With the recent diffusion of access to publicly available large language models (LLMs), common interest in generative artificial-intelligence-based applications for medical purposes has skyrocketed. The increased use of these models by tech-savvy patients for personal health issues calls for a scientific evaluation of whether LLMs provide a satisfactory level of accuracy for treatment decisions. This observational study compares the concordance of treatment recommendations from the popular LLM ChatGPT 3.5 with those of a multidisciplinary tumor board for breast cancer (MTB). The study design builds on previous findings by combining an extended input model with patient profiles reflecting patho- and immunomorphological diversity of primary breast cancer, including primary metastasis and precancerous tumor stages. Overall concordance between the LLM and MTB is reached for half of the patient profiles, including precancerous lesions. In the assessment of invasive breast cancer profiles, the concordance amounts to 58.8%. Nevertheless, as the LLM makes considerably fraudulent decisions at times, we do not identify the current development status of publicly available LLMs to be adequate as a support tool for tumor boards. Gynecological oncologists should familiarize themselves with the capabilities of LLMs in order to understand and utilize their potential while keeping in mind potential risks and limitations.

Activating PIK3CA mutations, present in up to 40% of hormone receptor-positive (HR⁺), human epidermal growth factor receptor 2-negative (Her2⁻) breast cancer (BC) patients, can be effectively targeted with the alpha isoform-specific PI3K inhibitor Alpelisib. This treatment significantly improves outcomes for HR⁺, Her2⁻, and PIK3CA-mutated metastatic BC patients. However, acquired resistance, often due to aberrant activation of the mTOR complex 1 (mTORC1) pathway, remains a significant clinical challenge. Our study, using in vitro and orthotopic xenograft mouse models, demonstrates that constitutively active mTORC1 signaling renders PI3K inhibitor-resistant BC exquisitely sensitive to various drugs targeting cancer metabolism. Mechanistically, mTORC1 suppresses the induction of autophagy during metabolic perturbation, leading to energy stress, a critical depletion of aspartate, and ultimately cell death. Supporting this mechanism, BC cells with CRISPR/Cas9-engineered knockouts of canonical autophagy genes showed similar vulnerability to metabolically active drugs. In BC patients, high mTORC1 activity, indicated by 4E-BP1^T37/46 phosphorylation, correlated with p62 accumulation, a sign of impaired autophagy. Together, these markers predicted poor overall survival in multiple BC subgroups. Our findings reveal that aberrant mTORC1 signaling, a common cause of PI3K inhibitor resistance in BC, creates a druggable metabolic vulnerability by suppressing autophagy. Additionally, the combination of 4E-BP1^T37/46 phosphorylation and p62 accumulation serves as a biomarker for poor overall survival, suggesting their potential utility in identifying BC patients who may benefit from metabolic therapies.

Purpose: Large language models (LLM) show potential for decision support in breast cancer care. Their use in clinical care is currently prohibited by lack of control over sources used for decision-making, explainability of the decision-making process and health data security issues. Recent development of Small Language Models (SLM) is discussed to address these challenges. This preclinical proof-of-concept study tailors an open-source SLM to the German breast cancer guideline (BC-SLM) to evaluate initial clinical accuracy and technical functionality in a preclinical simulation. Methods: A multidisciplinary tumor board (MTB) is used as the gold-standard to assess the initial clinical accuracy in terms of concordance of the BC-SLM with MTB and comparing it to two publicly available LLM, ChatGPT3.5 and 4. The study includes 20 fictional patient profiles and recommendations for 5 treatment modalities, resulting in 100 binary treatment recommendations (recommended or not recommended). Statistical evaluation includes concordance with MTB in % including Cohen’s Kappa statistic (κ). Technical functionality is assessed qualitatively in terms of local hosting, adherence to the guideline and information retrieval. Results: The overall concordance amounts to 86% for BC-SLM (κ = 0.721, p < 0.001), 90% for ChatGPT4 (κ = 0.820, p < 0.001) and 83% for ChatGPT3.5 (κ = 0.661, p < 0.001). Specific concordance for each treatment modality ranges from 65 to 100% for BC-SLM, 85–100% for ChatGPT4, and 55–95% for ChatGPT3.5. The BC-SLM is locally functional, adheres to the standards of the German breast cancer guideline and provides referenced sections for its decision-making. Conclusion: The tailored BC-SLM shows initial clinical accuracy and technical functionality, with concordance to the MTB that is comparable to publicly-available LLMs like ChatGPT4 and 3.5. This serves as a proof-of-concept for adapting a SLM to an oncological disease and its guideline to address prevailing issues with LLM by ensuring decision transparency, explainability, source control, and data security, which represents a necessary step towards clinical validation and safe use of language models in clinical oncology.

Background: Precision oncology treatments are being applied more commonly in breast and gynecological oncology through the implementation of Molecular Tumor Boards (MTBs), but real-world clinical outcome data remain limited. Methods: A retrospective analysis was conducted in patients with breast cancer (BC) and gynecological malignancies referred to our center’s MTB from 2018 to 2023. The analysis covered patient characteristics, next-generation sequencing (NGS) results, MTB recommendations, therapy received, and clinical outcomes. Results: Sixty-three patients (77.8%) had metastatic disease, and forty-four patients (54.3%) had previously undergone three or more lines of systemic treatment. Personalized treatment recommendations were provided to 50 patients (63.3%), while 29 (36.7%) had no actionable target. Ultimately, 23 patients (29.1%) underwent molecular-matched treatment (MMT). Commonly altered genes in patients with pan-gyn tumors (BC and gynecological malignancies) included TP53 (n = 42/81, 51.9%), PIK3CA (n = 18/81, 22.2%), BRCA1/2 (n = 10/81, 12.3%), and ARID1A (n = 9/81, 11.1%). Patients treated with MMT showed significantly prolonged progression-free survival (median PFS 5.5 vs. 3.5 months, p = 0.0014). Of all patients who underwent molecular profiling, 13.6% experienced a major clinical benefit (PFSr ≥ 1.3 and PR/SD ≥ 6 months) through precision oncology. Conclusions: NGS-guided precision oncology demonstrated improved clinical outcomes in a subgroup of patients with gynecological and breast cancers.

Purpose: The aim of this study was to analyze the prevalence of breast cancer in relation to body height and to investigate associations between body height and breast cancer in Germany. Methods: This retrospective cohort study included 135,741 adult women followed in one of 161 gynecology practices in Germany between January 2019 and December 2021. The 3 year prevalence of breast cancer (ICD-10: C50) during the study period was shown in relation to body height, which was included in this study as a five-category variable for women: ≤ 160 cm, 161–165 cm, 166–170 cm, 171–175 cm, > 175 cm. The associations between height and breast cancer were analyzed using logistic regression models adjusted for age and BMI. Results: The prevalence of breast cancer increased from 5.1% in women ≤ 160 cm to 6.8% in women > 175 cm in the age group 51–60, and from 9.2% in women ≤ 160 cm to 12.2% in women 171–175 cm in the age group > 60 years. The OR for breast cancer was 1.18 (95% CI 1.12–1.24) for every 10 cm increase in height. Compared to height ≤ 160 cm, the OR for height 166–170 cm was 1.26 (1.15–1.39), for 171–175 cm 1.43 (1.27–1.61), and for > 175 cm 1.49 (1.28–1.74). Conclusion: The results of this study suggest that greater body height in women is significantly related to an increased breast cancer risk.

Purpose: The aim of this study was to analyze the persistence of women on tamoxifen (TAM) and aromatase inhibitors (AIs) in Germany, and to investigate possible determinants of non-persistence. Methods: The present retrospective cohort study was based on the IQVIA longitudinal prescription database (LRx). The study included women with an initial prescription of TAM or AIs (anastrozole, letrozole, and exemestane) between January 2016 and December 2020 (index date). Kaplan–Meier analyses were performed to show the persistence for TAM and AI, using a therapy gap of 90 or 180 days, respectively. A multivariable Cox proportional hazards regression model was further used to estimate the relationship between non-persistence and drug prescription (AI versus TAM), age, and the specialty of the physician initiating therapy (gynecologist, oncologist, or general practitioner). Results: Up to 5 years after the index date, only 35.1% of AI and 32.5% of TAM patients were continuing therapy when therapy discontinuation was defined as at least 90 days without therapy. Using a 180-day therapy gap, 51.9% of AI and 50.4% of TAM patients remained on therapy after 5 years. Cox regression models reveal that initial therapy with TAM (HR 1.06, 95% CI 1.04–1.07), therapy initiation by oncologists (HR 1.09, 95% CI 1.07–1.11), or general practitioners (HR 1.24, 95% CI 1.21–1.27) and age ≤ 50 (HR 1.08, 95% CI 1.06–1.10) were significantly associated with an increased risk of therapy discontinuation. Conclusion: Overall, the present study indicates that persistence rates are low in all age groups for both TAM and AI treatment. We found several factors (e.g., physician specialty, younger age, and type of endocrine therapy) to be associated with an increased risk for non-persistence.