Objective To refine the spectrum of anti-Ku-associated disease, a condition that is equivocally described by current diagnostic criteria for connective tissue diseases.

Methods Among 42 consecutive patients harbouring anti-Ku antibodies, subgroups with similar phenotypes and prognosis were delineated without an a priori diagnosis using hierarchical clustering analysis of the cumulative clinico-biological features recorded during the follow-up. Features present at baseline that most efficiently predicted the outcomes were then identified using a sensitivity–specificity sum maximisation approach.

Results Clinico-biological features were clustered into three groups. Glomerulonephritis and ILD, the two fatal complications in this cohort, were unequally distributed between the three clusters that additionally differed on six clinico-biological features.

Among features present at baseline, elevated serum level of creatine kinase (CK) and anti-dsDNA antibodies were generally mutually exclusive and most efficiently predicted the cluster belonging at last follow-up. Anti-Ku patients with elevated CK had a 22-fold higher risk of ILD while anti-Ku patients with anti-dsDNA antibodies had a 13-fold higher risk of glomerulonephritis

Conclusion “Anti-Ku with elevated CK” syndrome and “anti-Ku with anti-dsDNA” syndrome represent two distinct entities that are important to recognise in order to best tailor patient care.