Personal symptom ecosystem predicts progression of chronic health conditions (CHCs) in adult survivors of childhood cancer: A report from the St. Jude Lifetime Cohort study (SJLIFE).

person Yiwang Zhou St. Jude Children's Research Hospital, Memphis, TN info_outline Yiwang Zhou, Madeline Horan, Jaesung Choi, Rachel T Webster, Tara M. Brinkman, Daniel A. Mulrooney, Deo Kumar Srivastava, Gregory T. Armstrong, Kirsten K. Ness, Melissa M. Hudson, I-Chan Huang

Authors person Yiwang Zhou St. Jude Children's Research Hospital, Memphis, TN info_outline Yiwang Zhou, Madeline Horan, Jaesung Choi, Rachel T Webster, Tara M. Brinkman, Daniel A. Mulrooney, Deo Kumar Srivastava, Gregory T. Armstrong, Kirsten K. Ness, Melissa M. Hudson, I-Chan Huang Organizations St. Jude Children's Research Hospital, Memphis, TN Abstract Disclosures Research Funding National Cancer Institute Background: Survivors of childhood cancer experience a range of interconnected symptoms, forming a personal symptom ecosystem with unique structures and dynamics. Unlike prior research analyzing the main effect of each symptom, we created personal symptom networks to predict the onset/worsening of CHCs for cancer survivors. Methods: We analyzed data collected from 2007-2020 among 4044 adult survivors of childhood cancer enrolled in SJLIFE. At baseline and follow-up (FU), individual CHCs were clinically assessed, severity-graded (CTCAE), and classified into organ-based groups: cardiac, pulmonary, musculoskeletal and neurological. Progression of CHCs from baseline to FU was defined as onset (grade 0-1 to grade 2-4) or worsening (grade 2 to 3-4; grade 3 to 4). Baseline symptoms in 9 domains (cardiac, pulmonary, sensation, nausea, movement, pain, fatigue, anxiety, depression) and personal factors (age, sex, education, smoking) were self-reported; neighborhood adversity was assessed by the Social Vulnerability Index (SVI); treatment data were sourced from medical records. The Ising model incorporating personal/treatment/SVI factors as covariates was used to develop personal symptom ecosystems. The effect of symptoms on the CHC progression included both the influence of a symptom domain (mean main effect) and its interaction with other symptom domains (ecosystem effect). These effects were examined using LASSO regularized logistic regressions on the progression of CHC groups. Results: The mean age of survivors at baseline was 30.3±8.6 years, 52.2% were male; the mean years from baseline to FU were 4.3±1.7. Survivors with lower education, smoking, and living in poorer neighborhoods reported interconnected pain-depression (effect size [ES] 0.35), anxiety-fatigue (ES 0.57), and cardiac-fatigue (ES 0.35) symptom domains. Symptom domains significantly impacting the progression of CHC groups were cardiac symptoms on the pulmonary (OR 1.2) and neurological CHC groups (OR 1.5); movement symptoms on the musculoskeletal (OR 1.4) and neurological CHC groups (OR 1.7); pain (OR 2.4) and fatigue (OR 1.8) on the neurological CHC group. Significant ecosystem effects of symptom domains impacting the progression of CHC groups were pain on the cardiac CHC group (OR 1.2); pulmonary symptoms on the pulmonary CHC group (OR 1.2); cardiac symptoms on the musculoskeletal CHC group (OR 1.4). The 95% confidence intervals were omitted since LASSO does not estimate standard deviation for variables identified as significant. Conclusions: The symptom network approach shows promising ecosystem effects of individual symptom domains beyond its main effects on CHC progression. This study improves our understanding of dynamic, interconnected symptoms underlying CHC progression, and highlights avenues for future intervention research.