Association Between Chronic Psychological Stress and Disease Course Outcomes in Pancreatic Cancer

Study Design and Population This is a multicenter, prospective, observational cohort study designed to investigate whether chronic psychological stress (anxiety, depression, and sleep disturbance) is independently associated with disease course outcomes in patients with pancreatic cancer. Study Procedures At baseline (day 0), participants complete validated self-report instruments: the Hospital Anxiety and Depression Scale (HADS), Patient Health Questionnaire-9 (PHQ-9), Generalized Anxiety Disorder-7 (GAD-7), and Pittsburgh Sleep Quality Index (PSQI). Clinical data are extracted from medical records (demographics, tumor stage, CA19-9, treatment history, comorbidities, etc.). A 2 mL remnant peripheral blood sample (collected during routine morning phlebotomy) is obtained for serum separation and storage at -80 °C; samples will be shipped to the central laboratory (Zhongshan Hospital) for batch analysis of biomarkers. Follow-up assessments occur at key clinical time points (neoadjuvant therapy mid-point, adjuvant therapy mid-point, end of adjuvant therapy, 12-month long-term follow-up, and at disease progression/recurrence). Psychological scales are repeated at each time point. Clinical outcomes (recurrence, progression, treatment modifications, adverse events, survival status) are captured every 3 months via medical records or telephone contact. Quality Assurance and Data Management Quality assurance plan: All sites receive standardized training on study procedures, scale administration, and GCP principles. Regular multicenter meetings and annual on-site monitoring/remote auditing by the coordinating center (Fudan University Zhongshan Hospital) are conducted. Data validation: A web-based electronic data capture (EDC) system is used with automated range and logic checks. Queries are generated for out-of-range or inconsistent entries. Source data verification (SDV): A random sample of 10% of participants will undergo SDV against original medical records. For key endpoints (OS, PFS, and psychological scores), 100% SDV will be performed. Data dictionary: A comprehensive data dictionary defines all variables, including coding dictionaries (MedDRA for adverse events, WHO Drug Dictionary for concomitant medications). Standard Operating Procedures (SOPs): SOPs govern recruitment, informed consent, psychological assessment, biospecimen handling, data entry, adverse event reporting, and protocol amendments. Sample Size Assessment The sample size of 320 patients was calculated based on the primary outcome (OS) using a two-sided alpha of 0.05, power of 80%, a hazard ratio (HR) of 1.5 (high vs. low psychological stress), an event rate of 67.5%, and 10% loss to follow-up. The Schoenfeld formula yields at least 192 death events and 285 completed patients; after accounting for dropouts, 317 patients are needed, rounded to 320. The HR of 1.5 is justified by recent literature and represents a clinically meaningful medium effect size while maintaining feasibility. Handling of Missing Data For OS, missing survival dates are minimized by active follow-up. For covariates and scale items with random missingness (\<15%), multiple imputation (MI) will be used, with complete-case analysis as sensitivity. Statistical Analysis Plan Primary analysis: Kaplan-Meier curves, log-rank tests, and multivariable Cox proportional hazards models (adjusting for study center, age, TNM stage, surgery, ECOG score, etc.) will be used to estimate the HR for OS. The proportional hazards assumption will be tested by Schoenfeld residuals. Secondary analyses: Similar Cox models will be applied to PFS. Longitudinal trajectories of psychological scores (HADS, PHQ-9, GAD-7, PSQI) will be modeled using generalized estimating equations (GEE). Subgroup and sensitivity analyses: Pre-specified subgroups (by TNM stage, surgery type, age, site) will be examined with interaction tests. Sensitivity analyses will use alternative definitions of psychological stress (e.g., continuous z-score) and different missing data methods. Bias assessment: E-values will be calculated for unmeasured confounding. Selection bias will be assessed by comparing completers vs. dropouts; worst-case scenario sensitivity analyses will be performed. Ethics and Dissemination The protocol (version V1.4, dated 20260302) has been approved by the ethics committees of all participating centers. Results will be published in peer-reviewed journals regardless of outcome direction.