Abstract
As idiopathic pulmonary fibrosis emerges as an important public health problem, there is a need to coordinate data on incidence and mortality globally. This study aims to systematically assess all available studies to investigate the global burden of disease.
Medline and Embase databases were searched systematically for all population-based studies of incidence or mortality of idiopathic pulmonary fibrosis. Clinical case series and prevalence studies were excluded. The search was supplemented using the Google search engine, hand-searching of references and conference abstracts. Data were extracted independently by two authors using a pre-specified proforma, with assessment of methodological quality.
34 studies were identified, providing data from 21 countries from 1968–2012. 28 studies reported incidence data and eight reported mortality data. In studies from the year 2000 onwards, we estimated a conservative incidence range of 3–9 cases per 100 000 per year for Europe and North America. Incidence was lower in East Asia and South America. The majority of studies showed an increase in incidence over time.
The incidence of idiopathic pulmonary fibrosis is increasing worldwide and rates are coming together across countries. Current data suggest incidence is similar to that of conditions such as stomach, liver, testicular and cervical cancers.
Abstract
Incidence of IPF varies worldwide but seems to be increasing to rates around 3–9 per 100 000 per year http://ow.ly/KSeMZ
Introduction
The incidence of idiopathic pulmonary fibrosis (IPF) has been reported in several studies worldwide and appears to be increasing [1–3], but different methodologies of case ascertainment and classification systems have prevented valid comparison between studies. A small number of published reviews have examined incidence and prevalence data from certain countries [4–8] but, to date, there has been no comprehensive systematic review of international incidence and mortality data in IPF.
This study aims to review all population-based studies of incidence and mortality of IPF worldwide, in an attempt to define the global burden of disease.
Methods
Data sources and searches
This review was registered on the PROSPERO international prospective register of systematic reviews on 6 February, 2014 (registration number CRD42014007452), with a pre-specified protocol. We aimed to include all original studies and abstracts assessing incidence or mortality of IPF, with no restrictions on language that might exclude certain areas of the world. We excluded prevalence studies and clinical case series, and focussed only on population-based studies with a specified denominator population. We excluded studies that examined interstitial lung disease (ILD) other than IPF (for example, in association with connective tissue disease) but included studies that looked broadly at ILD with possible sub-classification.
We searched Medline (1946 to present) and Embase (1974 to latest) using OvidSP, with the latest search in June 2014. We used the terms “idiopathic pulmonary fibrosis”, “interstitial lung disease”, “pulmonary fibrosis”, “cryptogenic fibrosing alveolitis”, “usual interstitial pneumonia” and “idiopathic fibrosing alveolitis”, combined with “incidence”, “mortality”, “death” and “epidemiology”. The full search strategy is detailed in the online supplementary material. We supplemented this search using the Google search engine with combinations of the search terms. We hand-searched abstract lists from the American Thoracic Society, British Thoracic Society and European Respiratory Society annual conferences, and screened reference lists of selected articles, as well as review articles identified. We also planned to review the websites of national statistics agencies for routine mortality data; this process was completed separately prior to the main review and has been published separately [9].
Study selection
All stages of paper selection/elimination were performed independently by two authors (screening of titles (John Hutchinson and Richard Hubbard), review of abstracts (John Hutchinson and Tricia McKeever), and screening of papers and additional sources (John Hutchinson and Andrew Fogarty)), with disagreements resolved in each case by discussion. Non-English texts were translated using Google Translate with a plan for more comprehensive translation in case of any uncertainty as to the message or relevance.
Data extraction and quality assessment
Data extraction took place using a pre-designed form that was piloted on four different studies independently by two reviewers (John Hutchinson and Tricia McKeever) before use by the same two reviewers for the remaining studies, with review of scoring and final agreement reached by consensus discussion. Data extracted included region and time period of study, source of data, condition studied, case definition, age of cases, exclusion criteria and incidence and mortality figures as provided. An assessment of methodological quality was made using a scoring system developed by consensus based on previous tools [10–12] (table s1 in the online supplementary material).
Meta-analysis of results was contemplated if data were suitably homogenous, but with an expectation that variable methodologies might make descriptive analysis more appropriate.
Results
1934 titles were screened, with selection of 109 abstracts and the full texts of 35 of these were obtained for review (fig. 1). These were supplemented by 13 records identified via additional sources (five from conference proceedings, seven from direct internet searching or reference lists of citations and review articles, one published by our group prior to this review). From this total of 48 studies, 32 were selected for inclusion in the analysis. Two further studies were added after a repeat search. One study in abstract form [13] was replaced with the full version [14] after this was published during the review process of the current work.⇓
Flow diagram of search process.
Incidence of IPF over time according to various studies (countries). Included studies used variable case definitions (see table 1). Where broad and narrow criteria for IPF were reported, narrow criteria have been plotted. Where incidences were reported only as a range over several years, the latest years have been plotted. 95% confidence intervals plotted where provided. For the Raghu (2014) study [2], lowest incidence estimate plotted. For the Agabiti (2014) study [14], most definitive estimate plotted.
The included papers covered 21 countries, with data from 1968 to 2013. Eight studies examined mortality from IPF [1, 9, 15–20] and 28 reported incidence [1–3, 14, 19, 21–43]. Most studies were from Europe and North America (25 studies), with a minority from Asia (five studies) and South America (two studies). Two multi-national studies included data from Oceania. Quality scores varied but eight studies scored full marks, with 29 (85%) out of the 34 studies scoring on at least half of the available criteria.
For clarity, incidence and mortality data were grouped by type of study: large pre-existing databases; local record systems; questionnaire surveys of physicians; and routine mortality statistics. Incidence data for a number of studies have been plotted in figure 2.
Large databases
The most frequent data sources used were pre-existing large databases (13 studies) (table 1). There were four studies from Europe: three from the UK and one from Denmark. All sampled nationwide health databases. The UK studies showed an incidence of 4.6 per 100 000 per year (95% confidence intervals (CI) 4.3–4.9) for 1991–2003 [21], 7.44 per 100 000 per year (95% CI 7.12–7.77) for 2000–2008 [1], and 8.65 per 100 000 per year (95% CI 8.40–8.90) for 2000–2012 [22]. The Danish study, in contrast, reported a decreasing incidence of IPF, with a crude incidence of 7.27 per 100 000 per year (95% CI 6.97–7.57) for 1995–2000, and 5.28 per 100 000 per year (95% CI 5.01–5.56) for 2001–2005 [23].
Incidence of IPF in studies using large databases, local records, and questionnaire surveys
Four studies from North America, all used insurance claims databases. Raghu et al. [2] introduced broad and narrow diagnostic criteria (see table 1), and reported an age-adjusted incidence of IPF, extrapolated to the overall US population, of 16.3 per 100 000 per year (broad criteria) and 6.8 per 100 000 per year (narrow criteria). A later study from Raghu et al. [43] examined Medicare data from 2001–2011 and reported an incidence of IPF of 93.7 per 100 000 per year in people aged 65 years and older, limiting comparison to other studies. This incidence estimate also used a slightly broader case definition. The incidence remained stable over the time period under study. The authors defined broad and narrow subgroups similarly to their previous study [2], and reported lower incidence rates of 31.1–43.0 per 100 000 per year and 15.9–31.1 per 100 000 per year, respectively.
Two other insurance datasets provided less precise incidence estimates from North America. Ehrlich et al. [27] reported the age-adjusted incidence of “pulmonary fibrosis” in diabetics and non-diabetics in California as 14 per 100 000 per year and 9 per 100 000 per year respectively. Saad et al. [28] reported the incidence of interstitial lung disease (but not IPF specifically) in a Canadian cohort, and the authors were able to provide an incidence of IPF specifically for 2006 of 36.6 per 100 000 per year (probable cases) (P. Ernst, McGill University, Montreal, Canada, personal communication, May 2014).
Four studies from East Asia also used insurance and claims databases, and all reported lower incidence rates, ranging from 1.2–4.16 per 100 000 per year [3, 24–26]. Lai et al. [3] noted the severity of IPF was higher and survival lower in Taiwan than in other studies, suggesting milder cases were not being captured. Data from the other three studies were more limited or needed extrapolation but the incidence estimates calculated were all in a similar range. One additional study using large databases was from Brazil, where the incidence of IPF was calculated to be 0.26 per 100 000 per year in 1996, rising to 0.48 per 100 000 per year in 2010 [19].
Local records
Nine studies were classified as using local records to arrive at incidence statistics (table 1). Coultas et al. [29] investigated the incidence of ILD in New Mexico in 1988–1990 with thorough attempts to locate all cases. The crude incidence of IPF was calculated as 10.7 per 100 000 per year in males, and 7.4 per 100 000 per year in females. In a later study, Fernandez Perez et al. [30] investigated the incidence of IPF in Minnesota from 1997 to 2005, again with efforts to identify and verify all cases using international criteria [44, 45]. Overall age- and sex-adjusted incidence in residents aged 50 years and older was 17.4 per 100 000 per year (95% CI 12.4–22.4) (broad criteria), and 8.8 per 100 000 per year (95% CI 5.3–12.4) (narrow criteria). In contrast to UK data, incidence appeared to decrease in later years (2003–2005) but case numbers were low.
Four studies from after the year 2000 examined incidence of IPF in regions of Europe [14, 31, 32, 42]. The incidence ranged from 1.3 per 100 000 per year in Denmark [42], to 7.5 per 100 000 per year (coding criteria) or 9.3 per 100 000 per year (after additional case review) in Italy [14]. Three older European studies examined incidence of IPF using earlier case terminology [33–35], with incidence of cryptogenic fibrosing alveolitis ranging from 0.74–1.28 per 100 000 per year in the Czech Republic from 1984–1998 [34], to 4.3 per 100 000 per year in Norway over 1984–1998 [33].
Questionnaire surveys
Six studies estimated the incidence of IPF across a country by surveying pulmonary physicians (table 1) [36–41]. The highest incidence of IPF was reported in the most recent study, by Musellim et al. [37] from Turkey, where an estimated incidence of 4.69 per 100 000 per year could be calculated and the lowest incidence of IPF was from Flanders, Belgium, from 1992–1996, where Thomeer et al. [41] reported an incidence of IPF of only 0.22 per 100 000 per year.
Routine mortality statistics
Routine mortality statistics were used in eight studies (table 2). Two studies compared mortality data across countries [9, 15], three explored data from the USA [16–18], one looked specifically at the UK [1] and two reported data from Brazil [19, 20]. All studies commented on change over time.
Mortality from idiopathic pulmonary fibrosis (IPF) in studies using routine mortality statistics
Hubbard et al. [15] examined mortality from pulmonary fibrosis in seven countries, predominantly from the 1980s. Crude incidence rates were reported graphically for each country over time. Mortality was highest in the UK (>1 per 100 000) and lowest in Germany and the USA (<0.2 per 100 000). There was an increase in rate ratios over time in most countries, but no change in Germany or New Zealand, and a fall in the USA.
We recently reported more contemporaneous data from ten countries [9]. Using broad codes, age-standardised rates ranged from 4.68 per 100 000 (Sweden) to 13.36 per 100 000 (Northern Ireland), with an increase in all countries over time. For more specific codes (available for selected countries), mortality varied from 4.64 per 100 000 (Spain) to 8.28 per 100 000 (England and Wales). Multiple-cause mortality data (IPF listed anywhere on the death certificate, rather than only underlying cause of death) were available for three countries, and found to be higher, at 12.98 per 100 000 in England and Wales (2010) and 9.37 per 100 000 in the USA. There was less variation between countries in this analysis than previously and, while mortality increased year on year in the UK, multiple-cause mortality data for the USA plateaued from 2003 onwards.
Three studies looked at multiple-cause mortality in the USA using death certificate reports. Age-adjusted mortality increased from 3.2 per 100 000 in 1979, to 3.65 per 100 000 in 1991 [16], with an overall rate of 5.08 per 100 000 for 1992–2003 [17] and 7.57 per 100 000 for 1999–2003 [18].
Navaratnam et al. [1] explored mortality of IPF in the UK from 1968–2008 and found that overall age-standardised mortality was 2.54 per 100 000 (95% CI 2.52–2.56) with a change from 0.92 per 100 000 in 1968–1972 to 5.10 per 100 000 in 2005–2008. The year on year increase in mortality was calculated at 5%.
Two Brazilian studies reported lower levels of mortality from IPF. Rufino et al. [19] noted an increase in mortality from 0.65 per 100 000 in 1996 to 1.21 per 100 000 in 2010 and Fortuna et al. [20] noted an increase in mortality in the southern Brazilian state of Rio Grande do Sul, from 0.22 per 100 000 (1970s), to 0.48 per 100 000 (1990s).
Overall incidence and mortality by geographic region
Most studies came from Europe and North America. In Europe, the highest rates were reported in the UK [1, 22], with a strong increase over time. Lower rates were noted in Scandinavia [23, 33] and southern Europe [36, 38–40], although some of these studies were likely subject to under-reporting, and a more recent study from Italy had a higher incidence [14]. In the USA, mortality statistics were lower and estimates using narrow criteria suggested an incidence of 5–8 per 100 000 [2, 17, 18, 30]. Both incidence and mortality studies from South America suggested a low incidence (0.4–1.2 per 100 000) [19, 20]. Insurance claims-based incidence studies from East Asia also showed a low incidence (1.2–3.8 per 100 000) [3, 24, 25], although routine mortality statistics from Japan suggested a higher incidence (adjusted mortality rate of 10.26 per 100 000 for broad coding). Adjusted mortality statistics from Oceania ranged from 5.08–6.49 per 100 000 [9].
Overall incidence over time
The majority of studies reporting temporal trends in incidence of IPF showed an increase over time. Studies from the 1980s tended to have lower rates [15, 16, 34, 35], while later studies using similar data showed far higher rates [1, 17]. Increasing incidence rates were particularly evident in UK datasets [1, 21], but also noted in South America [19, 20], East Asia [3] and Europe [34]. However, mortality data from the USA appeared to plateau in some studies, and a decline was noted in studies from the USA [15, 30] and Denmark [23].
Summary statistics
Due to variation in study methodology, lack of confidence intervals for most studies, and differing time periods, formal meta-analysis to derive summary statistics was not possible. Attempts using those studies with confidence intervals produced a very high I2 statistic of >98%, suggesting extremely high heterogeneity (values >75% considered “high”) [47], and this was also the case when we created roughly estimated confidence intervals from available raw data from other incidence studies.
The overall range of incidence statistics varied from 0.22–93.7 per 100 000 per year. In an attempt to deal with potential outliers and describe an estimate applicable to Europe and North America, we excluded studies from Asia and South America (different populations), and also questionnaire surveys with likely underreporting. We excluded older studies (with data prior to year 2000) and used narrow (rather than broad) criteria to limit over-diagnosis. This yielded a range of 2.8–9.3 per 100 000 per year, as an estimate of IPF incidence.
Discussion
This review summarises 34 studies of IPF incidence and mortality, and draws together different types of work from across the world. Varying study methodologies, time periods and case definitions makes summary statistics difficult, but incidence ranges from 0.2 per 100 000 per year to 93.7 per 100 000 per year, with a tighter range of 3–9 per 100 000 per year based on conservative estimates from Europe and North America. Incidence rates increased over time in most countries, and appear to be coalescing worldwide, but seem to be lower in Asia and South America. Current data suggest the incidence of IPF worldwide is comparable to that of several malignancies, including stomach, liver, testicular and cervical cancers [48, 49].
Different study designs have different strengths and weaknesses. Large dataset studies and routine mortality statistics benefitted from large numbers of patients, but at the expense of clinical verification of diagnoses, with potential for misclassification. Some databases were also not representative of the underlying population. Mortality studies rated highly on quality scoring and allowed comparison across countries, but a major limitation was that IPF might not be the underlying cause of death, or may have been misdiagnosed in life. Many countries only report the most common respiratory causes of death, such as pneumonia or chronic obstructive pulmonary disease (COPD), and full International Classification of Diseases 10th Revision (ICD-10) codes are used infrequently, hence the broad codes used to identify cases in some of our countries will almost certainly classify other diseases as IPF. Whether this is counterbalanced by underreporting on death certificates is unclear.
Local records studies covered smaller geographical regions, but where possible, diagnoses were verified by review of clinical records, in some cases with external review using international diagnostic criteria. While potentially more accurate, this approach limits the size of the population under review. The most detailed assessment was probably by Fernandez Perez et al. [30], but this identified only 24 cases of narrowly defined IPF over 8 years. Incidence statistics from these studies may also be difficult to apply to the wider population. Quality assessment varied considerably due to limited information regarding methodology and case verification in some studies.
The lower incidence found in questionnaire surveys undoubtedly reflects inadequate reporting of cases from participating centres. The highest level was reported most recently, in Turkey in 2013 [37]; this may reflect increased effectiveness of questionnaire surveys in the internet age, where electronic registration and widespread awareness of international guidelines may enhance uptake and participation. Despite criticism of these studies, later questionnaires have provided greater detail on subdivisions of idiopathic interstitial pneumonia (such as non-specific interstitial pneumonia and cryptogenic organising pneumonia [13]) than has been possible using ICD-10 coding, and therefore give some idea of the proportion of cases of IPF that might be over-diagnosed using routine coding studies.
The ideal incidence study would sample a large dataset, but with an attempt at validating clinical diagnoses by review of records. The study by Agabiti et al. [14] from Lazio, Italy, followed this approach in a region of 4.7 million people with sampling from six hospitals, and also highlighted additional cases with review of less specific codes. Alternatively, with greater acceptance of international guidelines, more widespread use of imaging technology and greater education of clinicians, there may be less uncertainty regarding diagnosis of interstitial lung disease. Part of the difficulty assessing epidemiological studies in IPF results from the varying classification methods used, which have altered over time, making ILD less robust a diagnosis than conditions such as breast cancer or myocardial infarction. Consolidating international diagnostic criteria, as has happened with COPD, should help to address this.
In most studies, the incidence of IPF appears to be increasing over time, although two good quality studies in Denmark [23] and the USA [30] showed a decrease. Low patient numbers may limit the reliability of the observed decline demonstrated in the US study, and in the Danish study there was a possibility that prevalent cases may have been included in the earlier time period, with more cases of “other” interstitial lung disease in the later time period suggesting diagnostic transfer. The increasing incidence seen in UK studies seems unlikely to be purely due to coding issues, and more needs to be done to assess the reasons behind international variation in incidence.
There are several explanations why incidence and mortality of IPF may vary across countries. The lower incidence in South America may be due to under-diagnosis or under-reporting on death certificates. Both studies here used routine data, and the level of industrialisation in Brazil means that other diseases may have more of a focus in healthcare terms. In East Asia, the higher severity of disease in study subjects from insurance datasets likely reflects exclusion of milder cases, and may explain the lower incidence than in western countries. The higher mortality data from Japan in our study of death certification data may give weight to this [9]. However, the coding used here was broad, and sub-classification suggested the majority of cases were recorded as “unspecified interstitial lung disease” rather than “IPF”, which might imply a different spectrum of interstitial lung disease in East Asia.
In this review, we have attempted to include all incidence and mortality studies, including those presented purely as conference abstracts. We placed no language restrictions on our search strategy; however, it is likely that some studies in other languages will have been missed by using English search terms. By focussing solely on population-based studies, we will have excluded a number of clinical case series from tertiary centres in more diverse areas of the world that may have provided indications of incidence, but it was considered that using these reports would require too many assumptions to be reliable. The fact that we included studies with differing methodologies and time periods did limit our ability to pool incidence statistics, but we felt our scope was appropriate.
Overall, available data is relatively consistent with regards the incidence and mortality of IPF worldwide. Most variation is likely a result of heterogeneity in study design, although there are trends that warrant further investigation, such as the apparent reduced incidence in East Asia and the contrast between incidence increasing in the UK and plateauing in the USA. Further studies should ideally be designed to allow appropriate comparisons across countries and we have proposed recommendations for future work (accepting the limitations researchers may face achieving these) (table s2 in the online supplementary material).
In summary, we have comprehensively searched for available data on incidence and mortality from idiopathic pulmonary fibrosis worldwide. Although different study methodologies limit comparisons, incidence does appear to be increasing in most regions worldwide, and rates are coming together. The variation between countries across different studies may reflect a transition in IPF incidence across the World from high to low incidence, and understanding why this is so may provide useful insights into the cause or causes of the disease.
Footnotes
Editorial comment in: Eur Respir J 2015; 46: 604-606 [DOI: 10.1183/13993003.00958-2015]
This review is registered at PROSPERO (www.crd.york.ac.uk/PROSPERO/) with identifier number CRD42014007452.
This article has supplementary material available from erj.ersjournals.com
Conflict of interest: Disclosures can be found alongside the online version of this article at erj.ersjournals.com
- Received October 6, 2014.
- Accepted March 19, 2015.
- Copyright ©ERS 2015
References
