Discussion
In this review, we synthesised the evidence for the incidence and prevalence of ILDs from studies published between 2015 and 2021. Considering the changing ILD nomenclature and the desire to reflect more current estimates, in this review, we decided to restrict the study period to past 6 years. We took this conscious effort with the aim to limit the heterogeneity across reported estimates. We evaluated 39 incidence and 78 prevalence estimates for individual ILD disorders that were distributed globally. We noted an increase in the number of studies investigating non-IPF ILDs and more specifically autoimmune ILDs in recent years. There was a 6-fold rise in the autoimmune ILDs studies, in 2021 when compared with 2015 (18 vs 3 studies, respectively). This increase in non-IPF ILD studies may be related to the emergence of antifibrotic therapies for non-IPF fibrosing lung diseases.91–93 Interestingly, the publication trend for IPF has remained unchanged.
This review revealed considerable inconsistencies in the incidence and prevalence estimated of the main ILD subgroups. The reported prevalence of IPF ranged from 7 to 1650 per 100 000 persons,2 16 an approximately 800-fold difference across case definitions, despite most studies reporting IPF prevalence in the general population. The incidence and prevalence estimates reported by Zhang et al16 were a notable outlier; this study was based on the USA veterans’ healthcare database which included mostly White patients aged over 70 years—the demographic in which IPF is most common. Aside from this study, the majority of studies reported a prevalence of IPF ranging from 7 to 42 per 100 000 persons across different case definitions.2 17
Unlike prevalence, we found considerable inconsistencies in how the incidence of IPF is reported. An important factor is the lack of uniformity in reporting units. Half of the studies reported incidence using person-years, whereas others reported per 100 000 person-years. We were, therefore, unable to compare incidence estimates in a similar fashion to prevalence. It is also important to note that changes in diagnostic guidelines for IPF over the years may have made it more challenging to accurately estimate its burden and temporal trends.94–96
For non-IPF subgroups, such as autoimmune ILDs, there were wide variations in prevalence estimates between countries and within different healthcare settings in the same country. Overall, the variation in prevalence and incidence estimates was even greater for non-IPF ILDs than IPF. This can be attributed to several factors. First, in clinical practice, it is common for the clinical presentation and serological autoantibody profiles to result in overlap syndromes. Autoimmune conditions can coexist and patients with occupational ILDs may also have autoimmune conditions. Such fluidity of diagnoses at a clinical level reflects the challenges in estimating non-IPF ILDs. Second, the denominator more frequently differs for non-IPF ILDs, resulting in lack of standardised reporting. Unlike IPF, for which there are published validated algorithms to identify ‘true’ cases in the general population.18 24 97 For non-IPF ILDs, studies relied on disease registries or were conducted at single/multispecialist clinics.
Majority of the autoimmune-related ILD estimates were in RA and SSc ILD. When assessing SSc ILD prevalence, we observed a wide range (26.1% to 88.1%)37 44 in reported estimates, but when studies were dichotomised into single-centre studies and multicentre studies, it became clear that the highest variability was contributed by single centre studies (SSc prevalence, 31.2%–88.1%).43–46 Owing to a smaller number of studies reporting incidence, we were unable to observe whether the same challenge existed.
The prevalence of silicosis ranged from 5.6%65 to 37%61 in workers exposed to silica. Occupational ILD studies were conducted at a factory, in a neighbourhood with proximity to industries, a registry or multicentre settings. Therefore, lack of generalisability and applicability of findings only to certain populations contributed largely to the wide variabilities of these reported estimates. The geographical distribution of occupational ILD papers alludes to dominance of exposure related ILDs in low-income and middle-income countries in Asia and South America (42.8% were in Asia).
While historical diagnostic classification has been founded on underlying aetiology or clinical pathways, there is now a growing emphasis on disease behaviour.98 99 Attention has focused on a subgroup of ILD patients who go on to develop a PF phenotype. IPF is the archetypal PF ILD but other ILDs such as chronic hypersensitivity pneumonitis (HP), SSc ILD can exhibit ‘IPF-like’ behaviour, including rapid decline in lung function and early mortality.100 The epidemiology of PF ILD is particularly challenging to examine as accepted guidelines on definition and diagnosis have yet to be published The reported prevalence of PF ILDs (per 100 000 persons) was 19.4 in France and 57.8 in the USA.88 89 The future direction of research will likely focus on PF ILD as a phenotype which transcends previously adhered-to diagnostic labels and is associated with poorer outcomes and increased mortality.100 101
Among the 39 studies reporting ILD incidence (online supplemental figure E6), most studies were categorised as medium risk (n=25/39, 64.1%). Two studies were categorised as high-risk primarily because of lack of information on ILD diagnosis and poor quality of reporting estimates (ie, descriptive statistics were not reported, were incomplete or did not include proper measures of dispersion).
Similarly, there were 78 prevalence assessments (online supplemental figure E7) of which approximately 18% (n=14/78) were categorised as high risk, 64.1% (n=50/78) as medium risk and 18% (n=14/76) as low risk. Most studies assessed as high risk were studies reporting autoimmune ILDs, mainly because of ILD diagnosis, single-centre studies or small sample size. Most of the studies reporting prevalence based on large healthcare datasets or disease registries were classified as low risk.
There are several strengths of this systematic review. We have provided an assessment of the incidence and prevalence of several ILD conditions globally and have grouped ILDs based on their aetiology to allow the appraisal of incidence and/prevalence at a disease level with as much granularity as possible. This review underlines the need for standardisation of diagnostic classifications for non-IPF ILDs—the narrower estimates for IPF provide the evidence that clear and consistent diagnostic guidelines are of great clinical utility. Guidelines have recently emerged for the diagnosis of HP102 103 which we envisage will further improve the epidemiological reporting of this important condition, although incorporation of guidelines into routine clinical practice and then into epidemiological estimates takes time. Cross-specialty guideline groups will undoubtedly improve standardisation of reporting for autoimmune driven ILDs.
It is possible that genetic differences between individuals from different ethnic backgrounds may play a role in the global variability in incidence and prevalence. For example, the MUC5B promoter polymorphism (rs35705950) is the dominant risk factor for IPF104 and is also a key risk factor for other ILDs such as RA.105 This gain of function polymorphism is frequent in those of European decent but almost completely absent in those of African ancestry.106 As more research is performed unravelling the complex interplay between genetics and environment in the development of ILD, it is likely that genetic variability will be found to play an important role in the global variability of ILD.
Despite the strengths, there are limitations to this systematic review. The certainty of the ILD case definition varied across studies. It was not always possible to be sure of how reliable the ascertainment method was. However, we attempted to reflect the differences in the ILD diagnostic methods in our risk of bias quality assessment. Along with the uncertainty in the diagnosis of ILD, there were different disease definitions used across studies. Therefore, in this review due to high heterogeneity, in how ILD was defined, we were unable to perform a meta-analysis. In this review, we have only included studies reporting ILD estimates in general populations, registries or populations with a specific disorder of interest. For single-centre studies reporting incidence and/or prevalence of autoimmune or exposure ILDs, the estimates were not generalisable and this has been reflected in the risk of bias quality assessment score. This review is limited to English publications only. However, due to high volume of papers found with the study period, we are confident it has a minimal effect on the overall conclusion.107