluble mesothelin was present in 17.6% of AML and 2.7% of ALL cases (all adults), but not in healthy individuals. In AML, it correlated with CD33 expression and inv(16)/ t _(16;16) . In conclusion, mesothelin is rare in ALL but enriched in specific AML subtypes and not prognostic for survival.

The ability to escape immune surveillance is a hallmark of malignancy. Programmed death ligand 1 (PD-L1) facilitates tumor progression by binding to the immune inhibitory receptor known as programmed cell death protein 1 (PD1) on immune cells, resulting in suppression of the cytotoxic T lymphocyte function. The degree of PD-L1 expression may have a prognostic value in some cancer types, and it may vary according to the genetic makeup and the ethnicity of patients. The expression level of PD-L1 in 63 cases of primary head and neck squamous cell carcinoma (HNSCC) tumor tissues was evaluated using immunohistochemistry (IHC). Also, PD-L1 association with various clinicopathologic characteristics and overall survival was studied. The positive expression rate of PD-L1 in HNSCC was 85.7%, 60.3%, and 52.3% of the total number of cases using combined positive score (CPS) ≥ 1, CPS ≥ 5, and CPS ≥ 20 cutoff values, respectively. Statistical analysis revealed no significant relationship between the expression of PD-L1 protein and clinicopathological features except for tobacco use using a cutoff CPS ≥ 20. The log-rank chi-square results showed that PD-L1 was not a significant factor affecting the 4-year overall survival of HNSCC patients. Also, the overall survival rate was not significantly affected by the patient's age, tumor differentiation, tumor size, and lymphovascular invasion. However, survival curves demonstrated lower overall survival in HNSCC female patients, disease recurrence, and positive perineural invasion. Our findings showed relatively high PDL-1 expression in most HNSCC patients. No significant association was found between PD-L1 protein expression and overall survival.

Background: The 21-gene recurrence score (Oncotype DX) guides adjuvant chemotherapy decisions in early-stage estrogen receptor (ER)-positive, human epidermal growth factor receptor-2 (HER2)-negative breast cancer. However, its high cost and limited availability motivate the development of simplified predictive models using routinely reported pathology parameters. The aim of this study was to develop and validate a practical, rule-based algorithm that predicts Oncotype DX recurrence score (RS) category using only histologic grade and progesterone receptor (PR) expression percentage. Methods: This retrospective study included 528 patients with ER+/HER2- early breast cancer who underwent Oncotype DX testing. Cases were randomly assigned to a learning (n = 377) and validation (n = 151) set. Univariate analysis and receiver operating characteristics (ROC) curves were used to determine PR% cut-offs within each histologic grade to stratify patients into low-risk (RS ≤ 25) or high-risk (RS > 25) categories. A stepwise algorithm was derived from these parameters and tested in the validation cohort. Results: Histologic grade and PR% were significantly associated with RS. Grade 1 tumors were uniformly low-risk regardless of PR%. In grade 2, PR ≥ 60% achieved 100% sensitivity for low RS; in grade 3, PR < 40% achieved 100% sensitivity for high risk. The algorithm confidently stratified ∼ 65% of cases. In the validation set, the model showed 87.5% sensitivity, 100% specificity, 100% positive predictive value (PPV), 99% negative predictive value (NPV), and 99% overall accuracy. Conclusion: This simplified algorithm accurately predicts Oncotype DX RS category using only histologic grade and PR%. It enables confident risk stratification in most patients without molecular testing, offering a low-cost, practical tool for clinical decision-making, particularly in resource-limited settings.

Objective: Tumor-infiltrating lymphocyte (TIL) assessment is recognized as an important prognostic and therapeutic biomarker in several tumors, including triple-negative breast cancer (TNBC). TILs face several challenges, including low to intermediate interobserver concordance. The use of tissue microarray (TMA) in TIL assessment and its relationship/concordance with whole sections (WS) has been studied only rarely. The aim of this study is to evaluate the reproducibility of TIL assessment in TNBC among various users and to compare TIL assessment using TMA versus whole tissue sections (WS). Material and Methods: Hematoxylin and eosin-stained sections of TMAs were prepared and assessed for quality and adequacy. Corresponding WS slides for all cases were retrieved and independently scored for TILs by one junior and two senior pathologists. The assessment followed the guidelines of the International TILs Working Group. TIL scores were categorized into three groups: Low: <20%; intermediate: 20–49%; and high: ≥50%. A total of 100 cases were collected; however, 76 cases were evaluable after excluding inadequate samples or cases with missing information. Results: Fleiss’ kappa was used to assess interobserver agreement among the three raters for both WS and TMA datasets. The Intraclass Correlation Coefficient (ICC) was used to assess the degree of agreement among the three raters for continuous variables before categorization, and Cohen’s kappa was applied to both WS and TMA datasets to assess interobserver agreement between any two raters. The interobserver agreement for WS analysis (Cohen’s kappa) was fair, while for TMA, it was moderate. The ICC also showed moderate agreement. Cohen’s kappa for TMA versus WS for each of the three observers ranged from fair to strong. Fleiss’ kappa showed fair agreement for WS versus TMA. Conclusion: Interobserver variability in TIL scoring remains a challenge. TMAs improve consistency but at the cost of reduced correlation with WS assessments.

Background: Information regarding the use of digital pathology (DP) in developing countries is limited. Additionally, the knowledge and attitudes/perceptions of pathologists are mainly unknown. In this study, we aim to assess the knowledge and attitudes of Jordanian pathologists on DP and artificial intelligence (AI).Methods: A digital survey consisting of 32 questions was constructed using Google Forms and sent to practicing pathologists across all sectors in Jordan. The results were analyzed using descriptive statistics.Results: Forty pathologists representing university hospitals, the Ministry of Health, the Royal Medical Services (RMS), and the private sector (PS) participated in the study. 69.2% of participants had average/above-average knowledge of DP. 77.8% of participants without scanners were interested in obtaining one if funds were available, and 85% were likely or very likely to use it for diagnostic purposes. In comparison, 92.5% were very likely to use it for consultation. Cases diagnosed using DP represent 10%. 85% of participants expressed interest in attending sessions at a national congress on DP, and 37.5% currently use AI platforms. Approximately 65% of people with DP didn't follow any guidelines. Seventy-one percent and twenty-nine percent of the guidelines used were from the College of American Pathologists (CAP) and the Royal College of Pathologists (RCP), respectively. At the same time, all pathologists believed the Jordanian Pathologists Society should develop its guidelines. 76.9% thought that a lack of funds was the primary obstacle to adopting DP. In comparison, a lack of infrastructure and experience ranked second, with 40% indicating a lack of interest or a preference for glass slides as obstacles. As for the primary use of DP, 86.8%, 73.7%, 63.2%, 50%, 44.7%, and 44.7% would use it for consultation, education, research, diagnosis, archiving cases, and tumor boards, respectively.Conclusions: Although digital pathology and slide scanners are limited in Jordan, most pathologists are willing to adopt their use, provided that the significant challenges of a lack of funding and inadequate infrastructure are addressed. The primary uses of DP in Jordan seem to be related to consultations and research.