Abstract Background Lateral epicondylitis or tennis elbow is a frequent condition with long-lasting symptoms. In order to identify predictors for treatment success and pain in lateral epicondylitis, we used data from a randomized controlled trial. This trial investigated the efficacy of physiotherapy alone or combined with corticosteroid injection for acute lateral epicondylitis in general practice. Methods The outcomes treatment success and pain score on VAS were assessed at 6, 12, 26 and 52 weeks. We ran a univariate binary logistic regression with generalized estimating equations (GEE) and subsequently an adjusted multilevel logistic regression to analyze the association between potential prognostic indicators and the outcome success/ no success. To assess the changes in pain score we used a two-level multilevel linear regression (MLR) followed by an adjusted MLR model with random effects. Results The most consistent predictor for reduced treatment success at all time points was a high Pain Free Function Index score signifying more pain on everyday activities. Being on paid sick-leave and having a recurring complaint increased short term treatment success but gave decreased long-term treatment success. The patients reporting symptoms after engaging in probable overuse in an unusual activity, tended towards increased treatment success at all time-points, but significant only at 12 weeks. The most consistent predictor of increased pain at all time points was a higher overall complaints score at baseline. Conclusions: Our results suggest that in treating acute lateral epicondylitis, a consideration of baseline pain, a registration of the patient’s overall complaint on a VAS scale and an assessment of the patient’s perceived performance in everyday activities with the Pain Free Function Index can be useful in identifying patients that will have a more protracted and serious condition. Trial registration ClinicalTrials.gov Identifier: NCT00826462. Date of registration January 22, 2009. The Trial was prospectively registrated.