Chest ACCP Career Connection
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Guest Access | Sign In via User Name/Password
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Wasfi, Y. S.
Right arrow Articles by Maier, L. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Wasfi, Y. S.
Right arrow Articles by Maier, L. A.
(Chest. 2006;129:1234-1245.)
© 2006 American College of Chest Physicians

A New Tool To Assess Sarcoidosis Severity*

Yasmine S. Wasfi, MD, PhD; Cecile S. Rose, MD, MPH; James R. Murphy, PhD; Lori J. Silveira, MS; Jan C. Grutters, MD, PhD; Yoshikazu Inoue, MD, PhD; Marc A. Judson, MD, FCCP and Lisa A. Maier, MD, MSPH, FCCP

* From the Pulmonary, Allergy and Critical Care Division (Dr. Wasfi), Department of Medicine, University of Pennsylvania School of Medicine, Philadelphia, PA; the Division of Environmental and Occupational Health Sciences (Drs. Rose and Maier, and Ms. Silveira), Department of Medicine, and the Division of Biostatistics (Dr. Murphy), National Jewish Medical and Research Center, Denver, CO; Heart Lung Center Utrecht (Dr. Grutters), Department of Pulmonology, St. Antonius Hospital, Utrecht, the Netherlands; the Department of Diffuse Lung Diseases and Respiratory Failure (Dr. Inoue), Clinical Research Center, National Hospital Organization Kinki-Chuo Chest Medical Center, Sakai, Osaka, Japan; and the Division of Pulmonary and Critical Care Medicine (Dr. Judson), Department of Medicine, Medical University of South Carolina, Charleston, SC.

Correspondence to: Lisa A. Maier, MD, MSPH, FCCP, National Jewish Medical and Research Center, 1400 Jackson St, Room G-216, Denver, CO 80206; e-mail: maierl{at}njc.org

Abstract

Study objectives: Sarcoidosis is a granulomatous disorder primarily affecting the lung, but with frequent extrapulmonary organ involvement. There are no comprehensive scoring systems for sarcoidosis disease severity. Our goal was to develop and validate an objective and comprehensive sarcoidosis disease severity scoring system.

Design: Three sarcoidosis experts reviewed clinical data on 104 patients with biopsy-confirmed sarcoidosis. Each expert independently scored disease severity using a visual analog scale. Interrater agreement was assessed. Univariate analysis was performed, and those variables with p values ≤ 0.25 were used in backward regression multivariable analysis. A model was obtained including variables with a p value of ≤ 0.15 to predict severity scores. This model was subsequently validated using an independent panel of three additional international experts.

Setting: Granuloma clinic at National Jewish Medical and Research Center.

Patients: A total of 104 patients with biopsy-confirmed sarcoidosis.

Interventions: None.

Measurements and results: Pairwise assessment of interrater agreement yielded high degrees of correlation with Spearman correlation coefficients of 0.86 to 0.89 and an intraclass correlation coefficient of 0.87. Univariate analysis showed that smoking status, immunosuppressive therapy, percent predicted for diffusing capacity of the lung for carbon monoxide (DLCO), FEV1, FVC, and total lung capacity, FEV1/FVC ratio, disease duration, sites of organ involvement, and African-American race were associated with mean severity score. The multivariable model included cardiac and neurologic involvement, current therapy with noncorticosteroid immunosuppressive agents, DLCO percent predicted, FEV1/FVC ratio, African-American race, FVC percent predicted, and skin involvement. This model was validated using additional reviewer scores yielding Spearman correlation coefficients of 0.66 to 0.76 and an intraclass correlation coefficient of 0.74.

Conclusions: We derived an objective disease severity scoring system that incorporates data on demographics, pulmonary function, and organ involvement to produce a whole-body sarcoidosis assessment. This preliminary tool has potential applicability in the assessment of disease severity in sarcoidosis research.

Key Words: granuloma • health status indicators • interstitial lung diseases • pulmonary sarcoidosis

Sarcoidosis is a systemic granulomatous disease of unknown etiology that primarily affects the lungs. A challenging aspect of studying and clinically managing this disease is its phenotypic diversity.1 Sarcoidosis can produce granulomatous inflammation in virtually any organ, hampering the assessment of disease severity. In addition, seemingly minor amounts of inflammation in some organs, such as the heart, may have devastating consequences, while more extensive inflammation at other sites, such as the skin, is usually not fatal. Existing severity scoring systems for sarcoidosis and other interstitial lung diseases have limited applicability in the assessment of sarcoidosis disease severity, as they do not incorporate extrapulmonary sites of organ involvement and/or have not been validated in an adult sarcoidosis population.

Most sarcoidosis severity scoring systems have relied on the chest radiograph Scadding stage, as defined in Table 1 .2 Associations have been found between the Scadding staging system and disease course, with the likelihood of spontaneous remission decreasing with increasing Scadding stage.234567 In addition, lung function abnormalities have been observed in only 20% of patients with stage I disease, compared with 40 to 70% of patients with stage II, III, or IV disease.289 Muers et al10 designed a more detailed assessment of radiographic severity with a score that incorporated the following four categories of abnormalities, each of which was graded as to extent and profusion: reticulonodular shadows; mass opacities; confluence; and fibrosis. They found significant correlations between the reticulonodular and fibrosis scores and the dyspnea score, spirometry values, and diffusing capacity of the lung for carbon monoxide (DLCO) at several time points, with a higher score correlating with worse dyspnea and lower lung function. A shortcoming of this system is that it requires establishing two readers with acceptable levels of interobserver reliability, which may limit the generalization of its use. Both the Scadding and Muers radiographic scores are limited in their assessment of sarcoidosis severity by the disparities that may be seen between chest radiograph appearance and functional impairment, and by the focus only on pulmonary disease.


View this table:
[in this window]
[in a new window]

  		Table 1.. Distribution of Sarcoidosis Cases by Demographics, Smoking Status, Scadding Stage, and Therapy

 
Two previous studies have devised scoring systems specifically to follow sarcoidosis disease severity over time. Gedalia and coworkers11 developed a global disease severity score to assess the efficacy of methotrexate as a steroid-sparing agent in children with sarcoidosis. The authors assigned a point value to each of nine sites of organ involvement and/or symptoms. The British Thoracic Society12 developed a relativistic sarcoidosis severity score that focused exclusively on the lung for its study of long-term corticosteroid treatment of sarcoidosis. This score assigned point values of + 1, –1, or 0, respectively, to indicate improvement, deterioration, or no change in the categories of lung function, breathlessness, and chest radiographic evidence of fibrosis. In both of these studies, scores were useful in discriminating among treatment groups. However, neither score provided a multiorgan assessment of disease severity in an adult sarcoidosis population.

Our goal was to devise and validate a comprehensive sarcoidosis severity scoring system that would incorporate readily available objective clinical parameters at a single time point, including the extent of pulmonary disease, the extent of extrapulmonary organ involvement, patient demographic variables, disease duration, and treatment. We wanted to create an instrument that would be straightforward to use with limited clinical information, and that would not rely on more subjective assessments such as patient symptoms. We used the opinion of an expert panel, expressed as a visual analog score of global sarcoidosis severity, as our "gold standard." Clinical parameters were then assessed for their contribution to a model that would predict this score. The score was subsequently validated using an international expert panel. The multivariable model provides a potential research tool for measuring overall disease severity in sarcoidosis at a single time point, using objective clinical data.

Methods and Materials

Study Population
Sarcoidosis patients (n = 104) were recruited from the granuloma clinic at National Jewish Medical and Research Center between January 2002 and August 2004. All patients met the diagnostic criteria for sarcoidosis established by the American Thoracic Society consensus panel.2 All study subjects were adults between the ages of 18 and 70 years, and provided their informed consent. The consent and the protocol were approved by the National Jewish Medical and Research Center Institutional Review Board.

Clinical Assessment
Study participants completed a modified version of the American Thoracic Society respiratory questionnaire, which includes information on smoking status, gender, self-reported race and ethnicity, age, and medication use.13 Information about medication use included treatment with prednisone, whether that treatment was current or in the past, and the current use of noncorticosteroid immunosuppressive agents, including methotrexate, cyclophosphamide, cyclosporine, azathioprine, chlorambucil, pentoxifylline, and thalidomide. Medical records were reviewed to extract clinical indicators of disease severity, including pulmonary function parameters (ie, FEV1, FVC, FEV1/FVC ratio, DLCO, and total lung capacity [TLC] as raw numbers and percent predicted) and the assessment of 15 sites of organ involvement using the A Case Control Etiologic Study of Sarcoidosis (ACCESS) assessment instrument.14 This instrument defines the criteria for the definite, probable, and possible involvement of each of the following organs or organ systems: lungs; nervous system (CNS or peripheral); nonthoracic lymph node; kidney; heart; skin; eye; liver; bone marrow; spleen; bone/joint; ear/nose/throat; parotid/salivary gland; muscles; and calcium metabolism. Individuals with definite or probable involvement, as defined by these criteria, were considered to have involvement at that site. During chart review, the biopsy date was also recorded and used to define the date of diagnosis and to calculate disease duration in years. Disease duration was defined as the difference between the biopsy date and the date of enrollment.

For each subject, a chest radiograph obtained within 1 year of study enrollment was reviewed and staged by two pulmonologists using the Scadding staging system.2 If discrepant interpretations occurred, a third reviewer was used to determine the chest radiographic stage.

Panel Assessments of Sarcoidosis Severity
The initial expert panel consisted of three sarcoidosis experts from National Jewish Medical and Research Center (L.M., C.R., and Y.W.), who met to review and score cases. The clinical history for each subject was summarized based on a brief review of the medical record, in most cases by the subject’s treating physician. The summaries included information on disease manifestations, biopsy site, treatment efforts (ie, therapeutic agent, dose, and duration), therapeutic response, general clinical course, and smoking history, but not race or ethnicity. Each initial panel member was also provided with a summary data sheet containing the pulmonary function and organ involvement data described above and the chest radiograph Scadding stage. Each panel member then independently rated disease severity using a visual analog scale (VAS) ranging from 0 to 10. This scale consisted of a 5-inch horizontal line, with one end marked with a 0, which was defined as an asymptomatic individual with disease discovered incidentally and not requiring treatment, and the other end marked with a 10, which was defined as the presence of severe and/or life-threatening end-organ dysfunction, for example, requiring consideration of transplantation. The distance (in inches) from the 0 mark to the panelist mark was measured and multiplied by 2 to obtain the value for that expert’s severity assessment.

In order to assess the validity of the disease severity model derived from the initial panel scores, a panel of three independent outside sarcoidosis specialists (J.G., Y.I., and M.J.) reviewed the same clinical information for each subject that had been presented to the initial panel. Specifically, the sessions during which the initial panel reviewed the cases were recorded and transcribed. The transcribed information consisted of the clinical history as presented by the clinician and as described above, edited into paragraph form. This was provided to the validation panel reviewers, along with the summary data sheet described above. The validation panel reviewers then scored the subjects independently using a VAS, as the initial reviewers had.

Statistical Analysis
Descriptive statistics were performed for demographic and clinical characteristics, and for the visual analog scores of each initial panel expert. We used three methods for the pairwise assessment of interrater agreement among the initial panel experts. First, differences between the visual analog scores for each pair of reviewers were calculated; a Student t test was used to assess whether the mean difference for each pair of reviewers was different from 0. Second, Spearman correlation coefficients were calculated for each pair of reviewers. Finally, an intraclass correlation coefficient was calculated for all three reviewers to assess the overall agreement among the three reviewers. This statistic is a measure of reliability and is equivalent to the appropriate average of the correlations among all pairs of tests. These pairwise comparisons were also performed for the validation panel experts.

The statistical analysis, which was performed in order to transform the highly consistent but subjective initial panel scores into a severity model using objective clinical measures, was completed in two steps. First, univariate analysis was used to compare the clinical variables and the initial panel mean severity score. For normally distributed continuous variables (ie, TLC percent predicted, DLCO percent predicted, FEV1 percent predicted, FVC percent predicted, and age), Pearson correlation coefficients were performed. For nonnormally distributed continuous variables (ie, FEV1/FVC ratio and disease duration), Spearman correlation coefficients were derived. For dichotomous variables (ie, gender, smoking status [ever/never], current steroid use [yes/no], noncorticosteroid immunosuppressive use within 1 month [yes/no], sites of organ involvement [present/absent], and ethnicity [Hispanic, non-Hispanic]), a Student t test was performed. For categoric variables (ie, chest radiograph Scadding stage, and race), a one-way analysis of variance was performed. Due to small cells, these two variables were analyzed as dichotomous; it is this analysis that is reported in the "Results" section.

Backward multiple linear regression was then performed (SAS, version 8.1; SAS Institute; Cary, NC) to derive the model for disease severity score. The initial panel mean score was the dependent variable, and variables with a p value of ≤ 0.25 in the univariate analyses were the independent variables. In addition, categories of organ involvement in which < 10% of the total group was affected were excluded, unless there was a compelling clinical reason to include this site. Variables with a significant nonnormal distribution were log-transformed prior to inclusion in the regression model. The potential for multicollinearity among the spirometric variables was considered. It is true that these variables are correlated, with correlation coefficients of 0.85 (FEV1 and FVC), 0.61 (FEV1 and FEV1/FVC ratio), and 0.18 (FVC and FEV1/FVC ratio). Among these, the pair that raises some concern for multicollinearity is FEV1 and FVC. However, in the context of sarcoidosis, which may present with either restrictive or obstructive lung disease, each of these parameters may contribute independently to an overall characterization of lung function. In addition, our expectation was that should more than one of these variables actually represent the same aspect of variance in disease severity, one of them would be removed from the model. Therefore, we included all three variables in the regression. We also performed an evaluation of variance inflation factors for the variables in the final models to evaluate for the possibility of multicollinearity. As a general rule, variables with a variance inflation factor of > 10 may be collinear with, and therefore replaced by, a linear combination of other independent variables. To remain in the final model, variables had to have a p value of ≤ 0.15.

The final step in the analysis was the validation of scores derived from the final model by comparing them to the validation panel scores. This was achieved using the same techniques described above for the pairwise assessment of agreement within the initial panel. Specifically, differences between the model-predicted scores and the scores of each validation panel reviewer were calculated; a Student t test was used to assess whether the mean difference from the model-predicted scores for each reviewer was different from 0. Second, Spearman correlation coefficients were calculated for the model-predicted scores and each reviewer’s scores. Finally, an intraclass correlation coefficient was calculated.

Results

Characterization of Study Population
The study population consisted of 39 men and 65 women. The demographic features of this group are presented in Table 1. Of note, the race distribution of sarcoidosis patients in this population reflects the local demographics for Colorado. The clinical characteristics of the patient population are summarized in Tables 1and 2 , and reflect a wide range of chest radiograph Scadding stage and organ involvement. All but one subject had pulmonary involvement, with a variety of lung function abnormalities. Approximately half of the subjects had a decreased FEV1 percent predicted, and 41% had a decreased FVC. The vast majority of subjects (88%) had a normal TLC, while 60% had a decreased diffusion capacity. There was a wide range of disease duration, from 1 month to > 22 years. The distribution was skewed toward shorter duration, with a mean time of 5.3 years and a median time of 4 years. At study enrollment, 59% of subjects were receiving prednisone, and 31% were receiving noncorticosteroid immunosuppressive agents, including methotrexate, cyclophosphamide, and azathioprine, as detailed in Table 1. We did not use anti-tumor necrosis factor agents for the treatment of sarcoidosis at the time of study initiation; therefore, these agents were not included in our questionnaire.


View this table:
[in this window]
[in a new window]

  		Table 2.. Organ Involvement in Sarcoidosis Cases*

 
Description of Rater Data
The distribution of scores of the three initial panel reviewers is demonstrated in Figure 1 . Three statistical tests were performed to compare these scores. First, the distribution of pairwise differences between the reviewer scores was evaluated. The differences were all normally distributed with a mean not statistically different from 0. The differences were within 1 point (± 1) in 58 to 65% of subjects, and within 2 points (± 2) in 80 to 89% of subjects. Second, the pairwise assessment of interrater agreement yielded high degrees of correlation, ranging from 0.86 to 0.89, as demonstrated in Figure 2 . Finally, the intraclass correlation coefficient among the initial reviewers was similarly high at 0.87.


Figure 1
View larger version (12K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1.. Distribution of initial panel reviewer VASs. NJ = National Jewish Medical and Research Center.

 

Figure 2
View larger version (23K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2.. Pairwise correlations of initial panel scores with 90% confidence intervals for reviewers 1 and 2 (top left, a), reviewers 2 and 3 (top right, b), and reviewers 1 and 3 (bottom, c). See Figure 1 legend for expansion of abbreviation.

 
The distribution of scores of the three validation panel reviewers is demonstrated in Figure 3 . The three statistical tests described above were used to compare these scores. The distributions of pairwise differences among reviewer scores were all normally distributed, but for two of the pairs the mean was statistically different from 0 (actual values, 0.37 and 0.6). The differences were within 1 point (± 1) in 38 to 54% of subjects, and within 2 points in 78 to 81% of subjects. The pairwise assessment of interrater agreement again yielded high degrees of correlation with Spearman {rho} correlation coefficients of 0.73 to 0.80. Finally, the intraclass correlation coefficient was also high at 0.75.


Figure 3
View larger version (14K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3.. Distribution of validation panel reviewer VASs.

 
Derivation of Sarcoidosis Severity Score
Univariate Analysis: Univariate analyses were performed comparing the clinical variables with the mean score from the initial panel, using Student t test or analysis of variance for independent categoric variables or correlation coefficients for continuous independent variables. All of the variables evaluated and the p values for their associations with the initial panel mean score are summarized in Table 3 . We found significant associations with modest correlations between the initial panel mean severity score and the TLC percent predicted, DLCO percent predicted, FEV1 percent predicted, FVC percent predicted, FEV1/FVC ratio, race, smoking status, disease duration, current corticosteroid therapy, current or recent noncorticosteroid immunosuppressive therapy, cardiac involvement, neurologic involvement, and skin involvement. These were all of the variables that were initially entered into the multivariable analysis.


View this table:
[in this window]
[in a new window]

  		Table 3.. Results of Univariate Analyses Comparing Mean Initial Panel Score and Clinical Variables*

 
Multivariable Analysis: Backward regression analysis was used to derive the multivariable model, which is summarized in Table 4 . These results can also be summarized as the following equation:

Formula
where C = 1 if there is cardiac involvement, 0 if not; N = 1 if there is neurologic involvement, 0 if not; IS = 1 if receiving noncorticosteroid immunosuppression therapy, 0 if not; AA = 1 if the subject was African American, 0 if not; and skin = 1 if there was skin involvement, 0 if not.


View this table:
[in this window]
[in a new window]

  		Table 4.. Multivariable Model for Sarcoidosis Disease Severity Score*

 
The most statistically significant contributors to the final model were cardiac involvement, neurologic involvement, and the use of a noncorticosteroid immunosuppressive agent (listed above) within 30 days of study participation. These three variables and the DLCO percent predicted had the largest standardized parameter estimates, and therefore the greatest impact on the final model-predicted scores. Of note, the analysis of the final variables in this model revealed variance inflation factors far < 10, indicating no collinearity among the final variables.

Since full pulmonary function test results are often not available on some clinical subjects and many research subjects, whereas spirometry usually is, we also derived a version of the score from the same data, but excluding both TLC and DLCO percent predicted from the regression analysis. This modified version of the score is summarized in Table 4. These results can again be presented in the form of the following equation:

Formula
with variables C, N, IS and AA defined as above.

Aside from the excluded lung function variables, the model resulting from this analysis differed from the first in that skin involvement did not contribute significantly enough to remain in the model. The most statistically significant contributors to this model were similar to the first model, with cardiac involvement, neurologic involvement, and immunosuppression, but in this case they also included FVC percent predicted. The variables with the largest standardized parameter estimates were cardiac involvement, FVC percent predicted, and neurologic involvement. The evaluation of variance inflation factors again revealed no evidence of collinearity among the variables in this final model.

Validation of the Sarcoidosis Severity Score
The first model in Table 4 was used to calculate severity scores for each of the study subjects. These calculated scores were then compared to the scores obtained from the validation panel of outside reviewers by employing the same three analytic methods used to compare the scores of reviewers on the initial panel. First, the distribution of pairwise differences between the predicted scores and each reviewer’s scores was evaluated. The differences were all normally distributed. The mean of the differences was not statistically different from 0 for two of the three reviewers, with the third reviewer having a statistically different mean of 0.37. The differences were within 1 point (± 1) in 45 to 55% of subjects, and within 2 points (± 2) in 70 to 84% of subjects. Second, the pairwise assessment of agreement (Spearman {rho}) between the predicted scores and the reviewer scores yielded high degrees of correlation, ranging from 0.66 to 0.76. Finally, the intraclass correlation coefficient was 0.74. These analyses were repeated for the modified version of the score (Table 4), which were obtained with only spirometric lung function data. The results were similar, with the differences between the predicted and reviewer scores all being normally distributed; with means not statistically different from 0 except for the third reviewer, who had a mean difference again of 0.37. The differences were within 1 point (± 1) in 43 to 53% of subjects, and within 2 points (± 2) in 73 to 81% of subjects. Pairwise assessments of agreement (Spearman {rho}) between the predicted scores and reviewer scores were similar, with values of 0.66 to 0.76. Finally, the intraclass correlation coefficient was 0.73, again validating this modified version of the score.

Discussion

Summary
Sarcoidosis is a disease with great phenotypic diversity, with the potential for numerous sites of organ involvement, varying levels of disease activity, and varying degrees of long-term organ dysfunction resulting from periods of active inflammation and, potentially, fibrosis. Our purpose in this study was to create a research tool that would incorporate these many aspects of the disease into a single measure of severity that could be derived from readily available clinical information from a single time point. Such a severity scale, if found to be valid and robust across a diverse group of sarcoidosis clinicians, could prove to be useful in studies that examine genetic or other risk factors for disease severity.

Using a clinically diverse population of sarcoidosis subjects, we first demonstrated the utility of expert opinion as a "gold standard" measure in this disease by confirming very high levels of agreement among the scores assigned by the initial panel of reviewers. These highly consistent subjective scores were then used to derive a multivariable model for a severity score based on objective clinical data. This model was then validated by the comparison of model-predicted scores to scores assigned by a validation panel of outside experts who independently reviewed the same clinical data as the initial panel. Components of the multivariable model included current therapy with noncorticosteroid immunosuppressive agents, specific sites of organ involvement (ie, neurologic, cardiac, and skin), several parameters of lung function (ie, DLCO percent predicted, FVC percent predicted, and FEV1/FVC ratio) and African-American race. We then derived a second simplified model (excluding DLCO percent predicted and TLC percent predicted from the analysis), which was also validated by comparison to the scores assigned by the validation panel. Interestingly, this model was similar to the first, including noncorticosteroid immunosuppressive therapy; neurologic and cardiac involvement, but not skin involvement; two measures of lung function (FVC percent predicted and FEV1/FVC ratio); and African-American race.

Use of Expert Panel as "Gold Standard"
One of the greatest challenges in this endeavor was to define a "gold standard" for sarcoidosis severity. By analogy, researchers have previously developed severity scores for idiopathic pulmonary fibrosis (IPF), including the clinical, radiographic, and physiologic (CRP) scoring system1516 and the composite physiologic index (CPI).17 Each of these systems defined a "gold standard" for severity, consisting of histopathologic findings, survival, or extent of disease as determined by CT scan. The updated CRP scoring system and the CPI then determined the clinical, radiologic, and/or physiologic variables that best predicted the "gold standard" outcome. These "gold standard" outcomes are, however, not as applicable for the assessment of multisystem sarcoidosis, for several reasons. Sarcoidosis is fatal in only a small minority of patients, making survival an impractical measure of sarcoidosis disease severity, as there would be too few end points. Sarcoidosis patients are uncommonly diagnosed by video-assisted thoracoscopic or open lung biopsy; even if this material were available for review, there is no evidence to confirm a correlation between the extent of granulomatous inflammation seen on biopsy specimens and functional impairment. Similarly, extent of disease as determined by CT scan does not necessarily correlate with functional abnormalities in sarcoidosis. Finally, even if lung biopsy and chest CT findings were highly correlated with sarcoidosis severity, they only address the pulmonary manifestations of the disease. Thus, a novel, more comprehensive measure of severity was required.

The use of expert opinion to define the clinical components of a severity score is well-established. In fact, the original IPF CRP score by Watters et al15 was devised by a team of IPF experts and then was correlated with another score based on expert opinion of IPF histopathologic findings. Similarly, scoring systems derived based on expert opinion have been used in patients with cystic fibrosis,1819 and were subsequently validated using outcomes such as survival and sensitivity for changes in clinical course. Sarcoidosis severity assessments have been applied mainly in studies of treatment efficacy or genetic associations, and have included a variety of primary or secondary individual clinical measures to assess disease status. These assessments have included symptoms, chest radiograph appearance, lung function, laboratory studies of blood and BAL fluid, quality-of-life measures, changes in therapy, gallium scan findings, and miscellaneous measures such as the severity or extent of other organ involvement.202122232425262728293031 Other approaches have defined severity based on chronicity (resolving vs persistent disease), the occurrence of relapses, or whether the initial presentation was consistent with Lofgren syndrome, an acute, typically spontaneously resolving form of sarcoidosis.3233343536 Our approach integrates evidence-based and experience-based clinical variables that were readily accessible in our sarcoidosis patient population at a single time point, as well as potential confounders such as smoking status, to create the first comprehensive and quantitative tool to assess sarcoidosis severity. Of note, the information provided to the expert panels included all of the same categories of variables used by the authors of the original IPF CRP system, including clinical status (level of dyspnea in CRP), chest radiograph appearance, and physiologic data, including exercise physiology findings when available.

Another aspect of our use of expert opinion that deserves brief comment is the use of regression analysis based on scores from VASs in order to derive our final models. While we did not identify an example in the pulmonary literature, studies in other disciplines have used VASs as a tool to assess the severity of symptoms or illness, and then performed regression analyses on these data in order to determine the variables that predict these scores. For example, a study of patients with chronic hepatitis C evaluated predictors of patient-perceived severity by asking them to make an assessment of their own disease severity on a VAS. The authors then used a regression analysis to determine which of a number of demographic, clinical, and psychological factors predicted this perceived severity.37 A slightly different approach was used in a study38 in which patients were asked to rate their pain on a VAS at the time of their emergency department visit and again 1 week later. Regression analysis was then used to determine the predictors of persistent pain at 1 week, which was defined as a VAS score of > 3.38 Global assessments of disease status by clinicians have also been used.39 There is, therefore, ample precedent in the literature for the use of a VAS, as we have done in this study.

Putting Multivariable Model Variables Into Context: Do They Make Sense?
Our models both extend and diverge from previous investigations of sarcoidosis severity. In a number of studies assessing sarcoidosis outcome or progression, measures of lung function have been associated with disease severity. As in our study, Mana et al36 found a decreased FVC at the time of diagnosis (< 80% predicted) to be one of several variables that are predictive of disease persistence over ≤ 5 years of follow-up. A Danish study40 examined mortality in a cohort of 254 patients with pulmonary sarcoidosis who had been followed up for a median of 27 years and found that abnormal lung function at diagnosis (ie, FEV1, ≤ 50% predicted; TLC, ≤ 80% predicted; or FEV1/FVC ratio, < 70%) significantly increased the risk of death from sarcoidosis or sarcoidosis-related diseases. Our findings also emphasize the importance of DLCO percent predicted in assessing sarcoidosis severity, as it was a highly significant component of our multivariable model, in addition to FVC percent predicted and FEV1/FVC ratio. This is not surprising, especially when considering the contribution of DLCO to the original CRP and CPI scores for IPF. It is also interesting to note that among the spirometric variables, of the two that were most highly correlated, FEV1 and FVC percent predicted, only one remained in the model. Thus, as we expected, the regression analysis accounted for the similarity in what these variables were measuring by removing one from the model.

It is interesting that African-American race was a significant contributor to both models, even though race was not included in the data provided to reviewers during scoring. The association of African-American race with more severe disease again confirms what has been reported in the literature. Previous studies24142 have shown that African Americans have a higher rate of extrathoracic involvement, particularly with features associated with a poor prognosis such as lupus pernio, chronic uveitis, and cystic bone disease, along with more chronic disease and a higher rate of relapse.

It is noteworthy that the Scadding stage seen on chest radiographs was not significantly associated with the mean initial panel score in the univariate analysis. This is of particular interest, as a number of studies support the prognostic relevance of Scadding stage, demonstrating significant morbidity associated with stages II, III, and IV disease but only rare serious disease sequelae in stage I disease. Others have described4043 an inverse relationship between Scadding stage and the likelihood of spontaneous remission,2 as well as excess mortality in patients with higher disease stage seen on the chest radiograph. However, other studies3641 have not found chest radiograph appearance to be a significant predictor of disease persistence. It is possible that in these studies other variables such as lung function may have correlated highly with chest radiograph appearance, making its independent contribution to the outcome of interest negligible. Chest radiograph stage also may have not been associated in our univariate analysis because it is not a sensitive enough indicator of the underlying pulmonary abnormalities to provide a consistent measure of severity. If this were true, one might expect a chest CT scan, particularly high-resolution images, to be a better indicator of disease severity. Indeed, one study by Drent et al44 suggests that this is the case. These investigators44 used a high-resolution CT (HRCT) scan scoring system to evaluate the relationship between CT imaging and sarcoidosis disease severity, as measured by lung physiology, and found that several measures of lung function and exercise physiology were significantly correlated with the HRCT scores but not with the chest radiographic stage. It is possible that the inclusion of HRCT findings as a variable in our study might have altered the multivariable models; however, our goal was to create a scoring system based on commonly available clinical variables, and HRCT scans are not uniformly obtained in the clinical evaluation of patients with sarcoidosis.

Potential Limitations
To our knowledge, this study is the first to incorporate the full clinical spectrum of sarcoidosis into a single validated scoring system based on objective clinical variables. However, this severity score is preliminary and does have some limitations. One potential limitation of the primary model is that it includes some data (ie, DLCO) that may not be readily available in all patients or research subjects. We anticipated this potential problem by deriving and validating a second model using only spirometry findings, which yielded similar results when compared with the validation panel. The data suggest that the simplified spirometry model can be used to almost as good effect as the full model including data from full pulmonary function tests.

Although the scores of our validation panel correlated highly with our model-derived severity scores, the scores of one of our validation panel reviewers were statistically significantly different from the model-predicted scores. Our validation panel was international, so this difference may reflect differences in sarcoidosis clinical presentations around the world, or different physician assessment patterns internationally. For example, there may be specific clinical presentations or test results that were not considered in our model that are important in the assessment of severity by physicians in other parts of the world. In fact, the highest correlation between our validation panel reviewers and the model-predicted scores was with an American reviewer. This suggests that the multivariable models derived in this study may be best suited for use in US sarcoidosis populations or by US raters, and that further validation in European and Asian sarcoidosis populations may be needed to further strengthen the global application of this tool.

This score was designed to provide an assessment of subjects at a single time point and thus is, as yet, not proven as a measure of disease progression over time. Sarcoidosis is typically a disease that progresses over the course of months to years. We did not examine the natural history of or longitudinal changes in the clinical variables. We did attempt to account for the influence of time on disease by including disease duration in the analysis.

A potential bias in the initial scoring was that all of the reviewers on the initial expert panel were pulmonologists, and therefore potentially predisposed to give lung function abnormalities greater weight in their scoring. However, each of these reviewers also has great breadth of experience in the care of sarcoidosis patients, including an understanding of the contribution of other sites of organ involvement to disease severity.

While an important goal of this study was to incorporate the many different potential sites of organ involvement into a comprehensive score, there were several organs for which only a small number of subjects were affected. In some cases, this meant that sites of involvement with the requisite p value in the univariate analysis were not included in the multivariable analysis, due to the instability introduced into the model by small numbers. We therefore lost the ability to evaluate the contribution of those organs. In addition, since we defined organ involvement in terms of presence or absence, as defined by the ACCESS14 instrument, we could not evaluate the relative importance of different types of involvement of a particular site. For example, a patient with cardiac involvement resulting in heart failure would clearly be clinically assessed as having more severe disease than one with an asymptomatic conduction abnormality, but for the purposes of our analysis, these individuals would have been grouped together.

Finally, by including treatment in the model we have precluded its use as a clinical tool to guide therapeutic decision making or as a tool to predict prognosis. Our primary purpose was to create a reliable tool for sarcoidosis research. A similar approach, with assessment of disease progression over time would be reasonable to derive a more clinically applicable score.

Potential Applications
Before this preliminary tool can be more broadly applied, it will need to undergo further validation. The most important issue is the applicability of this tool to different populations. Our population was recruited from a tertiary care medical center, and so likely consisted of more severely affected individuals. In addition, only 15% of our population was African American, whereas in populations recruited from other sites around the United States, this percentage is likely to be much higher. Since this group is often clinically distinct, with more severe disease, the evaluation of the instrument in a larger population of African Americans will be important. Similarly, European and Japanese sarcoidosis populations present differently with varied pulmonary and other organ involvement. Those populations therefore need to be evaluated separately as well, with specific investigation and evaluation of the factors that those physicians consider most important in a global assessment of severity. Future refinements of this system should specifically investigate the types, doses, and duration of therapy; more detailed descriptions of organ involvement; CT imaging parameters; and, importantly, changes in disease status over time. The availability of additional data regarding disease progression and survival will also provide the opportunity to validate our models against these outcomes.

With this additional validation, we anticipate that this tool will be applicable primarily in studies of factors that may contribute to the severity of sarcoidosis. For example, several gene variants, particularly in the major histocompatibility complex have been associated with sarcoidosis, and with specific forms of sarcoidosis. The HLA-DRB1*03/HLA-DQB1*0201 haplotype has been associated with milder forms of disease, particularly Lofgren syndrome, while the HLA-DRB1*15/HLA-DQB1*0602 haplotype has been associated with more severe disease, as defined by chronicity and specific clinical findings such as uveitis, chest radiograph stage, and lung function abnormalities.45 With this scale, it should be possible to better map the severity of disease in relation to genotype.

Acknowledgements

The authors thank Bevin Luna, Juliana Barnard, Trudi Madigan, Deborah Corliss, and Janice Herrell for their role in enrolling patients; Michele Cooper for technical assistance; Lee Newman for insightful discussion and careful review of the manuscript; Mary Solida for her tireless care of the patients; and the patients for their willingness to participate in this work.

Footnotes

Abbreviations: ACCESS = A Case Control Etiologic Study of Sarcoidosis; CPI = composite physiologic index; CRP = clinical, radiographic, and physiologic; DLCO = diffusing capacity of the lung for carbon monoxide; HRCT = high-resolution CT; IPF = idiopathic pulmonary fibrosis; TLC = total lung capacity; VAS = visual analog scale

This research was supported by the Parker B. Francis Fellowship and by National Institutes of Health grant M01 RR00051. None of the authors are involved in any organization with a direct financial interest in the subject of the manuscript.

Received for publication July 27, 2005. Accepted for publication November 4, 2005.

References

  1. Newman, LS, Rose, CS, Maier, LA (1997) Sarcoidosis. N Engl J Med 336,1224-1234[Free Full Text]
  2. American Thoracic Society, European Respiratory Society, World Association of Sarcoidosis and Other Granulomatous Disorders.. Statement on sarcoidosis: joint statement of the American Thoracic Society (ATS), the European Respiratory Society (ERS) and the World Association of Sarcoidosis and Other Granulomatous Disorders (WASOG) adopted by the ATS Board of Directors and by the ERS Executive Committee, February 1999. Am J Respir Crit Care Med 1999;160,736-755[Free Full Text]
  3. Scadding, JG Prognosis of intrathoracic sarcoidosis in England: a review of 136 cases after five years’ observation. BMJ 1961;5261,1165-1172[ISI][Medline]
  4. Siltzbach, LE, James, DG, Neville, E, et al Course and prognosis of sarcoidosis around the world. Am J Med 1974;57,847-852[CrossRef][ISI][Medline]
  5. Romer, FK Presentation of sarcoidosis and outcome of pulmonary changes. Dan Med Bull 1982;29,27-32[ISI][Medline]
  6. Hillerdal, G, Nou, E, Osterman, K, et al Sarcoidosis: epidemiology and prognosis: a 15-year European study. Am Rev Respir Dis 1984;130,29-32[ISI][Medline]
  7. Neville, E, Walker, AN, James, DG Prognostic factors predicting the outcome of sarcoidosis: an analysis of 818 patients. Q J Med 1983;52,525-533[ISI][Medline]
  8. Keogh, BA, Hunninghake, GW, Line, BR, et al The alveolitis of pulmonary sarcoidosis: evaluation of natural history and alveolitis-dependent changes in lung function. Am Rev Respir Dis 1983;128,256-265[ISI][Medline]
  9. Hunninghake, GW, Gilbert, S, Pueringer, R, et al Outcome of the treatment for sarcoidosis. Am J Respir Crit Care Med 1994;149,893-898[Abstract]
  10. Muers, MF, Middleton, WG, Gibson, GJ, et al A simple radiographic scoring method for monitoring pulmonary sarcoidosis: relations between radiographic scores, dyspnoea grade and respiratory function in the British Thoracic Society Study of Long-Term Corticosteroid Treatment. Sarcoidosis Vasc Diffuse Lung Dis 1997;14,46-56[Medline]
  11. Gedalia, A, Molina, JF, Ellis, GS, Jr, et al Low-dose methotrexate therapy for childhood sarcoidosis. J Pediatr 1997;130,25-29[CrossRef][ISI][Medline]
  12. Gibson, GJ, Prescott, RJ, Muers, MF, et al British Thoracic Society Sarcoidosis study: effects of long term corticosteroid treatment. Thorax 1996;51,238-247[Abstract]
  13. Ferris, BG Epidemiology Standardization Project (American Thoracic Society). Am Rev Respir Dis 1978;118,1-120[ISI][Medline]
  14. Judson, MA, Baughman, RP, Teirstein, AS, et al Defining organ involvement in sarcoidosis: the ACCESS proposed instrument; ACCESS Research Group—a case control etiologic study of sarcoidosis. Sarcoidosis Vasc Diffuse Lung Dis 1999;16,75-86[ISI][Medline]
  15. Watters, LC, King, TE, Schwarz, MI, et al A clinical, radiographic, and physiologic scoring system for the longitudinal assessment of patients with idiopathic pulmonary fibrosis. Am Rev Respir Dis 1986;133,97-103[ISI][Medline]
  16. King, TE, Jr, Tooze, JA, Schwarz, MI, et al Predicting survival in idiopathic pulmonary fibrosis: scoring system and survival model. Am J Respir Crit Care Med 2001;164,1171-1181[Abstract/Free Full Text]
  17. 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;167,962-969[Abstract/Free Full Text]
  18. Matouk, E, Ghezzo, RH, Gruber, J, et al Internal consistency reliability and predictive validity of a modified N. Huang clinical scoring system in adult cystic fibrosis patients. Eur Respir J 1997;10,2004-2013[Abstract]
  19. Matouk, E, Ghezzo, RH, Gruber, J, et al Construct and longitudinal validity of a modified Huang clinical scoring system in adult cystic fibrosis patients. Eur Respir J 1999;13,552-559[Abstract]
  20. Pietinalho, A, Ohmichi, M, Lofroos, AB, et al The prognosis of pulmonary sarcoidosis in Finland and Hokkaido, Japan: a comparative five-year study of biopsy-proven cases. Sarcoidosis Vasc Diffuse Lung Dis 2000;17,158-166[ISI][Medline]
  21. Pietinalho, A, Tukiainen, P, Haahtela, T, et al Oral prednisolone followed by inhaled budesonide in newly diagnosed pulmonary sarcoidosis: a double-blind, placebo-controlled multicenter study; Finnish Pulmonary Sarcoidosis Study Group. Chest 1999;116,424-431[Abstract/Free Full Text]
  22. du Bois, RM, Greenhalgh, PM, Southcott, AM, et al Randomized trial of inhaled fluticasone propionate in chronic stable pulmonary sarcoidosis: a pilot study. Eur Respir J 1999;13,1345-1350[Abstract]
  23. Gupta, SK Treatment of sarcoidosis patients by steroid aerosol: a ten-year prospective study from Eastern India. Sarcoidosis 1989;6,51-54[Medline]
  24. Spiteri, MA, Newman, SP, Clarke, SW, et al Inhaled corticosteroids can modulate the immunopathogenesis of pulmonary sarcoidosis. Eur Respir J 1989;2,218-224[Abstract]
  25. Erkkila, S, Froseth, B, Hellstrom, PE, et al Inhaled budesonide influences cellular and biochemical abnormalities in pulmonary sarcoidosis. Sarcoidosis 1988;5,106-110[Medline]
  26. Baltzan, M, Mehta, S, Kirkham, TH, et al Randomized trial of prolonged chloroquine therapy in advanced pulmonary sarcoidosis. Am J Respir Crit Care Med 1999;160,192-197[Abstract/Free Full Text]
  27. Baughman, RP, Winget, DB, Lower, EE Methotrexate is steroid sparing in acute sarcoidosis: results of a double blind, randomized trial. Sarcoidosis Vasc Diffuse Lung Dis 2000;17,60-66[ISI][Medline]
  28. Lower, EE, Baughman, RP Prolonged use of methotrexate for sarcoidosis. Arch Intern Med 1995;155,846-851[Abstract]
  29. Pia, G, Pascalis, L, Aresu, G, et al Evaluation of the efficacy and toxicity of the cyclosporine A-flucortolone-methotrexate combination in the treatment of sarcoidosis. Sarcoidosis Vasc Diffuse Lung Dis 1996;13,146-152[Medline]
  30. Muller-Quernheim, J, Kienast, K, Held, M, et al Treatment of chronic sarcoidosis with an azathioprine/prednisolone regimen. Eur Respir J 1999;14,1117-1122[Abstract]
  31. Baughman, RP, Judson, MA, Teirstein, AS, et al Thalidomide for chronic sarcoidosis. Chest 2002;122,227-232[Abstract/Free Full Text]
  32. Grunewald, J, Eklund, A, Olerup, O Human leukocyte antigen class I alleles and the disease course in sarcoidosis patients. Am J Respir Crit Care Med 2004;169,696-702[Abstract/Free Full Text]
  33. Sharma, SK, Balamurugan, A, Pandey, RM, et al Human leukocyte antigen-DR alleles influence the clinical course of pulmonary sarcoidosis in Asian Indians. Am J Respir Cell Mol Biol 2003;29,225-231[Abstract/Free Full Text]
  34. Seitzer, U, Gerdes, J, Muller-Quernheim, J Evidence for disease phenotype associated haplotypes (DR.TNF) in sarcoidosis. Sarcoidosis Vasc Diffuse Lung Dis 2001;18,279-283[ISI][Medline]
  35. Berlin, M, Fogdell-Hahn, A, Olerup, O, et al HLA-DR predicts the prognosis in Scandinavian patients with pulmonary sarcoidosis. Am J Respir Crit Care Med 1997;156,1601-1605[Abstract/Free Full Text]
  36. Mana, J, Salazar, A, Pujol, R, et al Are the pulmonary function tests and the markers of activity helpful to establish the prognosis of sarcoidosis? Respiration 1996;63,298-303[ISI][Medline]
  37. Constant, A, Castera, L, Quintard, B, et al Psychosocial factors associated with perceived disease severity in patients with chronic hepatitis C: relationship with information sources and attentional coping styles. Psychosomatics 2005;46,25-33[Abstract/Free Full Text]
  38. Johnston, CC, Gagnon, AJ, Pepler, CJ, et al Pain in the emergency department with one-week follow-up of pain resolution. Pain Res Manag 2005;10,67-70[Medline]
  39. Schanberg, LE, Anthony, KK, Gil, KM, et al Family pain history predicts child health status in children with chronic rheumatic disease. Pediatrics 2001;108,e47[Abstract/Free Full Text]
  40. Viskum, K, Vestbo, J Vital prognosis in intrathoracic sarcoidosis with special reference to pulmonary function and radiological stage. Eur Respir J 1993;6,349-353[Abstract]
  41. Israel, HL, Karlin, P, Menduke, H, et al Factors affecting outcome of sarcoidosis: influence of race, extrathoracic involvement, and initial radiologic lung lesions. Ann N Y Acad Sci 1986;465,609-618[Abstract]
  42. Judson, MA, Baughman, RP, Thompson, BW, et al Two year prognosis of sarcoidosis: the ACCESS experience. Sarcoidosis Vasc Diffuse Lung Dis 2003;20,204-211[ISI][Medline]
  43. Milman, N, Selroos, O Pulmonary sarcoidosis in the Nordic countries 1950–1982: II. Course and prognosis. Sarcoidosis 1990;7,113-118[Medline]
  44. Drent, M, De Vries, J, Lenters, M, et al Sarcoidosis: assessment of disease severity using HRCT. Eur Radiol 2003;13,2462-2471[CrossRef][ISI][Medline]
  45. Grutters, JC, Sato, H, Welsh, KI, et al The importance of sarcoidosis genotype to lung phenotype. Am J Respir Cell Mol Biol 2003;29,S59-62




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Wasfi, Y. S.
Right arrow Articles by Maier, L. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Wasfi, Y. S.
Right arrow Articles by Maier, L. A.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS