Background Inception cohorts aim to describe chronic diseases from diagnosis and over years of follow-up. Axial spondyloarthritis (axSpA) diagnosis might be challenging during the first years of the disease. Thus, identifying the features that will be associated with a confirmed diagnosis over time is key. Objectives To assess the frequency and the predisposing factors for a change of an initial diagnosis in an inception axSpA cohort. Methods DESIR is an ongoing national multicentre inception axSpA cohort with currently 12.5 years of follow-up. At the entry visit and confirmed at each visit, the diagnosis of axSpA was based on the opinion of the treating rheumatologist. Follow-up was interrupted in case of a change in this initial diagnosis. Multiple imputation was used to estimate the probability of a change in the initial diagnosis of axSpA for each patient lost to follow-up. Factors predisposing to an unchanged diagnosis of axSpA were then assessed using a multivariate logistic regression model on the imputed data sets. Results Of the 708 patients included, over 10 years of follow-up, 45 (6.4%) were excluded due to a diagnosis change and 300 (42.4%) patients were lost to follow-up. Based on the imputation of these 300 patients, a change in their initial axSpA diagnosis was estimated in 42 (14.0%). Factors predisposing to an unchanged initial axSpA diagnosis during follow-up were (ORs (95% CIs)): radiographic sacroiliitis: 17.0 (4.1 to 71.0); psoriasis: 5.3 (2.0 to 14.3); CRP≥6 mg/L: 2.7 (1.3 to 5.3); good NSAID response: 2.5 (1.5 to 4.2); HLA B27+: 2.0 (1.3 to 3.3); anterior chest wall pain: 2.0 (1.2 to 3.3) and female sex: 1.9 (1.2 to 3.0). Conclusion These data suggest that a change in diagnosis in recent onset axSpA exists, but is not frequent, and is less likely to occur in the presence of objective features at baseline.