Can Physical Activities Reduce Postoperative Pain in Adults

METHODS:

This is a single center, randomized prospective control trial (PA patients group vs control non-PA patients group). Information on the incidence and severity of pain and other sufferings of each patient were periodically recorded at time-points of 1, 4, 7, 10 and 24 hours after various types of surgical operations. Our study was aimed to determine beneficial effects on postoperative pain for patients after performing preoperative PA for 6 to 8 weeks vs non-PA patients. The investigators employed the recommendations of the American College of Sport Medicine and the World Health Organization for adults to divide our PA patients group into moderate-intensity as 30-60 min∙d-1 (≥150 min∙wk-1 ) and vigorous-intensity as 20-60 min∙d-1 (≥75 min∙wk-1). The severity of postoperative pain will be measured prospectively at 1, 4, 7, 10 and 24 hours after the surgical operations for the PA patients group and the control non-PA patients group. The operations were performed under general anesthesia (GA) with endotracheal intubation or inhalation through laryngeal mask.

The procedures of GA will be discussed and decided by one of our anesthesiologists of the Chia-Yi Christian Hospital together with patients/caregivers at the Pre-Anesthesia Consultation Clinic. The investigators used the American Society of Anesthesiologists physical status scoring system for risk stratification, the approaches of Apfel's preventive strategy of postoperative nausea and vomiting (PONV) prophylaxis, perioperative multi-modal pain management in addition to other appropriate elements in ERAS. PONV defined as nausea, vomiting or retching within 24 h of surgery.

Statistical Analysis:

Data will be analyzed using SPSS Version 22.0 and Mplus Version 8.2. Data on demographics, pain scores, PONV frequency and severity or sufferings were analyzed using descriptive statistics and frequency distributions.

Latent growth models are constructed through structural equation modeling using maximum likelihood estimations. The investigators will examine three unconditional growth models: (1) a linear model with the slope factor fixed at 0,1, 2, 3, and 4, (2) a quadratic model with the linear slope factor fixed at 0, 1, 2, 3, and 4, and the quadratic slope factor fixed at 0, 1, 4, 9, 16, and (3) an unspecified curve model with the slope factor loadings at the following settings: (a) wave 1 set to 0; (b) wave 2 set to 1; (c) wave 3, wave 4, and wave 5 allowed to be estimated freely.21 After choosing the model that had best fitted our data, the investigators introduce the predictive variable of PA, and the 9 time-invariant covariates and 15 time-varying covariates. The 9 time-invariant covariates are control variables of intercept and slope factors. The 15 time-varying covariates are repeated exogenous predictors of the pain scores. The pain scores in each waves are assigned three covariates. The investigators run the model including the predictor, 9 time-invariant variables and 15 time-variant covariates and all variables are allowed to correlate with one another.

The performance of model fitting are evaluated using goodness of fit indices, including the following: χ2 statistic, root mean square error of approximation (RMSEA), comparative fit index (CFI), standardized root mean square residual (SRMR), Akaike information criterion (AIC), and Bayesian information criterion (BIC). An RMSEA of 0.0 indicates a model with an exact fit and the investigators will adopte RMSEA of ≤0.08 and preferably ≤0.05 to indicate good fitting of models. CFI has a range from 0 to 1.00, the investigators will adopte CFI ≥0.90 or preferably ≥0.95 to indicate good fitting models. SRMR ≤0.05 indicated good-fitting model and values as high as 0.08 are considered acceptable. Although the investigators will perform the chi-square test in our results of model fit, its known sensitivity to sample size necessarily excludes its use as proper measure of goodness of fit.

AIC and BIC are useful for model selection, but do not provide absolute measurement of model fitting. The best-fitted model are those with smallest AIC and BIC values.