Associations of albuminuria with interstitial lung abnormalities in older community-dwelling adults confounded by age

Abstract

Background

Pulmonary microvascular dysfunction has been suggested to be an early feature of interstitial lung changes, which may precede interstitial lung disease. The prospective association of albuminuria, a marker of endothelial dysfunction, with interstitial lung abnormalities (ILA) and high-attenuation areas (HAA) remains unexplored.

Methods

The study included participants with available spot urinary albumin-creatinine ratio (UACR) and computed tomography data for ILA and HAA enrolled in two independent cohorts, Multi-Ethnic Study of Atherosclerosis (MESA; n=2248) and Age Gene/Environment Susceptibility (AGES)-Reykjavik (n=3509). HAA were defined as the percentage of imaged lungs with attenuation between −600 and −250 HU (MESA only). Regression modelling was performed to assess the associations of UACR with ILA and HAA, adjusted for anthropometric and demographic variables and kidney function. Cox proportional-hazard models were used to examine whether ILA modified the association between albuminuria and all-cause mortality.

Results

Log-transformed UACR was significantly associated with ILA, with an OR 1.21 (95% CI 1.12–1.30) in MESA and OR 1.13 (95% CI 1.06–1.21) in AGES-Reykjavik. In multivariable-adjusted models incorporating age, albuminuria was no longer associated with ILA, nor with ILA progression in AGES-Reykjavik. In MESA, higher levels of albuminuria were associated with greater HAA (mean increase of 1.01% per 1-unit increment in log-transformed UACR, 95% CI 1.01–1.02%), even after adjusting for covariates including age. Albuminuria was more strongly associated with death among those with ILA in MESA, but not in AGES-Reykjavik.

Conclusions

Albuminuria was not associated with ILA after accounting for chronological age. Our findings suggest that there may be a common systemic pathology of ageing that underlies albuminuria and interstitial lung changes.

Shareable abstract

In two population-based cohorts, no independent relationship between albuminuria and ILA was found, with age emerging as a key confounding factor. The results suggest there may be a common systemic pathology of ageing that underlies albuminuria and ILA. https://bit.ly/3BLwTfL

Introduction

Interstitial lung disease (ILD) refers to a group of diffuse parenchymal lung disorders with inflammation of varying degrees and sometimes fibrosis [1]. The pathogenesis of ILD may involve pulmonary microvascular dysfunction that precede the onset of pulmonary fibrosis by years [2–5]. Systemic markers of endothelial function may provide further insight as to the relationship between vascular integrity and ILD.

Increased albuminuria, an established marker of kidney damage and chronic kidney disease (CKD), correlates with systemic microvascular dysfunction, including dysfunction within the pulmonary circulation [5–7]. Epidemiological data indicate a high prevalence of moderate-to-severe reduction in kidney function (i.e. estimated glomerular filtration rate (eGFR) <60 mL·min⁻¹·1.73 m⁻²) among individuals with ILD, suggesting that CKD may serve as an antecedent risk factor for ILD development [8]. Older age is a shared risk factor for both ILD and CKD, and may confound this relationship. Validated pre-clinical models of subclinical pulmonary fibrosis are lacking and make it challenging to ascertain whether risk factors, such as albuminuria, truly contribute to the development of pulmonary fibrosis.

Computed tomography (CT) assessments that detect early signs of lung fibrosis in humans can partly address this gap. The best-studied definition of such signs is interstitial lung abnormalities (ILA), a term used to describe CT features of ILD in cohort study participants without clinically diagnosed ILD [9]. The presence of ILA on CT is associated with lower lung function and worse survival [10, 11]. Additionally, ILA share risk factors with ILD, such as older age, smoking and genetic polymorphisms, including the MUC5B common promoter variant (rs35705950) [12–15]. While ILA has shown varying associations with CKD defined by reduced kidney function across population-based cohorts [16], their association with albuminuria remains unknown. The detection of early signs of lung fibrosis on CT images has also been done using quantitative methods [17]. One such definition is that of high-attenuation areas (HAA), an automated CT densitometric method that quantifies lung areas that correlate with ILD-related CT patterns. HAA associate with mortality and hospitalisation due to ILD and the MUC5B promoter variant [18–20]. We hypothesised that higher levels of albuminuria were associated with increased odds of ILA and more HAA on CT, and tested this in large, population-based cohorts. Additionally, we explored whether the presence of ILA modified the association between albuminuria and mortality.

Methods

Study population

The Multi-Ethnic Study of Atherosclerosis (MESA) is a prospective cohort study that enrolled 6814 men and women between the ages of 45 and 84 years without known cardiovascular disease from six communities in the USA between 2000 and 2002 [21]. MESA participants attended the most recent follow-up visits in 2010–2012 (exam 5) and 2016–2018 (exam 6).

The Age Gene/Environment Susceptibility (AGES)-Reykjavik (Age, Gene/Environment Susceptibility-Reykjavik) study is a longitudinal birth cohort derived from the previous Reykjavik study, and included 5764 individuals born in Reykjavik (Iceland) who were aged 66–96 years at the time of enrolment (2002–2006) [22], with follow-up examinations being conducted from 2007 to 2011.

Both MESA and AGES-Reykjavik had no selection criteria related to lung disease, respiratory symptoms, smoking history or kidney disease. Written informed consent, including consent for genetic studies, was obtained from all participants. The University of Virginia research ethics committee (IRB-HSR#23936) and Icelandic Bioethics Committee (VSN-00-063) approved this study. More detailed cohort information can be found in the supplementary methods.

The timing of recruitment, albuminuria and ILA/HAA measurements, and follow-up assessments in both MESA and AGES-Reykjavik cohorts are shown in figure 1.

FIGURE 1.

Participant flow for the a) Multi-Ethnic Study of Atherosclerosis (MESA) and b) Age Gene/Environment Susceptibility (AGES)-Reykjavik studies. UACR: urinary albumin-creatinine ratio; ILA: interstitial lung abnormality; HAA: high-attenuation areas.

Albuminuria

Spot urinary albumin-creatinine ratio (UACR) measurements obtained during MESA exam 5 and the AGES-Reykjavik baseline examination were used for our analyses (supplementary table S1). Urine sampling adhered to standardised protocols, and the obtained samples were processed in central laboratories (supplementary table S2). While laboratory technologies differed between the two studies, the measurements were conducted similarly: urine albumin concentration was measured by nephelometry or immunoturbidometry, and urine creatinine concentration assessed by the Jaffé method [23, 24]. Both parameters were used to calculate UACR.

Interstitial lung abnormalities

CT data from MESA exam 5 and both AGES-Reykjavik baseline and follow-up examinations were included in our analyses (supplementary table S1). ILA were identified by visual assessments of full-lung CT (MESA) or chest CT scans (AGES-Reykjavik) and defined as nondependent radiological abnormalities affecting >5% of any lung zone, including ground-glass or reticular abnormalities, nonemphysematous cysts, honeycombing and traction bronchiectasis, as per 2020 Fleischner Society criteria (see supplementary methods for additional information) [9]. Changes affecting <5% of any lung zone were considered indeterminate and excluded from the analysis. In line with the 2020 Fleischner Society criteria, centrilobular abnormalities were excluded from the definition of ILA, as these are generally considered a common, nonprogressive finding, typically unrelated to fibrosis [9, 14, 25].

A subset of the AGES-Reykjavik participants underwent repeated CT scans with ILA assessments at the follow-up examination. For participants displaying ILA in at least one of the initial or subsequent CT scans, both sets of images were compared simultaneously [15]. ILA progression was defined as an increase in affected lung areas with nondependent ground-glass or reticular abnormalities, nonemphysematous cysts, honeycombing or traction bronchiectasis, or new appearance of at least one such abnormality [15]. All paired CT evaluations were determined by consensus among three readers, as described previously [15].

High-attenuation areas

Image attenuation was measured on full-lung CT scans from MESA exam 5 and semi-automatically segmented and corrected by a trained technician without knowledge of other participant information using the Pulmonary Analysis Software Suite at the University of Iowa's Advanced Pulmonary Physiomic Imaging Laboratory (Iowa City, IA, USA) (supplementary methods) [26]. HAA were defined as the percentage of imaged lungs with attenuation values between −600 and −250 HU [18].

Covariates and laboratory methods

Details regarding the covariates of interest and laboratory methods can be found in the supplementary methods. Briefly, sex, race and ethnicity, and smoking (history and pack-years) were self-reported in both cohorts. Hypertension was defined as systolic or diastolic blood pressure readings ≥140 or ≥90 mmHg, respectively, or by hypertension medication use. Diabetes mellitus was defined by fasting glucose levels ≥126 mg·dL⁻¹, self-reported physician's diagnosis, or by diabetic medication use [27]. CT scanner parameters included scanner model, radiation dose, total lung volume imaged and percentage of emphysema (only available in MESA), defined as the percentage of lung voxels <−950 HU. The genotyping methods for MUC5B (rs35705950) single nucleotide polymorphism in MESA and AGES-Reykjavik are detailed in the supplementary methods. Based on the number of copies of the risk allele MUC5B rs35705950-T, participants were classified into two groups: those with no copies and those with one or two copies.

Mortality ascertainment

MESA participants or their surrogates were contacted at 9–12-month intervals to obtain all-cause death data. This information was supplemented by a review of the National Death Index. All-cause mortality was adjudicated from exam 5 until 31 December 2018. In the AGES-Reykjavik study, all-cause mortality was ascertained until 31 January 2023, with survival status determined by death certificates and data collected from Registers Iceland, the official civil registry.

Statistical analysis

Participant characteristics were compared within subgroups using ANOVA, Kruskal–Wallis test, or pairwise Chi-squared test, as appropriate. Covariates were considered in models a priori on the basis of their consistent associations with UACR, ILA and HAA, and mortality [28–30]. UACR, HAA and percentage emphysema in natural log-transformation were used in all regression models due to their right-skewed distribution, following the approach of previous studies [20, 24, 29]. In MESA, missing cigarette pack-year data at exam 5 for ILA (n=34) and HAA (n=44) were addressed using a missing cigarette pack-year category in regression models, in accordance with previous studies [31]. In AGES-Reykjavik, pack-years were imputed using k-nearest neighbour imputing for participants reporting cigarette smoking, but reporting 0 pack-years. A generalised linear mixed model approach with random intercepts for clustering effect of study sites was used to examine the cross-sectional association of UACR with ILA and HAA in MESA, while logistic regression models were used to examine the cross-sectional association of UACR with ILA in AGES-Reykjavik. The first model was unadjusted. The second model was adjusted for potential confounders, which included age, sex, weight, height, hypertension medication use, including the use of angiotensin-converting enzyme inhibitors and angiotensin receptor blockers, insulin or other diabetic medication use, smoking (history and pack-years), total cholesterol and eGFR. Additionally, models in MESA were adjusted for waist circumference and race and ethnicity while models in AGES-Reykjavik were adjusted for abdominal circumference. HAA models were adjusted for CT scanner parameters in addition to the aforementioned covariates. The estimated effects in the HAA analysis were exponentiated to report outcomes as percentage differences in HAA to improve interpretability. Effect modification was examined in stratified models and by multiplicative interaction terms. Since age was a significant modifier of the association between UACR and ILA/HAA, its role was further explored through stratified analyses using 75 years as a cut-off. The cut-off was chosen because AGES-Reykjavik comprised an older population at enrolment compared to MESA. For the final analysis, we included participants with complete covariate data. Sensitivity analyses were conducted, stratified by race and ethnicity in MESA and by albuminuria range in both cohorts (i.e. by excluding participants with UACR levels above clinically defined cut-off values for severely increased albuminuria (>300 mg·g⁻¹) or nephrotic-range albuminuria (>2200 mg·g⁻¹) as per the 2012 Kidney Disease: Improving Global Outcomes CKD consensus guideline [5]). Exploratory analyses were performed to examine potential interactions between albuminuria and genetic risk for ILD. In MESA, the MUC5B risk allele status was assessed in a two-way interaction term with UACR, adjusting for the principal components of genetic ancestry in MESA. In AGES-Reykjavik, the stratified analysis by MUC5B risk allele count did not include the interaction term. In AGES-Reykjavik, logistic regression was used to investigate the association of baseline UACR with ILA progression. Adjustment was performed for a subset of covariates (age, sex, weight, height, eGFR, smoking history and pack-years) considering the limited number of participants available for ILA progression analysis.

The relationship of UACR and ILA status with time to all-cause death was examined in survival analysis. The survival probabilities of UACR ≤10 mg·g⁻¹ were estimated using the Kaplan–Meier method and their differences were assessed by log-rank tests. A UACR cut-off of ≥10 mg·g⁻¹ has been shown to independently predict all-cause mortality in the general population [32]. This association was further explored using Cox proportional hazards models, adjusting for age, sex and race and ethnicity in MESA, and weight, height, study site, total family gross income (in MESA), educational level, total intentional exercise, waist or abdominal circumference, smoking (history and pack-years), statin and hypertension medication use, insulin or diabetes medication use, total cholesterol, high-density lipoprotein cholesterol, eGFR and Agatston calcium score. To evaluate whether the presence of ILA on CT modified the association between albuminuria and death, we tested the interactive effect of ILA status and UACR in the Cox regression. p-values <0.05 were considered statistically significant for main and interactive effects. Analyses were conducted in SAS version 9.4 (SAS Institute Inc.) and R Foundation for Statistical Computing version 4.1.4 (Vienna, Austria).

Results

The characteristics of the participants with ILA and albuminuria data are summarised in table 1. <3% had missing covariate data in each cohort (supplementary table S3 shows comparison of included and excluded participants). In MESA and AGES-Reykjavik, 11.9% and 9.3% of the included participants, respectively, had ILA. Participants with ILA in each cohort tended to be older and were more likely to have a history of smoking, higher levels of albuminuria and lower eGFR values than those without ILA. AGES-Reykjavik had a higher proportion of men among those with ILA than in those without ILA. In MESA, hypertension was more common among those with ILA. Diabetes was more common among participants in MESA than in AGES-Reykjavik.

TABLE 1.

Characteristics of participants with assessments of albuminuria and interstitial lung abnormalities

	MESA# (n=2248)			AGES-Reykjavik# (n=3509)
	No ILA	ILA	p-value	No ILA	ILA	p-value
Participants	1982	266		3183	326
Age, years	68.3±9.0	74.3±9.2	<0.001	75.9±5.4	78.1±5.6	<0.001
Female	1024 (51.7)	147 (55.3)	0.30	1888 (59.3)	147 (45.1)	<0.001
Height, cm	165.6±10.0	163.8±10.1	0.002	166.9±9.2	167.9±9.8	0.08
Weight, kg	78.0±17.7	76.1±15.8	0.036	76.0±14.6	76.7±15.6	0.40
Body mass index, kg·m−2	28.4±5.5	28.4±5.3	0.73	27.2±4.4	27.1±4.7	0.83
Race and ethnicity			0.69			NA
Non-Hispanic White	783 (39.5)	113 (42.5)		100	100
Asian American	293 (14.8)	33 (12.4)		0	0
African American	482 (24.3)	64 (24.1)		0	0
Hispanic	424 (21.4)	56 (21.1)		0	0
Hypertension¶	1142 (57.6)	181 (68.3)	<0.001	2567 (80.6)	267 (81.9)	0.64
Diabetes mellitus+	797 (40.2)	112 (42.1)	0.55	395 (12.4)	47 (14.4)	0.34
Smoking status			0.002			<0.001
Never	998 (50.4)	104 (39.1)		1442 (45.3)	89 (27.3)
Former	849 (42.8)	138 (51.9)		1368 (43.0)	183 (56.1)
Current	135 (6.8)	24 (9.0)		373 (11.7)	54 (16.6)
Smoking§ pack-years	0 (0–9.0)	1.0 (0–20.3)	<0.001	0 (0–17)	11 (0–27)	<0.001
ACEi or ARB medication	650 (32.8)	106 (39.8)	0.027	899 (28.2)	88 (27.0)	0.68
UACR, mg·g−1 creatinine	5.8 (3.5–12.8)	8.3 (4.0–19.8)	<0.001	2.5 (1.2–6.1)	3.3 (1.5–11.0)	<0.001
UACR categories			0.011			<0.001
<30 mg·g−1 creatinine	1750 (88.3)	219 (82.3)		2951 (92.7)	284 (87.1)
30–300 mg·g−1 creatinine	192 (9.7)	36 (13.5)		198 (6.2)	40 (12.3)
≥300 mg·g−1 creatinine	40 (2.0)	11 (4.1)		34 (1.1)	2 (0.6)
eGFRƒ, mL·min−1·1.73 m−2	82.3±17.0	75.1±18.5	<0.001	68.8±15.6	66.8±16.5	0.03
MUC5B risk allele status##			0.09			<0.001
0 copies of risk allele	1766 (89.1)	225 (84.6)		2501 (78.7)	174 (53.5)
1–2 copies of risk allele	216 (10.9)	41 (15.4)		676 (21.3)	151 (46.5)

Data are presented as n, mean±sd, n (%) or median (interquartile range), unless otherwise stated. MESA: Multi-Ethnic Study of Atherosclerosis; AGES: Age Gene/Environment Susceptibility; ILA: interstitial lung abnormalities; ACEi: angiotensin-converting enzyme inhibitor; ARB: angiotensin receptor blocker; UACR: urinary albumin–creatinine ratio; eGFR: estimated glomerular filtration rate; NA: not applicable. ^#: characteristics from MESA (exam 5, 2010–2012) and AGES-Reykjavik (baseline, 2002–2006) are limited to participants with available data on UACR and ILA status, and complete covariate data; ^¶: defined as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, or current hypertension medication use; ⁺: defined as self-reported physician diagnosis, fasting glucose ≥126 mg·dL⁻¹, or diabetic medications use; ^§: for 39 AGES-Reykjavik participants reporting cigarette smoking but no pack-years, pack-year count was k-nearest-neighbour-imputed using other covariates; ^ƒ: in both MESA and AGES-Reykjavik, the eGFR was calculated using the 2021 Chronic Kidney Disease Epidemiology Collaboration creatinine equation, which does not include adjustment for race; ^##: data on MUC5B status (promoter polymorphism of the gene encoding mucin 5B (rs35705950)) were missing for 26 and seven of the included participants in MESA and AGES-Reykjavik, respectively.

Association between albuminuria and ILA

Higher levels of albuminuria were associated with the presence of ILA in unadjusted analyses. A 1-unit increment in log-transformed UACR was associated with an odds ratio for ILA of 1.21 (95% CI 1.12–1.30) in MESA and 1.13 (95% CI 1.06–1.21) in AGES-Reykjavik (table 2). In multivariable adjusted models, albuminuria was no longer significantly associated with ILA in either cohort. Age emerged as a key confounding covariate (supplementary table S4). In both cohorts, participants who were aged ≥75 years were more likely to have hypertension, a history of smoking, higher albuminuria levels and lower eGFR (supplementary table S5). An association between albuminuria and ILA was observed in the older age group in MESA (OR 1.10, 95% CI 1.00–1.21), but this association was attenuated when participants with severe or nephrotic-range albuminuria were excluded (supplementary table S6). In contrast to MESA, higher albuminuria was associated with higher odds of ILA in the AGES-Reykjavik younger age group (OR 1.15, 95% CI 1.01–1.33). No effect modification was observed when the participants were stratified by race and ethnicity in MESA (supplementary table S7) or MUC5B risk allele count in both cohorts (supplementary table S8).

TABLE 2.

Cross-sectional associations between albuminuria and interstitial lung abnormalities (ILA)

	MESA			AGES-Reykjavik
	Participants, n	OR for ILA per 1-unit increment in log-transformed UACR (95% CI)	p-value	Participants, n	OR for ILA per 1-unit increment in log-transformed UACR (95% CI)	p-value
Unadjusted model	2248	1.21 (1.12–1.30)	<0.001	3509	1.13 (1.06–1.21)	<0.001
Adjusted model#	2248	1.07 (0.97–1.18)	0.19	3509	1.05 (0.98–1.13)	0.19

Numbers of Multi-Ethnic Study of Atherosclerosis (MESA) participants (exam 5, 2010–2012) and Age Gene/Environment Susceptibility (AGES)-Reykjavik (baseline, 2002–2006) are limited to those with available data on urinary albumin-creatinine ratio (UACR), ILA status and complete covariate data. UACR was log-transformed prior to regression. p-values for ILA interaction with log-transformed UACR and age=0.56. ^#: adjusted for age, sex, weight, height, hypertension medication use, angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use, insulin or other diabetic medication use, smoking (history and pack-years), total cholesterol and estimated glomerular filtration rate. Race and ethnicity and waist circumference were additionally adjusted for in MESA, while abdominal circumference was additionally adjusted for in AGES-Reykjavik.

Among AGES-Reykjavik participants who underwent repeated CT scans after ∼5 years, 304 had ILA either at baseline or at follow-up and were included for analysis of ILA progression. Baseline albuminuria was not associated with ILA progression over time in the adjusted model (supplementary table S9).

Association between albuminuria and HAA

Among MESA participants who underwent HAA assessment, <2% had missing covariate data. Baseline characteristics of this sample are shown in supplementary tables S10 and S11. In the multivariable adjusted model, higher albuminuria was associated with higher HAA. A 1-unit increment in log-transformed UACR was associated with a mean percentage increase in HAA of 1.01% (95% CI 1.01–1.02%) (table 3). Excluding age as a covariate did not significantly modify the relationship between albuminuria and HAA (supplementary table S12). Among those included in HAA analysis, participants aged ≥75 years were more likely to have hypertension, a positive smoking history, higher albuminuria levels and lower eGFR (supplementary table S13). The association between albuminuria and HAA was more pronounced in the older age group (supplementary table S14), among Asian and Black and African American participants (supplementary table S15), and among those with no copies of the MUC5B risk allele (supplementary table S16). Stratification by clinically defined albuminuria ranges did not modify the overall association.

TABLE 3.

Cross-sectional association between albuminuria and high-attenuation areas (HAA) in the Multi-Ethnic Study of Atherosclerosis (MESA)

	Participants, n	Mean % difference in HAA (95% CI) per 1-unit increment in log-transformed UACR	p-value
Unadjusted model	2844	1.05 (1.03–1.06)	<0.001
Adjusted model#	2844	1.01 (1.01–1.02)	<0.001

UACR: urinary albumin-creatinine ratio. ^#: adjusted for age, sex, race and ethnicity, height, weight, waist circumference, smoking history, cigarette pack-years, hypertension medication use, angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use, insulin or other diabetic medication use, total cholesterol, estimated glomerular filtration rate, and computed tomography scanner parameters (scanner model, radiation dose, total imaged lung volume, percentage emphysema).

Albuminuria, ILA status and mortality

Over a median 9.2 years in MESA and 12.6 years in AGES-Reykjavik, a total of 325 (14.0%) and 2787 (79.3%) deaths occurred, respectively (supplementary table S17). In both cohorts, the absolute mortality rates were higher and within a similar range among participants with ILA or higher levels of albuminuria (≥10 mg·g⁻¹) than in those who did not meet either of these criteria. The presence of both ILA and higher levels of albuminuria was associated with a greater risk of death in both cohorts (figure 2; log-rank p<0.001).

FIGURE 2.

Kaplan–Meier estimates of survival in a) the Multi-Ethnic Study of Atherosclerosis (MESA) and b) Age Gene/Environment Susceptibility (AGES)-Reykjavik by interstitial lung abnormality (ILA) status and urinary albumin–creatinine ratio (UACR) levels. The number of participants at risk in MESA (from exam 5 through of 2018) and AGES-Reykjavik (from baseline through follow-up) are shown by by ILA status and UACR levels.

The multivariable Cox regression in MESA showed that the presence of ILA modified the association between UACR and mortality (p for interaction=0.01) (table 4). A 1-unit increment in log-transformed UACR was associated with HR 1.92 (95% CI 1.54–2.39) among those with ILA compared with 1.16 (95% CI 1.05–1.29) in those without ILA. These findings in MESA were not replicated in AGES-Reykjavik (p for interaction=0.42).

TABLE 4.

Association of albuminuria with mortality by interstitial lung abnormalities (ILA) status

	Participants	Hazard ratio per 1-unit increment in log-transformed UACR (95% CI)	p-value for interaction
MESA			0.01
No ILA	1714	1.16 (1.05–1.29)
ILA present	220	1.92 (1.54–2.39)
AGES-Reykjavik			0.42
No ILA	3142	1.11 (1.08–1.14)
ILA present	317	1.05 (0.98–1.13)

All hazard ratios and 95% confidence intervals are for the coefficient of continuous urinary albumin-creatinine ratios (UACR) using separate Cox regressions with participants stratified by ILA status. p-values are for the interaction term in a Cox regression which includes the direct effects and interaction effect of UACR and ILA with all participants included. Models are adjusted for age, sex, weight, height, study site (in MESA), total family gross income (in MESA), educational level, total intentional exercise, smoking (history and pack-years), statin and hypertension medication use, insulin or other diabetes medication use, total cholesterol, high-density lipoprotein cholesterol, estimated glomerular filtration rate and Agatston calcium score. Race and ethnicity and waist circumference were additionally adjusted for in MESA, while abdominal circumference was additionally adjusted for in AGES-Reykjavik. In MESA, 1934 participants were included in the adjusted model (407 excluded due to missing covariate data), while in AGES-Reykjavik 3459 participants were included in the adjusted model (57 excluded due to missing covariate data). MESA: Multi-Ethnic Study of Atherosclerosis; AGES: Age Gene/Environment Susceptibility.

Discussion

Contrary to our hypothesis, we did not find an association between albuminuria and ILA or its progression after adjustment for age. Conversely, elevated levels of albuminuria were associated with greater HAA, even after accounting for covariates including age. Taken together, our results do not support albuminuria as a marker of lung-specific endothelial dysfunction. Our findings reinforce the importance of accounting for age and ageing-related factors, as further work is conducted in identifying risk factors for ILD.

Prior studies have identified other biomarkers of endothelial dysfunction to be associated with ILA and HAA, which include adhesion molecules and D-dimer levels in the blood [33, 34]. Despite research strongly suggesting that albuminuria is an indirect marker of systemic endothelial dysfunction in humans and is even associated with worse lung function [29], albuminuria does not appear to be independently associated with ILA. This may be partly explained by the relationship of chronological age to both ILA and CKD. Older age is one of the strongest risk factors for ILA [13], and may reflect biological ageing mechanisms that contribute to recurrent lung injury and fibrosis [35, 36]. Similarly, these ageing biological processes have been attributed to other fibrosing diseases like CKD [37], wherein albuminuria serves as a marker for accelerated kidney telomere shortening [38]. In aggregate, these findings support the notion that ILA and albuminuria may serve as independent markers of systemic biological ageing.

Notably, the association between albuminuria and ILA varied across different age groups in both cohorts. These differing findings may be attributed to age differences between the cohorts, since the mean age of the younger age group in AGES-Reykjavik was higher than that of the younger age group in MESA. Thus, our findings suggest that although age was an important shared risk factor for ILA and albuminuria in the overall cohort, an age-related association between albuminuria and ILA cannot be ruled out; specifically, albuminuria and endothelial dysfunction may be more relevant to the presence of ILA in adults aged <75 years, before the occurrence of overlapping contributions by age-related factors. Furthermore, while a combination of clinical and genetic risk factors has been posited to contribute to the development of ILD, we did not find albuminuria to be more strongly associated with ILA among carriers of the MUC5B risk allele.

Albuminuria is a recognised risk factor for worse mortality, and we found this among adults with ILA on their CT scan in MESA. There may be systemic inflammatory and even fibrosing process with multiorgan involvement that can have an impact on overall health. Indeed, the prevalence of comorbidities tend to be higher among adults with ILA than those without ILA [16]. However, this was not replicated in AGES-Reykjavik, and thus our interpretation remains speculative. Our study emphasises the need for further research dedicated to imaging and biomarkers in this population to noninvasively evaluate multiorgan fibrosis and better understand the shared pathophysiology. Such research could lead to more precise identification of individuals at risk and guide the development of targeted interventions.

In contrast to ILA, albuminuria was associated with more HAA on CT, although associations were modest. One explanation for this may be that HAA are not as specific to ILD-related radiological patterns as ILA. While HAA are associated with ILD hospitalisation and death and genetic risk factors for ILD, they may still be confounded by adiposity and atelectasis [18–20, 39]. Alternatively, HAA could be detecting smouldering inflammation in the lungs that correlates with albuminuria, but does not reflect ILD pathology, as indicated by our ILA findings.

The inclusion of data from two independent population cohorts and countries extends the generalisability of our findings to those specific demographics [9, 40, 41]. However, the study had several limitations. While we explored the relationship between albuminuria and ILA/HAA, the rationale is limited by the indirect nature of potential disease mechanisms and shared risk factors, impacting the interpretation of these findings. Differences between both cohorts, such as variations in self-reported race and ethnicity, age, diabetes status and kidney measures, may have influenced direct comparisons. Although HAA showed a significant association with albuminuria, this was not replicated in an independent cohort, which limits the robustness of our finding. Furthermore, as noted, HAA may reflect other causes of microvascular injury, rather than ILD-specific pathology, potentially confounding the association with albuminuria. Given the low prevalence of nephrotic-range proteinuria in our cohort, pulmonary–renal syndromes (which are associated with severe proteinuria and diffuse alveolar haemorrhage) are unlikely. More nuanced, texture-based CT measurements will be informative in the future to distinguish water content from blood-related density changes in the lungs [42, 43]. It is important to note that antifibrotic therapies, which are known to potentially increase albuminuria, were unlikely to have influenced our findings, as these therapies were not widely approved until after 2014, which post-dates our albuminuria data collection period. Albuminuria was quantified by the results of a single spot urine test, raising questions regarding reproducibility and chronicity. In addition, the inability to replicate the MESA mortality findings in AGES-Reykjavik may be partially explained by the higher age and correspondingly elevated mortality rate in the AGES cohort. The age difference probably influenced both the overall risk profile and mortality rates, impacting the observed associations and limiting comparability between cohorts. Finally, the potential for unmeasured or residual confounding factors cannot be ruled out. For instance, although adjustments were made for diabetes and hypertension medication, including renin–angiotensin system blockers, all of which are known to influence albuminuria levels, we did not consider other drugs that may reduce albuminuria. Despite these limitations, the study provides valuable information for the field, and future research may address these limitations to further refine our understanding of the complex associations between markers of endothelial dysfunction and interstitial lung changes. Integration of biomarkers related to biological ageing and cellular senescence with omics platforms (i.e. RNA-sequencing, proteomics) are important research considerations and will require further collaborative efforts.

In conclusion, our findings did not show an independent relationship between albuminuria and ILA, with age emerging as a key confounding factor. Although these results are hypothesis-generating, there may be a common systemic pathology of ageing that underlies albuminuria and interstitial lung changes.

Supplementary material

Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.

Supplementary material

01221-2024.SUPPLEMENT

Data availability

The data used for the work are available from MESA and the Icelandic Heart Association upon approval by the MESA and the Icelandic Heart Association committees.