- Split View
-
Views
-
Cite
Cite
José María Pego-Reigosa, Ana Lois-Iglesias, Íñigo Rúa-Figueroa, María Galindo, Jaime Calvo-Alén, Jacobo de Uña-Álvarez, Vanessa Balboa-Barreiro, Jesús Ibáñez Ruan, Alejandro Olivé, Manuel Rodríguez-Gómez, Antonio Fernández Nebro, Mariano Andrés, Celia Erausquin, Eva Tomero, Loreto Horcada Rubio, Esther Uriarte Isacelaya, Mercedes Freire, Carlos Montilla, Ana I. Sánchez-Atrio, Gregorio Santos-Soler, Antonio Zea, Elvira Díez, Javier Narváez, Ricardo Blanco-Alonso, Lucía Silva-Fernández, María Esther Ruiz-Lucea, Mónica Fernández-Castro, José Ángel Hernández-Beriain, Marian Gantes-Mora, Blanca Hernández-Cruz, José Pérez-Venegas, Ángela Pecondón-Español, Carlos Marras Fernández-Cid, Mónica Ibáñez-Barcelo, Gema Bonilla, Vicenç Torrente-Segarra, Iván Castellví, Juan José Alegre, Joan Calvet, José Luis Marenco de la Fuente, Enrique Raya, Tomás Ramón Vázquez-Rodríguez, Víctor Quevedo-Vila, Santiago Muñoz-Fernández, Teresa Otón, Anisur Rahman, Francisco Javier López-Longo, Relationship between damage clustering and mortality in systemic lupus erythematosus in early and late stages of the disease: cluster analyses in a large cohort from the Spanish Society of Rheumatology Lupus Registry, Rheumatology, Volume 55, Issue 7, July 2016, Pages 1243–1250, https://doi.org/10.1093/rheumatology/kew049
- Share Icon Share
Abstract
Objectives. To identify patterns (clusters) of damage manifestations within a large cohort of SLE patients and evaluate the potential association of these clusters with a higher risk of mortality.
Methods. This is a multicentre, descriptive, cross-sectional study of a cohort of 3656 SLE patients from the Spanish Society of Rheumatology Lupus Registry. Organ damage was ascertained using the Systemic Lupus International Collaborating Clinics Damage Index. Using cluster analysis, groups of patients with similar patterns of damage manifestations were identified. Then, overall clusters were compared as well as the subgroup of patients within every cluster with disease duration shorter than 5 years.
Results. Three damage clusters were identified. Cluster 1 (80.6% of patients) presented a lower amount of individuals with damage (23.2 vs 100% in clusters 2 and 3, P < 0.001). Cluster 2 (11.4% of patients) was characterized by musculoskeletal damage in all patients. Cluster 3 (8.0% of patients) was the only group with cardiovascular damage, and this was present in all patients. The overall mortality rate of patients in clusters 2 and 3 was higher than that in cluster 1 (P < 0.001 for both comparisons) and in patients with disease duration shorter than 5 years as well.
Conclusion. In a large cohort of SLE patients, cardiovascular and musculoskeletal damage manifestations were the two dominant forms of damage to sort patients into clinically meaningful clusters. Both in early and late stages of the disease, there was a significant association of these clusters with an increased risk of mortality. Physicians should pay special attention to the early prevention of damage in these two systems.
Cardiovascular and musculoskeletal damage is associated with increased mortality, even in early stages of SLE.
Physicians should pay special attention to the early prevention of cardiovascular and musculoskeletal damage in SLE.
Introduction
SLE is a complex systemic rheumatic disease in which several domains should be assessed: disease activity, organ damage and health-related quality of life [ 1 ]. As survival in SLE patients has improved over the past decades [ 2 , 3 ], evaluation of organ damage has become more relevant.
Damage in SLE is defined as an irreversible change, not related to the active inflammation that occurs after the diagnosis of the disease, which is present for at least 6 months [ 4 ]. The SLICC/ACR Damage Index (SDI) has been demonstrated to be a valid and reliable measure of damage in SLE [ 4–6 ].
Various studies have shown that damage predicts future mortality in SLE patients [ 7 , 8 ]. However, the majority of these studies use the global SDI score to analyse the association between damage and mortality [ 9–15 ]. Little is known about clustering of the different damage manifestations and the impact of the different domains of the SDI on survival [ 16 , 17 ]. Furthermore, the existing studies analyse samples of SLE patients in which the number of deceased patients is small so it is difficult to draw significant conclusions regarding mortality.
Therefore, the primary objective of our present study was to identify patterns (clusters) of damage manifestations within a large cohort of SLE patients. In addition, we sought to evaluate the potential association of these clusters with the risk of mortality.
Materials and patients
Research study network
The Registry of Systemic Lupus Erythematosus Patients of the Spanish Society of Rheumatology (RELESSER) is a hospital-based registry that consists of two stages. The first one is a cross-sectional stage of which the main objective is description of the characteristics and comorbidities of patients diagnosed with SLE in Spain. There is a longitudinal follow-up stage over time, with repeated yearly visits. The RELESSER Registry was conducted by the Systemic Autoimmune Diseases Study Group of the Spanish Society of Rheumatology and it involved 45 Rheumatology departments. All investigators signed a written commitment before participation in the RELESSER Registry. Informed consent was obtained from all patients who participated in the longitudinal stage. The RELESSER Registry was approved by the local Ethics Committees of the participating centres in accordance with the Declaration of Helsinki’s guidelines for research on humans. Our study did not require ethical approval.
Study design
This is a national, multicentre, descriptive study of a cohort with a cross-sectional analysis at the time of the last medical visit of every patient (or death, if applicable). A detailed description of its methodology has been provided elsewhere [ 18 ]. Briefly, a specific protocol was created to collect data on approximately 400 variables per patient. The information was obtained by reviewing clinical histories and electronically collecting the data. In order to minimize missing data, all investigators were encouraged to carry out a census of their SLE patients and to fill in any missing data. In order to ensure data homogeneity and quality, every item in the protocol had a highly standardized definition; a previous training course for investigators was carried out to avoid information bias; and all investigators had online access to guidelines on how to complete the protocol. The first patient was entered into the Registry in October 2011. Electronic data collection finished in August 2012. Then, a professional monitor with experience in rheumatologic studies performed a review of the database to detect missing or inconsistent data; these were then discussed with the principal investigators and sent to the subinvestigators for additions and corrections.
Patients
We included unselected consecutive patients who fulfilled the following inclusion criteria: (i) age ⩾16 years and (ii) ⩾4 ACR 1997 criteria for the classification of SLE [ 19 , 20 ]. There were no specific exclusion criteria.
In order to avoid selection bias, patients were widely and homogeneously spread across Spain. Virtually all patients with SLE treated in our country are referred to hospitals, thus avoiding the possibility of centre selection bias.
Variables and definitions
Data for ∼400 variables per patient were collected in the RELESSER Registry [ 18 ]. For the current study, the following variables were included in the analyses: demographic features (age, gender and ethnicity), chronologic data (time of first symptom of the disease, time of diagnosis of SLE, time of follow-up, date of every damage event and date of death) and cumulative manifestations of damage (using the definitions of the SDI) at the time of the last medical visit of the patient (or death, if applicable).
Statistical analysis
Cluster analysis was carried out by applying k -means statistical analysis (partitioning method) to identify groups of SLE patients with similar patterns of damage manifestations by the end of follow-up. The method starts with k clusters (fixed a priori ) and then moves patients between clusters with the goal of minimizing variability within clusters and maximizing variability between clusters. Euclidean distance was used as a measure of similarity between the profiles of damage of two patients.
We ran the k -means analysis with 2, 3, 4 and 5 clusters and the outputs were compared with each other. A plot of the total within-groups sums of squares against the number of clusters was used to choose the optimal number of clusters.
Three clusters of patients were finally compared. The comparison of clusters was performed for the whole cohort and for the subgroup of patients with disease duration <5 years. Results were expressed as mean ( s . d .) for continuous variables, and as number of patients (percentages) for binary and categorical variables. The Analysis of Variance (ANOVA) test was used to compare continuous variables. The chi-squared test was employed to compare the frequencies of categorical variables among the three groups of patients. Statistical significance was concluded when P < 0.05. All analyses were performed with R Statistical Software, version 3.1.1. (R Foundation for Statistical Computing, Vienna, Austria).
Results
There were 3656 SLE patients in our cohort. The demographics and the global information of the cohort with respect to damage and mortality are summarized in Table 1 . Data about the involvement of every SDI domain in our cohort are presented in Table 2 .
Patient characteristic . | Value . |
---|---|
Age at diagnosis, mean ( s.d.) , years | 35.2 (14.7) |
Sex, n (%) | |
Male | 360 (9.7) |
Female | 3296 (90.3) |
Race/ethnicity, n (%) | 3307 (93.2) |
Caucasian | |
Afro-Caribbean | 8 (0.2) |
Latin American | 185 (5.2) |
Asian | 21 (0.6) |
Others | 29 (0.8) |
Disease duration, mean ( s.d.) , months | 120.2 (87.7) |
SDI score, mean ( s.d.) | 1.2 (1.7) |
Patients with some damage, n (%) | 1391 (38.1) |
Number of SDI domains affected, mean ( s.d.) | 0.7 (1.1) |
Death, n (%) | 207 (5.7) |
Patient characteristic . | Value . |
---|---|
Age at diagnosis, mean ( s.d.) , years | 35.2 (14.7) |
Sex, n (%) | |
Male | 360 (9.7) |
Female | 3296 (90.3) |
Race/ethnicity, n (%) | 3307 (93.2) |
Caucasian | |
Afro-Caribbean | 8 (0.2) |
Latin American | 185 (5.2) |
Asian | 21 (0.6) |
Others | 29 (0.8) |
Disease duration, mean ( s.d.) , months | 120.2 (87.7) |
SDI score, mean ( s.d.) | 1.2 (1.7) |
Patients with some damage, n (%) | 1391 (38.1) |
Number of SDI domains affected, mean ( s.d.) | 0.7 (1.1) |
Death, n (%) | 207 (5.7) |
SDI: SLICC/ACR Damage Index; n: number.
Patient characteristic . | Value . |
---|---|
Age at diagnosis, mean ( s.d.) , years | 35.2 (14.7) |
Sex, n (%) | |
Male | 360 (9.7) |
Female | 3296 (90.3) |
Race/ethnicity, n (%) | 3307 (93.2) |
Caucasian | |
Afro-Caribbean | 8 (0.2) |
Latin American | 185 (5.2) |
Asian | 21 (0.6) |
Others | 29 (0.8) |
Disease duration, mean ( s.d.) , months | 120.2 (87.7) |
SDI score, mean ( s.d.) | 1.2 (1.7) |
Patients with some damage, n (%) | 1391 (38.1) |
Number of SDI domains affected, mean ( s.d.) | 0.7 (1.1) |
Death, n (%) | 207 (5.7) |
Patient characteristic . | Value . |
---|---|
Age at diagnosis, mean ( s.d.) , years | 35.2 (14.7) |
Sex, n (%) | |
Male | 360 (9.7) |
Female | 3296 (90.3) |
Race/ethnicity, n (%) | 3307 (93.2) |
Caucasian | |
Afro-Caribbean | 8 (0.2) |
Latin American | 185 (5.2) |
Asian | 21 (0.6) |
Others | 29 (0.8) |
Disease duration, mean ( s.d.) , months | 120.2 (87.7) |
SDI score, mean ( s.d.) | 1.2 (1.7) |
Patients with some damage, n (%) | 1391 (38.1) |
Number of SDI domains affected, mean ( s.d.) | 0.7 (1.1) |
Death, n (%) | 207 (5.7) |
SDI: SLICC/ACR Damage Index; n: number.
SDI domain . | n (%) . |
---|---|
Musculoskeletal | 504 (13.8) |
Ocular | 311 (8.6) |
Cardiovascular | 292 (8.0) |
Renal | 225 (6.1) |
Neuropsychiatric | 221 (6.0) |
Malignancy | 170 (4.7) |
Peripheral vascular | 163 (4.4) |
Pulmonary | 132 (3.7) |
Cutaneous | 124 (3.3) |
Premature gonadal failure | 92 (2.5) |
Diabetes | 88 (2.4) |
Gastrointestinal | 71 (1.9) |
SDI domain . | n (%) . |
---|---|
Musculoskeletal | 504 (13.8) |
Ocular | 311 (8.6) |
Cardiovascular | 292 (8.0) |
Renal | 225 (6.1) |
Neuropsychiatric | 221 (6.0) |
Malignancy | 170 (4.7) |
Peripheral vascular | 163 (4.4) |
Pulmonary | 132 (3.7) |
Cutaneous | 124 (3.3) |
Premature gonadal failure | 92 (2.5) |
Diabetes | 88 (2.4) |
Gastrointestinal | 71 (1.9) |
SDI: SLICC/ACR Damage Index; n: number.
SDI domain . | n (%) . |
---|---|
Musculoskeletal | 504 (13.8) |
Ocular | 311 (8.6) |
Cardiovascular | 292 (8.0) |
Renal | 225 (6.1) |
Neuropsychiatric | 221 (6.0) |
Malignancy | 170 (4.7) |
Peripheral vascular | 163 (4.4) |
Pulmonary | 132 (3.7) |
Cutaneous | 124 (3.3) |
Premature gonadal failure | 92 (2.5) |
Diabetes | 88 (2.4) |
Gastrointestinal | 71 (1.9) |
SDI domain . | n (%) . |
---|---|
Musculoskeletal | 504 (13.8) |
Ocular | 311 (8.6) |
Cardiovascular | 292 (8.0) |
Renal | 225 (6.1) |
Neuropsychiatric | 221 (6.0) |
Malignancy | 170 (4.7) |
Peripheral vascular | 163 (4.4) |
Pulmonary | 132 (3.7) |
Cutaneous | 124 (3.3) |
Premature gonadal failure | 92 (2.5) |
Diabetes | 88 (2.4) |
Gastrointestinal | 71 (1.9) |
SDI: SLICC/ACR Damage Index; n: number.
Cluster analysis
Among the 3656 patients, three subgroups with particular patterns of damage were identified by k -means cluster analysis. Cluster 1 included 2949 (80.6%) patients, cluster 2 included 415 (11.4%) patients and cluster 3 included 292 (8.0%) patients. Characteristics within the clusters and P-values for between-cluster comparisons are shown in Table 3 .
. | Cluster 1, n = 2949 (80.6%) . | Cluster 2, n = 415 (11.4%) . | Cluster 3, n = 292 (8.0%) . | P-values . |
---|---|---|---|---|
Age at diagnosis, mean | ||||
( s.d.) , years | 34.4 (14.1) a,b | 36.7 (15.7) b,c | 40.3 (15.7) a,c | <0.001 |
Sex, n (%) | ||||
Male | 257 (8.7) b | 40 (9.7) b | 56 (19.2) a,c | <0.001 |
Female | 2686 (91.3) | 374 (90.3) | 236 (80.8) | |
Race/ethnicity, n (%) | ||||
Caucasian | 2644 (92.4) | 384 (95.8) | 279 (96.9) | 0.07 |
Afro-Caribbean | 8 (0.3) | 0 (0) | 0 (0) | |
Latin American | 163 (5.7) | 14 (3.5) | 8 (2.8) | |
Asian | 20 (0.7) | 1 (0.2) | 0 (0) | |
Others | 26 (0.9) | 2 (0.5) | 1 (0.3) | |
Disease duration, mean ( s.d.) , months | 109.9 (81.3) a,b | 167.1 (98.9) c | 154.9 (100.2) c | <0.001 |
Damage, n (%) | ||||
Ocular | 171 (5.8) a,b | 76 (18.3) c | 64 (21.9) c | <0.001 |
Neuropsychiatric | 123 (4.2) a,b | 47 (11.3) b,c | 51 (17.5) a,b | <0.001 |
Renal | 132 (4.5) a,b | 36 (8.7) b,c | 57 (19.5) a,b | <0.001 |
Pulmonary | 58 (2.0) a,b | 33 (8.0) b,c | 41 (14.0) a,b | <0.001 |
Cardiovascular | 0 (0) b | 0 (0) b | 292 (100) a,b | <0.001 |
Peripheral vascular | 88 (3.0) a,b | 37 (8.9) c | 38 (13.0) c | <0.001 |
Gastrointestinal | 44 (1.5) a,b | 15 (3.6) c | 12 (4.1) c | <0.001 |
Musculoskeletal | 0 (0) a,b | 415 (100) b,c | 89 (30.5) a,c | <0.001 |
Cutaneous | 56 (1.9) a,b | 35 (8.4) c | 33 (11.3) c | <0.001 |
Diabetes | 56 (1.9) b | 12 (2.9) b | 20 (6.8) a,c | <0.001 |
Malignancy | 114 (3.9) a | 38 (9.2) c | 18 (6.2) | <0.001 |
Premature gonadal failure | 48 (1.6) a,b | 28 (6.7) c | 16 (5.5) c | <0.001 |
SDI score, mean ( s.d.) | 0.7 (1.1) a,b | 2.6 (1.8) b,c | 3.8 (2.4) a,c | <0.001 |
Patients with damage, n (%) | 684 (23.2) a,b | 415 (100) c | 292 (100) c | <0.001 |
Number of SDI domains affected, mean ( s.d.) | 0.3 (0.6) a,b | 1.9 (1.0) b,c | 2.5 (1.4) a,c | <0.001 |
Death, n (%) | 102 (3.7) a,b | 45 (10.8) b,c | 60 (20.5) a,c | <0.001 |
. | Cluster 1, n = 2949 (80.6%) . | Cluster 2, n = 415 (11.4%) . | Cluster 3, n = 292 (8.0%) . | P-values . |
---|---|---|---|---|
Age at diagnosis, mean | ||||
( s.d.) , years | 34.4 (14.1) a,b | 36.7 (15.7) b,c | 40.3 (15.7) a,c | <0.001 |
Sex, n (%) | ||||
Male | 257 (8.7) b | 40 (9.7) b | 56 (19.2) a,c | <0.001 |
Female | 2686 (91.3) | 374 (90.3) | 236 (80.8) | |
Race/ethnicity, n (%) | ||||
Caucasian | 2644 (92.4) | 384 (95.8) | 279 (96.9) | 0.07 |
Afro-Caribbean | 8 (0.3) | 0 (0) | 0 (0) | |
Latin American | 163 (5.7) | 14 (3.5) | 8 (2.8) | |
Asian | 20 (0.7) | 1 (0.2) | 0 (0) | |
Others | 26 (0.9) | 2 (0.5) | 1 (0.3) | |
Disease duration, mean ( s.d.) , months | 109.9 (81.3) a,b | 167.1 (98.9) c | 154.9 (100.2) c | <0.001 |
Damage, n (%) | ||||
Ocular | 171 (5.8) a,b | 76 (18.3) c | 64 (21.9) c | <0.001 |
Neuropsychiatric | 123 (4.2) a,b | 47 (11.3) b,c | 51 (17.5) a,b | <0.001 |
Renal | 132 (4.5) a,b | 36 (8.7) b,c | 57 (19.5) a,b | <0.001 |
Pulmonary | 58 (2.0) a,b | 33 (8.0) b,c | 41 (14.0) a,b | <0.001 |
Cardiovascular | 0 (0) b | 0 (0) b | 292 (100) a,b | <0.001 |
Peripheral vascular | 88 (3.0) a,b | 37 (8.9) c | 38 (13.0) c | <0.001 |
Gastrointestinal | 44 (1.5) a,b | 15 (3.6) c | 12 (4.1) c | <0.001 |
Musculoskeletal | 0 (0) a,b | 415 (100) b,c | 89 (30.5) a,c | <0.001 |
Cutaneous | 56 (1.9) a,b | 35 (8.4) c | 33 (11.3) c | <0.001 |
Diabetes | 56 (1.9) b | 12 (2.9) b | 20 (6.8) a,c | <0.001 |
Malignancy | 114 (3.9) a | 38 (9.2) c | 18 (6.2) | <0.001 |
Premature gonadal failure | 48 (1.6) a,b | 28 (6.7) c | 16 (5.5) c | <0.001 |
SDI score, mean ( s.d.) | 0.7 (1.1) a,b | 2.6 (1.8) b,c | 3.8 (2.4) a,c | <0.001 |
Patients with damage, n (%) | 684 (23.2) a,b | 415 (100) c | 292 (100) c | <0.001 |
Number of SDI domains affected, mean ( s.d.) | 0.3 (0.6) a,b | 1.9 (1.0) b,c | 2.5 (1.4) a,c | <0.001 |
Death, n (%) | 102 (3.7) a,b | 45 (10.8) b,c | 60 (20.5) a,c | <0.001 |
a Significantly different from cluster 2.
b Significantly different from cluster 3.
c Significantly different from cluster 1. SDI: Systemic Lupus International Collaborating Clinics/ACR Damage Index.
. | Cluster 1, n = 2949 (80.6%) . | Cluster 2, n = 415 (11.4%) . | Cluster 3, n = 292 (8.0%) . | P-values . |
---|---|---|---|---|
Age at diagnosis, mean | ||||
( s.d.) , years | 34.4 (14.1) a,b | 36.7 (15.7) b,c | 40.3 (15.7) a,c | <0.001 |
Sex, n (%) | ||||
Male | 257 (8.7) b | 40 (9.7) b | 56 (19.2) a,c | <0.001 |
Female | 2686 (91.3) | 374 (90.3) | 236 (80.8) | |
Race/ethnicity, n (%) | ||||
Caucasian | 2644 (92.4) | 384 (95.8) | 279 (96.9) | 0.07 |
Afro-Caribbean | 8 (0.3) | 0 (0) | 0 (0) | |
Latin American | 163 (5.7) | 14 (3.5) | 8 (2.8) | |
Asian | 20 (0.7) | 1 (0.2) | 0 (0) | |
Others | 26 (0.9) | 2 (0.5) | 1 (0.3) | |
Disease duration, mean ( s.d.) , months | 109.9 (81.3) a,b | 167.1 (98.9) c | 154.9 (100.2) c | <0.001 |
Damage, n (%) | ||||
Ocular | 171 (5.8) a,b | 76 (18.3) c | 64 (21.9) c | <0.001 |
Neuropsychiatric | 123 (4.2) a,b | 47 (11.3) b,c | 51 (17.5) a,b | <0.001 |
Renal | 132 (4.5) a,b | 36 (8.7) b,c | 57 (19.5) a,b | <0.001 |
Pulmonary | 58 (2.0) a,b | 33 (8.0) b,c | 41 (14.0) a,b | <0.001 |
Cardiovascular | 0 (0) b | 0 (0) b | 292 (100) a,b | <0.001 |
Peripheral vascular | 88 (3.0) a,b | 37 (8.9) c | 38 (13.0) c | <0.001 |
Gastrointestinal | 44 (1.5) a,b | 15 (3.6) c | 12 (4.1) c | <0.001 |
Musculoskeletal | 0 (0) a,b | 415 (100) b,c | 89 (30.5) a,c | <0.001 |
Cutaneous | 56 (1.9) a,b | 35 (8.4) c | 33 (11.3) c | <0.001 |
Diabetes | 56 (1.9) b | 12 (2.9) b | 20 (6.8) a,c | <0.001 |
Malignancy | 114 (3.9) a | 38 (9.2) c | 18 (6.2) | <0.001 |
Premature gonadal failure | 48 (1.6) a,b | 28 (6.7) c | 16 (5.5) c | <0.001 |
SDI score, mean ( s.d.) | 0.7 (1.1) a,b | 2.6 (1.8) b,c | 3.8 (2.4) a,c | <0.001 |
Patients with damage, n (%) | 684 (23.2) a,b | 415 (100) c | 292 (100) c | <0.001 |
Number of SDI domains affected, mean ( s.d.) | 0.3 (0.6) a,b | 1.9 (1.0) b,c | 2.5 (1.4) a,c | <0.001 |
Death, n (%) | 102 (3.7) a,b | 45 (10.8) b,c | 60 (20.5) a,c | <0.001 |
. | Cluster 1, n = 2949 (80.6%) . | Cluster 2, n = 415 (11.4%) . | Cluster 3, n = 292 (8.0%) . | P-values . |
---|---|---|---|---|
Age at diagnosis, mean | ||||
( s.d.) , years | 34.4 (14.1) a,b | 36.7 (15.7) b,c | 40.3 (15.7) a,c | <0.001 |
Sex, n (%) | ||||
Male | 257 (8.7) b | 40 (9.7) b | 56 (19.2) a,c | <0.001 |
Female | 2686 (91.3) | 374 (90.3) | 236 (80.8) | |
Race/ethnicity, n (%) | ||||
Caucasian | 2644 (92.4) | 384 (95.8) | 279 (96.9) | 0.07 |
Afro-Caribbean | 8 (0.3) | 0 (0) | 0 (0) | |
Latin American | 163 (5.7) | 14 (3.5) | 8 (2.8) | |
Asian | 20 (0.7) | 1 (0.2) | 0 (0) | |
Others | 26 (0.9) | 2 (0.5) | 1 (0.3) | |
Disease duration, mean ( s.d.) , months | 109.9 (81.3) a,b | 167.1 (98.9) c | 154.9 (100.2) c | <0.001 |
Damage, n (%) | ||||
Ocular | 171 (5.8) a,b | 76 (18.3) c | 64 (21.9) c | <0.001 |
Neuropsychiatric | 123 (4.2) a,b | 47 (11.3) b,c | 51 (17.5) a,b | <0.001 |
Renal | 132 (4.5) a,b | 36 (8.7) b,c | 57 (19.5) a,b | <0.001 |
Pulmonary | 58 (2.0) a,b | 33 (8.0) b,c | 41 (14.0) a,b | <0.001 |
Cardiovascular | 0 (0) b | 0 (0) b | 292 (100) a,b | <0.001 |
Peripheral vascular | 88 (3.0) a,b | 37 (8.9) c | 38 (13.0) c | <0.001 |
Gastrointestinal | 44 (1.5) a,b | 15 (3.6) c | 12 (4.1) c | <0.001 |
Musculoskeletal | 0 (0) a,b | 415 (100) b,c | 89 (30.5) a,c | <0.001 |
Cutaneous | 56 (1.9) a,b | 35 (8.4) c | 33 (11.3) c | <0.001 |
Diabetes | 56 (1.9) b | 12 (2.9) b | 20 (6.8) a,c | <0.001 |
Malignancy | 114 (3.9) a | 38 (9.2) c | 18 (6.2) | <0.001 |
Premature gonadal failure | 48 (1.6) a,b | 28 (6.7) c | 16 (5.5) c | <0.001 |
SDI score, mean ( s.d.) | 0.7 (1.1) a,b | 2.6 (1.8) b,c | 3.8 (2.4) a,c | <0.001 |
Patients with damage, n (%) | 684 (23.2) a,b | 415 (100) c | 292 (100) c | <0.001 |
Number of SDI domains affected, mean ( s.d.) | 0.3 (0.6) a,b | 1.9 (1.0) b,c | 2.5 (1.4) a,c | <0.001 |
Death, n (%) | 102 (3.7) a,b | 45 (10.8) b,c | 60 (20.5) a,c | <0.001 |
a Significantly different from cluster 2.
b Significantly different from cluster 3.
c Significantly different from cluster 1. SDI: Systemic Lupus International Collaborating Clinics/ACR Damage Index.
Cluster 1 was the largest (80.6% of the total of patients) and included fewer individuals with damage (23.2% vs 100% in clusters 2 and 3, P < 0.001). The most frequently affected system was the ocular system (5.8%). Patients were younger at the time of SLE diagnosis compared with those in clusters 2 and 3 (P < 0.001). Patients in cluster 1 had the shortest disease duration.
All patients in cluster 2 had musculoskeletal damage, and it was most frequently associated with ocular (18.3% of patients) and neuropsychiatric (11.3%) damage. The mean values of damage in cluster 2 were higher than in cluster 1 both for number of organ systems affected and for the SDI score (P < 0.001, for both comparisons).
Cluster 3 was the smallest group (8.0% of the total). All patients in cluster 3 had cardiovascular damage and this was the only group with cardiovascular damage. It was more frequently associated with musculoskeletal (30.5% of patients), ocular (21.9%) and renal (19.5%) damage. The mean number of organ systems with damage and the mean SDI score in cluster 3 were higher than in clusters 1 and 2 (P < 0.001, for both comparisons). There was no significant difference in disease duration in patients in cluster 3 compared with patients in cluster 2.
There were a total of 207 deaths. Patients in cluster 3 had the highest mortality, while cluster 1 had the lowest mortality (P < 0.001). As expected, cardiovascular causes of death were significantly more common in cluster 3. The main causes of death in each cluster are presented in supplementary Tables S1 and 2 available at Rheumatology Online. The comparison between clusters for patients with disease duration <5 years presented the same results regarding damage involvement and mortality ( Table 4 ).
. | Cluster 1, n = 891 (90.3%) a . | Cluster 2, n = 54 (5.5%) a . | Cluster 3, n = 41 (4.2%) a . | P-values . |
---|---|---|---|---|
Age at diagnosis, mean ( s.d.) , years | 35.5 (14.5) b,c | 42.1 (19.0) d | 41.2 (18.5) d | <0.001 |
Disease duration, mean ( s.d.) , months | 27.8 (17.8) | 31.7 (18.3) | 29.7 (17.2) | 0.255 |
SDI score, mean ( s.d.) | 0.5 (0.9) b,c | 2.0 (1.4) c,d | 2.7 (2.1) b,d | <0.001 |
Patients with damage, n (%) | 138 (15.5) b,c | 54 (100) 1 | 41 (100) d | <0.001 |
Number of SDI domains affected, mean ( s.d.) | 0.2 (0.5) b,c | 1.5 (0.9) d | 1.8 (0.9) d | <0.001 |
Death, n (%) | 33 (3.7) b,c | 4 (7.4) d | 7 (17.1) d | <0.001 |
. | Cluster 1, n = 891 (90.3%) a . | Cluster 2, n = 54 (5.5%) a . | Cluster 3, n = 41 (4.2%) a . | P-values . |
---|---|---|---|---|
Age at diagnosis, mean ( s.d.) , years | 35.5 (14.5) b,c | 42.1 (19.0) d | 41.2 (18.5) d | <0.001 |
Disease duration, mean ( s.d.) , months | 27.8 (17.8) | 31.7 (18.3) | 29.7 (17.2) | 0.255 |
SDI score, mean ( s.d.) | 0.5 (0.9) b,c | 2.0 (1.4) c,d | 2.7 (2.1) b,d | <0.001 |
Patients with damage, n (%) | 138 (15.5) b,c | 54 (100) 1 | 41 (100) d | <0.001 |
Number of SDI domains affected, mean ( s.d.) | 0.2 (0.5) b,c | 1.5 (0.9) d | 1.8 (0.9) d | <0.001 |
Death, n (%) | 33 (3.7) b,c | 4 (7.4) d | 7 (17.1) d | <0.001 |
a Percentage calculated for those patients with <5 years of disease duration (n = 986 patients).
b Significantly different from cluster 2.
c Significantly different from cluster 3.
d Significantly different from cluster 1. SDI: SLICC/ACR Damage Index, n: number.
. | Cluster 1, n = 891 (90.3%) a . | Cluster 2, n = 54 (5.5%) a . | Cluster 3, n = 41 (4.2%) a . | P-values . |
---|---|---|---|---|
Age at diagnosis, mean ( s.d.) , years | 35.5 (14.5) b,c | 42.1 (19.0) d | 41.2 (18.5) d | <0.001 |
Disease duration, mean ( s.d.) , months | 27.8 (17.8) | 31.7 (18.3) | 29.7 (17.2) | 0.255 |
SDI score, mean ( s.d.) | 0.5 (0.9) b,c | 2.0 (1.4) c,d | 2.7 (2.1) b,d | <0.001 |
Patients with damage, n (%) | 138 (15.5) b,c | 54 (100) 1 | 41 (100) d | <0.001 |
Number of SDI domains affected, mean ( s.d.) | 0.2 (0.5) b,c | 1.5 (0.9) d | 1.8 (0.9) d | <0.001 |
Death, n (%) | 33 (3.7) b,c | 4 (7.4) d | 7 (17.1) d | <0.001 |
. | Cluster 1, n = 891 (90.3%) a . | Cluster 2, n = 54 (5.5%) a . | Cluster 3, n = 41 (4.2%) a . | P-values . |
---|---|---|---|---|
Age at diagnosis, mean ( s.d.) , years | 35.5 (14.5) b,c | 42.1 (19.0) d | 41.2 (18.5) d | <0.001 |
Disease duration, mean ( s.d.) , months | 27.8 (17.8) | 31.7 (18.3) | 29.7 (17.2) | 0.255 |
SDI score, mean ( s.d.) | 0.5 (0.9) b,c | 2.0 (1.4) c,d | 2.7 (2.1) b,d | <0.001 |
Patients with damage, n (%) | 138 (15.5) b,c | 54 (100) 1 | 41 (100) d | <0.001 |
Number of SDI domains affected, mean ( s.d.) | 0.2 (0.5) b,c | 1.5 (0.9) d | 1.8 (0.9) d | <0.001 |
Death, n (%) | 33 (3.7) b,c | 4 (7.4) d | 7 (17.1) d | <0.001 |
a Percentage calculated for those patients with <5 years of disease duration (n = 986 patients).
b Significantly different from cluster 2.
c Significantly different from cluster 3.
d Significantly different from cluster 1. SDI: SLICC/ACR Damage Index, n: number.
Discussion
In a large national SLE cohort, we have observed several clinical patterns of damage manifestations and their association with different levels of risk of mortality. In our study, we used one of the statistical techniques of cluster analysis to identify groups of lupus patients with similar damage patterns and to describe the clinical differences between these groups regarding damage and mortality.
The general purpose of these techniques was to detect clusters in observations (or variables) and to assign those observations to the clusters. In addition to their identification, it is equally interesting to determine how the clusters are different, that is, to determine the specific damage variables that vary and how they vary in regard to patients in different clusters. In our study, we used the k -means clustering method that, given a fixed number of clusters, assigns observations to those clusters so that the means across clusters (for all damage variables) are as different from each other as possible.
Our cluster analysis identified three distinct clusters based on damage manifestations, which we then explored for their association with mortality, both overall and for the subgroup of patients with disease duration <5 years. The first cluster (cluster 1) was formed of patients without damage or with little damage (⩽1 SDI domain per patient). It was by far the largest one, with four out of five patients belonging to this cluster. In clusters 2 and 3, there were no patients without damage. All patients in cluster 2 had musculoskeletal damage and none had cardiovascular damage. All patients in cluster 3 had cardiovascular damage, with about one-third of the total having musculoskeletal damage. We found significant differences between clusters, not only in the proportion of patients with some kind of damage, but also in the level of damage in terms of the number of SDI domains affected and the mean SDI score. To date, there is no study in the literature that groups SLE patients according to their patterns of damage. From our results, we can establish that there is a majority of patients who are virtually free of damage and who are mainly women with an earlier age at diagnosis of SLE. On the other hand, there is a small proportion (almost 10% of SLE patients) with cardiovascular damage that is frequently associated with damage in other systems, particularly the musculoskeletal, ocular and renal systems. These patients include men in a higher proportion than that in SLE patients overall, and they are diagnosed with SLE at a significantly more advanced age.
The association between subsets of patients and disease outcomes such as mortality has rarely been reported in the literature. Only one study has compared the mortality rate between subsets of patients [ 21 ]. In a large Chinese group of SLE patients, To et al. [ 21 ] identified three clinical patterns of organ manifestations that may influence the long-term prognosis of the disease. However, there is a notable difference between their results and the findings of our study because the vast majority of clinical features that they analysed were not damage manifestations but activity features.
In our study, we examined the mortality of the patients in every damage cluster. Of particular note, there were significant differences in mortality rates between them, mortality being higher in the clusters with more damage accrual. A number of studies have showed that damage measured by SDI predicts future mortality in SLE patients [ 9–15 ]. However, the impact of the different domains of the SDI on survival has hardly been evaluated [ 16 , 17 ]. Stoll et al. [ 16 ] reported that pulmonary damage at 1 year after diagnosis significantly predicted death within 10 years of diagnosis. Data from LUMINA, a multiethnic cohort, showed that the renal domain of the SDI is associated with a shorter time to death [ 17 ]. However, our current study is the first one that analyses the impact of damage clustering on survival. Our principal finding is that the presence of cardiovascular damage, which we observed in cluster 3, is the most important feature in defining this highest mortality risk cluster, whose mortality rate was double that of cluster 2 and five times that of cluster 1. In particular, we observed that coronary heart disease (SDI items: angina or myocardial infarction) was present in almost 40% of our patients in cluster 3 (supplementary Table S3, available at Rheumatology Online). This finding of our current study along with our recent description of a high prevalence of premature cardiovascular disease [ 22 ] imply that developing cardiovascular damage is one of the worst prognostic features that there is for Spanish patients with SLE and that, as a consequence, it should be aggressively prevented and treated.
Nevertheless, we also found that patients in cluster 2, a cluster characterized by predominant musculoskeletal damage had a three times higher mortality rate compared with cluster 1, the latter being characterized by limited or absent damage. This is one of the most important and original findings of our study because it implies that developing musculoskeletal damage is a marker of a poor prognosis group. Among the musculoskeletal damage manifestations, we found that osteoporosis (as defined by SDI: osteoporosis with fracture or vertebral collapse, excluding avascular necrosis) was the most frequent one, being present in almost 40% of the patients in cluster 2 (supplementary Table S3, available at Rheumatology Online). As osteoporosis and bone fractures are frequently associated with corticosteroid treatment [ 23 ], the prudent use of this therapy by clinicians and the early use of measures to prevent corticosteroid-induced osteoporosis should be recommended.
Even though the disease duration in cluster 1 was relatively long, it was shorter than that in clusters 2 and 3. Because of this and the fact that incident damage occurs at a higher rate in the first few years of the disease [ 8 , 15 ], we carried out the comparison between clusters for those patients with disease duration <5 years. In this comparison, in which disease duration was similar in the three clusters, we obtained identical results regarding damage and mortality to those in the overall sample. Therefore, our study demonstrates that the association between damage and mortality already exists in the early stages of the disease. This original finding has an important clinical implication given that it implies that it is crucial to prevent and reduce the development of damage promptly, even in the very early stages of the disease.
One of the limitations of this study was its cross-sectional design—a longitudinal study can be more adequate to assess mortality over time. However, carrying out a longitudinal mortality study in a disease such as SLE, with a relatively high survival rate at 10 years [ 2 , 3 , 24 , 25 ], would require a large number of patients and a long follow-up duration that would be challenging. On the other hand, our study has several strengths: the rigorous way that the clinical data were collected to make sure that it was comprehensive; the large number of patients derived from a large number of centres across the whole country with a long follow-up duration; and the appreciable number of deaths over time. In contrast, regardless of their design, previous studies that evaluated the association between damage and mortality usually analysed relatively small samples of SLE patients in which the number of deceased patients was limited, so it has been difficult to draw significant conclusions regarding mortality [ 9–13 , 15–17 ]. Finally, the statistical method we used is likely the main strength of our study: although cardiovascular damage was thought to be important in SLE, our study has proved it using cluster analysis, a method that does not start with any preconceived ideas.
In conclusion, our study identified different patterns (clusters) of damage manifestations within a large cohort of SLE patients. Cardiovascular and musculoskeletal damage were the two dominant forms of damage used to sort these patients into clinically meaningful clusters. Our study suggested that both cardiovascular and musculoskeletal damage manifestations have significant association with an increased risk of mortality, even in the early stages of the disease. Therefore, physicians should pay special attention right from diagnosis to the prevention of damage in these two systems.
We thank Juan Manuel Barrio and our colleagues in the Research Unit of the Spanish Society of Rheumatology who helped establish the database and provided excellent statistical support.
The authors would like to thank all the investigators who participated in this study: Inmaculada de la Torre Ortega, Luis Carreño Pérez, Patricia Carreira Delgado, Esther Rodríguez-Almaraz, Juan Antonio Martínez López, Olga Sánchez Pernaute, Txaro García de Vicuña Pinedo, Marta Valero Expósito, Paloma García de la Peña, Silvia Rodríguez Rubio, Jorge J. González Martín, Ana Pérez Gómez, Cristina Bohórquez, Atusa Morasat Hajkhan, Ana I. Turrión Nieves, José Luis Andreu Sánchez, Aline Lucice Boteanu, M. Luz Gamir Gamir and Patricia Richi Alberti (Madrid); José Carlos Rosas-Gómez de Salazar, Esteban Salas Heredia, Carlos Santos Ramírez and José M. Senabre Gallego (Villajoyosa Alicante); Paloma Vela-Casasempere and José Antonio Bernal (Alicante); Inmaculada Ros Vilamajó, Antonio Juan Mas and Claudia Murillo (Palma de Mallorca); Emma García Melchor (Badalona); María García Manrique, Carlos Galisteo Lencastre and Mireia Moreno Martínez-Losa (Sabadell); Raúl Menor Almagro (Jerez de la Frontera); Víctor Martínez Taboada, Miguel A. González-Gay Mantecón, Inés Pérez Martíny and M. del Carmen Bejerano (Santander); Ignacio Villa Blanco, Begoña Moreira, Elena Aurrecoechea and Teresa Ruiz Jimeno (Torrelavega); Ángeles Aguirre Zamorano (Córdoba); César Magro (Granada); M. Ángeles Acosta Mérida (Las Palmas de Gran Canaria); Cesar A. Egües Dubuc and Jorge Cancio Fanlo (San Sebastián); Carlos Vitovi and Alejandra López Robles (León); Rafael Benito Melero-González (Orense); M. Victoria Irigoyen Oyarzábal, M. Ángeles Belmonte López and Carmen M. Romero Barco (Málaga); María Rosario Oliva (Murcia); Claudia Stoye and María Concepción Fito-Manteca (Pamplona); Íñigo Hernández Rodríguez, Coral Mouriño Rodríguez and Bruno de Aspe de la Iglesia (Vigo); Ruth López González (Salamanca); Federico Navarro Sarabia, Francisco J. Toyos Sáenz de Miera, Julia Uceda Montañés, Raquel Hernández Sánchez and Rosalía Martínez Pérez (Sevilla); Beatriz Rodríguez Lozano (Tenerife); Eduardo Úcar Angulo, Olaia Fernández Berrizbeitia and Luis López Domínguez (Bilbao); Isabel de La Morena Barrio and Elia Valls (Valencia); Javier Manero Ruiz (Zaragoza); Sergio Machín and Javier Nóvoa (Gran Canaria).
The RELESSER Registry was supported by the Spanish Society of Rheumatology. Grants from GlaxoSmithKline, Roche, Union Chimique Belge and Novartis contributed to the support of the cross-sectional stage of the Registry. These study sponsors were not involved in the study design, in the collection, analysis and interpretation of data, in the writing of the report or in the decision to submit the paper for publication.
Acknowledgements
Dr J.M.P-R. is supported by the grant 316265 (BIOCAPS) from the European Union 7th Framework Programme (FP7/REGPOT-2012-2013.1) and FIS/ISCIII-Fondo Europeo de Desarrollo regional (FEDER) (Grant number PI11/02857).
Funding : No specific funding was received from any funding bodies in the public, commercial or not-for-profit sectors to carry out the work described in this manuscript.
Disclosure statement : F.J.L.-L. has received speaker fees from Abbvie, Roche Farma, BMS, Pfizer, UCB, MSD, Actelion and GST and has received research funding from Abbvie and GSK. All other authors have declared no conflicts of interest.
Supplementary data
Supplementary data are available at Rheumatology Online.
References
Comments