In the current study, we investigated the association and predictive significance of baseline demographic and clinical variables with sequentially obtained percentage predicted FVC measurements in a large prospective cohort of early SSc patients. Similar to our previous cross-sectional study [18
], a number of baseline characteristics were associated with FVC levels. Only ATA, however, was associated with lower serial FVC levels as well as with a higher decline rate in FVC at 3-year follow-up.
Our results confirm the previously reported association of ACA [11
] and ATA with FVC levels; that is, those patients with ACA have higher percentage predicted FVC, and those patients with ATA have lower percentage predicted FVC. Previous reports also have indicated that patients with anti-RNA polymerase III antibodies have extensive skin involvement but less frequent ILD [13
]. These antibodies, however, were not associated with higher percentage predicted FVC levels in our study. Presence of ATA was also a predictor of an accelerated decline rate in FVC within 3 years of enrollment; however, presence of ATA was not a significant predictor of the FVC decline rate on long-term follow-up (up to 11.5 years). This discrepancy can be partially explained by the fact that patients with rapidly progressive ILD had higher mortality and were not well represented in the long-term follow-up group.
The GENISOS cohort includes a considerable number of Hispanic patients (n
= 77) and African American patients (n
= 54), raising the possibility that our findings also are generalizable to nonwhite patient populations. In our study, African American patients had lower levels of percentage predicted FVC, even after adjusting for socioeconomic factors, antibody status and smoking habits. This observation might partially explain lower survival of nonwhite patients in the GENISOS cohort [36
] and in other studies [37
]. However, our results confirmed that African American ethnicity per se
does not predict the rate of decline in percentage predicted FVC [7
One study reported that SSc patients who were former smokers had an accelerated rate of decline in FVC [19
], but this finding has not been confirmed by other studies [7
]. We did not observe an association between smoking and percentage predicted FVC levels or its rate of decline. It is possible that smoking status influences progression of other PFT parameters such as the total lung capacity and DLco.
Two retrospective studies indicated that low baseline FVC levels predicted accelerated rate of decline in this variable over time [16
]. Both studies conducted a time to decline analysis rather than directly measuring the decline rate in FVC. Other studies, using an annualized rate of decline to measure percentage predicted FVC, reported no association between baseline FVC and the annualized rate of decline [7
]. Similarly, our study did not indicate that low baseline PFT parameters predicted a higher rate of decline in FVC on subsequent visits.
Some studies have indicated that the loss of volume on PFTs was greatest early in the course of disease and that the rate of decline in percentage predicted FVC decelerates over time [7
], while others could not confirm this finding [9
]. Disease onset was defined as the time of first symptom attributable to SSc in the first two studies [7
]. In our study, the overall rate of decline in percentage predicted FVC did not slow down during the follow-up time; similar results were seen in our subgroup analyses according to disease type and duration. We used the two most commonly adopted methods for calculation of disease duration in our study. The disease duration determined by both calculation methods was not a significant predictor of rate of decline in FVC. Similar to our findings, the rate of decline in FVC did not differ significantly in the Scleroderma Lung Study when patients were stratified according to disease duration [40
]. The frequency of patients with ≥10% decline in FVC, however, showed its highest increase in the first follow-up year and increased only at a slower pace to 45.9% after ≥8 years of follow-up in our study (Table ). This raises the possibility that a smaller subgroup of patients has an accelerated decline rate in the initial phase of disease. The FVC decline in these patients might stabilize in later follow-up years if they survive the initial rapid worsening of the ILD. However, this subgroup might not be well represented in patients with longer follow-up time because of its higher mortality rate.
We utilized a random-effect repeated-measurement model for our analyses. This approach utilizes all follow-up FVC measurements and accounts for the within-subject correlation. The approach therefore provides more accurate estimation of predictors for rate of decline than the time to decline or the annualized rate of decline analysis methods utilized in previous studies [7
Confirming findings of a previous study [21
], we observed that a faster rate of decline in percentage predicted FVC was a marker of poor survival. This finding emphasizes the importance of predictive parameters for progression of FVC. However, our findings also support the results of previous studies [7
] in which routinely obtained demographic and clinical parameters - except for ATA - failed to predict the rate of decline in FVC over time.
Consequently, our study provides further support for the identification of predictive biomarkers utilizing novel imagining and molecular techniques. Two previous reports indicated that the extent of reticular pattern on high-resolution computed tomography (HRCT) utilizing a semiquantitative scoring system predicts further decline in FVC [22
]. The fibrosis score based on HRCT was not captured in our study and is not yet available for routine clinical use, but has promise for future studies. Another study has indicated that high sedimentation rates at baseline predict a faster decline in percentage predicted FVC [14
]. Furthermore, SSc patients with ILD had more severe reflux disease on monometry and impedance PH monitoring in a cross-sectional study [42
]. The severity of esophageal involvement on monometry, however, was not predictive of the decline in FVC in a prospective study [17
Furthermore, a polymorphism in the IRF5
gene was associated with ILD in a cross-sectional sample of SSc patients [43
]. Others have reported that pneumoproteins such as surfactant protein D and Krebs von den Lungen-6 correlate with the presence of pulmonary fibrosis and alveolitis in SSc [44
]. However, large observational cohort studies with a collection of genetic data and serial blood samples are needed to investigate the predictive significance of these potentially useful biomarkers.
Cyclophosphamide is the only treatment for ILD that has been demonstrated to be effective in SSc [41
]. Considering that only a small portion of our patients were ever treated with cyclophophamide (8.27%), we believe our observations are close to the natural history of ILD in SSc. Furthermore, a subgroup analysis after exclusion of patients ever treated with cyclophosphamide did not change our overall results. We cannot, however, exclude that treatment with various potentially effective medications may have influenced our findings. Another limitation of the current study is that the majority of investigated patients did not undergo HRCT at enrollment, and therefore the predictive significance of the previously described semiquantitative scoring system on HRCT could not be investigated in our cohort [22
]. Furthermore, most of our patients were recruited from tertiary-care rheumatology clinics in South Texas, raising the possibility of a referral bias toward more severe disease.
Despite the present study's prospective design and relatively large sample size, we cannot exclude the possibility that we are still underpowered to identify more subtle clinical and demographic predictors of decline in percentage predicted FVC. Furthermore, it is possible that some of the observed associations are false-positive findings; we did not correct for multiple comparisons in order to decrease the likelihood of missing clinically important associations (β-error).