Use of Natural Language Processing to Infer Sites of Metastatic Disease From Radiology Reports at Scale.

PURPOSE: To evaluate natural language processing (NLP) methods to infer metastatic sites from radiology reports. METHODS: A set of 4,522 computed tomography (CT) reports of 550 patients with 14 types of cancer was used to fine-tune four clinical large language models (LLMs) for multilabel classification of metastatic sites. We also developed an NLP information extraction (IE) system (on the basis of named entity recognition, assertion status detection, and relation extraction) for comparison. Model performances were measured by F1 scores on test and three external validation sets. The best model was used to facilitate analysis of metastatic frequencies in a cohort study of 6,555 patients with 53,838 CT reports. RESULTS: The RadBERT, BioBERT, GatorTron-base, and GatorTron-medium LLMs achieved F1 scores of 0.84, 0.87, 0.89, and 0.91, respectively, on the test set. The IE system performed best, achieving an F1 score of 0.93. F1 scores of the IE system by individual cancer type ranged from 0.89 to 0.96. The IE system attained F1 scores of 0.89, 0.83, and 0.81, respectively, on external validation sets including additional cancer types, positron emission tomography-CT ,and magnetic resonance imaging scans, respectively. In our cohort study, we found that for colorectal cancer, liver-only metastases were higher in de novo stage IV versus recurrent patients (29.7% v 12.2%; P < .001). Conversely, lung-only metastases were more frequent in recurrent versus de novo stage IV patients (17.2% v 7.3%; P < .001). CONCLUSION: We developed an IE system that accurately infers metastatic sites in multiple primary cancers from radiology reports. It has explainable methods and performs better than some clinical LLMs. The inferred metastatic phenotypes could enhance cancer research databases and clinical trial matching, and identify potential patients for oligometastatic interventions.

Inferring cancer disease response from radiology reports using large language models with data augmentation and prompting.

OBJECTIVE: To assess large language models on their ability to accurately infer cancer disease response from free-text radiology reports. MATERIALS AND METHODS: We assembled 10 602 computed tomography reports from cancer patients seen at a single institution. All reports were classified into: no evidence of disease, partial response, stable disease, or progressive disease. We applied transformer models, a bidirectional long short-term memory model, a convolutional neural network model, and conventional machine learning methods to this task. Data augmentation using sentence permutation with consistency loss as well as prompt-based fine-tuning were used on the best-performing models. Models were validated on a hold-out test set and an external validation set based on Response Evaluation Criteria in Solid Tumors (RECIST) classifications. RESULTS: The best-performing model was the GatorTron transformer which achieved an accuracy of 0.8916 on the test set and 0.8919 on the RECIST validation set. Data augmentation further improved the accuracy to 0.8976. Prompt-based fine-tuning did not further improve accuracy but was able to reduce the number of training reports to 500 while still achieving good performance. DISCUSSION: These models could be used by researchers to derive progression-free survival in large datasets. It may also serve as a decision support tool by providing clinicians an automated second opinion of disease response. CONCLUSIONS: Large clinical language models demonstrate potential to infer cancer disease response from radiology reports at scale. Data augmentation techniques are useful to further improve performance. Prompt-based fine-tuning can significantly reduce the size of the training dataset.

COVID-19 Severity and Waning Immunity After up to 4 mRNA Vaccine Doses in 73 608 Patients With Cancer and 621 475 Matched Controls in Singapore: A Nationwide Cohort Study.

Importance: Despite patients with cancer being at risk of poor outcomes from COVID-19, there are few published studies for vaccine efficacy in this group, with suboptimal immunogenicity and waning vaccine efficacy described in small studies being a concern. Objective: To assess the incidence rate of severe COVID-19 disease outcomes associated with the number of vaccine doses received and the waning of protection over time. Design, Setting, and Participants: A prospective multicenter observational cohort study was carried out over 2 time periods (September 15, 2021, to December 20, 2021 [delta wave], and January 20, 2022, to November 11, 2022 [omicron wave]) predominated by SARS-CoV-2 delta and omicron variants, respectively. Overall, 73 608 patients with cancer (23 217 active treatment, 50 391 cancer survivors) and 621 475 controls matched by age, sex, race and ethnicity, and socioeconomic status were included. Exposure: Vaccine doses received, from zero to 4 doses, and time elapsed since last vaccine dose. Outcomes: Competing-risk regression analyses were employed to account for competing risks of death in patients with cancer. Main outcomes were incidence rate ratios (IRRs) of COVID-19 infection, hospitalization, and severe disease (defined as requirement for supplemental oxygen, intensive care, or death). The IRRs stratified by time from last vaccine dose served as indicators of waning of vaccine effectiveness over time. Results: The mean (SD) age of actively treated patients with cancer, cancer survivors, and controls were 62.7 (14.7), 62.9 (12.6), and 61.8 (14.7) years, respectively. Of 73 608 patients with cancer, 27 170 (36.9%) were men; 60 100 (81.6%) were Chinese, 7432 (10.1%) Malay, 4597 (6.2%) Indian, and 1479 (2.0%) were of other races and ethnicities. The IRRs for the 3-dose and 4-dose vs the 2-dose group (reference) for COVID-19 hospitalization and severe disease were significantly lower during both the delta and omicron waves in cancer and control populations. The IRRs for severe disease in the 3-dose group for active treatment, cancer survivors, and controls were 0.14, 0.13, and 0.07 during the delta wave and 0.29, 0.19, and 0.21 during omicron wave, respectively. The IRRs for severe disease in the 4-dose group during the omicron wave were even lower at 0.13, 0.10 and 0.10, respectively. No waning of vaccine effectiveness against hospitalization and severe disease was seen beyond 5 months after a third dose, nor up to 5 months (the end of this study's follow-up) after a fourth dose. Conclusion: This cohort study provides evidence of the clinical effectiveness of mRNA-based vaccines against COVID-19 in patients with cancer. Longevity of immunity in preventing severe COVID-19 outcomes in actively treated patients with cancer, cancer survivors, and matched controls was observed at least 5 months after the third or fourth dose.

Efficacy, patterns of use and cost of Pertuzumab in the treatment of HER2+ metastatic breast cancer in Singapore: The National Cancer Centre Singapore experience.

BACKGROUND: Pertuzumab is a humanized anti-human epidermal growth factor receptor 2 (HER2) monoclonal antibody found in a Phase III clinical trial to significantly improve median survival in HER2 positive metastatic breast cancer (MBC) when used in combination with a taxane and Trastuzumab, and its clinical efficacy has transformed the therapeutic landscape of HER2-positive breast cancer. There are currently few reports on the pattern of use and value of Pertuzumab in real world settings. Our study describes the clinical efficacy and treatment costs of Pertuzumab in HER2-positive MBC treated in a tertiary cancer centre in Singapore in a predominantly Asian population. AIM: To investigate the clinical efficacy and treatment costs of Pertuzumab in HER2-positive MBC in an Asian population in Singapore. METHODS: A retrospective study of 304 HER2-positive MBC patients seen at National Cancer Centre Singapore between 2011-2017 was conducted. Demographic and clinical data were extracted from electronic medical records. Clinical characteristics and billing data of patients who received Pertuzumab were compared with those who did not. RESULTS: Thirty-one (62.0%) of the fifty (16.4%) patients who received Pertuzumab as first-line therapy. With a median follow-up of 21.5 mo, there was a statistically significant difference in the median overall survival between Pertuzumab and non-Pertuzumab groups [51.5 (95%CI: 35.8-60.0) vs 32.9 (95%CI: 28.1-37.5) mo; P = 0.0128]. Two (4.88%) patients in the Pertuzumab group experienced grade 3 (G3) cardiotoxicity. The median treatment cost incurred for total chemotherapy for the Pertuzumab group was 130456 Singapore Dollars compared to 34523 Singapore Dollars for the non-Pertuzumab group. The median percentage of total chemotherapy costs per patient in the Pertuzumab group spent on Pertuzumab was 50.3%. CONCLUSION: This study shows that Pertuzumab use in the treatment of metastatic breast cancer is associated with a significantly better survival and a low incidence of serious cardiotoxicity. However, the proportionate cost of Pertuzumab therapy remains high and further cost-effectiveness studies should be conducted.