INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Idiopathic pulmonary fibrosis (IPF) is a term that has been used to describe a group of nongranulomatous interstitial lung diseases of unknown cause with different histopathology, responsiveness to therapy, and outcome. This group of diseases recently has been classified into subsets which include acute interstitial pneumonia, desquamative interstitial pneumonia, nonspecific interstitial pneumonia, and usual interstitial pneumonia (1). The most common of these is idiopathic usual interstitial pneumonia (IPF/UIP), which is characterized by relative unresponsiveness to corticosteroid and immunosuppressive therapy and a median survival of 3 to 5 yr after diagnosis (2). Prior studies have reported favorable responses to therapy in 25 to 30% of patients so treated (6), but most of these reports describe a heterogenous group of patients with IPF rather than the subset with IPF/UIP. When IPF/UIP is distinguished from the mixed group of IPF, the favorable responsiveness of patients with IPF/UIP to high-dose corticosteroid therapy and/or immunosuppressive drugs appears to be uncommon (11, 12).

Because of the poor objective clinical response to corticosteroid and immunosuppressive therapy and the high incidence of serious side effects associated with these programs (11), alternative therapies for the treatment of IPF/UIP have been explored. These programs focus on drugs that inhibit the evolution of fibrosis (15) or enhance apoptosis of lung fibroblasts (16) rather than on suppression of an inflammatory process. At Mayo Clinic Rochester (MCR), we have used colchicine, which has theoretical antifibrotic activity, as an alternative to a trial of high-dose prednisone therapy (12, 17, 18).

We have previously reported both retrospective and prospective trials of colchicine versus high-dose prednisone therapy for IPF/UIP. In both studies, we found a trend for better outcome using colchicine rather than prednisone (12, 18). In the prospective study, we also found that side effects were less with colchicine than with prednisone, and in addition, there was more rapid decline of pulmonary function during prednisone weaning in the prednisone treated group (12).

In recent years, we have treated a large group of IPF/UIP patients using a variety of treatment programs, including no treatment, colchicine only, and various combinations of lower-dose prednisone with or without colchicine or immunosuppressive agents. This retrospective study was designed to assess survival as a function of these different treatment programs, based on intent to treat in these patients. To enhance the clinical applicability of our findings, this study included previously untreated, previously treated, and currently treated patients. In addition, the clinical, radiographic, and histopathologic diagnostic criteria employed were similar to those used in current clinical practice at most secondary and tertiary referral centers. Each of these patients was thought to have IPF/UIP by the responsible clinician, and in each patient, either the surgical open lung biopsy (OLB) or computed tomography (CT) of the lungs or both was read by the clinical pathologist or clinical radiologist as being compatible with the diagnosis of IPF/UIP.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Records were reviewed for all patients coded as having pulmonary fibrosis (CPT code 515) or idiopathic pulmonary fibrosis (Current Procedural Terminology [CPT] code 516.3) who were seen at MCR during 1994 and 1995 (n = 1,411). In addition, we included patients seen with these diagnostic codes in the Division of Pulmonary and Critical Care Medicine at MCR during the year 1996 (n = 490). Patients were considered to have focal fibrosis, postinflammatory nonprogressive fibrosis, diffuse pulmonary fibrosis due to a known cause, or IPF/UIP by review of the clinical record. These records incorporate clinical data, laboratory studies, pulmonary function tests, radiographic interpretation, and biopsy reports. Focal fibrosis was defined as a solitary area of well circumscribed fibrosis in the lung without evidence of diffuse interstitial lung disease, and included localized radiation fibrosis. Postinflammatory fibrosis included single or multiple focal areas of fibrosis which on review of previous radiographs were thought to be the consequence of previous inflammatory processes because they did not progress over time, and included aspiration and infectious pneumonias. Known causes of diffuse pulmonary fibrosis for excluding patients from this study included connective tissue diseases, pneumoconioses, radiation therapy to the chest, fibrogenic drug therapy, neurofibromatosis, aspiration pneumonia, biopsy-proven nonspecific or desquamative interstitial pneumonia, or other diffuse infiltrative lung diseases such as sarcoidosis or hypersensitivity pneumonia.

Diagnosis of IPF/UIP required a balance of clinical, radiographic, and biopsy criteria. When surgical lung biopsy, read by an experienced pulmonary pathologist, and high-resolution CT scan (HRCT), read by an experienced pulmonary radiologist both were interpreted as showing typical findings of UIP, no clinical criteria other than the absence of an obvious cause were required. When CT findings were completely typical and no biopsy had been done, only the presence of typical bibasilar end-inspiratory crackles of "velcro" type and absence of an obvious cause were required. When CT findings were consistent with but somewhat atypical for a diagnosis of UIP and no biopsy had been done, then we also required the following clinical criteria: cough and/or dyspnea had to be present and of at least 6 mo duration; when previous chest radiographs were available, evidence of interstitial lung disease had to be present for at least a year; and if the interstitial lung disease had been present for 3 or more years evidence of disease progression by clinical, radiographic, or pulmonary function criteria had to be present.

Radiographic CT criteria for a diagnosis of UIP included the presence of a bilateral diffuse interstitial lung infiltrate with predominant reticular or honeycomb pattern, distributed preferentially to the lung bases and to peripheral and subpleural lung regions. Patients with predominant ground glass patterns or those with features of diseases other than UIP were excluded, as were those with fibrosis other than diffuse interstitial lung disease, including focal fibrosis and nonprogressive postinflammatory fibrosis.

Surgical lung biopsy criteria for a diagnosis of UIP included the presence of a patchy interstitial process with peripheral and subpleural predominance which demonstrated an appearance characterized by temporal heterogeneity, including areas of fibrosis, active fibroblastic foci and inflammation, and normal lung. The finding of granulomas, vasculitis, or any other specific interstitial lung disease than UIP excluded the patient.

Pulmonary function tests at MCR were performed using Medical Graphics (St. Paul, MN) equipment. Pulmonary function variables were expressed as percent predicted, and included TLC using plethysmography, vital capacity (VC), diffusing capacity for carbon monoxide (DL_CO) using the single-breath method, and alveolar volume (VA) using single-breath neon wash-in. Normal values were those of Miller and associates (19, 20).

Records were reviewed and data were recorded on a paper worksheet then later transferred to a computer spreadsheet by one of us (W.W.D.). Data included age, gender, vital status, place of residence, and whether or not referred because of interstitial lung disease. Lung biopsy results, CT scan findings, and the presence of crackles or digital clubbing on physical examination were recorded. When available, the presence or absence of progression of disease during the year prior to the index visit by review of and comparison with previous chest radiographs, pulmonary function tests prior to the index visit with decline of FVC by 15% and/or DL_CO by 20% from baseline at Mayo or at another facility, or symptoms (patient reported increase in cough or dyspnea) was documented. The history of the presence of a connective tissue disease, asbestos exposure, radiation therapy, therapy with drugs having the potential to cause pulmonary fibrosis, organic dust exposure, and smoking status, the date the patient was first seen with a diagnosis of pulmonary fibrosis within the study window (the index visit date), and the date of initial diagnosis of IPF/UIP at Mayo or elsewhere, were recorded. Treatment for pulmonary fibrosis, including oxygen therapy, prior to the first contact within the study window; and treatment for pulmonary fibrosis, including oxygen therapy, advised at the first MCR visit within the study window were documented.

Unless indicated otherwise, data are presented as mean ± 1 SD for continuous variables and percentages for discrete variables. Baseline patient characteristics were compared across recommended treatment groups using one-way analysis of variance for continuous variables and the chi-square test for discrete variables. For survival analysis, time zero is defined as the index visit, which was the date the patient was first seen at MCR during the study period (January 1, 1994 to December 31, 1996). Because survival after initial diagnosis or date of surgical biopsy has been described in other studies, we have included a descriptive summary of the time interval between the initial diagnosis and the index visit. Cumulative survival probabilities were estimated using the Kaplan-Meier method (21). Cox proportional hazards regression was used to identify variables associated with survival (22). Potential predictors considered in the survival analysis included continuous variables for age and baseline pulmonary function (VC, DL_CO, VA, Sa_O2 at rest); and binary indicator variables defining male gender; diagnosis by open lung biopsy; diagnosis by CT scan; presence of digital clubbing; smoking status (current or former versus never); evidence of disease progression at baseline (chest radiograph, pulmonary function, symptoms); prior pharmacologic treatment with prednisone, prior treatment with colchicine, prior treatment with cyclophosphamide or azathioprine); recommended treatment with the same drugs; prior O₂ therapy; and recommended O₂ therapy. TLC (plethysmography), FEV₁, and oxygen saturation during exercise were not considered in the survival analysis because these tests were frequently not performed in subjects having pulmonary function tests.

To assess the potential impact of referral practice on survival two additional potential predictor variables were included in the survival analysis: a binary variable indicating whether the patient was referred for interstitial lung disease; and a categorical variable describing the distance of the patient's primary residence to Rochester, Minnesota (< 100 miles, 100 to 249 miles, and > 250 miles). To identify a set of independent predictors of survival a multivariate analysis was performed using backward elimination of nonsignificant variables. After a multivariate model was derived from this approach, indicator variables defining the five recommended treatment groups were added to the model to obtain adjusted estimates for the relative risk (RR) associated with various treatments. For all survival analyses, time zero is defined as the date of the index visit, which is the date the patient was first seen with a diagnosis of IPF/UIP within the study window, January 1, 1994 to December 31, 1996. To describe survival from the date of first IPF/UIP diagnosis, all survival analyses were repeated for the subset of patients whose date of the first IPF/UIP diagnosis corresponded to the date of the index visit. In all cases, p values < 0.05 were considered statistically significant.

Surviving patients were contacted during 1998 for vital status, and were asked to complete a questionnaire describing current symptoms, exposures, and recent therapy to estimate the frequency of compliance with recommended treatment.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

During the study period (1994-1996), 1,901 patients were identified at MCR as having pulmonary fibrosis. Of the patients seen during 1994-1995, 34% were diagnosed as having IPF/ UIP, whereas 66% had clinical or radiographic findings inconsistent with IPF/UIP (31% had focal fibrosis, 19% had nonprogressive postinflammatory fibrosis, and 25% were thought to have fibrosis due to a known cause). For the entire study period (1994-1996), 581 patients were diagnosed by their physicians as having IPF/UIP. Of these 581 patients, 487 had either lung biopsy or CT scans that were read by the clinical pathologist or radiologist as being consistent with UIP. These 487 patients form the basis of this report.

Baseline patient characteristics are summarized in Table 1 according to recommended treatment. Treatment advised at the initial MCR visit within the study window included no treatment in 157 (32.2%), prednisone only in 54 (11.1%), colchicine only in 167 (34.3%), colchicine plus prednisone in 71 (14.6%), and other programs in 38 (7.8%). Oxygen therapy was recommended in 133 (27.7%) patients. Colchicine was used at a dosage of 0.6 mg once or twice a day as tolerated. Prednisone dosage usually had already been tapered to the range of 5 to 40 mg/d when the patients were initially seen or was being tapered from higher doses toward these maintenance levels. The dose and duration of previous prednisone therapy sometimes were incompletely documented in the record.

Baseline patient characteristics are summarized in Table 1 for each treatment group. Baseline pulmonary function tests showed a restrictive pattern in the majority, and are summarized in Table 2. TLC (plethysmography) was 68.8 ± 15.9%, VC was 69.1 ± 18.7%, DL_CO (single-breath method) was 50.8 ± 16.4%, VA (single-breath neon wash-in method) was 67.6 ± 13.4%, oxygen saturation (pulse oximeter) at rest was 92.4 ± 4.4%, and oxygen saturation during exercise was 85.3 ± 6.6%.

There were 250 patients (51.3%) who had not received drug treatment before the first MCR visit within the study window. Prednisone had been used in 202 patients (41.5%), colchicine in 70 (14.4%), cyclophosphamide in 30 (6.2%), azathioprine in nine (1.8%), hydroxychloroquine in three (0.6%), methotrexate in one (0.2%), an other agents in three (0.6%). In 123 of these patients (25.3%), only prednisone had been used, and in 24 patients (4.9%), only colchicine had been given. Oxygen therapy had been used by 74 patients (15.6%).

Median survival after the index visit for all patients with IPF/ UIP (n = 487) was 3.2 yr (Figure 1). For those patients whose initial diagnosis was at the initial visit (n = 190), median survival was 3.8 yr (Figure 2). From preliminary univariate analysis, for the group as a whole, worse survival was found to be associated with older age (p = 0.048, RR = 1.2 per decade), male gender (p = 0.001, RR = 1.7), patient referral for interstitial disease (p = 0.032, RR = 1.4), lower VC (p < 0.001, RR = 1.3 per 10 percentage points), lower DL_CO (p < 0.001, RR = 1.4 per 10 percentage points), lower VA (p < 0.001, RR = 1.4 per 10 percentage points), lower oxygen saturation at rest (p < 0.001, RR = 1.2 per 5 percentage points), evidence of disease progression by chest X-ray (p = 0.008, RR = 1.4), evidence of disease progression by pulmonary function (p < 0.001, RR = 1.8), evidence of disease progression by symptoms (p = 0.036, RR = 1.6), prior prednisone treatment (p = 0.001, RR = 1.5), prior cyclophosphamide or azathioprine treatment (p = 0.019, RR = 1.7), prior oxygen therapy (p < 0.001, RR = 2.2), and recommended oxygen therapy (p < 0.001, RR = 2.0). From univariate analysis, no significant difference was found when survival was compared across all recommended treatment groups simultaneously (p = 0.066). However, when compared with those for whom no therapy was advised, survival was worse for patients who were advised to take either prednisone only (RR = 1.5, p = 0.048) or prednisone plus colchicine (RR = 1.7, p = 0.011) (Table 3).

View larger version (8K): [in this window] [in a new window]   Figure 1.   Survival in years for the group as a whole (n = 487) after the index visit (the first visit with a diagnosis of IPF/UIP within the study window, January 1, 1994 to December 31, 1996). Median survival was 3.2 yr.

View larger version (8K): [in this window] [in a new window]   Figure 2.   Survival in years for the subset (n = 190) for whom the date of the initial diagnosis of IPF/UIP was the date of the index visit. Median survival was 3.8 yr.

From multivariate analysis, for the group as a whole, after eliminating nonsignificant variables, worse survival was found to be associated with older age (p = 0.024, RR = 1.2 per decade), male gender (p = 0.008, RR = 1.6), lower DL_CO (p < 0.001, RR = 1.4 per 10 percentage point decrease), lower VA (p < 0.001, RR = 1.2 per 10 percentage point decrease), and evidence of disease progression by pulmonary function (p = 0.021, RR = 1.5) (Table 4). After adjusting for age, gender, DL_CO, VA, and evidence of disease progression by pulmonary function, there was no evidence to indicate that survival was associated with recommended pharmacologic treatment (p = 0.720) or oxygen therapy (p = 0.700).

For 190 of the 487 patients seen during the study period, the date of the index visit corresponded to the date of the patient's initial IPF/UIP diagnosis. To describe survival after the date of first diagnosis, the survival analyses were repeated for this subset of patients. Median survival for these patients was 3.8 yr (Figure 2). From preliminary univariate analysis, worse survival was found to be associated with male gender (p = 0.004, RR = 2.3), lower VC (p = 0.002, RR = 1.3 per 10 percentage points), lower DL_CO (p < 0.001, RR = 1.3 per 10 percentage points), lower VA (p = 0.001, RR = 1.4 per 10 percentage points), prior prednisone treatment (p = 0.026, RR = 1.7), and recommended oxygen therapy (p = 0.020, RR = 2.3). From univariate analysis, no significant difference was found when survival was compared across all recommended treatment groups simultaneously (p = 0.089). From multivariate analysis, after elimination of nonsignificant variables, worse survival was found to be associated with male gender (p = 0.017, RR = 2.3), lower DL_CO (p = 0.019, RR = 1.2 per 10 percentage points), and lower VA (p = 0.033, RR = 1.2 per 10 percentage points (Table 4). After adjusting for gender, DL_CO, and VA, there was no evidence to indicate that survival was associated with recommended pharmacologic treatment (p = 0.845) or oxygen therapy (p = 0.534) (Table 3).

Questionnaires were returned with responses addressing compliance with recommended drug therapy by 214 patients. These respondents indicated that they had taken the recommended therapy for prednisone alone in 89%, for colchicine alone in 70%, and when prednisone and colchicine together were advised, they had taken prednisone in 88% and colchicine in 76% of instances.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this retrospective intent-to-treat analysis, we found no evidence to suggest that colchicine was any different than no therapy with respect to survival in the treatment of IPF/UIP. We conclude that a no treatment arm is an acceptable alternative for use in future studies of drug therapy for IPF/UIP. Although we found a significant decrease in survival in patients treated with either prednisone alone or prednisone plus colchicine compared with those treated with no therapy on univariate analysis, when corrections were made for age, gender, and disease severity by multivariate analysis, there was no difference in outcome. We therefore conclude that maintenance treatment using low to moderate doses of prednisone does not adversely impact survival when compared with no therapy when used for the treatment of IPF/UIP.

By itself, this study does not provide useful information regarding the efficacy of high-dose prednisone in the treatment of IPF/UIP. No placebo-controlled trials have been performed to evaluate the efficacy of high-dose prednisone in the treatment of IPF/UIP. Our previous retrospective (18) and prospective (12) studies of colchicine versus prednisone in the treatment of IPF/UIP suggested a trend to better outcome in patients treated with colchicine compared with those treated with prednisone. In addition, our prospective study found that treatment with colchicine results in fewer side effects than does high-dose prednisone (12). Selman and associates found that survival for patients treated with colchicine and prednisone together was not different than for prednisone alone (23). Others have reviewed the efficacy and complications of high-dose prednisone in the treatment of IPF/UIP (6, 11, 13, 14, 18). Taken together, these studies raise the question of whether no therapy would be more appropriate than trials of high-dose prednisone in the treatment of IPF/UIP, provided the diagnosis is certain.

Oxygen therapy has the theoretical potential to either prolong or shorten survival in patients with IPF/UIP. As the disease worsens, hypoxemia predictably causes or worsens pulmonary hypertension, which can lead to cor pulmonale and right heart failure. Oxygen therapy may reverse the hypoxic component of the pulmonary hypertension caused by IPF/UIP. However, improved survival associated with oxygen therapy in diffuse interstitial lung diseases has been extrapolated from results observed in studies of patients with emphysema (24, 25), and may not be relevant to patients with IPF/UIP. At the cellular level, IPF/UIP may in part be an oxidant-mediated disease (26), and oxygen therapy might be expected to increase tissue concentrations of toxic oxygen radicals. Alveolar extracellular concentrations of glutathione, the major antioxidant in alveolar fluid, are severely depleted in bronchoalveolar fluid from patients with IPF/UIP, and antioxidant therapy using N-acetylcysteine has been reported to improve the concentration of glutathione in the bronchoalveolar fluid of patients with IPF/UIP (27), and also to decrease the rate of decline of pulmonary function in IPF (28).

Our univariate analysis of survival showed that a significant predictor of a shortened survival (p  0.001) was oxygen therapy, either before the initial MCR visit, or advised at the time of the initial MCR visit. Winget and associates have reported similar findings in a group of IPF patients treated with intermittent intravenous cyclophosphamide (29). It was therefore unclear to us whether the decreased survival associated with oxygen therapy in IPF/UIP seen on univariate analysis was simply a marker for severity of disease or reflected an adverse effect on survival caused by the oxygen therapy. However, from a multivariate analysis which adjusted for age, gender, and disease severity, we found that oxygen therapy was not an independent variable associated with shortened survival. We therefore conclude that oxygen therapy does not appear to accelerate decline in IPF/UIP, and should be used as clinically indicated in the treatment of arterial hypoxemia, pulmonary hypertension, and cor pulmonale.

Although it may seem intuitively obvious that a history of radiographic, functional, or clinical worsening during the previous year is associated with a worse prognosis, the majority of the data addressing this issue relates to the change in pulmonary function or CT scans after a period of corticosteroid therapy (13, 30, 31). It is not yet clear whether the favorable response to corticosteroid therapy reported in some patients with IPF is a reflection of treatment, nonlinearity of the rate of decline of function as part of the natural history of IPF/UIP, or misdiagnosis of patients who actually have more steroid- responsive diffuse interstitial lung diseases such as nonspecific interstitial pneumonia or desquamative interstitial pneumonia (32). Studies describing steroid responsiveness if IPF have not always discriminated between these subsets of IPF (13, 30, 31). Our findings suggest that progression of pulmonary function impairment within the previous year predicts worse prognosis regardless of treatment advised.

IPF/UIP can be defined using several criteria. In this study, all patients were thought to have IPF/UIP by their clinical physicians, and either the lung biopsy and/or the CT scan was interpreted by the surgical pathologist or radiologist as showing findings consistent with UIP. The diagnosis of IPF/UIP was supported by clinical plus radiographic HRCT criteria in 94.7% of study subjects but OLB was done to confirm the diagnosis in only 20.3%. This surgical biopsy rate is higher than the 7.5 to 13% rate reported in two large clinical studies in the United Kingdom (3, 33) and the 11.1% rate reported in a study from New Mexico (34). The median survival of 3.2 yr for the group as a whole, and 3.8 yr for those whose initial diagnosis was at the index visit is similar to that reported previously in biopsy-confirmed series of patients with IPF/UIP (2, 4, 5), suggesting that the incidence of erroneous diagnosis in this study in small. Because the requirement for all patients with suspected IPF/UIP to have surgical lung biopsy appears to be unrealistic in clinical practice, the findings of a recent prospective multicenter study of CT plus OLB in diffuse interstitial disease (35, 36) are of considerable clinical importance. Previous studies comparing HRCT scans of the lungs with OLB diagnosis in IPF/UIP have suggested a high positive predictive value when the HRCT diagnosis is "definite" (13, 32). The multicenter study found that OLB confirmed UIP in 96% of those whose CT scans were read as showing "definite" UIP. However, only approximately 50% of patients with UIP on OLB in this study had "definite" HRCT findings, and the authors therefore estimated that OLB would be necessary in the remaining 50% (36). Although we were aware of these recommendations, we chose to use less restrictive criteria to enhance the applicability of our findings to usual clinical practice. Similarly, although quantitation of the extent of fibrosis in biopsies and CT scans has been shown to be useful in predicting outcome (13, 30), this is not standard clinical practice at this time, and was not included in this study.

There are obvious limitations inherent in the design of this investigation. As shown in Table 1, the choice of recommended treatment was likely influenced by many baseline patient characteristics. In general, prednisone was more likely to be advised for patients who were already taking prednisone, whereas either colchicine or no treatment was more likely to be advised for those who had already failed a trial of prednisone or for previously untreated patients. To address this limitation, we performed a multivariate analysis to identify characteristics associated with survival. From multivariate analysis after adjusting for age, gender, DL_CO, VA, and evidence of disease progression by pulmonary function, we found no difference in survival across treatment groups.

Although this study reflects a patient population seen at only one institution, patients referred and nonreferred, and patients living nearby as well as those coming from considerable distances were represented (Table 1). We found that treatment was more likely to be advised for patients referred to MCR for diagnosis and treatment of their lung disease compared with those not so referred. Although referral for interstitial disease was a significant predictor of worse survival by univariate analysis, it was not statistically significant on multivariate analysis. We found no differences in survival between those residing at a distance of more than 100 miles or more than 250 miles from Rochester, Minnesota compared with those living nearer.

We conclude that the effect of colchicine therapy, previously shown to be safer than prednisone therapy in terms of side effects (12), is probably not different than the effect of no therapy on survival in IPF/UIP. Low-dose corticosteroid therapy does not appear to be harmful in terms of survival, and can be used safely for other conditions such as asthma, which may coexist with IPF/UIP. Oxygen therapy does not appear to shorten survival when used for treatment of arterial hypoxemia, pulmonary hypertension, and cor pulmonale in IPF/UIP.