Laboratory tests as predictors of disease exacerbations in systemic lupus erythematosus. Why some tests fail

To evaluate whether changes in laboratory test values are either simultaneous with or precede disease exacerbations in patients with systemic lupus erythematosus (SLE).

At 9, 6, and 3 months preceding a flare in disease activity (defined as a rise of > or = 6 points in the modified SLE Disease Activity Index), laboratory tests were performed to measure patients' hematocrit levels, white blood cell, lymphocyte, and platelet counts, erythrocyte sedimentation rate, C1q binding, DNA binding, and levels of C3 and C4. Flares were classified as either present or absent, and were divided into renal, vasculitic, central nervous system, skin, serosal, and musculoskeletal subgroups. The predictive patterns were 1) the simultaneous change in the test value from the mean of 9, 6, and 3 months preceding a flare to the time of the flare; 2) the gradual change, following a linear time trend, in test results for the same time points; and 3) the change from the mean of 9 and 6 months to 3 months preceding a flare, as a measure of predictive ability. These analyses used repeated-measures analysis of variance models. Multiple linear regression was used to study the cross-sectional association of average-over-time differences in test results with patients' flare subgroup.

Among 202 patients with SLE (median followup 86.5 months), 83 flares occurred in 53 patients. Of 189 statistical contrasts performed, only 14 were significant (versus 10 expected), and the differences were of minor importance. Nonetheless, evaluation of all test results over each patient's observed disease course revealed significant differences between selected test values in association with specific types of flare.

Fluctuations in laboratory test values are poor predictors of disease exacerbations in SLE. Cross-sectional evaluation of some test results revealed differences at the time of flare for those patients who were destined to have different types of flares, because these values differed over the entire study period. This pattern explains the frequent cross-sectional association of disease activity with laboratory test results, and the inconsistent association of flares with recent changes in test values.