JCO Clinical Cancer Informatics最新文献

Purpose: Stage in multiple myeloma (MM) is an essential measure of disease risk, but its measurement in large databases is often lacking. We aimed to develop and validate a natural language processing (NLP) algorithm to extract oncologists' documentation of stage in the national Veterans Affairs (VA) Healthcare System.

Methods: Using nationwide electronic health record (EHR) and cancer registry data from the VA Corporate Data Warehouse, we developed and validated a rule-based NLP algorithm to extract oncologist-determined MM stage. To that end, a clinician annotated MM stage within over 5,000 short snippets of clinical notes, and annotated MM stage at MM treatment initiation for 200 patients. These were allocated into snippet- and patient-level development and validation sets. We developed MM stage extraction and roll-up algorithms within the development sets. After the algorithms were finalized, we validated them using standard measures in held-out validation sets.

Results: We developed algorithms for three different MM staging systems that have been in widespread use (Revised International Staging System [R-ISS], International Staging System [ISS], and Durie-Salmon [DS]) and for stage reported without a clearly defined system. Precision and recall were uniformly high for MM stage at the snippet level, ranging from 0.92 to 0.99 for the different MM staging systems. Performance in identifying for MM stage at treatment initiation at the patient level was also excellent, with precision of 0.92, 0.96, 0.90, and 0.86 and recall of 0.99, 0.98, 0.94, and 0.92 for R-ISS, ISS, DS, and unclear stage, respectively.

Conclusion: Our MM stage extraction algorithm uses rule-based NLP and data aggregation to accurately measure MM stage documented in oncology notes and pathology reports in VA's national EHR system. It may be adapted to other systems where MM stage is recorded in clinical notes.

Purpose: People with cancer experience poorly controlled symptoms that persist between treatment visits. Automated digital technology can remotely monitor and facilitate symptom management at home. Essential to digital interventions is patient engagement, user satisfaction, and intervention benefits that are distributed across patient populations so as not to perpetuate inequities. We evaluated Symptom Care at Home (SCH), an automated digital platform, to determine patient engagement, satisfaction, and whether intervention subgroups gained similar symptom reduction benefits.

Methods: 358 patients with cancer receiving a course of chemotherapy were randomly assigned to SCH or usual care (UC). Both groups reported daily on 11 symptoms and completed the SF36 (Short Form Health Survey) monthly. SCH participants received immediate automated self-care coaching on reported symptoms. As needed, nurse practitioners followed up for poorly controlled symptoms.

Results: The average participant was White (83%), female (75%), and urban-dwelling (78.6%). Daily call adherence was 90% of expected days. Participants reported high user satisfaction. SCH participants had lower symptom burden than UC in all subgroups: age, sex, race, income, residence type, diagnosis, and stage (all P < .001 effect size 0.33-0.65), except for stages I and II cancers. Non-White and lower-income SCH participants gained a higher magnitude of symptom reduction than White participants and higher-income participants. Additionally, SCH men gained higher SF36 mental health (MH) benefit. There were no differences on other SF36 indices.

Conclusion: Participants were highly satisfied and consistently engaged the SCH platform. SCH men gained large MH improvements, perhaps from increased comfort in sharing concerns through automated interactions. Although all intervention subgroups benefited, non-White participants and those with lower income gained higher symptom reduction benefit, suggesting that systematic care through digital tools can overcome existing disparities in symptom care outcomes.

Purpose: Early prediction of response to immunotherapy may help guide patient management by identifying resistance to treatment and allowing adaptation of therapies. This analysis evaluated a mathematical model of response to immunotherapy that provides patient-specific prediction of outcome using the initial change in tumor size/burden from baseline to the first follow-up visit on standard imaging scans.

Methods: We applied the model to 600 patients with advanced solid tumors who received durvalumab in Study 1108, a phase I/II trial, and compared outcome prediction performance versus size-based criteria with RECIST version 1.1 best overall response (BOR), baseline circulating tumor (ct)DNA level, and other clinical/pathologic predictors of immunotherapy response.

Results: In multiple solid tumors, the mathematical parameter representing net tumor growth rate at the first on-treatment computed tomography (CT) scan assessed around 6 weeks after starting durvalumab (α₁) had a concordance index to predict overall survival (OS) of 0.66-0.77 on multivariate analyses. This measurement of early tumor dynamics significantly improved multivariate OS models that included standard RECIST v1.1 criteria, baseline ctDNA levels, and other clinical/pathologic factors in predicting OS. Furthermore, α₁ was assessed consistently at the first on-treatment CT scan, whereas all traditional RECIST BOR groups were confirmed only after this time.

Conclusion: These results support further exploring α₁ as an integral biomarker of response to immunotherapy. This biomarker may be predictive of further benefit and can be assessed before RECIST response groups can be assigned, potentially providing an opportunity to personalize oncologic management.

Purpose: AML is a hematologic cancer that is clinically heterogeneous, with a wide range of clinical outcomes. DNA methylation changes are a hallmark of AML but are not routinely used as a criterion for risk stratification. The aim of this study was to explore DNA methylation markers that could risk stratify patients with cytogenetically normal AML (CN-AML), currently classified as intermediate-risk.

Materials and methods: DNA methylation profiles in whole blood samples from 77 patients with CN-AML in The Cancer Genome Atlas (LAML cohort) were analyzed. Individual 5'-cytosine-phosphate-guanine-3' (CpG) sites were assessed for their ability to predict overall survival. The output was validated using DNA methylation profiles from bone marrow samples of 79 patients with CN-AML in a separate data set from the Gene Expression Omnibus.

Results: In the training set, using DNA methylation data derived from the 450K array, we identified 2,549 CpG sites that could potentially distinguish patients with CN-AML with an adverse prognosis (intermediate-poor) from those with a more favorable prognosis (intermediate-favorable) independent of age. Of these, 25 CpGs showed consistent prognostic potential across both the 450K and 27K array platforms. In a separate validation data set, nine of these 25 CpGs exhibited statistically significant differences in 2-year survival. These nine validated CpGs formed the basis for a combined prognostic biomarker panel, which includes an 8-CpG Somatic Panel and the methylation status of cg23947872. This panel displayed strong predictive ability for 2-year survival, 2-year progression-free survival, and complete remission in the validation cohort.

Conclusion: This study highlights DNA methylation profiling as a promising approach to enhance risk stratification in patients with CN-AML, potentially offering a pathway to more personalized treatment strategies.

Purpose: Patients with advanced renal cell carcinoma (RCC) face significant challenges, stemming both from the complexities of the disease itself and the adverse effects of treatments. This study evaluated the feasibility and acceptability of a mobile health (mHealth) application tailored for education and symptom management of patients with advanced RCC receiving combined immune checkpoint inhibitor and tyrosine kinase inhibitor (ICI-TKI) therapy.

Methods: The primary end points were acceptability and feasibility. Acceptability was defined as the proportion of patients approached who consented to participate, setting a benchmark of at least 50% for this metric. Feasibility was gauged by the completion rate of the intervention among the participants; it required at least 50% of participants to fully complete the intervention and at least 70% to finish half of the administered questionnaires. The secondary end points included knowledge assessment and patient-reported outcomes (PROs). PROs were evaluated using validated instruments. To discern the changes between pre- and post-educational module quiz scores, we used the Wilcoxon signed-rank test. Time-course data of PROs were visualized using line plots and then compared using paired t-tests.

Results: From November 2022 to July 2023, 20 of 22 (90%) patients approached for the study consented and enrolled. Of the enrolled patients, 60% completed all questionnaires and knowledge assessments at every time point and 75% completed at least half of the surveys and questionnaires. Significant pre/post differences were noted in two of six quizzes in the knowledge assessment. This study population did not experience a significant change in PRO scores after starting therapy.

Conclusion: The mHealth application designed for education and symptom management in patients with advanced RCC undergoing combination ICI-TKI has proven to be both acceptable and feasible, meeting previous research benchmarks.

Purpose: Patients with epithelial ovarian cancer (EOC) have an elevated risk for venous thromboembolism (VTE). To assess the risk of VTE, models were developed by statistical or machine learning algorithms. However, few models have accommodated deep learning (DL) algorithms in realistic clinical settings. We aimed to develop a predictive DL model, exploiting rich information from electronic health records (EHRs), including dynamic clinical features and the presence of competing risks.

Methods: We extracted EHRs of 1,268 patients diagnosed with EOC from January 2007 through December 2017 at the National Cancer Center, Korea. DL survival networks using fully connected layers, temporal attention, and recurrent neural networks were adopted and compared with multi-perceptron-based classification models. Prediction accuracy was independently validated in the data set of 423 patients newly diagnosed with EOC from January 2018 to December 2019. Personalized risk plots displaying the individual interval risk were developed.

Results: DL-based survival networks achieved a superior area under the receiver operating characteristic curve (AUROC) between 0.95 and 0.98 while the AUROC of classification models was between 0.85 and 0.90. As clinical information benefits the prediction accuracy, the proposed dynamic survival network outperformed other survival networks for the test and validation data set with the highest time-dependent concordance index (0.974, 0.975) and lowest Brier score (0.051, 0.049) at 6 months after a cancer diagnosis. Our visualization showed that the interval risk fluctuating along with the changes in longitudinal clinical features.

Conclusion: Adaption of dynamic patient clinical features and accounting for competing risks from EHRs into the DL algorithms demonstrated VTE risk prediction with high accuracy. Our results show that this novel dynamic survival network can provide personalized risk prediction with the potential to assist risk-based clinical intervention to prevent VTE among patients with EOC.

Purpose: Longitudinal patient tolerability data collected as part of randomized controlled trials are often summarized in a way that loses information and does not capture the treatment experience. To address this, we developed an interactive web application to empower clinicians and researchers to explore and visualize patient tolerability data.

Methods: We used adverse event (AE) data (Common Terminology Criteria for Adverse Events) and patient-reported outcomes (PROs) from the NSABP-B35 phase III clinical trial, which compared anastrozole with tamoxifen for breast cancer-free survival, to demonstrate the tools. An interactive web application was developed using R and the Shiny web application framework that generates Sankey diagrams to visualize AEs and PROs using four tools: AE Explorer, PRO Explorer, Cohort Explorer, and Custom Explorer.

Results: To illustrate how users can use the interactive tool, examples for each of the four applications are presented using data from the NSABP-B35 phase III trial and the NSABP-B30 trial for the Custom Explorer. In the AE and PRO explorers, users can select AEs or PROs to visualize within specified time periods and compare across treatments. In the cohort explorer, users can select a subset of patients with a specific symptom, severity, and treatment received to visualize the trajectory over time within a specified time interval. With the custom explorer, users can upload and visualize structured longitudinal toxicity and tolerability data.

Conclusion: We have created an interactive web application and tool for clinicians and researchers to explore and visualize clinical trial tolerability data. This adaptable tool can be extended for other clinical trial data visualization and incorporated into future patient-clinician interactions regarding treatment decisions.

Purpose: Nutritional status is an established driver of cancer outcomes, but there is an insufficient workforce of registered dietitians to meet patient needs for nutritional counseling. Artificial intelligence (AI) and machine learning (ML) afford the opportunity to expand access to guideline-based nutritional support.

Methods: An AI-based nutrition assistant called Ina was developed on the basis of a learning data set of >100,000 expert-curated interventions, peer-reviewed literature, and clinical guidelines, and provides a conversational text message-based patient interface to guide dietary habits and answer questions. Ina was implemented nationally in partnership with 25 advocacy organizations. Data on demographics, patient-reported outcomes, and utilization were systematically collected.

Results: Between July 2019 and August 2023, 3,310 users from all 50 states registered to use Ina. Users were 73% female; median age was 57 (range, 18-91) years; most common cancer types were genitourinary (22%), breast (21%), gynecologic (19%), GI (14%), and lung (12%). Users were medically complex, with 50% reporting Stage III to IV disease, 37% with metastases, and 50% with 2+ chronic conditions. Nutritional challenges were highly prevalent: 58% had overweight/obese BMIs, 83% reported barriers to good nutrition, and 42% had food allergies/intolerances. Levels of engagement were high: 68% texted questions to Ina; 79% completed surveys; median user retention was 8.8 months; 94% were satisfied with the platform; and 98% found the guidance helpful. In an evaluation of outcomes, 84% used the advice to guide diet; 47% used recommended recipes, 82% felt the program improved quality of life (QoL), and 88% reported improved symptom management.

Conclusion: Implementation of an evidence-based AI virtual dietitian is feasible and is reported by patients to be beneficial on diet, QoL, and symptom management. Ongoing evaluations are assessing impact on other outcomes.

Recent clinical trials in oncology have used increasingly complex methodologies, such as causal inference methods for intercurrent events, external control, and covariate adjustment, posing challenges in clarifying the estimand and underlying assumptions. This article proposes expressing causal structures using graphical tools-directed acyclic graphs (DAGs) and single-world intervention graphs (SWIGs)-in the planning phase of a clinical trial. It presents five rules for selecting a sufficient set of adjustment variables on the basis of a diagram representing the clinical trial, along with three case studies of randomized and single-arm trials and a brief tutorial on DAG and SWIG. Through the case studies, DAGs appear effective in clarifying assumptions for identifying causal effects, although SWIGs should complement DAGs due to their limitations in the presence of intercurrent events in oncology research.

Purpose: The estimation of prognosis and life expectancy is critical in the care of patients with advanced cancer. To aid clinical decision making, we build a prognostic strategy combining a machine learning (ML) model with explainable artificial intelligence to predict 1-year survival after palliative radiotherapy (RT) for bone metastasis.

Materials and methods: Data collected in the multicentric PRAIS trial were extracted for 574 eligible adults diagnosed with metastatic cancer. The primary end point was the overall survival (OS) at 1 year (1-year OS) after the start of RT. Candidate covariate predictors consisted of 13 clinical and tumor-related pre-RT patient characteristics, seven dosimetric and treatment-related variables, and 45 pre-RT laboratory variables. ML models were developed and internally validated using the Python package. The effectiveness of each model was evaluated in terms of discrimination. A Shapley Additive Explanations (SHAP) explainability analysis to infer the global and local feature importance and to understand the reasons for correct and misclassified predictions was performed.

Results: The best-performing model for the classification of 1-year OS was the extreme gradient boosting algorithm, with AUC and F1-score values equal to 0.805 and 0.802, respectively. The SHAP technique revealed that higher chance of 1-year survival is associated with low values of interleukin-8, higher values of hemoglobin and lymphocyte count, and the nonuse of steroids.

Conclusion: An explainable ML approach can provide a reliable prediction of 1-year survival after RT in patients with advanced cancer. The implementation of SHAP analysis provides an intelligible explanation of individualized risk prediction, enabling oncologists to identify the best strategy for patient stratification and treatment selection.