METHODS

Section:

ChooseTop of pageAbstractMETHODS <<RESULTSDISCUSSIONReferencesCITING ARTICLES

Between January 1, 1978, and December 31, 1998, 104 patients met clinical and open-lung or thoracoscopic biopsy criteria for fibrotic IIP and make up the study population. These comprise all biopsied patients included in our files as cryptogenic fibrosing alveolitis, a term considered equivalent to idiopathic pulmonary fibrosis. A further six patients with cellular NSIP were excluded. Part of the cohort has been reported previously in studies of outcome against histopathologic diagnosis in IIP (10, 11). In patients presenting before 1987 (n = 42), clinical criteria comprised the following: (1) bilateral predominantly basal crackles, (2) evidence of predominantly basal bilateral lung fibrosis on chest radiography with no consolidation, (3) a restrictive functional defect (total lung capacity of less than 80% of predicted, in association with a restrictive FEV₁/FVC ratio) or isolated reduction in diffusing capacity (Dl_CO), and (4) no known primary cause for pulmonary fibrosis (17). After 1987 (n = 64), computed tomography criteria were used (predominantly basal/subpleural distribution; a mixture of reticular and ground-glass abnormalities, with traction bronchiectasis when ground-glass attenuation was prominent; the absence of consolidation or nodules) (18) in addition to clinical criteria.

Slides were reviewed by two pulmonary histopathologists (A.G.N. and T.V.C.) who were without knowledge of clinical data, and they were classified according to consensus criteria (2, 19). When classification differed between pathologists (κ coefficient of agreement = 0.46), a joint opinion on the overall histopathologic pattern was reached. When biopsies were taken from two sites (n = 58, 56%), both were interpreted; a pattern of UIP in one biopsy was taken to indicate an overall diagnosis of UIP.

Clinical data (smoking status, drug treatment, clinical findings, absence of previous environmental exposures, and connective tissue disease) were extracted from case records. Patients were categorized as nonsmokers, current smokers, or ex-smokers (a minimum of one cigarette a day for a minimum of 1 year, stopping at least 6 months before presentation). In all but two cases, both with mild disease, patients were treated during follow-up. The treatment regimen consisted of (1) high-dose prednisolone initially (40–60 mg daily), reducing to a maintenance average daily dose of 10 mg, or (2) prednisolone 10 mg daily in combination with azathioprine or cyclophosphamide. Changes were made during follow-up (introduction of or changes in immunosuppressive agents) as clinically warranted; for this reason and because efficacy differs little between these regimens (20, 21), treatment effects were not evaluated.

Analyzed lung function measurements consisted of FVC, FEV₁ (both measured using an Ohio water-seal spirometer; Ohio instruments, Atlanta, GA), and carbon monoxide Dl_CO using a single breath technique or a rebreathing technique with adjustment to single breath values (P.K Morgan respirometer; P.K Morgan, Chatham, Kent, UK). Results were expressed as percentages of predicted values (22). The recently defined composite physiologic index (CPI) (23), derived from FEV₁, FVC, and Dl_CO levels (CPI = 91.0 – [0.65 × percent predicted Dl_CO] − [0.53 × percent predicted FVC] + [0.34 − percent predicted FEV₁]), which corrects for the confounding effects of concurrent emphysema, was also evaluated. Measurement of earlobe capillary gases was performed on air at rest. Serial trends in pulmonary function tests (Dl_CO, FVC, FEV₁, all quantified as percentage change from baseline, and unadjusted changes in CPI scores) were defined in survivors at 6 months (4–8 months, n = 85) and at 1 year (9–15 months, n = 72). Trends in earlobe capillary gases (not repeated in a significant minority) were not evaluated.

Group comparisons were made using unpaired t tests, Wilcoxon's rank sum tests, or chi-squared statistics as appropriate. Survival was evaluated using proportional hazards regression (24) (STATA software; Computing Resource Center, Santa Monica, CA). The prognostic values of clinical indices, histopathologic diagnosis, functional indices at baseline, and serial functional trends were examined at presentation and after 6 and 12 months of follow-up.

DISCUSSION

Section:

ChooseTop of pageAbstractMETHODSRESULTSDISCUSSION <<ReferencesCITING ARTICLES

In fibrotic IIP, the histopathologic distinction between UIP and fibrotic NSIP provides useful prognostic information (8–11). However, the considerable overlap in longitudinal behavior between these two entities poses considerable difficulties in prognostic evaluation. We compare, for the first time, the prognostic values of short-term serial pulmonary function trends and the histopathologic diagnosis. After 6 months of follow-up, changes in functional indices (especially changes in gas transfer levels) were as powerful a determinant of outcome as the original histopathologic diagnosis. After 12 months of follow-up, the distinction between UIP and fibrotic NSIP provided no additional prognostic information, once serial pulmonary function trends had been taken into account. Thus, our findings confirm the anecdotal impression of many clinicians that in the difficult clinical problem of fibrotic IIP, longitudinal behavior is a more accurate determinant of outcome than evaluation at a single point in time. Changes in gas transfer levels provided the most accurate indication of survival, with no additional information gained by evaluating changes in other pulmonary function indices. However, changes in other indices at 1 year, including the recently described CPI (23), were also substantially more predictive of outcome than the histopathologic diagnosis.

A further novel aspect of this study was the separation of outcome into early and late mortality, not attempted formally in previous comparisons between UIP and NSIP. The fact that mortality in the first 2 years was primarily linked to the severity of pulmonary function impairment and not to the histopathologic diagnosis has important clinical implications for the initial management of fibrotic IIP. The role of thoracoscopic biopsy needs to be considered critically in the context of functionally severe disease, in which fibrosis appears to be established. In a minority of patients (approximately 25% in the present cohort) with Dl_CO levels less than 35% of predicted normally, biopsy evaluation provided no prognostically useful information. This cutoff was selected for two reasons. This level of functional impairment should stimulate consideration of transplantation (25) (in view of the need to avoid a delay in pretransplant investigations in inexorably progressive disease and the predicament of referring for transplantation too late). However, importantly, the risks of thoracoscopic biopsy may rise significantly as gas transfer levels approach 30% of predicted normal (26). The lack of a histopathologic diagnosis in this situation is stressful for clinicians. However, these data indicate that a UIP-like course is usual in this context, even when a diagnosis of fibrotic NSIP has been secured.

This observation is likely to explain a striking dichotomy in the determinants of early and late mortality in our cohort. During the first two years of follow-up, the level of functional impairment was the sole major determinant of survival in the combined cohort, and this can be ascribed to the strong link between early death and severe functional impairment, in UIP and fibrotic NSIP alike. In contrast, in individuals with less severe functional impairment, the longitudinal behavior of fibrotic NSIP appears to be truly heterogeneous, whereas UIP continues to progress in almost all cases: Thus, the histopathologic diagnosis was the cardinal determinant of late mortality. A decline in serial lung function indices at 1 year was much more prevalent in UIP than in NSIP, although a minority of NSIP cases with the longitudinal functional behavior of UIP at 1 year appear to have a UIP-like outcome.

From the overall survival profile in this study and the contrasts in longitudinal behavior within the NSIP cohort, it can thus be concluded that there are two important groups of patients associated with a histopathologic diagnosis of fibrotic NSIP. A large subset, identifiable by their functional severity at presentation or a decline in functional indices at 6 to 12 months despite treatment can be regarded, for practical therapeutic purposes, as equivalent to UIP. In contrast, a long-term survival of 40% in the fibrotic NSIP cohort is indicative of a separate important subset with a much better outcome.

Flaherty and colleagues have argued that the histopathologic distinction between UIP and (at least some cases of) fibrotic NSIP is flawed because of heterogeneity between and even within biopsied lobes in histopathologic classification (27). They showed that in patients with multiple biopsies, not all lobes showed the same histology: Those with UIP in one biopsy and NSIP in another were termed discordant UIP (approximately 25% of patients). The remaining 75% were either concordant UIP (approximately 50%) or concordant NSIP (approximately 25%). The authors suggested that some cases of NSIP may “evolve” to UIP, a hypothesis that has not been proven but might provide some support for the recent concept of “active” and “inactive” UIP (12). The observations of Flaherty and colleagues might also support our concept that there may be subgroups among patients with fibrotic NSIP, some of whom fare better than others. It appears clear that a separate entity of “true” fibrotic NSIP among the IIPs exists, with no evolution to UIP, judging from outcome findings in a subset of patients in this study. Thus, our findings highlight the importance of distinguishing between an NSIP/UIP overlap picture and true NSIP, a distinction probably best made using serial lung function trends.

The finding that serial trends in gas transfer were the most accurate determinants of outcome should be interpreted with caution. Serial trends in FVC were little inferior in this regard and may be more reliable than gas transfer trends in lung function laboratories without high exposure to interstitial lung disease, in view of the well recognized variability in gas transfer measurements (28). Serial trends in the recently defined CPI (23) did not add usefully to Dl_CO trends, although providing similarly strong prognostic information. This finding is not surprising; the value of the CPI lies in correcting for the confounding effects of emphysema, which is likely to be less prevalent in a younger population undergoing lung biopsy. Moreover, the value of an index at a single point in time does not always equate with its role in longitudinal monitoring. It is likely that the functional effects of emphysema change relatively little in 12 months; thus, changes in Dl_CO and FVC may be little affected by concurrent emphysema, despite major confounding by emphysema of measured values at presentation.

It should be stressed that our findings do not justify a policy of nonbiopsy at presentation in fibrotic IIP. First, our analysis presupposes that other histopathologic entities have been excluded, including inflammatory IIPs and hypersensitivity pneumonitis, and this conclusion cannot always be made confidently using noninvasive modalities. More importantly, the prognostic value of longitudinal functional behavior has been established in a treated cohort.

It appears possible, although unproven, that treatment is crucial in some patients with fibrotic NSIP, even if current regimens may be of limited value in UIP. In this cohort, the duration of dyspnea and levels of functional impairment were virtually identical at presentation in UIP and fibrotic NSIP, implying little difference in the rate of disease progression before presentation. Thus, it is possible that the good outcome in some NSIP patients in our cohort resulted from therapeutic intervention. If treatment is to be rationalized at presentation, immediate biopsy evaluation remains useful; however, our findings do suggest that biopsy is most useful when performed immediately, rather than when instigated following the observation of longitudinal behavior. It is also likely that biopsy evaluation will itself be refined, with additional prognostic information gleaned from the profusion of fibroblastic foci (29, 30) and clinical information (31) in proven UIP. Furthermore, lung function trends are especially useful when they show unequivocal change (decline or clear-cut stability). However, a large subset of patients exhibits marginal deterioration, which may represent either true decline or the “noise” of measurement. Thus, despite its cardinal prognostic role in the whole cohort in this study, longitudinal behavior over 1 year is not an infallible guide to outcome in every case.

In common with all histopathologic studies, the problem of selection bias needs to be considered. In the first half of the study period, a prospective policy was in place to biopsy all patients with suspected idiopathic pulmonary fibrosis. The group of patients with the most severe disease were necessarily excluded, but our findings indicate that in patients with more severe disease undergoing biopsy, the prognostic distinction between UIP and fibrotic NSIP adds little. Furthermore, in common with all biopsied cohorts from referral centers, the mean patient age was considerably lower than in populations drawn from secondary centers or the community. More important, in the last decade, computed tomography has been an increasingly important influence on the selection of cases to biopsy. In common with others (32, 33), we have tended to place most reliance on a computed tomography picture of predominantly peripheral subpleural honeycombing, as strongly indicative of underlying UIP, often obviating biopsy. Thus, there is undoubtedly some bias in the latter half of our cohort toward a biopsy diagnosis of fibrotic NSIP. However, in all of these regards, our patients may be typical of those in whom biopsy is generally performed, in the computed tomography era.

In conclusion, we have documented the considerable prognostic value in fibrotic IIP of serial pulmonary function trends at 6 and 12 months in a treated cohort. Longitudinal behavior is an invaluable prognostic supplement to lung biopsy at presentation. In treated patients, observed pulmonary function trends at 12 months may obviate biopsy in some cases. Our findings also disclose the marginal prognostic value of biopsy evaluation in functionally severe fibrotic IIP, underlining the potential importance of early biopsy evaluation in these difficult diseases.