PATIENTS AND METHODS
- Top of page
- PATIENTS AND METHODS
All patients who previously consented to participate in the Hopkins Lupus Cohort Study were included. The cohort comprises 1,500 patients with SLE (7). All patients were followed up prospectively at intervals of every 3 months, beginning at the time of entry into the cohort. Clinical features, serologic data, and damage accrual data were recorded at the time of entry into the cohort and were updated at subsequent visits. The following clinical features were defined according to the 1982 American College of Rheumatology (ACR) revised classification criteria for SLE (7): malar rash, discoid rash, photosensitivity, oral ulcer, arthritis, pleuritis, pericarditis, proteinuria (>0.5 gm/day), hemolysis, leukopenia (<4,000/mm3), lymphopenia (<1,500/mm3), and thrombocytopenia (<100,000/mm3).
Other clinical features were defined as follows. Raynaud's phenomenon was identified by blanching of fingers and/or toes induced by exposure to cold or stress, while livedo reticularis was characterized by reddish or cyanotic discoloration of the skin with a reticular pattern. Arthralgia was characterized by symptoms of joint pain but no signs of inflammation, while nephrotic syndrome was defined by daily proteinuria (>3 gm/day), and anemia was defined by hemoglobin concentration of <11.0 gm/dl in a woman and <12.0 gm/dl in a man, or a hematocrit of <33% in a woman and <36% in a man. Sjögren's syndrome consisted of dry eyes, confirmed by abnormal results of Schirmer's test, not attributable to medications (e.g., antidepressants, diuretics) or dry eyes or mouth and abnormal results of a salivary gland biopsy or dry eyes and mouth and the presence of anti-Ro and/or anti-La antibodies. Seizure, psychosis, organic brain syndrome (acute confusional state), meningitis (aseptic meningitis), depression, headache, and peripheral neuropathy were defined according to the ACR nomenclature and case definitions for neuropsychiatric lupus (8).
Stroke due to lupus was defined as a cerebrovascular event attributed to lupus activity rather than atherosclerosis, hypertension, or cardiac emboli. Myocarditis was defined as inflammation of the myocardium for which viral, bacterial, and drug causes were excluded. Gastrointestinal lupus was defined as colitis, vasculitis, or serositis of the abdominal cavity. Pancreatitis due to lupus was confirmed by imaging and/or the presence of raised levels of lipase/amylase. Arterial thrombosis included cerebrovascular accident, myocardial infarction, and transient ischemic events, confirmed by imaging, electrocardiography, cardiac enzymes, or clinical examination and history. Venous thrombosis was defined as deep vein thrombosis and/or pulmonary embolism, as ascertained by imaging. Items considered to reflect damage in SLE were defined and scored using the Systemic Lupus International Collaborating Clinics/ACR Damage Index (SDI) (9).
Baseline immunologic tests for 7 autoantibodies were performed at the first cohort visit. They included detection of anti-dsDNA by Crithidia assay; detection of anti-Ro, anti-La, anti-Sm, anti-RNP, and LAC by dilute Russell's viper venom time and confirmatory tests (10); and determination of IgG/IgM aCL. Before May 1999, the Ouchterlony double diffusion assay was performed for the detection of Ro, La, Sm, and RNP antibodies. Subsequently, enzyme-linked immunosorbent assay (ELISA) was used for detection of Ro, La, Sm, and RNP antibodies in patients enrolled after May 1999. The ELISA kits used were QUANTA Lite SS-A, QUANTA Lite SS-B, QUANTA Lite Sm, and QUANTA Lite RNP (all from Inova Diagnostics, San Diego, CA). The QUANTA Lite SS-A ELISA detects both 60-kd and 52-kd Ro antigens but does not allow differentiation of the 2 Ro antigens. The kits used for detection of aCL were ACA IgG, ACA IgM, and ACA IgA (all from Inova Diagnostics). IgG aCL levels >10 IgG phospholipid units and IgM aCL levels of >10 IgM phospholipid units were considered positive.
K-means cluster analysis (non-hierarchical clustering or Quick Cluster; SPSS version 10 software; SPSS, Chicago, IL) was used to identify groups of SLE patients with similar antoantibody patterns. Briefly, this first involves defining a disease metric with which to quantify the degree of similarity between autoantibody patterns in 2 patients. We used Euclidian distance (the square root of the sums of squared differences between patients with respect to each autoantibody). The initial centers for the clusters are chosen in a first pass of data, and patients are assigned to the closest center. Next, the cluster centers are recalculated based on the patients in the cluster, and the patients are reassigned. This iterative process continues until the clusters' means do not shift more than a given cutoff value or until the iteration limit is reached. Because we did not know in advance how many autoantibody clusters would be observed, we specified 3, 4, and then 5 clusters, respectively, in the K-means analysis, and ran the analysis several times. The outputs from the analyses with 3, 4, and 5 clusters of patients were then compared with each other with respect to the prevalence of the individual autoantibody. In order to be clinically meaningful in comparing the clinical features, the prevalence of the autoantibody should be statistically different between clusters. Finally, by clustering antibodies into 3 clusters, we identified 3 distinct groups of patients with very different autoantibody profiles.
In order for a patient with SLE to be eligible for the cluster analysis, information for all 7 selected autoantibodies (anti-dsDNA, anti-Ro, anti-La, anti-Sm, anti-RNP, LAC, and IgG/IgM aCL) had to be available at the time of the study. Eligible patients were then clustered by the K-means cluster procedure, based on these 7 autoantibodies as binary variables. Demographic variables (age at diagnosis, sex, ethnicity, annual household income, education, and smoking status), 32 clinical manifestations, and 27 items on the SDI were compared between the 3 autoantibody clusters. The conventional chi-square test was used to compare categorical variables. P values less than 0.05 were considered significant. All statistical analyses were performed using SPSS for Windows version 10.0.
- Top of page
- PATIENTS AND METHODS
This study is the largest observational study evaluating the clinical associations of autoantibody clusters in patients with SLE. The cluster analysis procedure, although rarely performed in SLE research, is an appropriate tool for examining serologic or clinical clusters in a heterogeneous disease such as SLE. All autoantibodies selected for the study were readily available and can be measured routinely in most medical centers. Thus, the observed associations are considered very clinically relevant and applicable in the daily management of patients with lupus.
The 3 distinct autoantibody clusters we observed have not been previously recognized in SLE. The prevalence of individual autoantibodies was significantly different among the 3 clusters. An exception was anti-dsDNA (as determined by Crithidia assay), a high proportion of which was present in 2 of the clusters (77% in cluster 2 and 73.2% in cluster 3). Our study confirmed the observations in several previous studies, in which autoantibodies tended to occur in pairs, such as anti-Sm and anti-RNP (2, 6), anti-Ro and anti-La (2, 11), and LAC and aCL (5).
African American patients with lupus have been reported to have more severe organ manifestations and poorer survival compared with Caucasian SLE patients (12, 13). This ethnic difference in lupus morbidity is likely multifactorial, involving the interplay of genetic, socioeconomic, and immunologic factors. Petri et al (14) and Ward et al (15) reported that noncompliance to medical therapy (as assessed by physician global assessment and percent of protocol visits) and/or low socioeconomic status, instead of African American ethnicity, was predictive of poor outcomes. However, few studies have examined the relationship between ethnicity and immunologic factors in terms of autoantibody profiles. Tikly et al (16) and Garcia et al (17) reported a high prevalence of anti-Sm and anti-RNP in African American patients. We also observed a similar association, but it failed to reach statistical significance. In addition, we also observed new associations not previously reported: the DNA/Ro/La cluster was associated with female sex (P = 0.002) and Asian ethnicity (P = 0.002), and the DNA/LAC/aCL cluster was associated with Caucasian ethnicity (P < 0.001). Although these associations need to be confirmed by further studies, they provide evidence that ethnicity differences in autoantibody clusters do exist and might be responsible for the divergent clinical presentations in various ethnic groups.
The Sm/RNP cluster has been previously linked to “the absence and the most benign form of SLE nephropathy” (6). Hoffman et al also reported that SLE patients with Sm/RNP antibodies had a lower prevalence of urine cellular casts (2). We observed that patients in the Sm/RNP cluster had the lowest incidence of renal manifestations (for proteinuria, P < 0.001 versus clusters 2 and 3; for nephrotic syndrome, P < 0.007 versus cluster 2). Of note, the Sm/RNP cluster also has the lowest percentage of anti-dsDNA (24.2%; P < 0.001). Because of the role of anti-dsDNA antibodies in disease activity (1, 18, 19) and in active lupus nephritis (20, 21), the relatively low frequency of renal involvement among patients in the Sm/RNP cluster may be attributable simply to the absence of anti-dsDNA rather than the presence of Sm/RNP. Nevertheless, this autoantibody cluster does provide a good indicator for the subset with the least frequent renal involvement and probably the most favorable renal prognosis.
Apart from the reported associations of LAC/aCL with thrombocytopenia (22) and hemolytic anemia (23), the relationship between autoantibody clusters and hematologic lupus has not been well established. We observed a negative association between Sm/RNP and anemia (P < 0.001), hemolytic anemia (P = 0.026 versus cluster 3), leukopenia (P = 0.013 versus cluster 2), lymphopenia (P < 0.001), and thrombocytopenia (P < 0.001). On the contrary, several cutaneous manifestations of lupus were more common in this cluster, including malar rash (P = 0.008 versus cluster 3), discoid lupus (P = 0.02 versus cluster 3), photosensitivity (P = 0.02 versus cluster 3), and Raynaud's phenomenon (increased trend but not statistically significant). The Sm/RNP cluster may represent the subset of SLE that is most benign, in which renal and hematologic manifestations are less common and the major manifestations are dermatologic. This is somewhat reminiscent of another report that the Sm/RNP cluster represented the subset of lupus patients with less major organ involvement (24).
The causative role of antiphospholipid antibodies, including LAC/aCL, in thromboembolism has been well described (5, 22, 25). Several groups of investigators have reported that high titers of IgG aCL were predictive of central nervous system involvement in patients with SLE (26, 27, 28). Correspondingly, we also observed the highest incidence of arterial thrombosis (P < 0.001) and venous thrombosis (P < 0.001), livedo reticularis (P < 0.001), and lupus-related stroke event (P = 0.016 versus cluster 1) in cluster 3 (DNA/LAC/aCL). However, in comparison with cluster 1 (Sm/RNP), cutaneous manifestations of lupus (malar rash, discoid lupus, photosensitivity, and Raynaud's phenomenon) were relatively less common in cluster 3 (DNA/LAC/aCL). Taken together, DNA/LAC/aCL represents a subset of SLE patients with predominantly neurologic and thrombotic events and in whom livedo reticularis is the main cutaneous manifestation.
Risk factors for organ damage in patients with SLE include older age at disease onset (29), longer duration of disease (30, 31), persistently active disease (31, 32), high cumulative dose of corticosteroid (33), and the use of cyclophosphamide (34). Antiphospholipid antibodies have been shown to be predictive of early damage in patients with SLE (35). Mikdashi and Handwerger observed that seizure and cerebrovascular accident (as measured using the SDI) were highly associated with antiphospholipid antibodies, whereas the presence of anti-dsDNA was predictive of polyneuropathy (32). However, Yee et al observed no significant association between organ damage and various autoantibodies (anti-Ro, anti-La, anti-Sm, anti-RNP, anti-dsDNA) (29). Of note, we observed strong associations between the DNA/LAC/aCL cluster and cerebrovascular accident, venous thrombosis, and an increased prevalence of cranial/peripheral neuropathy (although the difference was not statistically significant) as measured using the SDI. Interestingly, the incidence of osteoporotic fracture was also found to be highest in cluster 3 (DNA/LAC/aCL), but no similar trend was observed for other complications of corticosteroid therapy, such as cataract and avascular necrosis. Aside from the DNA/LAC/aCL cluster, the other clusters (Sm/RNP and DNA/Ro/La) seem to have a minimal role in predicting organ damage, except that an increased incidence of nephrotic syndrome was observed in patients in the DNA/Ro/La cluster. This supports results of an earlier study, in which autoantibodies other than antiphospholipid antibodies were not predictive of organ damage in SLE (29).
The current study is an exploratory analysis of 3 autoantibody clusters and their clinical correlates in a large cohort of patients with lupus. A patient in cluster 1, for instance, need not have only anti-Sm and anti-RNP antibodies. In fact, the interpretation of a “cluster” is that only the anti-Sm and anti-RNP antibodies were overrepresented in that particular cluster of patients. In other words, their autoantibody profiles were most similar, according to the clustering procedure. We characterized the individual clusters by these overrepresented autoantibodies and reported their clinical correlates. Although the laboratory technique for detection of Ro/La/Sm/RNP was changed from the Ouchterlony double diffusion assay to an ELISA in May 1999, a separate analysis of patients enrolled after this time point revealed the same autoantibody clusters. Thus, our observations are likely genuine and were not biased by this change in laboratory method. From a clinical standpoint, we can conclude that patients with these predefined autoantibody profiles might manifest the disease as we have reported. The present study cannot test the causal association between the individual autoantibodies and the specific clinical features and also cannot be used to predict the clinical manifestations in patients with other autoantibody combinations.
The results of our study confirm that in the setting of lupus, autoantibodies do exist in clusters. All of the subjects in the current study were unselected. Therefore, the clusters observed are likely to be a true reflection of different disease patterns and are likely to be generalizable to other centers. From the clinical standpoint, autoantibody clusters, similar to individual autoantibodies, help to differentiate between various subsets of SLE.