Objective To determine whether the pattern of skin involvement can predict clinical features, risk of restrictive lung disease (RLD) and survival in a large scleroderma (SSc) cohort.
Methods Demographic and clinical data collected over 30 years from 2205 patients with SSc were retrospectively analysed after subdividing subjects into four subtypes based on pattern of skin fibrosis: type 0 (no skin involvement), type 1 (limited to metacarpophalangeal joints), type 2 (distal to elbows/knees) and type 3 (proximal to elbows/knees). Clinical features associated with skin subsets were identified by regression analyses. Kaplan–Meier and Cox proportional hazards models were used to compare time to RLD and survival across subtypes.
Results The presence and severity of RLD were positively associated with skin subtype (p<0.001). RLD prevalence incrementally ranged from 51.9% in type 0 to 76.7% in type 3 (p<0.001). Type 2 SSc exhibited a distinct phenotype with intermediate risk for RLD relative to type 1 (higher, p<0.001) and type 3 (lower, p<0.001) and a unique autoantibody profile, with a prevalence of anticentromere antibodies lower than type 1 (28.9% vs 44.1%, p=0.001) and of anti-topoisomerase I antibodies similar to type 3 (32.8% vs 28.7%, p=0.38). These autoantibodies were also found to be significant negative (OR=0.33, p<0.001) and positive (OR=1.6, p=0.01) predictors of RLD risk, respectively. Mortality was also intermediate in type 2 patients relative to type 3 (p=0.0003) and type 1 (p=0.066).
Conclusions These data suggest that the current classification subdividing SSc into limited and diffuse cutaneous subtypes misclassifies an intermediate group of patients exhibiting unique autoantibody profile, disease course and clinical outcomes.
- Systemic Sclerosis
- Pulmonary Fibrosis
Statistics from Altmetric.com
Systemic sclerosis (scleroderma; SSc) is a chronic multisystem disease characterised by vascular dysfunction, immune activation and tissue fibrosis.1 These pathogenetic processes contribute to a broad spectrum of clinical phenotypes often associated with substantial morbidity and mortality.2 The name scleroderma derives from the Greek ‘scleros’ and ‘derma’ which literally mean ‘hard skin’. As a common and early manifestation of SSc, cutaneous fibrosis has been the major focus of diagnostic and classification criteria.
In 1980, preliminary criteria were defined to distinguish scleroderma from other diseases.3 Subsequently, numerous proposals for classification systems that divide scleroderma into disease subtypes based on the extent of skin involvement have been published.4–7 Studies comparing their clinical utility have generated conflicting results, often due to a relatively low number of patients investigated.8–11 For this reason, a dichotomous classification system has prevailed.12 ,13
Patients exhibiting fibrosis distal to the elbows and knees are included in the limited cutaneous disease (lcSSc) subset.6 In this group, a high prevalence of anticentromere antibodies (ACA) and an increased risk of developing pulmonary arterial hypertension are recognised.14 ,15 In contrast, more extensive skin fibrosis involving the proximal portion of the limbs or the trunk is classified as diffuse cutaneous scleroderma (dcSSc). Heterogeneous phenotypes within this subset can be further distinguished by autoantibody status. For example, patients with dcSSc with anti-Scl-70 antibodies have an increased risk of restrictive lung disease (RLD), while those with anti-RNA-polymerase III antibodies have an increased risk of scleroderma renal crisis, but develop RLD less often.14 ,16 Although the limited and diffuse classification system may be useful for research purposes, it does not fully account for the heterogeneous nature of SSc skin involvement. In particular, careful characterisation of patients with SSc suggests that at least two other subsets are consistently seen: an ‘intermediate’ subtype (skin fibrosis involving forearms and legs but not the trunk) and the ‘sine-scleroderma’ group with no skin fibrosis but typical visceral manifestations.17 ,18 These subtypes are usually ‘merged’ into the lcSSc group, but experientially present with distinct serological patterns and different clinical outcomes.19
Lung involvement is the leading cause of morbidity and mortality in SSc, with a prevalence ranging from 30% to 90% of patients.2 ,16 ,20 Although it is more common in dcSSc, no reliable association has been found between RLD and the extent or severity of skin involvement.21 ,22 In addition, the risk of lung disease and related outcomes in subjects with ‘intermediate’ skin phenotype has not been defined.
In this study, we sought to determine in a large single-centre cohort whether subdividing patients into four subtypes based on the pattern of skin involvement, in combination with their autoantibody status, could reliably stratify the risk of distinct clinical manifestations—in particular, the development and severity of interstitial lung disease and survival.
Patients and methods
Patients evaluated at the Johns Hopkins Scleroderma Center from December 1976 to March 2010 were enrolled into the longitudinal cohort after providing written informed consent if they met American College of Rheumatology criteria for SSc, or had at least three of five features of CREST syndrome (calcinosis, Raynaud's phenomenon (RP), oesophageal dysmotility, sclerodactyly, telangectasias), or had definite RP, abnormal scleroderma-pattern nailfold capillaries and the presence of an SSc-specific autoantibody (anti-Scl-70, ACA or anti-RNA-polymerase III).3 Of a total of 2348 patients, 100 were excluded from the analysis because key data were missing (ie, lung function tests, skin assessment and at least one follow-up visit after enrolment) and 43 because they had developed SSc before the age of 16. This study was approved by the Johns Hopkins institutional review board.
Demographic and clinical data including age, sex, ethnicity, smoking status, disease duration at first visit (defined from onset of the first non-RP symptom), scleroderma subtype, specific organ involvement and autoantibody status were recorded at the initial and follow-up visits. The severity of skin involvement was quantified using the modified Rodnan skin score (mRSS).23 The pattern of skin involvement was defined into four subtypes: type 0 if no detected cutaneous sclerosis; type 1 with sclerosis distal to the metacarpophalangeal joints, with or without face involvement; type 2 with skin changes more proximal but distal to elbows or knees and no trunk involvement; type 3 with sclerosis extending proximally to the elbows or knees and trunk. Patients were classified by the maximum extent of cutaneous involvement assessed during their clinical visits or as indicated by previous records. A patient's disease subtype designation was not changed if the skin involvement improved subsequently.
The presence and severity of RP, gastrointestinal, heart and renal organ involvement was assessed using the Medsger severity scale (MSS).24 Severe organ disease was defined as MSS ≥3. Muscle involvement was confirmed by raised muscle enzymes, abnormal electromyogram or biopsy-proved inflammation. Joint disease was defined by detection of synovitis, effusions, arthralgias or tendon friction rubs. The presence of sicca symptoms was identified by clinical criteria.25 Lung involvement was determined based on abnormal pulmonary function tests (PFTs). Measurements of forced vital capacity (FVC) and single-breath carbon monoxide transfer factor (TLCO) were calculated as a percentage of those predicted for an average person of the same height, weight, sex, age, race, and were standardised.26–28 The presence of RLD was defined by an FVC<80% of predicted, with severity categorised as normal (FVC>80%), mild (FVC 70–80%), moderate (FVC 50–69%), or severe (FVC<50%). For this study, echocardiographic evidence of pulmonary hypertension (ECHO-PH) was determined based on estimated right ventricular systolic pressure by Doppler echocardiography ≥45 mm Hg with no clinical evidence of congestive heart failure or thromboembolic disease.29
Patient demographics and clinical characteristics were compared across scleroderma subtypes, with p values calculated using χ2 test for binomial or categorical variables and analysis of variance for continuous variables. Pairwise comparisons were performed if statistically significant differences among SSc subtypes were present and a Bonferroni correction was used to determine significance. For each scleroderma subtype, Kaplan–Meier curves were calculated to visualise the time from SSc onset to development of RLD. Pairs of survival curves were tested for equality using the log rank test. The same analysis was repeated to compare subsets of patients defined by their autoantibody profile. Simple logistic regression was used to estimate the association between different skin subsets and the development of RLD. Multiple logistic regressions adjusting for sex, age at onset, disease duration (from SSc onset to the last visit or death), race, smoking status, muscle disease, anti-Scl-70 and ACA antibody were conducted to further assess the strength of this association.
The cumulative incidence of mortality was estimated by Kaplan–Meier analysis, and a log rank test was used to compare the incidence of death based on SSc skin subset and antibody status. The independent contribution of skin subtype to mortality was examined using Cox proportional hazards analyses, with adjustment for sex, age at scleroderma onset, smoking status and race. Reported p values are two sided with α=0.05. Statistical analyses were performed using Stata V.10.0 (Stata Corporation, College Station, Texas, USA).
Entry criteria were satisfied by 2205 patients. Table 1 summarises their sociodemographic and disease characteristics. The proportion of men was highest in the subtype with greater skin involvement (p<0.001). The mean age at onset was similar in the different subtypes, while disease duration at the first visit was significantly shorter in patients with type 3 SSc. African Americans exhibited type 3 SSc (205/367, 56%) more often than white subjects (561/1652, 34%) (p<0.001). Greater skin involvement was positively associated with increased prevalence of severe gastrointestinal disease (p=0.007), muscle inflammation (p<0.001) and joint involvement (p<0.001), especially with tendon friction rubs (p<0.001). Type 0 patients exhibited a pattern of disease manifestations and organ involvement similar to type 1, but overall with milder expression. Type 2 patients showed a unique clinical phenotype characterised by higher prevalence of calcinosis, sicca symptoms, more severe RP and heart disease. As expected, type 1 patients showed the highest prevalence of ECHO-PH (33%, p=0.005), followed by type 2 (29.9%) and type 3 (24.6%).
Skin involvement and lung disease in SSc
The presence of RLD was positively associated with the pattern of cutaneous involvement, with prevalence increasing by about 8% through each subsequent SSc skin subtype (p<0.001) (table 1). Accordingly, mean minimum TLCO and FVC were lower with increasing category of skin disease (p<0.001). The proportion of patients exhibiting established RLD at the time of their first available PFT incrementally ranged from 40% in type 0 group to 61% in type 3 (p<0.001) (figure 1A). About 41.8% of patients had their first PFT within 6 months of diagnosis and the average time from diagnosis to first PFT was 3.1±6.1 years. Analysis of RLD severity by SSc skin subtype showed that patients in higher categories of skin involvement more often developed moderate to severe RLD (p<0.001) (figure 1B).
The risk of developing RLD from the time of scleroderma onset is analysed in figure 2. Patients with type 2 SSc showed an intermediate and statistically distinct incidence of RLD over time compared with type 1 (p<0.0001) and type 3 (p<0.0001). We then evaluated the relevance of other variables potentially associated with an increased odds of developing lung disease (table 2). In unadjusted analyses, skin subtype was significantly associated with RLD risk. Relative to type 2, the odds of developing RLD for type 0 and type 1 patients were significantly reduced (OR=0.5, p=0.009 and OR=0.7, p=0.02 respectively), while for type 3 were significantly higher (OR=1.5, p=0.018). Among all predictors tested, increase risk for RLD was positively associated with anti-Scl-70 antibodies (OR=2.2, p<0.001) and muscle disease (OR=2.8, p<0.001), while ACA exhibited a protective effect (OR=0.27, p<0.001). After adjusting for the main demographic features (table 2), the regression model confirmed a lower risk for RLD in type 1 and type 0 relative to type 2 patients (OR=0.6, p=0.01 and OR=0.5, p=0.009), while no difference was detected between type 2 and type 3 (OR=1.3, p=0.2).
To determine how the presence of SSc-specific autoantibodies might influence the association between skin subtype and RLD, the multivariate regression model was further adjusted for anti-Scl-70 or ACA separately. After adjusting for anti-Scl-70, the risk for RLD was still reduced for type 1 patients relative to type 2 (OR=0.6, p=0.045), but not for type 0 compared with type 2 (OR=0.9, p=0.8). When controlling for ACA status, no difference for RLD risk was found among skin subtypes. When all significant predictors were included in the model, skin subtype was not significantly associated with the development of RLD. In contrast, anti-Scl-70 positivity was associated with 60% increased odds of developing RLD (p=0.01) and ACA with 70% reduction (p<0.001).
Autoantibody associations with scleroderma subtypes and RLD
We next analysed the autoantibody profile of the different skin subsets (figure 3A). The prevalence of ACA was significantly lower in type 2 than in type 1 patients (28.9% vs 44.1%, p=0.001), while that of anti-Scl-70 was higher (17% vs 32.8%, p<0.001). Interestingly, type 2 patients also differed significantly from type 3 in ACA prevalence (28.9% vs 4.4%, p<0.001), but not anti-Scl-70 (32.8% vs 28.7%, p=0.38).
To further determine the relevance of SSc-related autoantibodies as predictors of RLD, we stratified patients with SSc by autoantibody status (figure 3B–D). Type 0 patients were excluded from this analysis owing to their low number after stratification. Among ACA-positive patients (n=365) (figure 3B), those with type 2 SSc had an intermediate risk for RLD that was not significantly different from that of type 1 (p=0.4) or type 3 (p=0.3), although the latter had an increased RLD risk compared with type 1 (p=0.02). Among anti-Scl-70-positive patients (n=281) (figure 3C), type 1 and type 2 patients showed similar risk of RLD (p=0.2), but significantly lower than type 3 (p<0.0001 and p=0.006 respectively). Finally, in patients negative for both ACA and anti-Scl-70 (n=617) (figure 3D), the type 2 group exhibited an RLD risk similar to that of type 3 (p=0.5), which was greater than that of type 1 (p=0.018 and p<0.001, respectively). The same analyses conducted from the time of RP onset showed even stronger differences across skin subtypes (data not shown). Taken together, these data confirm that the pattern of skin involvement is an independent risk factor for RLD.
Survival differences among scleroderma subtypes
At the time of this analysis, 25% of patients were known to have died. Survival analysis by SSc subtype is shown in figure 4A. Overall mortality risk was increased in patients with type 3 SSc compared with type 2 and type 1 (p=0.0003 and p<0.0001, respectively). The survival curve for type 2 and type 1 subjects was similar during early phases of SSc, but over time type 2 patients showed an overall trend towards worse outcome (p=0.066). Surprisingly, the survival of patients with type 0 SSc was similar to that of type 3 (p=0.8).
Survival analysis by antibody status (figure 4B) showed that patients positive for ACA have significantly decreased death rates in comparison with those positive for anti-Scl-70 (p=0.0001) and with those negative for both autoantibodies (p<0.0001). The proportional hazards models (figure 4C) did not show a significant association between skin subtype and survival after autoantibodies were included. Also, anti-Scl-70 was not a significant predictor of mortality.
This large longitudinal cohort study provides evidence that the current division of scleroderma into two subtypes (limited and diffuse) may misclassify a group of patients with an intermediate degree of skin involvement, who present a unique serological profile and distinct clinical outcomes. This study also highlights the prognostic value of defining the pattern of skin involvement together with the autoantibody status to predict risk and severity for RLD as well as overall survival in SSc.
We found that type 2 patients show an ‘intermediate’ clinical phenotype compared with type 1 (milder) and type 3 (worse) for RLD prevalence, severity and time of onset. In addition, the risk of developing RLD is increased in type 2 patients compared with type 0 or type 1, suggesting that combining these groups in the lcSSc subset may not be appropriate. The detection of higher prevalence of anti-Scl-70 in patients with type 2 SSc, may partially explain the clinical features and outcomes detected in this intermediate subset.
The debate among SSc experts about which nosological system may provide the best clinical and prognostic value for patients with SSc has been relatively quiet for the past two decades.9–11 ,18 ,30–32 While some studies demonstrated the presence of an intermediate skin phenotype with unique clinical outcomes and survival rates not captured in the lcSSc/dcSSc classification, other investigations conducted on small cohorts have not shown significant differences using two or more skin subsets.11 ,18 ,33 Thus, a 1988 editorial co-signed by leading experts in the field has resulted in the subsequent general use of the limited and diffuse SSc subsets based largely on prevailing opinion.6
While some authors have shown an increase risk of RLD in patients with dcSSc compared with those with lcSSc, others found no difference.14 ,34–38 Based on our findings, it is possible that this may be due in part to the inclusion of a variable number of patients with intermediate skin phenotype in the lcSSc group. Mixed results have also been reported in studies analysing patients based on their mRSS, a scoring system assessing the severity of skin disease rather than the specific pattern of cutaneous involvement.22 ,39 The main challenge in using the mRSS is the subjective physician variability in examining the skin and the difficulty of uniformly measuring disease at the time of its maximal severity.
Our investigation confirms previous findings that a higher degree of skin disease is more prevalent among men and African Americans. It also emphasizes that RLD is an early and prevalent manifestation in scleroderma, as we found that a high number of patients who obtained their PFTs within 6 months from diagnosis had lung involvement.40–44 As previously shown, the value of FVC at the time of the first PFT is an important predictor of RLD progression and severity.45 ,46 Importantly, our study shows that progression to RLD after several years from disease onset can still be seen in all subtypes.
Scleroderma is associated with increased mortality and this investigation shows that the degree of skin involvement in association with autoantibody status can be an early predictor of survival. We found that patients negative for ACA and with type 3 skin disease have the greatest mortality risk and that survival in type 2 patients tends to decline faster than in type 1 patients. Our data also suggest that type 0 patients have death rates similar to those of type 3 patients despite their milder clinical phenotype.
The interpretation of these findings is limited by the small number of type 0 patients and the lack of knowledge about causes of death. There are several other limitations. This study is retrospective. One-quarter of our patients are non-white, which may limit the applicability of our findings to populations with different racial distribution. A referral bias toward more severely affected patients with SSc is possible. Data on other scleroderma-associated autoantibodies, including anti-RNA-polymerase III, anti-Pm/Scl, anti-Ku, antiphospholipid and anti-ribonucleoprotein (Th/To, U1, U3, U11/U12) antibodies are not available.47 ,48 Imaging studies to confirm the definite presence of interstitial lung disease are not recorded in our database and therefore there is the possibility that underlying muscle disease might have contributed to lower lung volumes. Also, we cannot fully exclude the possibility of a lead-time bias, as patients with more extensive SSc skin disease might have developed RLD earlier and not more often than other subjects. In allocating patients into skin subtypes, a small number of subjects with limited follow-up might have been misclassified because they had not reached their peak skin involvement. Finally, the multivariate regression models showed a lack of association between SSc subtypes and specific outcomes (ie, development of RLD and survival). This may be in part explained by the effects of colinearity since a well-established association is recognised between skin subtypes and specific autoantibodies such as anti-Scl-70 and ACA in SSc.
The strengths of this study include the availability of a large number of patients with SSc followed up at a single tertiary referral centre and the refined characterisation of their clinical phenotype over a long period of time (30 years) by physicians expert in the field of scleroderma.
In summary, this study suggests that a more refined phenotyping of SSc cutaneous involvement allows a classification of patients with SSc showing a greater power to predict clinical outcomes and survival than the traditional subdivision into limited and diffuse SSc. The improved prognostic value of more definite skin subsets together with their SSc-specific autoantibody patterns might facilitate the investigation and the management of this complex and highly heterogeneous disease.
We thank Adrianne Woods for database assistance, the Johns Hopkins Bayview Biostatistics, Epidemiology and Data Management (BEAD) Core for statistical help and the Scleroderma Research Foundation and the National Institute of Health (FB: NIH grant AR-055667) for their support.
Handling editor Tore K Kvien
Contributors All authors actively participated in the design, data analysis and writing of this manuscript and approved the submitted version.
Competing interests None.
Ethics approval Johns Hopkins University institutional review board.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.