University of Wollongong
Current issues in the analysis of weight loss trials
Batterham, Marijka1,Tapsell, Linda2, Charlton, Karen2
1 Statistical Consulting Centre, National Institute of Applied Statistics Research Australia, University of Wollongong
2 School of Medicine and Health, University of Wollongong
This presentation discusses analytical issues in research on the effectiveness of manipulating dietary composition to enhance weight loss. Our studies have demonstrated short term beneficial effects of dietary protein and polyunsaturated fats on fat metabolism compared with control diets. However, the effects of diet alone are not sustained in long term studies. Long term benefits have only been achieved with intensive interdisciplinary interventions. One of the main issues which prevails in our trials is attrition.
Attrition is common in weight loss studies and this may result in substantial missing data. We have previously demonstrated that frequently used methods of accounting for study attrition such as baseline observation carried forward, last observation carried forward and complete case analysis can provide inaccurate estimates in weight loss trials. Our current research suggests maximum likelihood based approaches represent the preferred approach when attrition rates are low and that multiple imputation should be implemented when attrition rates are high and effect sizes small.
Predicting attrition is important for future trial design. Our research demonstrates that initial weight loss (> 2% of initial weight) as well as a rapid and sustained weight loss using growth mixture modelling are associated with trial completion. Age may also predict attrition with older subjects less likely to drop out. We have also considered those successful in achieving the clinical target of 5% weight loss at trial completion and again early initial weight loss is a significant factor. Future weight loss trials should consider adaptive designs that can use this information, or provide targeted interventions for weight loss based on characteristics identified to predict attrition and weight loss success.