CT in Idiopathic Pulmonary Fibrosis: Diagnosis and Beyond

Transcription

1 Cardiopulmonary Imaging Review Gruden CT of Idiopathic Pulmonary Fibrosis Cardiopulmonary Imaging Review James F. Gruden 1 Gruden JF FOCUS ON: Keywords: CT, diagnosis, high-resolution CT, idiopathic pulmonary fibrosis, interstitial lung disease, usual interstitial pneumonitis DOI: /JR Received October 4, 2015; accepted without revision October 6, Department of Radiology, Cardiothoracic and ody Imaging, New York Presbyterian-Weill Cornell Medical Center, 525 E 68 St, Starr 8-37, New York, NY ddress correspondence to J. F. Gruden (jfg9007@med.cornell.edu). JR 2016; 206: X/16/ merican Roentgen Ray Society CT in Idiopathic Pulmonary Fibrosis: Diagnosis and eyond OJECTIVE. This review focuses on recent idiopathic pulmonary fibrosis guidelines with regard to CT-based diagnosis, the limitations and potential sources of diagnostic error in their clinical application, and proposals for future guideline modification and improvement. The review also addresses the use of CT in disease monitoring, prognostic assessment, and therapeutic clinical trials. CONCLUSION. Ongoing research and clinical experience highlight the need for updating existing guidelines frequently. T he merican Thoracic Society, European Respiratory Society, Japanese Respiratory Society, and Latin merican Thoracic ssociation recently issued joint evidence-based guidelines for the diagnosis and management of idiopathic pulmonary fibrosis (IPF) [1]. The Society of Thoracic Radiology and the Pulmonary Pathology Society also endorsed these guidelines. The document defines IPF as a chronic, progressive fibrosing interstitial pneumonia of unknown cause, occurring primarily in older adults, and limited to the lungs [1]. The clinical course is variable, but most patients suffer from increasing dyspnea and loss of pulmonary function, and the prognosis is poor [1, 2]. IPF is associated with the histopathologic pattern of usual interstitial pneumonitis (UIP) and, as such, represents a distinct clinical syndrome (effectively, idiopathic UIP ) as clarified in a prior consensus document [2]. diagnosis of IPF specifically requires the exclusion of other entities that may be associated with a UIP pattern, including occupational lung disease, connective tissue disorders, exposure to potential organic antigens (chronic hypersensitivity pneumonitis), or exposure to drugs known to cause pulmonary toxicity. Radiologists and pathologists identify findings that indicate UIP, but neither makes the diagnosis of IPF. UIP, like the other idiopathic interstitial lung disorders, is not a specific diagnosis but a radiographic and pathologic pattern; all patterns require a clinical context to define a more-specific diagnosis [3]. IPF is the clinical diagnosis in patients with UIP for whom other potential causes of this pattern have been excluded. The accuracy of this diagnosis improves when there is effective collaboration and communication among pulmonary medicine physicians, radiologists, and pathologists [1, 4, 5]. This review focuses on the most recent IPF guidelines, specifically with regard to the CT-based diagnosis of UIP (and subsequently of IPF), the limitations and potential sources of diagnostic error in their clinical application, and proposals for future guideline modification and improvement. The review will also address the increasing use of CT in persons with IPF beyond their initial diagnosis, particularly with regard to disease monitoring, prognostic assessment, and in therapeutic clinical trials [6]. CT and the Diagnosis of Usual Interstitial Pneumonitis and Idiopathic Pulmonary Fibrosis The new 2011 merican Thoracic Society, European Respiratory Society, Japanese Respiratory Society, and Latin merican Thoracic ssociation statement [1] represents significant advances in comparison with the 2000 document [2]. Importantly, the 2011 guidelines are evidence based rather than consensus based. This reliance on the literature and on scientific inquiry, rather than the collective opinion of a few experts, resulted in substantial modifications to the 2000 statement, many of which are relevant to practicing radiologists. JR:206, March

2 Gruden TLE 1: High-Resolution CT Criteria for Usual Interstitial Pneumonitis (UIP) Pattern. UIP (all four features) 1. Subpleural, basal predominant 2. Reticular abnormality 3. Honeycombing with or without traction bronchiectasis 4. bsence of findings inconsistent with UIP listed in (C). Possible UIP (all three features) 1. The findings in () with the exception of (3). C. Inconsistent with UIP (any of the seven features) 1. Upper or midlung predominance 2. Peribronchovascular predominance 3. Extensive ground-glass abnormality (greater than the extent of reticulation) 4. Profuse micronodules (bilateral, predominantly upper lobes) 5. Discrete cysts (multiple, bilateral, away from areas of honeycombing) 6. Diffuse mosaic attenuation/air-trapping (bilateral, in three or more lobes) 7. Consolidation in bronchopulmonary segment(s)/lobe(s) Note dapted with permission of the merican Thoracic Society. Copyright 2011 merican Thoracic Society. Raghu G, Collard HR, Egan JJ, et al. n official TS/ERS/JRS/LT statement: idiopathic pulmonary fibrosis: evidence-based guidelines for diagnosis and management. m J Respir Crit Care Med 2011; 183: [1]. The merican Journal of Respiratory and Critical Care Medicine is an official journal of the merican Thoracic Society. Most important, CT findings alone can now be used to diagnose UIP without the need for tissue confirmation. The new guidelines list three levels of CT-based certainty for the UIP pattern: UIP, possible UIP, and inconsistent with UIP (Table 1). If the CT findings are in either of the latter two groups, a surgical biopsy is recommended for accurate diagnosis. The guidelines also list four levels of certainty for pathologic diagnosis (UIP, probable UIP, possible UIP, and not UIP), with specific descriptions of the characteristics of each of these categories. The diagnosis of IPF thus requires exclusion of the other known causes of the UIP pattern, the presence of a UIP pattern on CT in the absence of biopsy, and specific combinations of CT and biopsy patterns in patients who do undergo biopsy [1]. These latter cases, in which there is some discrepancy between CT and pathologic findings, generally require multidisciplinary review. The central role of CT in UIP diagnosis is new and is an advance from the prior consensus statement, in which CT findings were simply one of many criteria (both major and minor) that were required for confident UIP diagnosis [2]. These major and minor criteria have been eliminated. The new reliance on CT reflects the results of many investigations of the CT appearance of UIP with pathologic and appropriate clinical correlation [7 11], but also extensive clinical experience with CT in the evaluation of patients with interstitial lung disease over the past decade. Surgical lung biopsy can now be avoided in patients with the evidence-based CT findings of UIP, which is an important advance. Limitations, Sources of Error, and Potential Modification CT Technique The guidelines, unfortunately, do not specifically address CT technique, but this topic is of great importance in the evaluation of all patients with diffuse lung disease, particularly in an attempt to render specific and accurate diagnoses. Current MDCT enables contiguous volumetric imaging of the lungs during a single breath-hold, with collimation between and 2.5 mm, and is the standard of care for all chest CT examinations. These are effectively contiguous high-resolution images of both lungs. The technique of high-resolution CT (HRCT) refers to thin sections (< 3 mm but usually mm) of Fig. 1 Four patients with honeycomb cysts in lungs., 63-year-old man with idiopathic pulmonary fibrosis (IPF). CT image obtained through lung bases with patient in prone position shows asymmetric honeycombing involving subpleural right lung, with thin-walled cystic lesions that share walls. There is no intervening parenchyma between cysts. Patient had long history of gastroesophageal reflux disease, and asymmetric IPF is common in this population (see text)., 54-year-old man with sarcoidosis. CT image of left lung obtained with patient in supine position shows that honeycomb cysts are visible in lingula and medially at left base adjacent to aorta. Traction bronchiectasis is also present posterolaterally in subpleural lung. oth findings indicate presence of fibrotic lung disease, and neither is specific to usual interstitial pneumonitis (UIP) pattern. (Fig. 1 continues on next page) 496 JR:206, March 2016

3 CT of Idiopathic Pulmonary Fibrosis C D Fig. 1 (continued) Four patients with honeycomb cysts in lungs. C, 47-year-old woman with scleroderma and fibrotic nonspecific interstitial pneumonitis. CT image obtained through lower lobes with patient in supine position shows bilateral cystic lesions posteriorly. Some share walls, whereas others are in single layer. Whether this is honeycombing or severe traction bronchiectasis is not clinically relevant (see text). Note right heart dilatation related to her known pulmonary hypertension. D, 68-year-old man with IPF. CT image obtained through lung bases with patient in supine position shows irregular peripheral reticular opacities and areas of traction bronchiolectasis. Reliance on honeycombing for diagnosis of radiologic UIP would result in interobserver disagreement in such cases. Other CT findings are more important to diagnosis (see text). the lung parenchyma; these were historically obtained in the pre-mdct era at intervals of 10 or 20 mm, such that most of the lungs were not visualized. This is no longer appropriate, given the additional information that is available with whole-lung scanning and additional reformatted images in nonaxial planes. This review uses the term CT in its current context, with contiguous high-resolution images depicting the lungs in their entirety during a volumetric scan. lthough the comparison between traditional HRCT, with gaps in imaging, and contiguous MDCT with regard to effects on diagnostic confidence in the diagnosis of UIP or other lung disorders has not been made, it is clear that contiguous images offer more information and should facilitate more accurate and complete assessment of pertinent imaging features. Unlike surgical lung biopsy, which depicts only tiny fragments of the lung, CT is able to depict both lungs in their entirety. Imaging with the patient in the prone position, at least through the lower lung zones, should also be routinely performed to allow detection of subtle abnormalities in the lung bases that may be obscured by or confused with normal dependent atelectasis [12]. Fig year-old woman with usual interstitial pneumonitis (UIP)., CT image obtained through upper lobes with patient in supine position shows peripheral irregular reticulation in upper lung zones crossing fissures in nonsegmental distribution. There is distortion of normal pulmonary lobules, indicating presence of fibrosis., CT image obtained through lung bases with patient in prone position shows peripheral reticular opacities with numerous foci of traction bronchiolectasis. ppearance of outer parenchyma is heterogeneous, with areas of normal lung alternating with areas of fine reticulation and areas of more advanced fibrosis. CT diagnosis is UIP despite absence of honeycombing (see text). The Problems With Reliance on Honeycombing CT diagnosis of UIP requires the presence of honeycombing (Table 1), and this is problematic for a number of reasons. There is confusion among even experienced thoracic radiologists as to the precise meaning of this term. The Fleischner Society definition is, clustered cystic airspaces, typically of comparable diameters on the order of 3 10 mm but occasionally as large as 2.5 cm. Honeycombing is usually subpleural and is characterized by well-defined walls [13]. This should allow distinction between honeycombing and peripheral traction bronchiectasis and bronchiolectasis; the latter represents dilated small airways that are separated from one another and from the pleural surface by intervening lung parenchyma, and there are no shared walls [14]. However, the distinction between confluent traction bronchiolectasis and honeycombing can be difficult, and this represents a significant source JR:206, March

4 Gruden Fig year-old man with usual interstitial pneumonitis (UIP) and idiopathic pulmonary fibrosis (IPF)., CT image obtained through upper lungs with patient in supine position shows subtle but definite peripheral reticulation. Some upper lobe reticulation is essential to diagnosis of UIP and is present in modified criteria (Table 2)., CT image obtained through costophrenic angles with patient in prone position shows peripheral reticulation with mild traction bronchiolectasis. There is no honeycombing. ppearance is heterogeneous with normal peripheral lung alternating with zones of minimal reticulation and other areas of more extensive fibrosis and bronchiolectasis. CT diagnosis of UIP led to clinical diagnosis of IPF. of interobserver disagreement [15] (Fig. 1). Interobserver agreement as to the presence of honeycombing is often less than optimal in studies of patients with UIP [16, 17]. This is unfortunate because honeycombing and traction bronchiolectasis actually have the same clinical significance: they indicate the presence of fibrosis. oth occur with peripheral and lower zone predominance in patients with UIP and both are often present in the same patient and on the same image (Fig. 1). The presence of honeycombing and the extent and severity of traction bronchiolectasis each also correlates with a poor prognosis in patients with IPF, and the severity of traction bronchiolectasis actually portends a poor prognosis irrespective of the degree of abnormality in the remaining lung [18]. Some investigators have emphasized the use of multiplanar reformatted images to distinguish between traction bronchiolectasis and honeycombing, emphasizing the importance of honeycombing to diagnosis and differential diagnosis [19]. lthough this may indeed be possible, it seems a futile exercise. The distinction between traction bronchiolectasis and honeycombing has no Fig year-old man for whom surgical lung biopsy showed usual interstitial pneumonitis (UIP). CT image through midlung zones shows multifocal ill-defined areas of consolidation with minimal architectural distortion. ppearance and distribution are not consistent with CT diagnosis of UIP. Such cases require multidisciplinary evaluation. Pathologic diagnosis is not always straightforward in patients with diffuse lung disease (see text). clinical relevance. There are other key CT findings in UIP that make reliance on honeycombing unnecessary (Fig. 1). The honeycombing requirement also ignores the presence of other typical and consistent CT features of UIP and places persons with traction bronchiolectasis but no honeycombing in the same management group (possible UIP) as those who lack either of these findings. This seems suboptimal, and other reviewers have also discussed this issue [20, 21]. Recent data indicate that a new category, patients with probable UIP, who have all of the specific CT findings other than honeycombing, probably also do not need a surgical biopsy for diagnosis [22]. In its current usage, honeycombing is a manifestation of end-stage parenchymal fibrosis of any cause and is not specific for UIP [19]. The cysts are subpleural in a posterior and lower lung distribution in most cases of UIP, which is important to the differential diagnosis. In addition, because honeycombing represents end-stage cystic destruction of the lung parenchyma, it implies advanced disease. The presence of honeycombing on CT has been repeatedly shown to be a poor prognostic sign [16, 18, 23 25]. n analysis of subgroups of patients treated with pirfenidone, one of the agents recently approved in Europe and subsequently the United States for treatment of IPF, showed that patients who had milder disease were most likely to benefit from therapy [26]. The contradiction between the need to diagnose UIP at an early stage and the requirement of end-stage findings for CT diagnosis was discussed in a recent review [27]. 498 JR:206, March 2016

5 CT of Idiopathic Pulmonary Fibrosis TLE 2: Proposed Guideline Revisions for Usual Interstitial Pneumonitis (UIP) Pattern. UIP (all four features) a 1. Subpleural, basal predominance, but with some upper lobe involvement 2. Reticular abnormality with lobular distortion in a nonsegmental distribution 3. Traction bronchiectasis or honeycombing 4. Heterogeneous appearance in the subpleural lung both on single images and on the examination as a whole. Probable UIP 1. The findings in () with the exception of (3) C. Possible UIP (either of the two features) 1. The findings in () with the exception of 1, 2, or 4 2. Significant associated emphysema such that distinction of smoking-related lung disease and UIP is difficult D. Unlikely UIP (either of the two features) 1. Lacking any two of (1 4) in () 2. typical features are the predominant or only finding Note Groups D could be coded with modifiers for a staging system based on extent and severity of the imaging findings (e.g., stage 1, 2, 3, and so forth). This may allow more standardization of patient groups to assess prognosis, outcome, and response to therapy. Coexistent emphysema should be reported, characterized, and quantified (see text). Care should be taken to avoid false-positive CT diagnoses; in particular, the confounding effects of emphysema should be recognized. ppropriate CT technique, including contiguous MDCT axial images of 2.5-mm collimation or less and prone imaging through the lung bases, is essential. a typical features should be evaluated in the appropriate clinical context. In summary, the guidelines currently require honeycombing for a specific CT diagnosis of UIP. Radiologists do not consistently define it and have trouble agreeing as to its presence even when using the same definition. Efforts made to distinguish it from traction bronchiolectasis result in some patients being placed in the possible UIP category and subjected to surgical biopsy or potentially precluded from receiving novel therapies or entering clinical trials. It occurs in fibrotic lung diseases other than UIP, and it indicates advanced disease with a poor prognosis. It seems that other CT findings that allow confident UIP diagnosis in the absence of honeycombing need to be identified and validated. Proposed Modifications to the CT Criteria for Usual Interstitial Pneumonitis Diagnosis recent study identified findings other than honeycombing that occurred in a biopsy-proven series of patients with UIP [28], and these findings were then validated in a prospective cohort with long-term clinical follow-up [29]. These findings, with slight modifications, are shown in Table 2. They add to or modify the existing guidelines in several ways. s stated in the original guidelines, the CT abnormalities in UIP have a peripheral and lower zone predominance, but there is a slight modification to this description: some upper lobe involvement is required for diagnosis [29]. Hunninghake and colleagues [10] found that the two CT findings most predictive of UIP were basal-predominant honeycombing and the presence of some reticulation in the upper lobes (present in 85% of the patients with UIP and 31% of those with other interstitial lung diseases). The existing requirement for reticular abnormality is also modified in Table 2 to include both distortion of the secondary lobules and a nonsegmental distribution. Lobular distortion indicates the presence of fibrosis; the lobules are shrunken, often triangular instead of polyhedral, and not of uniform size or shape. Thus, a process that infiltrates the interlobular septa without causing lobular distortion, such as pulmonary edema or other nonfibrotic process, cannot be confused with UIP. This lobular distortion is what the Hunninghake group referred to as irregular lines [10]. The distribution of CT abnormalities in UIP is nonsegmental and crosses the fissures without respect for the bronchopulmonary anatomy of the lung. Segmental peripheral fibrosis, with perhaps bilateral lower lobe and some posterior upper lobe involvement, can occur in other disorders with fibrosis, such as chronic aspiration pneumonitis. The third criterion simply reverses the word order of the phrase of the existing guidelines regarding honeycombing and traction Fig year-old man with usual interstitial pneumonitis (UIP) and idiopathic pulmonary fibrosis (IPF)., CT image obtained through lower lungs with patient in prone position shows peripheral reticulation but there is no definite traction bronchiolectasis or honeycombing. There was minimal upper lobe reticulation as well (not shown). ppearance of peripheral lung parenchyma is slightly heterogeneous, but findings are not sufficient to diagnose UIP with confidence (Table 2). Further study of patients with these CT findings is needed., Four years later, CT image obtained through lung bases with patient in prone position now more clearly shows peripheral reticulation with distortion and traction bronchiolectasis. ppearance of abnormalities in outer lung is heterogeneous, and upper lobe reticulation is more advanced (not shown). CT diagnosis was UIP, and clinical diagnosis was IPF. JR:206, March

6 Gruden Fig. 6 Two patients with scleroderma and nonspecific interstitial pneumonitis (NSIP)., 52-year-old woman with scleroderma and chronic progressive dyspnea. CT image obtained through lower lungs with patient in supine position shows ground-glass attenuation and reticulation superimposed on traction bronchiolectasis and bronchiectasis. Outer right lower lung is spared, and process is fairly homogeneous in appearance. CT findings are consistent with NSIP and not consistent with usual interstitial pneumonitis (UIP) pattern., 57-year-old woman with scleroderma and NSIP. CT image obtained through lung bases with patient in supine position shows extensive areas of homogeneous groundglass attenuation with distinct subpleural sparing in both bases. There is minimal traction bronchiolectasis at right base laterally. lthough not specific for NSIP, imaging features are not suggestive of UIP pattern. There is large hiatal hernia. bronchiolectasis. ecause these findings are almost always associated with one another, have the same significance (they indicate the presence of fibrosis), and are both independently correlated with a poor prognosis, they are grouped together. Rather than requiring honeycombing with or without bronchiolectasis, the modified guidelines simply require one or the other (Table 2). The last of the four modified criteria, all of which must be present for CT diagnosis of UIP, is the presence of a heterogeneous appearance to the CT abnormalities, both on the examination as a whole and on individual images. When analyzing the outer 1 2 cm of lung parenchyma, which is the predominant site of disease in UIP, areas of normal lung should alternate with areas of minimal reticulation, whereas other areas depict traction bronchiolectasis or honeycombing or both. Heterogeneity is a cardinal feature in the pathologic profile of UIP, indicating a temporally nonuniform process of recurrent injury and repair [1]. However, at the macroscopic level, few groups have described this as a CT feature of UIP [11, 28]. This is an important CT clue as to the underlying diagnosis (Fig. 2). The final proposed modification is to allow atypical imaging findings to be analyzed according to the clinical context. If the atypical features are the predominant or the only CT finding, then the classification could be unlikely UIP (Table 2). However, patients with characteristic findings of UIP may have some of the atypical CT features (e.g., minimal ground-glass attenuation away from fibrotic zones or a segmental area of consolidation) that should not negate a UIP diagnosis in all instances. Other investigators have presented the concept of a CT group called probable UIP, in addition to possible UIP, to include patients who meet the guidelines criteria for UIP diagnosis and who have traction bronchiolectasis but no honeycombing [22]. These patients can receive a certain UIP diagnosis in the modified guidelines in Table 2 if all of the other findings are present (Fig. 3). Patients with probable UIP in the modified guidelines have all the findings required for UIP diagnosis but there is no traction bronchiolectasis or honeycombing (Table 2). These patients still have a heterogeneous pattern of peripheral nonsegmental irregular reticulation with lower zone predominance, some upper lobe involvement, and a heterogeneous appearance in the outer 1 2 cm of lung parenchyma from apex to base. This indicates a lesssevere form of fibrosis and was an additional pattern noted in a previous biopsy-proven series [28]. Further study of the specificity of this pattern is needed before including these patients in the certain UIP category. The modified criteria eliminate the requirement for honeycombing, dispose of the need to distinguish between honeycombing and traction bronchiolectasis, and add CT features important to the accurate and specific diagnosis of UIP such that false-positive findings are avoided and the distinction between UIP and other diffuse lung diseases is more obvious. They allow at least some patients currently classified as having possible UIP to receive a definitive diagnosis (i.e., those with traction bronchiolectasis), and the proposed probable UIP designation may identify individuals with UIP before the advanced findings of fibrosis (traction bronchiolectasis and honeycombing) occur. The need for tissue sampling in patients with probable UIP would likely be clinically determined, and further assessment of the specificity and utility of this category would be helpful. typical CT ppearances in Usual Interstitial Pneumonitis and reas of Controversy Histologic UIP may be associated with atypical HRCT appearances, and in such cases in which surgical biopsy shows UIP, multidisciplinary discussion should be performed in an effort to ascertain a final diagnosis [1, 4, 5]. It should be emphasized that the pathologist receives only a small sample of lung parenchyma, and that even experienced lung pathologists may disagree as to the diagnosis of UIP in a significant number of cases [9, 12, 25] (Fig. 4). The disagreement between general community and pulmonary pathologists in the interpretation of biopsy specimens in patients with interstitial lung disease approaches 30 50% [30, 31]. In addition, multiple biopsies from the same patient may show different histologic patterns in roughly one-third of cases; this is particularly true of the distinction between UIP and nonspecific interstitial pneumonitis (NSIP) [32]. There are several current areas of interest and controversy concerning CT and the diagnosis and differential diagnosis of UIP (Fig. 5). 500 JR:206, March 2016

7 CT of Idiopathic Pulmonary Fibrosis Usual Interstitial Pneumonitis and Nonspecific Interstitial Pneumonitis Katzenstein and colleagues [33] described NSIP in 1994 as a distinct form of interstitial lung disease characterized by temporal homogeneity, in contrast to UIP, which shows a variegated temporally heterogeneous appearance. There are cellular and fibrotic forms of NSIP, with minimal to moderate degrees of fibrosis. NSIP is most often associated with underlying connective tissue disease but can also occur in an idiopathic form [3]. Several authors reported the CT findings in NSIP in contrast to those of UIP but this was after the year 2000 [9, 11, 25, 31, 34, 35]. Therefore, it is likely that investigations of the imaging and clinical findings in patients with UIP before this time included some individuals with NSIP. Most of the imaging descriptions of NSIP compared it to UIP and emphasized the more frequent occurrence of honeycombing in UIP with more ground-glass attenuation (i.e., increased density that does not obscure the underlying pulmonary vascular markings) in NSIP [9, 11, 25, 34, 35]. helpful characteristic in NSIP is subpleural sparing, with relatively normal lung present in the area immediately subjacent to the pleural surface in areas of otherwise extensive ground-glass attenuation and fibrosis [35]. This is best identified on images obtained with the patient in the prone position, but the finding is not present in all cases of NSIP. Only one study pointed out the patchy and heterogeneous appearance of UIP as a distinguishing feature different from that of NSIP [11]. The temporal homogeneity of NSIP noted pathologically is also visible on CT; the ground-glass attenuation with variable traction bronchiectasis or reticulation lacks the variegated heterogeneous appearance of UIP (Fig. 6). Traction bronchiolectasis and lobular distortion, indicating fibrosis, are often present in NSIP (although honeycombing is not common), but the appearance is homogeneous. Upper lobe involvement in NSIP is variable. The modified criteria for UIP diagnosis with CT should be better able to distinguish these entities because of the inclusion of peripheral heterogeneity and the need for some upper lobe involvement for a UIP diagnosis (Table 2). NSIP has a distinctly uniform and homogeneous appearance. Patients with NSIP also tend to be significantly younger than those with UIP; in fact, Fell and colleagues [36] showed that, in patients older than 65 years with peripheral reticulation on CT, the diagnostic likelihood of UIP or IPF exceeded 95% even in the absence of honeycombing. Connective tissue serologic analysis also often helps to differentiate the two entities. Unfortunately, there is a well-known subset of individuals in whom the CT features suggest NSIP but the surgical biopsy findings show UIP [9, 23, 37]. The ultimate diagnosis in such cases remains unclear, but some investigators suggest that patients with UIP and a CT appearance more suggestive of NSIP have a better prognosis than do patients with concordant pathologic and CT findings [23, 37]. Further study of individuals with such discordant radiologic-pathologic findings is needed (Fig. 7). It is important to recognize that the modified criteria are designed to avoid false-positive diagnoses; patients with NSIP would be unlikely to meet all criteria for a CT diagnosis of UIP, but it is not clear whether such individuals would ever be included in the probable UIP or possible UIP categories. Clinical evaluation would be important in such cases, and further study of the role of a CT pattern and classification system independent of histologic findings is needed. Usual Interstitial Pneumonitis and Chronic Hypersensitivity Pneumonitis Chronic hypersensitivity pneumonitis (HP) often has a distinctive histologic appearance characterized by bronchiolocentric interstitial pneumonia, noncaseating granulomas, and cellular bronchiolitis, but it can result in a histologic UIP pattern, and, for this reason, a careful exposure history is essential before the diagnosis of IPF [1, 38, 39]. CT of chronic HP usually has distinct features, including areas of heterogeneous lung attenuation with lobular areas of decreased attenuation indicating air trapping (more evident on expiratory scanning), mid-to-upper lung zone predominance, centrilobular nodules, and some component of central peribronchovascular reticulation and distortion. Honeycombing can occur but is not usually predominant in the lower zones and costophrenic angles [35] (Fig. 8). The incidence of peripheral heterogeneity has not been reported, to my knowledge. Fig year-old man with progressive dyspnea., CT image obtained through lower lungs with patient in supine position shows ground-glass attenuation and reticulation with homogeneous appearance. There is mild traction bronchiolectasis in both lungs. ppearance is not consistent with usual interstitial pneumonitis (UIP) pattern but more suggestive of nonspecific interstitial pneumonitis (NSIP) (see text)., Coronal reformatted CT image shows homogeneous mild ground-glass attenuation and reticulation in lower lungs with minimal traction bronchiolectasis. Note lack of upper lung involvement. CT is consistent with NSIP pattern, but surgical biopsy showed UIP. Ultimate clinical course, diagnosis, and management of these patients remains controversial (see text). JR:206, March

8 Gruden Fig year-old woman with chronic progressive dyspnea and cough., CT image through upper lungs shows heterogeneous lung attenuation with peripheral reticulation. Parenchymal abnormalities are diffuse and not confined to peripheral parenchyma. lthough clinical presentation suggested idiopathic pulmonary fibrosis, CT findings are more suggestive of chronic hypersensitivity pneumonitis (HP) (see text)., CT image through lung bases shows more pronounced heterogeneous lung attenuation with numerous hyperinflated secondary lobules (particularly on right) alternating with areas of reticulation and architectural distortion. CT findings suggest chronic HP. Patient owned several birds for many years. Chronic HP can clinically mimic IPF, and, when fibrosis is present, the disease is irreversible and often progressive. In one study, patients with fibrosis on either surgical biopsy or CT had a median survival duration of 4.9 years, whereas the median survival duration in those without fibrosis was 16.9 years; patients in whom no inciting antigen could be identified also had a markedly worse prognosis [39]. This area causes significant confusion currently and needs to be addressed and clarified: if chronic HP can result in a UIP histologic profile and no inciting antigen is found, how can one distinguish IPF from chronic HP with no known cause? In addition, if patients with chronic HP and fibrosis have a prognosis similar to that of UIP, is the distinction clinically and therapeutically relevant? CT examination that meets all the criteria for UIP in Table 2 and lacks any of the CT features to suggest chronic HP in a patient with no clear exposure history would seem to represent IPF, but this is an area of debate. Usual Interstitial Pneumonitis and Emphysema and Smoking-Related Lung Diseases Cigarette smoking is a risk factor for the development of IPF and for emphysema, and a significant number of patients with IPF are current or former smokers [1, 40]. Paraseptal emphysema, which usually but not always occurs in association with centrilobular emphysema, is thin-walled, cystic, and subpleural and can be similar in size to honeycomb cysts [14]. It can occur in the upper lung zones and paramediastinal regions, but can also be present in the lower zones and occur in clusters. Smoking-related interstitial fibrosis is often present adjacent to areas of paraseptal emphysema in the lower lung zones; this is a CT pattern of homogeneous peripheral basal ground-glass attenuation without honeycombing or traction bronchiolectasis that corresponds to a histologically homogeneous and localized fibrosis composed of hyalinized eosinophilic ropy collagen that thickens alveolar septa confined to the subpleural parenchyma [41]. However, according to the Fleischner definition [13], paraseptal emphysema can be confused with honeycombing, and the combination of smoking-related basilar fibrosis with paraseptal emphysema could be falsely interpreted as UIP (Fig. 9). This is yet another difficulty with the guideline s reliance on honeycombing for a UIP diagnosis: false-positive findings can occur particularly in the setting of emphysema and smoking-related pulmonary fibrosis. The presence of emphysema (centrilobular, paraseptal, or both) has a significant effect on reader accuracy (41%) in the diagnosis of fibrosing interstitial lung disease, largely attributable to the difficulty in differentiating honeycombing from paraseptal emphysema [42]. To complicate matters further, there is now an entity called combined pulmonary fibrosis and emphysema. The initial cohort was heterogeneous and included patients with centrilobular and paraseptal emphysema in the upper lobes associated with reticulation, with or without honeycombing in the lower lobes [43]. Not all patients had UIP-type fibrosis, however. Current usage of the term combined pulmonary fibrosis and emphysema should be confined to patients with IPF in particular who have significant concomitant emphysema. Patients with both entities, particularly those with a predominance of paraseptal emphysema, are difficult to evaluate with CT using the existing guidelines. The modified guidelines have not been evaluated in this setting. However, difficult cases are placed in the modified possible UIP category. gain, false-positive definite diagnoses should be avoided. In a recent study of 47 patients with combined pulmonary fibrosis and emphysema, all but one (n = 46) were male, and most had paraseptal emphysema; honeycombing was present in 76% of the patients, and nearly half had lung cancer [44]. Most former smokers with IPF do not have emphysema, and, conversely, there is no fibrosis in most patients with emphysema. The combination of the two may reflect unique individuals exposure to cigarette smoke but requires further study [45]. Certainly, accurate CT classification and identification of patients with combined pulmonary fibrosis and emphysema as a defined cohort is essential. Ryerson and colleagues [46] used a cutoff amount of 10% emphysema (i.e., percentage of 502 JR:206, March 2016

9 CT of Idiopathic Pulmonary Fibrosis lung volume affected) in patients with IPF and found this to distinguish patients with physiologically significant emphysema. Patients with combined pulmonary fibrosis and emphysema had less physiologic restriction (as measured by forced vital capacity [FVC]) and worse gas exchange (lower diffusing capacity for carbon monoxide) than did patients with IPF alone, and individuals with mild emphysema (< 10% lung volume) had clinical and physiologic findings similar to those with no emphysema [46]. Interestingly, 30% of all patients with UIP and IPF in that series had at least some emphysema, illustrating the common coexistence of these two conditions [46]. In an autopsy series of patients with combined pulmonary fibrosis and emphysema, emphysema, and IPF, Inomata and colleagues [47] described thick-walled cystic lesions more than 10 mm in diameter, occurring in the lower lobes adjacent to areas of peripheral honeycombing and reticulation, in nearly three-quarters of patients with combined pulmonary fibrosis and emphysema but in none of the patients with emphysema alone or IPF (Fig. 10). Enlargement of these cystic spaces over time was a sign of progression of the pulmonary fibrosis [47]. Future validation and analysis of these findings is needed. From a clinical standpoint, it is important to account for the presence and extent of emphysema in choosing the optimal longitudinal physiologic parameter to predict outcome; combined pulmonary fibrosis and emphysema was reported in 25% of patients with IPF in a recent study, and different pulmonary function parameters predicted outcomes in patients with combined pulmonary fibrosis and emphysema compared with patients with IPF alone [48]. Wells and colleagues [49] reported a composite physiologic index to follow patients with IPF that was an improved outcome measure when compared with pulmonary function tests alone; as such, this was an improvement over a previous system that did not account for the presence of emphysema [50]. However, in the study of Schmidt and colleagues [48], the composite physiologic index was a better outcome index only in patients without emphysema, whereas a serial change in forced expiratory volume at 1 second was more predictive of mortality in patients with combined pulmonary fibrosis and emphysema. It is, therefore, clinically important to be aware of combined pulmonary fibrosis and emphysema as an entity and to not confuse paraseptal emphysema with smoking-related fibrosis as IPF. When a patient with IPF has significant emphysema (> 10% of lung volume), this should be reported along with the type (paraseptal or centrilobular) and approximate extent (quantity). This may help our clinical colleagues more accurately interpret functional data as they monitor their patients. The spiration Issue study of 32 patients with asymmetric IPF (which can occur) showed that 63% had gastroesophageal reflux disease (GERD), 11 had a hiatal hernia, and 15 of 16 with a side-sleeping preference slept on the more fibrotic side; patients with asymmetric disease also had twice the incidence of reflux as those with symmetric IPF [51]. More-acute exacerbations of IPF also occurred in those with asymmetric disease and were often unilateral, affecting the more-fibrotic lung [51]. These findings suggest that, in some patients, particularly those with asymmetric disease, reflux of gastric acid may contribute to both the underlying fibrosis and to acute exacerbations of the disease. recent retrospective analysis of patients with IPF enrolled in several clinical trials showed that there were no confirmed acute exacerbations in patients receiving antiacid therapy [52]. Further study of the role of GERD in the development of pulmonary fibrosis and the presence or absence of differences in imaging features and clinical outcomes in patients with GERD-induced fibrosis versus symmetric and presumed idiopathic disease would be of interest. CT and Disease Monitoring and Staging Monitoring Patients With Idiopathic Pulmonary Fibrosis and Quantification of Disease IPF progresses in an unpredictable manner. Some patients remain stable for extended periods, some exhibit a progressive decline in respiratory status at variable intervals, and others suffer acute exacerbations that have a high mortality rate [53]. The best current method to monitor progression is to identify a decline in FVC or diffusing capacity for carbon monoxide; pulmonary function tests are generally monitored at 6- to 12-month intervals [1, 54]. Functionally stable patients show a decline in FVC of a maximum of 5% of the baseline value over the time period [55 57]. decline in FVC of 5 10 percentage points indicates prognostically relevant progression, whereas a decline of 10 percentage points or more in 6 months is associated with a fourfold to eightfold increased mortality within 1 year [55 57]. The extent of oxygen desaturation on a 6-minute walk test may also provide prognostic information [58]. However, there are limitations to pulmonary function testing, including a degree of effort dependence and the effect of other cardiopulmonary pathologic abnormalities, particularly emphysema, both of which are present in a significant number of patients with IPF. This led to the development of other more complex methods for monitoring patients with IPF, although neither achieved widespread clinical application [48, 49]. Currently, there is no single or combination of widely accepted objective measures to assess Fig year-old man., CT image through upper lung zones shows peripheral thin-walled cystic spaces that vary in size, consistent with paraseptal emphysema. Note that there is no centrilobular emphysema. There is also no peripheral reticulation to suggest usual interstitial pneumonitis (UIP)., CT image through midlungs shows minimal homogeneous ground-glass attenuation in lower lungs posteriorly with small subpleural cystic foci related to paraseptal emphysema. Imaging features are most consistent with minimal smoking-related fibrosis (see text). Patient was asymptomatic. CT criteria for UIP diagnosis are not present (Table 2). False-positive diagnoses are to be avoided. JR:206, March

10 Gruden clinical course and response to therapy in patients with IPF. ecause of the heterogeneous disease course, it is difficult to identify patients at increased risk of short-term mortality who may need more urgent consideration for lung transplantation. Similarly, there is no definite way to accurately gauge individual response to therapy, particularly in the short term. ecause the only available medical therapies primarily reduce the rate of pulmonary function decline, rather than reverse existing deficits, slow declines in pulmonary function are difficult to analyze in a specific patient. C Fig year-old man., CT image through lower lungs is normal. There was paraseptal and centrilobular emphysema in upper lungs (not shown)., CT image obtained through lower lungs with patient in supine position shows very fine peripheral reticulation particularly on right. Tiny focus of traction bronchiolectasis is present at right base. Overall CT findings are nonspecific. C, Three years later, CT image obtained through lower lungs with patient in supine position shows new peripheral reticulation in both lower lungs. D, CT image through lower lung zones shows progression of peripheral reticulation and traction bronchiolectasis. There is also thin-walled cystic lesion at left base (see text). Patient had developed progressive dyspnea and imaging features suggest combined pulmonary fibrosis and emphysema. Smoking-related fibrosis and its relationship to usual interstitial pneumonitis is area of ongoing research. CT has advantages as a monitoring tool and morphologic severity indicator. CT findings correlate more strongly with functional impairment than do histologic findings [49]. The study is performed in a single breath-hold, and even very ill patients can undergo the examination. The relative contributions of fibrosis, emphysema, and superimposed other pathologic abnormalities (e.g., pleural effusion, heart failure, or pneumonia) can be identified and quantified, either subjectively or objectively. There are many studies addressing the correlation between fibrosis diagnosed on CT and clinical outcomes. Shin and colleagues [12] identified a high fibrotic score at CT (defined as the extent of reticulation plus honeycombing) as an indicator of poor survival in patients with UIP and fibrotic NSIP. That study was a two-reader subjective quantification of reticulation and honeycombing estimated to the nearest 5% in each lung on studies performed with 1-mm collimation at 10-mm intervals [12]. Lynch and colleagues [16], also using gapped nonvolumetric CT and two readers, found that a higher overall extent of fibrosis (sum total of reticulation and hon- D 504 JR:206, March 2016

11 CT of Idiopathic Pulmonary Fibrosis eycombing) on CT was a strong independent predictor of mortality in patients with IPF. Lee and colleagues [17], in a study of patients with minimal or no honeycombing and thus with milder disease, found that patients with a subjective amount of pulmonary fibrosis of up to 35% of the whole lung volume had significantly better survival rates than did those with an overall pulmonary fibrosis extent of more than 35% of lung volume. That study also used an image protocol with noncontiguous scanning. Other investigators have reported similar findings, although none used contiguous current MDCT protocols, which is a significant limitation [18, 25, 59]. The most interesting study regarding the use of CT as a monitoring tool is that of Oda et al. [60], who studied 98 patients who received a diagnosis of IPF between 2008 and Two observers quantified CT findings specifically described in the consensus guidelines: reticulation, traction bronchiolectasis, honeycombing, and normal lung. lthough the investigators also evaluated noncontiguous limited CT images, an elevated CT fibrosis score predicted a poor prognosis at 6-month followup; even in patients with no significant FVC decline, an elevated CT fibrosis score independently heralded a worse prognosis. The authors concluded that CT could be used to monitor patients with IPF and may identify a subset of patients with a poor prognosis (and hence more urgent need for transplant consideration) despite a stable FVC. Despite the abundant data on CT analysis of extent of fibrosis and outcome correlation, the studies are subjective and use outdated techniques with large gaps in imaging. Most analyze patients with extensive disease and are retrospective, with long-term followup and mortality as an endpoint. None has a large number of CT readers. What is needed is more thorough analysis that can detect and potentially quantify subtle short-term changes in patients receiving therapy and allow comparison of one therapy to another within relatively short periods. The overall extent of fibrosis including reticulation, traction bronchiolectasis, honeycombing, and ground-glass attenuation in the fibrotic zones is gaining acceptance as an appropriate measure of monitoring patients with interstitial lung disease [61]. Radiologists must make this possible. Early studies of computerized objective quantitative measurements in CT again suffered from the technical limitations of outdated CT protocols with gaps in imaging [62, 63]. The best predictor of mortality in these studies remained the radiologist s subjective estimation of the visual extent of fibrosis on CT; the more quantitative measures and more sophisticated histogram computer analysis did not add additional information [63]. recent study showed the feasibility of more advanced texture-based computer analysis, but also applied the program to datasets with large gaps between individual slices, which may explain some of the limitations in the investigation [64]. recent study finally applied a textureanalysis approach to volumetric MDCT in an effort to classify and quantify various findings in the lungs of patients with UIP and IPF [65]. The authors showed objective results comparable to those of consensus expert radiologic analysis in this setting and the potential to provide objective measures of disease progression. However, the computer still cannot determine whether an area of groundglass attenuation is due to dependent atelectasis or fibrosis and may struggle with distinguishing between other adjacent or confluent abnormalities, and the subjective input of an experienced radiologist is still necessary. However, this technique offers real potential for rapid quantitative analysis and may assist the radiologist in the identification of areas of true subtle changes over time, particularly when the needed analysis involves hundreds of individual images. This is an area of ongoing and important investigation. CT Staging of Disease Quantification of disease extent and staging of disease may or may not be identical. Goh and colleagues [66] proposed a staging system for interstitial lung disease associated with scleroderma; the system was based on an initial subjective assessment of overall disease extent (any abnormality) at five parenchymal levels on nonvolumetric CT. n overall extent of disease threshold of 20% separated the cohort into two groups with 10-year survival rates of 67% and 43%, respectively. n FVC threshold of 70% corresponded to the optimal HRCT threshold of 20%, as judged by linear regression, and users therefore simply had to initially group patients into less than 10% or greater than 30% overall parenchymal involvement with use of the FVC threshold of greater or less than 70% in the remaining cases to group patients for prognostic purposes. lthough this may be useful because it is straightforward and allows stratification for mortality purposes, it is too crude to assess for subtle change in treatment trials, and it does not take into account the initial pattern of disease. s we have seen, data in patients with IPF show differences in outcome in the setting of traction bronchiectasis or honeycombing, but also show conflicting reports on outcomes in patients with histologic UIP and atypical CT findings. It is also not clear how associated emphysema affects response to therapy or outcome. staging system based on initial pattern with severity and emphysema moderators may be useful. For example, using the modified CT criteria in Table 2, subjective (or objective) modifiers could be added to each category to better describe the relevant findings. Group, diagnostic of UIP, could have modifiers quantifying the overall extent of disease in 10% increments or in the outer 10 mm of lung parenchyma and have subdivisions (1, 2, and 3) and so on. n emphysema modifier could be added according to the overall amount and type of emphysema (E1, E2, and E3). Similar modifiers could be applied to the other groups. This would allow patients with similar imaging patterns and extent of disease to be grouped and analyzed together and would allow more meaningful comparisons. It would also allow clearer identification of the types of patients in new treatment trials. CT and Clinical Trials Recent treatment trials, including those that led to the approval of pirfenidone and nintedanib for the treatment of IPF, used CT primarily as a diagnostic tool for study entry in patients with CT findings, as specified in Table 1 [67, 68]. Unfortunately, serial changes in FVC were the primary endpoint for efficacy. Disease-specific mortality differences would have required many more patients and a much longer study period to show significant differences between treatment and placebo arms and would be prohibitively expensive. However, the presence of emphysema and other cardiopulmonary disease can significantly alter pulmonary function tests, and little information is available with regard to the specific CT appearance in patients at baseline; however, at least two-thirds of the patients in each trial were current or former smokers [67, 68]. Future clinical trials will compare new therapies to the recently approved treatments and will require detection of small differences in response at relatively short intervals. The study of Oda et al. [60] comes to mind, in which CT showed differences in fibrosis score even in JR:206, March