Accuracy of High Resolution CT in Assessing ... - Science Direct

PATHOLOGY

Original Paper

RESEARCH AND PRACTICE © Urban & Fischer Verlag http://www.urbanfischer.de/journals/prp

Accuracy of High Resolution CT in Assessing Idiopathic Pulmonary Fibrosis Histology by Objective Morphometric Index* Ivany A. L. Schettino*, Alexandre M. Ab’Saber**, Robin Vollmer***, Paulo H. N. Saldiva**, Carlos R. R. Carvalho*, Ronaldo A. Kairalla*, Vera L. Capelozzi** *Division of Respiratory Diseases – Heart Institute (InCor), and **Department of Pathology University of São Paulo Medical School, Brazil; ***Laboratory Medicine, VA Medical Center, and Department of Pathology, Duke University Medical Center, Durhan, NC, USA

Summary To determine the accuracy of HRCT in assessing histology by objective morphometric index, twenty-five biopsy specimen-proved UIP were correlated with high-resolution CT (HRCT) by morphometric analysis. The scans were evaluated for the presence and extent of normal parenchyma, ground-glass attenuation, linear opacities, consolidation, honeycombing, vessels and bronchiectasis, and overall extent of histology involvement for normal parenchyma, honeycombing, alveolar septal inflammation, fibrosis, vessels, and bronchiectasis/bronchiolectasis. The comparison between morphometric measurements showed a strong correlation between HRCT and histologic parameters for extension (%) of normal tissue (p = 5 × 10–5), honeycombing (p = 6 × 10–5), and vessels (p = 0.0047). HRCT consolidation strongly correlated with alveolar septal inflammation (p = 0.015), whereas HRCT linear opacities had the highest correlation with histology for bronchiectasis or bronchiolectasis (p = 0.03). These associations also demonstrated that there was considerable residual scatter about the linear relationships found. By contrast, neither the ground glass patterns nor the bronchioectatic patterns determined by CT were associated with any histologic observation (p < 0.1). There was a borderline negative relationship between vessels deter*Financial support by the following Brazilian agencies: FAPESP, CNPq, CAPES and Lim05-HCFMUSP. Presented as poster discussion at the 1997 American Thoracic Society International Conference, May 16–21, San Francisco, California Pathol. Res. Pract. 198: 347–354 (2002)

mined by CT and histologic fibrosis (p = 0.069), i.e., the percentage of vessel patterns determined by CT was found to be lower when fibrosis was prominent histologically. Our results showed that HRCT patterns, usually employed to provide information about activity (ground glass) and fibrosis (consolidation) in IPF, failed to correlate with histology. On the other hand, chronic cystic lesions had a good correlation with histology. This finding suggests that in patients without a diffuse honeycomb pattern on HRCT, a lung biopsy may provide additional information. The more important limitation of our study was the lack of correlation related to the proximity of the biopsy site to the HRCT location evaluated by morphometry. Key words: Computed tomography – High-resolution CT – Usual interstitial pneumonia – Idiopathic pulmonary fibrosis – Morphometry – Lung biopsy

Introduction High-resolution computed tomography (HRCT) allows for a detailed examination of the pulmonary parenchyma and is one of the most promising non-invasive methods to asAddress for correspondence: Vera Luiza Capelozzi, Department of Pathology, University of São Paulo Medical School, Av. Dr. Arnaldo 455, São Paulo, Zip Code 01246-903, Brazil. Fax (+55) 5096-0761. E-mail: [email protected] 0344-0338/02/198/5-347 $15.00/0

348 · I. A.L. Schettino et al.

sess staging and prognosis of idiopathic pulmonary fibrosis (IPF) [3, 21–23]. IPF is characterized on HRCT by a pattern of reticular opacities, often associated with traction bronchiectasis, ground-glass attenuation, and honeycombing [1, 12, 19, 20]. Nevertheless, the meaning of each separate lesion and particularly its correlation with histology in usual interstitial pneumonia (UIP) remains unclear. Previous works related reticular abnormalities to separate patterns, such as fibrosis or traction bronchiectasis or bronchiolectasis, observed in open lung biopsy (OLB) specimens [12, 19]. Other authors [5, 11, 14, 18] have found that ground glass opacities seen on HRCT imply a better prognosis and response to treatment for patients with IPF, but the results are contradictory. Some studies have demonstrated that areas of ground-glass attenuation usually correspond to alveolitis on OLB [17], while other studies have found that areas of ground-glass attenuation usually represent fibrosis [9]. As for a fibrotic pattern on OLB, which is well detected on HRCT [3, 21–23], it possibly needs to be adequately quantified to be indisputably related to treatment response and prognosis. The same should be true of cystic areas on HRCT. The apparent discrepancy between these studies may be due to different and subjective methods of determining both HRCT and pathologic alterations. Also, one may argue that in some studies comparisons were made between the whole slice of lung represented in a tomogram when the biopsy specimen hardly is expected to depict more than the peripheral subpleural region of the lung. Previous studies have not used quantitation systems in both CT and the histologic specimen, nor have they performed their analyses using only the peripheral outline of the tomogram of the lung lobe from which the biopsy had been taken. To our knowledge, only three studies dealing with correlations between CT of the lungs and histology used a high-resolution technique with thin sections (e.g., 1 to 2 mm collimation [15, 20], and only one, an experimental study with rabbits, applied a quantitative method to analyze both histopathologic and tomographic alterations [9]. We have previously used quantitative methods in interstitial lung diseases for correlation purposes [2, 5, 7], and we believe that they could be successfully used to assess IPF/UIP. The aim of this study was to determine the accuracy of HRCT patterns (ground-glass attenuation, irregular linear opacities, consolidation and cystic lesions) in assessing IPF histologic patterns (fibrosis, septal alveolar inflammation, honeycombing and bronchio-bronchiolectasis) by objective morphometric index.

Materials and Methods Study population

Retrospectively, from May 1982 to March 1998, 112 patients were found to have pulmonary fibrosis by open lung biopsy. In 89 patients, idiopathic interstitial pneumonia (Usual Interstitial

Pneumonia - UIP, Descamative Interstitial Pneumonia/Respiratory Bronchiolitis-Interstitial Lung Disease – DIP/RB-ILD, Acute Interstitial Pneumonia – AIP, Non-Specific Interstitial Pneumonia – NSIP) was diagnosed using a comprehensive pathologic and clinical designation [23] by the Pathology and Pulmonary Departments of HCFMUSP, and after carefully reviewing the medical records to exclude patients with evidence of connective tissue disease and/or environmental/inhalational exposure known to cause interstitial lung disease (ILD), including a careful history for animal, dust, smoke and mold exposure either at home or in the work place. For the purpose of the present study, we also excluded DIP/RB-ILD, AIP and NSIP (that were diagnosed in 3, 6 and 13 cases, respectively). In only 25 of fthe remaining 67 patients with UIP pattern were the open lung biopsy slides available for review, as well as the clinical, functional data, and HRCT from the hospital case records. This population included 14 men and 11 women with a mean age of 59.7 + 14 years (range 29 to 74 years). In all subjects, specimens of open-lung biopsy, performed along the peripheral 2 cm of parenchyma in the same lobe (usually right lower lobe), were available for review. Smoking history was available in 18 of the 21 patients (93.9%). None of the patients was taking steroids or immunosuppressive drugs at the time the lung biopsy was undertaken. At the time of diagnosis, they were thoroughly evaluated clinically, including complete pulmonary function tests. The HRCT was accomplished 1 to 30 days before the lung biopsy. The diagnosis was established according to previously described clinical and histologic criteria [10, 13]. Histology morphometry

According to a standardized protocol established by our group, immediately following the biopsy, the specimens are inflated with 10 per cent neutral buffered formalin, using a syringe and a needle inserted directly into the specimen and injection through the pleura. Six sections (3 µm thick) from each paraffin block were stained with hematoxylin-eosin. The slides were examined by two of us (AMAbS and VLC) using light microscopy, and classified as UIP according to the recommended criteria [10]. In all sections of the specimen investigated, the histologic pattern of UIP was characterized by alterations in pulmonary acinus and represented by the alternating zones of the following morphologic components: 1. alterations in airspace configuration (alveolar collapse, ductal hyperinflation – Fig. 1, panel A, D and G); 2. alveolar septal inflammation (Fig. 1, panel A); 3. reparative process (fibrosis and fibroblastic foci – Fig. 1, panel D); 4. acinus remodeling (honeycomb change and bronchiolectasis – Fig. 1, panels D and G) and 5. normal acinus. The extension and distribution of these morphologic components (structure) were not homogeneous along the sections. Thus, a morphometric analysis was done to determine the areal fraction occupied by each morphologic component of UIP in sections. The morphometric analysis was carried out using a conventional stereological point-counting method [8, 25] in the sections of biopsy specimens with the aid of an ocular (numerical aperture, 10×) containing a reticulated grid with 100 points and 50 lines. Counting was done using a cascade progressive sampling approach [8, 25], totaling 1,000 points counted for

CT and Biopsy Morphometry of Idiopathic Pulmonary Fibrosis · 349

each specimen evaluated in two different magnifications. At a magnification of 100×, five non-coincident microscopic fields per section were studied by counting a total of 500 points covering a total area of 1250 µm2/section. At a magnification of 400×, used for evaluating alveolar septal inflammation, the same was done covering an area of 312,5 µm2/section. The areal fraction occupied by the morphologic components in UIP through the section was determined according to the mathematical formulas [8, 25] (1) AA (struct/sect) = Σ points N (struct) × 100 / Σ N points (sect) = Total Area of Structure × 100 / Total Area of Section = Areal Fraction in % (2) Astruct = Nstruct × 100 / 500 = Areal Fraction in % where struct represents the UIP components and sect the total number of points counted • Assessment of areal fraction occupied by the normal acinus – Aacin - in section: The areal fraction of normal acinar compartment (Aacin) was assessed at 100fold microscopic magnification by counting the number points overlying normal alveolar tissue (ducts, sacs and alveoli) in five non-coincident fields in the section, according to the formula in (2): Aacin = Nacin × 100 / 500 = Areal Fraction of Normal Acinus in %. • Assessment of areal fraction occupied by the UIP components in section: The areal fraction of UIP components (fibrosis, septal inflammation, honecomb, bronchiolectasis and vessels) was assessed at 100fold microscopic magnification by counting the number points overlying each morphologic component in section, defined by the formula (2), where struct represents the components of UIP: Afibr = Nfibr × 100 / 500 = Areal Fraction of Fibrosis in %. HRCT scans morphometry

The CT scans (Phillips Tomoscan LX) were obtained using high-resolution technique with 1-mm section thickness at 10mm intervals, and reconstructed with a high-resolution algorithm and a 512 × 512 matrix. The images were photographed using window width of 1200 HU, window level of – 800 HU and 6 images per 14-× 17 inch film. In all cases investigated, the HRCT presentation of UIP was characterized by the following components (structure): 1. normal parenchyma; 2. ground-glass attenuation – hazy increase in lung opacity which does not cause obscuration of underlying structures; 3. consolidation – defined as areas with increased attenuation and loss of blood vessels and bronchi margins (Fig. 1 – panel B); 4. cystic lesions – defined as round structures with a very low density, with well-defined walls and air in the interior (Fig. 1 – panel D); 5. irregular linear opacities – this variable included interlobular septal thickening and intralobular lines, therefore including any linear opacity of 1–3 mm, visible at the lung parenchyma and distinct from bronchovascular bundles (Fig 1 – panel F); 6. vessels, and 7. bronchi. The extension and distribution of these alterations were not homogeneous along the pulmonary acini. Thus, morphometric analyses were done to determine the areal fraction occupied by each component (structure) in HRCT.

The morphometric analyses of the components on HRCT were performed using a conventional stereological pointcounting method with a standard grid of 100 cm2 containing 361 points 5 mm equidistant [8, 25], previously applied to measurements of HRCT by Kairalla (data not published). One image per patient was analyzed by two different observers (IASL and RAK). The observers of HRCT and histology were blinded to each other’s results. Although the positioning of the grid illustrated in Fig. 1 (panel F) covers the whole lung parenchyma, an effort was made to ascertain the proximity of the biopsy site to the HRCT location for morphometric evaluation, using the entire peripheral 2 cm of parenchyma outlined on the slice tomogram of the lobe from which the biopsy was taken. This procedure was choosen because of the patchy nature of the IPF/UIP disease and predominant involvement of the peripheral (subpleural) regions. As shown in Fig. 1, the main CT patterns in IPF/UIP, such as ground-glass attenuation, consolidation (panel B), cystic lesions (panel D), irregular linear opacities (panel F), as well as vessels and bronchi, are all present in the peripheral regions where the outline was performed. The areal fraction occupied by UIP components (struct) on HRCT (sect) was determined according the mathematical formula (2) used for histology: Astruct = Nstruct × 100 / sect = Areal Fraction in %, where struct represents UIP components and sect the number of points counted on the outlined HRCT. • Assessment of areal fraction occupied by UIP components on HRCT: The areal fraction occupied by UIP components on HRCT (normal parenchyma, ground-glass attenuation, consolidation, cystic lesions, irregular linear opacities, vessels and bronchi) was determined by formula (2): Aground-glass = Nground-glass × 100 / sect = Areal Fraction (%) Statistical analysis

The statistical analysis consisted, in part, of a general linear model analysis [16]. As CT findings are generally thought to reflect the underlying histologic pathology, the CT observations made were considered dependent variables, and the histologic observations were considered independent variables. Thus, several models were developed to test the relationship between each CT variable and one or more histologic variables. The software used was that of S-PLUS (version 2000, MathSoft, Inc., Seattle, WA, USA).

Results Table 1 summarizes the percentages for CT and histologic observations of this study. It shows that there was a wide range in abnormality, irrespective of whether CT or histology was used. In general, the percentage of normal tissue determined by CT was closely associated with that determined by histology. For example, linear regression demonstrated that these two percentages were closely associated with one another (p = 5 × 10–5), although the percentage of normal tissue determined by CT was greater


Fig. 1. (Panels A to I). Histologic and CT illustration of the main correlations obtained after comparison between morphometric parameters in UIP. Panels A, D and G show the UIP-pattern, characterized by temporal heterogeneous appearance at low magnification with alternating areas of alveolar septal inflammation (ASI between arrows), fibrosis (HFIB), bronchiolectasis (HBR) and honeycomb change (HHC), all of them quantified using a Weibel grid shown in panel A. (H&E stain, panel D and B 40×, panel G 100×). Panels C, F and I show different combinations of the CT parameters measured, such as linear opacities (LO) and consolidation (CO), often associated with traction bronchiectasis, honeycombing (CTHC) and ground-glass attenuation in peripheral, basal and patchy distribution. Normal areas of lung parenchyma were also evident. Table 1. Summary of results of individual variables Variable (percentage)

Type of diagnosis

Mean

Range

Normal Ground glass Linear opacity Consolidation Honeycomb Vessels Bronchiectasis1 Normal Fibrosis Septal inflammation Honeycomb Vessels Bronchiectasis1

CT CT CT CT CT CT CT Histology Histology Histology Histology Histology Histology

22 31 2 15 14 14 2 9 30 29 12 15 5

0–74 0–64 0–10 0–46 0–65 3–45 0–14 0–39 6–67 3–68 0–53 0–34 0–17

1

Bronchiectasis as used here includes bronchiolectasis

by an average of 6% compared with that determined by histology. Nevertheless, the left upper plots in Figure 2 show that there was considerable scatter about the linear relationship, which is indicated by the line on the plot. The main results appear in Table 2, which shows the statistical associations between CT and histologic observations. Four of the CT observations were significantly associated with one or more histologic variables. For example, linear opacity was associated with bronchiectasis; consolidation was associated with alveolar septal inflammation; the honeycomb pattern examined by CT was associated with that determined by histology; and vessels ascertained by CT were associated with those determined by histology. Three of these associations are also illustrated by the remaining three plots in Fig. 1, which also demonstrates that there was considerable residual scatter about the linear relation-


Fig. 2. Upper left: percentage of normal tissue investigated by histology versus CT (r = 0.72; p < 0.01, kappa 0.3). Upper right: percentage of alveolar septal inflammation examined by histology versus percentage of consolidation by CT (r = 0.67; p = 0.01, kappa 0.2). Lower left: percentage of honeycomb pattern investigated by histology versus CT (r = 0.71; p < 0.01, kappa 0.3). Lower right: percentage bronchiectasis/bronchiolectasis by histology versus percentage of linear opacity by CT (r = –0.42; p = 0.03, kappa 0). The line on each plot provides the regression line relationships found by general linear model analysis. Table 2. P-values for associations between CT observations and histologic observations1 Histologic Variables Fibrosis Septal Inflam HoneyComb Vessels Bronchioectasis

CT Variables G. Glass

Lin. Opac.

Consolid.

HoneyComb

Vessels

Bronch.

ns ns ns ns ns

ns ns ns ns 0.03

ns 0.015 ns ns ns

ns ns 6 × 10–5 ns ns

0.069 (–) ns ns 0.0047 ns

ns ns ns ns ns

1)

Entries in the Table are p-values obtained from a general linear analysis of how the CT observations as dependent variables are related to the histologic observations as independent variables. “ns” stands for not significant, i.e., the p-value exceeded 0.1. G. Glass = ground glass change; Lin. Opac. = linear opacity; Consolid. = consolidation; Bronch. = both bronchiectasis and bronchiolectasis; Inflam. = inflammation.

ships found. By contrast, neither the ground glass patterns nor the bronchioectatic patterns determined by CT were associated with any histologic observation (p > 0.1). The results suggest that after controlling for histo-

logic vessels, there was a borderline negative relationship between vessels determined by CT and histologic fibrosis (p = 0.069). i.e., the percentage of vessel pattern ascertained by CT was found to be lower when fibrosis


was prominent histologically. Otherwise, we found no multivariate relationships between CT observations and histologic observations.

Discussion Idiopathic pulmonary fibrosis (IPF) represents an idiopathic condition characterized by progressive interstitial fibrosis, leading to restrictive lung disease, shortness of breath, and hypoxemia [6, 14, 21–22]. Patients with IPF have histologic features of the UIP-pattern, represented by dense fibrosis, scattered fibroblastic foci, a patchy nature, and a peripheral acinar distribution. Histologically, one of the most characteristic features of UIP is the heterogeneous appearance at low magnification with alternating areas of normal lung, interstitial inflammation, fibrosis, and honeycomb change [10]. Pathological assessment is thought to be important to determine the diagnosis and to provide additional information on the activity and stage of the disease. However, in routine practice, pulmonary biopsy is not possible in every case; thus, the analysis of the extension and severity of pulmonary involvement must be determined by other methods. High-resolution CT provides useful information about the distribution patterns and anatomic variability of UIP. Although it is clear that HRCT does detect and clarify the pattern and extent of tissue and organ abnormalities in the thorax, the question of whether it can substitute or decrease the need of open lung biopsy is presently unsettled. The main point is that there is no general agreement about the correlation between histology and CT parameters. The apparent discrepancy among the various studies assessing the correlation between CT and histology may be due to subjectivity inherent to inter- and intraobserver variability. By choosing morphometric parameters, we intended to minimize subjective bias as often as possible. In this study, we have shown that by means of a rigorous morphometric method, it is possible to quantify the extent of CT changes in UIP and to verify their correlation with biopsy findings. The analyses of the results clearly show that the question of how CT correlates with histology needs to be clarified. This is manifested by the scatter in the plots of Figs. 1 and 2, and the results raise several issues for discussion. The first issue deals with the ground-glass attenuation increasingly recognized in HRCT images. Our results showed that ground glass pattern was not related to any histologic pattern, either suggesting that it was an unimportant CT observation in this study, or that its histologic counterpart was not observed. It is possibly related to alveolar collapse or alveolar desquamation. Although it has been shown to most commonly reflect the presence of inflammation [15, 17, 18], ground-glass pat-

tern on CT has not been satisfactorily correlated with pathology. In a recent study, the ground-glass attenuation corresponded to inflammation in 65% and to fibrosis in 54% of cases of diffuse lung disease [20] and did not correlate with alveolitis in any patient with UIP [27]. Also, it can be a non-specific finding that may represent fibrosis or septal alveolar inflammation. In fact, the absence of a specific histologic correlation can eventually be interpreted as evidence of potential reversible disease [15]. These statements may explain the results of studies that have shown a greater likelihood of response to treatment in patients with ground-glass opacities [6, 11, 26, 27]. In another retrospective study, Leung et al. [15], studying 11 patients with interstitial pulmonary disease, concluded that parenchyma opacification on thin-section CT scans represents a non-specific finding of diseases affecting the air spaces, interstitium or both, and they suggested that it usually indicates reversible disease. Again, the subjectivity present in the procedures to obtain correlations may be responsible for the disagreements. On the other hand, even morphometric techniques, which minimize bias of subjectivity, as applied by Hirose et al. [9] in rabbits with bleomycin/oxygen-induced lung inflammation, were not accurate, and therefore do not allow for a correlation between hazy increased density of ground glass on HRCT and any histologic parameter. The second point in this study is that bronchiectasis determined by CT does not appear to be related to any histologic observation, including histologic bronchiectasis. This can be a manifestation of failure of CT in interpretation. Possibly, this means that CT needs to develop additional criteria. Our results appear to provide a partial answer to the question, because linear opacity by CT appears to be closely related to histologic bronchiectasis (p = 0.03). In a recent study, the presence of traction bronchiectasis and bronchiolectasis was correlated with areas of fibrosis in 11 of 13 patients [20]. The findings that linear opacity is associated with bronchiectasis and consolidation with alveolar septal inflammation suggest that it is necessary to reevaluate the current concept that both are markers of CT fibrosis and correlate with histologic fibrosis. Müller et al. [17] first studied the correlation between histology and tomographic findings in patients with IPF. Although they reported a correlation between the CT and biopsy findings, they concluded that despite high-resolution, the CT images are not sensitive enough to accurately detect microscopic details of pulmonary parenchyma inflammation. Certain characteristic image features, such as the bibasilar subpleural fibrotic changes of IPF, indeed supplement other conventional histologic findings. In this context, Nishimura et al. [19], studying 46 patients with IPF, concluded that HRCT can identify well-perilobular distribution of fibrosis when they are associated with cystic lesions. Semiquantitative methods were employed by Wells


[26] and Kazerooni [12], who found a strong correlation between CT reticular pattern and histological fibrosis. In addition, the results may be biased by the subjectivity inherent to the method. The third point in our analysis is that none of the CT observations appear to be related very closely to the percentage of fibrosis found by histology. The established extent of histologic fibrosis correlated neither with CT consolidation nor with linear opacities. By contrast, CT consolidation was the spatial presentation of histologic inflammation, while the CT linear opacities were the marker for enhanced spatial perception of bronchiectasis or bronchiolectasis, and both of these histologic markers were related to parenchyma remodeling and irreversible alterations in UIP. Similar findings were reported in the experimental study of Hirose et al. [9]. They found an inverse correlation between linear opacities on tomographic scans and bronchiectasis or bronchiolectasis on tissue morphometry. However, this coefficient of correlation was low, and the parameter linear opacities were present in few patients. The results obtained suggest that after controlling for histologic vessels, there was a borderline negative relationship between vessels determined by CT and histologic fibrosis (p = 0.069), i.e., the percentage of vessel patterns determined by CT was found to be lower when histologic fibrosis was prominent. Finally, the fourth point to be discussed in this work is the fact that clinicians increasingly obtain HRCT as a routine diagnostic test shortly after being confronted with a patient with IPF, which may be justified on the strong correlations between honeycombing on CT and honeycombing on histology obtained by morphometric techniques employed in this work. In addition, as a consequence of the extension of HRCT cystic changes found in pulmonary parenchyma and its association with linear opacities, the differential diagnosis of IPF will get too remote from other rare cystic disorders, such as LAM, histiocytosis-× or tuberous sclerosis. Moreover, both findings may have implications in the diagnosis and establishment of the therapeutic protocol in patients without clinical conditions to be submitted to open lung biopsy. Our study has limitations that could be due to the lack of correlation between the HRCT and the biopsy findings. One critical issue that should be addressed is related to the proximity of the biopsy site to the HRCT location evaluated by morphometry. Besides the effort made to ascertain this, performing the analysis using only the peripheral outlined HRCT of the lung lobe from which the biopsy was taken, the patchy and heterogeneous distribution of the lesions in UIP may lead to misinterpretation. Aside from not being ideal, the procedure employed does not invalidate the study if we remember that in routine practice the pathologist also has limitations when analyzing the biopsy specimen. The second limitation is the retrospective nature of our study. Despite these limitations, our re-

sults showed that HRCT patterns, usually employed to provide information about activity (ground glass) and fibrosis (consolidation) IPF, failed to correlate with histology. On the other hand, chronic cystic lesions had a good correlation with histology. This finding suggests that for patients without a diffuse honeycombing pattern on HRCT, a lung biopsy may provide additional information. The more important limitation of our study was the lack of correlation related to the proximity of the biopsy site to the HRCT location that was evaluated by morphometry. Acknowledgements. We are grateful to Prof. Nestor Muller for his comments on this paper.

References 1. Akira M, Sakatatani M, and Ueda E (1993) Idiopathic pulmonary fibrosis: progression of honeycombing at thin-section CR. Radiology 189: 687–691 2. Barbas Filho JV, Ferreira M , Sesso A, Kairala RA, Carvalho CRR, Capelozzi VL (1998) Evidence of type II pneumocytes apoptosis in the pathogenesis of Idiopathic Pulmonary Fiprosis (IPF) / Usual Interstitial Fibrosis (UIP). Journal Of Clinical Pathology 54: 132–138 3. Carrington C, Gaensler E, Coutu R, Fitzgerald M, Gupta R (1978) Natural history and treated course of usual and desquamative interstitial pneumonia. N Engl J Med 298: 801–809 4. Crystal R, Fulmer J, Roberts W, Moss M, Line B, Rweynolds HY (1976) Idiopathic pulmonary fibrosis: Clinical, histologic, radiographic, physiologic, scintigraphic, cytologic and biochemical aspects. Med Ann Intern 85: 769–788 5. Deheinzelin D, Capelozzi V, Kairalla R, Barbas Filho JV, Saldiva PH, Carvalho CRR (1996) Interstitial Lung Disease in Primary Sjogren Syndrome. Clinical-Pathological Evaluation and Response to Treatment. Am J Respir Crit Care Med 154: 794–799 6. Gay S, Kazerooni E, Toewea G (1998) Idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 157 (4): 1063–1072 7. Hoelz C, Negri EM, Lichtenfels AJFC, Conceição GMS, Barbas CSV, Saldiva PHN, Capelozzi VL (2001) Morphometric Differences in Pulmonary Lesions in Primary and Secondary ARDS. Pathol Res Pract 197: 521–530 8. Gundersen HJG, Bendtesen TF, Korbo L, Marcussen N, Moller A, Nielsen K, Nyengaard JR, Pakkernberg B, Sorenses FB, Vesterby A, West MJ (1988) Some new, simple and efficient stereological methods and their use in pathological research and diagnosis. APMIS 96: 379–94 9. Hirose N, Lynch DA, Cherniack RM, Doherty DE (1993) Correlation between high-resolution computed tomography and tissue morphometry of the lung in bleomycin-induced pulmonary fibrosis in the rabbit. Am Rev Respir Dis. 147(3): 730–738 10. Idiopathic pulmonary fibrosis: diagnosis and treatment. International consensus statement (2000) Am J Respir Crit Care Med 161: 646–664 11. Jin Seong Lee M, Jung-GI IM M, Joong MO AHN M, Yang Min Kim M, Man Chung Han M (1992) Fibrosing alveolitis:

354 · I. A.L. Schettino et al. prognostic implication of ground-glass attenuation at highresolution CT. Computed Tomography 184: 451–454 12. Kazerooni EA, Martinez FJ, Flint A, Jamadar DA, Gross BH, Spizarny DL, Cascade PN, Whyte RI, Lynch JP, Toews (1997) Thin-section CT obtained at 10-mm increments versus limited three-level thin-section CT for idiopathic pulmonary fibrosis: correlation with pathologic scoring. AJR Am J Roentgenol 169: 977–983 13. King T Jr (1991) Diagnostic advances in idiopathic pulmonary fibrosis. Chest 100: 238–241 14. King TE Jr (1993) Idiopathic pulmonary fibrosis. In: Book MY (Ed) Interstitial Lung Disease. 2nd ed., pp 367–404. St. Louis 15. Leung A, Miller R, Müller N (1993) Parenchymal opacification in chronic infiltrative lung diseases: CT-pathologic correlation. Radiology 188: 209–214 16. Mccullagh P, Nelder JA (1989) General linear models. 2nd ed. London: Chapman & Hall 17. Müller N, Miller R, Webb W, Evans K, and Ostrow D (1986) Fibrosing alveolitis: CT-pathologic correlation. Radiology 160: 585–588. 18. Müller N , Staples C, Miller R, Vedal S , Thurlbeck W and Ostrow D (1987) Disease activity in idiopathic pulmonary fibrosis: CT and pathologic correlation. Radiology 165: 731–734. 19. Nishimura K, Kitaichi M, Izumi T, Nagai S, Kanaoka M and Itoh H (1992) Usual interstitial pneumonia: histologic correlation with high-resolution CT. Radiology 182: 337–42 20. Remy-Jardin M, Giraud F, Remy J, Copin M, Gosselin B and Duhamel A (1993) Importance of ground-glass attenu-

21. 22. 23. 24. 25.

26.

27.

ation in chronic diffuse infiltrative lung disease: pathologic-CT correlation. Radiology 189: 693–698 Scadding J, Hinson K (1967) Diffuse fibrosing alveolitis (diffuse interstitial fibrosis of the lungs): correlation of histology at biopsy with prognosis. Thorax 22: 291–304 Turner Warwick M, Burrows B and Johnson A (1980) Cryptogenic fibrosing alveolitis: Clinical features and their influence on survival. Thorax 35: 171–180 Turner-Warwick M, Burrows B, and Johnson A (1980) Cryptogenic fibrosing alveolitis: Response to corticosteroid treatment and effect on survival. Thorax 35: 593–599. Webb W, M_ller N, Naidich D (1993) Standardized terms for high-resolution lung CT: a proposed glossary. J Thorac Imag 8: 167–175 Weibel ER, Cruz-Orive LM (1977) Morphometric Methods. In: Crystal RG, West JB (Eds) The Lung. 2nd ed., pp 333–344. Scientific Foundation, Lippincott – Raven Publishers, Philadelphia Wells A, Hansell D, Corrin B (1992) High-resolution computed tomography assessment of disease activity in the fibrosing alveolitis of systemic sclerosis: a histopathological correlation. Thorax 47: 738–742. Wells A, Rubens M, Du Bois R, Hansell D (1993) Serial CT in fibrosing alveolitis: prognostic significance of the initial pattern. AJR 161: 1159–1165.

Received: April 11, 2001 Accepted in revised version: February 25, 2002