[ Original Research Diffuse Lung Disease ]

[ Original Research Diffuse Lung Disease ] Predicting Mortality in Systemic Sclerosis-Associated Interstitial Lung Disease Using Risk Prediction Models Derived From Idiopathic Pulmonary Fibrosis Christopher J. Ryerson, MD ; Darragh O Connor, BSc ; James V. Dunne, MD ; Fran Schooley ; Cameron J. Hague, MD ; Darra Murphy, MD ; Jonathon Leipsic, MD ; and Pearce G. Wilcox, MD BACKGROUND: Mortality risk prediction tools have been developed in idiopathic pulmonary fibrosis, however, it is unknown whether these models accurately estimate mortality in systemic sclerosis-associated interstitial lung disease (SSc-ILD). METHODS: Four baseline risk prediction models the Composite Physiologic Index, the Interstitial Lung Disease-Gender, Age, Physiology Index, the du Bois index, and the modified du Bois index were calculated for patients recruited from a specialized SSc-ILD clinic. Each baseline model was assessed using logistic regression analysis with 1-year mortality as the outcome variable. Discrimination was quantified using the area under the receiver operating characteristic curve. Calibration was assessed using the goodness-of-fit test. The incremental prognostic ability of additional predictor variables was determined by adding prespecified variables to each baseline model. RESULTS: The 156 patients with SSc-ILD completed 1,294 pulmonary function tests, 725 6-min walk tests, and 637 echocardiograms. Median survival was 15.0 years from the time of SSc-ILD diagnosis. All baseline models were significant predictors of 1-year mortality in SSc- ILD. The modified du Bois index had an area under the receiver operating characteristic curve of 0.84, compared with 0.77 to 0.81 in the other models. Calibration was acceptable for the modified du Bois index, but was poor for the other models. All baseline models include FVC and 6-min walk distance was identified as an additional independent predictor of 1-year mortality. CONCLUSIONS: The modified du Bois index has good discrimination and calibration for the prediction of 1-year mortality in SSc-ILD. FVC and 6-min walk distance are important independent predictors of 1-year mortality in SSc-ILD. CHEST 2015; 148 ( 5 ): 1268-1275 Manuscript received January 1, 2015; revision accepted May 1, 2015; originally published Online First May 21, 2015. ABBREVIATIONS: 6MWD 5 6-min walk distance; 6MWT 5 6-min walk test; AUROC 5 area under the receiver operating characteristic; CPI 5 Composite Physiologic Index; CTD-ILD 5 connective tissue diseaseassociated interstitial lung disease; D lco 5 diffusion capacity of the lung for carbon monoxide; ILD 5 interstitial lung disease; ILD-GAP 5 Interstitial Lung Disease-Gender, Age, Physiology; IPF 5 idiopathic pulmonary fibrosis; PFT 5 pulmonary function test; RVSP 5 right ventricular systolic pressure; SSc 5 systemic sclerosis; SSc-ILD 5 systemic sclerosis-associated interstitial lung disease AFFILIATIONS: From the Department of Medicine (Drs Ryerson, Dunne, and Wilcox, Mr O Connor, and Ms Schooley), Centre for Heart Lung Innovation (Drs Ryerson and Wilcox), and Department of Radiology (Drs Hague, Murphy, and Leipsic), The University of British Columbia, Vancouver, BC, Canada. FUNDING/SUPPORT: The study was funded in part by the British Columbia Lung Association. Dr Ryerson is supported by a Career Investigator Award from the Michael Smith Foundation for Health Research. Mr O Connor was supported by a student research fellowship sponsored by InterMune. CORRESPONDENCE TO: Christopher J. Ryerson, MD, St. Paul s Hospital, 1081 Burrard St, Ward 8B, Vancouver, BC, V6Z 1Y6, Canada; e-mail: chris.ryerson@hli.ubc.ca 2015 AMERICAN COLLEGE OF CHEST PHYSICIANS. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians. See online for more details. DOI: 10.1378/chest.15-0003 1268 Original Research [ 148 # 5 CHEST NOVEMBER 2015 ]

Fibrotic interstitial lung disease (ILD) is a common complication of systemic sclerosis (SSc), 1,2 with a reported prevalence of 25% to 90%. 3 ILD is the leading cause of mortality in SSc 4 ; however, prognostication of SScassociated ILD (SSc-ILD) remains challenging given the substantial variability in disease course. Previous studies have shown that mortality in SSc-ILD is influenced by a number of patient-specific, ILD-specific, and SSc-specific factors,5 however, it is unknown how these variables should be used in combination to accurately estimate mortality risk in SSc-ILD. SSc-ILD shares many clinical features and prognostic variables with idiopathic pulmonary fibrosis (IPF), a chronic progressive fibrotic ILD of uncertain etiology. 6 Previous studies have described mortality risk prediction tools in IPF, 7-10 including one model that has been validated in a large cohort of patients with a heterogeneous collection of connective tissue disease-associated ILD (CTD-ILD). 11 Th is study included a small number of patients with SSc-ILD, but did not report model performance in this population. SSc-ILD also has several distinct features compared with IPF, and it is, therefore, unknown whether risk prediction models derived in IPF can also be used to accurately estimate mortality risk in SSc-ILD. We conducted a retrospective analysis using a large cohort of patients with SSc-ILD to determine whether previously validated IPF mortality risk prediction tools accurately estimate 1-year mortality in SSc-ILD. Our secondary goal was to compare previously derived IPF models to a similar novel model derived in SSc-ILD. We show that 1-year mortality in SSc-ILD can be predicted using previously derived risk prediction tools; however, calibration of most models was poor, and additional data are required to determine how these tools should be incorporated into clinical practice. Materials and Methods Study Population Patients were identified from a specialized SSc-ILD clinic in which they were assessed by a multidisciplinary team consisting of a pulmonologist, rheumatologist, and specialized nurse. Patients were included if they fulfilled American College of Rheumatology (ACR) diagnostic criteria for SSc 12,13 and had fibrotic ILD with radiologic or pathologic findings consistent with nonspecific interstitial pneumonia or usual interstitial pneumonia. 6,14 There were no exclusion criteria. All patients provided written informed consent for inclusion in a prospective ILD database (University of British Columbia ethics No. H10-03435 and H14-02858). Measurements Patients underwent pulmonary function tests (PFTs) according to established criteria for measurement of spirometry, lung volumes, and diffusion. 15-17 Patients completed 6-min walk tests (6MWTs) following established procedures, 18 including use of a forehead oxygen saturation probe. PFTs and 6MWTs are typically performed at 6-month intervals at our center. Echocardiography was performed annually, including estimation of the right ventricular systolic pressure (RVSP) based on the velocity of the tricuspid regurgitant jet. Vital status was determined at the time of analysis for all patients, excluding one patient who had moved out of province who was censored at the date last known to be alive. Mortality Risk Prediction Models Four baseline risk prediction models were calculated, including the Composite Physiologic Index (CPI), 8 the ILD-Gender, Age, Physiology (ILD-GAP) Index, 11 the du Bois index, 9 and the modified du Bois index. 10 All four models were derived based on patients with IPF. The CPI was derived to predict the extent of fibrosis on CT scan 8 a n d is also associated with mortality in IPF. The CPI was calculated based on simultaneous measurements of FVC, FEV 1, and diffusion capacity of the lung for carbon monoxide (D lco ). The Gender, Age, Physiology Index was derived from a prospective cohort of patients with IPF, 7 and the ILD-GAP Index was subsequently validated in several other fibrotic ILD subtypes, including patients with a variety of CTD-ILDs. 11 The ILD-GAP Index is calculated based on the ILD diagnosis, patient sex, age at the time of assessment, FVC, and D lco. The du Bois index was derived in the placebo arms from two randomized controlled trials of mild to moderate IPF, 19,20 with the purpose to estimate 1-year mortality. The du Bois index is calculated based on patient age, history of respiratory hospitalization in the preceding 24 weeks, FVC, and change in FVC in the preceding 24 weeks. 9 The modified du Bois index expanded the original du Bois index to also include 6-min walk distance (6MWD) and change in 6MWD in the preceding 24 weeks. 10 For the du Bois indexes, change in FVC and 6MWD were both calculated based on changes that occurred over intervals of 16 to 32 weeks (ie, 24 8 weeks), allowing for the irregular follow-up intervals typical of clinical practice. Statistical Analysis Mortality risk prediction models were calculated as previously described. 8-11 For the primary analysis, each baseline model was included as a single predictor variable in a logistic regression analysis with 1-year mortality as the outcome variable. Discrimination was quantified using the area under the receiver operating characteristic (AUROC) curve for prediction of 1-year mortality. An AUROC curve 0.70 is generally considered acceptable for clinical use. Calibration (the extent to which observed event rates match expected event rates in population subgroups) was assessed using the Hosmer-Lemeshow goodness-of-fit test, in which a low P value indicates poor discrimination. The observed 1-year mortality rate in our SSc-ILD cohort was compared with the score-specific expected 1-year mortality provided in the original ILD- GAP Index and du Bois publications. 9,11 The performance of each model for the prediction of time to death was also assessed using a Cox proportional hazards analysis, reporting discrimination using the c-statistic. Individual patients could provide multiple data points for each model if repeated measurements were available. The incremental prognostic ability of additional potential predictor variables was determined by adding these individual variables to each of the baseline models. These potential predictors of mortality were prespecified and included commonly measured variables that have been identified as possible prognostic variables in SSc-ILD or in other ILD populations. These variables included age, sex, smoking history and pack-years, BMI, measures of pulmonary physiology, 6MWD, and RVSP. A stepwise logistic regression analysis was performed to determine which additional variables provided independent prognostic information when added to the baseline models, retaining variables with a P value,.05. journal.publications.chestnet.org 1269

The baseline models were qualitatively compared with a novel model derived from our study population. This novel model was derived by first identifying unadjusted predictors of 1-year mortality as listed in the previous paragraph. Variables associated with mortality on unadjusted analysis were then included in a stepwise logistic regression model with backward elimination, retaining variables that were independently associated with 1-year mortality. A two-sided P value,.05 was considered significant for all analyses. All data analysis was performed using STATA 11.0 (StataCorp). Results Study Population Th e study population included 156 patients with SSc who were enrolled in a prospective ILD cohort, with a diagnosis of SSc-ILD established between June 1997 and November 2013. Baseline characteristics are summarized in Table 1. These patients underwent a total of 1,294 PFTs, 725 6MWTs, and 637 echocardiograms. Patients were followed for up to 15.5 years (mean 6.1 years). A total of 53 patients were deceased at the time of data analysis; no patients underwent lung transplant. Median survival was 15.0 years from the time of SSc-ILD diagnosis (Fig 1 ). Mortality Prediction Using Baseline Models The CPI, ILD-GAP Index, du Bois index, and modified du Bois index were each significant predictors of 1-year mortality in SSc-ILD. The modified du Bois index showed the best discrimination, with an AUROC curve of 0.84, compared with an AUROC curve of 0.77 to 0.81 in the other models (Fig 2, Table 2 ). Discrimination for each model was lower for prediction of overall mortality TABLE 1 ] Patient Characteristics at the Time of SSc-ILD Diagnosis Characteristic Value Age, mean (SD), y 54.5 (13.2) Male sex, % 16 Current or previous smoker, % 42 Pack-y, median (interquartile range) 0 (0-12) BMI, mean (SD), kg/m 2 25.5 (5.1) Antinuclear antibody positive, % 98 Anti-Scl-70 positive, % 34 Anti-centromere positive, % 22 FVC % predicted, mean (SD) 81 (20) FEV 1 % predicted, mean (SD) 80 (20) TLC % predicted, mean (SD) 86 (21) D LCO % predicted, mean (SD) 59 (20) 6MWD, mean (SD), m 389 (132) Pulmonary artery systolic pressure, mean (SD), mm Hg 36 (13) 6MWD 5 6-min walk distance; D LCO 5 diffusion capacity of the lung for carbon monoxide; SSc-ILD 5 systemic sclerosis-associated interstitial lung disease; TLC 5 total lung capacity. (ie, Cox proportional hazards analysis), ranging from a c-statistic of 0.69 in the modified du Bois model to 0.76 in the CPI. Mortality is shown in Figure 3, stratified by score-specific subgroups determined from the ILD-GAP Index and du Bois index. The ILD-GAP Index has previously established indexspecific estimates for 1-year mortality in patients with CTD-ILD,11 and the original du Bois index provided 1-year mortality estimates for patients with IPF. 9 The observed mortality rate in SSc-ILD was similar to these expected 1-year mortality rates ( Table 3 ); however, there was less consistent agreement between observed and expected 1-year mortality rates in patients with a moderate risk of mortality according to the ILD-GAP Index (ie, ILD-GAP Index of two to three points), and in those with a low to moderate risk of mortality according to the du Bois index (ie, du Bois index 29 points). Calibration (goodness of fit) was poor for the prediction of 1-year mortality using the CPI, ILD-GAP Index, and du Bois index, but was acceptable for the modified du Bois index ( Table 2 ). Mortality Prediction Using Modified Models Unadjusted predictors of mortality in SSc-ILD included respiratory hospitalization, several PFT measurements, 6MWD, and RVSP measured by echocardiography ( Table 4 ). Stepwise regression showed that 1-year mortality in SSc-ILD was independently predicted by a low FVC and low 6MWD (AUROC curve, 0.88). The model-adjusted hazard ratios for each predictor variable Figure 1 Survival in systemic sclerosis-associated interstitial lung disease. 1270 Original Research [ 148 # 5 CHEST NOVEMBER 2015 ]

Figure 2 Receiver operating characteristic curve for each baseline model. CPI 5 Composite Physiologic Index; ILD-GAP 5 Interstitial Lung Disease-Gender, Age, Physiology. are also shown in Table 4. This table shows the independent association of these predictor variables with 1-year mortality when each variable is added to the stated baseline model. PFT measurements, 6MWD, and RVSP were associated with mortality when added to the baseline models. There was no independent association of autoantibody status with mortality ( P..15 for all four baseline models), and RVSP remained an independent predictor of mortality when adjusting for anti-centromere antibody status. We used stepwise regression analysis to assess for independent association of these additional variables with the baseline model ( Table 5 ). Discrimination of the baseline CPI and modified du Bois models improved modestly with inclusion of additional independent prognostic variables, with greater improvements in discrimination for the ILD-GAP Index and du Bois index. Discussion Th e similar clinical features and mortality risk factors in SSc-ILD and IPF suggest that a single mortality risk TABLE 2 ] Discrimination and Calibration of Baseline Models for Prediction of 1-Year Mortality Model Discrimination (AUROC) Calibration a (Goodness of Fit) CPI 0.81, 0.00005 ILD-GAP Index 0.77 0.01 du Bois index 0.79 0.004 Modified du Bois index 0.84 0.19 AUROC 5 area under the receiver operating characteristic curve; CPI 5 Composite Physiologic Index; ILD-GAP 5 Interstitial Lung Disease- Gender, Age, Physiology. a Calibration (goodness of fit) was assessed using the Hosmer-Lemeshow Test, where a P value,.05 indicates poor fit. prediction tool may be useful in both of these populations despite other differences between these populations. Several prediction models have been derived from IPF, but none of these have been sufficiently evaluated in patients with SSc-ILD. In this study, we have applied these tools to a large cohort of patients with SSc-ILD, showing that 1-year mortality in SSc-ILD can be predicted with good discrimination, but often with poor calibration. We have identified the modified du Bois model as a potential model for use in SSc-ILD based on its acceptable discrimination and calibration; however, additional novel prediction models may provide superior performance. These findings have significant implications for clinical practice, however, several important questions remain. The modified du Bois index had the best discrimination and calibration for the prediction of 1-year mortality in SSc-ILD. 10 This is not surprising for several reasons. First, the du Bois index was derived to predict 1-year mortality, 9 while the CPI was derived to predict the extent of radiologic fibrosis, 8 and the ILD-GAP model was derived to predict time to death. 7 Our choice of 1-year mortality as the primary end point, therefore, favored the du Bois indexes over the CPI and ILD-GAP Index. Conversely, the modified du Bois index had a lower discrimination for prediction of time to death compared with the CPI and ILD-GAP Index, reflecting that the CPI and ILD-GAP were intended for both short and longer-term prognostication while the du Bois indexes were not developed for this purpose. Second, the modified du Bois index incorporates change in FVC and 6MWD, while the CPI and ILD-GAP Index models are both based on measures taken at a single point in time. The modified du Bois index, therefore, captures important prognostic information conveyed by the rate of decline, although this improved prognostication comes at the expense of requiring longitudinal data for risk assessment. Finally, the modified du Bois model includes the 6MWD, which incorporates the functional consequences of several pulmonary, cardiac, and musculoskeletal diseases that are all likely contributors to mortality in patients with SSc-ILD. The importance of the 6MWD is further illustrated by its addition to the CPI and ILD-GAP Index as a key variable that provides independent prognostic information beyond these baseline models, as well as the inclusion of 6MWD in an unbiased stepwise regression that identified 6MWD and FVC as the only independent predictors of 1-year mortality. Despite the superior calibration and discrimination of the modified du Bois model, additional data are journal.publications.chestnet.org 1271

Figure 3 A, B, Survival in systemic sclerosis-associated interstitial lung disease stratified by the ILD-GAP Index (A) and du Bois index (B). Survival is grouped according to the index values identified in each graph. See Figure 2 legend for expansion of abbreviation. required to determine how this model should be used in SSc-ILD, as well as in IPF. RVSP was not an independent predictor of mortality in a novel model that included FVC and 6MWD, nor was it an independent predictor of mortality in three of the four baseline models when 6MWD was also included as a predictor variable. We believe 6MWD was a more important prognostic variable than RVSP because it measures the cumulative impact of ILD, pulmonary hypertension, and other comorbidities, thus providing more prognostic information than the RVSP alone. This finding does not diminish the importance of pulmonary hypertension in SSc-ILD, but does demonstrate that TABLE 3 ] Expected and Observed Mortality Using Previous IPF Prediction Models Index Expected 1-y Mortality, % Observed 1-y Mortality in SSc-ILD, % ILD-GAP Index 0-1 3.1, in CTD-ILD 3.4 2-3 8.8, in CTD-ILD 15.0 4-5 18.2, in CTD-ILD 22.8 6 a 33.5, in CTD-ILD... du Bois index 0-4, 2, in IPF 3.2 8-14 2-5, in IPF 1.0 16-21 5-10, in IPF 11.6 22-29 10-20, in IPF 10.5 30-33 20-30, in IPF 23.8 34 b. 30, in IPF 42.1 CTD-ILD 5 connective tissue disease-associated interstitial lung disease; IPF 5 idiopathic pulmonary fibrosis. See Table 1 and 2 legends for expansion of other abbreviations. a The observed 1-y mortality could not be calculated for patients with an ILD-GAP Index 6 due to an insufficient number of patients with these scores. b Patients with a du Bois index 34 were grouped since scores 34 were uncommon in patients with SSc-ILD. mortality can be predicted in SSc-ILD without using pulmonary hypertension-specific variables such as RVSP. We chose 1-year mortality as our primary end point for multiple reasons. First, model performance typically degrades over time, and thus the accuracy in predicting early events is superior to the prediction of later events. For this reason, most prediction models concentrate on the prediction of short-term events. Second, there are established estimates for 1-year mortality for the ILD- GAP Index (in a connective tissue disease population) and in the du Bois index (in an IPF population), allowing comparison of observed to expected 1-year mortality in this SSc-ILD population. Third, estimation of long-term mortality has less clinical relevance for decision-making. Timing of lung transplantation, escalation of therapy, and initiation of palliative measures are all critical decisions that are primarily based on short-term prognosis. We found that model performance was worse when conducting a time-to-event analysis (ie, Cox regression). This indicates the need to reassess prognosis at regular intervals based on updated clinical measurements, rather than relying on remote prognostic estimates that become less accurate over time. None of the baseline models compared favorably to a novel risk prediction model derived in this SSc-ILD population. Although this unvalidated model is subject to overfitting, its superior performance suggests that estimation of mortality risk in SSc-ILD may be improved by using prediction tools specifically derived in this population. We identified reduced FVC and reduced 6MWD as the only independent predictors of 1-year mortality in this novel model, suggesting that these are key variables that should be evaluated in future studies. Previous studies have also identified radiologic extent of fibrosis and a usual interstitial pneumonia pattern as prognostic variables in SSc-ILD. 21-23 Evaluating 1272 Original Research [ 148 # 5 CHEST NOVEMBER 2015 ]

TABLE 4 ] Predictors of 1-Year Mortality Adjusting for Baseline Prediction Models Model-Adjusted HR Predictor Variable Unadjusted HR CPI ILD-GAP Index du Bois Index Modified du Bois Index Age 0.97 (0.81-1.17) 1.10 (0.83-1.38)......... P 5.77 P 5.60 Male sex 1.33 (0.76-2.35) 0.29 (0.12-0.71)... 0.64 (0.28-1.47) 2.24 (0.69-7.28) P 5.32 P 5.006 P 5.29 P 5.18 Current or previous smoker 1.34 (0.85-2.12) 1.25 (0.69-2.28) 1.20 (0.71-2.03) 1.25 (0.62-2.52) 0.88 (0.28-2.73) P 5.21 P 5.47 P 5.49 P 5.54 P 5.83 Pack-y a 1.14 (0.99-1.31) 1.32 (1.11-1.57) 1.15 (0.99-1.35) 1.15 (0.90-1.46) 1.01 (0.60-1.69) P 5.08 P 5.002 P 5.08 P 5.26 P 5.97 BMI 0.91 (0.86-0.97) 0.93 (0.86-0.99) 0.97 (0.91-1.03) 0.94 (0.87-1.01) 1.01 (0.92-1.11) P 5.002 P 5.04 P 5.26 P 5.11 P 5.81 Respiratory hospitalization 10.8 (5.3-21.9) 2.35 (0.80-6.90) 7.00 (2.94-16.7)...... P,.0005 P 5.12 P,.0005 FVC % predicted a 0.61 (0.54-0.70)............ P,.0005 FEV 1 % predicted a 0.61 (0.54-0.70)... 0.79 (0.67-0.94) 0.79 (0.64-0.98) 0.78 (0.57-1.08) P,.0005 P 5.007 P 5.03 P 5.13 TLC % predicted a 0.63 (0.54-0.73) 1.39 (1.09-1.77) 0.81 (0.67-0.98) 0.79 (0.62-1.00) 0.59 (0.40-0.87) P,.0005 P 5.008 P 5.03 P 5.05 P 5.008 D LCO % predicted a 0.41 (0.33-0.51)...... 0.48 (0.34-0.69) 0.55 (0.32-0.93) P,.0005 P,.0005 P 5.03 6MWD a 0.90 (0.88-0.93) 0.93 (0.89-0.97) 0.91 (0.88-0.94) 0.92 (0.88-0.96)... P,.0005 P,.0005 P,.0005 P,.0005 RVSP a 1.57 (1.35-1.83) 1.31 (1.04-1.65) 1.64 (1.33-2.01) 1.75 (1.27-2.41) 1.66 (0.82-3.38) P,.0005 P,.0005 P,.0005 P,.0005 P 5.16 Data are reported as HR (95% CI), adjusting for the specified model. Empty cells represent variables that were included in the specified model. HR 5 hazard ratio; RVSP 5 right ventricular systolic pressure. See Table 1 and 2 legends for expansion of other abbreviations. a Hazard ratio reported for a 10-unit change in predictor variable. journal.publications.chestnet.org 1273

TABLE 5 ] Mortality Prediction Using Modified Prediction Models Baseline Model Variables in Baseline Model Baseline Model AUROC Curve a Variable(s) Added to Baseline Model b Modified Model AUROC Curve CPI FVC, FEV 1, D LCO 0.85 6MWD 0.87 ILD-GAP Index Sex, age, FVC, D LCO 0.82 6MWD 0.87 du Bois index Age, respiratory hospitalization, FVC, D FVC 0.83 RVSP 0.91 Modified du Bois index Age, respiratory hospitalization, FVC, D FVC, 6MWD, D 6MWD 0.80 D LCO 0.83 D 6MWD 5 change in 6MWD; D FVC 5 change in FVC. See Table 1, 2, and 4 legends for expansion of other abbreviations. a The baseline AUROC was based on the cohort of patients with available measurements for the variables added to the baseline model (ie, the baseline AUROC could be slightly different than the AUROC listed in Table 2 ). b All variables were independent predictors of 1-y mortality and improved the net reclassification index for prediction of 1-y mortality when added to the baseline model ( P,.05). radiologic features was beyond the scope of this study since our primary goal was to compare mortality risk prediction tools that were previously derived in other fibrotic ILD. Our secondary goal was to compare these previous models to a novel model derived in our SSc-ILD cohort that contained easily and commonly measured variables. Objective radiologic features (eg, fibrosis score) are not routinely measured in clinical practice, and we, therefore, did not include such variables in our analysis. We similarly did not include subjective patientreported symptoms (eg, severity of gastroesophageal reflux) or patient questionnaires (eg, dyspnea score) since these variables do not have a standard approach to measurement and are not routinely quantified in clinical practice. Future studies may be useful to evaluate these variables, however, mortality risk prediction tools that include these measurements are less likely to be clinically useful. In conclusion, prognostication in fibrotic ILD would be simplified by the ability to use a single mortality risk prediction tool that is accurate in all fibrotic ILD subtypes. We show that the modified du Bois index has good discrimination and calibration for the prediction of 1-year mortality in SSc-ILD and that discrimination is also acceptable for the CPI, ILD-GAP, and original du Bois indexes. Our study provides objective data that indicate mortality prediction does not depend on specific SSc-ILD subtypes, thus simplifying the approach to prognostication in SSc-ILD. Additional studies are required to determine how to translate these models into specific mortality estimates in SSc-ILD and how these estimates should then be used in clinical practice. Future studies are also needed to determine whether novel mortality risk prediction tools can substantially improve prognostication in patients with SSc-ILD. 1274 Original Research [ 148 # 5 CHEST NOVEMBER 2015 ]

Acknowledgments Author contributions: C. J. R. had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis, including and especially any adverse effects. C. J. R., D. O., J. V. D., F. S., C. J. H., D. M., J. L., and P. G. W. contributed substantially to the study design, data analysis and interpretation, and the writing of the manuscript. Conflict of interest: None declared. Role of sponsors: The sponsors had no role in study design, analysis, interpretation, or writing of the manuscript. Other contributions: We would like to acknowledge and thank the patients who allowed us to conduct this research. References 1. Rubin LJ. Introduction: diagnosis and management of pulmonary arterial hypertension: ACCP evidence-based clinical practice guidelines. Chest. 2004 ;126(1_suppl):7S-10S. 2. Benan M, Hande I, Gul O. The natural course of progressive systemic sclerosis patients with interstitial lung involvement. Clin Rheumatol. 2007 ;26(3):349-354. 3. White B. Interstitial lung disease in scleroderma. Rheum Dis Clin North Am. 2003 ;29(2):371-390. 4. Steen VD, Medsger TA. Changes in causes of death in systemic sclerosis, 1972-2002. Ann Rheum Dis. 2007 ;66(7): 940-944. 5. Winstone TA, Assayag D, Wilcox PG, et al. Predictors of mortality and progression in scleroderma-associated interstitial lung disease: a systematic review. Chest. 2014 ;146(2):422-436. 6. Raghu G, Collard HR, Egan JJ, et al ; ATS/ ERS/JRS/ALAT Committee on Idiopathic Pulmonary Fibrosis. An official ATS/ ERS/JRS/ALAT statement: idiopathic pulmonary fibrosis: evidence-based guidelines for diagnosis and management. Am J Respir Crit Care Med. 2011 ;183(6): 788-824. 7. Ley B, Ryerson CJ, Vittinghoff E, et al. A multidimensional index and staging system for idiopathic pulmonary fibrosis. Ann Intern Med. 2012 ;156(10):684-691. 8. Wells AU, Desai SR, Rubens MB, et al. Idiopathic pulmonary fibrosis: a composite physiologic index derived from disease extent observed by computed tomography. Am J Respir Crit Care Med. 2003 ;167(7):962-969. 9. du Bois RM, Weycker D, Albera C, et al. Ascertainment of individual risk of mortality for patients with idiopathic pulmonary fibrosis. Am J Respir Crit Care Med. 2011 ;184(4):459-466. 10. du Bois RM, Albera C, Bradford WZ, et al. 6-Minute walk distance is an independent predictor of mortality in patients with idiopathic pulmonary fibrosis. Eur Respir J. 2014 ;43(5):1421-1429. 11. Ryerson CJ, Vittinghoff E, Ley B, et al. Predicting survival across chronic interstitial lung disease: the ILD-GAP model. Chest. 2014 ;145(4):723-728. 12. van den Hoogen F, Khanna D, Fransen J, et al. 2013 classification criteria for systemic sclerosis: an American College of Rheumatology/European League against Rheumatism collaborative initiative. Arthritis Rheum. 2013 ;65(11):2737-2747. 13. Preliminary criteria for the classification of systemic sclerosis (scleroderma). Subcommittee for scleroderma criteria of the American Rheumatism Association Diagnostic and Therapeutic Criteria Committee. Arthritis Rheum. 1980 ;23(5):581-590. 14. Travis WD, Costabel U, Hansell DM, et al ; ATS/ERS Committee on Idiopathic Interstitial Pneumonias. An official American Thoracic Society/European Respiratory Society statement: Update of the international multidisciplinary classification of the idiopathic interstitial pneumonias. Am J Respir Crit Care Med. 2013 ;188(6):733-748. 15. Macintyre N, Crapo RO, Viegi G, et al. Standardisation of the single-breath determination of carbon monoxide uptake in the lung. Eur Respir J. 2005 ; 26 (4):720-735. 16. Miller MR, Hankinson J, Brusasco V, et al ; ATS/ERS Task Force. Standardisation of spirometry. Eur Respir J. 2005 ;26(2): 319-338. 17. Wanger J, Clausen JL, Coates A, et al. Standardisation of the measurement of lung volumes. Eur Respir J. 2005 ; 26 (3):511-522. 18. ATS Committee on Proficiency Standards for Clinical Pulmonary Function Laboratories. ATS statement: guidelines for the six-minute walk test. Am J Respir Crit Care Med. 2002 ;166(1):111-117. 19. King TE Jr, Albera C, Bradford WZ, et al ; INSPIRE Study Group. Effect of interferon gamma-1b on survival in patients with idiopathic pulmonary fibrosis (INSPIRE): a multicentre, randomised, placebo-controlled trial. Lancet. 2009 ;374(9685):222-228. 20. Raghu G, Brown KK, Bradford WZ, et al ; Idiopathic Pulmonary Fibrosis Study Group. A placebo-controlled trial of interferon gamma-1b in patients with idiopathic pulmonary fibrosis. N Engl J Med. 2004 ;350(2):125-133. 21. Goh NS, Desai SR, Veeraraghavan S, et al. Interstitial lung disease in systemic sclerosis: a simple staging system. Am J Respir Crit Care Med. 2008 ;177(11):1248-1254. 22. De Santis M, Bosello SL, Peluso G, et al. Bronchoalveolar lavage fluid and progression of scleroderma interstitial lung disease. Clin Respir J. 2012 ;6(1): 9-17. 23. Moore OA, Goh N, Corte T, et al. Extent of disease on high-resolution computed tomography lung is a predictor of decline and mortality in systemic sclerosis-related interstitial lung disease. Rheumatology (Oxford). 2013 ;52(1):155-160. journal.publications.chestnet.org 1275