Objectives: The broad aim of this study was to assess the homogeneity of patients included in trials of nonspecific low back pain (NSLBP). To do this, we investigated the consistency and clarity of criteria used to identify and exclude participants with serious pathologies and nerve root compromise in randomized controlled trials, investigating interventions for NSLBP. Methods: We searched Medline database for randomized controlled trials of low back pain (LBP) published between 2000 and 2009. We then randomly selected and screened trials for inclusion until we had 50 eligible trials. Data were extracted on the criteria used to identify cases of serious conditions (e.g. cancer, fracture) and nerve root involvement. Results: The majority of papers (35/50) explicitly excluded patients with serious pathology. However, the terminology used and examples given were highly variable. Nerve root involvement was an exclusion criterion in the majority but not all studies. The criteria used for excluding patients with nerve root involvement varied greatly between studies. The most common criteria were 'motor, sensory or reflex changes' (nine studies), followed by 'pain radiating below the knee' (five studies) and 'reduced straight leg raise which reproduces leg pain' (five studies). In half of the included studies, the criteria used, while alluding to nerve root involvement, were not explained adequately for us to determine the types of patients included or excluded. Discussion: The inconsistent and unclear criteria used to identify cases of serious pathology and nerve root compromise means that published trials of LBP likely include heterogeneous patient populations. This trait limits our ability to make comparisons across trials or pool studies. Standardization and consensus is important for future research.