Although the Roland disability questionnaire (RDQ) is a commonly used questionnaire for patients with low back pain (LBP), several important issues remain understudied. The purposes of this study were to investigate the impact that several parameters (length of test-retest interval, interpretation of the global perceived effect (GPE) scores, methods to estimate standard error of measurement (SEM), intervention and baseline scores) have on various measurement properties, such as agreement (Limits of Agreement (LOA), SEMagreement), responsiveness (area under the ROC curve, sensitivity, specificity) and interpretability (optimal cut-off point, Minimal Detectable Change), of the RDQ. RDQ was administered four times to 212 patients with chronic non-specific LBP (first consultation (T-1), eligibility check (T0), randomization/start of treatment (T1) and end of treatment (T2)). They also completed a GPE scale at T2. To study the influence of length of test interval between T-1, T0 and T1, six time intervals were assessed. Analyses of the impact of parameters on responsiveness and interpretability were based on the GPE score and on RDQ score changes (T1 versus T2). Results revealed that agreement parameters decreased with increasing time interval between test-retest. They confirmed the influence of the interpretation of the GPE as well as of the method to calculate the SEM on the magnitude of the Minimal Detectable Change and ROC parameters. Baseline RDQ scores and the way to cluster patients with regard to baseline scores influenced mainly the optimal cut-off point and responsiveness parameters; intervention did not affect the magnitude of the Minimal Detectable Change.