Zhou Yiwang, Deshpande Samira, Horan Madeline R, Choi Jaesung, Mulrooney Daniel A, Ness Kirsten K, Hudson Melissa M, Srivastava Deo Kumar, Huang I-Chan
Department of Biostatistics, St. Jude Children's Research Hospital, Memphis, TN, USA.
Department of Pediatrics, School of Medicine, Wake Forest University, Winston-Salem, NC, USA.
Commun Med (Lond). 2025 Aug 29;5(1):377. doi: 10.1038/s43856-025-01105-0.
Childhood cancer survivors experience persistent and evolving symptom burden post-therapy. Network analysis can help uncover the complex symptom patterns. However, current network analyses often rely on cross-sectional data and focus on average symptom patterns among survivors, overlooking individual heterogeneities.
We introduced an autoregressive logistic model with covariates to account for individual heterogeneities in network estimation and to construct personal temporal symptom networks. Simulation experiments were conducted to validate the robustness of this method in constructing personal temporal symptom networks. We also applied the autoregressive logistic model with covariates to longitudinal symptom data from a random sample of 2000 adult survivors of childhood cancer in the St. Jude Lifetime Cohort Study (SJLIFE).
Simulation studies demonstrate that the proposed method reliably recovers personal temporal symptom network structures under various conditions. In the real data application, older age, female sex, lower educational attainment, annual personal income <$20,000, and receipt of chemotherapy and/or radiation therapy are associated with stronger connections between symptoms at baseline and the first follow-up.
We demonstrate that the logistic autoregressive model with covariates effectively estimates personal temporal symptom networks for childhood cancer survivors, enabling personalized symptom monitoring and informing tailored symptom management strategies.
儿童癌症幸存者在治疗后经历着持续且不断演变的症状负担。网络分析有助于揭示复杂的症状模式。然而，当前的网络分析通常依赖横断面数据，且侧重于幸存者中的平均症状模式，忽略了个体异质性。
我们引入了一种带有协变量的自回归逻辑模型，以在网络估计中考虑个体异质性，并构建个人时间症状网络。进行了模拟实验，以验证该方法在构建个人时间症状网络方面的稳健性。我们还将带有协变量的自回归逻辑模型应用于圣裘德终身队列研究（SJLIFE）中2000名儿童癌症成年幸存者随机样本的纵向症状数据。
模拟研究表明，所提出的方法在各种条件下都能可靠地恢复个人时间症状网络结构。在实际数据应用中，年龄较大、女性、教育程度较低、个人年收入<$20,000以及接受化疗和/或放疗与基线和首次随访时症状之间更强的关联有关。
我们证明，带有协变量的逻辑自回归模型能有效地估计儿童癌症幸存者的个人时间症状网络，实现个性化症状监测，并为量身定制的症状管理策略提供依据。
