Extracting Systemic Anticancer Therapy and Response Information From Clinical Notes Following the RECIST Definition.

PURPOSE: The RECIST guidelines provide a standardized approach for evaluating the response of cancer to treatment, allowing for consistent comparison of treatment efficacy across different therapies and patients. However, collecting such information from electronic health records manually can be extremely labor-intensive and time-consuming because of the complexity and volume of clinical notes. The aim of this study is to apply natural language processing (NLP) techniques to automate this process, minimizing manual data collection efforts, and improving the consistency and reliability of the results. METHODS: We proposed a complex, hybrid NLP system that automates the process of extracting, linking, and summarizing anticancer therapy and associated RECIST-like responses from narrative clinical text. The system consists of multiple machine learning-/deep learning-based and rule-based modules for diverse NLP tasks such as named entity recognition, assertion classification, relation extraction, and text normalization, to address different challenges associated with anticancer therapy and response information extraction. We then evaluated the system performances on two independent test sets from different institutions to demonstrate its effectiveness and generalizability. RESULTS: The system used domain-specific language models, BioBERT and BioClinicalBERT, for high-performance therapy mentions identification and RECIST responses extraction and categorization. The best-performing model achieved a 0.66 score in linking therapy and RECIST response mentions, with end-to-end performance peaking at 0.74 after relation normalization, indicating substantial efficacy with room for improvement. CONCLUSION: We developed, implemented, and tested an information extraction system from clinical notes for cancer treatment and efficacy assessment information. We expect this system will support future cancer research, particularly oncologic studies that focus on efficiently assessing the effectiveness and reliability of cancer therapeutics.

Optimizing Word Embeddings for Patient Portal Message Datasets with a Small Number of Samples.

Background: Patient portal messages often relate to specific clinical phenomena (e.g., patients undergoing treatment for breast cancer) and, as a result, have received increasing attention in biomedical research. These messages require natural language processing and, while word embedding models, such as word2vec, have the potential to extract meaningful signals from text, they are not readily applicable to patient portal messages. This is because embedding models typically require millions of training samples to sufficiently represent semantics, while the volume of patient portal messages associated with a particular clinical phenomenon is often relatively small. Objective: We introduce a novel adaptation of the word2vec model, PK-word2vec, for small-scale messages. Methods: PK-word2vec incorporates the most similar terms for medical words (including problems, treatments, and tests) and non-medical words from two pre-trained embedding models as prior knowledge to improve the training process. We applied PK-word2vec on patient portal messages in the Vanderbilt University Medical Center electric health record system sent by patients diagnosed with breast cancer from December 2004 to November 2017. We evaluated the model through a set of 1000 tasks, each of which compared the relevance of a given word to a group of the five most similar words generated by PK-word2vec and a group of the five most similar words generated by the standard word2vec model. We recruited 200 Amazon Mechanical Turk (AMT) workers and 7 medical students to perform the tasks. Results: The dataset was composed of 1,389 patient records and included 137,554 messages with 10,683 unique words. Prior knowledge was available for 7,981 non-medical and 1,116 medical words. In over 90% of the tasks, both reviewers indicated PK-word2vec generated more similar words than standard word2vec (p=0.01).The difference in the evaluation by AMT workers versus medical students was negligible for all comparisons of tasks' choices between the two groups of reviewers (p = 0.774 under a paired t-test). Conclusions: PK-word2vec can effectively learn word representations from a small message corpus, marking a significant advancement in processing patient portal messages.

Travel-Time Disparities in Access to Proton Beam Therapy for Cancer Treatment.

Importance: Proton beam therapy is an emerging radiotherapy treatment for patients with cancer that may produce similar outcomes as traditional photon-based therapy for many cancers while delivering lower amounts of toxic radiation to surrounding tissue. Geographic proximity to a proton facility is a critical component of ensuring equitable access both for indicated diagnoses and ongoing clinical trials. Objective: To characterize the distribution of proton facilities in the US, quantify drive-time access for the population, and investigate the likelihood of long commutes for certain population subgroups. Design, Setting, and Participants: This population-based cross-sectional study analyzed travel times to proton facilities in the US. Census tract variables in the contiguous US were measured between January 1, 2017, and December 31, 2021. Statistical analysis was performed from September to November 2023. Exposures: Drive time in minutes to nearest proton facility. Population totals and prevalence of specific factors measured from the American Community Survey: age; race and ethnicity; insurance, disability, and income status; vehicle availability; broadband access; and urbanicity. Main Outcomes and Measures: Poor access to proton facilities was defined as having a drive-time commute of at least 4 hours to the nearest location. Median drive time and percentage of population with poor access were calculated for the entire population and by population subgroups. Univariable and multivariable odds of poor access were also calculated for certain population subgroups. Results: Geographic access was considered for 327 536 032 residents of the contiguous US (60 594 624 [18.5%] Hispanic, 17 974 186 [5.5%] non-Hispanic Asian, 40 146 994 [12.3%] non-Hispanic Black, and 195 265 639 [59.6%] non-Hispanic White; 282 031 819 [86.1%] resided in urban counties). The median (IQR) drive time to the nearest proton facility was 96.1 (39.6-195.3) minutes; 119.8 million US residents (36.6%) lived within a 1-hour drive of the nearest proton facility, and 53.6 million (16.4%) required a commute of at least 4 hours. Persons identifying as non-Hispanic White had the longest median (IQR) commute time at 109.8 (48.0-197.6) minutes. Multivariable analysis identified rurality (odds ratio [OR], 2.45 [95% CI, 2.27-2.64]), age 65 years or older (OR, 1.09 [95% CI, 1.06-1.11]), and living below the federal poverty line (OR, 1.22 [1.20-1.25]) as factors associated with commute times of at least 4 hours. Conclusions and Relevance: This cross-sectional study of drive-time access to proton beam therapy found that disparities in access existed among certain populations in the US. These results suggest that such disparities present a barrier to an emerging technology in cancer treatment and inhibit equitable access to ongoing clinical trials.

Overcome the Limitation of Phenome-Wide Association Studies (PheWAS): Extension of PheWAS to Efficient and Robust Large-Scale ICD Codes Analysis.

The Phenome-wide association studies (PheWAS) have become widely used for efficient, high-throughput evaluation of relationship between a genetic factor and a large number of disease phenotypes, typically extracted from a DNA biobank linked with electronic medical records (EMR). Phecodes, billing code-derived disease case-control status, are usually used as outcome variables in PheWAS and logistic regression has been the standard choice of analysis method. Since the clinical diagnoses in EMR are often inaccurate with errors which can lead to biases in the odds ratio estimates, much effort has been put to accurately define the cases and controls to ensure an accurate analysis. Specifically in order to correctly classify controls in the population, an exclusion criteria list for each Phecode was manually compiled to obtain unbiased odds ratios. However, the accuracy of the list cannot be guaranteed without extensive data curation process. The costly curation process limits the efficiency of large-scale analyses that take full advantage of all structured phenotypic information available in EMR. Here, we proposed to estimate relative risks (RR) instead. We first demonstrated the desired nature of RR that overcomes the inaccuracy in the controls via theoretical formula. With simulation and real data application, we further confirmed that RR is unbiased without compiling exclusion criteria lists. With RR as estimates, we are able to efficiently extend PheWAS to a larger-scale, phenome construction agnostic analysis of phenotypes, using ICD 9/10 codes, which preserve much more disease-related clinical information than Phecodes.

SEETrials: Leveraging Large Language Models for Safety and Efficacy Extraction in Oncology Clinical Trials.

Background: Initial insights into oncology clinical trial outcomes are often gleaned manually from conference abstracts. We aimed to develop an automated system to extract safety and efficacy information from study abstracts with high precision and fine granularity, transforming them into computable data for timely clinical decision-making. Methods: We collected clinical trial abstracts from key conferences and PubMed (2012-2023). The SEETrials system was developed with four modules: preprocessing, prompt modeling, knowledge ingestion and postprocessing. We evaluated the system's performance qualitatively and quantitatively and assessed its generalizability across different cancer types- multiple myeloma (MM), breast, lung, lymphoma, and leukemia. Furthermore, the efficacy and safety of innovative therapies, including CAR-T, bispecific antibodies, and antibody-drug conjugates (ADC), in MM were analyzed across a large scale of clinical trial studies. Results: SEETrials achieved high precision (0.958), recall (sensitivity) (0.944), and F1 score (0.951) across 70 data elements present in the MM trial studies Generalizability tests on four additional cancers yielded precision, recall, and F1 scores within the 0.966-0.986 range. Variation in the distribution of safety and efficacy-related entities was observed across diverse therapies, with certain adverse events more common in specific treatments. Comparative performance analysis using overall response rate (ORR) and complete response (CR) highlighted differences among therapies: CAR-T (ORR: 88%, 95% CI: 84-92%; CR: 95%, 95% CI: 53-66%), bispecific antibodies (ORR: 64%, 95% CI: 55-73%; CR: 27%, 95% CI: 16-37%), and ADC (ORR: 51%, 95% CI: 37-65%; CR: 26%, 95% CI: 1-51%). Notable study heterogeneity was identified (>75% I 2 heterogeneity index scores) across several outcome entities analyzed within therapy subgroups. Conclusion: SEETrials demonstrated highly accurate data extraction and versatility across different therapeutics and various cancer domains. Its automated processing of large datasets facilitates nuanced data comparisons, promoting the swift and effective dissemination of clinical insights.

The evolving posology and administration of immune checkpoint inhibitors: subcutaneous formulations.

Immune checkpoint inhibitors (ICIs) have transformed cancer care. Recently, atezolizumab gained its first global approval in a subcutaneous (SC) formulation by the UK medicines regulator, being notable as the first time an FDA- and/or European Medicines Agency (EMA)-approved ICI has been licensed via this administration route. Here, we discuss this approval, other SC ICIs in development, and the benefits and challenges of this administration route, including potential implications for patient care.

Oral antivirals for COVID-19 among patients with cancer.

Purpose: Immunocompromised individuals, such as those diagnosed with cancer, are at a significantly higher risk for severe illness and mortality when infected with SARS-CoV-2 (COVID-19) than the general population. Two oral antiviral treatments are approved for COVID-19: Paxlovid® (nirmatrelvir/ritonavir) and Lagevrio® (molnupiravir). There is a paucity of data regarding the benefit from these antivirals among immunocompromised patients with cancer, and recent studies have questioned their efficacy among vaccinated patients, even those with risk factors for severe COVID-19. Methods: We evaluated the efficacy and safety of nirmatrelvir/ritonavir and molnupiravir in preventing severe illness and death using our database of 457 patients with cancer and COVID-19 from Brown University-affiliated hospitals. 67 patients received nirmatrelvir/ritonavir or molnupiravir and were compared to 56 concurrent controls who received no antiviral treatment despite being eligible to receive it. Results: Administration of nirmatrelvir/ritonavir or molnupiravir was associated with improved survival and lower 90-day all-cause and COVID-19-attributed mortality (p<0.05) and with lower peak O2 requirements (ordinal odds ratio [OR] 1.52, 95% confidence interval [CI] 0.92-2.56). Conclusion: Acknowledging the small size of our sample as a limitation, we concluded that early antiviral treatment might be beneficial to immunocompromised individuals, particularly those with cancer, when infected with SARS-CoV-2. Larger-scale, well-stratified studies are needed in this patient population.

Genomic Characterization and Clinical Outcomes of Patients with Peritoneal Metastases from the AACR GENIE Biopharma Collaborative Colorectal Cancer Registry.

Peritoneal metastases (PM) are common in metastatic colorectal cancer (mCRC). We aimed to characterize patients with mCRC and PM from a clinical and molecular perspective using the American Association of Cancer Research Genomics Evidence Neoplasia Information Exchange (GENIE) Biopharma Collaborative (BPC) registry. Patients' tumor samples underwent targeted next-generation sequencing. Clinical characteristics and treatment outcomes were collected retrospectively. Overall survival (OS) from advanced disease and progression-free survival (PFS) from start of cancer-directed drug regimen were estimated and adjusted for the left truncation bias. A total of 1,281 patients were analyzed, 244 (19%) had PM at time of advanced disease. PM were associated with female sex [OR: 1.67; 95% confidence interval (CI): 1.11-2.54; P = 0.014] and higher histologic grade (OR: 1.72; 95% CI: 1.08-2.71; P = 0.022), while rectal primary tumors were less frequent in patients with PM (OR: 0.51; 95% CI: 0.29-0.88; P < 0.001). APC occurred less frequently in patients with PM (N = 151, 64% vs. N = 788, 79%) while MED12 alterations occurred more frequently in patients with PM (N = 20, 10% vs. N = 32, 4%); differences in MED12 were not significant when restricting to oncogenic and likely oncogenic variants according to OncoKB. Patients with PM had worse OS (HR: 1.45; 95% CI: 1.16-1.81) after adjustment for independently significant clinical and genomic predictors. PFS from initiation of first-line treatment did not differ by presence of PM. In conclusion, PM were more frequent in females and right-sided primary tumors. Differences in frequencies of MED12 and APC alterations were identified between patients with and without PM. PM were associated with shorter OS but not with PFS from first-line treatment. SIGNIFICANCE: Utilizing the GENIE BPC registry, this study found that PM in patients with colorectal cancer occur more frequently in females and right-sided primary tumors and are associated with worse OS. In addition, we found a lower frequency of APC alterations and a higher frequency in MED12 alterations in patients with PM.

Introducing a Comprehensive Score of Systemic Anticancer Treatment Relevance.

Treatment patterns in systemic anticancer therapy (SACT) are extremely varied and complex. While professional society guidelines exist that suggest recommended treatment strategies, these guidelines are produced through an extremely laborious and sometimes opaque manual process, making it impossible for such guidelines to cover all relevant treatment scenarios. To complement these manually curated guidelines, we leveraged a database of 5818 clinical trials and 7012 supporting references from 1943-present to calculate a quantifiable "relevance score". In a pilot evaluation, this score was strongly associated with professional society guideline recommendations, while also providing relevance information on thousands of additional therapies. We show that this score also accurately illustrates trends in SACT adoption over time. We foresee that this score, which comprehensively evaluates the relevance of SACT overall and by cancer subtype, will have utility for clinical practitioners as well as researchers in real-world data.

Telehealth and Technology: New Directions in Cancer Care.

ABSTRACT: Telehealth is a broad concept that refers to any delivery of health care in real time using technologies to connect people or information that are not in the same physical location. Until fairly recently, telehealth was more aspiration than reality. This situation changed radically due in part to the COVID-19 pandemic, which led to a near-overnight inability for patients to be seen for routine management of chronic health conditions, including those with cancer. The purpose of this brief narrative review is to outline some areas where emerging and future technology may allow for innovations with specific implications for people with a current or past diagnosis of cancer, including underserved and/or historically excluded populations. Specific topics of telehealth are broadly covered in other areas of the special issue.

Graphical Presentations in Systemic Anticancer Treatment Network Meta-Analyses: A Systematic Review.

Network meta-analysis (NMA) draws conclusions about indirect comparisons of randomized clinical trials and is considered high-level evidence. Most NMA publications make use of network plots to portray results. Network plots are complex graphics that can have many visual attributes to portray useful information, such as node size, color, and graph layout. We analyzed the network plots from 162 NMAs of systemic anticancer therapy efficacy using a set of 16 attributes. Our review showed that the current state of network plot data visualizations within the NMA space lacks diversity and does not make use of many of the visual attributes available to convey information. More thoughtful design choices should be placed behind these important visualizations, which can carry clinical significance and help derive treatment plans for patients.

Frontal and anterior temporal hypometabolism post chemoradiation in head and neck cancer: A real-world PET study.

BACKGROUND AND PURPOSE: Adverse neurological effects after cancer therapy are common, but biomarkers to diagnose, monitor, or risk stratify patients are still not validated or used clinically. An accessible imaging method, such as fluorodeoxyglucose positron emission tomography (FDG PET) of the brain, could meet this gap and serve as a biomarker for functional brain changes. We utilized FDG PET to evaluate which brain regions are most susceptible to altered glucose metabolism after chemoradiation in patients with head and neck cancer (HNCa). METHODS: Real-world FDG PET images were acquired as standard of care before and after chemoradiation for HNCa in 68 patients. Linear mixed-effects voxelwise models assessed changes after chemoradiation in cerebral glucose metabolism quantified with standardized uptake value ratio (SUVR), covarying for follow-up time and patient demographics. RESULTS: Voxelwise analysis revealed two large clusters of decreased glucose metabolism in the medial frontal and polar temporal cortices following chemoradiation, with decreases of approximately 5% SUVR after therapy. CONCLUSIONS: These findings provide evidence that standard chemoradiation for HNCa can lead to decreased neuronal glucose metabolism, contributing to literature emphasizing the vulnerability of the frontal and anterior temporal lobes, especially in HNCa, where these areas may be particularly vulnerable to indirect radiation-induced injury. FDG PET shows promise as a sensitive biomarker for assessing these changes.

Cancer Informatics 2023: Data Sharing and Federating Learning Point Towards New Collaborative Opportunities.

OBJECTIVE: To summarize significant research contributions on cancer informatics published in 2022. METHODS: An extensive search using PubMed/MEDLINE was conducted to identify the scientific contributions published in 2022 that address topics in cancer informatics. The selection process comprised three steps: (i) ten candidate best papers were first selected by the two section editors, (ii) external reviewers from internationally renowned research teams reviewed each candidate best paper, and (iii) the final selection of three best papers was conducted by the editorial board of the Yearbook. RESULTS: The three selected best papers demonstrate advances in federated learning, drug synergy prediction, and utilization of clinical note data. CONCLUSION: Cancer informatics continues to mature as a subfield of biomedical informatics. Applications of informatics methods to data sharing and federated approaches are especially notable in 2022.

DeepPhe-CR: Natural Language Processing Software Services for Cancer Registrar Case Abstraction.

PURPOSE: Manual extraction of case details from patient records for cancer surveillance is a resource-intensive task. Natural Language Processing (NLP) techniques have been proposed for automating the identification of key details in clinical notes. Our goal was to develop NLP application programming interfaces (APIs) for integration into cancer registry data abstraction tools in a computer-assisted abstraction setting. METHODS: We used cancer registry manual abstraction processes to guide the design of DeepPhe-CR, a web-based NLP service API. The coding of key variables was performed through NLP methods validated using established workflows. A container-based implementation of the NLP methods and the supporting infrastructure was developed. Existing registry data abstraction software was modified to include results from DeepPhe-CR. An initial usability study with data registrars provided early validation of the feasibility of the DeepPhe-CR tools. RESULTS: API calls support submission of single documents and summarization of cases across one or more documents. The container-based implementation uses a REST router to handle requests and support a graph database for storing results. NLP modules extract topography, histology, behavior, laterality, and grade at 0.79-1.00 F1 across multiple cancer types (breast, prostate, lung, colorectal, ovary, and pediatric brain) from data of two population-based cancer registries. Usability study participants were able to use the tool effectively and expressed interest in the tool. CONCLUSION: The DeepPhe-CR system provides an architecture for building cancer-specific NLP tools directly into registrar workflows in a computer-assisted abstraction setting. Improved user interactions in client tools may be needed to realize the potential of these approaches.

Normalization of drug and therapeutic concepts with Thera-Py.

Objective: The diversity of nomenclature and naming strategies makes therapeutic terminology difficult to manage and harmonize. As the number and complexity of available therapeutic ontologies continues to increase, the need for harmonized cross-resource mappings is becoming increasingly apparent. This study creates harmonized concept mappings that enable the linking together of like-concepts despite source-dependent differences in data structure or semantic representation. Materials and Methods: For this study, we created Thera-Py, a Python package and web API that constructs searchable concepts for drugs and therapeutic terminologies using 9 public resources and thesauri. By using a directed graph approach, Thera-Py captures commonly used aliases, trade names, annotations, and associations for any given therapeutic and combines them under a single concept record. Results: We highlight the creation of 16 069 unique merged therapeutic concepts from 9 distinct sources using Thera-Py and observe an increase in overlap of therapeutic concepts in 2 or more knowledge bases after harmonization using Thera-Py (9.8%-41.8%). Conclusion: We observe that Thera-Py tends to normalize therapeutic concepts to their underlying active ingredients (excluding nondrug therapeutics, eg, radiation therapy, biologics), and unifies all available descriptors regardless of ontological origin.

Next-generation phenotyping: introducing phecodeX for enhanced discovery research in medical phenomics.

MOTIVATION: Phecodes are widely used and easily adapted phenotypes based on International Classification of Diseases codes. The current version of phecodes (v1.2) was designed primarily to study common/complex diseases diagnosed in adults; however, there are numerous limitations in the codes and their structure. RESULTS: Here, we present phecodeX, an expanded version of phecodes with a revised structure and 1,761 new codes. PhecodeX adds granularity to phenotypes in key disease domains that are under-represented in the current phecode structure-including infectious disease, pregnancy, congenital anomalies, and neonatology-and is a more robust representation of the medical phenome for global use in discovery research. AVAILABILITY AND IMPLEMENTATION: phecodeX is available at https://github.com/PheWAS/phecodeX.

Burden of AML, 1990-2019: Estimates From the Global Burden of Disease Study.

PURPOSE: AML accounts for 80% of acute leukemia in adults. While progress has been made in treating younger patients in the past 2 decades, there has been limited improvement for older patients until recently. This study examines the global and European Union (EU) 15+ trends in AML between 1990 and 2019. METHODS: We extracted age-standardized incidence rates (ASIRs), age-standardized death rates (ASMRs), and disability-adjusted life years, stratified by sex from the Global Burden of Disease Study database, and mortality-to-incidence ratio (MIR) were computed. Trends were compared using Joinpoint regression. RESULTS: The findings show a global increase in AML incidence for both sexes from 1990 to 2019. In the EU15+ countries, most countries exhibited an increase in ASIR for both sexes. Joinpoint revealed that globally for male patients, ASIR steadily increased until 2010, remained stable until 2015 followed by a decline till 2019. Similar trends were observed in female patients. For ASMR, although there was an increase globally and in most EU15+ countries, there was a statistically significant decrease in mortality rates globally and in the majority of EU15+ countries in recent years. MIR improved in both sexes globally. On age stratification, AML burden was highest among older groups (55 years and older), while the lowest rates were observed in younger than 20 years. CONCLUSION: The findings from our study indicate a global rise in AML incidence and mortality in both sexes and decrease in MIR from 1990 to 2019 suggesting a better survival. However, on Joinpoint analysis, there is no change in MIR in women in the past decade and past 4 years in men indicating plateau in survival trends despite recent advances.

Factors Associated With Altmetric Attention Scores for Randomized Phase III Cancer Clinical Trials.

PURPOSE: Altmetric Attention Scores (Altmetrics) are real-time measures of scientific impact and attention through various public outlets, including news, blogs, and social media. Herein, we aimed to describe and characterize the relationship between Altmetrics, conventional impact metrics, and features of published cancer clinical trials. METHODS: We identified two-arm phase III cancer randomized clinical trials with a superiority end point and publication date between 2015 and 2020 from HemOnc and tabulated the following data: Altmetric, study positivity, US Food and Drug Administration (FDA) registration trial status, cancer site/category, treatment context (curative or palliative), trial design, primary end point type, experimental/control arm modality, and journal tier. We further collected conventional bibliometrics including the number of citations and relative citation ratio (RCR) for all published studies. Multiple linear regression modeling identified clinical trial factors predictive of Altmetrics, with alpha = .05 defining statistical significance. RESULTS: Altmetrics were found for 681 (98%) of 698 publications, with a median score of 38.5 (IQR, 13-132.8). FDA registration studies (ß [95% CI], 84.7 [48.8 to 120.6]; P < .001), studies reporting on curative (as opposed to palliative) interventions (-29 [-53.7 to -4.4]; P = .02), genitourinary trials (73.2 [28.1 to 118.2]; P = .001), studies published in tier 1 journals (P < .001), and those with an increased number of citations per year (0.81 [0.66 to 0.95]; P < .001) were significantly associated with increased engagement as measured by Altmetrics. Furthermore, there was a strong correlation between all collected bibliometrics and Altmetrics (R2 = 0.63, 0.68, and 0.67; P < .001 for citation count, citations per year, and RCR, respectively). CONCLUSION: FDA registration trials describing curative interventions, studies published in traditionally defined high-impact journals, and genitourinary trial publications tend to have the greatest Altmetrics. We observed a strong relationship between Altmetrics and conventional bibliometrics. The significance and consequences of these relationships warrant further investigation.

Natural Language Processing to Automatically Extract the Presence and Severity of Esophagitis in Notes of Patients Undergoing Radiotherapy.

PURPOSE: Radiotherapy (RT) toxicities can impair survival and quality of life, yet remain understudied. Real-world evidence holds potential to improve our understanding of toxicities, but toxicity information is often only in clinical notes. We developed natural language processing (NLP) models to identify the presence and severity of esophagitis from notes of patients treated with thoracic RT. METHODS: Our corpus consisted of a gold-labeled data set of 1,524 clinical notes from 124 patients with lung cancer treated with RT, manually annotated for Common Terminology Criteria for Adverse Events (CTCAE) v5.0 esophagitis grade, and a silver-labeled data set of 2,420 notes from 1,832 patients from whom toxicity grades had been collected as structured data during clinical care. We fine-tuned statistical and pretrained Bidirectional Encoder Representations from Transformers-based models for three esophagitis classification tasks: task 1, no esophagitis versus grade 1-3; task 2, grade ≤1 versus >1; and task 3, no esophagitis versus grade 1 versus grade 2-3. Transferability was tested on 345 notes from patients with esophageal cancer undergoing RT. RESULTS: Fine-tuning of PubMedBERT yielded the best performance. The best macro-F1 was 0.92, 0.82, and 0.74 for tasks 1, 2, and 3, respectively. Selecting the most informative note sections during fine-tuning improved macro-F1 by ≥2% for all tasks. Silver-labeled data improved the macro-F1 by ≥3% across all tasks. For the esophageal cancer notes, the best macro-F1 was 0.73, 0.74, and 0.65 for tasks 1, 2, and 3, respectively, without additional fine-tuning. CONCLUSION: To our knowledge, this is the first effort to automatically extract esophagitis toxicity severity according to CTCAE guidelines from clinical notes. This provides proof of concept for NLP-based automated detailed toxicity monitoring in expanded domains.