Context.—Familial idiopathic interstitial pneumonia (F-IIP) describes the unexplained occurrence of diffuse parenchymal lung disease in related individuals. Prevailing wisdom suggests that the histopathology of F-IIP is indistinguishable from that of idiopathic pulmonary fibrosis, namely, usual interstitial pneumonia (UIP).
Objective.—To define the histopathology of F-IIP in lung tissue samples.
Design.—Tissue sections from 30 patients with F-IIP, enrolled in a national research program, were evaluated by 3 pulmonary pathologists using 15 predefined histopathologic features. Each feature was recorded independently before a final diagnosis was chosen from a limited list dichotomized between UIP or “not UIP.” These 2 groups were then compared to survival.
Results.—The consensus diagnosis for the F-IIP cohort was an unclassifiable parenchymal fibrosis (60%), with a high incidence of histopathologic honeycombing, fibroblast foci, and smooth muscle in fibrosis. Usual interstitial pneumonia, strictly defined, was identified in less than half of the F-IIP cases (range, 23%–50%). Interobserver agreement was fair (κ = 0.37) for 2 observers for the overall diagnosis of UIP. Findings unexpected in UIP were prevalent. The survival for the entire F-IIP cohort was poor, with an estimated mortality of 93% and a median age at death of 60.9 years. Subjects with UIP had a shorter survival and younger age at death.
Conclusions.—Pulmonary fibrosis was the dominant histopathology identified in our patients, but diagnostic features of UIP were seen in less than 50% of the samples. Overall survival was poor, with mortality accelerated apparently by the presence of a UIP pattern of disease.
The diffuse parenchymal lung diseases encompass a broad spectrum of predominantly nonneoplastic inflammatory conditions that preferentially involve the lung interstitium (often referred to clinically as “interstitial lung diseases”). These conditions can be related to infection, a variety of inhaled or systemic exposures, systemic autoimmune disease, and even neoplasms.1–3 A subset of diffuse lung diseases has no identifiable etiology despite a comprehensive evaluation and is referred to as the “idiopathic interstitial pneumonias” (IIPs). By international multidisciplinary consensus, the IIPs include the clinical disorders of idiopathic pulmonary fibrosis (IPF), cryptogenic organizing pneumonia, respiratory bronchiolitis–associated interstitial lung disease/desquamative interstitial pneumonia, lymphoid interstitial pneumonia, and acute interstitial pneumonia.4
Kindred associations have been described for diffuse lung diseases of both known and unknown (idiopathic) etiology, strongly implicating a heritable genetic influence in affected individuals. The reported conditions include sarcoidosis,5 a variety of rheumatic diseases,6–9 lymphangioleiomyomatosis,10 the Hermansky-Pudlak syndrome11 neurofibromatosis,12 and familial hypercalcemic hypocalciuria.13
The most commonly described genetic predisposition in diffuse lung disease is for a familial form of IPF.14–17 The occurrence of IPF in large family kindreds as well as in twins18 adds further support for a yet-to-be identified genetic predisposition to chronic, particularly fibrosing, interstitial lung disease. Nevertheless, recent evidence suggests considerable pathologic diversity in F-IIP. In 2005, Steele et al19 reported on the largest study of familial interstitial pneumonia from the United States (111 families having 2 or more affected family members) in which 45% of the pedigrees demonstrated features of more than 1 histopathologic subtype of IIP. In this study, ever tobacco smoking was associated with risk of disease, suggesting that both genetic and environmental factors may influence the type of lung disease that supervenes in these families. When using this US cohort to conduct a genome-wide search for susceptibility alleles by linkage analysis, followed by association-based fine mapping, a single nucleotide polymorphism (SNP) in the promoter region of MUC5B was shown to increase the risk of developing pulmonary fibrosis among family members, by 6-fold if heterozygous for the SNP and by 22-fold if homozygous. Similar increased risk of disease was identified in sporadic cases of idiopathic pulmonary fibrosis with the MUC5B SNP.14
Despite studies emphasizing considerable heterogeneity of disease in F-IIP, many investigators still hold on to the concept that the histopathology of F-IIP is indistinguishable from that seen in idiopathic pulmonary fibrosis.14,17,20 No rigorous study has yet focused on the specific pathologic attributes of F-IIP, using predefined pathologic criteria and independent slide review by more than 1 pathologist. In our study, 3 experienced pulmonary pathologists evaluated lung tissue samples from 30 subjects with F-IIP for the presence, distribution, and severity of individual histopathologic features and then rendered an overall diagnostic impression dichotomized first between usual interstitial pneumonia (UIP) or “not UIP.”
MATERIALS AND METHODS
Study Patients With F-IIP
Thirty lung tissue specimens (29 surgical lung biopsies and 1 autopsy specimen) were randomly selected from a large cohort of familial interstitial pneumonia patients who had complete phenotype data including medical history, diffusing capacity for carbon monoxide (DLCO), chest computed tomography, and availability of lung tissue blocks.19 The 30 patients had a mean age of 51.96 years (median, 51 years; SD, 11.22) and represented 21 separate pedigrees. Sex was equally divided and the cohort was predominantly white (27 white, 1 Hispanic, 2 unknown). The mean DLCO at presentation was 43.75% predicted (median, 41% predicted; SD, 16.71). Each subject had a definite clinical diagnosis of IIP with the criteria established by the American Thoracic Society and the European Respiratory Society.4 All tissue slides for each subject were deidentified and assigned a subject number before analysis.
Selection of Histopathologic Features
Four were selected by their common occurrence in UIP as it is diagnosed today and further validated by current international multidisciplinary consensus (peripheral lobular accentuation of fibrosis, histopathologic honeycombing, fibroblast foci, and smooth muscle proliferation in dense fibrosis).4 The remaining 11 features were chosen as potential markers of other forms of IIP.4 All 15 variables were evaluated first for their presence or absence (primary descriptor), followed by 1 or more secondary descriptors related to extent, intensity, and cellular composition, as appropriate. For each case, 31 total descriptors were evaluated. Each pathologist was asked to collect data on all of the individual findings from the biopsy sections before rendering a global pathologic pattern diagnosis from a predefined list (>900 individual observations for each pathologist throughout the course of the study) (Table 2).
We purposely did not include any non-UIP IIP diagnoses (eg, lymphoid interstitial pneumonia, nonspecific interstitial pneumonia [NSIP]) in the not-UIP list of global diagnosis choices given our a priori desire to collect objective descriptions of the dominant histopathologic findings for the not-UIP group.
A booklet of data collection sheets was created for each of the study pathologists for use during a pretest training session. These booklets included high-resolution color photomicrographs illustrating each parameter to be evaluated, along with images demonstrating semiquantitative aspects of the feature (eg, mild bronchiolitis versus moderate bronchiolitis). All 3 reviewers met over a multiheaded microscope where the data sheets were discussed in the context of a random set of IIP surgical biopsies.
Three reviewers independently collected observations on each subject's biopsy slides and entered them in specially prepared booklets. The booklets had data collection sheets for each parameter, identical to the ones used in pretest training except for the use of black-and-white photographs as a reminder of the histopathologic parameter being assessed. There was wide variability in the number of tissue slides available for each surgical biopsy (1 to >10). In cases with multiple slides (potentially representing more than 1 biopsy sample, as all slides for a given subject were assigned a single unique identifier), all slides were reviewed in assessing the individual features and arriving at an overall histopathologic pattern diagnosis.
Establishing Observer Consensus
To control for variability between observers, during the analysis phase of the study, a “consensus result” was established for every histopathologic feature when 2 or more observers were in agreement. For fibroblast foci and vascular adventitial thickening, for which data entries were numerical, an average was determined. For secondary features, if all 3 reviewers chose different values, the result was entered as “no consensus.” If any parameter had only 1 data entry, the result was entered as “no consensus.” If consensus of 2 reviewers determined that a variable was absent, entries for the third reviewer's related modifier features were considered “not applicable.”
Agreement for the pathology reviewers' interpretations of the primary descriptor variables was estimated by Cohen κ coefficient.21 A nonparametric test was applied to survival, given the relative small sample size in the 2 comparison groups (non UIP = 17; UIP = 11). Age at death was evaluated for the entire cohort and was compared between groups by using the Wilcoxon rank-sum test.
Histopathologic Patterns (Features) in F-IIP
The consensus results for all parameters are presented in Table 3. The UIP morphologic pattern, defined as the co-occurrence of peripheral lobular accentuated fibrosis with temporal heterogeneity, histopathologic honeycombing, fibroblast foci, and smooth muscle proliferation in dense peripheral lobular fibrosis (Figure 2, A through D), was present in the F-IIP cohort, but the complete pattern was judged to be present by the individual reviewers in 23% to 50% of the cases.
For the 17 not-UIP cases, all but 1 case fell within the spectrum of “unclassifiable pulmonary fibrosis” (Table 3, footnote “a”). Some cases might be considered forms of NSIP, except for the high prevalence of histopathologic honeycombing (unexpected in NSIP).4,22,23 Example images from the not-UIP cases are presented in Figure 3, A through D. The most commonly identified individual features in the F-IIP biopsy specimens are presented in Table 4.
Considerable variability was identified between pathologists for the 15 primary variables, and more dramatically, for the descriptor variables. This diversity was not unexpected given the subjective nature of the observations and the high prevalence of advanced fibrosis and remodeling. Disagreement between pathologists regarding the presence or absence of 1 or more features was present to a variable degree in every case. The κ scores for combinations of reviewers, regarding several major features, are presented in Table 5.
For the global assessment of UIP or not UIP, a κ score of 0.37 was achieved, for 2 observers, consistent with fair interobserver agreement.21 Among the UIP-associated features, agreement was greatest for the presence or absence of fibroblast foci (0.71), subpleural fibrosis (0.64), and histopathologic honeycombing (0.45) and least for smooth muscle hyperplasia (0.07).
Survival for our study cohort was poor, with 28 of 30 subjects deceased (mortality of 93%), a mean age of death of 55 years, and a median age of death of 60.9 years. This may be in part a reflection of the very low mean DLCO values for the group, obtained by pulmonary function studies performed at the time of subject identification. Sixteen of 28 deaths resulted from respiratory failure caused by chronic pulmonary disease (57%), although this figure may be a low estimate, given missing data on specific cause of death in 12 of 28. There was a single surviving patient in each of the UIP and not-UIP groups. The age at death was significantly younger in the UIP pattern group, with a median age of 51.0 years for patients with a UIP pattern and 64.1 years for those with a pattern of not UIP (P = .03), indicating that the UIP pattern is associated with a decreased relative survival by 10 years. Kaplan-Meier survival curves comparing these 2 groups are presented in Figure 4, A and B.
The exact prevalence of familial forms of IIP is uncertain. Previously, familial forms of IIP were thought to be rare,24 although descriptions have appeared in the literature since 1907.25 Over the years, several case reports and clinical studies have described individual patients and small clinical series of familial diffuse lung disease (61 probands in total). On the basis of published accrual methods, most study subjects were identified through 1 or more index case(s) in which other family members had unexplained lung disease or through prospective national searches for kindred-linked lung disease.14–16,20,26–29
In our study, we were able to dismiss the prevailing hypothesis that all subjects with F-IIP have histopathology indistinguishable from sporadically occurring IPF, namely, usual interstitial pneumonia. Most of our patients had individual histopathologic features commonly associated with UIP, but most of our study subject lung biopsies would not qualify for inclusion in clinical trials targeting IPF today. Many different mechanisms of injury to the lung can result in advanced lung remodeling with fibrosis. Usual interstitial pneumonia, as defined today, is a distinctive form of pulmonary fibrosis characterized by the critical presence of so-called temporal heterogeneity, implying the coexistence of advanced microscopic fibrosis (with destruction of alveolar architecture) and uninvolved lung in the same biopsy section. This arrangement of fibrosis at the microscopic level produces a variegated appearance to the biopsy section at scanning magnification. Focal proliferations of immature fibroblasts covered by reactive type II pneumocytes (fibroblast foci) can be seen at the interface between fibrosis and preserved lung. Fibroblast foci are considered to be the leading edge of ongoing injury in UIP and are required to be present for diagnosis. However, fibroblast foci are not unique to UIP and occur in such disparate conditions as NSIP and recurrent pneumothorax in otherwise young, healthy individuals. Importantly, identification of only 1 of these expected findings is not sufficient for a diagnosis of the UIP pattern in current practice.30
To add stringency to the histopathologic assessments in this study, each reviewer was asked to identify the presence or absence of individual histopathologic features as an initial part of the slide evaluation. Once these were recorded, each reviewer then chose a global histopathologic assessment from a list of choices dichotomized by whether the overall pattern was UIP or not UIP. Using current international multidisciplinary criteria, 60% of our F-IIP cases did not qualify as UIP, mainly because of a lack of temporal heterogeneity. Their biopsy findings were felt to represent a form of unclassifiable fibrotic lung disease.
Whether or not this is biopsy selection bias or a true manifestation of the underlying pathologic process is impossible to judge accurately, given the inherent constraints of the study material (eg, mostly advanced disease being sampled). Importantly, cases having this “unclassified fibrosis” frequently had other changes that would reasonably argue against a UIP diagnosis, such as diffuse alveolar septal fibrosis resembling NSIP, multifocal peribronchiolar fibrosis, and/or diffuse chronic inflammation in the alveolar parenchyma. None of our study patients had granulomas in their biopsy results. The frequent finding of non-UIP pathologic features supports the observation of different types of IIP being identified within families, as reported by Steele et al.19 The observation of both UIP and non-UIP features among family members in their study supports the hypothesis that multiple factors, genetic or environmental, determine the phenotype in a given individual.
Given the potential permutations and substantial differences in observer opinion regarding the presence or absence of specific features (Table 3), it is not surprising that the breakdown by reviewer of UIP versus not UIP was variable (range, 23%–50%). To be fair, most of the lung specimens had significant fibrosis and no study to date has evaluated interobserver variability within a cohort for which “difficult to classify” lung fibrosis is the dominant pattern. Furthermore, none of the cases in our study had pathologic features entirely characteristic of an alternate idiopathic interstitial pneumonia, but some of the non-UIP cases would have been reasonably classified as NSIP in at least some areas of the sections evaluated. The high prevalence of advanced histopathologic honeycombing remodeling in these cases limited our ability to assign this diagnosis.22 Another interesting finding in our study was the occurrence of arterial adventitial fibrosis (> 1∶1 ratio comparing media thickness to adventitia thickness involving small pulmonary arteries <200 µm in diameter) in 40% of our subjects. This parameter was included on the basis of unpublished data from one of the authors (C.D.C., verbal communication, 2002) of a relationship between this finding in lung biopsies and presence of a systemic autoimmune disease. Despite this implied relationship, none of the patients in this study had clinical manifestations, or serology findings, to suggest undiagnosed connective tissue disease.
Additional histopathologic features (Table 3) that argued against a diagnosis of UIP were chronic bronchiolitis (60%), NSIP-like diffuse alveolar septal fibrosis with preservation of alveolar architecture (40%), peribronchial fibrosis (37%), diffuse chronic parenchymal inflammation (33%), and intra-alveolar eosinophils (27%). As might be expected, some of these features were also those most likely to be seen in patients with pulmonary fibrosis simulating UIP of IPF, namely, systemic autoimmune disorders (such as rheumatoid arthritis) and chronic hypersensitivity pneumonitis to inhaled organic antigen.
Potential flaws in our study include biopsy sampling bias and the lack of radiologic correlation. Sampling bias could have played a definite role regarding a confident diagnosis of UIP, based on published data showing potential discordances between biopsy samples from different sites in the same patient (eg, one biopsy specimen showing UIP, while another did not).31 We cannot exclude a similar sampling bias regarding our F-IIP biopsies. An independent assessment of radiologic imaging was not included as a parameter in our study, given our initial design as a pure assessment of histopathology.
The κ score for the global assessment of UIP or not UIP was 0.37, consistent with fair interobserver agreement (range, 0.19–0.37) for 2 observers. This finding is similar to that identified in a previous study of interstitial lung disease diagnosis by Nicholson et al,32 in which specialist pathologists were participants. In their study, 10 experienced pulmonary pathologists evaluated independently a set of lung biopsy specimens representing a wide spectrum of diffuse parenchymal lung diseases. No clinical or imaging information was provided. The observers were asked to select from a list of defined acute and chronic diffuse lung diseases, including the IIPs. The κ scores for the diagnosis of UIP were fair, similar to our results. Interestingly, the κ scores were worst for the diagnosis of NSIP, underscoring the need for better criteria for this IIP and rigorous application of these during assessment. As one might expect, Nicholson et al32 found that the κ scores were significantly better for diagnoses having distinctive histopathologic attributes (eg, the granulomas of sarcoidosis). This effect is also evident in our study, where agreement on the presence of specific features, such as fibroblast foci, was quite good (Table 3). Among the UIP-associated features we evaluated, the highest κ scores were obtained for the presence of fibroblast foci (κ = 0.71) and the lowest, for the presence of smooth muscle hyperplasia in fibrosis (κ = 0.07). This latter discordance may have been methodological, as the lower threshold for identifying this feature was never established (ie, how much is enough for including the presence of this finding).
Survival of F-IIP Cohort
Overall survival for this F-IIP cohort was poor (93% mortality by 10 years) and not unexpected given the high prevalence of advanced fibrosis with lung remodeling. Most deaths were attributable to respiratory failure from pulmonary disease, despite missing data on specific cause of death in 12 subjects. The atypical histopathologic features arguing against a diagnosis of UIP did seem to have a survival influence however, as those patients with F-IIP who achieved a consensus histopathologic pattern of UIP had a significantly earlier age at death by 10 years. The veracity of this finding will require confirmation in a larger cohort of patients with F-IIP.
The pathologic process underlying F-IIP was most frequently one of advanced and unclassifiable interstitial fibrosis with a high incidence of histopathologic honeycombing, smooth muscle hyperplasia in fibrosis, and fibroblast foci, in varying combinations. The biopsy specimens frequently lacked temporal heterogeneity, a critical and distinctive attribute of UIP. Moreover, even when 1 or 2 features were suggestive of UIP, non-UIP features were often present, such as diffuse chronic inflammation, diffuse alveolar septal fibrosis, alveolar space eosinophils, squamous metaplasia, and/or multifocal peribronchiolar fibrosis. The extent of histopathologic fibrosis may explain why historical studies of familial IIP concluded that the pathology was “UIP” at a time when the criteria for this diagnosis were less strictly applied. In our study, UIP by current criteria was identified by consensus in only 40% of cases, and this subset of patients had a significantly earlier age at death than the remainder of the study population, arguing for a more aggressive phenotype.
The authors are grateful to the patients and families who participated in this study. We thank Ms Janet Talbert of National Jewish Health for assistance in data entry and Henry Tazelaar, MD, for critical review of the final manuscript.
From the Department of Pathology, Mayo Clinic Arizona, Scottsdale (Dr Leslie); the Departments of Pathology (Dr Cool) and Medicine (Dr Schwartz), University of Colorado School of Medicine, Aurora; the Department of Pathology, Division of Pathology Clinical Services (Dr Sporn) and Division of Pulmonary, Allergy, and Critical Care Medicine (Drs Steele and Wahidi), Duke University Medical Center, Durham, North Carolina; the Division of Biostatistics and Bioinformatics (Dr Curran-Everett) and the Department of Medicine (Dr Brown), National Jewish Health, Denver, Colorado.
Travel and accommodation expenses to Denver, Colorado, for case review was supported by a grant from the National Heart, Lung, and Blood Institute (NHLBI) of the National Institutes for Health and Duke University Medical Center, Durham, North Carolina (T.A.S.) and by Mayo Clinic Arizona, Scottsdale (K.O.L.).
Statistical support for Mayo Clinic statisticians (Amylou Dueck, PhD, and Yu-Hui Chang, PhD) was provided by Mayo Clinic Arizona, Scottsdale. The authors have no relevant financial interest in the products or companies described in this article.
This study was presented in part at the biennial meeting of the Pulmonary Pathology Society, Annecy, France, June 16, 2005.