Objective. Chronic inflammatory back pain (IBP) is frequently reported in axial SpA (axSpA) but also in the general population. We evaluated a recently proposed two-step referral system for early recognition of axSpA in primary care and compare it with other combinations of symptoms and SpA-related items. Methods. Consecutive chronic back pain patients <= 45 years of age answered a questionnaire and were seen by a primary care physician who decided whether HLA-B27 needed to be determined. They were then referred to a rheumatologist who made the diagnosis. Generally sticking to the two-step system with HLA-B27 as an additional option, combinations with a sensitivity >90% and a likelihood ratio >4 were compared. Results. A total of 326 patients were included, 46 of whom were diagnosed with axSpA (14.1 %). The sensitivity of the strategy was 87%, the specificity was 56.8% and the positive and negative predictive values were 24.8% and 96.4%, respectively. A 'good response to NSAIDs', 'morning stiffness >30 min' and 'elevated C-reactive protein' performed best, with a sensitivity of 91%, specificity of 67%, positive predictive value of 31% and negative predictive value of 98%. On that basis, only three patients had to be seen by a rheumatologist to diagnose one. Conclusion. The earlier proposed referral system worked well but was outperformed by other combinations with high sensitivity and better specificity, which deserve to be prospectively studied.