In idiopathic pulmonary fibrosis, the quantitation of disease severity using pulmonary function tests is often confounded by emphysema. We have identified the composite physiologic index (CPI) most closely reflecting the morphologic extent of pulmonary fibrosis. Consecutive patients with a clinical/computed tomography (CT) diagnosis of idiopathic pulmonary fibrosis (n = 212) were divided into group I (n = 106) and group II (n = 106). The CPI was derived in group I (by fitting pulmonary function tests against disease extent on CT) and was tested in Group II. The formula for the CPI was as follows: extent of disease on CT = 91.0 − (0.65 × percent predicted diffusing capacity for carbon monoxide [DLCO]) − (0.53 × percent predicted FVC) + (0.34 × percent predicted FEV1). In group II, the CPI correlated more strongly with disease extent on CT (r2 = 0.51) than the individual pulmonary function test (DLCO the highest value, r2 = 0.38). A subanalysis demonstrated that the better fit of the CPI was ascribable to a correction of the confounding effects of emphysema. Mortality was predicted more accurately by the CPI than by a pulmonary function test in all clinical subgroups, including a separate cohort of 36 patients with histologically proven usual interstitial pneumonia (CPI, p < 0.0005; FVC, p = 0.002; PO2, p = 0.002). In conclusion, a new CPI, derived against CT and validated using split sample testing, is a more accurate prognostic determinant in usual interstitial pneumonia than an individual pulmonary function test.
In idiopathic pulmonary fibrosis (IPF), there is an unmet need for an accurate noninvasive measure of disease severity. As well as refining prognostic evaluation, a reliable measure of severity would improve the precision with which disease is monitored and would allow accurate stratification in clinical studies. Currently, no single method of quantifying IPF is wholly satisfactory. Pulmonary function tests reflect the histologic severity of disease better than symptoms or chest radiography (1). Among routine indices, the diffusing capacity for carbon monoxide (DlCO) has the strongest correlation with the morphologic extent of disease, histologically (2) and on computed tomography (CT) (3, 4). However, the interpretation of lung function tests is confounded by coexistent emphysema, present in over 20% of patients (4), and results in a spurious preservation of lung volumes and devastating depression of gas transfer (4–6). A combination of six lung function variables makes up the majority of the clinical–radiographic–physiologic (CRP) score (7); however, the presence and severity of emphysema is not taken into account by the CRP score. A subsequent attempt to adjust for emphysema, in constructing a composite lung function score against the extent of fibrosis on CT, was limited by insufficient numbers of patients to allow split-sample validation (4). Recently, a new CRP score has been developed using multivariate methods, fitting a number of variables, including lung function indices, against survival (8). However, the use of this index depends on the integration of clinical data and plain radiographic scores.
In this study, a composite physiologic index (CPI) was constructed against the morphologic severity of disease in order to calibrate the quantification of pulmonary fibrosis using pulmonary function tests in isolation. The CPI was derived in 106 patients with a clinical and CT diagnosis of IPF and was tested in a further 106 subjects. The prognostic value of the CPI was compared with individual functional variables in all clinical subgroups and in a separate earlier cohort of 36 patients with biopsy-proven usual interstitial pneumonia (UIP).
The population used to generate the CPI comprised consecutive patients presenting with a clinical diagnosis of IPF between December 1990 and December 1996 (n = 212, age 62.2 + 10.6 years, 158 males). Diagnostic criteria comprised (1) bilateral basal or widespread crackles, (2) a restrictive ventilatory defect or isolated depression of DlCO, (3) CT appearances compatible with IPF (discussed later here), and (4) no environmental exposure to a fibrogenic agent (4). Patients with connective tissue diseases were excluded.
A separate population, reported previously (9), was used to evaluate further the prognostic strength of the CPI. All open-lung biopsy diagnoses of IPF between January 1, 1979, and January 1, 1989, were reviewed by A.G.N. and T.V.C. Patients with UIP (n = 36) were studied.
CT criteria for inclusion in the study consisted of (1) predominantly basal and subpleural distribution of disease, (2) a variable mixture of reticular abnormalities and ground-glass attenuation (when ground-glass attenuation was prominent [in a minority of cases], evidence of underlying fine fibrosis, consisting of traction bronchiolectasis, was required), (3) the absence of consolidation, nodular abnormalities, or other parenchymal abnormalities (except for centrilobular emphysema).
The extent of disease was defined on CT as described previously (10, 11); 1.5-mm (n = 121) or 3-mm (n = 91) sections were acquired supine, at full inspiration, at 10-mm intervals (window center = −550 Hounsfield units (HU); window width = 1,500 HU), using an electron beam scanner (Imatron Inc., San Francisco, CA). Images were evaluated by two thoracic radiologists at five levels; the overall extent of disease (i.e., the combined extent of reticular abnormalities and ground-glass attenuation) was scored subjectively to the nearest 5%. At each level, an estimation was also made of the proportions of ground-glass attenuation and reticular abnormalities, and the coarseness of reticular abnormalities, graded as (1) predominantly fine intralobular fibrosis, (2) predominantly microcystic honeycombing (comprising air spaces less than 4 mm in diameter), and (3) predominantly macrocystic honeycombing (comprising air spaces more than 4 mm in diameter). From these observations, the overall extent of disease (mean of five levels, expressed as percentage of abnormal lung), the proportion of ground-glass attenuation, and the coarseness of fibrosis (maximum summed score for five levels = 15) were computed. Based on analyses in a previous patient cohort (4), no weighting factor was used to adjust for volume differences between sections (see online supplement).
The extent of emphysema (permeative destruction of lung parenchyma, resulting in decreased attenuation, without visible walls and of nonuniform distribution [12]) was estimated to the nearest 5% at the same five levels. The overall extent of emphysema was not included in formal analyses but was used in the formal matching process (see Statistical Analysis).
In preliminary analyses, stepwise regression was performed to identify the independent CT determinants of reduction in FVC and DlCO levels (in separate models). FVC and DlCO levels were both independently negatively linked to the extent of disease on CT; exclusion of the coarseness of fibrosis and the proportion of ground-glass attenuation had a minimal effect on equation explanatory powers. These observations, reproducing findings in earlier cohorts (13, 14), were considered to justify the use of the extent of disease on CT, as in multiple studies in previous cohorts (4, 10, 11, 15–18), as the independent CT variable, against which to calibrate a CPI.
Based on a recent comparison of CT findings between UIP and nonspecific interstitial pneumonia (NSIP), in patients presenting with the clinical features of IPF (19), in keeping with recommendations made by the American Thoracic Society and the European Respiratory Society (ATS/ERS) diagnostic committee (20), a retrospective decision was made to categorize CT appearances as (1) typical of UIP (a coarseness score of 8.0 or higher, a proportion of ground-glass attenuation that was less than 20%, n = 118) or (2) compatible with UIP, but fibrotic NSIP possible (n = 94). Based on a review of clinical and bronchoalveolar lavage (BAL) data (performed in a subgroup), 62 patients were identified who met either histologic (n = 32) or clinical/CT/BAL ATS/ERS criteria for UIP (20).
Pulmonary function tests (the percentage predicted) (21), performed within 1 month of CT, included FEV1, FVC, total lung capacity (TLC), residual volume; DlCO, corrected for Hb concentration, and also adjusted for alveolar volume (KCO), using a single-breath technique (n = 91) or a rebreathing technique with adjustment to single-breath values (n = 121). Earlobe capillary gases were performed on air at rest. The alveolar-arterial oxygen gradient was calculated (22).
Analyses were performed using STATA software (Stata data analysis software; Computing Resource Centre, Santa Monica, CA). Group comparisons were made using Student's t-test, Wilcoxon's rank-sum test, or chi-squared testing as appropriate. A p value of less than 0.05 was considered statistically significant. Univariate correlations between the extent of disease on CT and pulmonary function indices were examined using Pearson's product-moment correlation.
In order to exclude undue influence due to outlying observations, the CPI was developed and validated using split-sample testing. The cohort was divided into group I (n = 106) and group II (n = 106). Patients with pulmonary fibrosis but no emphysema (n = 136) were matched in pairs for the extent of disease (within 5%) and were assigned randomly to group I or group II. In patients with coexistent IPF and emphysema (n = 76), matching was based on the extent of emphysema. In group I, stepwise regression was used to generate the weighted combination of lung function variables fitting best with the extent of pulmonary fibrosis on CT (the dependent variable) (23). In order to minimize collinearity between variables, a p value of less than 0.10 was required for retention of individual functional indices. The FEV1, FVC, TLC, residual volume, DlCO, KCO, Po2 and the alveolar-arterial oxygen gradient were examined in the model. The derived formula was used to calculate the CPI. In group II patients, univariate correlations between CT disease extent and the CPI were then compared with correlations between CT disease extent and individual functional indices. These comparisons were repeated in patients with a histologic diagnosis of UIP at thoracoscopic biopsy, in those meeting CT diagnostic criteria for UIP (20), and in those meeting full ATS/ERS diagnostic criteria for UIP (either histologic or clinical/CT criteria) (20).
The prognostic values of individual pulmonary function indices, the extent of disease on CT, and the CPI were compared in the combined group I/group II cohort, in patients meeting CT diagnostic criteria for UIP, and in those meeting full ATS/ERS diagnostic criteria for UIP, using proportional hazards survival analysis (23). In each subgroup, the CPI was compared with each of the other variables in turn, in separate models. A survival analysis was also performed in the separate cohort of 36 patients with histologically proven UIP, reported previously (9). In a subgroup of 30 patients performing maximal exercise testing, the prognostic values of the CPI, the physiologic component of the original CRP score (7), and the physiologic component of the new CRP score (8) were compared.
The extent of interstitial lung disease on CT evaluation in the entire population (n = 212) was 54.6 ± 21.2% of the lung area. In 76 patients with concurrent emphysema (36%), the median extent of emphysema was 10.5% (range, 1–69%). The single-determination SD for the two observers was 6.5% for the extent of IPF and 6.6% for the extent of emphysema. There was excellent agreement on the presence or absence of emphysema (k = 0.94). Group I (derivation of CPI) and group II (testing of CPI) did not differ significantly in demographic, clinical, lung function, or CT findings (Table 1)
Group 1 | Group 2 | UIP at Biopsy | |
---|---|---|---|
Number | 106 | 106 | 36 |
Age* | 61.8 ± 10.9 | 62.6 ± 10.3 | 57.2 ± 7.1 |
Male/female* | 79/27 | 79/27 | 33/3 |
Smoking history Current smokers | 11 | 7 | 5 |
Ex-smokers | 63 | 80 | 28 |
Nonsmokers | 27 | 17 | 3 |
Inadequate data | 3 | 2 | 0 |
Clinical features Mean ± SD duration of dyspnea, months | 30.3 ± 32.2 | 29.2 ± 29.6 | 20.2 ± 17.6 |
Mean ± SD dyspnea score | 2.7 ± 1.5 | 2.4 ± 1.5 | |
1.5-mm CT collimation, DLCO measured via rebreathing method | 57/106 | 64/106 | |
CT features Mean ± SD extent of IPF, % | 54.9 ± 20.3 | 54.2 ± 22.1 | |
Number of patients with CT emphysema | 38/106 | 38/106 | |
Median extent of emphysema, range | 10.5% (1.5–69.5) | 10.0% (1.0–66.5) | |
Lung function tests DLCO, % predicted* | 36.0 ± 15.1 | 39.0 ± 15.8 | 43.5 ± 17.6 |
PO2, kpa* | 9.0 ± 2.0 | 9.2 ± 1.9 | 10.4 ± 1.7 |
FEV1, % predicted | 68.3 ± 18.6 | 74.1 ± 19.3 | 75.3 ± 12.6 |
FVC, % predicted | 67.5 ± 22.0 | 72.4 ± 22.0 | 71.5 ± 16.1 |
TLC, % predicted | 63.0 ± 17.6 | 65.6 ± 18.4 | 70.3 ± 17.3 |
The CPI was derived in group I. The extent of IPF on CT was independently related to percent predicted DlCO (regression coefficient = 0.65; 95% confidence interval = −0.89, −0.42; p < 0.0005), percent predicted FVC (regression coefficient = −0.53; 95% confidence interval = −0.78, −0.28; p < 0.0005), and percentage predicted FEV1 (regression coefficient = 0.34; 95% confidence interval = 0.06, 0.62; p = 0.02). Thus, the final formula for the CPI was as follows:
On reanalysis, after removal of six “outliers” (identified using leverage versus residual plots), the same three variables, with the same negative (DlCO, FVC) or positive (FEV1) weighting were retained in the equation (DlCO, p < 0.0005; FVC, p < 0.0005; FEV1, p < 0.005).
The analysis was repeated in group II. DlCO and FVC levels were independently negative, and FEV1 levels were positive, related to the extent of disease on CT (p < 0.0005 for all three variables). Thus, the same variables were retained as in the CPI (derived from group I). In addition, in group II, Po2 levels were independently weakly linked to disease extent (p = 0.06).
Variations in CT and DlCO measurement protocols (3-mm CT collimation/single breath DlCO method versus 1.5-mm collimation/rebreathing DlCO method adjusted to single breath values) had no independent influence on relationships between structure and function when included in stepwise regression models.
The CPI was calculated in group II. Univariate correlations between physiologic indices (including the CPI) and the extent of IPF on CT are given in Table 2
r | r2 | |
---|---|---|
CPI | 0.71 | 0.51 |
DLCO | −0.61 | 0.37 |
KCO | −0.26 | 0.07 |
FEV1 | −0.37 | 0.14 |
FVC | −0.57 | 0.32 |
TLC | −0.56 | 0.31 |
VA | −0.55 | 0.30 |
RV | −0.38 | 0.14 |
PO2 | −0.48 | 0.23 |
A-ag | 0.40 | 0.16 |
In order to determine whether the superior relationship between function impairment and the morphologic extent of disease provided by the CPI could be ascribed to adjustment for the confounding effect of emphysema, the CPI was re-evaluated in the subgroup of 136 patients with no emphysema on CT. The correlation between CPI and the extent of disease on CT (r2 = 0.57) (Figure 4)
differed little from the correlation between the percent predicted DlCO and the extent of disease on CT (r2 = 0.55) (Figure 5) ; both relationships were substantially stronger than other functional–morphologic correlations, as shown in Table 3r | r2 | |
---|---|---|
CPI | 0.76 | 0.57 |
DLCO | −0.75 | 0.55 |
KCO | −0.44 | 0.19 |
FEV1 | −0.49 | 0.24 |
FVC | −0.61 | 0.37 |
TLC | −0.58 | 0.34 |
VA | −0.60 | 0.36 |
RV | −0.28 | 0.08 |
PO2 | −0.57 | 0.32 |
A-ag | 0.58 | 0.34 |
In order to minimize the inclusion of nonbiopsied patients with diagnoses other than UIP, the CPI was re-evaluated in patients with a histologic diagnosis of UIP (n = 32), in patients meeting ATS/ERS criteria for a diagnosis of IPF on CT (n = 118), and in patients meeting full ATS/ERS clinical/CT/ bronchoscopic or histologic criteria (n = 62). As shown in Table 4
Diagnosis of UIP at Biopsy (n = 32) | Confident CT Diagnosis of IPF (n = 118) | ATS/ERS Diagnosis of IPF (n = 62) | |
---|---|---|---|
CPI, r2, p value | 0.61, 0.0005 | 0.47, < 0.0005 | 0.39, < 0.0005 |
DLCO, r2, p value | 0.51, < 0.0005 | 0.31, < 0.0005 | 0.20, < 0.0005 |
KCO, r2, p value | 0.08, NS | 0.03, NS | 0.00, NS |
FEV1, r2, p value | 0.27, 0.002 | 0.09, < 0.001 | 0.15, 0.001 |
FVC, r2, p value | 0.47, < 0.0005 | 0.30, < 0.0005 | 0.34, < 0.0005 |
TLC, r2, p value | 0.36, < 0.0005 | 0.31, < 0.0005 | 0.24, < 0.0005 |
VA, r2, p value | 0.41, < 0.0005 | 0.28, < 0.0005 | 0.29, < 0.0005 |
RV, r2, p value | 0.03, NS | 0.16, < 0.0005 | 0.04, 0.02 |
PO2, r2, p value | 0.14, 0.05 | 0.17, < 0.0005 | 0.09, 0.02 |
A-ag, r2, p value | 0.14, 0.05 | 0.10, < 0.001 | 0.07, 0.04 |
In the combined Group I/Group II cohort, follow-up was complete in 197 cases (93%) and was censored at the date of last follow-up in 15 cases (7%). There was a median survival of 22 months, and the 5-year survival was 23% in the entire cohort (23% in patients with CT appearances typical of UIP, 28% in patients meeting full ATS/ERS diagnostic criteria for UIP).
On univariate analysis, in the combined cohort, increased mortality was associated with increasingly extensive disease on CT (p < 0.0005), greater functional impairment (DlCO, FVC, TLC, FEV1, alveolar volume (VA), Po2, alveolar-arterial oxygen gradient all p < 0.0005; Kco p < 0.005) and higher CPI scores (p < 0.0005). When compared with individual functional indices in separate models, CT disease extent was a more powerful prognostic determinant than other variables, with the exception of DlCO levels and the CPI. This finding persisted when analyses were performed in patients with CT features typical of UIP and in those meeting full ATS/ERS diagnostic criteria for UIP.
In the combined cohort and in both subgroups, the CPI was the most powerful prognostic determinant in all analyses. In all three groups, when compared with each other variable in turn (in bivariate models), the CPI remained predictive of mortality (p < 0.0005 in all analyses). The extent of disease on CT, DlCO, FVC, TLC, FEV1, and VA had no independent prognostic significance, when evaluated together with the CPI. However, Po2 levels and the alveolar-arterial oxygen gradient were both significantly independently predictive of mortality (0.005 < p < 0.05 in different subgroups), although their prognostic value was much lower than that of the CPI.
In the separate cohort of patients with biopsy-proven UIP, follow-up was complete in 33 cases and censored at date of last follow-up in three cases (median survival, 28 months; 5-year survival, 16%). When functional variables were analyzed in separate proportional hazards models (Table 5)
All Patients (n = 36) | Patients Able to Exercise (n = 30) | |
---|---|---|
CPI | 0.092 (0.043, 0.141), p < 0.0005 | 0.083 (0.027, 0.140), p = 0.004 |
DLCO, % predicted | −0.022 (−0.050, 0.008), p = 0.11 | −0.017 (−0.047, 0.012), p = 0.25 |
KCO, % predicted | 0.004 (−0.010, 0.019), p = 0.58 | 0.005 (−0.010, 0.021), p = 0.52 |
FEV1, % predicted | −0.020 (−0.055, 0.016), p = 0.27 | −0.004 (−0.043, 0.034), p = 0.83 |
FVC, % predicted | −0.050 (−0.081, −0.019), p = 0.002 | −0.039 (−0.074, −0.005), p = 0.02 |
TLC, % predicted | −0.033 (−0.059, −0.006), p = 0.01 | −0.028 (−0.056, −0.001), p = 0.04 |
Alveolar-arterial O2 gradient, kpa | 1 (0.11, 0.50), p = 0.002 | 0.33 (0.02, 0.65), p = 0.04 |
PO2, kpa | −0.36 (−0.60, −0.13), p = 0.002 | −0.20 (−0.55, 0.14), p = 0.25 |
CRP score, physiologic component | 0.044 (0.008, 0.080), p = 0.02 | |
New CRP score, physiologic component | 1 (0.03, 0.23), p = 0.009 |
The CPI is a new “severity variable” that reconciles functional severity and the global morphologic extent of disease. Our observations show that the CPI correlates with the extent of pulmonary fibrosis on CT more strongly and is linked to mortality (in histologically proven UIP and in IPF diagnosed using clinical and CT criteria) more closely than individual pulmonary function indices. The CPI accounts for coexisting emphysema (a major confounding influence on pulmonary function indices [4–6]). In patients without emphysema on CT, the CPI reflected the extent of fibrosis no better than DlCO levels.
The CPI quantifies the functional defect attributable to pulmonary fibrosis, while excluding that ascribable to emphysema. The emergence of DlCO as a key component of the CPI was predictable, as DlCO levels best reflected the extent of fibrosis in univariate analysis, as in previous histologic (2) and CT (3, 4) studies. The integration of FVC and FEV1 levels in the CPI, with a negative weighting for the latter, is equally logical. In essence, a higher (“restrictive”) FEV1 level results in an increased CPI score, because in that case, reductions in DlCO and FVC are largely or wholly due to fibrosis rather than emphysema (4). The role of spirometric volumes in the CPI was underscored by the fact that DlCO levels were not linked to survival in patients with proven UIP, whereas FVC levels provided statistically significant prognostic information, as reported in several recent studies (17, 24, 25). In contrast, in younger IPF patients referred for transplantation (26) with little coexistent emphysema (Egan J.J., personal communication), DlCO levels were the most accurate prognostic indicator.
The physiologic component of the original CRP score (7), consisting of six variables, differs from the CPI in two essential respects: The CRP score was constructed from univariate correlations between histologic severity and lung function variables, without adjustment for collinearity between lung function indices, and does not take the effects of emphysema into account. Thus, for a given percent predicted FVC, the restrictive picture of a relative preservation of FEV1 results in a lower CRP score (implying less severe disease) but a higher CPI score (indicating more severe disease). The CPI has advantages over the original CRP score in routine clinical practice. It is easier to generate, requiring solely the measurement of gas transfer and spirometric volumes. Moreover, the original CRP score does not predict survival consistently in IPF (27); in the present study, it was inferior to the CPI in this regard.
A new CRP score, recently published by King and colleagues (8), generated against survival in a large cohort of patients with UIP, includes a simplified physiologic component, derived from TLC and arterial oxygen levels on maximal exertion. Unlike the original CRP score, collinearity between lung function indices was addressed, and the simplicity of the physiologic component is attractive. Despite major method differences in the derivations of the new CRP score and the CPI in this study, the components of the two composite indices are remarkably similar. In effect, exercise data are replaced by DlCO levels in the CPI, but these two measures are closely correlated in UIP (8). The inclusion of spirometric volumes in the CPI, with a negative weighting for FEV1, provides essentially the same adjustment as TLC levels (which are relatively preserved in patients with concurrent emphysema). Thus, the new CRP score and the CPI have in common a single measure of gas exchange and a lung volume component that is modulated in opposite directions by pulmonary fibrosis and emphysema.
The CPI offers two distinct advantages. First, it is generated in isolation, without radiologic or clinical components. In contrast, the use of the new CRP score depends on the inclusion of other variables, including the presence of finger clubbing and International Labour Organization (ILO) scoring of chest radiographs, which provide a very similar proportion of the total score (up to 18%) to the combined contribution of TLC levels and exercise data (up to 22%). Given the lack of ready availability of ILO scoring skills and the subjective nature of clubbing, the CPI is likely to be easier to validate in routine practice, and more attractive to clinicians.
A second advantage of the CPI index is the use of DlCO levels, rather than exercise data. In many patients with IPF, the CRP score cannot be used because maximal exercise tests are precluded by cardiac disease or resting hypoxia. The most compelling argument for maximal exercise data is that their exclusion weakened the prognostic strength of the new CRP score (8). However, this conclusion was based on reanalysis after their removal from the final multivariate equation, without the substitution of a surrogate variable (such as DlCO). Because exercise data were an independent determinant of outcome when grouped with other selected variables, their removal (without replacement) necessarily diminishes the prognostic strength of the new CRP score. It is uncertain whether DlCO levels and exercise data would provide equivalent prognostic information in the CPI and new CRP score. The minor apparent prognostic advantage of the CPI over the new CRP score, in patients with biopsy-proven UIP, must be interpreted with caution in view of the small sample size.
The reproducibility of the CPI and new CRP scores is a crucial consideration. Measurement variation in gas transfer between lung function laboratories is an important caveat (28). In this study, the use of single breath and rebreathing measurements in separate subgroups had no independent effect on correlations between structure and function. However, other differences between laboratories, including calibration and quality control measures, were not captured. Importantly, spirometric volumes have excellent intrasubject agreement, which can only increase the reproducibility of the CPI, compared with DlCO measurement in isolation. Precisely the same reservations apply to the reproducibility of arterial oxygen tension at maximal exercise, not previously evaluated in IPF; no definition of “significant deterioration” (over and above the “noise” of measurement) has been developed for clinical use.
CT has major advantages as a morphologic index of severity. The entire spectrum of IPF can be evaluated, including patients too compromised to undergo a lung biopsy. Split-sample testing requires a large population; analyses of a smaller subgroup undergoing biopsy could not have generated a precise composite physiologic score. Moreover, CT abnormalities correlate more strongly with functional impairment than histologic findings. Functional–histologic correlations in IPF are limited by small numbers, the grouping of IPF with other interstitial diseases, and the effects of concurrent emphysema (29–31). In studies with large numbers of IPF patients (8, 32), stronger correlations are reported, but these are consistently weaker than observed relationships between lung function abnormalities and CT scores in IPF (3, 4, 16, 33). The discrepancy between CT–functional and histologic–functional relationships is largely ascribable to the regional nature of lung biopsy; it is not surprising that functional impairment is related more closely to global extent than regional severity. A further advantage of CT is that once overall disease extent has been quantified, the degree of functional impairment is not independently influenced by other CT features, including the coarseness of a reticular pattern (13), the proportions of ground-glass and reticular patterns (14), or disease distribution (13). These observations, reproduced in the current cohort in preliminary analyses (data not shown), justify the simple CT scoring system used in many previous studies (4, 10, 11, 15–18). Finally, technical differences between two CT scanners did not influence our findings; collimation thickness had no independent effect on functional–morphologic relationships.
An important consideration, in comparing composite indices between populations, is differences in diagnostic criteria. The new CRP score was developed in patients with a histologic diagnosis of UIP, and the authors highlight the possibility of major selection bias, in view of the exclusion of patients with severe disease, who were unable to undergo surgical biopsy (8). However, an equally important drawback with clinical and CT diagnostic criteria is the recent recognition of a significant subgroup of “IPF” patients with NSIP, with a better prognosis than those with UIP (9, 34, 35). Patients with no histologic diagnosis necessarily include a subgroup with NSIP, presenting with clinical features of IPF.
However, in the particular subgroup of NSIP patients with clinical features of IPF, the course is often progressive (9), and it has been argued that NSIP may sometimes be a precursor of UIP (36). In this study, no patient had consolidation or nodules on CT (reported in the larger unselected population of NSIP patients [37]), and CT abnormalities were those previously regarded as compatible with IPF (i.e., a variable mixture of reticular abnormalities and ground-glass attenuation, in a predominantly subpleural, basal distribution compatible with IPF). Possible CT differences in our population between UIP and NSIP, including the coarseness of fibrosis, the proportions of ground-glass and reticular patterns, and distribution, have no independent functional effects in IPF, once the extent of disease on CT has been scored (13, 14). It is now known that among patients presenting with the clinical features of IPF, those with a histologic diagnosis of fibrotic NSIP are characterized by a greater proportion of ground-glass attenuation and relatively less microcystic or macrocystic honeycombing than those with UIP (19). It is now accepted that diagnostic CT features for UIP include subpleural basal honeycombing and limited ground-glass attenuation (16). When analysis was confined to patients meeting these CT criteria, the advantage of the CPI, in its correlation with the CT extent of disease, was preserved, and remained robust in subgroups with histologic proof of UIP and those meeting full ATS/ERS criteria for IPF. Moreover, the high mortality in all patient subgroups in this study was strikingly similar to mortality in reported UIP populations. Thus, it is unlikely that undiagnosed NSIP had a major confounding effect. As this population and that studied by King and colleagues (8) are complementary, each posing different selection problems, the similarities between the CPI and the new CRP score are reassuring.
In conclusion, we report a new CPI for use in IPF, which reflects the extent of disease more accurately than individual lung function indices and provides more powerful prognostic information in patients with a clinical diagnosis of IPF and in those with histologically proven UIP. The major advantages of the CPI are the ease with which it can be generated, using spirometric volumes and measures of gas transfer, and its applicability without the need for formal radiographic scoring. The greatest utility of the CPI is likely to lie in the staging of disease severity in clinical practice and in a wide variety of clinical studies.
1. | Keogh BA, Crystal RG. Pulmonary function testing in interstitial pulmonary disease: what does it tell us? Chest 1980;78:856–865. |
2. | Cherniack RM, Colby TV, Flint A, Thurlbeck WM, Waldron JA Jr, Ackerson L, Schwarz MI, King TE Jr. Correlation of structure with function in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 1995;151:1180–1188. |
3. | Staples CA, Muller NL, Vedal S, Abboud R, Ostrow DN, Miller RR. Usual interstitial pneumonia: correlation of CT with clinical, functional and radiologic findings. Radiology 1987;162:377–381. |
4. | Wells AU, King AD, Rubens MB, Cramer D, du Bois RM, Hansell DM. Lone cryptogenic fibrosing alveolitis: a functional-morphologic correlation based on extent of disease on thin section computed tomography. Am J Respir Crit Care Med 1997;155:1367–1375. |
5. | Wiggins J, Strickland B, Turner-Warwick M. Combined cryptogenic fibrosing alveolitis and emphysema: the value of high resolution computed tomography in assessment. Respir Med 1990;84:365–369. |
6. | Docherty MJ, Pearson MG, O'Grady EA, Pellegrini V, Calverley PM. Cryptogenic fibrosing alveolitis with preserved lung volumes. Thorax 1997;52:998–1002. |
7. | Watters LC, King TE, Schwarz MI, Waldron JA, Stanford RE, Cherniack RM. 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. |
8. | King TE, Tooze JA, Schwarz MI, Brown KR, Cherniack RM. Predicting survival in idiopathic pulmonary fibrosis: scoring system and survival model. Am J Respir Crit Care Med 2001;164:1171–1181. |
9. | Nicholson AG, Colby TV, du Bois RM, Hansell DM, Wells AU. The prognostic significance of the histologic pattern of interstitial pneumonia in patients presenting with the clinical entity of cryptogenic fibrosing alveolitis. Am J Respir Crit Care Med 2000;162:2213–2217. |
10. | Wells AU, Rubens MB, du Bois RM, Hansell DM. Serial CT in fibrosing alveolitis: prognostic significance of the initial pattern. Am J Roentgenol 1993;161:1159–1165. |
11. | Wells AU, Hansell DM, Rubens MB, Cullinan P. black CM, du Bois RM. The predictive value of appearances on thin section computed tomography. Am Rev Respir Dis 1993;148:1076–1082. |
12. | Austin JHM, Muller NL, Friedman PJ, Hansell DM, Naidich DP, Remy-Jardin M, Webb WR, Zerhouni EA. Glossary of terms for CT of the lungs: recommendations of the Nomenclature Committee of the Fleischner Society. Radiology 1996;200:327–331. |
13. | Chan TYK, Hansell DM, Rubens MB, du Bois RM, Wells AU. Cryptogenic fibrosing alveolitis and the fibrosing alveolitis of systemic sclerosis: morphological differences on computed tomography. Thorax 1997;52:265–270. |
14. | Wells AU, Rubens MB, du Bois RM, Hansell DM. Functional impairment in fibrosing alveolitis: relationship to reversible disease on thin section computed tomography. Eur Respir J 1997;10:280–285. |
15. | Wells AU, Cullinan P, Hansell DM, Rubens MB, Black CM, Newman-Taylor AJ, Du Bois RM. Fibrosing alveolitis associated with systemic sclerosis has a better prognosis than lone cryptogenic fibrosing alveolitis. Am J Respir Crit Care Med 1994;149:450–454. |
16. | Wells AU, Hansell DM, Rubens MB, King AD, Cramer D, Black CM, du Bois RM. Fibrosing alveolitis in systemic sclerosis: indices of lung function in relation to extent of disease on computed tomography. Arthritis Rheum 1997;40:1229–1236. |
17. | Wells AU, Hansell DM, Rubens MB, Cailes JB, Black CM, du Bois RM. Functional impairment in lone cryptogenic fibrosing alveolitis and fibrosing alveolitis associated with systemic sclerosis: a comparison. Am J Respir Crit Care Med 1997;155:1657–1664. |
18. | Wells AU, Hansell DM, Haslam PL, Rubens MB, Cailes JB, Black CM, du Bois RM. Bronchoalveolar lavage cellularity: lone cryptogenic fibrosing alveolitis compared with the fibrosing alveolitis of systemic sclerosis. Am J Respir Crit Care Med 1998;157:1474–1482. |
19. | MacDonald SLS, Rubens MB, Hansell DM, Copley SJ, Desai SR, du Bois RM, Nicholson AG, Colby TV, Wells AU. Nonspecific interstitial pneumonia and usual interstitial pneumonia: comparative appearances and diagnostic accuracy of high-resolution computed tomography. Radiology 2001;221:600–605. |
20. | American Thoracic Society. Idiopathic pulmonary fibrosis: diagnosis and treatment: International Consensus Statement: American Thoracic Society (ATS) and the European Respiratory Society (ERS). Am J Respir Crit Care Med 2000;161:646–664. |
21. | Quanjer PH. Standardised lung function testing. Clin Respir Physiol 1983;19:1–95. |
22. | Riley RL, Cournand A. Ideal alveolar air and the analysis of ventilation-perfusion relationships in the lungs. J Appl Physiol 1949;1:825–847. |
23. | Kleinbaum DG, Kupper LL, Muller R. Applied regression analysis and other multivariate methods. 1988; Kent, Boston. |
24. | Schwartz DA, Helmers RA, Galvin JR, Van Fossen DS, Frees KL, Dayton CS, Burmeister LF, Hunninghake GW. Determinants of survival in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 1994;149:450–454. |
25. | Erbes R, Schaber T, Loddenkemper R. Lung function tests in patients with idiopathic pulmonary fibrosis: are they helpful for predicting outcome? Chest 1997;111:51–57. |
26. | Mogulkoc N, Brutsche MH, Bishop PW, Greaves SM, Horrocks AW, Egan JJ. Pulmonary function in idiopathic pulmonary fibrosis and referral for lung transplantation. Am J Respir Crit Care Med 2001;164:103–108. |
27. | Gay SE, Kazerooni EA, Toews GB. Idiopathic pulmonary fibrosis: predicting response to therapy and survival. Am J Respir Crit Care Med 1998;157:1063–1072. |
28. | Snow M, Stein C, Fallat R. Interlaboratory variability of pulmonary function tests. Am Rev Respir Diagn 1983;2:43–45. |
29. | Fulmer JD, Roberts WC, von Gal ER, Crystal RG. Morphologic-physiologic correlates of the severity of fibrosis and degree of cellularity in idiopathic pulmonary fibrosis. J Clin Invest 1979;63:665–676. |
30. | Green GM, Graham WGB, Hanson JS, Gump DW, Phillips CA, Brody AR, Sylvester DW, Landis JN, Davis GS, Craighead JE. Correlated studies of interstitial pulmonary disease. Chest 1976;69:263–284. |
31. | Chinet T, Jaubert F, Dusser D, Danel C, Chretien J, Huchon GJ. Effects of inflammation and fibrosis on pulmonary function in diffuse lung fibrosis. Thorax 1990;45:675–678. |
32. | Cherniack RM, Colby TV, Flint A, Thurlbeck WM, Waldron JA Jr, Ackerson L, Schwarz MI, King TE Jr. Correlation of structure and function in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 1995;151:1180–1188. |
33. | Xaubet A, Agusti C, Luburich P, Roca J, Monton C, Ayuso MC, Barbera JA, Rodriguez-Roisin R. Pulmonary function tests and CT scan in the management of idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 1998;158:431–436. |
34. | Bjoraker JA, Ryu JH, Edwin MK, Myers JL, Tazelaar HD, Schroeder DR, Offord KP. Prognostic significance of histopathological subsets in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 1998;157:199–203. |
35. | Travis WD, Matsui K, Moss J, Ferrans VJ. Idiopathic nonspecific interstitial pneumonia: prognostic significance of cellular and fibrosing patterns–survival comparison with usual interstitial pneumonia and desquamative interstitial pneumonia. Am J Surg Pathol 2000;24:19–33. |
36. | Flaherty KR, Travis WD, Colby TV, Toews G, Kazerooni EA, Gross BH, Jain A, Strawderman RL, Flint A, Lynch JP, et al. Histopathologic variability in usual and nonspecific interstitial pneumonias. Am J Respir Crit Care Med 2001;164:1722–1727. |
37. | Hartman TE, Swensen SJ, Hansell DM, Colby TV, Myers JL, Tazelaar HD, Nicholson AG, Wells AU, Ryu JH, Midthun DE, et al. Nonspecific interstitial pneumonia: variable appearance at high resolution chest CT. Radiology 2000;217:701–705. |