JCO Clinical Cancer Informatics最新文献

{"title":"Timing and Method of Patient-Provider Communication for Abnormal Hepatocellular Carcinoma Screening Results in Cirrhosis.","authors":"Jeremy Louissaint, Beverly Kyalwazi, John Deng, Timothy P Hogan, Robert W Turer, Elliot B Tapper, David E Gerber, Bryan D Steitz, Sarah R Lieber, Amit G Singal","doi":"10.1200/CCI-24-00269","DOIUrl":"https://doi.org/10.1200/CCI-24-00269","url":null,"abstract":"<p><strong>Purpose: </strong>Patients with cirrhosis undergo frequent abdominal imaging including semiannual hepatocellular carcinoma (HCC) screening, with results released immediately via the patient portal. We characterized time from patient review to patient-provider communication (PPC) for patients with abnormal liver imaging results.</p><p><strong>Methods: </strong>We identified patients with cirrhosis enrolled in the patient portal with a new abnormal liver lesion (LI-RADS, LR) on ambulatory liver ultrasound (US) or multiphasic computed tomography/magnetic resonance imaging. Imaging findings were grouped into low-risk (US-2, LR-2), intermediate-risk (US-3, LR-3), and high-risk (LR-4, LR-5, LR-M, LR-TIV) results. We extracted three date-time events from the electronic health record, including result release to the patient, patient review of the result, and result-related PPC. We compared communication methods and the median time with PPC after patient review of results between groups.</p><p><strong>Results: </strong>The cohort included 133 patients (median age, 62 years, 56% male) with 34 (25.6%) low-risk, 61 (45.9%) intermediate-risk, and 38 (28.6%) high-risk results. PPC for high-risk results was predominantly via telephone calls (60.5%), whereas portal messages were most commonly used for low- and intermediate-risk results (61.8% and 45.9%, respectively; <i>P</i> < .001). For patients who reviewed their result on the portal, most (79.3%) reviewed the result before PPC, among whom the median time between review and PPC was 55.8 (IQR, 22.0-219.0), 167 (IQR, 42.7-324.0), and 47.3 (IQR, 25.8-78.8) hours for low-, intermediate-, and high-risk results, respectively (<i>P</i> = .02).</p><p><strong>Conclusion: </strong>Portal-based review of abnormal imaging results by patients before provider communication is common, including results concerning a new HCC diagnosis. Further studies are needed to evaluate patient-reported outcomes, such as psychological distress, associated with this method of disclosing cancer-related results.</p>","PeriodicalId":51626,"journal":{"name":"JCO Clinical Cancer Informatics","volume":"9 ","pages":"e2400269"},"PeriodicalIF":3.3,"publicationDate":"2025-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144052306","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Data Events Are Safety Events: High-Reliability Organization Approach to Improving Data Quality and Safety.","authors":"Kelly Merriman, Emily Yu, Andrea Hawkins-Daarud, Kerin Adelson, Kenna Shaw, Dan Shoenthal, Jay Patel, Janna Baganz, Andy Futreal, David A Jaffray, Jose Rivera, Caroline Chung","doi":"10.1200/CCI-24-00273","DOIUrl":"https://doi.org/10.1200/CCI-24-00273","url":null,"abstract":"","PeriodicalId":51626,"journal":{"name":"JCO Clinical Cancer Informatics","volume":"9 ","pages":"e2400273"},"PeriodicalIF":3.3,"publicationDate":"2025-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12058365/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144057168","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Error Reduction in Leukemia Machine Learning Classification With Conformal Prediction.","authors":"Mariya Lysenkova Wiklander, Dave Zachariah, Olga Krali, Jessica Nordlund","doi":"10.1200/CCI-24-00324","DOIUrl":"https://doi.org/10.1200/CCI-24-00324","url":null,"abstract":"<p><strong>Purpose: </strong>Recent advances in machine learning have led to the development of classifiers that predict molecular subtypes of acute lymphoblastic leukemia (ALL) using RNA-sequencing (RNA-seq) data. Although these models have shown promising results, they often lack robust performance guarantees. The aim of this study was three-fold: to quantify the uncertainty of these classifiers, to provide prediction sets that control the false-negative rate (FNR), and to perform implicit error reduction by transforming incorrect predictions into uncertain predictions.</p><p><strong>Methods: </strong>Conformal prediction (CP) is a distribution-agnostic framework for generating statistically calibrated prediction sets whose size reflects model uncertainty. In this study, we applied an extension called conformal risk control to three RNA-seq ALL subtype classifiers. Leveraging RNA-seq data from 1,227 patient samples taken at diagnosis, we developed a multiclass conformal predictor ALLCoP, which generates statistically guaranteed FNR-controlled prediction sets.</p><p><strong>Results: </strong>ALLCoP was able to create prediction sets with specified FNR tolerances ranging from 7.5% to 30%. In a validation cohort, ALLCoP successfully reduced the FNR of the ALLIUM RNA-seq ALL subtype classifier from 8.95% to 3.5%. For patients whose subtype was not previously known, the use of ALLCoP was able to reduce the occurrence of empty predictions from 37% to 17%. Notably, up to 34% of the multiple-class prediction sets included the <i>PAX5</i>alt subtype, suggesting that increased prediction set size may reflect secondary aberrations and biological complexity, contributing to classifier uncertainty. Finally, ALLCoP was validated on two additional RNA-seq ALL subtype classifiers, ALLSorts and ALLCatchR.</p><p><strong>Conclusion: </strong>Our results highlight the potential of CP in enhancing the use of oncologic RNA-seq subtyping classifiers and also in uncovering additional molecular aberrations of potential clinical importance.</p>","PeriodicalId":51626,"journal":{"name":"JCO Clinical Cancer Informatics","volume":"9 ","pages":"e2400324"},"PeriodicalIF":3.3,"publicationDate":"2025-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144175805","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Optimizing Strategy for Lung Cancer Screening: From Risk Prediction to Clinical Decision Support.","authors":"Hao Dai, Yu Huang, Xing He, Tiancheng Zhou, Yuxi Liu, Xuhong Zhang, Yi Guo, Jingchuan Guo, Jiang Bian","doi":"10.1200/CCI-24-00291","DOIUrl":"https://doi.org/10.1200/CCI-24-00291","url":null,"abstract":"<p><strong>Purpose: </strong>Low-dose computed tomography (LDCT) screening is effective in reducing lung cancer mortality by detecting the disease at earlier, more treatable stages. However, high false-positive rates and the associated risks of subsequent invasive diagnostic procedures present significant challenges. This study proposes an advanced pipeline that integrates machine learning (ML) and causal inference techniques to optimize lung cancer screening decisions.</p><p><strong>Materials and methods: </strong>Using real-world data from the OneFlorida+ Clinical Research Consortium, we developed ML models to predict individual lung cancer risk and estimate the benefits of LDCT screening. Explainable artificial intelligence techniques were applied to identify key risk factors, ensuring transparency and trust in the model's predictions. Causal ML methods were used to estimate individualized treatment effects of LDCT screening, answering the critical what-if question regarding risk reduction from LDCT.</p><p><strong>Results: </strong>We defined a high-risk cohort of 5,947 patients who underwent LDCT, along with matched controls, to evaluate the framework. Our models demonstrated predictive performance with AUCs of 0.777 and 0.793 for 1-year and 3-year risk predictions, respectively. Causal modeling showed a consistent reduction in lung cancer risk across different subgroups due to LDCT. Specifically, the doubly robust model showed an average risk reduction of 9.5% for males and 12% for females. Age-stratified results indicated a 9.5% reduction for individuals age 50-60 years, a 7.5% reduction for those age 60-70 years, and the largest reduction of 15.1% for the 70-80 age group.</p><p><strong>Conclusion: </strong>Integrating ML and causal inference into clinical workflows offers a robust tool for enhancing lung cancer screening. This pipeline provides accurate risk assessments and actionable insights tailored to individuals, empowering clinicians and patients to make informed screening decisions. The differential risk reduction across subgroups highlights the importance of personalized screening in improving outcomes for populations at risk of lung cancer.</p>","PeriodicalId":51626,"journal":{"name":"JCO Clinical Cancer Informatics","volume":"9 ","pages":"e2400291"},"PeriodicalIF":3.3,"publicationDate":"2025-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12061033/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144057452","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Application of Generative Artificial Intelligence for Physician and Patient Oncology Letters-AI-OncLetters.","authors":"Adel Shahnam, Udit Nindra, Nadia Hitchen, Joanne Tang, Martin Hong, Jun Hee Hong, George Au-Yeung, Wei Chua, Weng Ng, Ashley M Hopkins, Michael J Sorich","doi":"10.1200/CCI-24-00323","DOIUrl":"https://doi.org/10.1200/CCI-24-00323","url":null,"abstract":"<p><strong>Purpose: </strong>Although large language models (LLMs) are increasingly used in clinical practice, formal assessments of their quality, accuracy, and effectiveness in medical oncology remain limited. We aimed to evaluate the ability of ChatGPT, an LLM, to generate physician and patient letters from clinical case notes.</p><p><strong>Methods: </strong>Six oncologists created 29 (four training, 25 final) synthetic oncology case notes. Structured prompts for ChatGPT were iteratively developed using the four training cases; once finalized, 25 physician-directed and patient-directed letters were generated. These underwent evaluation by expert consumers and oncologists for accuracy, relevance, and readability using Likert scales. The patient letters were also assessed with the Patient Education Materials Assessment Tool for Print (PEMAT-P), Flesch Reading Ease, and Simple Measure of Gobbledygook index.</p><p><strong>Results: </strong>Among physician-to-physician letters, 95% (119/125) of oncologists agreed they were accurate, comprehensive, and relevant, with no safety concerns noted. These letters demonstrated precise documentation of history, investigations, and treatment plans and were logically and concisely structured. Patient-directed letters achieved a mean Flesch Reading Ease score of 73.3 (seventh-grade reading level) and a PEMAT-P score above 80%, indicating high understandability. Consumer reviewers found them clear and appropriate for patient communication. Some omissions of details (eg, side effects), stylistic inconsistencies, and repetitive phrasing were identified, although no clinical safety issues emerged. Seventy-two percent (90/125) of consumers expressed willingness to receive artificial intelligence (AI)-generated patient letters.</p><p><strong>Conclusion: </strong>ChatGPT, when guided by structured prompts, can generate high-quality letters that align with clinical and patient communication standards. No clinical safety concerns were identified, although addressing occasional omissions and improving natural language flow could enhance their utility in practice. Further studies comparing AI-generated and human-written letters are recommended.</p>","PeriodicalId":51626,"journal":{"name":"JCO Clinical Cancer Informatics","volume":"9 ","pages":"e2400323"},"PeriodicalIF":3.3,"publicationDate":"2025-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144032285","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Prognostic Significance of the BALAD Serological Model in Systemic Therapies for Hepatocellular Carcinoma: A Personalized Approach to the Prediction of Survival Benefit.","authors":"Hidenori Toyoda, Atsushi Hiraoka, Hironori Ochi, Kunihiko Tsuji, Kazuto Tajiri, Joji Tani, Toshifumi Tada, Tomomi Okubo, Masanori Atsukawa, Masashi Hirooka, Ei Itobayashi, Takeshi Hatanaka, Kazuya Kariyama, Toru Ishikawa, Hidekatsu Kuroda, Koichi Takaguchi, Hisashi Kosaka, Kazuhito Kawada, Satoru Kakizaki, Yutaka Yada, Chikara Ogawa, Takashi Nishimura, Satoshi Yasuda, Akihiro Deguchi, Asahiro Morishita, Norio Itokawa, Taeang Arai, Akemi Tsutsui, Atsushi Naganuma, Hirayuki Enomoto, Masaki Kaibori, Kazuhiro Nouso, Yoichi Hiasa, Takashi Kumada, Tomoyuki Akita, Junko Tanaka, Philip J Johnson","doi":"10.1200/CCI-24-00175","DOIUrl":"10.1200/CCI-24-00175","url":null,"abstract":"<p><strong>Purpose: </strong>The BALAD model, a scoring system for staging hepatocellular carcinoma (HCC), is based on five serum markers: bilirubin, albumin, <i>lens culinaris</i> agglutinin-reactive alpha-fetoprotein [AFP], AFP, and des-gamma-carboxy prothrombin. It has shown good ability to predict survival in patients with HCC irrespective of stage and treatment, a high BALAD value being associated with a poor prognosis. However, its prognostic significance is unclear in patients with advanced unresectable HCC (uHCC) who undergo systemic therapies. We assessed the prognostic ability of BALAD in this subpopulation.</p><p><strong>Methods: </strong>In a multicenter cohort of 1,510 patients with advanced uHCC treated with first-line systemic therapies, the baseline BALAD score was calculated on the basis of pretreatment serum levels. Overall survival (OS), progression-free survival (PFS), overall response rate (ORR), and disease control rate (DCR) were calculated and related to the BALAD score.</p><p><strong>Results: </strong>In all, 502 patients were treated with sorafenib, 435 with lenvatinib, and 573 with atezolizumab plus bevacizumab. Irrespective of treatment regimen, OS, PFS, ORR, and DCR were all independently negatively correlated with the BALAD score. The beneficial effects of specific systemic therapy regimens differed according to the BALAD score.</p><p><strong>Conclusion: </strong>The BALAD score had good prognostic ability for predicting OS and PFS in patients with advanced uHCC who underwent systemic therapies and was associated with treatment response. Application of the BALAD score offers increased precision in the prediction of outcome both for individual patients and for specific subgroups of patients with HCC.</p>","PeriodicalId":51626,"journal":{"name":"JCO Clinical Cancer Informatics","volume":"9 ","pages":"e2400175"},"PeriodicalIF":3.3,"publicationDate":"2025-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144008002","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Use of Large Language Models in Clinical Cancer Research.","authors":"Kenneth L Kehl","doi":"10.1200/CCI-25-00027","DOIUrl":"https://doi.org/10.1200/CCI-25-00027","url":null,"abstract":"<p><p>Artificial intelligence (AI) is increasingly being applied to clinical cancer research, driving precision oncology objectives by gathering clinical data at scales that were not previously possible. Although small, domain-specific models have been used toward this end for several years, general-purpose large language models (LLMs) now enable scalable data extraction and analysis without the need for large, labeled training data sets. These models support several applications, including building clinico-omic databases, matching patients to clinical trials, and developing multimodal foundation models that integrate text, imaging, and molecular data. LLMs can also streamline research workflows, from automating documentation to accelerating clinical decision making. However, data privacy, hallucination risks, computational costs, regulatory requirements, and validation standards remain significant considerations. Careful implementation of AI tools will therefore be an important task for cancer researchers in coming years.</p>","PeriodicalId":51626,"journal":{"name":"JCO Clinical Cancer Informatics","volume":"9 ","pages":"e2500027"},"PeriodicalIF":3.3,"publicationDate":"2025-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144103094","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development and Validation of an ¹⁸F-Fluorodeoxyglucose Positron Emission Tomography-Computed Tomography-Based Imaging Score to Predict 12-Week Life Expectancy in Advanced Chemorefractory Colorectal Cancer.","authors":"Zelda Paquier, Jennifer Dhont, Thomas Guiot, Hugo Levillain, Gabriela Critchi, Rita Saude Conde, Francesco Sclafani, Patrick Flamen, Nick Reynaert, Erwin Woff, Alain Hendlisz","doi":"10.1200/CCI-24-00207","DOIUrl":"https://doi.org/10.1200/CCI-24-00207","url":null,"abstract":"<p><strong>Purpose: </strong>Managing chemorefractory metastatic colorectal cancer (mCRC) requires a meticulous equilibrium between the efficacy and toxicity of interventions, a task compounded by the constrained life expectancy of the patient. While existing prognostic tools, such as the Colon Life nomogram, primarily focus on general patient conditions or a single diagnostic modality, they do not fully integrate the potential predictive value of multimodal data. This study aims to develop and validate an Imaging Score, integrating clinical and imaging features derived from whole-body ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) positron emission tomography-computed tomography (PET-CT), predicting death probability within 12 weeks from treatment initiation for refractory disease.</p><p><strong>Materials and methods: </strong>The development cohort comprises 254 patients from three clinical trials. Nine clinical variables and six imaging variables were assessed. After optimal subset selection through recursive Feature Elimination with cross-validation, a support vector classifier-trained machine learning model generated the Imaging Score. Validation was performed on a real-life patient cohort (n = 74). Model performance was assessed on discrimination (Harrell C-index) and calibration.</p><p><strong>Results: </strong>Final prognostic features included whole-body metabolically active tumor volume, Eastern Cooperative Oncology Group performance status, visceral fat density, number of metastatic sites, body mass index, maximum standardized distance, and months since diagnosis. The Imaging Score demonstrated robust discriminative ability in both the development (C-index, 0.797) and validation (C-index, 0.714) sets, outperforming the Colon Life nomogram that tended to overestimate 12-week mortality.</p><p><strong>Conclusion: </strong>The Imaging Score, integrating ¹⁸F-FDG PET-CT imaging with clinical parameters, is an effective prognostic tool for patients with chemorefractory mCRC. This combination of imaging biomarkers with clinical factors improves discrimination, enhancing its potential for clinical decision making, patient stratification for chemorefractory treatments, and trial eligibility.</p>","PeriodicalId":51626,"journal":{"name":"JCO Clinical Cancer Informatics","volume":"9 ","pages":"e2400207"},"PeriodicalIF":3.3,"publicationDate":"2025-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12058364/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144053303","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Impact of Providing an Automated Telephone Option to Report Weekly Patient-Reported Outcome Measures in the PRO-TECT Trial (AFT-39) on Disparity Gaps in Symptom Management and Outcomes.","authors":"Angela M Stover, Allison M Deal, Brenda Ginos, Amylou Dueck, Patricia A Spears, Jennifer Jansen, Philip Carr, Sydney Henson, Antonia V Bennett, Mattias Jonsson, Claire Snyder, Ethan Basch","doi":"10.1200/CCI-25-00046","DOIUrl":"https://doi.org/10.1200/CCI-25-00046","url":null,"abstract":"<p><strong>Purpose: </strong>Many trials ask patients to complete patient-reported outcome measures (PROMs) via the web, excluding patients unable to use/access the Internet. The PRO-TECT trial (AFT-39, ClinicalTrials.gov identifier: NCT03249090) also offered a telephone interface option (interactive voice response [IVR]). We compared patients choosing IVR versus web on alert rates to nurses and clinical outcomes to determine if a telephone option can close disparities in symptom management.</p><p><strong>Methods: </strong>PRO-TECT randomized 26 community oncology practices to the PROM intervention arm where concerning symptoms generated automated alerts to nurses. IVR and web patients were compared for social determinants of health (SDOH) using analysis of variance and chi-square tests. After accounting for clustering and confounders, we used generalized estimating equations to compare alert rates, mixed models for quality of life (QOL) at 3 months, and Cox regression for emergency visits and survival at 12 months.</p><p><strong>Results: </strong>Among 593 patients, 215 (36%) chose IVR and 378 (64%) chose web. IVR patients were older (65.2 <i>v</i> 60.8 years) and were more often rural residents (32% <i>v</i> 23%), Black (27% <i>v</i> 11%), and with less education (54% <i>v</i> 27% ≤high school; all <i>P</i> < .01). Patients choosing IVR had more surveys with concerning symptoms (49% <i>v</i> 37%) and nurses felt clinical attention was warranted more often (4.8 surveys <i>v</i> 3.4 surveys; all <i>P</i> < .001) but ultimately experienced similar benefits as web in QOL, emergency visits, and survival.</p><p><strong>Conclusion: </strong>One third of community patients choose a telephone option over the web for reporting PROMs during cancer care. These patients are disproportionately from SDOH backgrounds at risk of poor clinical outcomes and have higher symptom management needs but ultimately experience similar clinical benefits as patients choosing the web. PROM programs should offer web alternatives to close disparities in symptom management.</p>","PeriodicalId":51626,"journal":{"name":"JCO Clinical Cancer Informatics","volume":"9 ","pages":"e2500046"},"PeriodicalIF":3.3,"publicationDate":"2025-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144132045","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Assessing Algorithmic Fairness With a Multimodal Artificial Intelligence Model in Men of African and Non-African Origin on NRG Oncology Prostate Cancer Phase III Trials.","authors":"Mack Roach, Jingbin Zhang, Osama Mohamad, Douwe van der Wal, Jeffry P Simko, Sandy DeVries, Huei-Chung Huang, Songwan Joun, Edward M Schaeffer, Todd M Morgan, Jessica Keim-Malpass, Emmalyn Chen, Rikiya Yamashita, Jedidiah M Monson, Farah Naz, James Wallace, Jean-Paul Bahary, Derek Wilke, Sonny Batra, Gregory B Biedermann, Sergio Faria, Lindsay Hwang, Howard M Sandler, Daniel E Spratt, Stephanie L Pugh, Andre Esteva, Phuoc T Tran, Felix Y Feng","doi":"10.1200/CCI-24-00284","DOIUrl":"https://doi.org/10.1200/CCI-24-00284","url":null,"abstract":"<p><strong>Purpose: </strong>Artificial intelligence (AI) tools could improve clinical decision making or exacerbate inequities because of bias. African American (AA) men reportedly have a worse prognosis for prostate cancer (PCa) and are underrepresented in the development genomic biomarkers. We assess the generalizability of tools developed using a multimodal AI (MMAI) deep learning system using digital histopathology and clinical data from NRG/Radiation Therapy Oncology Group PCa trials across racial subgroups.</p><p><strong>Methods: </strong>In total, 5,708 patients from five randomized phase III trials were included. Two MMAI algorithms were evaluated: (1) the distant metastasis (DM) MMAI model optimized to predict risk of DM, and (2) the PCa-specific mortality (PCSM) MMAI model optimized to focus on prediction death in the presence of DM (DDM). The prognostic performance of the MMAI algorithms was evaluated in AA and non-AA subgroups using time to DM (primary end point) and time to DDM (secondary end point). Exploratory end points included time to biochemical failure and overall survival with Fine-Gray or Cox proportional hazards models. Cumulative incidence estimates were computed for time-to-event end points and compared using Gray's test.</p><p><strong>Results: </strong>There were 948 (16.6%) AA patients, 4,731 non-AA patients (82.9%), and 29 (0.5%) patients with unknown or missing race status. The DM-MMAI algorithm showed a strong prognostic signal for DM in the AA (subdistribution hazard ratio [sHR], 1.2 [95% CI, 1.0 to 1.3]; <i>P</i> = .007) and non-AA subgroups (sHR, 1.4 [95% CI, 1.3 to 1.5]; <i>P</i> < .001). Similarly, the PCSM-MMAI score showed a strong prognostic signal for DDM in both AA (sHR, 1.3 [95% CI, 1.1 to 1.5]; <i>P</i> = .001) and non-AA subgroups (sHR, 1.5 [95% CI, 1.4 to 1.6]; <i>P</i> < .001), with similar distributions of risk.</p><p><strong>Conclusion: </strong>Using cooperative group data sets with a racially diverse population, the MMAI algorithm performed well across racial subgroups without evidence of algorithmic bias.</p>","PeriodicalId":51626,"journal":{"name":"JCO Clinical Cancer Informatics","volume":"9 ","pages":"e2400284"},"PeriodicalIF":3.3,"publicationDate":"2025-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144065285","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}