Return to work (RTW) is the primary goal in the rehabilitation of patients with chronic low back pain. In spite of expensive rehabilitative efforts, many patients do not RTW. To increase cost effectiveness, predictive tests for non-RTW are needed to select patients for rehabilitation. The reliability of these tests must be high, to prevent exclusion of patients who might improve. This study evaluates the reliability and predictive validity of four tests and the following psychosocial factors for non-RTW: nationality, off-work duration, unemployment and work load. It was designed as a prospective cohort study of 99 patients with chronic low back pain. Upon entry, physical work load, time off work, unemployment and nationality were recorded. The study investigated four tests with an anticipated prognostic value for non-RTW: the Numeric Pain Rating Scale (NRS, 9-10 of a maximum of 10), the Step Test and Pseudo Strength Test (precipitous cessation) and Behavioural Signs. After 12 months, the RTW rate was obtained from the physicians responsible for sick-listing by postal survey. The response rate regarding RTW was 91% at 1 year. The RTW rate at 1 year was 20%. All investigated tests significantly correlated with non-RTW. Regression analysis showed that the best prediction of non-RTW was obtained when at least two out of the four tests were positive (positive predictive value 0.97, sensitivity 0.45). Unemployment, time off work, nationality and physical work load were less predictive. The results show that the combination of the four prognostic tests allows a very reliable prognosis of non-RTW. The cost effectiveness of rehabilitation aiming at RTW will, therefore, be increased by excluding patients with two or more positive tests.