Page 58 - When surgery alone won’t cut it - Valerie Maureen Monpellier

Chapter 3
Statistical analyses
Descriptive statistics were used to summarize patients’ baseline and follow-up char- acteristics. Changes in body weight and scores of Baecke and DEBQ were analysed using repeated measures ANOVA. Quartiles for preoperative physical activity and eat- ing style were calculated and a MANCOVA analysis was conducted to assess the different weight loss trajectories. Creating weight loss trajectories in four groups rang- ing from low preoperative physical activity to high preoperative physical activity and four groups ranging from low preoperative emotional, external and restrained eating to high preoperative emotional, external and restrained eating. Plots were created to visualize the differences between these groups. All above analyses were performed using SPSS software, version 23 (IBM Corp. Released 2015. IBM SPSS Statistics for Windows, Version 23.0. Armonk, NY: IBM Corp.).
Regression analyses were used to investigate the relationship between baseline self-reported physical activity / eating style and weight change (TWLmax, 36mWR and 48mWR) and between physical activity and eating style at follow-up and weight regain at 36 and 48 months.
A linear mixed model was conducted to assess the association between physical activity and %TWL over all the follow-up moments. This model compared Baecke scoring at the specific follow-up moment with the baseline score, thus focussing on how Baecke scores changed. The relationship between Baecke scores and %TWL over all follow-up moments was assessed with random intercept, thereby the model takes into account different intercepts for each patient. Then baseline score, gender and age were added to the model as fixed effects. In the last part of the model effect modification of time on Baecke was assessed. First calculating if the interaction of Baecke and time had a significant impact on the model and then assessing on which follow-up moments this was. The same model was performed for each of the eating styles (separately). All assumptions for regression analysis were met. These analyses were performed using STATA, version 13 (StataCorp. 2013. Stata 13 Base Reference Manual. College Station, TX: Stata Press.).
Findings were considered statistically significant if the p-value was <0.05.
RESULTS
Study population
The patients included in the analysis (n=4,569) had a mean age of 47.1 years (±10.7), which was higher compared to those who were excluded from the analysis (mean age 44.3 (±11.2), p<0.001). More females (81.7%) were observed in the included popu- lation compared to the excluded population (69.2%, p<0.001). Mean follow-up (FU) since surgery was 38.7 months (range 24 – 59 months), which was shorter compared to the excluded population (42.3 months, p<0.001).
58