School of Public Health, University of Alberta, Edmonton, Alberta, Canada.
Department of Pediatrics, Vanderbilt University Medical Center, Nashville, Tennessee.
Cancer Res. 2022 Aug 16;82(16):2940-2950. doi: 10.1158/0008-5472.CAN-22-0418.
Therapy-related pulmonary complications are among the leading causes of morbidity among long-term survivors of childhood cancer. Restrictive ventilatory defects (RVD) are prevalent, with risks increasing after exposures to chest radiotherapy and radiomimetic chemotherapies. Using whole-genome sequencing data from 1,728 childhood cancer survivors in the St. Jude Lifetime Cohort Study, we developed and validated a composite RVD risk prediction model that integrates clinical profiles and polygenic risk scores (PRS), including both published lung phenotype PRSs and a novel survivor-specific pharmaco/radiogenomic PRS (surPRS) for RVD risk reflecting gene-by-treatment (GxT) interaction effects. Overall, this new therapy-specific polygenic risk prediction model showed multiple indicators for superior discriminatory accuracy in an independent data set. The surPRS was significantly associated with RVD risk in both training (OR = 1.60, P = 3.7 × 10-10) and validation (OR = 1.44, P = 8.5 × 10-4) data sets. The composite model featuring the surPRS showed the best discriminatory accuracy (AUC = 0.81; 95% CI, 0.76-0.87), a significant improvement (P = 9.0 × 10-3) over clinical risk scores only (AUC = 0.78; 95% CI: 0.72-0.83). The odds of RVD in survivors in the highest quintile of composite model-predicted risk was ∼20-fold higher than those with median predicted risk or less (OR = 20.01, P = 2.2 × 10-16), exceeding the comparable estimate considering nongenetic risk factors only (OR = 9.20, P = 7.4 × 10-11). Inclusion of genetic predictors also selectively improved risk stratification for pulmonary complications across at-risk primary cancer diagnoses (AUCclinical = 0.72; AUCcomposite = 0.80, P = 0.012). Overall, this PRS approach that leverages GxT interaction effects supports late effects risk prediction among childhood cancer survivors.
This study develops a therapy-specific polygenic risk prediction model to more precisely identify childhood cancer survivors at high risk for pulmonary complications, which could help improve risk stratification for other late effects.
治疗相关的肺部并发症是儿童癌症长期幸存者发病的主要原因之一。限制性通气缺陷（RVD）很常见，胸部放射治疗和放射模拟化疗后风险增加。利用圣裘德终身队列研究中的 1728 名儿童癌症幸存者的全基因组测序数据，我们开发并验证了一种综合的 RVD 风险预测模型，该模型将临床特征和多基因风险评分（PRS）结合在一起，包括已发表的肺表型 PRS 和一种新的针对 RVD 风险的、反映基因-治疗相互作用（GxT）的幸存者特异性药物/放射基因组 PRS（surPRS）。总的来说，这种新的针对治疗的多基因风险预测模型在独立数据集的多个指标中显示出了优越的区分准确性。surPRS 在训练（OR = 1.60，P = 3.7×10-10）和验证（OR = 1.44，P = 8.5×10-4）数据集均与 RVD 风险显著相关。具有 surPRS 的综合模型显示出最佳的区分准确性（AUC = 0.81；95%CI，0.76-0.87），与仅基于临床风险评分的模型（AUC = 0.78；95%CI：0.72-0.83）相比有显著提高（P = 9.0×10-3）。在预测风险最高五分位数的幸存者中，RVD 的可能性是预测风险中位数或更低的幸存者的约 20 倍（OR = 20.01，P = 2.2×10-16），超过了仅考虑非遗传风险因素的可比估计值（OR = 9.20，P = 7.4×10-11）。纳入遗传预测因子还可以选择性地改善高危原发性癌症诊断的肺部并发症风险分层（AUCclinical = 0.72；AUCcomposite = 0.80，P = 0.012）。总的来说，这种利用 GxT 相互作用的 PRS 方法支持儿童癌症幸存者的晚期效应风险预测。
本研究开发了一种针对治疗的多基因风险预测模型，以更精确地识别出患肺部并发症风险高的儿童癌症幸存者，这有助于改善其他晚期效应的风险分层。
