Abstract
Objective. To analyze survival, prognostic factors, and causes of death in a large cohort of patients with systemic sclerosis (SSc).
Methods. From 1991 to 2010, 947 patients with SSc were treated at 2 referral university centers in Brazil. Causes of death were considered SSc-related and non-SSc-related. Multiple logistic regression analysis was used to identify prognostic factors. Survival at 5 and 10 years was estimated using the Kaplan-Meier method.
Results. One hundred sixty-eight patients died during the followup. Among the 110 deaths considered related to SSc, there was predominance of lung (48.1%) and heart (24.5%) involvement. Most of the 58 deaths not related to SSc were caused by infection, cardiovascular or cerebrovascular disease, and cancer. Male sex, modified Rodnan skin score (mRSS) > 20, osteoarticular involvement, lung involvement, and renal crisis were the main prognostic factors associated to death. Overall survival rate was 90% for 5 years and 84% for 10 years. Patients presented worse prognosis if they had diffuse SSc (85% vs 92% at 5 yrs, respectively, and 77% vs 87% at 10 yrs, compared to limited SSc), male sex (77% vs 90% at 5 yrs and 64% vs 86% at 10 yrs, compared to female sex), and mRSS > 20 (83% vs 90% at 5 yrs and 66% vs 86% at 10 yrs, compared to mRSS < 20).
Conclusion. Survival was worse in male patients with diffuse SSc, and lung and heart involvement represented the main causes of death in this South American series of patients with SSc.
Systemic sclerosis (SSc) is a diffuse connective tissue disease characterized predominantly by fibrosis and vascular abnormalities affecting skin and internal organs. Many studies have examined the causes of death and the predictors of poor outcome in SSc1. Steen and Medsger, analyzing the Pittsburgh SSc database from 1972–2002, found a significant change in the causes of death in SSc2. The frequency of deaths due to scleroderma renal crisis (SRC) decreased over the 30-year period, from 42% to 6% of the SSc-related deaths; whereas lung involvement (pulmonary fibrosis and pulmonary hypertension) is now the main cause of SSc-related deaths2. Another study from the same group indicated that the severity of the disease predicted early death; in patients with severe skin (modified Rodnan skin score > 40), lung (forced vital capacity < 55% predicted), heart (severe arrhythmia or congestive heart failure), gastrointestinal (GI; malabsorption, pseudo-obstruction), or kidney involvement, the 9-year cumulative survival was 38%, while it was 78% in patients who had mild organ involvement3. A recent study analyzing causes of death and risk factors in a large worldwide series of 5860 patients with SSc prospectively followed in the European League Against Rheumatism Scleroderma Trials and Research (EUSTAR) cohort found that 55% of the 234 deaths were attributed directly to SSc [35% to pulmonary fibrosis, 26% to pulmonary arterial hypertension (PAH) and 26% to cardiac causes], and 41% to non-SSc causes (mainly infections, malignancies, and cardiovascular causes). Independent risk factors for mortality included proteinuria, the presence of PAH based on echocardiography, forced vital capacity below 80% of normal, diffusing capacity of the lung, patient age at the onset of Raynaud’s phenomenon, and the modified Rodnan skin score4.
Epidemiologic variables associated with a poor prognosis are male sex5,6,7,8 and older age at onset5,8,9,10,11,12,13,14,15,16. Among the clinical variables associated with a worse prognosis are diffuse skin involvement6,7,12,13,16,17 and internal organ involvement affecting lung6,7,8,9,10,11,12,13,15,18,19,20,21,22,23,24, kidney7,8,11,12,14,16,17,20,21, heart7,11,12,14,17,18,20,21,22,23,24, and GI tract8. Laboratory variables include anemia10,13, increased erythrocyte sedimentation rate (ESR)7,13,16,18,19, abnormal urine sediment5,18,19, hypergammaglobulinemia5, and anti-Scl-70 antibody20 as factors of poor prognosis.
The series of Tuffanelli and Winkelmann, analyzing 727 patients with SSc, showed a 5-year survival of 70% and a 10-year survival of 59%25. This survival gradually improved in the last decades; the best results were associated with a lower percentage of diffuse SSc (dcSSc) and severe organ involvement in the different series. Many studies analyzed longterm survival in SSc. Al-Dhaher, et al, in a cohort of 185 patients with SSc in Canada, found a 5-year survival rate of 90% and a 10-year survival rate of 82%26. Joven, et al observed survival rates of 85%, 75%, and 55% at 5, 10, and 20 years27 in 204 patients in Spain. Nihtyanova, et al, analyzing 520 patients with SSc from the Royal Free Hospital in London, showed that the 5-year survival among patients with dcSSC improved from 69% in a historical cohort (1990–93) to 84% in a contemporary cohort (2000–03), whereas 5-year survival among patients with limited SSc (lcSSc) remained unchanged (93% and 91%, respectively)28. Hashimoto, et al, analyzing 405 Japanese patients with SSc 1973–2009, found an overall 10-year survival rate of 88%, associated to a significantly higher mortality in the SSc population compared to the general population (standardized mortality ratio 2.76)29. A study from Taiwan, analyzing 1479 patients with SSc 2003–07, found a 5-year survival rate of 82.3%; the main risk factors for death were male sex, older age, diagnosis of cancer, and endstage renal disease30.
A recent systematic review and metaanalysis of cohort studies published from January 1960 to June 2010 showed that SSc is still a very severe disease, reflected by a standardized mortality ratio of 3.5; heart and lung involvement were the main causes of death in a total of 2691 patients31.
The objective of our study was to report and analyze survival, causes of death, and prognostic factors in a Brazilian series of 947 patients followed at 2 referral university centers.
MATERIALS AND METHODS
Nine hundred forty-seven consecutive patients with SSc treated at the SSc outpatient clinics of the University of São Paulo and University of Campinas between January 1991 and December 2010 were analyzed. These patients lived predominantly in an industrial region of about 20 million inhabitants in the State of São Paulo (southeast Brazil). The 2 universities used the same protocol of investigation for patients with SSc.
All patients fulfilled the American College of Rheumatology classification criteria for SSc32 and were divided into the subtypes dcSSc and lcSSc, according to LeRoy, et al33. Patients were also diagnosed as having SSc sine scleroderma according to established criteria34.
All patients had appointments at least once a year, depending on the severity of their disease. Patients who did not attend a scheduled medical consultation for > 1 year were contacted.
For the patients who did not die at the university hospitals of the University of São Paulo or Campinas, death certificates were requested from the families; if necessary, attending physicians at the time of death were contacted about its specific causes.
Each death was classified as primarily related to SSc or not. The SSc-related causes of death were (1) lung involvement such as interstitial lung disease (ILD), determined by radiograph, high-resolution computed tomography, pulmonary function tests, or hypoxemia requiring oxygen supplementation; (2) lung involvement such as pulmonary hypertension (PH), defined by right heart catheterization with mean pulmonary arterial pressure > 25 mm Hg, or by Doppler echocardiogram with systolic pulmonary artery pressure > 40 mm Hg, not associated to pulmonary fibrosis; (3) ILD + PH: patients with ILD who also presented PH in the followup, and both involvements were considered associated to death; (4) heart involvement such as congestive cardiac failure, arrhythmias, or conduction defects (not attributable to other cardiac conditions) requiring treatment; (5) renal: deaths during the SRC or related to dialysis; (6) GI: severe involvement of the esophagus or small bowel, resulting in malabsorption or malnutrition; and (7) multiorgan failure: severe simultaneous damage from > 1 SSc-related organ involvement, in which it was difficult to determine the primary organ that was the cause of death.
The non-SSc-related causes of death included (1) infection: septicemia as the main cause of death, independent of SSc activity; (2) cardiovascular or cerebrovascular disease: patients who died of myocardial infarction or stroke, or of complications from chronic non-scleroderma atherosclerotic heart disease; (3) cancer: patients who died of cancer, independent of severe organ involvement; (4) sudden death: a sudden death without prior SSc heart disease and without obvious etiology; and (5) other: a variety of common causes of death not related to SSc.
Statistical analysis
Pearson chi-square test (categorical variables), t-test (numeric variables), and log-rank test (survival curve) were used to identify the prognostic factors. Multivariate logistic regression analysis was performed to define predictive factors of death. Survival at 5 and 10 years was estimated using Kaplan-Meier method.
RESULTS
According to the clinical variant, 533 patients (56.4%) presented with limited SSc, 294 (31%) dcSSc, 79 (8.3%) SSc sine scleroderma, and 41 (4.3%) an overlap syndrome. There were 838 (88.5%) female and 109 (11.5%) male patients, as well as 712 (75.2%) whites and 235 (24.8%) African Brazilians (black patients of unmixed ancestry and mulattos, i.e., originating from the mixture of white and black individuals).
Demographic and clinical variables related to the 947 patients with SSc are presented in Table 1.
Clinical, demographic, and immunological data in 947 patients. Data are number (%) unless otherwise indicated.
One hundred sixty-eight patients (17.7% of the whole group) died during the followup; 110 deaths (65.5%) were considered related to SSc. Among those deaths, we found a predominance of lung involvement (48.1%), heart involvement (24.5%), renal crisis (10.9%), and multiorgan failure (9.1%; Table 2). Among the deaths related to lung involvement, ILD and PH were the causes in 16.3% each; in 15.5% of the cases, both conditions were present and could be considered associated to death. The 5 deaths caused by GI involvement were related to esophageal stenosis (4 patients) and malabsorption syndrome (1 patient). Three other patients died from unknown causes: 2 patients (1 with lcSSc and the other with dcSSc) using intravenous cyclophosphamide were hospitalized with acute severe bleeding and died in 3 days; another patient with dcSSc presented severe pancytopenia while taking azathioprine (with a history of intravenous cyclophosphamide) and died in 5 days; none of these patients were submitted to necroscopic study.
Causes of death. Data are number.
Most of the 58 deaths not related to SSc were associated with infection (septicemia, independent of SSc activity), cardiovascular and cerebrovascular disease (4 with congestive heart failure, 2 with myocardial infarction, 1 with arrhythmia, and 1 with stroke), and cancer (lung in 3 patients, esophagus, colon, stomach, ovary, and breast in 1 patient each; Table 2). Twelve patients (20.7%) with active disease presented sudden death not related to a specific organ involvement; none of these patients were submitted to necroscopic study; and 6 cases presented other causes: autoimmune liver cirrhosis in 2 patients, tuberculosis in another 2 patients (1 with lung tuberculosis and the other with neurotuberculosis), amyloidosis, and polymyositis.
Multivariate logistic regression showed that the variables associated to death were male sex (p = 0.003, OR 2.35, 95% CI 1.34−4.12), modified Rodnan skin score (mRSS) > 20 (p = 0.016, OR 1.71, 95% CI 1.11−1.65), osteoarticular involvement (p < 0.001, OR 4.38, 95% CI 2.72−7.07), lung involvement (p < 0.001, OR 4.2, 95% CI 2.7–6.5), and renal crisis (p < 0.001, OR 9.96, 95% CI 3.74–26.54).
Overall survival rate was 90% for 5 years and 84% for 10 years (Figure 1). Clinical variants, mRSS > 20, and sex showed significant differences regarding survival. Patients with dcSSc presented worse survival than lcSSc (85% vs 92% at 5 years, respectively, and 77% vs 87% at 10 years; p < 0.001; Figure 2). Patients with SSc sine scleroderma presented a 5-year survival of 89%, similar to lcSSc; because patients with SSc sine scleroderma were diagnosed after 2000, we do not have their 10-year survival rates. Male sex was associated with worse survival compared to female sex (77% vs 90% at 5 years and 64% vs 86% at 10 years; p < 0.001; Figure 3). Modified RSS > 20 (83% vs 90% at 5 years and 66% vs 86% at 10 years, compared to mRSS < 20; p < 0.001) also presented worse prognosis (Figure 4). White patients presented no significant statistical differences in survival when compared to the African Brazilians (90% vs 89% at 5 years and 84% vs 81% at 10 years; p = 0.786; Figure 5).
Overall survival.
Clinical variants of survival.
Sex differences and survival.
Survival and Rodnan skin score.
Survival and race.
DISCUSSION
Although clinical outcomes appear to have improved in recent years, SSc continues to cause substantial mortality. Our study confirmed the previous observations showing that lung involvement is now the most frequent cause of death in SSc2,3,4. In this series of 947 patients with SSc, treated at the 2 largest SSc centers in Brazil, 48.1% of the deaths were related to lung involvement. ILD and PH were each found in 16.3% of the patients. In 15.5% of the patients (17 cases) there was ILD (affecting at least 20% of the lung area) associated to PH (pulmonary arterial systolic pressure by echocardiogram > 40 mm Hg); as both conditions could be considered associated to death, we decided to specify ILD + PH as the causes of death. A study from the Johns Hopkins SSc cohort has recently shown that ILD associated to PH has a significantly worse prognosis compared to isolated PAH (1-, 2-, and 3-year survival rates 82%, 46%, and 39% vs 87%, 79%, and 64%, respectively; p < 0.01); in a multivariable analysis, ILD-associated PH was associated with a 5-fold increase in risk of death compared with PAH35.
Steen and Medsger observed that 33% of the SSc-related deaths were related to ILD, as well as that PH, independent of ILD, also represented an important cause of death in SSc2. Hachulla, et al, analyzing the risk factors for death in a series of 546 French patients with SSc, found that 32.2% of the deaths were associated to PH, justifying the yearly echocardiographic screening24. The analysis of the large EUSTAR databank showed that lung involvement caused 61% of the deaths, considering 35% for ILD and 26% for PAH4. A US study showed that pulmonary fibrosis was the major predictor of poor hospitalization outcomes in patients with SSc36. Nihtyanova, et al demonstrated that the significant improvement in survival in the last decade in patients with dcSSc was associated with better and more complete ascertainment of lung complications as a result of systematic annual screening28. Similar results emphasizing lung involvement as the most frequent cause of death were also observed in SSc series in populations of different ethnicity, published in the last decade in Spain15, Hungary16, and Thailand14.
Heart involvement, especially congestive heart failure and cardiac arrhythmia, represented an important cause of death in our study, causing 24.5% of the SSc-related deaths, in a similar percentage (26%) to that observed in the EUSTAR databank4. An Italian study has also emphasized the importance of heart involvement as a significant cause of death in SSc37. Steen and Medsger observed no significant changes in heart-related deaths in SSc from 1972 to 20022. Recent studies have emphasized the importance of cardiopulmonary manifestations as prognostic factors in SSc, in adult24 and juvenile23 patients. Cardiovascular disease also represents an important cause of death unrelated to SSc.
Renal involvement, especially SRC, was the leading cause of death in SSc for a long time, until the advent of the angiotensin-converting enzyme inhibitor captopril2. In the last 3 decades, there was a progressive decrease in the percentage of patients with SRC in SSc. In our series, SRC represented the cause of death in 10.9% of the SSc-related deaths. Corticosteroids have been implicated in the exacerbation of SRC38. In our series, only 3 of the 12 patients with SSc who died of SRC had been taking prednisone > 15 mg daily (for treatment of symptomatic pericarditis) in the 3 months before the onset of SRC. Another important cause of death was multiorgan failure (9.1%); some of these patients had dcSSc with SRC and heart and/or pulmonary insufficiency. In fact, visceral involvement (lung, heart, kidney, GI tract, and multiorgan failure) was the main factor responsible for 97.3% of the SSc-related deaths in our study, emphasizing the need to search for the most frequent visceral manifestations systematically.
Although esophageal involvement is present in 60% to 90% of patients with SSc, it is not a frequent cause of death4. Because the high frequency of esophageal involvement makes esophageal examination mandatory in patients with SSc, and early use of proton pump inhibitors and prokinetic drugs can be indicated, the progression to severe stenosis and Barrett esophagus, with subsequent malnutrition and death, is probably becoming rarer. Barrett esophagus can represent a risk factor for esophageal adenocarcinoma in SSc39. Intestinal involvement is not frequent, and the repeated use of antibiotics in cases of severe diarrhea improved survival and quality of life in these patients with SSc. Involvement of the GI tract represented 4.6% of SSc-related deaths in our study.
Infection is another important cause of morbidity and mortality in SSc40. The increased risk of infections is the result of immune abnormalities and manifestations affecting vital organs associated with the SSc itself and with treatment, particularly in those receiving corticosteroids and immunosuppressive therapy. We considered infection a non-related cause of death, independent of SSc activity.
The occurrence of cancer in the context of SSc is predominantly associated with ILD41. A recent nationwide population-based cohort study in Denmark, analyzing 2040 patients with SSc for a median followup of 6.4 years, found that SSc was a risk factor for cancer, especially in men [standardized incidence ratio (SIR) 2.2] and particularly smoking- and alcohol-related cancers. These included hematological cancers (SIR 2.5), lung cancer (SIR 1.6), and immune-related cancers (SIR 1.4)42. Cancer was responsible for 13.8% of the non-SSc-related deaths in this series, affecting 8 patients. Six different sites (lung, esophagus, stomach, colon, ovary, and breast) were observed; in 2 patients (1 with lung and another with ovary cancer), the neoplasia developed a few years after the use of cyclophosphamide. In our study, only 2 patients with active SSc died of cancer (1 with esophageal and 1 with lung cancer).
Specific autoantibodies such as anti-Scl-70 (with ILD) and anti-RNA-polymerase III (with SRC) are also important prognostic markers for SSc43. Various other serum markers are currently being studied. Genetic factors are also prognostic predictors of mortality in SSc. One study, analyzing the HLA class II genotyping in a group of 250 patients with early SSc, found that HLA alleles DRB1*0802 and DQA1*0501 were significant predictors of mortality44. Our group has recently found significant good prognosis related to HLA-G in SSc45.
The overall survival rate in our study was 90% for 5 years and 84% for 10 years. These rates are compatible with studies that have shown an improvement in SSc survival in the last decades2,3,4,26,27,28,29,30. The worse results in patients with dcSSc and mRSS > 20 were expected. Patients with SSc sine scleroderma present survival rates similar to patients with lcSSc; it is not clear whether SSc sine scleroderma is a specific clinical variant of SSc or an initial stage of lcSSc.
Studies have shown that clinical and serologic features can be strongly influenced by ethnic background in patients with SSc46,47,48,49. African Americans were more likely to present diffuse skin involvement, digital ulcers, and impaired lung function, associated to anti-RNP (anti-U1-RNP and anti-U3-RNP/fibrilarin) compared to North American whites47,48. Our study did not show statistical difference in survival rates between Brazilian whites and African Brazilians. The main reason for this finding could be that the African Brazilian population is predominantly mixed; in the demographic census of 2010 in Brazil, 47.7% of the people declared themselves as whites (predominant European ancestry, with Portuguese, Spanish, or Italian origin), 43.6% mestizos (most mixed white/black), 7.6% black (pure origin), 0.7% Asian ancestry (predominantly Japanese), and 0.4% Indians.
Male sex was a significant factor of worse prognosis in our study, with a survival rate of 77% in 5 years and 64% in 10 years. These results confirm that male sex, although affected by SSc in a lower percentage compared to females, represents a factor of bad prognosis5,6,7,8.
Our study showed that lung and heart involvement are the worst prognostic factors and the main causes of death in this large South American series of patients with SSc, and that survival rates were worse in male patients with dcSSc.
Footnotes
-
Dr. Sampaio-Barros received a research grant from Federico Foundation, Switzerland.
- Accepted for publication June 25, 2012.
