The need to treat rheumatoid arthritis patients aggressively early in the disease course to avoid permanent radiological damage may result in overtreating a proportion of patients who would have achieved low disease activity with minimal treatment. The prediction of disease course and severity at diagnosis would therefore allow the adaptation of treatment regimes to specific patient needs. First, we review briefly advances made in the identification of non-genetic markers of erosive disease. Secondly, we present a review of the literature on the role of human leukocyte antigen (HLA)-DRB1 rheumatoid arthritis susceptibility alleles in predicting erosive disease or the extent of radiological damage. Finally, studies of non-HLA rheumatoid arthritis susceptibility single nucleotide polymorphisms as putative markers of radiological destruction are reviewed. Currently, no demographic, clinical, laboratory or genetic predictor of disease severity can be reliably used to guide clinical decisions. We highlight specific methodological issues inherent to genetic severity studies and suggest avenues for future research. TOUCH BRIEFINGS 2012 ; 2012573674