INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Idiopathic pulmonary fibrosis (IPF) is associated with significant morbidity and mortality despite aggressive therapeutic attempts. Significant toxicity is associated with currently employed therapies, including steroids (1, 2) and cytotoxic agents. Accordingly, a diverse group of clinical variables have been evaluated for their ability to predict outcome of IPF with or without therapy in an effort to identify those patients most likely to benefit from treatment. The efficacy of these therapies is difficult to determine. Widely different approaches have been used to define response including mortality and changes in pulmonary function. The magnitude of change in pulmonary function required to be judged a responder has varied from 10 to 20%.

Histologic findings of open lung biopsies have been felt to be of particular importance in predicting response to therapy and prognosis. A pattern of desquamative interstitial pneumonitis has been reported to have a better outlook than one of usual interstitial pneumonitis (3, 4). Furthermore, in a retrospective study a more cellular histologic picture has been associated with better long-term prognosis (2). More recently, because of expense and risk, the need to routinely perform a surgical lung biopsy has been questioned (5). No prospective trials have confirmed the predictive value of pretreatment open lung biopsy.

Pretherapy pulmonary function has been studied extensively. A worse survival has been associated with a lower pretreatment diffusing capacity (6, 7), TLC (8), and FVC (7, 8, 10, 11). A lower survival has been reported in patients who experience greater reduction in pulmonary function during the initial course of therapy (12, 13). Pulmonary function has been shown to correlate poorly with histologic severity in confirmed IPF (14, 15), however. Clinical (dyspnea scale), radiographic features (chest radiograph), and weighted pulmonary function parameters including exercise testing have been combined to generate a clinical/radiographic/physiologic score (CRP) (16). The correlation of CRP score and histologic severity is improved, although clinical validation and use of this staging technique during prolonged therapy is limited (16, 17). Accordingly, diverse groups of clinical variables have been evaluated for their ability to predict outcome in IPF with varying results.

Recently, high-resolution computed tomography (HRCT) has been shown to more clearly delineate the anatomy of the lungs when compared with conventional CT (18, 19). Retrospective studies have demonstrated a higher likelihood of response to therapy in those individuals with areas of "ground glass" attenuation on HRCT (4, 20).

We have tested the hypothesis that initial diagnostic studies can identify those patients with IPF more likely to respond to corticosteroid therapy. We have examined the ability of initial diagnostic studies to predict outcome in patients with biopsy-proven IPF. The outcomes evaluated include short-term improvement as determined by CRP scoring. Long-term outcomes evaluated include survival and change in pulmonary function. The initial diagnostic studies obtained included CRP, HRCT, and quantitative pathologic scoring. Specifically we postulated that: (1) pulmonary function testing would identify patients with greater disease severity and greater responsiveness to therapy, (2) open biopsy would identify greater inflammatory change and a greater response to therapy, and (3) radiographic studies, including HRCT, would identify greater "ground glass" attenuation and therefore greater response to therapy.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Recruitment

All patients referred to the University of Michigan for prospective enrollment in the Specialized Center Of Research (SCOR) protocol studying IPF were considered. Suspicion of IPF was based on symptoms, physiologic abnormalities, or radiographic findings. None of the patients had undergone previous open lung biopsy or had received therapy. Patients were excluded if they were found to have a disease other than IPF during the enrollment workup. Disorders that excluded patients include: collagen vascular disease, pneumoconiosis, sarcoidosis, cancer, lymphoma, eosinophilic granuloma, hypersensitivity pneumonitis, respiratory bronchiolitis, and lymphangioleiomyomatosis. Additionally, patients were excluded if they were unable, or unwilling, to undergo open or video-assisted thoracoscopic (VAT) lung biopsy.

Physiologic Assessment

Physiologic assessment was performed before open lung biopsy and before the initiation of therapy. Pulmonary function tests, including spirometry, lung volumes, and diffusing capacity of the lungs for carbon monoxide (DL_CO), were performed on the same day but before cardiopulmonary exercise testing (CPET). All spirometric studies were performed using a calibrated pneumotachograph (Medical Graphics Co., St. Paul, MN) and values were expressed as a percentage of the predicted values published by Morris and coworkers (23). Lung volumes were measured in a whole-body plethysmograph, using the technique of Dubois and coworkers (24). Data were expressed as a percentage of the predicted values published by Goldman and Becklake (25). Diffusing capacity was measured using the technique of Ogilvie and coworkers (26), corrected for measured hemoglobin, and expressed as a percentage of the predicted values published by Crapo and Morris (27).

CPETs were performed on an electronically braked, calibrated cycle ergometer. Work load was increased by 20 watts per minute until maximal symptom-limited exercise was achieved. Expired gases and ventilation were measured using a calibrated metabolic cart (2001 System; Medical Graphics Co., or Collins CPXII, Warren E. Collins, Inc., Braintree, MA). Arterial blood gases were obtained at rest and every 2 min during exercise via an indwelling radial artery catheter. Alveolar-arterial oxygen tension difference [p(A  a)O₂] was calculated for each sample by the alveolar gas equation, using the measured respiratory quotient during the same time-point of exercise. Oxygen saturation was measured by co-oximetry of the blood gas sample. Twelve lead electrocardiograms and noninvasive measurements of blood pressure were recorded every minute of exercise. The values of Jones and coworkers were used to predict maximal exercise oxygen consumption (O₂) (28).

CRP Scoring System

Clinical severity was assessed for each patient using a previously developed CRP composite score for IPF (16). The total CRP score ranges from 0 to 100 points (100 being the most severe disease) based on the following seven variables: level of dyspnea, 0-20 points; chest radiograph, 0-10 points; spirometry (FVC, 0-12 points, FEV₁, 0-3 points); lung volume 0-10 points; diffusion capacity (corrected for alveolar volume), 0-5 points; resting p(A  a)O₂, 0-10 points; O₂ saturation corrected for maximal achieved O_2max, 0-30 points.

HRCT Protocol

All HRCTs were performed with 1.0- or 1.5-mm-thick sections taken at 1-cm intervals throughout the entire thorax and were reconstructed using a high spatial frequency algorithm. No intravenous contrast was administered. The scans were performed using a General Electric Advantage CT/T scanner (G.E. Medical Systems, Milwaukee, WI). Four experienced thoracic radiologists independently evaluated all HRCTs (P.N.C., B.H.G., D.L.S., E.A.K.). All HRCTs were obtained 1 to 4 wk before open lung biopsy. The radiologists scored ground glass opacity (CT-alv) and reticular opacity (CT-fib) on a scale of 0-5. These scores were also summed into a total CT score (CT-Tot). Table 1 outlines the scoring system used for the evaluation of the HRCT scans which has been described previously (29). For the purpose of analysis, each lobe was scored by the interpreters and the mean of all lobes was incorporated into a fibrotic, ground glass, and total score for each patient.

Biopsy Technique

All patients underwent bronchoscopy with bronchoalveolar lavage and transbronchial biopsies before open lung biopsy. Patients were referred for open lung biopsy when the results of transbronchial biopsy did not reveal a clear-cut alternative diagnosis. Open lung biopsy was performed by formal thoracotomy or video-assisted thoracoscopy. Similar sized biopsies were obtained by either technique. Biopsies were obtained from all three lobes on the right side or from the upper and lower lobe on the left with exclusion of the lingula. The surgeon was asked to biopsy areas involved with ground glass opacity and/or reticular opacity. A clinical pathologist suggested the diagnosis of IPF.

Pathologic Scoring

All open lung biopsy specimens were subsequently reviewed by a specialist in lung pathology (AF) with expertise in scoring histologic abnormality in IPF. The pathologist was not aware of the radiographic scoring for individual patients at the time of pathologic scoring. Each specimen was processed in routine fashion (30). Four histologic sections were prepared from each paraffin block. Individual slides were stained with hematoxylin and eosin, pentachrome stain (which demonstrates differential staining of elastic tissue, collagenized connective tissue, and mucopolysaccharide-rich stroma), Prussian blue (iron stain), and a trichrome stain (which demonstrates differential staining of collagenized connective tissue and smooth muscle). A grade of 0-5 was applied to (1) the extent and (2) severity of cellularity in the alveolar wall (interstitium); (3) metaplastic cells lining the air spaces, including foci of honeycomb lung; (4) the extent and (5) severity of cellularity in the alveolar space (desquamation); (6) interstitial "young connective tissue"; (7) interstitial fibrosis; (8) honeycomb cysts; (9) metaplastic smooth muscle in the stroma; (10) myointimal mural thickening in the vessel walls. A score of zero to 2 was used to assess (11) airway luminal granulation tissue, (12) air space granulation tissue; (13) airway wall inflammation; and (14) airway wall fibrosis. These 14 characteristics were separated into four factor scores: fibrosis, cellularity, desquamation, and granulation/connective tissue. The scores for each lobe biopsied were averaged. The total pathology score represented the sum of the four scores (15).

Therapy

All patients were treated with prednisone (1 mg/kg/d) for 3 mo after confirmatory open lung biopsy and baseline studies were completed. Changes in the CRP score at 3 mo were used to assess therapeutic response. Responders (> 10 point drop in CRP score), stable individuals (± 10 point change in CRP score) or nonresponders (> 10 point increase in CRP score or death) were identified. Patients who died were classified as nonresponders. Corticosteroid therapy was continued (in a tapering dose over 18 mo) only among patients exhibiting a favorable response to initial therapy. In all others, the corticosteroids were rapidly tapered and discontinued within 4 wk. Patients failing prednisone at the 3-mo evaluation were crossed-over to oral cyclophosphamide (2 mg/kg/d) for a 6-mo course of therapy.

Statistical Analysis

Analysis of variance. Data were expressed as mean ± SE throughout. To examine the early response to therapy the groups were initially segregated according to the change in 3-mo CRP score or death within the first 3 mo of steroid treatment. The means of each ranked parameter including pulmonary function, radiographic studies, and pathologic indices were compared between each group using analysis of variance (ANOVA). Post hoc analysis for pairwise comparisons were done using the Bonferroni adjustment for inflated alpha error. Dichotomous variables were compared using Fisher exact test. To examine longer term follow-up, similar analyses were performed with nonresponders regrouped as a greater than 10-point rise in CRP during the initial 3-mo steroid trial or death during the follow-up period. Follow-up concluded with patient death or last evaluation (no patient was lost to follow-up). In these analyses adjustment was made for age, smoking history, time to onset of illness, length of follow-up, and peak exercise p(A  a)O₂. Initial FVC, FEV₁, TLC, and DL_CO were compared in the surviving patients at baseline before therapy and at last follow-up using paired Student's t tests.

Those individuals responding to steroids or remaining stable were grouped and compared with those individuals not responding to high-dose steroids or dying during long-term follow-up using Student's t test. Clinical features, physiologic data, radiographic data, and pathologic scoring were compared in this way.

Receiver operating curves. To better define the value of all predictors in identifying those individuals most likely to die during follow-up, comparison of ranked parameters was performed by receiver operating curve (ROC) analysis. The area under the ROC represents the probability that a randomly chosen diseased subject is correctly ranked with greater suspicion than a randomly chosen, nondiseased subject (31). This analysis also allows a graphical representation of sensitivity and specificity for a given diagnostic feature or study and well-defined endpoint.

ROC curves were created with individual clinical, physiologic, and pathologic parameters added to CT-fib score in an effort to improve prediction of mortality. The area under the curve for the composite ROC curves was compared with that of the ROC curve for the CT-fib score alone using two-tailed Student's t test.

Survival analysis. Survival curves obtained from Kaplan-Meier analysis were used to compare independent variables (clinical features, pulmonary function, pathologic scoring, and HRCT scoring) with the endpoints of survival and death. In these analyses adjustment was made for age, smoking history, time to onset of illness, and peak p(A  a)O₂. A similar analysis was performed with pulmonary function as the endpoint. For this analysis, we defined deterioration in pulmonary function as a > 10% drop in FVC, TLC, or DL_CO as used in prior studies (1, 2, 9, 11, 12). Similar analysis was performed using a 20% drop in DL_CO. In all analyses, p < 0.05 was considered significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Characteristics

Thirty-eight patients formed the study group, including 17 men and 21 women with a mean age of 54.6 ± 2.2 yr and 2.6 ± 0.6 mo of symptoms prior to histologic diagnosis. The histologic picture was that of usual interstitial pneumonitis in 37 and desquamative interstitial pneumonitis in one. Twenty-seven were active or ex-smokers (24.8 ± 3.1 pack-years) while 11 were nonsmokers. The mean CRP score for the group as a whole was 44.3 ± 3.1 with a clinical score of 9.8 ± 1.0, radiographic score of 5.1 ± 0.4, and physiologic score of 29.0 ± 2.7. The latter consisted of mean FVC percentage of predicted of 69.7 ± 2.5, FEV₁ percentage of predicted of 78.3 ± 2.9, TLC percentage of predicted of 72.9 ± 2.3, FRC percentage of predicted of 76.4 ± 3.1, DL_CO percentage of predicted of 49.9 ± 2.4, DL_CO/VA percentage of predicted of 79.7 ± 3.5, resting p(A  a)O₂ of 19.7 ± 1.9, and exercise score of 24.7 ± 1.8. The total pathology score was 24.3 ± 1.2 with a cellular score of 6.7 ± 0.3, desquamative score of 5.2 ± 0.3, granulation score of 1.7 ± 0.2, and fibrosis score of 10.8 ± 0.7. The HRCT ground glass score was 1.2 ± 0.2, fibrotic score 1.6 ± 0.1, and total score of 2.8 ± 0.2.

The results of therapy are shown in Figure 1. After 3 mo of corticosteroid therapy, 10 of 38 (26%) responded to prednisone therapy, 14 of 38 (37%) remained stable while an additional 14 of 38 (37%) were considered nonresponders. Seven of 14 patients in the nonresponder group died of respiratory failure while receiving high-dose steroid therapy. The duration of follow-up included 26.0 ± 2.9 mo in responders, 20.9 ± 3.0 mo in nonresponders, and 39.4 ± 3.9 mo in stable individuals. There was no difference in the percentage of smokers or ex-smokers between the three groups.

View larger version (11K): [in this window] [in a new window]   Figure 1.   Short-term response and longer term follow-up in 38 patients with biopsy-proven IPF treated with immunosuppressive therapy.

Short-term Response to Steroid Therapy

The response to 3 mo of high-dose corticosteroid therapy was defined by the change in CRP score. Table 2 illustrates the various parameters comprising the CRP score as well as the various pathologic scores in the three groups as defined by response to initial 3 mo of steroid therapy. Those individuals experiencing a response to steroids were younger than those who were nonresponders (p < 0.05). A higher CRP score was evident in the responders than in the those who remained stable or were nonresponders (p < 0.05). No difference in CRP score was noted between the stable and nonresponders groups. The overall physiologic score was higher in responders than in stable individuals (p < 0.05). Of individual pulmonary function studies, only the FEV₁ as a percentage of predicted was statistically different between the groups (responder 68.7 ± 5.15, nonresponder 89.07 ± 4.35, stable 74.36 ± 4.35, p = 0.01. Higher cellular and desquamative pathology scores were seen in responders compared with the stable group (p < 0.05). No difference was seen in fibrotic pathology scores between the three groups or in any pathology score between the nonresponder and stable groups.

HRCT Scoring and Response to 3 mo of Steroid Therapy

Although not graphically presented, the pretreatment ground glass score was statistically higher in the responder group than in the nonresponder and stable groups (p = 0.004). A higher fibrotic score in the nonresponder group approached statistical significance when compared with the responder or stable groups (p = 0.067).

Long-term Response to Therapy

A major change occurred in the long-term evaluation of outcome as three of the subjects not initially defined as nonresponders (two responders and one stable) died of respiratory failure during the course of follow-up. The data were reexamined with their pretreatment tests moved into the "nonresponder" category. Table 3 illustrates the tabulated results expressed using the same format as in Table 2. A greater separation between the responder group and the other two groups is noted. Specifically, the CRP score and physiologic scores rose in the responder group whereas the chest X-ray score (radiographic score) rose in the nonresponder group. The pretreatment CRP was higher in the responder group than in the stable or nonresponder groups (p < 0.05). No difference was noted between the stable and nonresponder groups. In addition, FVC, FEV₁, and TLC were statistically lower in the responder group compared with the nonresponder and stable groups. The radiographic score was higher in the nonresponder than either the responder or stable groups (p < 0.05). Additional changes were seen when comparing pathology scores among the groups. A statistically higher cellular, desquamative, and fibrotic score was noted in the responder group compared with the stable group (p < 0.05), whereas the nonresponder group demonstrated a statistically higher fibrotic score than the other two groups (p < 0.05). Furthermore, the total pathology score was higher in the nonresponder group than in the stable individuals (p < 0.05). Figure 2A illustrates the various pretreatment HRCT parameters among the three groups. There was a higher ground glass score in the responder group compared with the other two groups (p = 0.0007) while the fibrotic score was now greater in the nonresponder group compared with the stable and responder groups (p = 0.002). Figure 2B illustrates the pretherapy HRCT findings grouped by survival status at final follow-up. The group surviving demonstrated a higher CT-alv (p = 0.02) and lower CT-fib score (p = 0.003) than those who died during the course of follow-up.

View larger version (39K): [in this window] [in a new window]   Figure 2.   (A) Pretherapy HRCT scoring grouped by longer term follow-up. Responders (solid columns) experienced a 10-point drop in CRP score during 3 mo of steroid therapy and long-term survival; nonresponders (cross-hatched columns) experienced a 10-point rise in CRP score during 3 mo of steroid therapy or death during all treatment; and stable individuals (columns with diagonals) experienced a less than 10-point change in CRP score during 3 mo of steroid therapy and long-term survival. *p < 0.05. (B) Pretherapy HRCT scoring grouped by survival status at last follow-up. The solid columns represent those alive at last follow-up, and the columns with diagonals represent those patients who have died. *p < 0.05.

The group of patients demonstrating an initial response to high-dose steroid therapy or remaining stable during steroid therapy (responder + stable) were significantly different from the group that did not respond to steroid therapy or who died during long-term follow-up (nonresponder). The nonresponder group was older (62.3 ± 2.1 versus 48.6 ± 0.4 yr, p = 0.0006), and exhibited a higher chest X-ray score (6.2 ± 0.3 versus 4.2 ± 0.6, p = 0.003), HRCT fib score (2.0 ± 0.1 versus 1.2 ± 0.2, p = 0.001), pathology fibrosis score (13.3 ± 0.7 versus 8.7 ± 1.0, p = 0.001), and total pathology score (27.5 ± 1.5 versus 21.8 ± 1.6, p = 0.01). No difference was seen in CRP score (43.7 ± 3.6 versus 44.7 ± 5.0, p = 0.87), HRCT ground glass score (1.0 ± 0.1 versus 1.4 ± 0.3, p = 0.19), or pathology cellularity (7.0 ± 0.5 versus 6.4 ± 0.4, p = 0.29).

ROC Analysis to Define the Ability of Pretreatment Tests to Predict Death during Follow-up

To better define the ability of pretreatment diagnostic studies to predict the likelihood of death during follow-up, we performed ROC analysis comparing all initial diagnostic parameters (clinical, radiographic, and pathologic) in those patients who died (n = 10) versus those still living during a mean of 34.0 ± 2.5 mo of follow-up (n = 28). Clinical parameters including dyspnea and duration of symptoms did not reliably predict death. In addition, no pulmonary function parameter, individually or as weighted in the CRP, demonstrated a statistically significant ability to predict death during follow-up. Only the initial HRCT fibrotic score (Figure 3A, p = 0.009) and fibrotic pathology score (Figure 3A, p = 0.03) demonstrated statistical significance. The HRCT ground glass opacity score did not serve as an independent predictor of death during follow-up. A HRCT fibrotic score  2.0 demonstrated a sensitivity of 80% and specificity of 85% while a pathology fibrotic score  16.0 demonstrated a sensitivity of 84% and specificity of 67% in predicting death during therapy. Table 4 illustrates the sensitivity and specificity of various values for these two parameters. The addition of physiologic measurements, CRP score, or pathologic findings to the pretreatment HRCT fibrotic score did not improve the ability to predict death during follow-up. Figure 3B illustrates the ROC curve for CT-fib and the ROC curve for the CT-fib plus CRP score. Both curves demonstrate a significant ability to predict death although no difference was seen between the two (p = 0.31). Although the peak p(A  a)O₂ correlated strongly with pathologic cellularity (r = 0.45, p = 0.008), granulation (r = 0.50, p = 0.003, desquamation (r = 0.40, p = 0.02), and fibrosis (r = 0.72, p < 0.00001, it did not demonstrate a significant ability to predict death during follow-up.

View larger version (13K): [in this window] [in a new window]   Figure 3.   (A) ROC analyses of CT-fib score (thin line) and pathologic fibrosis (thick line) versus long-term survival in 38 patients with IPF (10 deaths during follow-up). Both are statistically significant (HRCT fib, p = 0.009 and pathology fibrosis, p = 0.03). (B) ROC analyses of HRCT fib score (thin line) and CT-fib score plus CRP score (dashed line) versus long-term survival in 38 patients with IPF (10 deaths during follow-up). The difference in the area under the curve for both ROCs is not statistically significant (p = 0.31).

Response to 3 mo of Steroid Therapy and Long-term Survival

The effect of early steroid response on long-term survival is illustrated in Figure 4. A Kaplan-Meier survival analysis demonstrates a significantly better survival in the responder and stable groups compared with the nonresponder group (p = 0.01). No difference in survival was evident between the responder and stable groups. Figure 5 illustrates a Kaplan-Meier survival analysis for the patients as grouped by the pretherapy HRCT fib score ( 2 or < 2). A statistically significant difference is noted with much improved survival over the course of follow-up for those subjects with a pretherapy HRCT fib score < 2 (p = 0.008). The analysis remained similar when age, smoking history, time to onset of disease, and peak p(A  a)O₂ were adjusted for (data not shown).

View larger version (10K): [in this window] [in a new window]   Figure 4.   Kaplan-Meier survival analysis grouped by response to initial 3 mo of high-dose corticosteroids. A significantly better survival is seen in those individuals responding to (thick line) and those demonstrating little change (dashed line) during 3 mo of steroid therapy versus those not responding to steroid therapy (thin line), p = 0.01.

View larger version (11K): [in this window] [in a new window]   Figure 5.   Kaplan-Meier survival analysis grouped by initial HRCT fib score. A significantly better survival is seen in those individuals with a score  2 (thick line) versus those with an initial score > 2 (dashed line), p = 0.008.

Response to 3 mo of Steroid Therapy and Long-term Pulmonary Function

We examined the relationship between the response to a 3-mo trial of steroid therapy and a > 10% decrease in FVC, TLC, and DL_CO individually, in any grouping of two, or in all three pulmonary function parameters in the surviving patients in the three response groups. No significant difference was seen between the initial response to 3 mo of steroid therapy and a subsequent fall in FVC or TLC. Figure 6 demonstrates that a greater likelihood of a 10% drop in DL_CO was noted in those individuals not responding to initial steroid therapy. To more completely define predictors of long-term treatment failure, we grouped death or dropping pulmonary function as indices of treatment failure. A greater likelihood of death or > 10% drop in FVC was noted in those patients not responding to 3 mo of high-dose steroids (p = 0.004). A similar outcome was noted for death or a > 10% drop in TLC (p = 0.005) and death or a > 10% drop in DL_CO (p = 0.0003). In addition, a similar result was documented with a 20% drop in DL_CO (data not shown). Of interest, those individuals remaining stable after an initial course of steroids demonstrated the best course.

View larger version (13K): [in this window] [in a new window]   Figure 6.   Kaplan-Meier analysis of those individuals demonstrating > 10% drop in DLCO during longer term follow-up in 28 patients grouped by the initial response to 3 mo of steroid therapy. A significantly greater proportion of individuals showing a greater than 10-point rise in CRP score (thin line) demonstrated a > 10% drop in DLCO during longer term therapy compared with those individuals demonstrating a greater than 10-point drop in CRP (thick line) or a less than 10-point change in CRP score (dashed line) during three months of steroid therapy.

The responder or stable groups demonstrated no change in baseline to last measured FVC, FEV₁, TLC, or DL_CO as an absolute or percentage of predicted. The surviving nonresponders experienced a statistically significant drop in FVC (2.7 ± 0.1 to 2.4 ± 0.2 L, p = 0.047) and DL_CO (49.4 ± 4.1 to 40.6 ± 3.2 percentage of predicted, p = 0.01).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Idiopathic pulmonary fibrosis remains a difficult therapeutic dilemma with a limited response to immunosuppressive therapy. The development of a therapeutic plan for IPF is difficult. Five-year survival is estimated at 30 to 50% (32) with only 6 to 54% of patients responding to steroids or immunosuppressive therapy (32). In addition, steroid therapy is associated with a high incidence of adverse side effects. Predicting the likelihood of response to such therapy has proven difficult (33) but could potentially avoid significant toxicity associated with immunosuppressive therapy in these patients (1, 2, 9). We tested the hypothesis that initial clinical, radiographic, or pathologic features could predict response to a defined regimen of immunosuppressive therapy.

Several important findings emerge from this study of patients with biopsy-proven IPF: (1) Pretherapy clinical features, including a lower age, worse pulmonary function and exercise gas exchange, are statistically different in individuals likely to respond to a 3-mo trial of high-dose steroids than in those who remain stable or are nonresponders, but do not predict longer term mortality; (2) pretherapy open lung biopsy findings of increased cellularity define individuals likely to respond to a 3-mo trial of high-dose steroids while lesser fibrosis appears to be associated with a lower mortality; (3) pretherapy HRCT is the best technique to define individuals likely to respond to a 3-mo trial of high-dose steroids (higher pretherapy HRCT ground glass scores) and have lower longer term mortality (lower pretherapy HRCT fibrosis scores); and (4) response to a 3-mo trial of high-dose corticosteroids is associated with improved long-term survival.

Pulmonary function (including exercise gas exchange) is more impaired prior to therapy in individuals who demonstrate a short-term response to 3 mo of corticosteroids. There were few differences in individual pulmonary function parameters between patients who responded, failed (nonresponders), or remained stable during 3 mo of steroid therapy, however. This may reflect the poor correlation between individual pulmonary function studies and histologic findings in IPF (14, 15). When grouped into a physiologic score (16), clear differences were evident. When failure was defined as a drop in CRP score or death during longer term follow-up, more dramatic differences were noted. The group who failed steroid therapy had statistically lower physiologic scores. The limitation of physiologic parameters to accurately predict longer term survival is highlighted by the three individual patients (two short-term responders and one stable after 3 mo of steroid therapy) who died during the course of follow-up. Importantly, ROC analysis failed to reveal a level of pretreatment pulmonary function that demonstrated clinically acceptable sensitivity and specificity to define survival. Much of the mortality in our study occurred early during the course of therapy and likely reflects a referral bias to our tertiary referral center. On the other hand, the greater severity of disease in many patients allows us to draw conclusions regarding identification of more advanced disease and a higher likelihood of mortality.

Prior investigators have attempted to correlate the outcome of therapy with pulmonary function parameters. In retrospective studies, worse survival and/or response to therapy has been associated with lower pretreatment diffusing capacity (6, 7), TLC (9, 10), FVC (10), or higher pretreatment FVC (7) (when analysis was limited to patients with an FVC < 90% predicted) (11), and a higher FEV₁/FVC ratio (12). A lack of consensus is evident from these studies, however. These studies are limited by a number of factors, including: lack of standardization of therapy between treatment groups (6, 7, 10, 12); presence of systemic disease in a significant percentage of the population studied (6, 9) with a possible improvement in survival (34); and varying diagnostic criteria without consistent use of lung biopsy (6, 8, 9, 11). Our study evaluated a patient population with open biopsy-proven IPF and utilized a standard protocol of prospective treatment and follow-up. We show that there is a patient population that, prior to treatment, has lower pulmonary function (FVC, TLC, and DL_CO) and a better response to 3 mo of steroid therapy. However, these parameters, individually or grouped in a CRP score, were unable to reliably predict survival during longer term therapy and follow-up.

We used a composite clinical, radiographic, and physiologic scoring system to assist in the assessment of disease. A similar composite score has been shown to improve the correlation with histologic abnormalities in IPF (16). Short-term response to therapy using changes in CRP scoring have been reported (16, 17) although the use of such a CRP score during longer term therapy has not been reported. Furthermore, independent validation of this composite score has not been reported. We clearly show statistically significant differences in pretreatment CRP scores in individuals exhibiting varying responses to short-term steroid therapy. A statistically higher CRP score is seen in those more likely to respond to such immunosuppressive therapy. However, no absolute, pretreatment CRP score could be identified which exhibited an acceptable sensitivity or specificity in predicting either response or survival. Although peak exercise p(A  a)O₂ correlated strongly with histologic indices, it was not an independent predictor of response to therapy or long-term survival. These findings should not be interpreted as negating the value of physiologic parameters in the management of patients with IPF as we used changes in CRP score to guide therapy.

The cellular pathology score was higher in patients who responded to steroid therapy than those who remain stable. When failure was defined as a rise in CRP score (16) during 3 mo of steroid therapy or death during long-term follow-up, clearer differences in pathologic findings were noted. Patients who were nonresponders demonstrated lower pathology cellular but higher pathologic fibrotic scores than those who did not respond or remained stable during steroid therapy. Those who responded to therapy demonstrated higher cellular, desquamative and higher fibrotic scores than patients who remained stable. Our findings validate prior retrospective studies which suggested a greater response to steroids in patients with a more cellular lung biopsy (2, 3). ROC analysis documented that a parenchymal fibrosis score  16 predicted death with clinically acceptable sensitivity and specificity. This finding could be of considerable value in determining therapy for patients with IPF. Patients with fibrosis scores of  16 could be advised that present therapy is unlikely to be of benefit; such patients would avoid the side effects and cost of currently accepted therapies. Such patients could be ideal candidates for novel therapies if they can understand the potential risks, benefits, and implications of these studies and consent to participate in such trials.

An important finding of our study is that HRCT findings predict mortality with greater sensitivity and specificity than histopathological features from open lung biopsy. We have found that a HRCT fibrotic score > 2 is highly predictive of greater mortality. Furthermore, we document a greater ground glass score and lower fibrotic score in individuals responding to 3 mo of steroid therapy. The addition of pulmonary function, pathologic data, or HRCT ground glass scores to HRCT fibrosis scores did not improve the ability to predict survival. The scoring system used in the current study has demonstrated low interobserver variability (kappa statistic 0.51-0.83) (29). Furthermore we have demonstrated an excellent correlation between HRCT fib scores and pathology fibrosis score but a lesser, although statistically significant, correlation between HRCT ground glass opacity and pathologic cellularity scores (29). The latter could represent sampling error as the surgeon was asked to biopsy areas with abnormality on HRCT and/or visual inspection but not based on percutaneous CT-guided needle localization. As such, this potential error may have impacted on the ability of HRCT ground glass scores to predict long-term outcomes. Higher HRCT ground glass scores were seen in those patients responding to 3 mo of high-dose steroids. Our study strongly supports the ability of HRCT to predict response to immunosuppressive therapy and longer term survival. All patients in our study had histologic confirmation of IPF and treatment was guided by changes in CRP score. As such, our data cannot be interpreted as suggesting that HRCT should be the sole diagnostic modality in patients with suspected pulmonary fibrosis.

Prior retrospective studies have correlated histologic findings in IPF and HRCT. In this respect, the amount of "ground glass opacity" has been correlated with the degree of pulmonary function (18, 35) and pathologic abnormality (19). Previous studies have also examined the ability of HRCT findings to predict outcome of therapy (18, 21, 22). In a retrospective study of 76 patients with IPF, a CT picture of predominant ground glass attenuation was associated with improved survival (21). All of these studies are limited by their retrospective nature (18, 20), lack of consistent histologic diagnosis (18, 20), and variable immunosuppressive therapy and follow-up (18, 20). More recently, 12 patients with usual interstitial pneumonitis (UIP) were contrasted with 11 suffering from desquamative interstitial pneumonitis (DIP) (4). The latter were more likely to respond to therapy with improvement in ground glass opacity, while those with UIP progressed to fibrosis despite therapy. This study was also limited by varying treatment regimens and follow-up data. In the current study we examined a large group of patients with untreated, open-biopsy-proven IPF (37 UIP and 1 DIP) and no systemic disease, treated in consistent fashion with standardized therapy and follow-up. Furthermore, HRCT scans were independently interpreted by four thoracic radiologists blinded to clinical history and biopsy findings.

Our data support the prognostic value of a therapeutic trial of high-dose corticosteroid therapy. Those individuals who did not respond after 3 mo of therapy suffered greater mortality during follow-up. Furthermore, these surviving nonresponders were more likely to demonstrate greater than 10% reductions in DL_CO, TLC, and FVC. This magnitude of change in pulmonary function has been used by others to define long-term response to therapy (1, 2, 9, 11, 12). The drop in FVC and DL_CO was statistically significant in surviving nonresponders between the initial pulmonary function testing and that at the last follow-up. Our data support previous retrospective data suggesting an improved survival and pulmonary function in those patients responding to prednisone therapy (2).

An additional important finding of our study is the long-term stability noted in many individuals who demonstrated little change after an initial 3-mo course of steroids. These patients had a significantly lower mortality than nonresponders and appeared remarkably similar to those patients who respond to steroid therapy. Furthermore, pulmonary function appeared preserved in the majority of these individuals during a mean of 39 mo follow-up. In this respect these patients resemble those individuals who respond to an initial 3-mo trial of steroids. We cannot exclude potential late decrements in pulmonary function with longer follow-up. The therapy of the groups in the current study differed depending on the initial response to steroid therapy. Those individuals not responding or remaining stable were subsequently treated with oral cyclophosphamide therapy based on previous literature suggesting a poorer survival in these individuals (2) and the potential advantage of cyclophosphamide therapy in patients with IPF (9). An examination of alternative modalities (azathioprine or colchicine) cannot be assessed on the basis of the data provided in this study. Further study appears warranted to better refine alternative therapy in patients not responding to initial high-dose steroid therapy.

This prospective study of patients with open-biopsy-proven IPF has generated several important findings: (1) Pretherapy clinical features, including pulmonary function and exercise testing, are statistically different in individuals likely to respond to a 3-mo trial of high-dose steroids but do not predict longer term mortality. (2) Pretherapy open lung biopsy findings of increased cellularity can identify individuals likely to respond to a 3-mo trial of high-dose steroids, while decreased fibrosis appears associated with a lower mortality. (3) Higher pretherapy HRCT ground glass scores identified individuals likely to respond to a 3-mo trial of high-dose steroids, whereas lower pretherapy fibrotic scores identified individuals with better long-term survival. (4) Response to short-term, high-dose prednisone therapy is associated with improved survival. (5) More than one-third of patients demonstrated stability during short-term prednisone therapy and stability with longer term follow-up. These findings suggest that greater physiologic abnormality, when interpreted in conjunction with HRCT and possibly open biopsy findings, will aid in identifying those patients most likely to benefit from immunosuppressive treatment strategies.