Browse Articles

Tumor board–based multi-agent LLMs with guideline retrieval and consensus deliberation: Implications for hepatocellular carcinoma clinical reasoning.

e16015 Background: Hepatocellular carcinoma (HCC), requires multidisciplinary decision-making across specialities. Multidisciplinary team (MDT) care has been associated with improved outcomes in HCC cohorts, yet major care-process gaps persist. Conventional single-call large language model (LLM) inference for clinical decision support may be constrained by limited guideline grounding. We evaluated whether an MDT-analogous multi-agent consensus architecture augmented with guideline retrieval improves LLM accuracy on validated HCC clinical reasoning tasks versus standard single-call inference. Methods: We developed a multi-agent system comprising three specialist agents (hepatology, oncology, radiology personas) and a supervisor agent. The system used retrieval-augmented generation (RAG) over a curated guideline corpus (AASLD, EASL, ESMO, BSG) implemented in a vector store, retrieving the top-5 passages per case. Performance was benchmarked against baseline single-call inference on a validated 88-item HCC Script Concordance Test (SCT) validated by 3 hepatologists with more than 10 years of experience. We evaluated a top large model (gpt-5.2) and a top small model (qwen3_vl_30b_thinking) selected from an initial benchmarking of 10 LLMs. Results: Baseline single-call accuracy ranged from 61.8% to 74.9% with no significant heterogeneity (Friedman p = 0.44). Large versus small model classes showed no significant difference (69.9% vs 67.4%; Wilcoxon p = 0.26). The multi-agent consensus architecture produced significant improvements for both model classes. For the large model (gpt-5.2), accuracy improved from 74.2% to 80.3% (+6.1 percentage points; 95% CI, 75.1-85.4%; p = 0.020). For the small model (qwen3-vl-30b), accuracy improved from 55.4% to 64.8% (+9.4; 95% CI, 57.0-72.5%; p = 0.029). Notably, the small model demonstrated the largest absolute gain; item-level analysis showed 17 previously incorrect items converted to correct versus only 3 errors introduced (McNemar p = 0.004). Run-to-run consistency was maintained or improved (gpt-5.2: 93%; qwen3-vl-30b: 55% to 73%). Conclusions: A tumor board-inspired multi-agent LLM architecture with guideline retrieval significantly improved HCC clinical reasoning accuracy for both large and small models, suggesting clinical relevance. The larger proportional benefit in smaller models raises the possibility of cost-effective deployment using open-weight architectures. Prospective validation should assess impact on tumor board efficiency, guideline adherence rates, and patient outcomes. Summary comparison table. Metric GPT-5.2 Solo GPT-5.2 Consensus Qwen Solo Qwen Consensus Accuracy 74.2% 80.3% 55.4% 64.8% 95% CI 68.1-79.8% 75.1-85.4% 48.8-62.0% 57.0-72.5% Wilcoxon p - 0.020 - 0.029 McNemar p - 0.19 - 0.004 Consistency 95.8% 93.0% 54.9% 73.2%

Artificial intelligence and nanotechnology in hepatocellular carcinoma: Implications for molecular pathology and precision diagnostics

Circularly Polarized 1540 nm Short‐Wave Infrared Electroluminescence from Er‐Based Halide LEDs with 3.06% Record Efficiency

ABSTRACT Er 3+ ‐doped 1540 nm light‐emitting diodes (LEDs) are critical components in optical communications C‐band, non‐trunk communication, bioimaging, and sensing. However, integrating high luminous efficiency with tailored circularly polarized luminescence (CPL) in such LEDs remains a critical challenge. Here, we demonstrate efficient 1540 nm short‐wave infrared (SWIR) electroluminescence with distinct CPL in Cs 3 ErCl 6 nanocrystals (NCs)‐based LEDs via a synergistic strategy of Sb 3+ /Y 3+ co‐doping and camphor ligand modification. Y 3+ doping modulates lattice symmetry, inducing Stark splitting of the Er 3+ energy level and enhancing luminescence intensity. Sb 3+ introduction triggers efficient self‐trapped excitons emission at 530 nm, whose energy levels match Er 3+ states to boost energy transfer efficiency. Subsequently, camphor ligand exchange passivates Er 3+ ‐related defects, increasing the 1540 nm photoluminescence quantum yield to 35.7% and endowing NCs with CPL (asymmetry factor: −3.67 × 10 −2 ) via camphor's chiral structure. SWIR LEDs based on camphor‐modified Cs 3 Er 0.7 Y 0.3 Cl 6 : Sb 3+ NCs exhibit a record‐high external quantum efficiency of 3.06% at 1540 nm, and first, demonstrate electrically‐driven circularly polarized 1540 nm emission with asymmetry factor of −3.08 × 10 −2 . This work presents a synergistic doping‐ligand strategy for Er‐based halide optoelectronics, offering a versatile platform to develop high‐performance long‐wavelength devices with integrated efficient emission and tailored polarization, crucial for advancing next‐generation optical communication and bioimaging.

Rural-urban survival differences in advanced non–small cell lung cancer in the modern treatment era: A SEER-based analysis.

e13735 Background: Rural-urban disparities in cancer outcomes are well documented; however, whether recent therapeutic advances have translated into equitable survival gains for patients with advanced non-small cell lung cancer (NSCLC) remains unclear. Beginning 2015, immune checkpoint inhibitors became standard for advanced NSCLC. This study examined rural-urban differences in overall survival (OS) across treatment eras. Methods: Data were obtained from the SEER 17 database and included adults with distant-stage NSCLC. Patients were classified as rural or urban and stratified by era of diagnosis (pre-2015 vs post-2015). OS was analyzed using Kaplan–Meier methods and compared using the log-rank test. Multivariable Cox proportional hazards models adjusted for age, sex, race, income, and marital status, and included an interaction term between rural residence and era of diagnosis. Results: A total of 248,598 adults were identified; 14.8% resided in rural areas and 85.2% in urban areas. Overall, 59.3% were diagnosed in the pre-2015 era and 40.7% in the post-2015 era. The population was predominantly male and aged ≥65 years, with a median OS (mOS) of 6 months. Patients in rural areas had inferior OS compared with urban residents (mOS 5 vs 6 months; P < 0.001). On multivariable analysis, rural residence was independently associated with increased mortality. Increasing age, male sex, Black race, lower household income, and single or previously married status were also associated with higher mortality (Table 1). In era-stratified analyses, rural-urban survival differences persisted and widened after 2015. In the pre-2015 era, mOS was 5 months in both groups (P < 0.001). In the post-2015 era, mOS increased to 7 months in urban patients but remained 5 months in rural patients (P < 0.001). In interaction-adjusted analysis, post-2015 diagnosis was associated with reduced mortality overall, while rural patients experienced an additional mortality risk in the post-2015 era. Conclusions: Among adults with distant-stage NSCLC, rural residence was associated with inferior overall survival compared with urban residence. Survival gains after 2015 were primarily observed in urban patients, resulting in widening rural-urban disparities. These findings highlight geographic inequities in advanced lung cancer outcomes and the need to improve access to cancer care for rural populations. Multivariate analysis. Variable Hazard Ratio (HR) 95% CI P value Rural (vs Urban) 1.04 1.03–1.06 <0.001 Post-2015 (vs Pre-2015) 0.79 0.78–0.80 <0.001 Rural × Post-2015 interaction 1.04 1.02–1.07 <0.001 Age (per year) 1.02 1.02–1.02 <0.001 Male (vs Female) 1.25 1.24–1.26 <0.001 Black (vs White) 1.04 1.02–1.06 <0.001 Asian (vs White) 0.81 0.79–0.83 <0.001 ≥$75,000 (vs <$75,000) 0.93 0.92–0.93 <0.001 Single (vs Married) 1.25 1.24–1.27 <0.001 Previously married (vs Married) 1.18 1.17–1.19 <0.001

Comparison of survival after chemotherapy vs no chemotherapy in elderly patients with pancreatic cancer.

e16442 Background: Pancreatic cancer disproportionately affects older adults, yet real-world outcomes associated with chemotherapy in this population remain incompletely characterized. We evaluated clinical outcomes among patients aged ≥70 years with pancreatic cancer treated with chemotherapy compared with those who did not receive chemotherapy. Methods: We conducted a retrospective cohort study using the TriNetX Global Collaborative Network, comprising electronic health records from 170 healthcare organizations. Patients aged ≥70 years with a diagnosis of pancreatic cancer (ICD-10 C25) were included. The chemotherapy cohort included patients who received gemcitabine- or fluoropyrimidine-based regimens, including FOLFIRINOX or FOLFOX, after cancer diagnosis. Patients without documented chemotherapy exposure comprised the comparator cohort. Propensity score matching (1:1) was performed to balance demographics and comorbidities. Outcomes assessed from 30 days to 5 years after index event included all-cause mortality, hospitalization, emergency department visits, ICU admission, venous thromboembolism (VTE), infection, small bowel obstruction, and biliary tract obstruction. Kaplan–Meier survival and measures of association were used. Results: After matching, 3,612 patients in the chemotherapy cohort and 3,438 patients in the non-chemotherapy cohort were analyzed. Median survival in the chemotherapy cohort was 606 days, whereas median survival was not reached in the non-chemotherapy cohort; survival probability at the end of follow-up was 25.0% versus 64.0%, respectively. Chemotherapy was associated with significantly worse overall survival (hazard ratio [HR] 2.51, 95% CI 2.30–2.74; log-rank p < 0.001). Chemotherapy-treated patients experienced higher healthcare utilization and complication rates, including hospitalization (32.3% vs 9.4%; risk ratio [RR] 3.44), ED visits (10.8% vs 3.9%; RR 2.79), infections (17.4% vs 4.2%; RR 4.15), ICU admissions (9.1% vs 2.1%; RR 4.29), VTE (12.1% vs 2.3%; RR 5.16), SBO (4.5% vs 1.6%; RR 2.78), and biliary tract obstruction (7.7% vs 1.6%; RR 4.75), p < 0.001. Conclusions: In older adults with pancreatic cancer, chemotherapy was associated with significantly worse survival and substantially higher treatment-related morbidity and healthcare utilization. These findings highlight the need for careful patient selection and shared decision-making when considering chemotherapy in this vulnerable population.

Net clinical value of neoadjuvant immune checkpoint inhibitors in early triple-negative breast cancer: Integrating efficacy heterogeneity and treatment burden.

e12659 Background: Neoadjuvant immune checkpoint inhibitors (ICIs) improve pathological complete response (pCR) rates in early-stage triple-negative breast cancer (TNBC). In curative-intent settings, however, treatment escalation should be guided not only by efficacy signals but also by baseline clinical risk and downstream treatment burden. Whether neoadjuvant ICIs provide a consistent net clinical value across TNBC risk groups remains uncertain. Methods: We performed a trial-level meta-analysis of randomized phase II–III trials comparing neoadjuvant chemotherapy with or without ICIs in early TNBC. Odds ratios (ORs) for pCR and hazard ratios (HRs) for event-free survival (EFS) were pooled using inverse-variance models. Prespecified subgroup analyses were conducted according to nodal status and tumor stage to explore efficacy heterogeneity. To contextualize efficacy findings in a curative-intent setting, treatment discontinuation due to adverse events was evaluated as a competing clinical consideration, estimating absolute excess risk and numbers needed to harm (NNH). Results: Over 3,000 patients with early TNBC were included. ICIs significantly increased pCR rates, with marked heterogeneity across clinical risk groups. The pCR benefit was pronounced in node-positive disease (OR 2.57, 95% CI 1.76–3.75) but attenuated and not statistically significant in node-negative patients (OR 1.29, 95% CI 0.93–1.80). By contrast, EFS benefit was observed across risk categories, including both node-positive (HR 0.63, 95% CI 0.47–0.86) and node-negative disease (HR 0.65, 95% CI 0.46–0.93), highlighting a discordance between early response endpoints and long-term outcomes in lower-risk patients. Across trials, ICI-containing regimens were associated with clinically relevant excess treatment discontinuation due to adverse events, with absolute increases ranging from 3% to 11%, corresponding to one additional treatment discontinuation for every 9–36 patients treated. Conclusions: In early TNBC, neoadjuvant ICIs do not deliver uniform net clinical value across patient subgroups. While long-term outcome improvements extend across risk categories, heterogeneous early efficacy gains combined with non-negligible treatment discontinuation challenge universal escalation strategies, particularly in lower-risk disease. Integrating long-term benefit with treatment burden is essential to define the true clinical value of neoadjuvant immunotherapy and to support a risk-adapted, value-oriented approach rather than routine intensification.

Characterization of inavolisib-associated hyperglycemia in metastatic hormone receptor–positive (HR+) breast cancer.

e13055 Background: Inavolisib + palbociclib + fulvestrant is approved in patients with HR+, HER2-, PIK3CA -mutant MBC based on INAVO120, which demonstrated significant progression-free survival and overall survival benefit in the first line compared with palbociclib + fulvestrant. In INAVO120, 63.4% of patients developed any-grade hyperglycemia (HG) and 6.8% developed grade 3-4 HG despite eligibility requirement of fasting glucose ≤126mg/dl and hemoglobin A1c (HbA1c) ≤6%. Here we describe the incidence and treatment of inavolisib-associated HG in a single center cohort. Methods: Patients with MBC who received inavolisib as standard care from 10/2024–1/2026 at Memorial Sloan Kettering Cancer Center and had ≥1 on-treatment glucose were included in this retrospective study. Patient and tumor characteristics, pretreatment body mass index (BMI), HbA1c, on-treatment glucose levels, and details on HG management were abstracted. HG was graded per CTCAE v4.0. Results: 40 female patients were included in this study. Median age was 65 (range 39-92). 21 patients (53%) received inavolisib in the first-line setting and 11 (28%) received it in the second-line (range 1-11 th line). Median pretreatment BMI was 26.2 kg/m 2 (range 20.3-47.0). Among 31 (78%) patients with pretreatment HbA1c levels available, median HbA1c was 5.5% (range 4.7-6.9%); 17 (55%) had normal HbA1c, 10 (32%) had HbA1c in the prediabetes range (5.7-6.4%), and 4 (12.9%) had HbA1c in the diabetes range (≥6.5%). 23 patients (58%) developed HG of any grade; 4 (10%) developed grade 1, 10 (25%) developed grade 2, and 9 (23%) developed grade 3 HG. The median time to onset of any-grade HG was 42 days (range 4-226). Among those who developed HG, 11 (48%) received anti-hyperglycemic medications; 7 patients received 1 or 2 anti-hyperglycemic agents whereas 4 patients required ≥3 anti-hyperglycemic agents. In 10/11, metformin was the first anti-hyperglycemic initiated and in 7, an SGLT2 inhibitor was the next anti-hyperglycemic initiated. 4 patients were treated with insulin, 1 of whom had HbA1c in the prediabetes range and 2 of whom had HbA1c in the diabetes range before inavolisib. 12 patients were referred to endocrinology. Among the total cohort, inavolisib was held until resolution of HG in 10 patients (25%) and dose-reduced due to HG in 7 patients (18%). 4 patients (10%) discontinued inavolisib due to HG; 3 of these patients required insulin, 2 had prior inavolisib dose reductions, 1 had pretreatment HbA1c in the diabetes range and 2 had HbA1c in the prediabetes range. Conclusions: Real-world rates of HG with inavolisib in this single-center cohort underscore the importance of baseline metabolic assessment, proactive glucose monitoring, and management to minimize inavolisib interruptions. As a limitation, fasting status was not uniformly documented so some glucose values may have been non-fasting, affecting HG grading.

Prevalence of symptomatic skeletal events (SSE) with reduced denosumab (Dmab) dose density (every 12 weeks versus every 4 weeks): A randomized phase III non-inferiority trial SAKK 96/12 REDUSE.

1004 Background: Clinical guidelines recommend Dmab 120 mg every 4 weeks (Q4W) for patients (pts) with bone metastases. While dosing every 12 weeks (Q12W) is established for zoledronic acid based on phase III trials, prospective trials with Dmab evaluating the efficacy endpoint of skeletal-related events (SRE) for a Q12W schedule are missing. The results of an early dose-finding study of Dmab demonstrated a similar suppression of bone-turnover markers for 60 mg Q12W as for 120 mg Q4W at weeks 13 and 25, suggesting that Q12W dosing might be sufficient. The SAKK 96/12 REDUSE trial investigated whether a de-escalated Dmab schedule is non-inferior to the standard Q4W regimen in preventing symptomatic skeletal related events (SSEs; symptomatic pathological fracture, radiation or surgery to bone, or spinal cord compression). Methods: In this international, multicenter, randomized, phase III trial, pts with metastatic breast cancer (BC) or castration-resistant prostate cancer (CRPC) and at least three bone metastases were randomized 1:1. Arm A (standard) received Dmab 120 mg Q4W. Arm B (reduced) received 4 loading doses Q4W, followed by Q12W dosing. The primary endpoint was time to first on-trial SSE. The sample size of 1,380 pts was calculated to demonstrate non-inferiority using a log-rank test with a non-inferiority margin of 1.20 (80% power, 5% type I error). Secondary endpoints included time to first and subsequent on-trial SSE, and toxicity focusing on hypocalcemia and osteonecrosis of the jaw (ONJ). A health economic analysis is ongoing. Results: 1,380 pts were enrolled between 7/2014 and 3/2024 (784 and 596 with BC and CRPC, respectively). Median follow-up time was 37 months (mo) (42 mo for BC and 29 mo for CPRC). The trial met its primary endpoint, demonstrating that denosumab Q12W is non-inferior to Q4W. Median time to first on-trial SSE for arm A was 56.6 mo (95% CI (40.7, not reached [NR]) and for arm B 56.5 mo (95% CI 47.7, NR). The stratified HR and 90% CI calculated using a Cox regression model is 1.023 (0.874, 1.197) with the upper limit remaining below 1.20. Also, for time to first and subsequent on-trial SSE, non-inferiority was shown with a stratified HR 1.043 (90% CI 0.907, 1.198). Safety analysis showed less toxicity for the Q12W arm: Hypocalcemia occurred in 46% (Arm A) vs 30% (Arm B), and ONJ was reported in 8.5% vs 6.9% of pts, respectively. Conclusions: The SAKK 96/12 REDUSE trial demonstrates that a Dmab Q12W schedule, following a 3-month loading phase, is non-inferior to the Q4W regimen in pts with bone-metastatic BC or CRPC. The observed lower rates in ONJ and hypocalcemia indicate a more favorable safety profile for the Q12W schedule. These data indicate that Dmab Q12W represents the new standard of care in pts with bone metastasis from BC or CRPC, offering comparable efficacy while reducing toxicity, treatment burden, and costs. Clinical trial information: NCT02051218 .

Improving dermatologic surveillance in solid organ transplant recipients through a navigation program.

e21554 Background: Solid organ transplant recipients (SOTRs) face markedly elevated skin cancer risk, with cutaneous squamous cell carcinoma incidence increased up to 65-fold compared to the general population.[1] Skin cancer mortality in this population reaches 35 per 100,000 person-years and is 9-fold higher than in the general population, comparable to or exceeding mortality from colon or breast cancer.[1][2] Despite expert consensus guidelines supporting dermatologic surveillance to reduce morbidity and mortality, screening rates remain low, with prior studies reporting only 18.6-27% of SOTRs undergoing routine surveillance.[3][4][5] Barriers include perceived lack of evidence, limited dermatology access, and variable referral patterns. Methods: In order to improve surveillance rates in SOTR at our center, we developed a focused navigation program in which the transplant team shares updated information of transplant recipients monthly with a dedicated patient coordinator who contacts patients to assess whether they have scheduled dermatology appointments internally or externally. For patients without scheduled appointments, the navigator facilitates scheduling with dermatology. Between May 2023 and July 2024, 489 SOTRs were referred, including kidney (236), liver (133), heart (93), lung (10), and multiorgan (13) transplant recipients. Results: Of 489 referred patients, 21% had dermatology appointments scheduled through the navigation program, representing patients who had not previously scheduled surveillance. Further surveillance visits were scheduled by the dermatology clinic. Among the remaining 79%, 45% were already established with dermatology within our institution. Of those not scheduled, 25% declined referral, 15% were lost to follow-up, 4% expired, and 11% cited other reasons. Ongoing work will further assess skin cancer incidence, stage, and outcome in both patients who participated in surveillance and those who did not. Conclusions: This navigation program demonstrates a practical approach to improving dermatologic surveillance in SOTRs, successfully engaging patients who had not previously scheduled screening. The navigation tool addresses the gap between guideline recommendations and clinical practice by providing systematic outreach and scheduling assistance. However, significant barriers persist, including patient refusal and loss to follow-up, highlighting the need for enhanced patient education and multidisciplinary coordination. This model may serve as a template for institutions seeking to improve skin cancer surveillance in this high-risk population, though adaptations may be necessary for resource-limited settings where risk-stratified approaches tools may be more useful.[6][1]

Adenoma detection following screening colonoscopy and stool-based colorectal cancer screening tests with follow-up colonoscopy: A systematic and network meta-analysis.

e15647 Background: Colorectal cancer (CRC) screening offers an opportunity to prevent invasive disease by detecting and removing precancerous lesions. Colonoscopy-based detection of adenomas and serrated polyps is strongly associated with reduced risk of interval CRC. While the performance of noninvasive CRC screening tests is well described, comparative evidence on adenoma detection across screening colonoscopy and stool-based strategies with follow-up colonoscopy remains limited. We conducted a systematic review to compare adenoma detection across CRC screening modalities. Methods: MEDLINE (PubMed) and Embase were searched from January 1, 2000, through July 10, 2025, for peer-reviewed, U.S.-based studies reporting colonoscopy among average-risk adults aged ≥45 years, who underwent screening colonoscopy or follow-up colonoscopy after a positive stool-based screening test [fecal immunochemical test (FIT) or multi-target stool DNA (mt-sDNA) test]. The primary outcome was the detection of one or more adenomas at colonoscopy. High-risk or surveillance cohorts, AI-assisted colonoscopy, diagnostic accuracy trials, and non-U.S. studies were excluded. Results: Among 7,710 de-duplicated records, 147 studies underwent full-text review, and 47 studies were included in quantitative synthesis, with reported adenoma detection for screening colonoscopy alone (n = 31), follow-up colonoscopy after a positive FIT alone (n = 4), follow-up colonoscopy after a positive mt-sDNA test alone (n = 4), and for more than one of the screening strategies of interest (n = 8). Across included studies, overall adenoma detection differed by screening pathway. Follow-up colonoscopy after a positive mt-sDNA test demonstrated the highest adenoma detection (66.3%; 95% CI, 61.2-71.1), with lower adenoma detection observed for colonoscopy after a positive FIT (50.6%; 95% CI, 43.4-57.8) and screening colonoscopy alone (33.8%; 95% CI, 30.9-36.8) (Table 1). Conclusions: Detection of adenomas was highest following colonoscopy performed after a positive mt-sDNA test compared with FIT-positive follow-up or screening colonoscopy alone. These findings highlight differences in downstream detection across CRC screening tests and suggest that, in settings where colonoscopy resources are constrained, initial mt-sDNA screening with timely follow-up colonoscopy may help prioritize patients most likely to benefit. CRC screening modality No. of Studies Adenoma detected Study population Proportion, % (95% CI) Heterogeneity, I² (%) Between-study variance (τ 2 ) P-value* mt-sDNA 8 17,436 29,098 66.3 (61.2-71.1) 94.6 0.0924 <0.0001 FIT 11 17,588 32,630 50.6 (43.4-57.8) 94.0 0.2187 <0.0001 Screening colonoscopy 38 2,351,801 6,378,891 33.8 (30.9-36.8) 99.8 0.1710 NA *P-value represents adenoma detection vs. screening colonoscopy adenoma detection.

Extracellular and immune matrix signature of interstitial fluid as mediator of pathological characteristics of colorectal cancer.

e15653 Background: The interstitial fluid (ISF) in tumor microenvironment harbors abundant proteins and metabolites critically influencing tumor progression, yet its comprehensive molecular profile in colorectal cancer (CRC) remains uncharacterized. We aim to elucidate the proteomic and metabolomic characteristics of colorectal cancer interstitial fluid and investigate their correlation with clinical pathological features. Methods: We collected tumor tissue and matched normal adjacent tissues (NAT) samples from 47 treatment-naïve CRC surgery patients. ISF was extracted through low-speed centrifugation. Multi-omics analysis involving 4D-DIA proteomics and targeted metabolomics (93 metabolites) was performed on all samples (47 tissue samples and their paired 47 interstitial fluid samples, encompassing tumor/NAT). Bioinformatic analysis subsequently comprised differential expression, consensus clustering, and pathway enrichment. Results: We quantified 8,642 proteins in ISF, with significant enrichment for extracellular matrix (ECM) and immune-related proteins compared to tissue lysates. Tumor ISF exhibited upregulated ECM organization but downregulated complement and adaptive immune pathways compared to NAT ISF. Consensus clustering based on 754 immune- and ECM-related proteins within ISF defined three subtypes: High-Cellular immunity/Low-ECM (HCLE, n = 18), High-inflammation/High-ECM (HiHE, n = 22), and Low-immune/Low-ECM (LiLE, n = 7). The HiHE subtype was characterized by concurrent unstable ECM and inflammatory signaling, significantly associated with low differentiation (p = 0.012) and KRAS mutation (p = 0.037). 6 proteins (COL1A1, COL1A2, COL3A1, THBS2, CCDC80, VIM) were significantly overexpressed in the HiHE subtype (fold change > 2, p < 0.05) and were associated with poorer overall survival and disease-free survival in colorectal cancer. Metabolomics revealed elevated proline and hydroxyproline in tumor ISF, particularly in the HiHE subtype. Furthermore, hydroxylation ratios of major collagens (COL1A1, COL3A1) were significantly reduced in HiHE tumor ISF, indicating ECM instability despite high collagen content. Conclusions: This study establishes the first multi-omics of tumor ISF in CRC, revealing that tumor ISF serves as a dynamic reservoir of ECM remodeling proteins, complement, and antibodies. The subtyping based on ISF could provide a unique signature that tissue subtyping could not present. A subtype called HiHE, characterized by concurrent ECM hyperactivation and inflammatory signaling, is correlated with pathological features such as low differentiation.

Determinants of skin cancer prevention and early detection behaviors in Syria: A cross-sectional survey.

e22557 Background: Skin cancer prevention and early detection depend on sun-protective behaviors, skin self-examination, and timely care seeking. Evidence on screening behaviors and barriers in resource-limited settings remains limited. Methods: We conducted a national cross-sectional survey in Syria. Adults aged ≥18 years were eligible; individuals enrolled in or working in health-related fields were excluded using a screening item. The questionnaire assessed socio-demographics, knowledge, attitudes, practices, sources of information, perceived barriers, and Fitzpatrick skin type. Outcomes were dichotomized using predefined cutoffs. Multivariable binary logistic regression examined independent predictors of high practice. Institutional approval was obtained. Results: Among 2,033 participants (median age 25 years; IQR 21–34), 75.6% were female and 16.7% had health insurance. Only 0.9% reported ever undergoing skin cancer screening. Knowledge was moderate (median 4/8; IQR 3–5), with gaps in recognizing key risk factors and warning signs. Overall, 43.8% met criteria for good knowledge, 62.1% for positive attitude, and 50.7% for high practice. Early detection behaviors were limited: 69.4% reported not performing skin self-examination, and females were more likely than males to report high practice (57.5% vs 29.6%). The most commonly reported barriers were having no symptoms (68.8%), lack of knowledge about skin cancer (37.9%), lack of recommendations from healthcare professionals (30.5%), lack of health insurance (27.1%), and lack of time (27.1%). In multivariable analysis, high practice was associated with female sex, university education, medical sources of information, and positive attitude, while good knowledge was not associated. Conclusions: In this Syrian population, knowledge alone did not translate into stronger prevention and early detection practices. Attitudes, trusted medical information sources, and modifiable access- and recommendation-related barriers were more influential. Interventions should reframe screening as preventive behavior and strengthen feasible pathways through primary care and community settings to improve timely evaluation and early detection. Independent predictors of high preventive/early-detection practice. Predictor Category (vs reference) aOR 95% CI p Sex Female vs male 3.087 2.448–3.892 <0.001 Education level University vs secondary/lower 1.536 1.241–1.901 <0.001 Source of information Medical vs non-medical 1.673 1.330–2.105 <0.001 Attitude level Positive vs less-positive 1.904 1.571–2.308 <0.001

A randomized phase II trial of fruquintinib plus capecitabine versus capecitabine alone as maintenance therapy following first-line chemotherapy in metastatic colorectal cancer (mCRC).

3534 Background: Maintenance therapy with capecitabine (Cap) is widely recommended to unresectable mCRC patients (pts). Fruquintinib (Fru) is a highly selective inhibitor of VEGFR1/2/3. We sought to compare the therapeutic potential of Fru plus Cap versus Cap as maintenance therapy for mCRC. Methods: Eligible pts with histologically confirmed mCRC who achieved disease control (including CR/PR and SD) after at least six cycles of first-line standard chemotherapy were included in the study. During phase Ib, pts received Fru (4 mg, p.o. qd, 3 weeks on/1 week off, q4w) plus Cap (850 mg/m 2 , p.o. bid, d1-7 and d15-21, q4w). In phase II, pts were randomized in a 1:1 ratio to receive either Fru (RP2D, 3mg, 3 weeks on/1 week off) plus Cap or Cap alone. The primary endpoint was progression-free survival (PFS). Secondary endpoints included objective response rate (ORR), disease control rate (DCR) and safety. Results: As of November 15, 2025, 113 eligible patients were enrolled, including 6 in phase Ib and 107 in phase II. In this presentation, we report only the updated data from Phase II. 107 were randomly assigned to receive either fruquintinib plus capecitabine (Fru+Cap; n=52) or capecitabine alone (Cap; n=55). Baseline characteristics included median age: 61 years (range, 30-78) vs. 60 years (range, 31-80), male: 51.9% vs. 58.2%, left-sided colorectal primary tumors: 55.8% vs. 61.8% and RAS-mutant tumors: 53.8% vs. 52.7%. As first-line maintenance therapy, the median PFS1 (defined as time from randomization to disease progression or death) in the per-protocol population was 6.03 months with Fru+Cap versus 3.30 months with Cap (p=0.003). Furthermore, the Fru+Cap arm demonstrated significantly prolonged PFS2 (defined as the time from first receipt of first-line treatment to disease progression or death): 12.1 vs. 9.7 months, p=0.006. Consistent improvements were observed in objective response rate (ORR: 14.6% vs. 7.0%) and disease control rate (DCR: 75.6% vs. 51.2%). Grade ≥3 treatment-emergent adverse events (TEAEs) in Fru+Cap arm included hypertension (5.77%), palmar-plantar erythrodysesthesia syndrome (1.9%) and albuminuria (1.9%). Diarrhea (3.64%), palmar-plantar erythrodysesthesia syndrome (1.8%) and oral mucositis (1.8%) were the most frequently reported grade ≥3 TEAEs in Cap arm. Conclusions: PFS was significantly prolonged and ORR was improved in mCRC pts treated with Fru plus Cap as first-line maintenance treatment. The safety profile of the combination regimen was manageable and aligned with the expected toxicity patterns of the individual agents. Clinical trial information: NCT05451719 .

Correlates of alcohol-related cancer risk awareness in U.S. adults.

e22675 Background: Alcohol consumption is a modifiable risk factor for multiple malignancies, yet public understanding of this association remains limited in the United States. Identifying sociodemographic, behavioral, and perceptual correlates of awareness is essential to inform targeted cancer prevention communication strategies. Methods: We analyzed data from 15,826 U.S. adults who participated in the Health Information National Trends Survey (HINTS) 4, Cycle 3, through HINTS 7 (2013–2024). Participants were categorized by awareness of alcohol as a cancer risk factor (“Yes,” “No,” or “Don’t know”). Logistic regression models were used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for demographic, behavioral, and cancer-related belief variables associated with awareness. Results: The study population had a mean age of ~55 years. Overall, 42.2% of participants reported awareness that alcohol increases cancer risk. Awareness was significantly associated with higher educational attainment (OR = 1.20, 95% CI 1.14–1.26), with awareness reaching 55.5% among individuals with a bachelor’s degree or higher, compared with 17.9% among those with a high school education or less. Greater annual household income (OR = 1.11, 95% CI 1.07–1.15) and having ever sought information about cancer (OR = 1.78, 95% CI 1.59–1.99) were also significant predictors of awareness. Perceptually, awareness was associated with lower cancer fatalism; 38.0% of aware respondents strongly disagreed that little can be done to reduce cancer risk. Behaviorally, non-smokers were more likely to report awareness (OR = 1.22, 95% CI 1.09–1.37). Notably, a knowledge–behavior paradox was observed, as heavy alcohol consumption (≥9 drinks/week) was most prevalent among those who were aware (11.8%) compared with those responding “No” (10.1%) or “Don’t know” (7.2%). Conclusions: Awareness of alcohol as a cancer risk factor among U.S. adults remains suboptimal and is strongly patterned by socioeconomic status, health information-seeking, and cancer-related beliefs. These findings highlight persistent disparities in cancer risk knowledge and underscore the need for targeted public health messaging, to bridge gaps between awareness and behavioral risk reduction.

Peritumoral stiffness as a novel noninvasive biomarker: Predicting TACE response in intermediate-stage hepatocellular carcinoma via 2D-shear wave elastography.

4143 Background: Intermediate-stage hepatocellular carcinoma (HCC) demonstrates considerable biological and clinical heterogeneity, making it challenging to predict treatment response to transcatheter arterial chemoembolization (TACE). This multicenter retrospective study evaluated the association between preoperative tissue stiffness measurements obtained by 2D-shear wave elastography (2D-SWE) and TACE outcomes in intermediate-stage HCC, aiming to identify an optimal peritumoral stiffness cutoff value for prognostic stratification and clinical decision-making. Methods: A total of 115 patients with intermediate-stage HCC were enrolled. Preoperative 2D-SWE was performed to measure stiffness of the tumor (TSWE), peritumoral tissue (PSWE), and background liver parenchyma (LSWE). Cox proportional hazards regression and Kaplan-Meier analyses were conducted to determine whether 2D-SWE-derived stiffness parameters independently predicted TACE efficacy. Immunohistochemical staining, immunofluorescence, and RNA sequencing were performed to characterize the biological features of peritumoral tissue associated with varying PSWE values. Results: Mean stiffness values demonstrated a progressive gradient from liver parenchyma (LSWE: 13.0 kPa) through peritumoral tissue (PSWE: 18.9 kPa) to tumor tissue (TSWE: 25.6 kPa). Univariate and multivariate analyses identified primary tumor size and PSWE as independent prognostic factors for both overall survival (OS) and time to progression (TTP). A PSWE cutoff of 19.04 kPa achieved the highest discriminatory performance for predicting objective response rate (ORR) and disease control rate (DCR) (AUROC = 0.691 and 0.695, respectively). Patients with low PSWE demonstrated significantly superior OS (26.0 vs 17.5 months, P = 0.008) and TTP (11.6 vs 9.5 months, P = 0.026) compared to those with high PSWE. RNA sequencing revealed 778 differentially expressed genes enriched in angiogenesis and fibrosis-related pathways. Notably, patients with elevated PSWE exhibited enhanced stromal activation and angiogenesis in peritumoral regions, including increased collagen deposition, α-SMA, COL-I, VEGF expression, and CD31+ microvessel density, correlating with inferior TACE response. Conclusions: Preoperative peritumoral stiffness measured by 2D-SWE serves as a promising noninvasive prognostic biomarker for intermediate-stage HCC patients undergoing TACE, offering valuable insights for personalized treatment stratification.

Nimotuzumab combined with paclitaxel and cisplatin as first-line treatment for metastatic esophageal squamous cell carcinoma: A prospective, multicenter, double-blind, randomized controlled clinical study.

4055 Background: Chemotherapy for advanced esophageal squamous cell carcinoma (ESCC) is limited due to the lack of effective drugs. Such as traditional two drugs chemotherapy (5-Fluorouracil and Cisplatin) as first-line treatment for metastatic and recurrent ESCC has an efficacy of 25-35%. Nimotuzumab is an anti-epidermal growth factor receptor (EGFR) monoclonal antibody. Previous studies have shown that its combination with paclitaxel and cisplatin (TP regimen) has a good efficacy, with an objective response rate (ORR) of up to 55%, median overall survival (mOS) 13.9 months in some small-sample trials. Methods: Eligible patients from 36 centers in China, who were randomly assigned to receive nimotuzumab (400 mg once per week, up to 2 years) or placebo followed by TP regimen (paclitaxel 175 mg/m², cisplatin 60 mg/m², Day 1, with a 21-day cycle, up to 6 cycles) until disease progression or unacceptable toxicity. The primary end point was overall survival (OS) and the secondary end points were progression-free survival (PFS), response rates, quality of life (QoL) and safety. Results: A total of 640 patients were screened. 503 patients were enrolled and 161 patients with EGFR gene amplification tumors were eligible. In the full analysis set of 497 patients who had took at least one dose of medication, there were no differences in baseline characteristics. The mOS were 12 vs 11.5 months, and the median PFS were 5.6 vs 5.4 months respectively, suggesting survival benefit trend in the trial. ORR were 55.5% and 50.4% for both arms. In the EGFR gene amplification subgroup, the mOS was 13.3 vs 9.5 months (log-rank test, hazard ratio [HR] = 0.66 (95%CI, [0.47-0.93]), P = 0.016) for two groups. Patients who were with ESCC at stage IV, as well as who had not undergone surgery, radical radiotherapy, or neoadjuvant and adjuvant therapy showed significant survival benefits after the addition of nimotuzumab. Median PFS were 6.0 vs 5.5 months for two arms (log-rank test, HR = 0.63 [0.43-0.91], P = 0.014). Both OS and PFS were longer in the nimotuzumab group than in the placebo group. The ORR of the two groups were 61.3% and 44.0%, respectively (P = 0.029). There were no statistically significant differences in the QoL scores of the FAS population at each time point before and after treatment. The incidence of adverse events were comparable in the experimental group and the control group. Such as the incidence of serious adverse event (SAE) was 30.7% and 34.5%, serious adverse drug reaction (SADR) was 2.7% and 8.3%, the incidence of above grade 3 treatment emergent adverse event (TEAE) was 85.3% and 83.3%, respectively, etc. Conclusions: In patients with metastatic EGFR gene amplification ESCC, nimotuzumab plus TP as first-line treatment significantly improved OS and PFS with a good safety profile.

Influence of racial, socioeconomic, and tumor characteristics on potential years of life lost (PYLL) in renal cell carcinoma: A population-based SEER analysis.

e16532 Background: Traditional survival metrics may underestimate the burden of premature death. Potential years of life lost (PYLL) quantifies years lost due to early mortality. We evaluated PYLL in renal cell carcinoma (RCC) and assessed the influence of demographic, socioeconomic, and tumor-related factors. Methods: This retrospective population-based study used SEER 17 data (2000–2019). Adults who died from RCC were included (N = 30,850). RCC cases were identified using ICD-O-3 histology codes 8310, 8312, 8260, 8317, 8319, and 8320. PYLL was calculated as the difference between age at death and a reference age of 75 years, consistent with established cancer burden studies. Generalized linear models (GLM) with Gamma family distribution and log link were used for primary analysis; exponentiated coefficients represent percent change in PYLL. Ordinary least squares (OLS) regression provided absolute year estimates. Covariates included race/ethnicity, sex, income quantile, AJCC stage, tumor grade, and histologic subtype. All analyses were performed using Python 3.14 with statsmodels and pandas libraries. Results: In this study, the mean PYLL was 11.7 years per death from renal cell carcinoma, with significant racial/ethnic disparities persisting even after multivariable adjustment for clinical and demographic factors. Compared to non-Hispanic Whites, Black patients had 20.3% higher PYLL (+2.56 years; P0.001), Hispanic patients had 18.6% higher PYLL (+2.32 years; P0.001), Asian patients had 10.4% higher PYLL (+1.30 years; P0.001), and American Indian patients had 10.4% higher PYLL (+1.38 years; P = 0.003). Males had 4.5% higher PYLL than females (+0.55 years; P0.001). Income quantile was not significantly associated with PYLL (Q2 vs Q1: P = 0.251; Q3 vs Q1: P = 0.787). Stage IV disease had 7.0% higher PYLL than Stage I (+0.92 years; P0.001), while Stage II (−7.7%; P = 0.002) and Stage III (−5.1%; P = 0.001) had lower PYLL. Grade IV tumors had 21.0% higher PYLL than Grade II (+2.62 years; P0.001), Grade III had 11.2% higher PYLL (+1.42 years; P0.001), and Grade I had 6.0% lower PYLL (−0.70 years; P = 0.002). Among histologic subtypes, collecting duct carcinoma (8319) had the highest PYLL (+34.0%, +4.71 years; P0.001), followed by granular cell (8320; +15.4%, +1.98 years; P = 0.002), RCC-NOS (8312; +7.2%, +0.93 years; P0.001), and papillary (8260; +3.8%, +0.51 years; P = 0.022). Conclusions: Substantial racial/ethnic disparities in PYLL from RCC persist after multivariable adjustment, with Black and Hispanic patients experiencing disproportionately greater premature mortality. Income did not explain these disparities, implicating factors such as access to care, treatment quality, and tumor biology. These findings underscore the need for targeted interventions to reduce premature cancer deaths in high-risk populations.

Impact of broadening trial eligibility criteria on the inclusion of patients infected with HIV in cancer clinical trials.

1548 Background: In October 2017, an ASCO, Friends of Cancer Research (FoCR), and US FDA (ASCO/FoCR/FDA) task force recommended that common eligibility criteria be modified to make trials more inclusive. We examined whether patterns of HIV-related eligibility criteria changed over time in relation to these recommendations. Methods: Trial eligibility criteria were abstracted from ClinicalTrials.gov for any phase, US-based clinical treatment trials for patients with cancer from 2010-2024. Trials were classified as not excluding, conditionally excluding (ie, excluding under specific circumstances such as unstable or untreated cases), or completely excluding patients with HIV. Eligibility criteria were classified using OpenAI language models with prompt refinement and independent validation. Interrupted time series analyses were used to determine whether the recommendations in 2017 were associated with changes in HIV criteria. We also examined whether eligibility-criteria changes varied by phase and lead sponsor (federal vs. industry). Results: We evaluated 20,751 trials, mostly phase I-II (80.3%), with 7.8% federally and 40.6% industry funded, including breast (19.1%), lung (11.0%), prostate (16.2%), and colorectal (14.1%) cancers. The AI classification achieved 97% accuracy on the development sample and 94% on an independent validation sample. Overall, patients with HIV were wholly excluded in 28.9%, conditionally excluded in 26.1%, and not excluded in 45.0% of trials. The guidance was associated with an increase in conditional exclusion (13.1% to 29.1%; p<.001) and no exclusion (24.1% to 28.9%; p=.002), and a decrease in wholly exclusion (62.8% to 42.0%; p<.001). Similar shift patterns were observed across trials with different main funding sources (federal vs industry) and phases (I–II vs III) (Table). Conclusions: Following the ASCO/FoCR/FDA task force recommendations, HIV exclusion criteria shifted predominantly from complete exclusion to conditional exclusion, indicating progress toward more inclusive clinical trial designs. Acknowledgement: We thank Dr. Roy Burstein, PhD, at the Institute for Disease Modeling, Gates Foundation, for his invaluable advice. HIV exclusion criteria before and after the guidance. Nov 2017- Dec 2025 Jan 2010-Oct 2017, No. (%) Without guidance (%) With guidance (% ) p All C 4175 (48.1) 62.8 42.0 <.001 Cond 1905 (22.0) 13.1 29.1 <.001 N 2597 (30.0) 24.1 28.9 <.001 Federal led C 342 (41.5) 41.4 22.3 <.001 Cond 299 (36.2) 36.2 57.2 <.001 N 184 (22.3) 22.3 20.5 0.40 Industry led C 1589 (48.7) 61.1 40.2 <.001 Cond 532 (16.3) 12.6 25.5 <.001 N 1141 (34.9) 26.3 32.6 <.001 Phase 1-2 C 3742 (53.8) 69.3 46.6 <.001 Cond 1622 (23.4) 13.4 30.1 <.001 N 1581 (22.8) 17.3 22.5 <.001 Phase 3 C 958 (46.3) 52.0 29.1 <.001 Cond 442 (21.4) 10.2 26.2 <.001 N 666 (32.2) 37.8 44.7 0.10 C/Cond/N: completely/conditionally/not excluded.

Real-world toxicity of bispecific antibodies versus CAR-T cell therapy in diffuse large B-cell lymphoma and follicular lymphoma.

7033 Background: CAR-T cell therapy and bispecific antibodies (BsAbs) have transformed treatment of relapsed or refractory B-cell lymphomas. While CAR-T toxicities are well characterized, real-world comparative safety data for newer BsAbs remain limited. We compared early and delayed toxicities of CAR-T versus BsAbs in diffuse large B-cell lymphoma (DLBCL) and follicular lymphoma (FL). Methods: Using the TriNetX Research Network, we conducted a retrospective cohort study of adults with DLBCL or FL treated with BsAbs or CAR-T therapy. 1:1 propensity score matching was performed for demographics, comorbidities, laboratory values, and prior lines of chemotherapy. Outcomes included cytokine release syndrome (CRS) at 14 and 30 days; immune effector cell–associated neurotoxicity syndrome (ICANS) at 30 and 60 days; ICU admission within 90 days; infections within 90 days; hypogammaglobulinemia within 180 days; neutropenia within 90 days; and treatment-related mortality at 90 days. Analyses used Kaplan–Meier methods, hazard ratios (HR), risk ratios (RR), and 95% confidence intervals (CI). Results: A total of 1,110 CAR-T–treated and 897 BsAb-treated patients met inclusion criteria. After matching, 384 patients were included per cohort. Compared with CAR-T, BsAbs were associated with lower CRS at 14 days (RR 0.45, 95% CI 0.37–0.55) and 30 days (RR 0.52, 95% CI 0.44–0.62), and lower ICANS at 30 days (RR 0.30, 95% CI 0.20–0.47) and 60 days (RR 0.36, 95% CI 0.25–0.51). ICU admission within 90 days occurred less frequently with BsAbs (RR 0.56, 95% CI 0.40–0.78). Infection rates at 90 days (RR 1.06, 95% CI 0.87–1.30) and hypogammaglobulinemia at 180 days (RR 0.81, 95% CI 0.65–1.01) were similar between groups. Neutropenia within 90 days was lower with BsAbs (RR 0.55, 95% CI 0.49–0.61). 90-day overall survival was lower with BsAbs (HR 2.99, 95% CI 1.89–4.73), likely driven by treatment selection bias, as BsAbs are frequently used in patients ineligible for CAR-T due to poor performance status, comorbidities, or aggressive disease. Conclusions: In this real-world cohort of DLBCL and FL, BsAbs were associated with substantially lower early immune-mediated toxicity and healthcare utilization compared with CAR-T therapy, while infectious and delayed immune complications were similar. These findings provide important safety data for newer bispecific agents. Outcome Bispecific Antibodies(n = 384) CAR-T(n = 384) HR/RR[95% CI] CRS (0–14 days) 25.3% 56.3% 0.45 [0.37-0.55] CRS (0–60 days) 31.8% 59.6% 0.53 [0.45-0.63] ICANS (0–30 days) 8.1% 22.4% 0.30 [0.20-0.47] ICANS (0–60 days) 9.4% 26.0% 0.36 [0.25-0.51] ICU admission (0–90 days) 12.0% 21.4% 0.56 [0.40-0.78] Infections (0–90 days) 34.6% 32.6% 1.06 [0.87-1.30] Neutropenia (0–90 days) 49.2% 89.6% 0.55 [0.49-0.61] Hypogammaglobulinemia (0–180 days) 27.1% 33.6% 0.81 [0.65-1.01] Overall Survival (0–90 days) 81.4% 93.2% 2.99 [1.89 – 4.73]

Anthracycline-free vs anthracycline-containing dual anti-HER2 neoadjuvant therapy: Real-world outcomes.

e12676 Background: The incremental value of anthracyclines with dual anti-HER2 neoadjuvant therapy (NAT) for early HER2+ breast cancer remains uncertain. We compared pCR, post-NAT escalation, and recurrence outcomes. Methods: Retrospective single-institution cohort of stage I–III HER2+ breast cancer treated with dual anti-HER2 NAT followed by surgery. Exposure: anthracycline-containing vs anthracycline-free TCHP. Primary endpoint: pCR. Secondary endpoints: post-NAT systemic therapy in non-pCR (trastuzumab emtansine [T-DM1] vs trastuzumab± pertuzumab), completion/discontinuation with reasons, and recurrence/RFS. Results: Of 112 evaluable patients, 74 (66.1%) received anthracycline-containing NAT and 38 (33.9%) TCHP. Overall pCR was 50.0% (56/112): 48.6% (36/74) vs 52.6% (20/38). pCR by subgroup (anthracycline vs TCHP): HR+ 43.2% vs 42.9%; HR− 56.7% vs 64.7%; cN0 52.4% vs 75.0%; cN+ 47.2% vs 42.3%; stage I–II 53.3% vs 61.9%; stage III 40.7% vs 41.2%. Among non-pCR (n = 56), post-NAT therapy was T-DM1 in 71.4% (40/56) and trastuzumab±pertuzumab in 26.8% (15/56) ; T-DM1 uptake was 78.9% (30/38) after anthracycline-containing NAT vs 55.6% (10/18) after TCHP. Early discontinuation occurred in 20.0% (8/40) for T-DM1 and 26.7% (4/15) for trastuzumab±pertuzumab; discontinuation due to toxicity was 87.5% and 50.0%, respectively. Cardiac-toxicity discontinuation was rare (1 per NAT arm). Recurrence occurred in 7.1% (4/56) with pCR vs 17.9% (10/56) without pCR; by NAT regimen 14.9% (11/74) vs 7.9% (3/38). Conclusions: Anthracycline-containing and anthracycline-free dual anti-HER2 NAT achieved comparable pCR in routine practice. Post-NAT escalation with T-DM1 was common; discontinuation was mainly toxicity-driven and cardiac discontinuation was rare. Larger cohorts with longer follow-up are needed to clarify comparative recurrence and safety outcomes.