Published ahead of print on November 1, 2007, doi:10.1164/rccm.200611-1696OC
© 2008 American Thoracic Society doi: 10.1164/rccm.200611-1696OC
Computed Tomography Findings in Pathological Usual Interstitial PneumoniaRelationship to SurvivalDepartments of 1 Radiology and 2 Medical Physics, Osaka University Graduate School of Medicine, Osaka, Japan; 3 Department of Pathology, Mayo Clinic, Scottsdale, Arizona; 4 Division of Respiratory Medicine, Saiseikai Kumamoto Hospital, Kumamoto, Japan; 5 Department of Respiratory Medicine and Allergy, Tosei General Hospital, Seto, Aichi, Japan; 6 Department of Respiratory Medicine, Kanagawa Cardiovascular and Respiratory Center, Yokohama, Kanagawa, Japan; 7 Department of Radiology, Dokkyo University School of Medicine, Tochigi, Japan; 8 Department of Radiology, Kurume University School of Medicine, Kurume, Japan; and 9 Department of Radiology, University of British Columbia and Vancouver General Hospital, Vancouver, Canada Correspondence and requests for reprints should be addressed to Hiromitsu Sumikawa, M.D., Department of Radiology, Osaka University Graduate School of Medical, 2-2 Yamadaoka, Suita, Osaka, 565-0825, Japan. E-mail: h-sumikawa{at}radiol.med.osaka-u.ac.jp
Rationale: Patients with a clinicopathological diagnosis of idiopathic pulmonary fibrosis (IPF) may have typical findings of usual interstitial pneumonia (UIP) on computed tomography (CT) or nonspecific or atypical findings, including those often seen in nonspecific interstitial pneumonia. Objectives: The aims of this study were to revisit the high-resolution CT findings of IPF and to clarify the correlation between the CT findings and mortality. Methods: The study included 98 patients with a histologic diagnosis of UIP and a clinical diagnosis of IPF. Two observers evaluated the CT findings independently and classified each case into one of the following three categories: (1) definite UIP, (2) consistent with UIP, or (3) suggestive of alternative diagnosis. The correlation between the CT categories and mortality was evaluated using the Kaplan-Meier method and the log-rank test, as well as Cox proportional hazards regression models. Measurements and Main Results: Thirty-three of the 98 CT scans were classified as definite UIP, 36 as consistent with UIP, 29 as suggestive of an alternative diagnosis. The mean survival was 45.7, 57.9, and 76.9 months, respectively. There was no significant difference in survival among the three categories (all P > 0.05). Traction bronchiectasis and fibrosis scores were significant predictors of outcome (hazard ratios: 1.30 and 1.10, respectively; 95% confidence intervals: 1.18–14.2 and 1.03–1.19, respectively). Conclusions: In patients with IPF and UIP pattern on the biopsy, the pattern of abnormality on thin-section CT, whether characteristic of UIP or suggestive of alternative diagnosis, does not influence prognosis. Prognosis is influenced by traction bronchiectasis and fibrosis scores.
Key Words: interstitial pneumonia • idiopathic pulmonary fibrosis • high-resolution computed tomography • lung
According to the American Thoracic Society (ATS) and European Respiratory Society (ERS) 2002 Working Group Consensus Classification of Idiopathic Interstitial Pneumonias (IIPs), the primary role of thin-section computed tomography (CT) is to separate patients with characteristic findings of idiopathic pulmonary fibrosis (IPF)/usual interstitial pneumonia (UIP) from those with other IIPs (1). Previous reports have stated that the characteristic CT findings of UIP include the following: honeycombing, reticular opacities, ground-glass attenuation, and both basal and peripheral predominance, which is often associated with traction bronchiectasis and architectural distortion (2–4). Thin-section CT is useful for differentiating IPF/UIP from other IIPs (5–7). However, even experienced radiologists sometimes confuse IPF/UIP with other IIPs, especially nonspecific interstitial pneumonia (NSIP) (5, 6, 8, 9). This implies that the CT findings of patients with a histologic diagnosis of UIP and a clinical diagnosis of IPF include not only patients with characteristic findings of UIP on CT but also patients with an NISP pattern or nonspecific or atypical CT findings. Few previous reports have described the CT findings of UIP diagnosed according to the current histopathological criteria and the frequency of patients with IPF who have atypical CT findings. In addition, it is still unclear which CT findings influence the prognosis. This study had two aims: to revisit the thin-section CT findings of IPF and to clarify the correlation between CT findings and mortality.
Patients and Diagnoses The ethical review boards of the three institutions that contributed cases to the present study did not require the patients' approval or informed consent for the retrospective review of their records and images. One hundred and fifty-four patients who underwent surgical biopsies at three institutions and who met the clinical and histologic criteria for diagnosis recommended by the ATS/ERS consensus classification of the IIPs were identified (1). All 154 cases were originally diagnosed histologically as diagnostic of UIP by a lung pathologist at each of the contributing institutions. All biopsy specimens were also reviewed by a second lung pathologist with 32 years of experience and classified into the following four categories by the certainty of the diagnosis of UIP: (1) confident UIP, (2) probable UIP, (3) probably not UIP, and (4) confident not UIP. A confident diagnosis of UIP was made if all the ATS/ERS criteria were fulfilled: patchy involvement with clear evidence of chronic scarring/honeycombing and the presence of fibroblast foci in the absence of features against the diagnosis of UIP, such as granulomas and so forth. A confident diagnosis of "not UIP" was made if there were clear features of an alternative diagnosis, or if none of the ATS/ERS criteria for UIP were present. Diagnoses of "probable UIP" and "probably not UIP" were more subjective; most commonly the former represented cases of extensive honeycombing without good evidence of patchy involvement in the sample reviewed, and the latter were cases in which the histology was more suggestive of an alternative diagnosis. On the basis of this review, 112 cases (73%) were diagnosed as definite UIP, 19 cases (12%) as probable UIP, 16 cases (10%) as probably not UIP, and 7 cases (5%) as confidently not UIP. Only the 112 cases interpreted by the second lung pathologist as definitely being UIP were considered acceptable for the study. Of these, 14 had to be excluded because their CT scans were not available or they had the other diseases in the lungs. The study therefore included 98 patients who had a confident histologic diagnosis of UIP made by two independent lung pathologists and who had a clinical diagnosis of IPF. The study sample consisted of 71 men and 27 women, with a mean age of 63 years (range, 36–75 yr). Survival analysis, survival status, and follow-up periods were assessed by review of the medical records. The mean and median follow-up periods of all cases were 79 and 63 months, respectively. Forty-six patients died and 10 were lost to follow-up.
Thin-Section CT Images and Review The radiologist evaluated the presence, extent, and distribution of CT findings, which included the presence of ground-glass attenuation, airspace consolidation, nodules, interlobular septal thickening, thickening of bronchovascular bundles, intralobular reticular opacities, irregular interlobular septal thickening, nonseptal linear or platelike opacities, presence of honeycombing, cysts, emphysema, architectural distortion, traction bronchiectasis, and fibrosis score. The definitions of each CT finding are shown in Table 1.
The observers evaluated the extent of all radiologic abnormalities, excluding emphysema, that were present in both lungs to determine the percentage of lung parenchyma occupied by the disease. The lungs were divided into six zones (upper, middle, and lower on both sides); each zone was evaluated separately. The upper lung zone was defined as the area of the lung above the level of the tracheal carina, the lower lung zone was defined as the area of the lung below the level of the inferior pulmonary vein, and the middle lung zone was defined as the area of the lung between the upper and lower zones. When abnormal findings were present, the extent of lung involvement was evaluated visually and independently for each of the six lung zones. The score was based on the percentage of the lung parenchyma that showed evidence of the abnormality and was estimated to the nearest 5% of parenchymal involvement. The overall percentage of lung involvement was calculated by averaging the six lung zones. The extent of traction bronchiectasis was quantified by assessing the generations of the most proximal bronchial branches that were involved. Traction bronchiectasis was scored as follows: 0, none; 1, bronchial dilatation involving bronchi distal to the fifth generation; 2, bronchial dilatation involving fourth-generation bronchi; 3, bronchial dilatation involving bronchi proximal to the third-generation bronchi. These scores were assessed in each of the six lung zones and the overall score of traction bronchiectasis was obtained by adding the scores of the six lung zones. For each lung zone, architectural distortion was scored as 0 (absent) or 1 (present), and all of the scores were summed. To evaluate interstitial fibrosis, in each zone a fibrosis score was assigned: 0, none; 1, ground-glass attenuation without reticulation; 2, ground-glass and fine reticular opacity; 3, reticular opacity and microcysts less than 3 mm; or 4, coarse reticular opacity and large cysts more than 3 mm (6). These scores were assessed for each of the six lung zones and then summed. Furthermore, the extent of disease close to the hilum was scored as follows: 1, abnormal parenchyma distal to the fifth-generation bronchus; 2, abnormal parenchyma distal to the fourth-generation bronchus; 3, abnormal parenchyma proximal to the third-generation bronchus. The observers also provided an overall evaluation of their impression of the CT findings as to whether they were homogeneous or heterogeneous. After assessing the presence and extent of the findings, the observers evaluated their predominant distribution. The distribution was classified as being predominantly upper, lower, peripheral, dependent, or peribronchovascular. Asymmetrical predominance was considered to be present if the extent of findings or the degree of fibrosis was predominant on one side. The overall extent of various abnormal findings was obtained by averaging the evaluation by the two independent observers. Disagreements with respect to architectural distortion, traction bronchiectasis, fibrosis score, extent of disease toward the hilum, impression of diseases, and predominant distribution were resolved by consensus. After review of the findings, the CT scans in each case were classified by consensus between the four observers into those that were radiologically consistent with UIP or those that were suggestive of an alternative diagnosis based on the results of previous studies that reported the thin-section CT appearance of UIP (1–4). The detailed classification was as follows: (1) definite UIP, (2) consistent with UIP, or (3) suggestive of alternative diagnosis. The CT scan was classified as showing a definite UIP pattern when it demonstrated honeycombing in a predominantly peripheral and basal distribution. The CT was classified as consistent with UIP when it demonstrated a reticular pattern in a predominantly peripheral and basal distribution but only minimal or no honeycombing. The CT was classified as suggestive of alternative diagnosis when alternatives to UIP, such as NSIP, were more appropriate. The relationship between survival duration and these three CT categories as subtypes was evaluated.
Statistical Analysis Comparison between definite UIP and the other two CT categories of abnormality was made using univariate analysis. The extent of the individual CT patterns, traction bronchiectasis, disease distribution toward the hilum, and fibrosis score were assessed using the Mann-Whitney U test. The presence of architectural distortion, predominant distribution, and the overall impression was analyzed using Fisher exact test. Univariate and multivariate Cox proportional hazards regression models were used to identify independent CT predictors of outcome. On multivariate analysis, the variables were selected using a stepwise procedure including each one of the following CT findings: presence of ground-glass attenuation; airspace consolidation; nodules; interlobular septal thickening; thickening of bronchovascular bundles; intralobular reticular opacities; irregular interlobular septal thickening; nonseptal linear or platelike opacities; presence of honeycombing, cysts, emphysema, architectural distortion, or traction bronchiectasis; fibrosis score; the extent of disease close to the hilum; and upper, lower, peripheral, dependent, peribronchovascular, and asymmetric predominant distribution. Findings were retained if they contributed to the power of the regression equation (P < 0.10). Patient survival between the three CT categories (1, definite UIP; 2, consistent with UIP; and 3, suggestive of alternative diagnosis) was determined using the log-rank test and displayed using Kaplan-Meier curves. A P value of less than 0.05 was considered to indicate statistical significance.
Observer Agreement and Diagnoses Interobserver agreement (Table 2) was poor to good with respect to the extent of the various abnormalities (Spearman rank correlation coefficient, r = 0.13–0.79; P < 0.001–0.19), with poor to moderate agreements for the presence of architectural distortion and the predominant distribution (  = 0–0.56). Interobserver agreement was poor for the presence of nodules (r = 0.19), thickening of bronchovascular bundles (r = 0.15), nonseptal linear opacity (r = 0.13), and dependent distribution ( = 0). The interobserver agreement of CT diagnosis into consistent with UIP (definite or probable) or suggestive of alternate diagnosis (suggestive of NSIP or indeterminate) was moderate ( = 0.60).
The observers classified 33 cases (34%) as having definite UIP (Figure 1), 36 as consistent with UIP (Figure 2), 21 (21%) as having CT findings suggestive of NSIP (Figure 3), and 8 cases (8%) as having unclassified findings (Figure 4).
Predominant Findings The total CT findings of all cases are shown in Table 2. Ground-glass attenuation (23.4%) and intralobular irregular reticular opacities (11.8%) were the most common findings, whereas thickening of bronchovascular bundles (0.4%), intralobular regular reticular opacities (0.9%), and nonseptal linear opacities (0.8%) were the least common findings. Lower zone predominance (n = 77) and peripheral predominance (n = 76) were common. Seventy-four patients had symmetric bilateral distribution of findings and 24 had asymmetric distribution. The CT findings of each CT category are shown in Table 3. As compared with patients with CT findings classified as definite UIP, patients with findings consistent with UIP were more likely to have airspace consolidation (5.2%), interlobular septa (3.4%), regular intralobular reticular opacity (0.4%), and emphysema (3.0%), and less likely to have honeycombing (1.8%) and cysts (2.9%) (P < 0.05). Patients with CT findings suggestive of alternative diagnosis were more likely to have ground-glass attenuation (28.4%), airspace consolidation (6.2%), interlobular septal thickening (5.9%), regular reticular opacity (2.6%), and peribronchovascular distribution (n = 4), and less likely to have honeycombing (1.7%) (P < 0.05).
The results of Cox regression analysis for the relationship between the CT findings and prognosis are shown in Tables 2 and 4. On univariate analysis, all abnormalities, including airspace consolidation, honeycombing, architectural distortion, traction bronchiectasis, fibrosis score, and heterogeneous overall impression, were significant predictors (hazard ratios: 1.05, 1.10, 1.05, 1.72, 1.31, 1.12, and 4.30, respectively). On multivariate analysis, only traction bronchiectasis and fibrosis score were significant predictors of mortality (hazard ratios: 1.30 and 1.10, respectively).
CT Pattern of Abnormality and Mortality The Kaplan-Meier survival curves and their relation to the pattern of abnormality on CT are shown in Figure 4. The mean survivals of patients with CT findings interpreted as definite UIP, consistent with UIP, suggestive of alternative diagnosis categories were 45.7, 57.9, and 76.9 months, respectively, and the median survivals were 34.8, 43.4, and 112 months, respectively. The prognosis of definite UIP was not significantly different from that of the other two categories (log-rank test: P = 0.26 and 0.09, respectively).
All 98 patients in this study had a pathological diagnosis of definite UIP. The most common findings in the cases classified as having definite UIP radiologically were honeycombing, irregular intralobular reticular opacities, and ground-glass opacities. This result agrees with previous reports of CT findings of UIP (2, 4, 9, 10). However, in the present study, an asymmetric distribution was seen more often than expected, in 24 of 98 patients. An asymmetric distribution may reflect a temporal heterogeneity between the right and left lungs and may be a characteristic finding of UIP. However, further assessment, including cases with NSIP, is necessary to further clarify this issue. Cases in the probable UIP category radiologically had irregular reticular opacities and ground-glass opacities in the lower and peripheral lung. These findings are compatible with UIP but the lack of peripheral and basal honeycombing on CT precluded a confident diagnosis. The findings on CT may be indicative of early UIP. However, there was no difference in the mortality of definite UIP and consistent with UIP cases. Some cases with these findings had areas of airspace consolidation, which pathologically sometimes reflect airless fibrotic tissue or honeycombing filled with mucinous secretions (4, 11); and dependent atelectasis would cover the other findings. For such reasons, CT may underestimate the extent of honeycombing and thus the severity of fibrosis. As previously shown, there is overlap between the high-resolution CT (HRCT) findings of UIP and those of NSIP (6, 12–14). Therefore, not surprisingly, several of our patients had CT findings interpreted as suggestive of alternative diagnosis, and 21 of 29 cases with an alternative diagnosis had CT findings similar to NSIP. It must be pointed out that a confident diagnosis of NSIP requires surgical biopsy. It should also be noted that regions showing pathological NSIP are described in patients who have biopsy-confirmed UIP in surgical lung biopsies (15, 16). Cases with histologic regions showing both UIP and NSIP have a prognosis similar to patients with UIP and are therefore considered to have "histologic discordant UIP." Although all of the obtained biopsy specimens in the current study had characteristic histologic features of UIP, it is possible that other areas of the remaining lungs had findings of NSIP. Some cases having findings suggestive of alternative diagnosis had various CT findings, but in each case the findings on CT did not allow for classification into any of the categories of IIPs based on the ATS/ERS consensus (1). Hartman and colleagues reported that the appearance of NSIP on CT was variable and included the findings of UIP and those of other chronic infiltrative lung diseases (14). They suggested that the presence of a histologic spectrum of NSIP or sampling error of the biopsy specimens could explain the variability of CT findings (14). The findings of UIP would vary on CT, likely due to the variability in the histology of NSIP. In the current study, 29 of 98 patients (30%) with UIP had HRCT findings that were interpreted as more consistent with an alternate diagnosis. Thus, almost one-third of cases with histologic UIP had CT features suggestive of other diseases, especially NSIP. This result is compatible with previous reports (5, 6, 8, 9). In histologically confirmed UIP cases, there were some cases in which the CT findings were identical to those of NSIP. The current study only included cases that had surgical biopsy–proven diagnosis. In clinical practice, patients with characteristic CT findings of IPF/UIP in accordance with ATS/ERS consensus statement usually do not require biopsy (1). Therefore, our study is biased toward patients with atypical CT findings of IPF. This selection bias presumably accounts for the relatively high extent of ground-glass opacity (mean extent, 23.4%) and the presence of consolidation (mean extent, 4.4%) in our patients with UIP. The relatively high extent of ground-glass opacity also accounted for the high mean overall extent of abnormality (36%). The fact that the extent of ground-glass opacity was greater than reticulation (11.8%) is quite atypical for UIP, and reflects a skewed population, because patients with characteristic features of UIP seldom undergo surgical biopsy. The prognosis of IPF in our study was poor, and there was no significant difference in prognosis between cases with typical CT findings of UIP and those with atypical findings. Previous studies have shown that the prognosis of IPF/UIP is poor and that the prognosis of NSIP is better than that of IPF/UIP (15, 17–19). In histologically diagnosed UIP cases, Flaherty and colleagues reported that the prognosis of cases that were diagnosed as UIP based on both CT and histologic findings was worse than in cases with a histologic diagnosis of UIP but without CT features of UIP (20). However, in their previous report, the difference in prognosis was small, and the prognosis of cases with a histologic diagnosis of NSIP was much better (20). Our study included only the cases with definite diagnosis of UIP in histology, and the shorter survival time in the cases with atypical CT pattern could be overestimated. The relatively small number of cases may also have been a factor. There may also have been some preselection bias. Cases were selected solely on the basis of definite UIP histologically confirmed by at least two observers, which might have led to a smaller number of cases than would have resulted if some of the histologically "probable" UIP cases were included. However, the causes for the discrepancy between our study and Flaherty and colleagues' are not clear. The results of our study suggest that the prognosis of cases with histologic diagnosis of UIP is poor even if they have atypical CT findings. Various CT findings were found to influence outcome in our study. On univariate analysis, total amount of all abnormalities, airspace consolidation, honeycombing, architectural distortion, traction bronchiectasis, fibrosis score, and the overall impression of the CT findings influenced the outcome significantly, whereas, on multivariate analysis, only traction bronchiectasis and fibrosis score influenced outcome significantly. These findings are mainly associated with the degree and extent of lung fibrosis, and the results of the present study are similar to those of a previous report (21). Correlations between observers for traction bronchiectasis and fibrosis score were good (0.75 and 0.79, respectively). In the present study, the CT category was determined not only by specific CT findings but also by their distribution. In particular, the determination of concordant UIP was strongly based on the distribution of the findings. However, the distribution of the findings did not affect the mortality on both univariate and multivariate analyses. Therefore, the CT category is not helpful in predicting the outcome. Our study has several limitations. First, the study was retrospective. The treatments given patients were not identical and this may have influenced the evaluation of prognosis. Second, the study included only patients who had surgical biopsy, and thus had a selection bias with a higher proportion of patients with atypical CT findings of IPF/UIP than would be seen in daily clinical practice. In addition, the observers knew that all patients had clinical and histological UIP. This fact would influence the interpretation and diagnosis in HRCT. Thirty percent of all cases were diagnosed as not having UIP; however, the number of cases would increase in blind reading. Another limitation of this study is that it was limited to the correlation of the initial CT findings with survival data. It did not include correlations of survival or CT data with functional parameters. The study also included relatively few patients. Although there was no significant difference in survival between patients with CT findings consistent with UIP and those with findings more suggestive of NSIP, the mean survival of patients with CT findings interpreted as definite UIP was 45.7 months, compared with 57.9 months for patients with CT scans interpreted as being probable UIP, and 76.9 months for patients with CT scans interpreted as suggestive of alternative diagnosis. These results suggest a trend toward greater survival of patients with CT findings more suggestive of alternative diagnosis. Furthermore, there was a strong trend (P = 0.07) that the extent of disease in patients with CT findings more suggestive of alternative diagnosis was more extensive than that in those with CT findings typical of UIP. The lack of statistical difference between the various groups may be due to the relatively small number of patients in the study. Therefore, the study cannot be used to refute the findings by Flaherty and colleagues that patients with IPF who have atypical CT findings have a better prognosis than patients with characteristic CT findings of UIP. In addition, only one pathologist decided on the final pathological diagnosis in this study. All 154 cases were originally diagnosed histologically as diagnostic of UIP by a lung pathologist at each of the participating institutions, then all biopsy specimens were reviewed by a second lung pathologist. Therefore, more than two pathologists diagnosed the cases as UIP; however, they did not diagnose independently, and the interobserver variation could not be evaluated. The other limitation is that the histologic diagnoses based on surgical biopsies may have had a sampling error. The histologic diagnosis of multiple specimens from one patient may sometimes differ (15). Finally, some CT variables were collated with each other, such as honeycombing and fibrosis score. Therefore, honeycombing might not be a predictor of poor outcome in multivariate analysis, although honeycombing was a predictor in univariate analysis. In conclusion, cases with pathologically proven IPF/UIP had greater variability in their CT findings than expected; however, there was no significant difference in mortality between the various CT categories. Traction bronchiectasis and fibrosis score were predictors of poor outcome.
Originally Published in Press as DOI: 10.1164/rccm.200611-1696OC on November 1, 2007 Conflict of Interest Statement: None of the authors has a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Received in original form November 23, 2006; accepted in final form October 29, 2007
Related articles in AJRCCM:
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
AT A GLANCE COMMENTARY
