Roach Mack, Zhang Jingbin, Mohamad Osama, van der Wal Douwe, Simko Jeffry P, DeVries Sandy, Huang Huei-Chung, Joun Songwan, Schaeffer Edward M, Morgan Todd M, Keim-Malpass Jessica, Chen Emmalyn, Yamashita Rikiya, Monson Jedidiah M, Naz Farah, Wallace James, Bahary Jean-Paul, Wilke Derek, Batra Sonny, Biedermann Gregory B, Faria Sergio, Hwang Lindsay, Sandler Howard M, Spratt Daniel E, Pugh Stephanie L, Esteva Andre, Tran Phuoc T, Feng Felix Y
UCSF Medical Center, San Francisco, CA.
Artera, Santa Barbara, CA.
JCO Clin Cancer Inform. 2025 May;9:e2400284. doi: 10.1200/CCI-24-00284. Epub 2025 May 9.
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.
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.
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]; = .007) and non-AA subgroups (sHR, 1.4 [95% CI, 1.3 to 1.5]; < .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]; = .001) and non-AA subgroups (sHR, 1.5 [95% CI, 1.4 to 1.6]; < .001), with similar distributions of risk.
Using cooperative group data sets with a racially diverse population, the MMAI algorithm performed well across racial subgroups without evidence of algorithmic bias.
人工智能（AI）工具可能改善临床决策，也可能因偏差而加剧不平等。据报道，非裔美国（AA）男性前列腺癌（PCa）的预后较差，且在基因组生物标志物的开发中代表性不足。我们使用多模态AI（MMAI）深度学习系统，结合NRG/放射治疗肿瘤学组PCa试验中的数字组织病理学和临床数据，评估所开发工具在不同种族亚组中的通用性。
总共纳入了来自五项随机III期试验的5708例患者。评估了两种MMAI算法：（1）优化用于预测远处转移（DM）风险的远处转移（DM）MMAI模型，以及（2）优化用于关注存在远处转移时死亡预测（DDM）的PCa特异性死亡率（PCSM）MMAI模型。使用至DM时间（主要终点）和至DDM时间（次要终点），在AA和非AA亚组中评估MMAI算法的预后性能。探索性终点包括生化失败时间和使用Fine-Gray或Cox比例风险模型的总生存期。计算事件发生时间终点的累积发病率估计值，并使用Gray检验进行比较。
有948例（16.6%）AA患者、4731例非AA患者（82.9%）以及29例（0.5%）种族状态未知或缺失的患者。DM-MMAI算法在AA亚组（亚分布风险比[sHR]，1.2[95%CI，1.0至1.3]；P = 0.007）和非AA亚组（sHR，1.4[95%CI，1.3至1.5]；P < 0.001）中均显示出强烈的DM预后信号。同样，PCSM-MMAI评分在AA亚组（sHR，1.3[95%CI，1.1至1.5]；P = 0.001）和非AA亚组（sHR，1.5[95%CI，1.4至1.6]；P < 0.001）中均显示出强烈的DDM预后信号，且风险分布相似。
使用具有不同种族人群的合作组数据集，MMAI算法在不同种族亚组中表现良好，没有算法偏差的证据。
