OBJECTIVES: The objective of this study was to utilize the BOLD database to analyze percent BMI loss after bariatric surgery.  Of particular interest was the determination of predictors to help explain the large variation in bariatric surgery success. METHODS: The dataset extracted for analysis consisted of patients age 21 or older having their first bariatric surgery (laparoscopic adjustable gastric band (LAGB), Roux-en-y gastric bypass (RYGB), or vertical sleeve gastrectomy (VSG)) between January 1, 2007 and February 26, 2010, with pre-surgery BMI of at least 30, a baseline visit and at least one postoperative visit.  Subpopulations with postoperative visits between 9-15, 15-21, and 21-27 months after surgery were identified for 12-month, 18-month, and 24-month endpoints.  All available data relating to the procedure, demographics, comorbidities, and prior surgical history were considered as potential predictors of %BMI loss.  Regression models, using multiple model selection procedures, were fitted at each endpoint. RESULTS: The population consisted of 31,443 LAGB, 40,352 RYGB, and 2194 VSG patients, of whom 79% were female, 79% were Caucasian, and the mean age was 46 years.  Of the total 73,989 patients, 26,920 had a 12-month endpoint, 7245 had an 18-month endpoint, and 1774 had a 24-month endpoint.  Regression models explained 37 to 55% of the variance in %BMI loss, depending primarily on the endpoint, with the highest percent variance explained at the 12-month endpoint and the lowest at the 24-month endpoint.  Model selection methods made little difference in model fit.  The type of bariatric surgery performed was the most significant predictor in all models.  Other significant predictors in various models included key demographic variables and comorbidities, as well as baseline BMI. CONCLUSIONS: The data in the BOLD registry are sufficiently robust to enable the evaluation of predictors of bariatric surgery success including surgery type, comorbidities, baseline BMI and key demographic variables.