Interstitial lung diseases: an epidemiological overview

Introductory remarks on epidemiological studies

Diffuse interstitial (or parenchymal) lung diseases (ILDs) represent a very large group of more than 200 different entities, many of which are rare or “orphan” diseases. Much remains unknown or debatable for many of these ILDs, notably issues of prevalence, incidence and mortality rates.

Epidemiology can be defined as the study of the distribution of disease and of the factors that determine this distribution (see article by Antó and Cullinan 1 in this Supplement). This apparently simple definition includes a wide number of applications. The most important are; the measurement of the magnitude of health problems (including prevalence, incidence and associated burdens); the identification of geographical and temporal distributions (including the relevant patterns of clustering); the investigation of outbreaks and clusters; the study of natural history and aetiology (including both environmental and genetic determinants); the assessment of validity and reproducibility of diagnostic tests that define disease; and the identification and assessment of preventive and therapeutic measures.

The successful application of epidemiology to the previous topics is based on the appropriate use of a large array of concepts and methods, many of which have evolved over previous decades and are still evolving 2. There is now growing evidence that inconsistency of results between different studies is, in part, due to the presence of different types of bias ranging from selection to information bias or insufficient adjustment of confounding factors 3. Recently, the introduction of evidence-based approaches has resulted in a number of guidelines that are helpful for a critical appraisal of epidemiological investigations 4.

In ILD, the major types of epidemiological studies can be subdivided into the following categories.

Classification and definition of interstitial lung disease

Internationally accepted disease definitions and classifications are a prerequisite without which, useful comparisons cannot be made between population studies (see article by Antó and Cullinan 1 in this Supplement). Recently, a number of important consensus statements have been published 5–7, and it now appears likely that, for the first time, worldwide agreement will be reached on the classification of the idiopathic interstitial pneumonias (IIPs) (see Review by du Bois and Wells 8 in this Supplement).

In the whole group of ILDs, several types of classification have been applied, such as systematic classifications based on aetiology or on specific disease entities 9, 10, or alternative classifications based on clinical, histological, or radiological patterns 5, 11. Systematic classifications, e.g. the one by Crystal et al. 9 or the International Classification of Diseases (ICD)‐9 10, in which each individual ILD is mentioned only once, are appropriate for epidemiological studies. The classification by Crystal et al. 9 in ILD of known and unknown aetiology (table 1⇓), is useful in this regard and easy to apply. It is possible to refine this classification by subdividing ILD of unknown aetiology (65% of all ILD) into diseases confined to the lung (most frequently idiopathic pulmonary fibrosis (IPF)) and lung diseases associated with systemic diseases (such as sarcoidosis and connective tissue diseases). Alternative approaches include histological classifications, such as that proposed by Schwarz 11, based upon responsiveness to treatment, and classifications based upon clinical plus aetiological features at first presentation, of the sort recently proposed by the British Thoracic Society (BTS) 5. However, these two latter classifications 5, 11, in which individual ILD may be classifiable into more than one subgroup, cannot be adapted easily to epidemiological studies. For example, pulmonary sarcoidosis is most often associated with a good response to treatment. However, a smaller subgroup with end-stage fibrotic disease would be grouped separately in the histological classification of Schwarz 11, which is based upon the reversibility of disease. Similarly, in the BTS classification 5, which is based on clinical features at first presentation, hypersensitivity pneumonitis may be classified in the subgroup of “episodic ILD” as well as in that of “chronic ILD”.

In order to be able to compare the data from different epidemiological studies it is mandatory that the same definitions for the disease entities are applied. Some of these entities have highly specific histological features on the basis of which the diseases may be defined. Other entities, however, are morphologically nonspecific and the diagnosis has to rely on the combination of clinical (including radiological) and pathological presentation. One striking example of a group of diseases, in which a change in classification has had major consequences, is that of the IIP (see Review by du Bois and Wells 8 in this Supplement). Until some years ago, large registries of patients with ILD 12–14 and epidemiological studies on IPF 15, 16 were based upon the pathological classification of Liebow 17 from ∼30 years ago. However, it is now apparent that recent modifications made by Katzenstein and Myers 18 have increased the clinical relevance of disease classification. This evolution is not surprising since both classifications are separated by 3 decades of research into both aetiopathogenic mechanisms and histological specificity, and also to structure-function correlations and clinicopathological specificity. As a result, for some of the ILDs originally considered idiopathic, aetiology has been elucidated and other entities are now more accurately described in terms of spatial and temporal distribution in the lung parenchyma, which has led to new terminologies (see article by Verbeken 19 in this Supplement). In the recent International Consensus Statement on Nomenclature 6 a clear distinction has been made between IPF, histologically defined as usual interstitial pneumonia (UIP), and the other forms of IIP, especially the recently defined entity of nonspecific interstitial pneumonia (NSIP) (table 2⇓). Historically, the 5‐yr survival of patients with IPF was ∼50% 20, 21. However, it is now known that the subgroup with UIP has a 5‐yr survival of only 20% 22, whereas well over half of the cases with desquamative interstitial pneumonia (DIP) or NSIP have a better outcome 22–24. By contrast, acute interstitial pneumonia (AIP), first described by Hamman and Rich 25, 26 and subsequently characterized as a separate entity by Katzenstein et al. 27, has an appallingly bad prognosis. The grouping together of UIP, NSIP, DIP and AIP in large epidemiological series was unavoidable. However, the distinct natural histories and treated courses of these entities demand that further epidemiological studies be performed. Before this work can be undertaken, a final consensus must be reached on disease classification by the American Thoracic (ATS) and European Respiratory Society (ERS), which will hopefully be published this year.

Furthermore, it may be difficult to differentiate between IIP and other ILD of unknown or even known (but not recognized) aetiology, and thus, misclassifications are inevitable even in published data. AIP may be confused with other clinical entities characterized by rapidly progressive interstitial pneumonia 28. Furthermore, the ILD patterns associated with inflammatory bowel disease (IBD) (e.g. NSIP, DIP, and granulomatous or eosinophilic lung disease) may escape detection or be wrongly considered as unrelated to the IBD 29.

In addition, radiological or histological organizing pneumonia (OP) patterns associated with a variety of exogenous causes, may be wrongly considered as cryptogenic organizing pneumonia (COP; otherwise known as idiopathic bronchiolitis obliterans organizing pneumonia (BOOP)) 30, 31. Finally, chronic fibrotic ILD with honeycombing due to an exogenous cause that is no longer identifiable, can wrongly be classified as IPF because of its UIP pathological pattern.

Diagnostic criteria and techniques

Epidemiological studies require standardized and detailed diagnostic criteria. More recent technical diagnostic tools are more sensitive (e.g. bronchoalveolar lavage (BAL), high-resolution computed tomography (HRCT)), and some ILDs may now be easier to detect and diagnose. For instance, it has been suggested that the increase in mortality rates due to IPF reported by Johnston et al. 32, in several countries between 1979 and 1992, could, at least in part, be due to more accurate diagnosis.

Usually, a stepwise approach is taken in the diagnosis of ILD, starting with a noninvasive evaluation (detailed history, physical findings, blood tests, imaging techniques, lung function tests), followed by semi-invasive procedures (fibreoptic bronchoscopy with BAL and transbronchioloalveolar biopsies (TBB)) and then by more invasive, mostly surgical biopsy techniques, e.g. open lung biopsy or video-assisted thoracic surgery (VATS). Historically, the histopathological assessment of a surgical biopsy specimen has been viewed as the “gold standard” among the diagnostic tests in ILD, especially in IPF, but it is clear that only small, possibly nonrepresentative, samples can be obtained. Furthermore, high proportions of patients with IPF are too compromised (due to severe respiratory disease or associated cardiac disease) to undergo a surgical procedure. In a further large subgroup, either the patient declines the procedure or the physician does not believe that practical management would be influenced by findings at biopsy. As a result, few patients with IPF undergo surgical biopsy evaluation. For example, in two UK surveys of patients with IPF, only 7.5% 33 and 12% 34, respectively, had an open lung biopsy. It must also be emphasized that for large-scale epidemiological (especially multicentre) studies, detailed information on the reliability of the histopathological classification of ILD should be available. To date, no systematic studies on intra- and interobserver pathological variability have been performed. For these reasons, surgical biopsy evaluation has never been a satisfactory diagnostic “gold standard” for epidemiological studies.

Thus, it is clear that other diagnostic strategies must be developed for epidemiological studies, especially in IPF. Fortunately, it is now increasingly accepted that in IPF (and in most other ILDs), a secure diagnosis can often be made by HRCT, now considered an alternative means of evaluating morphological appearances. In some patients, a specific diagnosis can be obtained from the CT appearances alone; a correct first choice diagnosis is made by computed tomography (CT) in 75–90% of patients with various major ILDs, including sarcoidosis, silicosis, IPF, lymphangitis carcinomatosis, and Langerhans' cell histiocytosis 35–37 (table 3⇓). However, CT findings must be integrated with the clinical evaluation and other investigations in order to reach a specific diagnosis. In this regard, BAL continues to play a crucial diagnostic role. The BAL pattern may be lymphocytic in sarcoidosis (with an increase in the CD4/CD8 ratio in most patients) and extrinsic allergic alveolitis (with a normal or decreased CD4/CD8 ratio), neutrophilic and possibly eosinophilic in IPF, or mainly eosinophilic in a variety of disorders (table 4⇓). Surgical biopsy procedures are increasingly reserved for patients in whom the CT appearances are indeterminate, and those in whom CT findings and other clinical and investigative features are at odds. Thus, histological evaluation is a crucial, final diagnostic arbiter in difficult cases, but can no longer be used to select representative patient groups for epidemiological studies. A new “gold standard” definition has long been needed.

The ATS/ERS core committee has redefined diagnostic criteria for IPF 6. According to the new definition, IPF is now recognized as a distinct and specific form of chronic fibrosing interstitial pneumonia, limited to the lung and associated with UIP on surgical biopsy. In the absence of a surgical lung biopsy, the presence of all of the major, and three of the four minor diagnostic criteria listed in table 2⇑ of the Review by du Bois and Wells 8 in this Supplement, increases the likelihood of a correct clinical diagnosis of IPF. Notably, the new techniques, HRCT and BAL, are included among the four major criteria 6. In IPF, the HRCT shows patchy, predominantly subpleural, bilateral reticular abnormalities, often with associated traction bronchiectasis and bronchiolectasis and/or subpleural honeycombing. Ground-glass opacities may be present but should be limited in extent. The accuracy of a confident diagnosis of IPF made on HRCT by a trained observer appears to be ∼90% 32–34 and ∼70% in patients with histological UIP 38. Today, a surgical lung biopsy in suspected IPF can be limited to those patients who show an atypical pattern on HRCT or in BAL (e.g. a predominant lymphocytosis) or an atypical clinical presentation (e.g. young age or short duration of illness).

The diagnosis of sarcoidosis needs a compatible clinical picture of a systemic disease, histological demonstration of noncaseating granulomas and exclusion of other diseases capable of producing a similar histological or clinical picture 7. TBB is the recommended diagnostic procedure in most cases. In some cases of sarcoidosis, biopsy proof is not needed. Clinical and/or radiological features alone may be diagnostic for patients with stage I (reliability of 98%) or stage II (89%) disease 39. Furthermore, patients with the classic Löfgren's syndrome, patients with a BAL lymphocytosis and an increased CD4/CD8 ratio, or those with the appearance of a panda and lambda pattern on a total body gallium scan, may prevent the need for invasive diagnostic procedure (see Review by Costabel 40 in this Supplement). HRCT is not routinely indicated in patients with sarcoidosis, except for detection of complications if chest radiographic findings are atypical, or in those patients with a normal chest radiograph but a high clinical suspicion of the disease 7.

In extrinsic allergic alveolitis (hypersensitivity pneumonitis), biopsy is rarely required (see Review by Bourke et al. 41 in this Supplement). The diagnosis is usually based on: 1) systemic symptoms and signs (fever and weight loss together with dyspnoea, cough and crackles in acute cases associated with an appropriate exposure); 2) evidence of ILD (either by imaging studies or pulmonary function tests); and 3) evidence of exposure (demonstration of antigen in the environment, improvement of symptoms after withdrawal from antigen, or detection of antibody to a specific triggering agent in patients' serum and/or BAL).

In doubtful cases, supportive criteria are: 1) BAL with >40% lymphocytes; 2) HRCT pattern with widespread ground-glass attenuation mixed with areas of “mosaic perfusion”, variably associated with poorly defined centrilobular micronodules in a nonsmoker; 3) a positive inhalation challenge; and 4) compatible histological changes on open lung biopsy specimen. It is notable that in this ILD, the new techniques, HRCT and BAL, have also been included in the set of diagnostic criteria 42.

Also, for other entities of ILD (table 1⇑), diagnostic criteria have been redefined in recent decades. It is not possible to enumerate all of these in the present report. It must be emphasized that for all these diseases the most precise current definitions and criteria should be used in all epidemiological studies.

Sources/population groups for epidemiological studies

Epidemiological data may be obtained from different sources or population groups, using different study designs.

Population data

Few data are available on the epidemiology of ILD in the general population as few systematic large-scale registration programmes on ILD have been published. In a population-based study in Bernalillo County, New Mexico, Coultas et al. 12 found a prevalence of ILD of 81 per 10⁵ in males and 67 per 10⁵ in females, and an incidence of 32 per 10⁵ per yr in males and of 26 per 10⁵ per yr in females. They also calculated the prevalence and incidence of several subgroups of ILD (table 5⇓). A potential limitation of this study is the fact that in only ∼7% of the ILD patients, the diagnosis was confirmed by open lung biopsy. Despite this concern, this study shows that the incidence/prevalence of several ILDs, especially IPF, are much higher than previously published figures.

If the frequency data for ILD in the study of Coultas et al. 12 are compared with the registries by pneumologists in Flanders 13, 49, Germany 14 and Italy 50, the distributions show some remarkable similarities but also differences (see table 1⇑ in article by Thomeer et al. 51 in this Supplement). In the New Mexico registry, prevalences of sarcoidosis (12%) and especially hypersensitivity pneumonitis (0%) were lower, and prevalences of nonspecific fibrosis (32%), pneumoconiosis (14%) and ILD due to connective tissue disease (13%) were higher than in the European registries. Drug-induced ILDs are very low in all registries (1.9–3.3%) and are undoubtedly underestimated (see Review by Camus et al. 61 in this Supplement). A major problem of most registries, such as the aforementioned European ones 13, 14, 49, 50, is that they are often incomplete and underestimate the true incidence of drug-induced ILDs.

Mortality data

Official mortality data may be more systematic than other registries, but have several disadvantages. These include the possibility that classification may be incorrect due to the peculiarities of ICD codes, that mortality rates differ significantly for the different ILD (e.g. very low in sarcoidosis) and the fact that for some categories of systemic diseases, such as sarcoidosis or collagen vascular diseases, it is not mentioned whether the lung was involved and/or was the cause of death. Therefore, the type of data from these national mortality registries is quite different from that obtained from incidence or prevalence registries.

Coultas and Hughes 62 examined death certificates and state mortality data in patients in New Mexico with a clinical diagnosis of ILD before death, and concluded that an ILD was listed somewhere on the death certificate in <50% of the patients, and as an immediate cause of death for only 15%.

A critical analysis of mortality data is, therefore, not inappropriate. Most countries routinely code cause of death using death certificates and these provide a potential source of data on disease incidence for ILD. These data will obviously be most appropriate for diseases that usually lead to death, such as IPF, and are of little use for diseases with a good prognosis, such as sarcoidosis. Nevertheless, in 1980 Turner-Warwick et al. 20 found that in patients known to have cryptogenic fibrosing alveolitis (CFA, which is synonymous to IPF) in their lifetime, the cause of death was directly attributed to CFA in only 55%. In the ICD‐9 10, a specific disease code for CFA/IPF was introduced for the first time (516.3). Studies by Johnston and coworkers on the use of this new code for death certificates 32 and hospital admission data 15 in the UK, suggest that most patients who are coded as having CFA do have this disease, but that about half of the people known to have CFA are not coded correctly and many receive the less precise code of “postinflammatory fibrosis” (515) 32. In the USA, an underreporting of IPF/CFA is also likely due to use of the less precise ICD code 515 instead of 516.3. Despite the limitations of using these data, there does appear to be an increase in the annual number of deaths from CFA in the UK and a number of other industrialized countries 32, 63.

Strikingly, to the best of the authors' knowledge, large-scale mortality data comparing state mortality data and registries by pulmonologists for different ILDs have not been published. In the framework of a pilot study, the mortality data from the registry of the University Hospital Gasthuisberg of Leuven 64, have been compared with the data of the National Institute of Statistics of Belgium 65 (table 6⇓). In the University Hospital Gasthuisberg, the percentage of 5‐yr mortality was highest in IPF (44%, all groups together i.e. UIP, DIP, etc.) and lowest in sarcoidosis (2%); 8.4% of all registered ILD patients had died within 5 yrs due to IPF and only 0.62% due to sarcoidosis. Conversely, the national mortality statistics showed overall mortality rates for IPF (0.05 per 10⁵) that were similar to those for hypersensitivity pneumonitis (0.06 per 10⁵), but were much lower than those for sarcoidosis (0.15 per 10⁵) and connective tissue disease (0.39 per 10⁵); however, the percentage of mortalities that were due to failure of the lungs in these latter two diseases are not known.

Interstitial lung disease in children

The spectrum of paediatric ILD includes a large, heterogeneous group of rare disorders that differ from the adult forms of ILD 66. Because ILDs in children are rare, there are no epidemiological prevalence or incidence data, and most studies in the literature present only small numbers relative to those reported for ILD in adults. It is estimated that 20% of cases of ILD in children are caused by an infectious agent, a percentage much higher than that shown in adult ILD populations. Infectious aetiologies of ILD include adenovirus bronchiolitis/pneumonia leading to bronchiolitis obliterans, and chronic lung disease from influenza, Mycoplasma and Chlamydia infections. Recurrent aspiration secondary to gastroesophageal reflux is another known cause of ILD 67. The “classic” forms of idiopathic ILD (UIP, DIP, lymphoid interstitial pneumonia (LIP) and BOOP/OP) rarely occur in childhood 66, and often have a significant morbidity and a poor prognosis 68. The most frequent forms are LIP or follicular bronchiolitis and NSIP; the latters response to corticosteroids in childhood disease is similar to that in adult disease 69. Open lung biopsy appears to make a substantial contribution to the management of ILD in children, since it resulted in a change of management in 56% of cases in one study 69. Paediatric ILD should be considered as quite distinct from adult ILD in most cases and especially in IIPs.

Familial interstitial lung disease

The occurrence of familial forms of ILD suggests the potential importance of genetic influences such as major histocompatibility complex (MHC) alleles and fibrogenic gene polymorphisms in the aetiology (see Report of Working Group 2 by Verleden et al. 70 in this Supplement). Few epidemiological data are available on the frequency of familial ILD 71. In sarcoidosis, familial forms have been described in 1–8% of cases 45, 72, particularly mother/child, monozygote twins and same-sex pair associations. These studies of family relationships have not found a clear Mendelian pattern to the inheritance of sarcoidosis, but rather an inherited predisposition for the disease, which is probably polygenic.

Cases of familial IPF/CFA are rare and account for <1% of all cases in the UK. In most cases, the inheritance appears to be autosomal dominant, with variable penetrance, and one study has demonstrated evidence of subclinical alveolitis in asymptomatic first-degree relatives 73. Clinically, the familial disease appears to be indistinguishable from the idiopathic condition, but may present at a younger age, although this may be the result of active screening. Other inherited forms of ILD, such as tuberous sclerosis complex and the metabolic storage diseases, are more clearly familial 74, 75.

Sarcoidosis

From radiographic population screening programmes (especially for detection of tuberculosis), a global prevalence of 10–40 per 10⁵ and an incidence of 10 per 10⁵ has been estimated 44, 45. The incidence appears to be higher in some countries (Scandinavia), some population groups (e.g. 40 per 10⁵ in African Americans) and in some families 86–92. International pulmonary registries have illustrated differences in the presentation of sarcoidosis in different countries 86; in Asia the majority of cases presented with a radiological stage I (mediastinal and hilar lymphadenopathies), and a positive tuberculin skin test was found more frequently than in other countries. However, erythema nodosum has not been reported in the Japanese 93, 94, is rare in African Americans, is the presenting symptom in 18% of cases in Finland 94 and occurs in about 30% of British sarcoidosis patients 95. In the USA, a higher percentage of patients <40‐yrs-old was found (68% of the patients in USA) than in other countries, as well as a higher percentage of patients of African background (58% of all sarcoidosis patients in the USA).

Sarcoidosis is rarely reported in Portugal, India, Saudi Arabia or South America, partly because of the absence of mass screening programmes, and probably also because of the presence of other, more commonly recognized granulomatous diseases (tuberculosis, leprosy, fungal infection). Further, in Catalonia (Spain), an incidence of only 1.2 per 10⁵ inhabitants has been reported 96.

Few mortality data on sarcoidosis are available. A 5‐yr mortality rate of 2% has been reported by a referral university hospital 64 (table 6⇑). A review of autopsy records of Japanese nationwide data and of two American institutions (Mayo Clinic, Rochester and University of Southern California, Los Angeles) showed that in Japan, heart involvement was the most common cause of death, and in Western countries, lung involvement was most common, with a mortality of about 5% 96. Fleming 97 reported much higher mortality rates among patients with cardiac sarcoidosis in 1988.

Idiopathic pulmonary fibrosis/cryptogenic fibrosing alveolitis

For IPF, changes in definitions from the old histological classification by Liebow 17 to the recent adaptation by Katzenstein and Myers 18 and the ATS/ERS Consensus 6 have to be taken into account. In addition, the histological pattern of UIP may be found in conditions other than IPF, e.g. drug-induced or collagen vascular ILD, asbestosis, etc. (see the Review by du Bois and Wells 8 in this Supplement).

The exact incidence, prevalence and mortality of IPF are not known 98. Scott et al. 76 estimated a prevalence for IPF of 3–6 cases per 10⁵ in the general population. More recently, Coultas et al. 12 found a prevalence of IPF of 20 per 10⁵ in males and 13 per 10⁵ in females. They found an incidence of 11 per 10⁵ per yr in males and of 7 per 10⁵ per yr in females (table 7⇑), but the criteria that provided the basis for these data are not precisely defined. About 80% of the IPF patients are ≥65‐yrs-old.

Among seven countries, Hubbard et al. 16 found a large variation in mortality data of IPF/CFA (ICD 516.3) from 0.03–1.3 per 10⁵ and of postinflammatory fibrosis (ICD 515) from 0.6–1.7 per 10⁵ (table 8⇓), which they attributed, at least to some extent, to differences in classification. Indeed, the countries with higher frequencies for postinflammatory fibrosis tended to have lower frequencies for IPF and vice versa. In England, the mortality rate of IPF in 1979 was estimated at ∼1.5 per 10⁵ population in males (standardized for age) and 1 per 10⁵ in females 32. In Japan, the mortality rate for IPF was estimated to be 3.0 per 10⁵ 99. In the USA, Mannino et al. 100 found higher age-adjusted mortality rates in Whites than in African-Americans, but this finding has been attributed to incomplete reporting 6.

Death certificates of IPF may contain errors due to several causes: cases of IPF (ICD 516.3) may be coded as having died from other causes (e.g. infection, heart failure), and may be coded as postinflammatory pulmonary fibrosis (ICD 515). Panos et al. 101 found that of their IPF patients who died, respiratory failure was the cause in 39%, cardiovascular disease in 27%, lung cancer in 10% and other causes in 18%. For all these reasons, statistics by death certification generally under-report the diagnosis 15, 20, 32, 62, 76. Johnston et al. 32 and Hubbard et al. 16 found that mortality rates increased in most countries in the years after 1979, which in part may be due to improved diagnostic skills (see earlier).

In a cohort study in England, Hubbard et al. 102 found that in newly diagnosed cases of CFA/IPF (i.e. incident cases) median survival was only 2.9 yrs and that the expected life span in these cases was reduced to ∼7 yrs. Median survival for prevalent cases was 9 yrs compared to an expected 13 yrs. They concluded that the estimate of a median survival for CFA/IPF of ∼5 yrs in previous clinical case series 20, 21, 99, in which prevalent cases were also included, may thus have been influenced by this inclusion bias.

Hypersensitivity pneumonitis/extrinsic allergic alveolitis

The list of different aetiological agents and sources of antigens capable of inducing hypersensitivity pneumonitis (HP) or extrinsic allergic alveolitis (EAA) is already enormously long and continues to grow. However, as Selman 56 states, “the literature is replete with a number of HP-like disorders”. Indeed, some factors may cause HP and other forms of inhalation pathology, such as organic dust toxic syndrome (ODTS), chemical pneumonitis and inhalation fever. ODTS is much more frequent than hypersensitivity pneumonitis: it is at least 20–50 times more common than farmer's lung in Sweden 103. In many publications reporting high incidences of HP (e.g. farmer's lung), cases of ODTS may also have been included. However, the mechanism, evolution and prognosis of ODTS is quite different from that of HP.

The two most extensively studied forms of HP are farmer's lung and pigeon breeder's lung (table 7⇑). The prevalence of farmer's lung has been estimated to range from 10–200 per 10⁵ inhabitants in different zones of England 54 and in Finland 55, and 4–170 per 1,000 farmers in France 59 and the USA 77 (see Review by Bourke et al. 41 in this Supplement).

Prevalence of clinical disease in pigeon breeders has in the past, been estimated at ∼1 per 1,000 78, but more recently, prevalences of >10% and sensitization rates of 32% have been reported in those with regular high exposure 104.

The prevalence of HP varies markedly between countries, due to the local climate, season, geographical conditions, local customs, smoking history and presence of industrial manufacturing plants. As many as 50% of individuals exposed to environmental antigens that can cause hypersensitivity pneumonitis develop a lymphocytic alveolitis but remain asymptomatic 105, whereas only a few develop clinical symptoms of the disease. Cigarette smokers have a very low incidence of HP compared with nonsmokers 104, 106, which has been attributed to the low level of expression of costimulatory molecules (e.g. B7) on alveolar macrophages from smokers, and to their resistance to further upregulation 107. Furthermore, some cofactors might enhance the risk of developing the disease, such as endotoxin exposure or viral infection, as has been shown in a mouse model of HP where Senda virus infection increases susceptibility for up to 6 months 108. A highly significant increase in prevalence of farmer's lung in dairy farmers in the French Doubs province has been found with increasing altitude 109. A genetic predisposition, e.g. a particular HLA specificity 110, 111, may also enhance the risk of developing HP (see also Report of Working Group 3 by Nemery et al. 112 in this Supplement).

Interstitial lung disease in connective tissue disease

Lung involvement is now considered a major source of morbidity in connective tissue disorders (CTD). In systemic sclerosis (SSc) and polymyositis/dermatomyositis (PM/DM), pulmonary disease (including ILD, aspiration pneumonitis and pulmonary vascular disease) is now the most common cause of death. In rheumatoid arthritis (RA), up to 20% of fatalities are due to bronchopneumonia 113, 114. Therefore, it is surprising that estimates of the prevalence of pulmonary involvement vary widely in CTD. The major difficulty, which also applies to ILD in general, is that the frequency of diagnosing ILD is critically dependent upon the methods used to detect lung abnormalities (see Review by Lamblin et al. 115 in this Supplement).

Symptoms are unreliable in defining lung involvement in CTD. Exertional dyspnoea is common, especially in SSc (occurring in at least half of the patients) 116. However, dyspnoea cannot be used as a marker of ILD, as it often occurs when the work of locomotion is increased (in arthritis or myositis) without lung involvement. Furthermore, the opposite is sometimes true: the presence of severe systemic disease may prevent patients exercising sufficiently to experience dyspnoea, despite severe impairment in lung function tests. Thus, objective methods must be employed, but estimates of prevalence of ILD remain highly variable. In general, ILD is most commonly identified in histological or autopsy studies and in lung function surveys. CT often demonstrates abnormalities not seen on plain chest radiography. A highly variable proportion of patients have evidence of alveolitis on BAL (although it is not yet clear whether subclinical alveolitis is synonymous with early ILD in CTD).

The difficulty in defining prevalence is well-illustrated in individual CTD. In SSc, pulmonary fibrosis is one of the American Rheumatism Association's (ARA) diagnostic criteria for the disease 117; pulmonary fibrosis is found at autopsy in >75% of patients 118, 119 and lung function abnormalities are found in 90%. ILD is seen on chest radiography in 25–65% 82, with the highest prevalence amongst CTD; on CT, limited fibrosis is identified in a further subgroup with normal chest radiograph appearances 120. Thus, even in a disease in which ILD is present in the majority of patients, in most studies the prevalence may be variably estimated as lying between 25% and 90%.

In diseases in which clinically overt ILD is less common, such as RA, the data on prevalence are even more conflicting. In an open lung biopsy study of volunteers with RA (most without clinical evidence of lung involvement), histological abnormalities compatible with ILD were identified in 60% 121. This finding is consistent with the observations that gas transfer is reduced in 40% of unselected RA patients 122 and that lung disease is evident on CT in up to 50% 123, 124. However, in striking contrast, there is evidence of pulmonary fibrosis on chest radiography in only 1–20% (in three very large chest radiographic series of unselected patients with RA 83, 125, 126). Thus, it can be argued that clinically significant ILD is difficult to predict.

These findings are mirrored in ankylosing spondylitis (AS), systemic lupus erythematosus (SLE) and Sjögren's syndrome (SS). In general, plain chest radiographic series show a prevalence of pulmonary fibrosis of <10% 80, 127, 128; however, on CT, ILD is present in ≥30% of patients 129–131, and subclinical alveolitis on BAL is common 132, 133. The definition of the prevalence of pulmonary fibrosis in SS has been particularly hampered by failure to discriminate between primary and secondary SS in early series (lung disease in secondary SS being ascribable to the primary CTD), and variations in diagnostic criteria for SS 134. PM/DM is less well studied; ILD is clinically overt in up to 30% of patients 135, but no single definitive evaluation of large numbers of patients has been performed.

It may be concluded that through the use of sensitive diagnostic methods such as CT and BAL, evidence of ILD can be identified in a majority, or large minority of patients with CTD. However, with the use of chest radiography and traditional clinical evaluation only, ILD is identified in <10%, except in SSc and PM/DM. The dilemma in epidemiological studies is whether to view subclinical ILD, detectable only on CT or at BAL, as part of a continuum with clinically overt ILD. Present data do not justify this assumption: it is not clear whether subtle interstitial abnormalities identified using highly sensitive tests necessarily evolve into important fibrotic lung disease. However, appropriate longitudinal studies should be designed to address this issue.

It, therefore, appears to be difficult to provide meaningful data on prevalence of ILD in CTD. The prevalence of RA is estimated at 2% and evidence of ILD on chest radiography and routine lung function may be present in up to 20% of these patients 83 (table 7⇑). The prevalence of SLE is 40 per 10⁵ population 81 with an estimated clinically relevant ILD of 10% 80 and the prevalence of SSc is 10 per 10⁵ population 81 with ILD in 20–65% of these 82.

Drug-induced interstitial lung disease

Drug-induced lung diseases often have no pathognomonic signs or symptoms 136 and are underdiagnosed (see Review by Camus et al. 61 in this Supplement). Indeed, they account only for 2.5–3% of all ILD in several registries 12–14, 49, 50. In fact, some cases of presumed IIP may be due to unrecognized drug-induced ILD.

Several groups of drugs are especially prone to induce ILD; some examples are listed below.