Patients, Settings, and Study Procedures
This real world longitudinal study included consecutive esophageal cancer patients who underwent esophagectomy at Sichuan Cancer Hospital between April 2019 and March 2020. The study was approved by the Medical Ethics Committee and Clinical Trial Review Committee of the Sichuan Cancer Hospital (SCCHEC-02-2020-036). In this department, symptom severity was assessed for all patients undergoing esophagectomy as a clinical routine, so the informed consent was waived. The study was registered on ChiCTR.org.cn Web site (ChiCTR2000040780). For the current analysis, patients were excluded if they (1) had postoperative pathological non-squamous carcinoma; (2) underwent palliative esophagectomy; (3) had esophageal and gastric junction carcinoma; or (4) ≥4 times symptoms data missing.
Data Collection
Demographic and Clinical Characteristics
Demographic characteristics and clinical information were collected from medical records. Demographic characteristics included age, sex, and marital status; clinical information included cancer location, neoadjuvant therapy, type of surgery, TNM classification, Eastern Cooperative Oncology Group performance status (ECOG PS), Patient-Generated Subjective Global Assessment (PG-SGA) scores, postoperative hospital stay, and comorbid conditions.
Symptom Data
Symptom data were obtained from the Sichuan Cancer Hospital’s Case Management Registration Database based on REDCap (http://125.71.214.100:888/
redcap). Symptom data were collected at pre-surgery (baseline) and on days 1, 3, 5, 7, 14, 21, 30, and 90 after surgery. Symptoms were assessed using the Chinese version of the MD Anderson Symptom Inventory (MDASI-C). MDASI-C is a brief measure of 13 common cancer-related symptoms, and each symptom was rated on an 11-point scale, with 0 being ‘not present’ and 10 being ‘as bad as you can imagine’[22] .We defined mild symptom as 1-3, moderate symptom as 4-6, and severe symptom as 7-10[22]. MDASI-C is well-established validity and reliability in cancer patients[23], and can be used for postoperative high-frequency symptom collection by assessing the previous 24 hours symptoms[24].
Statistical Analyses
Data are summarised as means (±standard deviation [SD]), medians (quantiles), or frequencies (percentages), as appropriate. We used the percentage of the moderate-severe level (score ≥4) on day 1 after surgery to identify the 5 most-severe symptoms and profile symptom recovery from surgery.
Mixed-effect models with the maximum likelihood method were used to compare symptom scores at each postoperative time point with preoperative levels of the symptoms. Because the random effect of timepoints was estimated to be zero, the random effect of intercepts was included in all models. The fixed effects of all independent variables were age, sex, marital status, ECOG PS, neoadjuvant therapy, surgery type, comorbid conditions, PG-SGA scores, cancer location, and the interaction between time and length of hospitalisation. In the models, we considered time as a categorical variable. Variance structure types, such as unstructured, simple, first-order auto-regressive, and compound symmetric, were compared via Bayesian information criterion (BIC). The model with least BIC was preferred.
We defined ‘postoperative recovery’ as symptom recovery to the mild level; i.e., after surgery, the patient-reported MDAS-C symptom scores ≤3 at one measurement. Median and mean recovery days and 95% confidence intervals (CIs) were estimated using Kaplan–Meier analysis.
All tests were two sided, with 5% as the significance level. All statistical analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA).