Background: State-of-the art behavioral interventions can help people lose 8-10% of body weight over a 6 month period, typically the nadir for weight loss efforts. Unfortunately, 40- 60% of participating adults do not achieve this clinically significant weight loss goal. If early response to treatment accurately predicts successful weight loss, identifying early nonresponders and either augmenting or switching treatment approaches could be a potentially cost-effective approach to enhancing overall success rates. However, clinically useful definitions for early response/non-response have not been established. Purpose: To systematically compare alternative definitions for identifying early responders/non-responders to standard behavioral weight loss intervention based on weight loss during treatment, rate of weight loss, and program attendance, as well as the timing at which early responders/non-responders are identified. Methods:Weekly weight data collected during intervention sessions and at baseline and 6-month research assessments from four behavioral weight loss trials (n=492) were pooled. Receiver operating characteristic (ROC) curves were used to compare alternate definitions for early response/non-response. The accuracy in predicting successful weight loss at month 6 (10% weight loss) for eachwas quantified using the area under theROCcurve (AUC). Results: Early response/non-response definitions based on cumulative weight loss to date, either in pounds or as percent of starting weight lost, more accurately predict successful weight loss than rules based on rate of loss (1, 1.5, 2 lbs/week) or program attendance. ByWeek 3, pounds (AUC=.72) and percent lost (AUC=.73) accurately predicted success, and better than rate- (AUCs=.60-.66) or attendance-based rules (AUCs=.52-.53). We will demonstrate how to use model-predicted likelihoods (e.g., participants losing less than 3% of body weight only have a 43% success likelihood) to refine a non-response definition (e.g., 3% weight loss by 3 weeks) to meet specific objectives such as a high non-responder rate. Conclusions: Simpler definitions for early response/non-response based on amount or percent of weight loss during treatment predict successful weight loss at month 6 at least as accurately as more complicated definitions.