Survival and prognosis factors in systemic sclerosis: data of a French multicenter cohort, systematic review, and meta-analysis of the literature

Population

The French National Scleroderma Cohort includes 42 centers. The present analysis was restricted to five university hospitals, Lille, Paris (two centers), Nantes, and Lyon to ensure better quality of data, especially on survival data. These five centers participated in recruiting about two thirds of the National Cohort. Data were retrospectively collected before 2010 and then prospectively collected.

Patients were included between January 1, 2000, and December 31, 2013, if they met the following inclusion criteria: (i) be aged over 18, (ii) fulfill the ACR 1980 preliminary classification criteria [28] or ACR/EULAR 2013 classification criteria [29] for SSc, (iii) have at least one additional visit after the inclusion visit, and (iv) be incident cases, defined as patients having disease duration from time of diagnosis to enrolment in the study of less than 3 years. Patients were followed-up until July 1, 2016. Patients were considered as lost to follow-up if the vital status could not be ascertained. When possible, the vital status was ascertained by querying death registers at birth town councils.

Collected data and variable definition

Data collected at the inclusion visit were patient demographics, history of Raynaud phenomenon (RP) and first non-RP symptom, SSc subtype and modified Rodnan skin score (mRSS), auto-antibody profile, and organ involvement.

Disease onset was defined as the time of onset of first non-RP symptom. Interstitial lung disease was diagnosed on HRCT or chest x-ray. Pulmonary function tests including forced vital capacity (FVC) and diffusing capacity of the lungs for carbon monoxide (DLCO) were collected. Six-minute walking distance (6MWD) was collected. Pulmonary hypertension (PH) was suspected on a Doppler echocardiogram when systolic pulmonary arterial pressure (PAP) was estimated to be > 35 mmHg or maximum tricuspid regurgitant jet velocity > 2.8 m/s. Pulmonary arterial hypertension (PAH) was confirmed by right heart catheterization (RHC) when mean PAP was found to be ≥ 25 mmHg at rest, with mean pulmonary arterial wedge pressure ≤ 15 mmHg. EKG alterations, left ventricular ejection fraction (LVEF), diastolic dysfunction [30], valvular disease (excluding tricuspid valve regurgitation to avoid confounding with PH), and pericarditis were recorded according to the American Society of Echocardiography and European Association of Cardiovascular Imaging guidelines [30]. Scleroderma renal crisis was defined as new onset hypertension > 150/85 mmHg associated with a decrease in renal function or manifestations of malignant hypertension. Gastrointestinal tract involvement included reflux, dysmotility, constipation, or diarrhea; signs of bacterial overgrowth and/or malabsorption; and abnormal manometry and/or endoscopy test. Muscle involvement included myalgia and/or muscle weakness and/or elevation of creatinine kinase (CPK). Joint involvement included arthralgia, synovitis, and/or tendon friction rubs. Anemia was defined as a hemoglobin level < 12 g/dl. Smoking included self-reported current or former cigarette smoking.

Systematic review and meta-analysis

The meta-analysis was conducted according to MOOSE guidelines [31]. MEDLINE and EMBASE databases were queried by two of the authors (MRP and DL) using the following search terms: ((systemic sclerosis [Title]) OR (scleroderma, systemic[Title])) AND ((death) OR (mortality) OR (prognosis) OR (survival)). Cochrane did not retrieve additional abstracts. All records published before July 1, 2017, were included in the search. Language was restricted to English or French. Reference list of selected studies was hand-searched for additional relevant studies to be included in the meta-analysis.

Two of the authors (MRP and DL) independently screened the titles and abstracts of the retrieved records to identify eligible articles. The full text of eligible articles was read for inclusion in the meta-analysis. Selected articles were compared, and in case of disagreement, decisions were made by consensus.

Cohort studies of unselected adult SSc patients assessing routine clinical and laboratory prognosis factors and SMR were included. Studies which included patients diagnosed with SSc overlap with other connective tissue diseases were excluded.

Studies from the same centers were included if their respective study periods were different. If, for the same center, two studies covered an overlapping study period, data from the largest cohort were kept in the analysis. A study was recorded as an incident according to the authors’ definition.

Quality of the studies was assessed using the Newcastle-Ottawa scale [32].

Data were extracted and entered into a predefined spreadsheet table which included the following items: study design, length of follow-up, definition of disease onset, disease duration, SMR, and adjusted or, if unavailable, unadjusted hazard ratios (HR) for each studied prognosis factor.

Data analysis

Characteristics of the population were described using mean ± standard deviation or median (interquartile range (IQR)) in case of non-normality, for quantitative variables, and number (percentage) for qualitative variables. Comparisons between limited cutaneous SSc (lcSSc) and diffuse cutaneous SSc (dcSSc) patients were conducted using the Student t test or Wilcoxon test in case of non-normality for quantitative variables and Fisher’s exact test for qualitative variables.

Survival was estimated from diagnosis using the Kaplan-Meier method. Prognosis factors were assessed by Cox regression analysis in the non-adjusted analysis and subsequently adjusted for age, sex, and SSc subtype. The assumption that hazard ratios were constant over time was verified. SMR was calculated as the ratio of observed death in the cohort to the number of death of the French age/sex-matched population in 2014.

We calculated weighted pooled summary estimates of SMR and HR of prognosis factors. For each meta-analysis, we used the DerSimonian and Laird method. Accordingly, studies were considered to be a random sample from a population of studies. Heterogeneity was assessed using an I² statistic and a chi-square heterogeneity statistic. A random-effects model was used to combine data. The overall effect was estimated using a weighted average of individual effects, with weights inversely proportional to variance in observed effects. Publication bias was evaluated with a funnel plot and Egger’s test. The pooled SMR and HR were estimated with 95% confidence interval (CI). Meta-regression was used to assess the impact of mid-cohort year, the proportion of males, the proportion of diffuse cutaneous forms, and the prevalence of anti-Scl70 antibodies on SMR. The impact of diagnosis of PH by RHC on the association of PH with mortality was evaluated. Separate analyses were performed for (i) SMR according to whether a given study included incident cases only and (ii) HR of PH diagnosed by either echocardiography and/or RHC and PH diagnosed by RHC.

All analyses were performed using R software with the survival and metafor packages. p values less than 0.05 were considered significant.

French cohort study

Survival and standardized mortality ratio

During the study period, 104 deaths (16.6%) were recorded and 133 patients were lost to follow-up. Overall survival rates at 1, 3, 5, 10, and 15 years from diagnosis were 98.0% (95% CI 96.9–99.1%), 92.5% (90.4–94.7%), 85.9% (82.8–89.1%), 71.7% (66.3–77.5%), and 53% (33.8–83.4%) respectively. Survival rates for the diffuse and limited cutaneous subtypes are shown in Fig. 1 and in Additional file 1: Table S1. Overall SMR was 5.73 (95% CI 4.68–6.94).

Prognosis factors

No association was found for digital ulcers, gastrointestinal, articular, muscular involvement, and specific auto-antibodies after adjustment.

Male sex showed a trend towards worse outcome, but without reaching statistical significance (HR = 1.53; 95% CI 0.98–2.39; p = 0.06).

Meta-analysis: study selection

A total of 4128 citations were assessed for inclusion. After screening, 244 abstracts were deemed potentially relevant and the full-text copies were obtained. Of these articles, 44 studies, including our cohort, were included in the meta-analysis (Fig. 2). Eighteen articles were included in the SMR analysis, representing a total population of 11,719 patients [7–12, 15, 22–24, 26, 33–38]. Thirty-six studies were included in the prognosis factor analysis, representing a total of 26,187 patients [3, 7, 10–18, 20, 22–26, 34, 39–55]. No study was excluded based on poor quality. The main characteristics of the studies are summarized in Additional file 1: Table S2 and Table S3.

SMR meta-analysis

The pooled SMR for all studies was 3.45 (95% CI 3.03–3.94; I² = 88.8%; p(het) < 0.001). The pooled SMR for studies including only incident patients was 3.64 (95% CI 3.06–4.34; I² = 82.0%; p(het) < 0.001), and the pooled SMR for studies including prevalent patients was 3.28 (95% CI 2.69–3.99; I² = 91.6%; p(het) < 0.001). There was no funnel plot asymmetry, and Egger’s test failed to provide any evidence for small study effect, making publication bias unlikely. Meta-regression stratified by study type (incident or prevalent) did not show any association with SMR (p = 0.461), meaning no statistical difference between pooled SMR of incident and prevalent studies. Subsequent analyses were therefore conducted on all studies. Meta-regression analysis revealed a significant increase of SMR with proportion of dcSSc (p < 0.001) and prevalence of anti-Scl70 antibodies (p = 0.021). There was no association with male sex (p = 0.130). There was no significant association between SMR and mid-cohort year (p = 0.656) (Additional file 2: Figure S1. Prognosis factors meta-analysis).

Prognosis factor meta-analysis

Survival and SMR

With a mid-cohort year of 2008, our study population is the largest multicenter incident and well-phenotyped cohort study of SSc patients in France and is among the most recent published to date in the literature. The overall survival rates at 5 and 10 years from diagnosis were 85.9% and 71.7%, respectively, and are lower than those reported in other recent cohorts [7, 22, 24, 44, 54]. We also report one of the highest SMR of 5.73. These differences could be explained by a high heterogeneity between studies as well as methodological issues such as the inclusion of prevalent cases in many studies or differences in time origin from which survival time is calculated (from disease onset, diagnosis, or enrolment). It is usually admitted that studies including prevalent cases underestimate mortality and that better survival is observed in prevalent patients with longer disease duration prior to inclusion. Yet, our meta-analysis did not show a significant difference between pooled SMR of studies that included prevalent cases and those restricted to incident according to the authors’ definition. The high heterogeneity observed within studies with incident patients could be due to the definition of incidence and the proportion of males and patients with anti-Scl70 antibodies. This high heterogeneity could explain the lack of difference between incident and prevalent cohorts. Meta-regression showed a significant association between SMR and proportion of dcSSc (p < 0.001) and prevalence of anti-Scl70 antibodies (p = 0.021). Our high SMR of 5.73 could therefore be partly explained by the high proportion of anti-Scl70 antibodies (35%) in our population. Interestingly, there has been a debate whether or not the survival could have improved over time in SSc. Our study did not show any improvement of SMR over time, which is in line with the study of Elhai et al. [6]. However, considering life expectancy in OECD (Organisation for Economic Co-operation and Development) countries increased by 12 years from 1960 to 2014 [56], and SMR being the ratio of mortality in SSc cohorts to that of the general population, this suggests that all-cause mortality has decreased proportionately to the general population in SSc cohorts.

Prognosis factors

Prognosis factors have been assessed in many observational studies [7, 13–26] and have been recently reviewed. Our systematic review and meta-analysis, as well as two prior meta-analyses [5, 27] and a recent EUSTAR study [2], have identified the following characteristics as consistently associated with a worse prognosis: male gender; older age; dcSSc; lung and cardiac involvement, including PH and ILD; kidney involvement; and inflammation. These robust factors are included in a recent prognosis score [2] as well as in older ones [57, 58].

Besides these well-known prognosis factors, our cohort study identifies new ones: telangiectasia, 6MWD, valvular disease, cancer, and autoantibody status.

Telangiectasia was slightly associated with a higher mortality in our study population. In contrast, Poormoghim et al. [51] reported a non-significant, yet elevated HR of 1.44 in a smaller cohort of Iranian patients. An increased number of telangiectasia has been suggested to be a clinical marker of microvascular disease in SSc and is associated with an increased risk of PAH [59].

The 6MWD is a simple tool used to assess submaximal functional capacity. It is influenced by various disease parameters during SSc and lacks organ specificity [60]. While the 6MWD has been shown to be an independent predictor of mortality in idiopathic PAH, its prognosis value in SSc-PAH is less clear [61]. To our knowledge, we are the first to report a negative association of the 6MWD with survival (HR = 0.51) in SSc patients. This can be at least partly explained by the association between the 6MWD and PAH [60].

While cardiac involvement in SSc patients is robustly associated with a poor prognosis, conferring a nearly fivefold increased risk of mortality in our meta-analysis, no study had yet focused on the valvular manifestations of SSc. A recent article comparing echocardiography in SSc patients and a matched control population showed a greater frequency of valvular regurgitation and valvular replacement due to regurgitation [62]. Moreover, in a large multicenter French cohort, De Groote et al. [63] reported 6.7% mitral regurgitation and 2.5% aortic regurgitations. To our knowledge, we are the first to describe an association between valvular disease and survival in SSc. These data indicate that more attention should be paid to valvular disease in SSc patients and further studies are needed to confirm its prognostic significance.

An increased incidence of malignancy has been reported during SSc, especially lung and hematological cancer [64]. Cancer has also been described among the leading cause of non-SSc-related deaths [7, 11, 15, 65], and a temporal relation has been reported between the onset of cancer and SSc [66]. As expected, in our cohort as well as in the meta-analysis, malignancy was significantly associated with shorter survival.

In our cohort, we did not observe any association between anti-Scl70 antibodies or ACA and survival. Interestingly, the absence of association of anti-Scl70 antibodies with survival has been suggested by numerous studies [7, 17, 18, 23, 25] while the protective role of anti-centromere antibodies is better established [7, 18, 22, 25, 48]. Our meta-analysis confirms that the presence of ACA is associated with better survival (pooled HR = 0.58). Moreover, our meta-analysis suggests that the presence of anti-Scl70 antibodies could be indeed a predictor of mortality with a pooled HR of 1.38. Our analysis also highlights a probable publication bias. Small studies reporting a negative association of anti-Scl70 antibodies with death are notably lacking. Therefore, it is difficult to draw a firm conclusion on the role of anti-Scl70 antibodies as a prognosis factor in SSc.

The major strength of our study is the availability of detailed clinical and laboratory characteristics in a multicenter cohort of incident patients. The major strengths of our meta-analysis include (i) the first analysis of pooled HR of anti-Scl70/ACA antibodies, (ii) the separate analysis of pooled SMR in incident cohorts of SSc, and (iii) the separate analysis of pooled HR of PH diagnosed by RHC only.

The main limitation is a proportion of loss to follow-up of around 20% in our cohort, despite our attempts to collect information on participants who dropped out. These patients lost to follow-up had a higher prevalence of ILD and lower prevalence of ACA at baseline, leading to potential underestimation of mortality. A second limitation is the variable definition of incident patients among studies. Although we defined incident patients as newly diagnosed ones, these patients had relatively short disease duration from first non-RP symptom diagnosis, thus minimizing survivor bias. Because of missing values, the effect of specific treatments and other data such as type and stage of cancer could not be studied, and no multivariate analysis could be performed. In addition, multiple univariate tests are responsible for an inflation of the alpha risk. However, multiple test adjustments in such exploratory study, in addition for a rare disease, are not strictly required [67]. Finally, our study was performed in five selected referral centers and may therefore have focused on a subset of patients with more severe disease, which could limit the representativeness of our findings.