The objective of this study was to determine the frequency of loss to follow-up and the factors predictive of its occurrence in a systemic lupus erythematosus (SLE) multiethnic cohort. We studied SLE patients from the LUMINA cohort (Hispanics from Texas and from the Island of Puerto Rico, African-Americans and Caucasians). Loss to follow-up was defined as subjects who failed to attend two or more of the latest consecutive yearly study visits. The relationship between baseline features and loss to follow-up was examined by univariable and multivariable Cox regression analyses with loss to follow-up as the dependent variable. The retention rate in the cohort was estimated by the Kaplan-Meier method. Five-hundred and fifty-four patients with a mean (SD) follow-up of 3.4 (2.9) years were studied. One-hundred and fifty-eight (29%) met the definition of lost to follow-up. The cumulative loss to follow-up rate at five years was 36%. The cumulative loss to follow-up rate at five years was higher for the African-Americans. Patients lost to follow-up tended to be younger and more likely to have poor social support and higher levels of helplessness. They also tended to have more renal involvement and more active disease as per the Systemic Lupus Activity Measure-Revised. Disease activity (hazard ratio = 1.04, 95% confidence interval 1.01-1.07, P = 0.02) was the only variable independently contributing to loss to follow-up. Our data suggest that in longitudinal SLE studies, loss to follow-up does not occur at random and it differs between ethnic groups and is also particularly higher among patients with more active disease. Pro-active measures may need to be applied to decrease the probability of patients 'at risk' of becoming lost to follow-up and to preserve the integrity of the cohort. © 2006 Edward Arnold (Publishers) Ltd.