Background: Low back pain (LBP) is one of the most disabling and costly disorders affecting modern society, and approximately 90% of patients are labelled as having non-specific LBP (NSLBP). Several interventions for patients with NSLBP have been assessed in clinical trials, but heterogeneous reporting of outcomes in these trials has hindered comparison of results and performance of meta-analyses. Moreover, there is a risk of selective outcome reporting bias. To address these issues, the development of a core outcome set (COS) that should be measured in all clinical trials for a specific health condition has been recommended. A standardized set of outcomes for LBP was proposed in 1998, however, with evolution in COS development methodology, new instruments, interventions, and understanding of measurement properties, it is appropriate to update that proposal. This protocol describes the methods used in the initial step in developing a COS for NSLBP, namely, establishing a core domain set that should be measured in all clinical trials. Methods/Design: An International Steering Committee including researchers, clinicians, and patient representatives from four continents was formed to guide the development of this COS. The approach of initiatives like Core Outcome Measures in Effectiveness Trials (COMET) and Outcome Measures in Rheumatology (OMERACT) was followed. Participants were invited to participate in a Delphi study aimed at generating a consensus-based core domain set for NSLBP. A list of potential core domains was drafted and presented to the Delphi participants who were asked to judge which domains were core. Participant suggestions about overlap, aggregation, or addition of potential core domains were addressed during the study. The patients' responses were isolated to assess whether there was substantial disagreement with the rest of the Delphi panel. A priori thresholds for consensus were established before each Delphi round. All participants' responses were analysed from a quantitative and qualitative perspective to ascertain that no substantial discrepancies between the two approaches emerged. Discussion: We present the initial step in developing a COS for NSLBP. The next step will be to determine which measurement instruments adequately cover the domains.