To determine whether serum C3 or C4 is more likely to be normal during systemic lupus erythematosus (SLE) remission and abnormal during SLE relapse we studied twelve SLE patients who presented with severe nephritis. The patients were followed long term (12 to 77 months) through multiple relapses (N = 41) and remissions (N = 13) defined by protocol. A total of 471 serum samples were obtained at defined intervals during these relapses and remissions and were analyzed for C3 and C4 levels by two different methods: nephelometry (N) and radial immunodiffusion (R). During SLE remission (defined by protocol and without reference to serum complement levels), C3 measured by N-assay (C3N) and by R-assay (C3R) tended to be normal (specificity of 93% and 71%, respectively). By contrast, C4 measured by N-assay (C4N) and by R-assay (C4R) showed no such tendency (specificity of 50% for both C4N and C4R). During SLE relapse (defined by protocol and without reference to serum complement levels), C3N and C3R were more likely to be abnormal (sensitivity 95% and 85%, respectively) compared with C4N and C4R (sensitivity 56% and 54%, respectively, P less than 0.001 compared with corresponding values for the C3 assay). Analysis by receiver-operator characteristic (ROC) curves demonstrated that the reduced diagnostic sensitivity of C4 versus C3 is not explained by use of an inappropriate lower limits of normal (LLN) for C4.(ABSTRACT TRUNCATED AT 250 WORDS)