Objective For better management of RA, new biomarkers are needed to predict the development of different disease courses. This study aims to identify autoantibodies against epitopes on proteins in the joints and to predict disease outcome in patients with new onset RA.Methods Sera from new-onset RA patients from the Swedish BARFOT (Better Anti Rheumatic PharmacOTherapy) and TIRA-2 (Swedish acronym for 'tidiga insatser vid reumatoid artrit') cohorts (n = 1986) were screened for autoantibodies to selected peptides (JointIDs) in a bead-based multiplex flow immunoassay. Disease outcomes included Boolean remission 1.0, swollen joint count and radiographic destruction. Multivariate logistic regression and zero-inflated negative binomial models that accounted for clinical factors were used to identify JointIDs with the strongest potential to predict prognosis.Results Boolean remission was predicted with 42% sensitivity and 75% specificity in male patients positive for antibodies to a non-modified collagen type II (COL2) peptide at 12 months. When antibodies to a specific citrullinated cartilage oligomeric protein (COMP) peptide were absent and the patient was in Boolean remission at 6 months, the sensitivity was 13% and the specificity 99%. Positivity for the non-modified COL2 peptide also reduced the frequency of swollen joints by 41% and 33% at 6 and 12 months, respectively. Antibodies to CCP predicted joint destruction with low specificity (58%). Positivity for a COL2 and a glucose-6-phosphate dehydrogenase peptide in citrullinated forms increased specificity (86%) at the expense of sensitivity (39%).Conclusion Autoantibodies against joint-related proteins at RA diagnosis predict remission with high specificity and, in combination with clinical factors, may guide future treatment decisions.