Short-term exposure to PM_2.5 components and renal health: findings from the Veterans Affairs Normative Aging Study

Abstract

There is little evidence on the short-term impact of fine particulate matter (PM_2.5) on renal health, and the potential interactions and various influences of PM_2.5 components on renal health have not been examined. We investigated whether short-term (≤28 days) ambient PM_2.5 and 15 PM_2.5 components were associated with serum uric acid (SUA), blood urea nitrogen (BUN), estimated glomerular filtration rate (eGFR), and odds of incident chronic kidney disease (CKD) using both mixed-effect and Bayesian kernel machine regression (BKMR) models in the Normative Aging Study. This analysis included 2466 study visits from 808 older males enrolled during 1998–2016 with available data. BKMR showed positive relationships of PM_2.5 mixture with SUA and odds of CKD, and an inverse relationship with eGFR. In the 28-day exposure window, an interquartile range (IQR) increase in vanadium was associated with a 0.244-mg/dL higher SUA. IQR increases in sulfur and lead were associated with a 1.281- and 1.008-mL/min/1.73m² decrease in eGFR, respectively. The same change in sulfur was also associated with a 39% higher odds of CKD. Our findings provide solid evidence supporting short-term adverse effects of PM_2.5 on renal health and further highlight that components from oil combustion and regional pollution may be major contributors.

1. Introduction

Air pollution is a global public health issue that is estimated to contribute to ~8.9 million deaths worldwide per year [1]. Mounting epidemiological evidence suggests a relationship between ambient fine particulate matter <2.5μm in aerodynamic diameter (PM_2.5; particulate matter: PM) and impaired renal health outcomes, including declined estimated glomerular filtration rate (eGFR) and increased risk of incident chronic kidney disease (CKD) [2]. Previous studies have investigated the impacts of long-term (≥1 year) average concentrations of PM_2.5 on renal health. For instance, three large cohort studies showed that elevated annual concentrations of PM_2.5 were significantly correlated with increased risk of CKD among 1,164,057 adults enrolled in the U.S. Medicare program [3], 2,482,737 U.S. veterans [4], and 100,629 Asians [5]. In addition, Kim et al. [6] and Wang et al. [7] identified that higher monthly average PM_2.5 concentrations were related to higher risk of proteinuria in children and lower eGFR in healthy middle-aged women, respectively. Although annual average air pollution levels have been substantially lowered in the recent decade worldwide, short-term peaks of air pollution can harm human health and may cause short-term and acute changes in renal function. However, to the best of our knowledge, there is a dearth of research examining the short-term (days to weeks) effects of PM_2.5 exposures on the renal health of elderly adults.

PM_2.5 is not a homogenous pollutant but is a complex mixture of various organic compounds, ions, and metals [8, 9]. These PM_2.5 components have different physicochemical and toxicological characteristics, which may result in diverse effects on renal health and can further reflect specific sources of emissions [10]. Understanding their individual impact on renal health is imperative to provide a much clearer landscape of the nephrotoxicity of PM_2.5, which will allow policymakers to establish a more specific source-directed PM_2.5 control strategy for the aging population. Nevertheless, this knowledge gap still exists as no study has examined the toxicity of each PM_2.5 component and their renal effects by considering potential interactions and differential renal effects of the PM_2.5 mixture at the population level. Only a recent and relatively small study by Fang et al. investigated the relationship between personal exposure to PM_2.5 components in 72h and the decline of eGFR in a panel study of 71 older Chinese individuals using linear regression models [11].

Therefore, we performed this investigation to assess whether short-term (≤28 days) exposure of ambient PM_2.5 and 15 PM_2.5 components were associated with aberrant changes in three renal function biomarkers (serum uric acid [SUA], blood urea nitrogen [BUN], and eGFR) and the odds of CKD. We employed Bayesian kernel machine regression (BKMR), a popular Bayesian variable selection framework, to assess both the individual and cumulative relationship of PM components with renal health profiles. BKMR models the combined effects of different chemicals, while allowing for nonlinear effects as well as interactions among them, which may be closer to real-life scenarios [12]. Individuals are typically exposed to multiple air pollutants and their joint effects cannot be fully depicted by single- or two-pollutant linear models [13]. This investigation was conducted in the Normative Aging Study (NAS), which is an all-male study of older veterans living in the Greater Boston area.

2. Methods

2.1. Study design and population

Established in 1963, the NAS is a cohort of 2280 older male veterans from the Greater Boston area, where data on PM_2.5 components have been collected since 1998 [14, 15]. Participants were free of known chronic medical conditions at initial health screening and had detailed on-site study visits with physical examinations and questionnaires conducted every 3–5 years on a continuously rolling basis. Due to the small proportion of non-white participants in the NAS (n=29), only Caucasians were analyzed to increase the statistical power. In this cross-sectional analysis, a total of 2466 study visits from 808 participants with available data on PM levels and renal health outcomes collected during 1998–2016 were included. The NAS was approved by the institutional review boards of the Department of Veterans Affairs and Columbia University (New York, NY, USA), and each subject provided written informed consent before participation.

2.2. Data collection

Participants were asked to provide detailed information about their lifestyles, activity levels, and demographic factors as previously reported [14]. Body mass index (BMI [kg/m²]) was calculated from height and weight. Major diseases including cardiovascular diseases and diabetes were assessed based on medical histories and prior diagnoses. Blood samples were collected at each visit after overnight fasting. Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured once in each arm while the subject was seated, using a standard cuff.

2.3. Renal health outcomes

As previously described [16, 17], concentrations of SUA (mg/dL), BUN (mg/dL), and serum creatinine (Scr, mg/dL) were determined at each visit by Boehringer Mannheim (Boehringer-Mannheim Co., Indianapolis, IN, USA) / Hitachi (Hitachi Co., Tokyo, Japan) 747 analyzers. We calculated eGFR (mL/min/1.73 m²) at each visit using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [eGFR = 141 × min (Scr/0.9, 1)^−0.411 × max (Scr/0.9, 1) ^{− 1.209} × 0.993^Age] [18]. CKD cases were defined as individuals with an eGFR <60 mL/min/1.73 m².

2.4. Exposure assessments

Based on the reported composition and sources of PM_2.5 in the Greater Boston area [9], we focused on the mass of PM_2.5 (mg/m³) and 15 major species as trackers of seven pollution sources: regional pollution (sodium [Na], sulfur [S]), motor vehicles factors (black carbon [BC], iron [Fe], copper [Cu], zinc [Zn]), crustal/road dust (aluminum [Al], silicon [Si], calcium [Ca], titanium [Ti]), sea salt (chloride [Cl]), wood burning (potassium [K]), oil combustion (vanadium [V], nickel [Ni]), and other (lead [Pb]). The regional pollution is remotely transported secondary aerosols.

All concentrations of PM_2.5 and components were measured from fixed monitoring devices located at the Harvard University Countway Library (Boston, MA, USA), which was approximately 1 km from the examination site and a median distance of ~20 km away from participants’ residencies in the Greater Boston area. Given the longitudinal correlations between daily personal PM exposure (outdoor + indoor) and daily ambient concentrations were high in our study area because of a high air exchange rate of many participants’ residencies [19], we assumed that the measures of ambient PM concentrations could be the primary exposure metrics of PM for participants at their home addresses. Furthermore, because institutional agencies and policymakers have universally used ambient levels to set regulatory standards, our use of ambient PM levels is likely to facilitate policy translations. Daily PM_2.5 levels were measured using the Harvard Impactor Sampler. Daily BC levels were measured using aethalometers (Magee Scientific Inc., Berkeley, CA, USA). Daily concentrations of other PM_2.5 components were assessed using Energy Dispersive X-ray Fluorescence Spectrometers (Epsilon 5; PANalytical, Almelo, Netherlands). To estimate the short-term effects of PM exposures, we generated the PM levels on the day of each visit and used the daily values to calculate the multi-day averages for the 7, 14, 21, and 28 days prior to the study visits.

We also obtained temperature and relative humidity data from the National Weather Service Station at Logan Airport (Boston, MA, USA), located approximately 12 km from the examination center [20]. As participants’ residencies were spread out across the metropolitan area, we assumed that the monitored temperature and humidity could also serve as corresponding surrogates.

2.5. Statistical analysis

Descriptive statistics were used to summarize sociodemographic and lifestyle factors and renal health outcomes for all study visits and the visits by CKD status (yes/no).

We used two approaches to assess the association of PM_2.5 and its components with SUA, BUN, eGFR, and odds of CKD. Method 1 was to test the associations in single- (PM_2.5) and two-pollutant (PM components) models using mixed-effect linear regression for SUA, BUN, and eGFR, and logistic regression for odds of CKD. We used random participant-specific intercepts and fixed slopes accounting for the correlation of repeated measures of individuals in all models. We adjusted for basic covariates in all models including: age (years), BMI (underweight or normal weight/overweight/obese), smoking status (current/former/never smoker), alcohol intake (<2 drinks or ≥2 drinks per day), SBP, DBP, hypertension, stroke, coronary heart disease, diabetes, seasons of the visit (warm [Apr. – Sep.]/cold [Oct. – Mar.]), ambient temperature (°C), and relative humidity (%). Corresponding PM_2.5 concentrations were additionally added in two-pollutant models for PM components. Yielded estimates of PM on renal function traits were reported as changes per interquartile range (IQR) increase in exposure.

Method 2 was to select predictable PM_2.5 components for each renal health measure using BKMR, and then to estimate the effects of predictors simultaneously using mixed-effect linear or logistic regression models with random participant-specific intercepts. This statistical algorithm allows for a variable selection step and provision of measures of variable importance, which extracts components that might be predictive based on posterior inclusion probabilities (PIPs). The BKMR model was defined as:

$$E\left(Y_{ij}\right)=h\left(P{M}_{2.5ij},PM{\mathit{\text{ components}}}_{ij}\right)+\beta X_{ij}^{\text{'}}+{ϵ}_{ij},$$

where Y_ij is one of the four renal function measures of subject i at visit j, the function h() is a concentration-response function containing nonlinear and/or interactions of PM mass and components for subject i at visit j, $X_{ij}^{\text{'}}$ are all the basic covariates considered in the first approach, and ϵ_ij are the residuals. Components with a PIP >0.5 were considered potential predictors and simultaneously controlled for in subsequent mixed-effect linear or logistic regression models. Corresponding concentration-response relationships of the PM_2.5 mixture and each component were also estimated in the BMKR.

Additionally, we conducted four sensitivity analyses to test the robustness of our primary findings based on BKMR-selected air pollutants. First, as healthier study participants are more likely to participate in subsequent examinations over time, to evaluate validity of the missing at random assumption and assess the impact of potential selection bias caused by non-random unavailability for follow-up, we used inverse probability weighting (IPW) to correct for this potential survival bias as a sensitivity analysis [21]. Weighted models, simultaneously adjusting for the inverse probability weights and the previously introduced covariates, were used in the models for this sensitivity analysis. Second, to clarify whether confounding from other major air pollutants, such as ozone (O₃), nitrogen oxide (NOx), carbon monoxide (CO), and sulfur dioxide (SO₂) could influence our findings, we collected the corresponding levels of O₃, NO, NO₂, CO, and SO₂ to perform another sensitivity analysis. The models controlled for corresponding O₃, NO, NO₂, CO, and SO₂ concentrations with the basic covariates. Similarly, as socioeconomic position (SEP) may be another confounder for renal function decline, we also did another sensitivity analysis with an additional adjustment of the years of education, a commonly used surrogate of SEP, to test the impact of SEP on our main findings. We note that in this cohort mostly comprised of retirees, years of education is a better SEP indicator than annual income and has been consistently used to adjust for SEP status in previous analyses. Last, as our participants were elderly adults that may have higher proportions of CKD, we conducted a sensitivity analysis in CKD-free participants only to exclude the impact of CKD on our primary findings.

SAS version 9.4 TS1M5 (SAS Institute Inc., Cary, NC, USA) and R Version 4.0.2 (R Core Team, Vienna, Austria) with the ‘bkmr’ package were used to perform data cleaning and all analyses. A two-sided <0.05 p-value was considered statistically significant.

3. Results

3.1. Participants’ characteristics and air pollution distribution

Participants’ age (mean±standard deviation) across all study visits (n=2466) was 75.7±7.2 years (Table 1). About 64% of the visits were from former smokers, and 32% from never smokers. The majority of observations were from participants who were obese or overweight, consumed <2 drinks per day, and had hypertension. Nearly 30% of visits were from participants with CKD, who were considerably older and had higher proportions of other comorbidities and higher levels of SUA and BUN. Average PM levels and weather variants over the 28-day exposure window remained stable (Table 2). The 28-day average concentrations were 9.27±3.08 μg/m³ (IQR=4.09) for PM_2.5, 0.68±0.19 μg/m³ (IQR=0.29) for BC, 12.81±7.88°C (IQR=12.86) for temperature, and 67.67±5.97 % (IQR=8.45) for relative humidity. Levels of PM_2.5 and all PM components except Cl were correlated with each other across the exposure windows up to 28 days before the examinations (Figures 1, S1–S4; all p-values <0.05).

Table 1.

Characteristics of NAS participants at all study visits and by status of CKD.^a

Characteristic	All study visits (n = 2466)	CKD (eGFR < 60 mL/min/1.73 m2)
		Yes (n = 707)	No (n = 1759)	P-value
Age (years)	75.7 (7.2)	79.1 (6.5)	74.3 (7.0)	<0.0001
Smoking status c				0.12
Current smoker	95 (3.9)	20 (2.8)	75 (4.3)
Former smoker	1589 (64.4)	448 (63.4)	1141 (64.9)
Never smoker	782 (31.7)	239 (33.8)	543 (30.9)
BMI				0.08
Underweight or normal weight (<25.0)	570 (23.1)	158 (22.3)	412 (23.4)
Overweight (≥25 to <30)	1246 (50.5)	381 (53.9)	865 (49.2)
Obese (≥30.0)	650 (26.4)	168 (23.8)	482 (27.4)
Alcohol consumption (≥2 drinks per day)	466 (18.9)	108 (15.3)	358 (20.4)	0.0036
Major diseases
Hypertension	1879 (76.2)	610 (86.3)	1269 (72.1)	<0.0001
Stroke	208 (8.4)	78 (11.0)	130 (7.4)	0.0033
Coronary heart disease	865 (35.1)	342 (48.4)	523 (29.7)	<0.0001
Diabetes	389 (15.8)	140 (19.8)	249 (14.2)	0.0005
Season of visit				0.69
Warm season (Apr.–Sep.)	1076 (43.6)	304 (43.0)	772 (43.9)
Cold season (Oct.–Mar.)	1390 (56.4)	403 (57.0)	987 (56.1)
SBP (mm Hg)	129.4 (17.6)	127.9 (19.5)	130.0 (16.8)	0.09
DBP (mm Hg)	71.6 (10.7)	68.7 (11.0)	72.8 (10.4)	<0.0001
SUA (mg/dL)	6.1 (1.5)	6.8 (1.7)	5.9 (1.4)	<0.0001
BUN (mg/dL)	20.1 (6.9)	25.5 (9.0)	18.0 (4.3)	<0.0001
eGFR (mL/min/1.73 m2)	68.3 (16.2)	47.9 (10.1)	76.5 (9.7)	<0.0001

^a:Mean values (standard deviation) for continuous variables and n (%) for categorical variables; differences in characteristics between CKD groups (Yes / No) were tested for statistical significance by the Kruskal-Wallis test (continuous variables) and chi-square test (categorical variables)

NAS, Normative Aging Study; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; SUA, serum uric acid; BUN, blood urea nitrogen

Table 2.

Distribution of PM_2.5, PM components, temperature, and relative humidity in the 28-day exposure window before the visits of NAS participants.

Exposure (Unit)		Exposure window (days)
		Same day		7		14		21		28
		Mean (SD)	IQR	Mean (SD)	IQR	Mean (SD)	IQR	Mean (SD)	IQR	Mean (SD)	IQR
Air pollutant	PM2.5 (μg/m3)	9.57 (6.31)	6.84	9.23 (3.83)	4.87	9.17 (3.34)	4.35	9.22 (3.13)	4.14	9.27 (3.08)	4.09
	Na (μg/m3, 10−1)	1.71 (1.28)	1.46	1.70 (0.79)	1.03	1.69 (0.65)	0.90	1.70 (0.60)	0.84	1.71 (0.57)	0.82
	S (μg/m3, 10−1)	9.54 (8.76)	7.63	9.08 (5.14)	5.92	9.01 (4.39)	5.39	9.09 (4.10)	5.33	9.14 (4.02)	5.18
	Black carbon (μg/m3, 10−1)	7.77 (4.18)	5.28	6.84 (2.31)	3.33	6.80 (2.02)	2.97	6.79 (1.92)	2.89	6.80 (1.86)	2.92
	Fe (μg/m3, 10−2)	6.93 (3.83)	4.01	6.27 (2.38)	2.44	6.22 (2.11)	2.19	6.18 (1.90)	2.13	6.19 (1.79)	2.10
	Cu (μg/m3, 10−3)	3.59 (2.85)	3.04	3.37 (1.47)	1.53	3.37 (1.12)	1.16	3.35 (0.94)	1.00	3.35 (0.85)	0.91
	Zn (μg/m3, 10−2)	1.18 (1.24)	0.85	1.06 (0.58)	0.61	1.06 (0.50)	0.56	1.06 (0.49)	0.52	1.06 (0.47)	0.51
	Al (μg/m3, 10−2)	4.60 (3.17)	3.38	4.50 (2.74)	2.48	4.51 (2.57)	2.16	4.53 (2.41)	2.10	4.55 (2.31)	2.21
	Si (μg/m3, 10−2)	7.06 (5.40)	4.91	6.79 (4.57)	3.97	6.82 (4.36)	3.68	6.83 (4.04)	3.49	6.87 (3.83)	3.32
	Ca (μg/m3, 10−2)	3.21 (1.81)	1.82	2.90 (1.14)	1.22	2.89 (0.92)	0.98	2.90 (0.92)	0.98	2.91 (0.87)	0.98
	Ti (μg/m3, 10−3)	3.56 (2.77)	2.44	3.32 (1.65)	1.46	3.32 (1.43)	1.37	3.30 (1.27)	1.28	3.30 (1.17)	1.22
	Cl (μg/m3, 10−2)	1.07 (4.48)	0.33	1.39 (3.88)	0.56	1.33 (2.79)	0.86	1.33 (2.38)	1.01	1.33 (2.17)	1.05
	K (μg/m3, 10−2)	3.60 (2.56)	2.08	3.90 (3.70)	1.54	3.95 (2.88)	1.37	3.92 (2.33)	1.35	3.92 (2.03)	1.28
	V (μg/m3, 10−3)	3.01 (3.47)	3.36	2.98 (2.39)	3.00	3.01 (2.30)	3.09	3.04 (2.24)	3.04	3.05 (2.18)	3.02
	Ni (μg/m3, 10−3)	2.58 (3.43)	2.63	2.57 (2.48)	2.70	2.63 (2.50)	2.68	2.67 (2.48)	2.65	2.68 (2.41)	2.53
	Pb (μg/m3, 10−3)	5.03 (3.19)	4.09	5.11 (1.94)	2.49	5.13 (1.71)	2.04	5.14 (1.58)	1.77	5.16 (1.52)	1.90
Temperature (°C)		13.01 (8.64)	13.29	12.79 (8.04)	12.94	12.77 (7.93)	12.59	12.78 (7.90)	12.86	12.81 (7.88)	12.86
Relative humidity (%)		67.82 (15.73)	24.42	67.66 (9.00)	12.70	67.69 (7.26)	10.20	67.65 (6.45)	8.92	67.67 (5.97)	8.45

PM, particulate matter; NAS, Normative Aging Study; SD = standard deviation; IQR = interquartile range

Figure 1.

Correlation matrix of 28−day average levels of PM_2.5 and PM components

3.2. Association of PM_2.5 and components with renal health (Method 1)

We first evaluated the associations of PM_2.5 and its components with SUA, BUN, eGFR, and odds of CKD using single- (PM_2.5) and two-pollutant (PM components) models (Table S1). For instance, in the 28-day exposure window, an IQR increase in PM_2.5 was associated with an increase of 0.067 mg/dL in SUA, albeit the association was not statistically significant. Among the PM components, the two-pollutant models demonstrated that Na and Cl were correlated with lower SUA levels, whereas BC, V, and Ni were correlated with higher SUA levels (Figure 2). Although PM_2.5 of up to 28 days prior to the study visits was not associated with BUN, Cu was still related to elevated BUN levels (Figure 3). PM_2.5 also had a robust association with lower eGFR, and six PM components (Na, S, BC, Cu, K, and Pb) showed strong relationships with eGFR (Figure 4). However, the PM_2.5-CKD relationship was slightly attenuated comparatively, and only Na and S were positively correlated with higher odds of CKD (Figure 5).

Figure 2.

Associations of PM_2.5 and PM components with serum uric acid in the exposure window up to 28 days

Dots: Point estimates; Error bars: Confidence intervals;

Figure 3.

Associations of PM_2.5 and PM components with blood urea nitrogen in the exposure window up to 28 days

Dots: Point estimates; Error bars: Confidence intervals;

Figure 4.

Associations of PM_2.5 and PM components with eGFR in the exposure window up to 28 days

Dots: Point estimates; Error bars: Confidence intervals;

Figure 5.

Associations of PM_2.5 and PM components with CKD odds in the exposure window up to 28 days

Dots: Point estimates; Error bars: Confidence intervals;

3.3. Association of PM_2.5 and components with renal health (Method 2)

We tested the associations of PM_2.5 and PM components for each renal health measure in the 28-day exposure window using BKMR. We observed an increase in SUA and odds of CKD with the PM_2.5 mixture (Figure 6), and higher PM_2.5 quantiles were associated with a lower eGFR.

Figure 6.

Combined effects of PM_2.5 mixture on serum uric acid, blood urea nitrogen, eGFR, and CKD odds estimated by BKMR

Dots: Point estimates; Error bars: Confidence intervals;

We further found that higher levels of PM_2.5, Na, Al, Si, Ca, V, Fe, and Cu were initially identified as predictors of higher SUA levels (PIPs >0.5, Table S2), among which PM_2.5, Al, Si, Ca, and V were verified by the subsequent mixed-effect linear model. Particularly, V was the only component that showed robust associations with SUA in both methods and demonstrated a monotonic increasing pattern (Figure 7a). In Method 2, an IQR increase in V was associated with a 0.244-mg/dL higher SUA (Table 3). No components were significantly associated with BUN (Figure 7b). Lower eGFR was predicted by higher levels of S and Pb (Figure 7c), although their inverse associations with eGFR were slightly decreased compared with the findings of Method 1 (Table 3). IQR increases in S and Pb were associated with a 1.281- and 1.008-mL/min/1.73m² decreases in eGFR, respectively. S was also the only component significantly associated with higher odds of CKD in both methods (Figure 7d, Table 3). The same increase in S was associated with 39% higher odds (95% confidence interval [CI]: 1.14 – 1.68) of CKD, as exhibited in Method 2.

Figure 7.

Univariate exposure−response functions and 95% confidence interval for 28−day average level of each air pollutant (z−scored) with the other 15 pollutants fixed at the median

Table 3.

Association of 28-day average levels of PM_2.5 and PM components with renal health outcomes in the NAS.

Renal health outcome	Air pollutant	Method 1a		Method 2b
		Estimates per IQR change (SE)	p-value	Estimates per IQR change (SE)	p-value
SUA (mg/dL)	PM2.5	0.067 (0.044)	0.13	0.214 (0.069)	0.0020*
	Na	−0.182 (0.059)	0.0021*	−0.040 (0.074)	0.59
	Al	−0.048 (0.038)	0.22	−0.564 (0.143)	<0.0001*
	Si	0.010 (0.031)	0.76	0.527 (0.118)	<0.0001*
	Ca	−0.032 (0.040)	0.42	−0.160 (0.068)	0.018*
	V	0.130 (0.063)	0.04*	0.244 (0.074)	0.0009*
	Fe	−0.009 (0.045)	0.85	−0.075 (0.069)	0.28
	Cu	−0.012 (0.038)	0.74	0.048 (0.041)	0.25
BUN (mg/dL)	PM2.5	0.011 (0.199)	0.96	−0.239 (0.251)	0.34
	V	0.393 (0.280)	0.16	0.403 (0.291)	0.17
	Zn	0.080 (0.193)	0.68	0.003 (0.201)	0.99
eGFR (mL/min/1.73 m2)	S	−2.581 (0.979)	0.0085*	−1.281 (0.598)	0.032*
	Pb	−1.384 (0.542)	0.011*	−1.008 (0.453)	0.049*
		OR per IQR change (95% CI)	p-value	OR per IQR change (95% CI)	p-value
Odds of CKD	BC	1.10 (0.89 – 1.37)	0.36	0.82 (0.62 – 1.08)	0.15
	S	1.45 (1.06 – 1.99)	0.020*	1.39 (1.14 – 1.68)	0.0008*

^a:Method 1: Mixed-effect linear / logistic models for PM_2.5 or single PM components adjusting for age, body mass index, smoking status, alcohol consumption, systolic blood pressure, diastolic blood pressure, hypertension, stroke, coronary heart disease, diabetes, season, and corresponding temperature and humidity. Corresponding PM_2.5 levels were additionally added in models for PM components (two-pollutant model).

^b:Method 2: Mixed-effect linear / logistic models for predictable air pollutants selected by BKMR. Models included all selected air pollutants simultaneously and covariates used for Method 1.

PM, particulate matter; NAS, Normative Aging Study; IQR, interquartile range; SE, standard error; SUA, serum uric acid; BUN, blood urea nitrogen; eGFR, estimated glomerular filtration rate; CKD, chronic kidney disease; OR, odds ratio; CI, confidence interval

3.4. Sensitivity analyses

Four sensitivity analyses were conducted to validate the robustness of our primary findings (Table S3). The first controlling for IPW in mixed-effect linear and logistic regression models for Method 2 yielded essentially unchanged estimates of uric measurements for each air pollutant from those in the main findings, suggesting that our results were robust and not biased by the loss to follow-up. The second analysis, which additionally controlled for the corresponding O₃, NO, NO₂, CO, and SO₂ levels, considerably weakened the estimates of several components with SUA, although vanadium was still significantly associated with an elevated SUA with a slightly attenuated effect. For eGFR and CKD odds, the estimates were essentially unchanged compared to the primary findings. Essentially unchanged estimates were also observed in sensitivity analysis 3 with additional adjustment of years of education to clarify the potential bias from SEP. The 2^nd and 3^rd analyses fully elucidated that residual bias from other air pollutants or SEP may affect our primary findings only to a limited degree. The last sensitivity analysis was conducted in CKD-free participants, effects of key PM components for SUA, BUN, and eGFR were slightly attenuated, but most were in the same patterns as they were in the primary results. This indicates that the impact of reversal causation of CKD with renal function on our study was very minor.

4. Discussion

This study is the first to investigate the PM-renal health relationship while considering PM_2.5 as a mixture of multiple chemical components. We used a tiered statistical approach applying both single- and two-pollutant models as well as BKMR to explore independent associations of PM_2.5 and its components with renal function. Among 808 older adult males, we observed significant heterogeneity in the associations between PM_2.5 and four renal health measurements across clusters defined by 15 PM_2.5 component profiles. Both approaches exhibited that in the 28-day exposure window, an elevated level of V was associated with increased SUA, and higher levels of S and Pb were associated with decreased eGFR and corresponding higher odds of CKD. Our findings provide a comprehensive characterization of the possible associations of short-term exposure PM_2.5 and its components with renal health.

Our analyses confirmed findings of previous studies on the adverse renal effects caused by acute PM_2.5 exposure and also made several new observations. Consistent with the report by Fang et al. [11], our study showed a decline in eGFR with short-term exposure of PM_2.5 and demonstrated that this relationship may extend to up to 28 days after exposure. While the authors noted that the acute effects of PM_2.5 on kidney function were transient with considerable attenuation after 24h to 72h, our study instead showed that such effects may accumulate and then remain stable for weeks. Further studies are need to address this discrepancy and determine whether the change in eGFR may be an adaptive response to acute exposure or may persist and cause somewhat permanent renal function impairment. Although several studies using medium- or long-term exposure data have suggested robust positive associations of PM levels with SUA and BUN [7, 22, 23], there is little evidence on the acute effects of PM on both biomarkers. Only two population-based studies using short-term air pollutant data found much weaker positive air pollution–SUA correlations [24, 25], both of which were in line with the results of our study showing that short-term PM_2.5 exposure has a relatively weak relationship with SUA. These data indicate that the effects of PM on SUA and BUN might be chronic and latent. Additionally, in our sensitivity analysis with corresponding levels of other air pollutants (i.e. O₃, NO, and NO₂) that may be related to renal function alteration, the key PM components still demonstrated robust associations with each measure of renal function albeit the associations of other components or PM_2.5 with renal health profiles were considerably changed. We thus believe that O₃ and NOx may operate as confounders with respect to some of the associations with renal function, which may need further explorations by future relevant studies.

Ni, another key tracer of oil combustion, also demonstrated a similar but slightly weakened association with SUA in our study as the identified V–SUA relationship. Because Ni and V are co-emitted during combustion of fuel oils [26], these patterns highlight the fact that the pollutants from oil combustion are major contributors to elevated SUA. The discrepancy between the effects of V and Ni on SUA may be related to the features of V, a transition metal that further exhibits nephrotoxic properties including glomerulonephritis and pyelonephritis [27]. A previous study with a mouse model demonstrated that the effects of V inhalation could damage kidneys by producing histological changes located mainly in the urinary tubules [28]. The authors also found that such damages may be mediated by oxidative stress and therefore may interfere with glucose regulation. We additionally noted that other unmeasured PM_2.5 components along with components representing crustal/road dust (Al, Si, and Ca) also may be associated with an increased SUA as shown by the BKMR model. Given that oil combustion represents only 8% of PM_2.5 mass in the Greater Boston area and great decreases in oil combustion source types were observed in 2010 [9], these pieces of evidence together warrant further studies of other PM_2.5 components, especially those unmeasured, to explicitly understand the full landscape of the impact of PM components on SUA.

S and Pb were correlated with lower eGFR in our study, and S was further associated with higher odds of CKD, while Pb was only marginally related to the odds. Regional pollution was the major source contributor to PM_2.5 mass (~48%), and S was characterized as a reliable tracer for the slow atmospheric conversion of sulfur dioxide (SO₂) to sulfate particles [9]. Na was associated with eGFR and the odds of CKD in Method 1 because it is another tracer of regional pollution, which can be measured through residual of sodium sulfate left on filters [9]. Both elements suggest that regional pollution may play a critical role in decreased eGFR and CKD development. Regional pollution showed seasonal variability in the Greater Boston area, especially in summer when regional pollution peaks due to the increased conversion rate of emitted SO₂ to sulfate particles. Moreover, Pb on the particles could be short-lived radon progeny (²¹⁴Pb) and long-lived thoron progeny (²¹⁰Pb) and may produce particle radioactivity (PR), which could play a major role in renal function impairment [29]. A previous study by our research group tested this hypothesis and found that short-term low-level ambient PR demonstrated a robust negative association with eGFR and was related to higher odds of CKD in participants with the highest levels of two plasma inflammatory markers, C-reactive protein and fibrinogen [17].

The major strengths of our study are that it provides detailed information on a broad range of covariates and multiple measures of PM levels and renal function, and it utilizes two distinct methods including BKMR, a novel method that could evaluate the exposure-response relationships of the PM_2.5 mixture and components. Nonetheless, the study also has some limitations. First, even though each participant had up to seven visits, the cross-sectional nature of this study curbed exploration of a causal link of PM exposure with renal function impairment. Also, we selected only 15 PM components representing sources of pollution in the Greater Boston area; thus, we were unable to investigate other unmeasured PM components on renal health, limiting the generalization of our findings to other areas with different pollutant sources. Additionally, multiple comparisons may also potentially induce false-positive findings, although we observed consistencies of the significant associations between certain exposures and renal function across each time window in the two methods – these may indicate that our findings were not incidental. Additional studies using other statistical methods, such as weighted quantile sum regression and/or group least absolute and shrinkage and selection operator (LASSO) model may help verify the robustness of our findings [13]. We further acknowledge that our analysis is subject to measurement bias in that the levels of PM obtained from a single site in Boston that we utilized may differ from that at the participants’ address and/or their personal exposure in the combination of outdoor and indoor exposures. Nevertheless, a previous Boston study found that the correlations of ambient PM_2.5 concentrations and the corresponding personal PM exposure were fairly high because many of the subjects lived in old and leaky homes with a very high air exchange rate [19]. Another measurement bias from using the data of a single site will result in primarily Berkson-type measurement error, which may bias the standard errors but not the estimated associations of our primary findings [30]. Given the majority of NAS participants are retired and spend most of their time at their residencies, we believe that the discrepancies between the city-average and personal PM data are likely to be non-differential and to bias results toward the null, rather than causing the observed associations. Lastly, participants in this study were older white men, which suggests the possibilities that the results might not be generalized to other racial/ethnic groups or/and women; however, given our findings are in line with previous relevant studies in other populations and women [22], this possibility is much less likely.

5. Conclusions

In conclusion, we observed notable associations between specific PM_2.5 components and four renal health traits using BKMR. These findings provide solid evidence supporting the short-term adverse effect of PM exposure on renal health and further highlight that pollutants from oil combustion and regional pollution may be the major contributors to renal function impairment. Overall, this study demonstrates the complexity of assessing renal health risk associated with exposure to PM_2.5 mixture. Future multidisciplinary studies based on larger cohorts are needed to validate our findings and to determine the underlying biological mechanisms of different PM components.

Supplementary Material

all supplements

Figure S1 Correlation matrix of same day levels of PM_2.5 and PM components

Figure S2 Correlation matrix of 7−day average levels of PM_2.5 and PM components

Figure S3 Correlation matrix of 14−day average levels of PM_2.5 and PM components

Figure S4 Correlation matrix of 21−day average levels of PM_2.5 and PM components

Table S1 Association of PM_2.5 and PM components with renal health outcomes in the NAS up to 28 days (Method 1)

Table S2 Posterior inclusion PIPs of BKMR fully-adjusted models for the 28-day average PM_2.5 mixture

Table S3 Association of 28-day average levels of PM2.5 and PM components with renal health outcomes in sensitivity analyses