A Bivariate Mixed Model Approach in Characterizing the Evolution of Longitudinal Body Mass Index and Quality of Life Processes in Adolescents with Severe Obesity Following Bariatric Surgery: A 5-year follow-up of the TeenLABS cohort.

Obesity is identified as a major global health problem. Along with measuring body mass index (BMI), the most common metric for defining weight status, health related quality of life (HRQol) has been accepted as a routine method to evaluate how body weight may be impacted by psychosocial factors. The objective of the current study is to characterize the joint association of change in longitudinal BMI and HRQol following metabolic and bariatric surgery and to examine the correlation between these two outcomes measured concurrently over time. We identified the optimal modeling strategy by comparing four models, all of which involved the covariance structures appropriate for correlated outcomes, BMI and HRQol in a repeated measures analysis. The bivariate random effects models performed better than the univariate random effects models. Moreover, bivariate models with composite covariate structures had better model fit compared to the bivariate random slope models. The bivariate models with composite covariate structures reflected that changes in HRQol (and BMI) were most significant during the first 6 months, a clinically useful window to monitor changes in post-operative HRQol and BMI, and if there might need to be additional interventions or at least, closer monitoring.