Leveraging Large Language Models for Decision Support in Personalized Oncology.

Importance: Clinical interpretation of complex biomarkers for precision oncology currently requires manual investigations of previous studies and databases. Conversational large language models (LLMs) might be beneficial as automated tools for assisting clinical decision-making. Objective: To assess performance and define their role using 4 recent LLMs as support tools for precision oncology. Design, Setting, and Participants: This diagnostic study examined 10 fictional cases of patients with advanced cancer with genetic alterations. Each case was submitted to 4 different LLMs (ChatGPT, Galactica, Perplexity, and BioMedLM) and 1 expert physician to identify personalized treatment options in 2023. Treatment options were masked and presented to a molecular tumor board (MTB), whose members rated the likelihood of a treatment option coming from an LLM on a scale from 0 to 10 (0, extremely unlikely; 10, extremely likely) and decided whether the treatment option was clinically useful. Main Outcomes and Measures: Number of treatment options, precision, recall, F1 score of LLMs compared with human experts, recognizability, and usefulness of recommendations. Results: For 10 fictional cancer patients (4 with lung cancer, 6 with other; median [IQR] 3.5 [3.0-4.8] molecular alterations per patient), a median (IQR) number of 4.0 (4.0-4.0) compared with 3.0 (3.0-5.0), 7.5 (4.3-9.8), 11.5 (7.8-13.0), and 13.0 (11.3-21.5) treatment options each was identified by the human expert and 4 LLMs, respectively. When considering the expert as a criterion standard, LLM-proposed treatment options reached F1 scores of 0.04, 0.17, 0.14, and 0.19 across all patients combined. Combining treatment options from different LLMs allowed a precision of 0.29 and a recall of 0.29 for an F1 score of 0.29. LLM-generated treatment options were recognized as AI-generated with a median (IQR) 7.5 (5.3-9.0) points in contrast to 2.0 (1.0-3.0) points for manually annotated cases. A crucial reason for identifying AI-generated treatment options was insufficient accompanying evidence. For each patient, at least 1 LLM generated a treatment option that was considered helpful by MTB members. Two unique useful treatment options (including 1 unique treatment strategy) were identified only by LLM. Conclusions and Relevance: In this diagnostic study, treatment options of LLMs in precision oncology did not reach the quality and credibility of human experts; however, they generated helpful ideas that might have complemented established procedures. Considering technological progress, LLMs could play an increasingly important role in assisting with screening and selecting relevant biomedical literature to support evidence-based, personalized treatment decisions.

Frequency of Positive Familial Criteria in Patients with Adenocarcinoma of the Esophageal-Gastric Junction and Stomach: First Prospective Data in a Caucasian Cohort.

OBJECTIVES: Current prospective studies investigating the frequency of hereditary criteria in a Caucasian population for adenocarcinoma of the esophagogastric junction (AEG) and stomach (GC) are missing. Genetic testing criteria (screening criteria) for hereditary diffuse gastric cancer (HDGC) were updated in 2020, but do not address patients with intestinal histology (familial intestinal gastric cancer FIGC). Thus, we prospectively screened patients residing in Berlin newly diagnosed with AEG or GC for hereditary criteria to gain insights into the frequency of HDGC. METHODS: Prospective documentation of familial/clinical parameters in patients residing in Berlin with AEG or GC over three years was conducted. Besides HDGC criteria from 2015 and revised 2020, we also documented patients fulfilling these criteria but with intestinal type gastric cancer (FIGC). Statistical analysis was performed using X2-test. RESULTS: One hundred fifty-three patients were finally included (92 GC; male: 50 (n.s.); 61 AEG; male: 47; p = 0.007). Hereditary criteria for HDGC were detected in 9/92 (9.8%) (2015 criteria) and in 14/92 (15.2%) (2020 criteria) of GC patients (AEG: 2015 criteria 3/61 (4.9%) versus 4/61 according to 2020 criteria (6.5%)). Patients fulfilling hereditary criteria but with intestinal histology (FIGC) increased from 8.7% (2015) to 14.1%, respectively (2020) (AEG: 3.2% (2015) versus 6.6% (2020)). Hereditary criteria including intestinal histology were found in 29.3% (GC) and 13.1% (AEG) (p = 0.03) according to the 2020 criteria. CONCLUSIONS: HDGC criteria were found in 15.2% of GC patients according to the 2020 criteria. Percentage increased to 29.3% including patients with intestinal histology among the GC group, and was 13.1% in cases with AEG. These data indicate that family history seems to be of utmost importance in GC to further detect potential hereditary genetic risks. This equally applies for patients with intestinal subtype GC.

Prediction of clinical response to drugs in ovarian cancer using the chemotherapy resistance test (CTR-test).

BACKGROUND: In order to validate if the test result of the Chemotherapy Resistance Test (CTR-Test) is able to predict the resistances or sensitivities of tumors in ovarian cancer patients to drugs, the CTR-Test result and the corresponding clinical response of individual patients were correlated retrospectively. Results were compared to previous recorded correlations. METHODS: The CTR-Test was performed on tumor samples from 52 ovarian cancer patients for specific chemotherapeutic drugs. Patients were treated with monotherapies or drug combinations. Resistances were classified as extreme (ER), medium (MR) or slight (SR) resistance in the CTR-Test. Combination treatment resistances were transformed by a scoring system into these classifications. RESULTS: Accurate sensitivity prediction was accomplished in 79% of the cases and accurate prediction of resistance in 100% of the cases in the total data set. The data set of single agent treatment and drug combination treatment were analyzed individually. Single agent treatment lead to an accurate sensitivity in 44% of the cases and the drug combination to 95% accuracy. The detection of resistances was in both cases to 100% correct. ROC curve analysis indicates that the CTR-Test result correlates with the clinical response, at least for the combination chemotherapy. Those values are similar or better than the values from a publication from 1990. CONCLUSIONS: Chemotherapy resistance testing in vitro via the CTR-Test is able to accurately detect resistances in ovarian cancer patients. These numbers confirm and even exceed results published in 1990. Better sensitivity detection might be caused by a higher percentage of drug combinations tested in 2012 compared to 1990. Our study confirms the functionality of the CTR-Test to plan an efficient chemotherapeutic treatment for ovarian cancer patients.

Role of IGF-I in Primary Ovarian Cancer - A Study of the OVCAD European Consortium.

BACKGROUND/AIM: IGF-I (insulin growth factor 1) is crucially involved in cellular proliferation. Moreover, deregulation of IGF-I has been shown to be relevant in the carcinogenesis of various tumor entities. However, the impact of IGF-I in epithelial ovarian cancer (EOC) is unclear. In the present study, we investigated the predictive and prognostic role of circulatory IGF-I in primary EOC patients. PATIENTS AND METHODS: In the FP6 European Project "OVCAD", 275 consecutive primary EOC patients were enrolled. Patients were eligible if radical cytoreductive surgery and platinum-based chemotherapy were performed. Plasma IGF-I was detected using ELISA. RESULTS: Increased plasma IGF-I levels were more frequently found in well-differentiated epithelial ovarian carcinoma (p=0.0047). A weak correlation was observed between IGF-I levels and CA-125 in patients with serous EOC (p=0.04). No association between IGF-I expression and other clinico-pathological parameters was observed. CONCLUSION: IGF-I is overexpressed in patients with well-differentiated EOC. Further studies are warranted to elucidate the role of IGF-I in this sub-group of patients.

Lymph Node Involvement Pattern and Survival Differences of FIGO IIIC and FIGO IIIA1 Ovarian Cancer Patients After Primary Complete Tumor Debulking Surgery: A 10-Year Retrospective Analysis of the Tumor Bank Ovarian Cancer Network.

OBJECTIVE: The main goal of the current study was to compare survival differences among subgroups of primary ovarian cancer patients in International Federation of Gynecology and Obstetrics (FIGO) stages IIIC and IIIA1 after complete tumor debulking surgery. METHODS: A total of 218 patients with primary ovarian cancer who received complete cytoreductive surgery were included in the current retrospective analysis of the validated Tumor Bank Ovarian Cancer Network Database, which covers the periods January 2002 until December 2012. According to their tumor spread pattern, patients were divided into three groups: Group A (peritoneum only), Group B (peritoneum and lymph nodes), and Group C (lymph nodes only). Associations between groups and clinicopathological factors were analyzed using standard statistical procedures. RESULTS: The vast majority of patients were classified into Group B. Lymph node involvement was detected in 70.5 % of the cases where peritoneal implants presented ≥2 cm beyond the pelvis (Group A + B). The estimated 5-year overall survival (OS) rates were 47.4 % in Group A, 45.1 % in Group B, and 91.7 % in Group C (p < 0.01). In the subgroup analysis of Group B, both pelvic and para-aortic lymph node involvement was found in 57 % of patients. Patients in Group B who had para-aortic lymph node involvement only had better median progression-free survival (PFS) compared with patients with pelvic lymph node involvement only and pelvic and para-aortic lymph node involvement (28, 16, and 18 months, respectively; p = 0.02). The median OS differed significantly between patients with para-aortic lymph node involvement only versus patients with both pelvic and para-aortic involvement (68.5 vs. 46.7 months; p = 0.02). Three-year PFS was 90.0 % in FIGO IIIA1(i) and 62.6 % in FIGO IIIA1(ii) (hazard ratio 2.30, 95 % confidence interval 0.45-11.58). CONCLUSIONS: Patients with FIGO stage IIIC with lymph node involvement only had the best clinical outcome compared with patients in the same stage with peritoneal involvement only. Furthermore, involvement of both pelvic and para-aortic lymph nodes were of the same infrequency, and involvement of only para-aortic lymph nodes in this stage resulted in a better chance of survival than involvement of pelvic lymph nodes only or both pelvic and para-aortic lymph nodes simultaneously. In accordance with the revised FIGO classification of 2013, our study revealed that FIGO IIIA1(i) is prognostically better compared with FIGO IIIA1(ii).

The role of HE4 for prediction of recurrence in epithelial ovarian cancer patients-results from the OVCAD study.

Patients with epithelial ovarian cancer (EOC) are at high risk of tumor recurrence. Human epididymis protein 4 (HE4) has been shown to be overexpressed in EOC. The primary aim of our study was to evaluate the role of HE4 in predicting recurrence in EOC patients. Furthermore, we assessed the role of HE4 in predicting recurrence after second-line chemotherapy. We retrospectively analyzed data of 92 out of 275 primary EOC patients of the multicenter project "Ovarian Cancer: Diagnosis of a silent killer" (OVCAD). The concentrations of HE4 and CA125 were determined preoperatively and 6 months after the end of platinum-based first-line chemotherapy (FU) using ELISA and Luminex technique, respectively. The role of HE4 and CA125 for prediction of recurrence was determined using receiver operating characteristics (ROC) curves. Out of 92 patients included, 70 (76 %) were responders and 22 (23 %) non-responders in terms of response to platinum-based first-line chemotherapy. Median HE4 concentrations at follow-up (FU) differed between responders and non-responders (60.5 vs. 237.25 pM, p = 0.0001), respectively. The combined use of HE4 and CA125 at FU with cut-off values of 49.5 pM and 25 U/ml for HE4 and CA125, respectively, for predicting recurrence within 12 months after first-line chemotherapy performed better than HE4 or CA125 alone (area under the curve (AUC) 0.928, 95 % confidence intervals (CI) 0.838-1, p < 0.001). HE4 at FU could predict recurrence within 6 months after second-line chemotherapy (AUC 0.719, 95 % CI 0.553-0.885, p = 0.024). The combination of both elevated biomarkers revealed significantly worse estimated median progression-free survival (PFS; hazard ratio (HR) 8.14, 95 % CI 3.75-17.68, p < 0.001) and slightly worse PFS in those in whom only one biomarker was elevated (HR 1.46, 95 % CI 0.72-2.96, p = 0.292) compared to those patients in whom no biomarker was elevated. For the estimated median overall survival (OS), our analysis revealed similar results. HE4 in combination with CA125 performed better than CA125 and HE4 alone in predicting recurrence within 12 months after first-line chemotherapy.

Detection of soluble EpCAM (sEpCAM) in malignant ascites predicts poor overall survival in patients treated with catumaxomab.

EpCAM is an attractive target for cancer therapy and the EpCAM-specific antibody catumaxomab has been used for intraperitoneal treatment of EpCAM-positive cancer patients with malignant ascites. New prognostic markers are necessary to select patients that mostly benefit from catumaxomab. Recent data showed that soluble EpCAM (sEpCAM) is capable to block the effect of catumaxomab in vitro. This exploratory retrospective analysis was performed on archived ascites samples to evaluate the predictive role of sEpCAM in catumaxomab-treated patients. Sixty-six catumaxomab-treated patients with an available archived ascites sample were included in this study and tested for sEpCAM by sandwich ELISA. All probes were sampled before treatment start and all patients received at least one catumaxomab infusion. Overall survival, puncture-free survival and time to next puncture were compared between sEpCAM-positive and -negative patients. We detected sEpCAM in ascites samples of 9 patients (13.6%). These patients showed a significantly shorter overall survival. The prognostic significance of sEpCAM in ascites was particularly strong in patients with ovarian cancer. Puncture-free survival and time to next puncture were not significantly different between sEpCAM-positive and -negative patients. We propose sEpCAM in malignant ascites as a potential predictive marker in cancer patients treated with catumaxomab. Prospective studies with larger patients samples are urgently needed to confirm these findings and studies testing dose-intensified catumaxomab in patients with sEpCAM-positive ascites should be envisaged.

Inflammatory bowel disease (IBD)-like disease in a case of a 33-year old man with glycogenosis 1b.

BACKGROUND: Inflammatory bowel disease (IBD)-like conditions in glycogen storage disease (GSD) type Ib have been predominantly described in children. Signs and symptoms of GSD type Ib are hypoglycemia, pancytopenia and hepatosplenomegaly. Based on few published cases, there is evidence that granulocyte-colony stimulating factor (G-CSF) in patients with glycogenosis-related pancytopenia might ameliorate the IBD-like disease through leukocyte increase. CASE PRESENTATION: Here we firstly describe a case of an adult 33-year-old Caucasian male patient with GSD type Ib accompanied with IBD-like disease with persistent pancytopenia despite moderate-dose G-CSF treatment. Recent vomiting and abdominal discomfort were due to a high-grade stenosis in the transverse colon. A dose increase of the G-CSF successfully normalized his leukocyte count. However, the stenosis worsened and surgical therapy was needed. CONCLUSION: We suggest that symptomatic patients with GSD type Ib should undergo endoscopic examination in order to detect IBD-like disease and to initiate early treatment.

Preoperative HE4 and ROMA values do not improve the CA125 diagnostic value for borderline tumors of the ovary (BOT) - a study of the TOC Consortium.

BACKGROUND: Borderline tumors of the ovary (BOT) are a distinct entity of ovarian tumors, characterized by lack of stromal invasion. Recent studies postulated that the presence of invasive implants, incomplete staging, fertility sparing surgery and residual tumor after surgery are major prognostic factors for BOT. There are no biomarkers that can predict BOT or the presence of invasive implants. OBJECTIVE: The aim of our study was to assess the value of CA125 and HE4 alone, or within ROMA score for detecting BOT, and for predicting the presence of invasive implants. METHODS: Retrospective, monocentric study on 167 women diagnosed with BOT or benign ovarian masses. Serum HE4, CA125 levels and ROMA were assessed preoperatively. Due to low number of BOT with invasive implants, we performed an unmatched analysis (consecutive patients) and a matched analysis (according to age and histology) to compare BOT with invasive implants, BOT without invasive implants and benign disease. RESULTS: There were no significant differences in the HE4 and CA125 expressions in the three groups of patients (p = 0.984 and p = 0.141, respectively). The ROC analysis showed that CA125 alone is superior to ROMA and HE4 in discriminating patients with BOT with invasive implants from patients with benign diseases and BOT without invasive implants. A newly established score, ROMABOT, did not perform better than ROMA. The analysis of the matched groups revealed similar results as the analysis of all samples. CONCLUSIONS: Both HE4 and CA125 are not reliable biomarkers for the diagnosis of BOT or for predicting the presence of invasive implants.

Is endometrial cancer really a neurophobic tumor? A case report and review of the literature.

Brain metastases due to endometrial cancer are rare and usually occur in the context of widespread disease. We present a rare case of a 74-year-old woman with recurrent endometrial cancer in terms of a solitary brain lesion two years after initial diagnosis. She was treated with local resection of the brain metastasis and subsequent whole-brain radiotherapy. She then experienced relapse twice, presenting two solitary metastases at two different time points at the same location as at initial diagnosis, but never showed any signs of extracranial widespread disease. The patient has been alive for 13 months after detection of her initial brain metastasis. Despite the identification of some risk factors, there is still very limited knowledge why some patients develop brain metastases as the only sign of distant spread. Our review of the literature revealed that the combination of two treatment modalities yields higher survival rates than single treatment-alone, as was the case in the presented patient. Further case reports, as well as large and prospective studies, may contribute to a better understanding of the etiology and dynamics of this disease and allow better evaluation of treatment options.

Role of serum VEGFA, TIMP2, MMP2 and MMP9 in monitoring response to adjuvant radiochemotherapy in patients with primary cervical cancer--results of a companion protocol of the randomized NOGGO-AGO phase III clinical trial.

AIM: The aim of the current study was to analyze the type of variations in expression profiles of matrix metalloproteinase 2 (MMP2), matrix metalloproteinase 9 (MMP9), tissue inhibitor of metalloproteinase 2 (TIMP2), and vascular endothelial growth factor A (VEGFA) before and after radiochemotherapy in patients with locally advanced FIGO stage Ib-IIb cervical cancer. We analyzed the role of these biomarkers in monitoring response to treatment. PATIENTS AND METHODS: Serum from 72 patients with cervical cancer treated within a phase III trial with either simultaneous radiochemotherapy (S-RC) with cisplatin, or systemic paclitaxel and carboplatin followed by percutaneous radiation (PC-R) was analyzed by ELISA. Sera were obtained during surgery and after the end of adjuvant treatment. RESULTS: The median age at time of diagnosis was 46 years (range=30-71 years). The most common histological types were squamous cell (73.6%) and adenocarcinoma (25%). Thirty-five (48.6%) patients underwent surgery followed by S-RC and 37 (51.4%) patients were treated with surgery followed by PC-R. Five patients developed recurrence within six months after radiochemotherapy. VEGFA levels were significantly higher before and after adjuvant treatment in patients who developed early recurrence (p=0.001). An increase of more than 500 pg/ml VEGFA and a decrease of more than 9% of the pre-therapeutic value of TIMP2 were significantly associated with a higher risk of early recurrence (RR=8.5, 95% CI=1.8-39.8 and RR=11.0, 95% CI=2.5-48.2, respectively). TIMP2 expression and risk score for early relapse (which is calculated using values of VEGFA and TIMP2) were independent prognostic factors for overall survival (p=0.043, HR=0.96, 95% CI=0.93-0.99 and p=0.002, HR=1.09, 95% CI=1.03-1.15, respectively). CONCLUSION: Our results indicate a predictive value of VEGFA and TIMP2 in monitoring cervical cancer patients undergoing radiochemotherapy.

ERCC1 expression as a predictor of resistance to platinum-based chemotherapy in primary ovarian cancer.

AIM: The purpose of the present study was to investigate the possible association between Excision repair cross-complementing group 1 (ERCC1) score and platinum resistance in first-line chemotherapy for ovarian cancer. PATIENTS AND METHODS: ERCC1 Expression was determined using immunohisto-chemistry in 68 patients with platinum-responding tumor and 30 with platinum-resistant tumors. The primary end-point of this study was the association between the expression of ERCC1 protein with resistance to standard platinum-based chemotherapy in primary ovarian cancer. RESULTS: In pairwise comparisons, the overall survival (OS) for patients with ovarian cancer, who were non-responders to platinum-based chemotherapy with low or intermediate H-score for ERCC1 was better than that of non-responders with high H-score for ERCC1 [median OS=21 (16.8-25.2 months) and 28 (14.6-41.4 months) vs. 15 months (6.2-23.8 months), p-value=0.048, and p-value=0.017, respectively]. CONCLUSION: There were no significant differences in the progression-free survival between those with low, intermediate and high H-score for ERCC1. There is no statistically significant relationship between ERCC1 score and response to platinum-based chemotherapy in patients with primary ovarian cancer.

HE4 expression in plasma correlates with surgical outcome and overall survival in patients with first ovarian cancer relapse.

BACKGROUND: Epithelial ovarian cancer (EOC) remains the main cause of mortality due to gynecological malignancies. Optimal tumor debulking and platinum response are the most important prognostic factors for overall survival (OS) in primary EOC. In the setting of recurrence, the role of cytoreduction is not clear. A critical point is to predict preoperatively the subgroup of patients with optimal surgical outcome. The aim of the study was to analyze the predictive role of HE4 for surgical outcome and platinum response in EOC patients experiencing a first relapse. Secondary aims were the prognostic role of HE4 for OS and progression-free survival (PFS). METHODS: Plasma was obtained before secondary cytoreduction from 73 EOC patients. A total of 66.7 % underwent a total macroscopic tumor clearance; 86.3 % of the patients had disease that responded to platinum therapy. HE4 was detected by enzyme-linked immunosorbent assay. For statistical analysis, the chi-square test, Fisher's exact test, Kendall's tau b, and Mann-Whitney U test were used. OS, PFS rates, and respective 95 % confidence intervals (CI) were estimated according to the Kaplan-Meier method. RESULTS: At a HE4 cutoff value of 250 pMk, a sensitivity of 52 % and a specificity of 93.8 % (p = 0.001, 95 % CI 0.601-0.861) were reached in predicting total macroscopic tumor clearance. Plasma HE4 concentrations together with platinum response were the only independent prognostic factors for OS (p < 0.001, hazard ratio [HR] 18.77, 95 % CI 4.68-75.25; and p = 0.044, HR 3.33, 95 % CI 1.03-10.7, respectively). Together with ascites, HE4 was the only independent predictive factor for surgical outcome (p = 0.029, odds ratio [OR] 7.2, 95 % CI 1.22-42.19 and p = 0.036, OR 10.18, 95 % CI 1.16-88.69, respectively). CONCLUSIONS: HE4 is an independent predictive marker for surgical outcome and OS in patients with recurrent EOC. Larger population studies are needed to validate these results.

HE4 tissue expression in borderline ovarian tumors: a pilot study by the tumorbank ovarian cancer network.

AIM: Our purpose was to analyze the tissue expression of human epididymis protein-4 (HE4) in borderline tumors of the ovary (BOT) and to correlate it with histological subtypes and clinical features. PATIENTS AND METHODS: Tumor tissue samples from 25 patients with BOT were stained on tissue microarrays. The percentage of stained tumor cells was represented by grouped immunoreactivity scores (IRS) 0 to 4. RESULTS: The median patient age was 47 (range=22-73) years. Tumors in most patients (19/25) were staged-FIGO I and presented serous (52%) or mucinous (40%) histology. HE4 immunoreactivity occurred exclusively within the tumor cells. No association between grouped IRS and histological type, age, CA125 and FIGO stage was found. Correlation between HE4 positivity cells and HE4 IRS was significant (p<0.001). CONCLUSION: The role of HE4 in BOT remains unclear. Multicenter surveys are needed to more profoundly help in the understanding of the biological and clinical features of BOT.