Prevalence, Risk Factors, and Outcomes of Adult Interstitial Lung Abnormalities: A Systematic Review and Meta-Analysis

Abstract

Rationale

Incidental parenchymal abnormalities detected on chest computed tomography scans are termed interstitial lung abnormalities (ILAs). ILAs may represent early interstitial lung disease (ILD) and are associated with an increased risk of progressive fibrosis and mortality. The prevalence of ILAs is unknown, with heterogeneity across study populations.

Objectives

Estimate the pooled prevalence of ILAs in lung cancer screening, general population-based, and at-risk familial cohorts using meta-analysis; identify variables associated with ILA risk; and characterize ILA-associated mortality.

Methods

The study protocol was registered on PROSPERO (CRD42022373203), and Meta-analyses of Observational Studies in Epidemiology recommendations were followed. Relevant studies were searched on Embase and Medline. Study titles were screened and abstracts reviewed for full-text eligibility. Random effect models were used to pool prevalence estimates for specified subgroups and ILA-associated mortality risk. Risk of ILAs was estimated based on age, sex, and FVC. Quality assessment was conducted using an adapted Assessment Tool for Prevalence Studies.

Measurements and Main Results

The search identified 9,536 studies, with 22 included, comprising 88,325 participants. The pooled ILA prevalence was 7% (95% confidence interval [CI], 0.01–0.13) in lung cancer screening, 7% (95% CI, 0.04–0.10) in general population, and 26% (95% CI, 0.20–0.32) in familial cohorts. Pooled mortality risk was increased in those with ILAs (odds ratio, 3.56; 95% CI, 2.19–5.81). Older age, male sex, and lower FVC% were associated with greater odds of ILA.

Conclusions

Populations undergoing imaging for non-ILD indications demonstrate high ILA prevalence. Standardized reporting and follow-up of ILAs is needed, including defining those at greatest risk of progression to ILD.

At a Glance Commentary

Scientific Knowledge on the Subject

Interstitial lung abnormalities (ILAs) are incidental parenchymal findings detected on chest imaging performed for indications other than for interstitial lung disease (ILD). The prevalence of ILAs in lung cancer screening studies, observational general population cohorts, or subclinical ILD in at-risk familial cohorts is unknown.

What This Study Adds to the Field

This systematic review and meta-analysis reports a pooled ILA prevalence of 7% in lung cancer screening studies, 7% in studies of the general population, and 26% in at-risk familial cohorts. Across these studies, male sex, older age, and lower FVC% predicted were associated with an increased risk of ILA. The presence of ILA was also associated with increased odds of death. ILA is prevalent across these groups and associated with increased mortality risk.

Interstitial lung disease (ILD) comprises a large and heterogeneous group of pulmonary parenchymal disorders associated with lung inflammation and/or fibrosis. Chest computed tomography (CT) remains the key imaging modality to establish an ILD diagnosis, often informing a patient’s prognosis and management (1–3). Patients may be diagnosed at symptom onset with abnormal pulmonary function tests (PFTs) or radiographic evidence of fibrosis (4), and this can occur in late stages of the disease. Better methods to facilitate earlier diagnosis and initiation of treatment for ILD are needed to prevent or slow the progression of otherwise irreversible fibrosis (5).

The natural history of ILD is heterogeneous, depending on disease extent at diagnosis and diagnostic subtype, among other factors. Parenchymal abnormalities suggestive of ILD can be detected incidentally on chest CT scans ordered for other reasons. These incidental parenchymal abnormalities are termed interstitial lung abnormalities (ILAs) and may represent early, preclinical forms of ILD, or previously unrecognized disease, in some cases (6). Some patterns of ILA have been shown to progress to clinical ILD over time (7) and are associated with an increased risk of all-cause mortality (8). The detection of ILAs, therefore, presents an opportunity for risk factor modification and studies of early treatment interventions for patients early in their disease state, before the development of advanced fibrosis.

Lung cancer screening (LCS) programs aim to detect early-stage cancer in at-risk populations, and their importance has been demonstrated in several landmark trials (9–11). Incidental findings, such as ILAs, have been reported in up to 14% of individuals undergoing LCS (12). These cohorts offer a unique opportunity to capture patients at risk for ILD, given shared risk factors (smoking, older age) and the known association between ILD and an increased risk of lung cancer (13). Incidental findings on chest CT, including ILAs, have also been reported with similarly high frequency in general population–based cohort studies of middle aged and older adults undergoing CT imaging (14). At-risk relatives of patients with sporadic and familial pulmonary fibrosis are not considered to have ILAs according to the recent Fleischner position statement, given their increased baseline risk as family members of those affected by ILD (6). Such findings in undiagnosed family members have been referred to as preclinical or subclinical ILD (15). Several screening studies have been completed to understand the prevalence and risk of progression to ILD in this particular group (16, 17). Epidemiologic data to define the prevalence of ILAs across these three distinct groups are needed to ensure a standardized and equitable approach to diagnostic evaluation and follow-up and to inform the management of a potentially large volume of individuals with ILAs as chest imaging screening programs are implemented.

The objective of this study was to estimate the pooled prevalence of ILAs in LCS, general population, and at-risk familial cohorts, using meta-analysis. We further sought to identify risk factors for ILAs across these cohorts and estimate the pooled risk of mortality associated with ILAs.

Methods

Study Criteria and Selection

Studies involving adults ⩾18 years of age undergoing chest CT were eligible for inclusion. Radiographic studies must have been for the purpose of LCS, screening at-risk family members for ILD, or population-based cohorts that included a chest CT for indications other than identifying ILD. The prevalence of ILAs, as defined in the 2020 Fleischner position paper (6), “subclinical ILD,” “preclinical ILD,” “early ILD,” or a variation thereof had to be reported, and publications must have been in the English language. For simplicity, all related terms are referred to as ILAs within this manuscript, with the exception of the familial cohort, which is termed “subclinical ILD.” Review articles, case reports, studies reporting on fewer than 10 subjects, and conference abstracts were excluded. If multiple studies were published from the same cohort, the most recent or largest cohort was included, with the largest cohort prioritized.

Data Search and Extraction

The study protocol was registered on PROSPERO (CRD42022373203), and Meta-Analyses of Observational Studies in Epidemiology recommendations were followed (see Table E1 in the online supplement) (18). Relevant studies were searched on Embase and Medline databases from inception date until May 11, 2023 using a search strategy including keywords, free text, and MeSH terms (Table E2). Full references were imported into Covidence, with duplicates removed. Two reviewers (A.G.-O. and B.M.) independently screened study titles and abstracts for full-text eligibility. Full texts of eligible studies were reviewed independently by the two reviewers, with conflicts resolved by a third reviewer (K.A.J.). Data were extracted from full-text articles and supplemental material. Variables extracted, where available, included age, sex, race/ethnicity, definition of ILA used, prevalence and radiographic pattern of ILAs, incident lung cancer diagnosis, occupation and exposure history, smoking history (pack-years), concomitant chronic obstructive pulmonary disease diagnosis, family history of ILD or lung cancer, PFTs (FVC%, Dl_CO%), MUC5B single-nucleotide polymorphisms genotype, occurrence of ILA progression, and risk of all-cause mortality. Participants with an indeterminate pattern for ILAs were considered to not have ILAs for the purpose of the prevalence analysis, to minimize classification that might overestimate prevalence. When a study reported both “no ILA” and “indeterminate ILA” data, the meta-analysis for “no ILA” demographic data is reported as study A and “indeterminate for ILA” demographic data as study B. Corresponding authors were contacted to request missing demographic information, which was included when provided.

Risk of Bias Assessment

Two reviewers (A.G.-O. and B.M.) independently assessed the methodological quality using relevant components of an adapted Assessment Tool for Prevalence Studies (19). This tool reviews two domains, external and internal validity. Sections of each domain were reported as “yes” or “no.” The risk of bias of the overall domain was judged to be low, moderate, or high. The overall study was judged as low risk if both domains had a low risk of bias, with some concern of bias if there were moderate concerns within one domain, and high risk of bias if any domain was deemed to be at high risk of bias. Consensus with a third reviewer (K.A.J.) was used to resolve disagreements. Publication bias was not assessed.

Statistical Analysis

Study data were reported with either means and SDs or medians and ranges to describe continuous variables. Pooled prevalence of ILAs was estimated with a 95% confidence interval (CI) for each cohort (LCS, general population cohort, familial cohort), using a random effects model. The random effects models reported were generated using the method of DerSimonian and Laird (20), with the estimate of heterogeneity being taken from the inverse-variance fixed-effect model (21). Risk factors were determined using meta-regression, based on their associations with ILA. Age and FVC% predicted were reported as the mean difference between those with and without ILAs. Sex was reported as a risk ratio with 95% CI. When reported, study participant mortality was recorded as a binary variable, and the prognostic impact of ILAs on mortality was reported as an odds ratio (95% CI) using a random effects model (21). Between-study heterogeneity was analyzed using visual inspection of forest plots, calculation of Cochran’s chi-square test, and the I² statistic. P values < 0.05 were considered statistically significant. Data analysis was conducted using RevMan (Review Manager [RevMan] version 5.4; The Cochrane Collaboration, 2020) and STATA (StataCorp. 2021; Stata Statistical Software: release 17).

Results

Search Results

The initial search identified 9,536 studies, and 7,851 abstracts were reviewed after removal of duplicates (Figure 1). Of these, 87 studies underwent full-text review, and 35 studies were included in the systematic review. Several publications used the same cohort. Therefore, data were extracted from a total of 22 studies using the largest cohort or most recent publication (Table 1) (8, 14, 16, 17, 22–40). A total of 88,325 participants were included in the pooled prevalence analysis.

Figure 1.

Flow diagram of search progress, as per PRISMA guidelines.

Table 1.

Study Characteristics

First Author, Year (Reference)	Study Design	Country	Recruitment Period, Follow-Up Duration	Number of Participants	Population	ILD History Specifically Excluded	ILA Definition	CT Modality	ILA Frequency and Type [% (n)]
Lung cancer screening cohort
Hoyer, 2018 (22)	Secondary analysis of DLCST	Denmark	2004 to 2006, follow-up until 2016	1,990	Age 50–70 yr, 20 pack-year smoking history, former smokers quit after the age of 50 yr and within previous 10 yr, baseline FEV1 ⩾ 30%, climb 2 flights of stairs (36 steps)	No	Presence of centrilobular, pleural, paraseptal, GG attenuation, reticulation, or HC	LDCT, 120 kV and 40 mAs, thin slice reconstruction	16.7 (332) 115 GGO 155 nodular pattern 162 reticulation 12 honeycomb
Whittaker Brown, 2019 (23)	Secondary analysis of NLST	United States	April 2002 to 2004, median follow-up 6.6 yr (IQR, 6.2–6.9 yr)	25,041	Age 55–74 yr, 30 pack-year smoking history, former must have quit within 15 yr	Yes	Report includes reticular/reticulonodular opacities, HC, fibrosis, or scarring	LDCT	20.2 (5,053)
Sverzellati, 2011 (24)	Secondary analysis of MILD	Italy	September 2005 to September 2006, 3-yr follow-up	700	Age ⩾ 49 yr, current or former smoker (quit <10 yr), ⩾20 pack-year smoking history and no cancer history within 5 yr	No	Presence of UIP-like, OCIP-like	LDCT, 120 kV, 30 mAs, 0.75 mm collimation, 1 mm reconstructed thick sections	4.0 (28)
Lee, 2018 (25)	K-LUCAS pilot study secondary analysis	Korea	November 16 to March 2017	256	Asymptomatic current or ex-smokers aged 55–74 yr with a smoking history ⩾30 pack-years who had used tobacco within the last 15 yr	Unsure “those receiving treatment for ILD excluded”	None provided	CT chest thick ⩽3 mm and thin ⩽1.25 mm, dose < 3.0 mGy	1.2 (3)* 3 UIP
Mackintosh, 2019 (26)	Secondary analysis of Queensland Lung Cancer Screening Study	New Zealand	2007 to 2009, 5-yr follow-up	256	60–74 yr, current or <15-yr former smoker, ⩾30 pack-year history, FEV1 ⩾ 50%	Unsure “medical comorbidity or poor lung function excluded”	Report includes reticulation, ground glass, HC, consolidation, mosaicism, traction bronchiectasis, nodularity, cysts, and fibrosis	LDCT, 64 multidetector thin-section, 120 kV, 30 mA, 350 mm, 3.7-s scan time	7.8 (20) 9 reticulation 18 ground-glass changes 1 traction bronchiectasis
Salvatore, 2016 (27)	Secondary analysis of Mount Sinai lung cancer screening database	United States	2010 to 2014	951	40–85 yr or older at enrollment time	No	Presence of traction bronchiectasis, GGO with traction bronchiectasis, reticulations with TB, HC. If any findings present, image reviewed by all 3 radiologists and 2/3 must agree to confirm ILD presence	LDCT, 120 kVp, ⩽80 mAs	6.6 (63) 47 pre-HC 16 HC 52 peripheral 11 central 62 bilateral 1 unilateral
Hewitt, 2022 (28)	Secondary analysis of pilot lung cancer screening cohort	England	August 2018 to April 2021	1,853	Age 55–75 yr, ever-smokers, PLCOm2012 ⩾ 1.51% over 6 yr or LLPv2 5-year risk ⩾ 2.5%.	No	ILA > 5% extent of CT	LDCT 16 channel, total dose < 2 mSv, thickness ⩽1.25 mm	4.2 (78) 59 referred to ILD center, 43 assessed: 15 ILA 13 IPF 5 SR-ILD 4 HP 2 PPFE 1 sarcoid 1 LCH, 1 post-COVID 1 unclassifiable
Balata, 2023 (38)	Secondary analysis of pilot lung cancer screening cohort	England	June 2017 to October 2016	1,384	Ever-smokers aged 55–74 yr with PLCOm2012 score ⩾ 1.51%	No	ILA > 5% extent of CT	LDCT, total dose <3.0 mSv, thickness 1.25 mm	3.9 (54) 31 subpleural nonfibrotic 20 subpleural fibrotic 3 nonsubpleural nonfibrotic
Patel, 2023 (39)	Secondary analysis of LCS cohort	United States	January 2012 to September 2014, follow-up until September 2019	1,699	National Comprehensive Cancer Network Guidelines	Yes	ILA > 5% extent of CT	LDCT, 64 multidetector thin-section, 100 kV, 30–100 mA, 1.25–1.5 mm reconstructed	2.4 (41) 2 typical UIP 2 probable UIP 23 indeterminate UIP 14 alternative
Stuart, 2023 (40)	Secondary analysis of LCS cohort	Canada	April 2015 to December 2019	806	Age 55–79 yr with a PLCOm2012 risk score ⩾ 1.5% or with ⩾30 pack-year smoking history and having quit ⩽15 yr prior	No	Baseline CT screened for words “ILD”, “fibrosis”, “reticulation”, “traction bronchiectasis”, “HC”, “GGO”, or terms NSIP or UIP. CT reports re-reviewed and classified as non-subpleural, subpleural nonfibrotic, and subpleural fibrotic	LDCT	3.7 (30) 5 nonsubpleural 5 subpleural nonfibrotic 20 subpleural fibrotic
General population cohort
Putman, 2016 (8); original FHS ILA cohort: Araki, 2016 (14)	Secondary analysis of FHS MDCT2	United States	2008 to 2011, follow-up until December 2015; median, 4.0 yr	2,633	Men ⩾ 35 yr, women ⩾ 40 yr, nonpregnant, weight < 350 pounds	No	Presence of nondependent changes affecting >5% of any lung zone, including any combination of nondependent ground-glass or reticular abnormalities, diffuse centrilobular nodularity, nonemphysematous cysts, honeycombing, or traction bronchiectasis	Noncontrast thoracic CT, 0.63 mm	6.7 (177)
Axelsson, 2020 (29); (A) no ILA; (B) indeterminate Prevalence and mortality data: Putman, 2019 (30)	Secondary analysis of AGES-Reykjavik	Iceland	2002 to 2006, follow-up until August 2016	3,167; additional 2,153 participants included in mortality analyses (n = 5,320)	Randomly recruited surviving participants of Reykjavik study born between 1907 and 1935	No	Presence of nondependent ground-glass or reticular abnormalities, diffuse centrilobular nodularity, nonemphysematous cysts, honeycombing and traction bronchiectasis that affected >5% of any lung zone	Cardiac CT	7.0 (375)
Podolanczuk, 2016 (31)	Secondary analysis of MESA	United States	2000 to 2002, follow-up 2012	2,430	Adult 45–84 yr of age free of clinical cardiovascular disease at baseline from 6 communities	No	Presence of ground-glass, reticular abnormality, diffuse centrilobular nodularity, honeycombing, traction bronchiectasis, nonemphysematous cysts, or architectural distortion in ⩾5% of nondependent portions of the lung	Noncontrast cardiac CT, 64-slice, reconstructed 0.625 mm	12.6 (306)
Tsushima, 2010 (32)	Retrospective review of Nagano Japan health screening program	Japan	January 2004 to December 2004	3,061	Adults concerned about their health	No	Presence of HC, interlobular septal thickening, GGO, ill-defined subpleural line, combined pulmonary fibrosis, and emphysema	LDCT. If interstitial changes present, recruited for further testing, including HRCT 150 mA, 120 kVp, 1 mm collimation, prone imaging	2.6 (80) 35 interlobular septal thickening 15 ill-defined subpleural line 9 GGO 7 HC 14 CPFE
Buendía-Roldán, 2021 (33)	Lung aging program registry retrospective review	Mexico	2015 to 2019, follow-up 24 ± 18 mo	817 (564 normal scans, 173 other identified abnormalities, 80 ILAs)	Asymptomatic volunteers > 60 yr of age	No	Presence of GGO, reticular abnormalities, diffuse centrilobular nodules, HC, traction bronchiectasis, nonemphysematous cysts, or architectural distortion involving ⩾5% of nondependent portions of the lung	Supine and prone HRCT	9.8 (80) 38 subpleural nonfibrotic 24 subpleural fibrotic 18 subpleural nonfibrotic
Lee, 2023 (34)	Retrospective review of health screening cohort	Korea	2007 to 2010, follow-up until 2021	2,765	Age > 50 yr	No	Presence of any nondependent abnormality affecting ⩾5% of any lung zone, including ground-glass or reticular abnormalities, architectural distortion, traction bronchiectasis, nonemphysematous cysts, and HC	Noncontrast CT chest thickness 1–5 mm; 45% had prone images.	3.4 (94) 35 nonfibrotic 59 fibrotic
At-risk familial cohort
Salisbury, 2020 (35); mortality data: Steele, 2023 (36)	Retrospective review of prospective registry	United States	2008 to 2019, follow-up until 2019	336 participants, 157 kindred; follow-up, 493 participants	Siblings or offspring of individuals affected by ILD and members of families in which at least two relatives had ILD, at least one of whom was diagnosed with idiopathic pulmonary fibrosis. Eligible individuals were ⩾40 yr of age or within 10 yr of the age of diagnosis of the youngest affected family member	Yes	ILA categorized as none, early/mild, or extensive on the basis of the presence and extent of specific interstitial features, including GGO, intralobular reticular opacities, irregular thickening of interlobular septa, traction bronchiectasis, traction bronchiolectasis, and HC. ILA extensive >5% HC ⩾2 zones, of which >30% of total lung parenchyma had interstitial features	HRCT	22.9 (77) 74 early/mild 3 extensive
Rosas, 2007 (17)	Cross-sectional study	United States	Not provided	143 participants, 18 kindred	Age ⩾ 18 yr without familial IPF with two or more first-degree relatives with either open lung biopsy showing UIP and/or HRCT consistent with IPF as per ATS guidelines	Yes	Presence of reticular opacities, septal lines, reticulation, as defined by Avila et al. 2002 (41)	Non-contrast HRCT, end-inspiration prone	21.7 (31)
Hunninghake, 2020 (16)	Prospective registry	United States	November 2016 to September 2019	105 participants, 53 kindred	First-degree relatives, aged 45–85 yr and without known ILD, of patients without known connective tissue disease or sarcoidosis	Yes	Changes affecting >5% of any lung zone, including nondependent ground-glass or reticular abnormalities, diffuse centrilobular nodularity, nonemphysematous cysts, HC, or traction bronchiectasis	Single prone CT at full inspiration	31 (33) 29 ILD†
Lucas, 2023 (37)	First-degree relatives recruited from prospective Australian registry and from clinics	Australia	September 2019 to November 2019	15 participants, 12 kindred	Relatives ⩾50 yr old, except if ILD case diagnosed before age 50 then included if at least as old as the youngest diagnosed case	Yes	Presence of incidental findings >5% of any lung zone and potentially compatible with ILD (i.e., GGO or reticular abnormalities, traction bronchiectasis, HC, and nonemphysematous cysts)	HRCT	46.7 (7) 6 interlobular reticular opacities 4 GGO 1 HC 3 lung distortion 3 traction bronchiectasis 3 traction bronchiolectasis

Definition of abbreviations: AGES = Age, Gene/Environment Susceptibility; ATS = American Thoracic Society; CPFE = combined pulmonary fibrosis and emphysema; CT = computed tomography; DLCST = Danish Lung Cancer Screening Trial; FHS = Framingham Heart Study; GGO =ground-glass opacities; HC = honeycombing; HP = hypersensitivity pneumonitis; HRCT = high-resolution computed tomography; ILA = interstitial lung abnormalities; ILD = interstitial lung disease; IPF = idiopathic pulmonary fibrosis; IQR = interquartile range; K-LUCAS = Korean Lung Cancer Screening Project; LDCT = low-dose computed tomography; LLPv2 = Liverpool Lung Project model; MDCT2 = multidetector computed tomography; MESA = Multi-Ethnic Study of Atherosclerosis; MILD = Multicentric Italian Lung Detection trial; NSIP = nonspecific interstitial; NSLT = National Lung Screening Trial; OCIP = other chronic interstitial pneumonia; PLCO = Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial model; PPFE = pleuroparenchymal fibroelastosis; SR-ILD = smoking related-interstitial lung disease; TB = traction bronchiectasis; UIP = usual interstitial pneumonia.

^*Reported as other CT findings of ILD with a UIP pattern.

^† ILD defined as total lung capacity < 80% predicted or Dl_CO < 70% predicted.

Study Characteristics

A total of 10 studies and 30,909 participants were included from LCS cohorts. Studies recruited participants between 2002 and 2021, and median duration of follow-up was 5 years (range, 3–10 yr). Two studies reported all-cause mortality for those with and without ILAs on their initial screening CT. A total of eight studies and 56,817 participants were included in the general population cohort. Studies recruited participants between 2000 and 2021, with a median follow-up duration of 7.5 years (range, 2–11 yr). A total of five studies and 599 participants from 240 kindred were included in the at-risk familial cohort. Studies recruited participants between 2008 and 2019, with limited follow-up data. Study populations are presented in Table 1.

ILA Prevalence

In the LCS cohort, ILA prevalence was 7% (95% CI, 0.01–0.13), although it was 7% (95% CI, 0.04–0.10) in the general population cohort. Subclinical ILD prevalence was 26% (95% CI, 0.20–0.32) in the familial cohort (Figure 2). Variance due to heterogeneity was substantial in the overall cohort (I² = 99.5%; P < 0.0001), the LCS cohort (I² = 99.6%; P < 0.0001), and the general population cohort (I² = 98.3%; P < 0.0001), whereas it was lower in the familial cohort (I² = 53.4%; P = 0.09). The overall pooled prevalence for all cohorts was 10% (95% CI, 0.07–0.13; I² = 99.5%) (Figure E1). Sensitivity analysis was conducted to examine potential sources of heterogeneity in two outcome groups (LCS and general population). For the LCS group, we examined whether studies varied because of their definition of ILA and inclusion criteria. First, we removed studies that excluded known or treated ILD. Although the I² value dropped (97.2%), heterogeneity remained significant. By exploring heterogeneity according to ILA definition, no change in I² was observed either. Within the general population outcome group, we examined effects of using thin-section technique versus cardiac CT, but no change was observed in the heterogeneity estimate. Similarly, the study year did not affect the results, with I² values of 98.2% for the years 2010–2018 and 99.8% for the years 2019–2023 in the LCS group, whereas it was 99% for the years 2010–2016 and 97.3% for the years 2019–2022 in the general population group. Study location (North America vs. Europe) did not substantially change the heterogeneity results either.

Figure 2.

Pooled prevalence of interstitial lung abnormalities. (A) LCS. (B) GP. (C) FC. AGES = Age, Gene/Environment Susceptibility; CI = confidence interval; ES = effect size; FC = at-risk familial cohort; FHS = Framingham Heart Study; GP = general population cohort; LCS = lung cancer screening cohort.

Factors Associated with ILA Prevalence

In the overall cohort, 44.9% of the participants with ILAs present were female, the mean age was 65.2 ± 9.6 years, and the mean FVC% predicted was 99.1 ± 2.8, compared with those without ILAs who were 43.2% female, mean age 56.1 ± 20.5 years, with FVC% predicted 101.2 ± 4.4 (Table 2). Increased age (weighted mean difference, 2.22; 95% CI, 1.89 to 2.54), lower FVC% predicted (weighted mean difference, −1.72; 95% CI, −3.00 to −0.45), and male sex (risk ratio, 0.84; 95% CI, 0.71 to 0.94) were associated with increased prevalence of ILAs. Between-study heterogeneity was significant for age (I² = 96%; P < 0.00001) and sex (I² = 88%; P < 0.00001), but not FVC% predicted (I² = 21%; P = 0.27) (Figure 3). Other patient-specific variables, such as concomitant chronic obstructive pulmonary disease, ILA progression, or MUC5B polymorphism, were infrequently reported and thus not able to be analyzed in meta-analysis.

Table 2.

Cohort Demographics

	ILA Present	ILA Absent	Comparison (95% CI)	P Value*
Count	8,346	79,958	—	—
Sex, % female (n)
Total cohort	44.8 (3,097)	43.7 (16,737)	Risk ratio, 0.82 (0.71 to 0.93)	0.003
LCS	44.9 (2,506)	41.7 (10,603)
GP	42.1 (339)	42.7 (3,970)
FC	55.4 (82)	61.4 (277)
Age, mean ± SD
Total cohort	65.5 ± 8.8	57.3 ± 19.0	Weighted mean difference, 2.27 (1.95 to 2.58)	<0.00001
LCS	63.5 ± 4.1	60.7 ± 3.9
GP	71.9 ± 6.7	66.4 ± 11.0
FC	59.5 ± 10.6	42.6 ± 28.6
FVC% predicted, mean ± SD
Total cohort	97.9 ± 4.3	101.1 ± 4.1	Weighted mean difference, −2.12 (−3.37 to −0.87)	0.0009
LCS	96.6 ± 6.2	101.8 ± 2.0
GP	97.0 ± 5.7	98.0 ± 4.2
FC	99.9 ± 1.1	102.4 ± 5.9

Definition of abbreviations: CI = confidence interval; FC = at-risk familial cohort; GP = general population cohort; ILA = interstitial lung abnormalities; LCS = lung cancer screening cohort.

^*P values for comparisons.

Figure 3.

Patient-specific variables associated with interstitial lung abnormality (ILA) risk. (A) Age. (B) Male sex. (C) FVC% predicted. Studies marked (A) indicate meta-analysis of “no ILA” demographic data; studies marked (B) indicate meta-analysis of “indeterminate for ILA” data. CI = confidence interval; FC = at-risk familial cohort; FHS = Framingham Heart Study; GP = general population cohort; IV = inverse variance; LCS = lung cancer screening cohort; M–H = Mantel-Haenszel.

Mortality

The pooled mortality risk was higher in those with ILAs than those without (odds ratio [OR], 3.56; 95% CI, 2.19–5.81), over a median follow-up duration of 5 years (range, 2–11 yr). Between-study heterogeneity was high (I² = 93%; P < 0.001) (Figure 4). In the LCS cohort, ILAs were associated with increased risk of mortality based on three studies (OR, 2.41; 95% CI, 1.29–4.50) (I² = 90%; P < 0.001) (22, 23). In the general population cohorts, ILAs increased the risk of pooled mortality based on three studies (OR, 4.76; 95% CI, 2.12–10.67) (I² = 90%; P < 0.001) (8, 30, 34). Mortality risk was only available from one familial cohort (OR, 5.81; 95% CI, 2.23–15.16) (36).

Figure 4.

Association of ILAs and all-cause mortality. CI = confidence interval; FC = at-risk familial cohort; FHS = Framingham Heart Study; GP = general population cohort; ILAs = interstitial lung abnormalities; LCS = lung cancer screening cohort; M–H = Mantel-Haenszel.

Quality Assessment

Risk of bias was assessed for all 22 included studies (Figure E2). Three out of 10 LCS studies were judged as having high risk of bias, primarily due to lack of random selection and nonresponse bias. Three of the seven general population cohort studies were deemed at high risk of bias because of a limitation in their target population, sampling frame, and randomization of selection. Two of the familial cohort studies were deemed at high risk of bias and three at moderate risk, because of their limited external validity based on their sampling from a mostly registry-based, tertiary care center population. Definitions of subclinical ILD were fairly homogeneous across familial cohort studies, but definitions of at-risk family members varied by age cut-offs and the requirement of either one or two relatives to be diagnosed with a form of ILD.

Discussion

In this systematic review and meta-analysis, ILAs were identified with high prevalence in LCS participants and in the general population. Subclinical ILD was also highly prevalent in familial cohorts. ILAs are an important radiographic finding that may precede a clinical ILD diagnosis and, in and of themselves, are associated with an increased risk of mortality. Taken together, these findings inform the need for increased study and recognition of ILAs, standardized radiologic reporting guidelines, and evidence-informed referral and management pathways. With the increasing availability and use of chest CT, incidental findings such as ILAs are expected to become more frequently encountered over time. Our findings highlight the potential for screening of specific patient populations to identify individuals at risk for or with early forms of ILD.

The overall pooled ILA prevalence was slightly lower in the LCS and general population cohorts, whereas subclinical ILD was substantially higher in at-risk family members of patients with pulmonary fibrosis. The heterogeneity across these groups is likely indicative of their demographic and genetic risks and may be useful to inform differential approaches to ILA screening across different groups. It is unknown what proportion of those undergoing LCS or enrolled in population-based cohorts have a family history of ILD and thus may have familial PF. Future prospective ILA studies should ascertain whether participants have a family history of pulmonary fibrosis to better characterize these risks. Interestingly, a recent study from the United Kingdom reported an estimated 11% prevalence of residual lung abnormalities in patients who required and survived hospitalization for coronavirus disease (COVID-19) pneumonia (42). Although these residual lung abnormalities are not ILA equivalents, nor have their frequency of progression been well defined, the number indicates a substantial proportion of individuals at risk for fibrotic ILD, particularly on a global scale. In addition, the high proportion of residual lung abnormalities may further complicate the interpretation of CT scans from a risk assessment and healthcare capacity perspective. Ongoing work is needed to understand any differences and similarities between residual lung abnormalities and ILAs and to better characterize their risk of both development and progression.

Delays in ILD diagnosis are a major issue, associated with worse survival, even when accounting for disease severity and age at presentation (43). Earlier diagnosis provides an opportunity for prompt disease management, pharmacotherapeutic intervention, education, and enrollment in clinical trials (44–46). However, not all individuals with ILAs have radiographic progression, and even fewer develop clinical ILD. Some ILAs resolve or regress, although it is unclear if this occurs spontaneously or because of risk factor mitigation (14, 35, 36). The Fleischner paper recommends active monitoring of those with known risk factors for clinically relevant or progressive disease, which includes exposures known to cause ILD as well as fibrotic features on CT (6). Individuals with ILAs are recommended to have baseline PFTs (6). An FVC < 80% predicted and Dl_CO < 70% predicted may identify those at highest risk of a clinical ILD diagnosis and mortality (47). Ultimately, incorporating multiple parameters, including blood biomarkers and radiologic features, may best inform how risk factor modification and early treatment initiation may alter the disease trajectory in at-risk patients and improve outcomes, but a composite risk score has yet to be developed (48).

The high prevalence of ILAs in LCS cohorts highlights the potential to leverage existing screening programs to identify individuals at risk for ILD and enable early diagnosis of pulmonary fibrosis in individuals with certain risk factors. Yet, many ILAs remain unrecognized and underreported (49). Given the potential for ILAs to represent undiagnosed clinical disease, and their independent association with increased mortality, ILAs warrant standardized reporting, referral, and follow-up. A recent expert survey initiative recommended that the presence of honeycombing be reported as a clinically significant finding indicative of ILD when detected on LCS CTs (50). In addition, consensus was reached that a recommendation should be made to refer individuals to pulmonology on the detection of honeycombing, traction bronchiectasis, or bronchiolectasis. Currently, in some regions, ILAs are recommended to be flagged on radiology reports using an S-modifier, which identifies a significant or potentially significant finding, but this does not necessarily guarantee subsequent clinical evaluation, diagnostic testing, or follow-up. Equitable access to care and follow-up, particularly in such large populations as those undergoing LCS, is essential, and agreement on a standardized approach to recognition, reporting, and follow-up is urgently needed.

The evidence from this review may be of moderate certainty, because of estimated between-study heterogeneity. Although the observed heterogeneity in the pooled estimates raises concerns about the accuracy of the averaged estimates and their clinical value, the relatively high heterogeneity values have been anticipated because of the characteristics of the included studies. It has been previously discussed that meta-analysis of prevalence estimates often present with high heterogeneity, given variation in the contexts and uniqueness of the populations from which the individual proportions were obtained (51). This trend can be especially observed in studies estimating regional or national disease burden in diverse populations. Accordingly, considering the international background of the included studies and variations across their clinical settings, the pooled prevalence may still offer an appropriate estimate reflecting the burden of ILA. The meta-regression findings were likely influenced by the weighting of parent studies because of heterogeneous sample sizes. CT scan techniques (high-resolution CT vs. low-dose CT) also varied across studies, impacting the external validity of our results. Although our review process identified few studies that reported mortality, there was a consistent trend of increased ORs across included studies with overlapping CIs, consistent with prior publications (8). Study authors were contacted regarding missing data, which reduced the impact of missing information. Given there have only been a few published ILA studies, and all known studies at the time of the search were identified, the risk of nonreporting bias was believed to be low.

ILAs and subclinical ILD are prevalent in LCS, general population, and at-risk familial cohorts. Our findings inform the urgent need to characterize those at risk of active or progressive disease and to establish evidence-based recommendations on risk prediction, longitudinal follow-up, and management. A standardized approach to reporting ILAs should be developed and implemented to ensure appropriate recognition and management of individuals with ILAs.