Effects of Long-Term Air Pollution Exposure on Disease Outcomes and Hybrid Immune Responses in SARS-CoV-2 Breakthrough Infections: A Study of the Yichang COVID-19 Antibody Longitudinal Survey (YC-CALS) in China

Abstract

Background:

As immunity wanes and viral mutations continue, the risk of endemic SARS-CoV-2 breakthrough infections (BTIs) remains. Air pollution is considered a risk factor for respiratory infection, but evidence of its association with SARS-CoV-2 BTIs is limited.

Objectives:

We aimed to examine the effects of long-term exposure to air pollution on disease outcomes, immune responses, and antibody dynamics of SARS-CoV-2 BTIs.

Methods:

We gathered data on self-reported SARS-CoV-2 infections through questionnaires and measured IgG antibody levels using serological assays from a total of 6,875 participants from the Yichang COVID-19 Antibody Longitudinal Survey cohort in China. Air pollutant exposure [particulate matter (PM) with an aerodynamic diameter $\le 2.5\mathrm{\mu }\mathrm{m}$ (${\mathrm{PM}}_{2.5}$), PM with an aerodynamic diameter $\le 10\mathrm{\mu }\mathrm{m}$ (${\mathrm{PM}}_{10}$), PM with an aerodynamic diameter $\le 1\mathrm{\mu }\mathrm{m}$ (${\mathrm{PM}}_1$), ${\mathrm{SO}}_2$, ${\mathrm{NO}}_2$, ${\mathrm{O}}_3$, and CO] was quantified using validated models for the past 5 y (2018–2022). Logistic and linear regression models were applied to analyze the associations between air pollutant levels and SARS-CoV-2 BTIs, Long COVID, COVID-19 hospitalization, and antibody responses. Quantile g-computation was used to assess the combined effects of pollutant mixtures. A linear mixed model was used to evaluate the effect of air pollution on antibody dynamics.

Results:

Per interquartile range (IQR) increase in ${\mathrm{PM}}_{2.5}$, ${\mathrm{SO}}_2$, ${\mathrm{NO}}_2$, and CO, the adjusted odds ratios (ORs) for SARS-CoV-2 BTIs were 1.65 [95% confidence interval (CI): 1.30, 2.08], 1.30 (95% CI: 1.12, 1.50), 1.63 (95% CI: 1.20, 2.20), and 1.24 (95% CI: 1.06, 1.45). The ORs for ${\mathrm{PM}}_{2.5}$ were 1.78 (95% CI: 1.07, 3.02) and 2.02 (95% CI: 1.18, 3.54) for Long COVID and hospitalization. Per IQR increase in ${\mathrm{PM}}_1$ and ${\mathrm{NO}}_2$, IgG antibody percentages decreased by $-2.31\%$ (95% CI: $-4.49\%$, $-0.13\%$) and $-2.69\%$ (95% CI: $-5.35\%$, $-0.03\%$). Effects were stronger in older adults, those with comorbidities, and the undervaccinated. The combined effect on SARS-CoV-2 BTIs was mainly driven by ${\mathrm{PM}}_{2.5}$ (59.4%), and the impact on IgG response was largely attributed to ${\mathrm{NO}}_2$ (63.7%). Exposure to the highest levels of ${\mathrm{PM}}_{2.5}$ ($p=0.002$), ${\mathrm{PM}}_1$ ($p<0.001$), and ${\mathrm{NO}}_2$ ($p=0.002$) was associated with a faster IgG decline than the lowest.

Discussion:

Long-term exposure to air pollution increases the risk of SARS-CoV-2 BTIs and disease severity while weakening the immune response, particularly for vulnerable populations. https://doi.org/10.1289/EHP15660

Introduction

The coronavirus disease 2019 (COVID-19) pandemic has caused immeasurable damage to global health. As of 2 June 2024, the cumulative number of confirmed cases worldwide had reached $775.6\text{\hspace{0.17em}million}$, with $7.05\text{\hspace{0.17em}million}$ deaths.¹ Despite global progress in vaccination, the risk of breakthrough infections (BTIs) remains. In particular, following the adjustment of China’s dynamic zero-COVID policy in early December 2022, stringent measures, such as centralized isolation, contact tracing, and mass nucleic acid testing, were relaxed.² Although 90.1% of the population had been fully vaccinated,³ a significant increase in COVID-19 cases was observed within a short period.⁴ The World Health Organization (WHO) declared the end of the global health emergency of COVID-19 in May 2023.⁵ However, it may become endemic in the future due to the ongoing mutation of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2).⁶ This persistence is further compounded by waning hybrid immunity, which includes immunity from both vaccination and post-infection recovery.⁷ Identifying key modifiable factors that may increase the risk of SARS-CoV-2 infection, exacerbate disease severity, and influence immune responses is critical to preparing for similar epidemics in the future.

Exposure to air pollution plays a critical role in respiratory infections including COVID-19.⁸ Particulate matter (PM) exposure can facilitate SARS-CoV-2 transmission by acting as an environmental vehicle, facilitating entry into the respiratory tract.⁹ Air pollution increases susceptibility to respiratory infections through several biological pathways, including disruption of respiratory barriers, effects on macrophage-mediated host defense, imbalances in the respiratory microbiota, and alterations in cell receptors that viruses use for infection.¹⁰ In addition, exposure to air pollutants may increase the severity of COVID-19 patients by exacerbating inflammatory responses and oxidative stress.¹¹ Air pollution can also affect the immune system by weakening immune responses and disrupting the regulation of inflammation.^10,12

Despite recent studies examining the potential association between air pollution exposure and the outcomes of COVID-19, some limitations remain. Most previous studies have relied on ecological analyses using population-level data, neglecting individual-level variation and potential confounders, which limits the precision and interpretation of their results.^13–15 Although a few cohort studies using individual-level data have explored the potential link between air pollution and SARS-CoV-2 infection, disease severity, or immune responses,^16–20 these inquiries were primarily focused on a limited number of air pollutants and have not accounted for the combined effects of multiple pollutants.^16–20 Moreover, these studies often overlooked crucial individual-level factors such as smoking, vaccination status, and comorbidities, as well as environmental factors like temperature, relative humidity, Normalized Difference Vegetation Index (NDVI), and population density.^16–20 In addition, the research lacked comprehensive population data to thoroughly examine disease outcomes and the hybrid immune response from SARS-CoV-2 BTIs,¹⁸ which have become more relevant in the later stages of the pandemic. Furthermore, the impact of air pollutants on the dynamics of immune response has not been adequately reported.

Four months after adjusting the dynamic zero-COVID strategy, a large-scale investigation on SARS-CoV-2 BTIs and subsequent immune responses in the general population was launched in Yichang, China. To address the aforementioned limitations, we first examined the association between long-term exposure to air pollutants and the risk of disease outcomes, including SARS-CoV-2 BTIs, Long COVID, and COVID-19 hospitalizations. We then examined the correlation between air pollutants and IgG antibody responses to SARS-CoV-2 BTIs and the effect of air pollutants on IgG antibody dynamics.

Materials and Methods

Study Design and Population

This study was based on the cohort of the Yichang COVID-19 Antibody Longitudinal Survey (YC-CALS). Yichang, with a population of $4\text{\hspace{0.17em}million}$ at the end of 2023, is a representative city in central China, characterized by a diverse population structure and rich environmental and socioeconomic conditions. The air pollution emissions from local industries and transportation may vary across different regions within Yichang. All 14 districts of Yichang were included in the investigation (Figure S1). The study employed a combination of retrospective investigation and prospective cohort methods to assess the prevalence of SARS-CoV-2 infections following the lifting of strict COVID-19 control measures in China on 10 December 2022, alongside a follow-up survey to monitor SARS-CoV-2-specific antibody levels in the population.

The survey targeted community residents, special populations (including health care workers, epidemic control personnel, and nursing home staff), and early outbreak groups (from a university in Yichang in September 2022), encompassing a total of 6,875 participants (Figure 1). A stratified cluster sampling method was used for surveying community residents in Yichang. First, 1 or 2 communities were selected from each of the 14 districts and counties in Yichang using convenience sampling. Then, based on the residents’ registration information provided by the community management departments (including name, age, and residential address), residents were surveyed using stratified sampling according to four age groups ($<20$, 20–39, 40–59, and $\ge 60$ y of age), with a total of 4,804 participants. The survey of special populations, conducted with the assistance of relevant institutions, involved the direct distribution and collection of questionnaires, totaling 1,768 participants. A retrospective survey was conducted for the early outbreak group, with 303 participants surveyed.

Figure 1.

Flow diagram describing participation in the YC-CALS cohort at each study stage. YC-CALS refers to Yichang COVID-19 Antibody Longitudinal Survey. Note: BTIs, breakthrough infections.

The enrollment survey, conducted between 20 March and 30 March 2023, was carried out by staff from the local Centers for Disease Control and Prevention and in the Community Health Care Center. The survey collected data on demographic characteristics (age, sex, height, weight, occupation, smoking status, etc.), SARS-CoV-2 infection details since 10 December 2022 (nucleic acid test results, antigen test results, symptom onset time, symptom presentation, symptom duration, health care–seeking behaviors, etc.), COVID-19 vaccination history (number of doses, vaccine type, and vaccination dates), and comorbidities (metabolic diseases, cardiovascular diseases, malignant tumors, etc.) (Table S1). A notable approach was that, in conducting the analysis, we categorized the occupational variable based on the potential SARS-CoV-2 infection risk for occupations based on potential contact with infected individuals²¹ as follows: a) low risk included occupations such as homemakers, unemployed individuals, retirees, freelancers, farmers, construction workers; b) medium risk included occupations such as transportation workers, basic industrial workers, employees of institutions and enterprises, students, and teachers; and c) high risk included occupations such as health care workers, epidemic prevention and control personnel, and nursing home staff. General demographic and infection-related information was primarily obtained through self-reports; vaccination data were retrieved from the Yichang Health care Data Platform using ID information; comorbidity information was confirmed through self-reports and diagnostic certificates or records from medical institutions. Detailed information on the survey questionnaire is provided in the Supplementary Materials. In addition, trained medical staff collected venous blood samples from participants for serological testing to measure SARS-CoV-2 antibody levels at the time of the enrollment survey.

The antibody levels in the follow-up were conducted flexibly based on different regions and follow-up times in Yichang. Specifically, based on the geographical distribution of the 14 regions in Yichang, we divided these regions into several groups (each group consisting of 1 to four regions) and conducted follow-up surveys from April to December 2023. It should be noted that the follow-up times differed among the groups. Considering the small sample sizes in other groups and the differences in follow-up schedules, our study analyzed participants from four regions within Yichang (Wujiagang, Yiling, Dangyang, and Yidu) focusing on individuals who were part of the follow-up assessments in April, June, and October 2023 (Figure 1). During the first follow-up, only blood samples were collected to measure antibody levels. From May to June 2023, Yichang experienced a second wave of small-scale COVID-19 outbreaks. Therefore, during the second and third follow-ups, in addition to collecting blood samples, a brief questionnaire was conducted to assess individual SARS-CoV-2 reinfection.

The study protocol was approved by the Clinical Research Review Board of Huazhong University of Science and Technology Tongji Medical College (registration number: 2021-IEC-A209). All participants signed an informed consent form.

Study Participants

Among the 6,875 participants, 7 did not report height and 8 did not report weight. For these 15 participants, we imputed the missing values using the mean for the same age group and gender. In addition, six participants who reported SARS-CoV-2 BTIs did not report the date of occurrence, so we imputed the date using the median value from other participants with available data. Participants with complete data were divided into three categories in our analysis (Figure 1). We focused on SARS-CoV-2 BTIs occurring after the lifting of strict COVID-19 control measures on 10 December 2022. We excluded 303 cases of clustered infections on September 2022, as they took place prior to the policy change. In addition, we excluded 215 cases of early infections from 2020, considering that individuals with prior infections in early 2020 might have experienced reinfection after December 2022, which would introduce confounding by mixing the effects of air pollution on reinfections, and that the individuals with a first infection in 2020 and both vaccination and reinfection in 2022 may have had a different immune background from those with only vaccination and infection in 2022. Such difference may affect the impact of air pollution on antibody levels. We also excluded 61 unvaccinated individuals. In total, 6,296 participants were included in the analysis of the association between air pollution and BTIs. Moreover, 5,497 participants with SARS-CoV-2 BTIs were included in the analysis of the association between air pollution and Long COVID, COVID-19 hospitalization, and immune response, excluding 799 individuals who remained uninfected in the short-term following the lifting of strict COVID-19 control measures on 10 December 2022. In addition, from the 1,531 participants with SARS-CoV-2 BTIs from four specific regions in Yichang included in the antibody follow-up cohort, we excluded 340 (22.2%) individuals lost to follow-up and 211 (17.7%) cases of reinfection during the follow-up period (because reinfection impacts antibody levels). Ultimately, 980 participants with SARS-CoV-2 BTIs who completed all three serological follow-ups were included in the analysis of the association between air pollution and immune response dynamics.

Antibody Detection

The iFlash SARS-CoV-2 IgG chemiluminescent immunoassay kit (C86095G; YHLO Biotech) was used to measure IgG antibodies against SARS-CoV-2. The sensitivity and specificity of the kit are 90% and 95% for IgG. For COVID-19 screening, a combination of nucleocapsid (N) protein and spike (S) glycoprotein was used as the coated antigen to enhance sensitivity. IgG levels were positively correlated with relative light units (RLU) and expressed in arbitrary units per milliliter (AU/mL). Serum samples from both healthy individuals and confirmed COVID-19 patients were tested according to the manufacturer’s instructions.

Outcome Assessment

SARS-CoV-2 BTIs were defined as self-reported associated symptoms or self-reported positive nucleic acid and/or antigen tests or among vaccinated individuals. This definition applied during the 4 months from the date of lifting the dynamic zero-COVID policy on 10 December 2022 to the end of the enrollment survey on 30 March 2023. Specifically, based on the COVID-19 clinical management guidelines released by the WHO,²² we inquired whether the participants had experienced typical symptoms such as fever, cough, sore throat, and fatigue after December 2022. We also inquired whether individuals had conducted self-testing for nucleic acid and/or antigen. A notable finding was that not all individuals performed self-tests, which may have occurred only when they experienced relevant symptoms or had an infected person in their household. Given the rapid spread of the Omicron variant and the near absence of infections in China before the cancellation of the dynamic zero-COVID policy, the risk of infection was high among those who developed related symptoms after 10 December 2022, essentially qualifying them as infection cases. Detailed diagnostic information on SARS-COV-2 infection is presented in Figure S2. Based on the definition from the WHO,²³ Long COVID was defined as the persistence of any symptoms related to SARS-CoV-2 infection, such as cough, fatigue, and loss of taste, for more than 2 months following the initial infection. COVID-19 hospitalization was defined based on self-reporting by participants, who then provided corresponding medical records and supporting documentation. The time frame for investigating Long COVID and hospitalization was consistent with that for SARS-CoV-2 BTIs. IgG antibody levels were examined in participants in the enrollment survey, as well as dynamic changes in antibody levels in participants with three completed follow-up visits. The outcomes are shown in Figure S3 and Figure 1.

Air Pollution Assessment

Air pollutant exposure concentrations for ${\mathrm{PM}}_{2.5}$ (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{PM}}_{10}$ (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{PM}}_1$ (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$) (only for 2018–2021), ${\mathrm{NO}}_2$ (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{SO}}_2$ (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{O}}_3$ (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), and CO (${\mathrm{mg}/\mathrm{m}}^3$) from 2018 to 2022 at a $1\text{-km}²$ high resolution were obtained from the ChinaHighAirPollutants (CHAP) dataset (https://weijing-rs.github.io/product.html). The predicted concentrations of each pollutant in the CHAP dataset were derived based on ground monitoring data, satellite remote sensing products, and atmospheric reanalysis, using the advanced extreme random temporal and spatial trees model. The model was evaluated using 10-fold cross-validation and showed excellent performance in predicting pollutant concentrations. The $R^2$ values were 0.92 for ${\mathrm{PM}}_{2.5}$, 0.86 for ${\mathrm{PM}}_{10}$, 0.82 for ${\mathrm{PM}}_1$, 0.84 for ${\mathrm{NO}}_2$, 0.84 for ${\mathrm{SO}}_2$, 0.87 for ${\mathrm{O}}_3$, and 0.80 for CO. The detailed data used to estimate air pollution were described in previous studies.^24–29 Each participant’s residential address was standardized to geographic latitude and longitude coordinates. Annual pollutant exposure values were assigned to participants using the nearest neighbor matching method based on their residential addresses. The 5-y average air pollution concentrations before 2023 were calculated to represent long-term air pollution exposure.

Assessment of Other Covariates

We obtained datasets from the National Earth System Science Data Center, National Science & Technology Infrastructure of China (http://www.geodata.cn), which provided average temperature, relative humidity (only for 2018–2020), and NDVI at a $1\text{-km}²$ resolution for the years 2018–2022. We computed the annual averages of temperature, humidity, and NDVI for participants’ residences. We estimated the 5-y average population density per square kilometer for each participant up to 2023, using the LandScan Global database (https://landscan.ornl.gov), which combines geospatial science, remote sensing, and machine learning to provide high-resolution ($1\mathrm{km}²$) global population data.

Statistical Analyses

We calculated the Spearman correlation coefficient for the correlation between pollutants and other environmental factors. We used a logistic regression model to assess the association between air pollutants and outcomes, including SARS-CoV-2 BTIs, Long COVID, and hospitalizations reported by participants at enrollment survey in March 2023. We expressed the results as odds ratios (ORs) and 95% confidence intervals (CIs) for an interquartile range (IQR) increase in air pollutants. We identified confounding factors based on a directed acyclic graph (Figure S4). The impact of single-pollutant exposure on outcomes was estimated through four steps, with increasing adjustment for covariates: Model 1 adjusted for age groups ($<20$, 20–39, 40–59, $\ge 60$ y of age), sex (male, female), residence (urban, rural), region (north, south), and risk population (low risk, medium risk, high risk); Model 2 further adjusted for other environmental exposures, including temperature (continuous, °C), relative humidity (continuous, %), NDVI (continuous), and population density (continuous, $\text{person}/{\mathrm{km}}^2$); Model 3 further adjusted for BMI groups (normal, overweight, obese), smoking [nonsmoker, passive smoker, recently quit (ex-smokers who quit at least 1 y ago), $<10$ per day, 10–19 per day, $\ge 20$ per day], and comorbidities (none, present); and Model 4 (main model) further adjusted for vaccination doses (one, two, three, four doses), vaccine type (same type, different type) (some participants who received more than one dose may have received a mix of inactivated vaccines, protein vaccines, and adenovirus vaccines); “same type” indicates that all doses were of the same vaccine type, and “different type” indicates that at least one dose was of a different type. Participants who received only one dose were classified into the “same vaccine” group), and the interval between the last vaccination and current infection or enrollment investigation ($<3$, 3–6, 7 to $<12$, $\ge 12$ months). In the main model, we used restricted cubic splines (RCS) with three knots for each pollutant to explore the exposure–response relationship between pollutants and infection. We were unable to perform an RCS analysis for Long COVID and hospitalization cases because of the small sample size. Interaction analyses were performed by incorporating interaction terms between pollutants and the same covariates used in the subgroup analyses into the models. We conducted subgroup analyses on the association between air pollutants and SARS-CoV-2 BTIs based on variables such as age group, sex, smoking (nonsmoker, passive smoker, smoker), risk population, BMI group, comorbidities, residence, region, vaccination dose ($\le 2$ doses, $\ge 3$ doses), vaccine type, and the interval between the last vaccination and current infection or enrollment investigation ($<12$ months, $\ge 12$ months). We performed subgroup analyses to assess the relationship between air pollutants and Long COVID and hospitalization based on variables such as age group ($<60$, $\ge 60$ y of age), smoking (nonsmoker, smoker), risk population (low risk, medium/high risk), and BMI group (normal, overweight/obese). Due to the limited sample size of different vaccine types in the Long COVID and hospitalized populations, we did not perform subgroup analysis for vaccine type, and subgroup analyses for other variables were consistent with those for BTIs outcomes. To investigate whether the significant effects of air pollutants were confounded by other pollutants, we adopted a two-pollutant model to evaluate the independent association of each pollutant with outcomes. We also assessed the combined effects of air pollutant mixtures on outcomes in the multipollutant model using the quantile g-calculation algorithm based on the R package “qgcomp.”³⁰

We used a linear regression model to estimate the association between air pollutants and IgG antibody levels in participants who reported having SARS-CoV-2 BTIs at enrollment survey in March 2023. We expressed the results as the percentage change in IgG levels per IQR increase in pollutants. Models 1 to 3 were run in the same way as described above. In Model 4, additional factors of time from infection to sampling, Long COVID, and hospitalization were included. We performed the subgroup analysis of the association between air pollutants and IgG antibody levels used the same characteristic groupings as the BTIs subgroup analysis, and we additionally conducted subgroup analyses based on time from infection to sampling ($<50$ d, 50–99 d, $\ge 100$ d), Long COVID (none, present), and hospitalization (none, present). We also employed RCS analyses, as well as two-pollutant and multipollutant analyses, to assess this association.

We used a linear mixed model (LMM) based on the R package “glnmtab” to analyze the effect of pollutants on the dynamics of IgG antibodies in participants with four consecutive complete serological measurements of SARS-CoV-2 BTIs, excluding those reinfected during the follow-up period. We fitted the IgG levels into an LMM that incorporated subject-level random effects for both the intercept and slope across the days after infection.³¹ We divided the pollutants into four quartiles based on their distribution: low (Q1), low-middle (Q2), high-middle (Q3), and high (Q4) exposure groups. Based on Model 4, we additionally incorporated the Long COVID and hospitalization covariates, as well as the interaction between time and pollutants, into the LMM. The interaction term represented the rate of IgG level decline in different exposure groups.

Considering that SARS-CoV-2 infection was self-reported, we conducted sensitivity analyses on 3,615 participants who tested positive via nucleic acid and/or antigen testing, as well as 5,319 participants who self-reported symptoms. We further examined the effects of ${\mathrm{O}}_3$ exposure during the warm season (April–September) and cold season (October–March) to explore potential seasonal differences. Finally, given that varying levels of nonpharmaceutical interventions (e.g., city lockdowns, travel restrictions, and reduction in industrial production) were implemented in China during the COVID-19 pandemic from 2020 to 2022, which may have influenced air pollutant levels, we performed sensitivity analyses on the associations between the 3-y average concentrations of air pollutants (2020–2022) and disease outcomes. We used R (version 4.3.3; R Development Core Team) for statistical analyses, considered $p$-values $\le 0.05$ significant, and provide the R code in the Supplementary Materials.

Results

Characteristics of Participants

A total of 6,296 participants were included in the study, with an average age of 39.4 y, and 3,020 (48.0%) were male. Among them, 1,300 (20.6%) were smokers, 545 (8.7%) were obese, 902 (14.3%) had comorbidities, 4,729 participants (75.1%) had received at least three doses of the COVID-19 vaccine, and 4,677 (74.3%) participants had their last vaccination more than 12 months before their current SARS-CoV-2 infection or the late March 2023 survey. In total, 5,497 participants were identified as SARS-CoV-2 BTIs. In addition, 980 infected participants completed sequential serological assays at four time points, with median intervals of 90, 120, 191, and 302 d, from the date they were diagnosed with SARS-CoV-2 BTIs between 10 December 2022 and the end of March 2023. During the follow-up period, no additional vaccinations were administered to the participants. The characteristics of the three categories of participants [participants in enrollment survey in March 2023 ($n=\mathrm{6,296}$), participants self-reported having SARS-CoV-2 BTIs in December 2022 ($n=\mathrm{5,497}$), and participants with SARS-CoV-2 BTIs in three follow-up visits ($n=980$)] were similar, as detailed in Table 1. In addition, 126 (2.3%) of infected participants were classified as Long COVID, commonly reporting symptoms of cough [$n=63$ (50.0%)], fatigue [$n=18$ (14.3%)], and pharyngeal discomfort [$n=16$ (13.0%)]. Only 113 (2.1%) cases required hospitalization. The characteristics of participants with no infection, Long COVID, and those hospitalized are detailed in Table S2.

Table 1.

Characteristics of three participant groups in the YC-CALS study.

Categories	Participants in enrollment survey in March 2023d ($n=\mathrm{6,296}$)	Participants self-reported having SARS-CoV-2 BTIs in December 2022e ($n=\mathrm{5,497}$)	Participants SARS-CoV-2 BTIs in 3 follow-up visitse,f ($n=980$)
Current age [y ($\text{mean}\pm \text{SD}$)]	$39.4\pm 17.8$	$38.9\pm 17.2$	$39.6\pm 16.3$
Age group [$n$ (%)]
$<20$ y	1,115 (17.7)	920 (16.7)	156 (15.9)
20–39 y	2,179 (34.6)	2,020 (36.8)	338 (34.5)
40–59 y	2,006 (31.9)	1,789 (32.5)	339 (34.6)
$\ge 60$ y	996 (15.8)	768 (14.0)	147 (15.0)
Sex [$n$ (%)]
Male	3,020 (48.0)	2,523 (45.9)	508 (51.8)
Female	3,376 (52.0)	2,974 (54.1)	472 (48.2)
Smoking status [$n$ (%)]
Nonsmoker	4,886 (77.6)	4,330 (78.7)	745 (76.0)
Passive smoking	110 (1.7)	91 (1.7)	10 (1.0)
Recently quita	252 (4.0)	224 (4.1)	53 (5.4)
$<10$ per day	257 (4.1)	216 (3.9)	42 (4.3)
10–19 per day	731 (11.6)	593 (10.8)	123 (12.6)
$\ge 20$ per day	60 (1.0)	43 (0.8)	7 (0.7)
Risk populationb [$n$ (%)]
Low risk	1,834 (29.1)	1,541 (28.0)	265 (27.0)
Medium risk	3,210 (51.0)	2,790 (50.8)	494 (50.4)
High risk	1,252 (19.9)	1,166 (21.2)	221 (22.6)
BMI [$n$ (%)]
Normal ($18.5–24.9{\mathrm{kg}/\mathrm{m}}^2$)	3,988 (63.3)	3,473 (63.1)	628 (64.1)
Overweight ($25–29.9{\mathrm{kg}/\mathrm{m}}^2$)	1,763 (28.0)	1,543 (28.1)	266 (27.1)
Obese ($\ge 30{\mathrm{kg}/\mathrm{m}}^2$)	545 (8.7)	481 (8.8)	86 (8.8)
Number of comorbidities [$n$ (%)]
None	5,395 (85.7)	4,612 (83.9)	800 (81.6)
One	644 (10.2)	631 (11.5)	128 (13.1)
Two	196 (3.1)	193 (3.5)	34 (3.5)
More than three	62 (1.0)	61 (1.1)	18 (1.8)
Residence [$n$ (%)]
Urban	4,382 (69.6)	3,868 (70.4)	641 (65.4)
Rural	1,914 (30.4)	1,629 (29.6)	339 (34.6)
Region of Yicang [$n$ (%)]
North	1,647 (26.2)	1,388 (25.2)	407 (41.5)
South	4,649 (73.8)	4,109 (74.8)	573 (58.5)
Vaccination [$n$ (%)]
One dose	29 (0.5)	25 (0.4)	2 (0.2)
Two doses	1,158 (18.4)	968 (17.6)	174 (17.8)
Three doses	4,729 (75.1)	4,307 (78.4)	780 (79.6)
Four doses	380 (6.0)	197 (3.6)	24 (2.4)
Vaccine typec [$n$ (%)]
Same vaccine	6,110 (97.1)	5,395 (98.1)	966 (98.6)
Different vaccine	186 (2.9)	102 (1.9)	14 (1.4)
Time from last vaccination to infection or enrollment survey [$n$ (%)]
$<3$ months	393 (6.2)	242 (4.4)	33 (3.4)
3–6 months	74 (1.2)	36 (0.7)	4 (0.4)
$7\text{\hspace{0.17em}to}<12$ months	1,152 (18.3)	1,123 (20.4)	170 (17.3)
$\ge 12$ months	4,677 (74.3)	4,096 (74.5)	773 (78.9)
Time from BTIs of enrollment survey to each serological test {d [median (${\mathrm{P}}_{25}$, ${\mathrm{P}}_{75}$)]}
First serological test (end of March 2023)	—	90 (85, 95)	90 (85, 95)
Second serological test (end of April 2023)	—	—	120 (115, 125)
Third serological test (end of June 2023)	—	—	191 (186, 194)
Fourth serological test (end of October 2023)	—	—	302 (296, 307)

Note: YC-CALS refers to Yichang COVID-19 Antibody Longitudinal Survey. —, no data available. BMI, body mass index; BTIs, breakthrough infections.

^a “Recently quit” comprises ex-smokers who quit at least 1 y ago.

^b Low risk: occupations include homeworker, unemployed, retired, freelance, farmers, construction workers, etc. Medium risk: occupations include transportation and basic industrial workers, employees of institutions and enterprises, students, teachers, etc. High risk: occupations include health care workers, epidemic prevention and control management populations, and nursing home workers.

^c Vaccine type refers to whether individuals received the same type of vaccine or different types of vaccines. For participants receiving more than one dose, some may have received a mix of inactivated vaccines, protein vaccines, and adenovirus vaccines; “same type” indicates that all doses were of the same vaccine type, while “different type” indicates that at least one dose was of a different type. Participants who received only one dose were classified into the “same vaccine” group.

^d The enrollment survey was conducted at the end of March 2023.

^e SARS-CoV-2 BTIs were defined as self-reported positive nucleic acid or antigen tests or self-reported associated symptoms among vaccinated individuals. This definition applied during the 4 months from the date of lifting the dynamic zero-COVID policy on 10 December 2022, to the end of the enrollment survey on 30 March 2023.

^f Because the enrollment survey for SARS-CoV-2 BTIs and serology was conducted in late March 2023, serologic follow-up tests were conducted three times in April, June, and October. Excluding 211 cases who had reinfections during the follow-up period.

Air Pollution Exposure

The exposure levels of the 6,296 participants to air pollutants and other environmental factors are presented in Table 2. The median (IQR) concentrations of the seven air pollutants were: $42.16\left(6.64\right){\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ for ${\mathrm{PM}}_{2.5}$, $66.44\left(10.06\right){\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$, for ${\mathrm{PM}}_{10}$, $21.76\left(5.96\right){\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ for ${\mathrm{PM}}_1$, $9.50\left(1.00\right){\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ for ${\mathrm{SO}}_2$, $26.94\left(7.72\right){\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ for ${\mathrm{NO}}_2$, $0.87\left(0.05\right){\mathrm{mg}/\mathrm{m}}^3$ for CO, and $95.36\left(4.76\right){\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ for ${\mathrm{O}}_3$. The concentrations of ${\mathrm{PM}}_{2.5}$, ${\mathrm{PM}}_{10}$, ${\mathrm{NO}}_2$, and ${\mathrm{O}}_3$ exceeded the standards³² set by the 2021 WHO Air Quality Guidelines. The Spearman correlations among air pollutants and other environmental factors are shown in Figure S5.

Table 2.

Distribution of residential air pollutant exposure and other environmental factors among participants in the enrollment survey of the YC-CALS study ($n=\mathrm{6,296}$).

Exposure	$\text{Mean}\pm \text{SD}$	Median (${\mathrm{P}}_{25}$, ${\mathrm{P}}_{75}$)	IQR
Air pollutants
${\mathrm{PM}}_{2.5}$ (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$)	$39.33\pm 5.66$	42.16 (36.54, 43.18)	6.64
${\mathrm{PM}}_{10}$ (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$)	$62.24\pm 8.21$	66.44 (57.68, 67.74)	10.06
${\mathrm{PM}}_1$ (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$)	$21.99\pm 3.95$	21.76 (19.72, 25.68)	5.96
${\mathrm{SO}}_2$ (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$)	$9.62\pm 0.62$	9.50 (9.16, 10.16)	1.00
${\mathrm{NO}}_2$ (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$)	$25.88\pm 5.27$	26.94 (22.44, 30.16)	7.72
CO (${\mathrm{mg}/\mathrm{m}}^3$)	$0.85\pm 0.06$	0.87 (0.83, 0.88)	0.05
${\mathrm{O}}_3$ (${\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$)	$95.62\pm 3.98$	95.36 (93.48, 98.23)	4.76
Other environmental factors
Population density (person/${\mathrm{km}}^2$)	$\mathrm{2,985.49}\pm \mathrm{8,001.81}$	632.90 (185.40, 1,728.60)	1,543.20
NDVI	$0.47\pm 0.14$	0.48 (0.36, 0.58)	0.23
Relative humidity (%)	$69.63\pm 3.69$	68.63 (66.76, 72.98)	6.22
Temperature (°C)	$17.50\pm 0.95$	17.79 (17.57, 17.93)	0.36

Note: YC-CALS refers to Yichang COVID-19 Antibody Longitudinal Survey. CO, carbon monoxide; IQR, interquartile range; NDVI, Normalized Difference Vegetation Index; ${\mathrm{NO}}_2$, nitrogen dioxide; ${\mathrm{O}}_3$, ozone; PM, particulate matter; SD, standard deviation; ${\mathrm{SO}}_2$, sulfur dioxide.

Air Pollution and Outcomes of SARS-CoV-2 Breakthrough Infections

The effects of exposure to air pollutants on SARS-CoV-2 BTIs, Long COVID, and hospitalization are shown in Figure 2 and Table S3–S5. Among the 6,296 participants, all pollutants were positively associated with BTIs in Model 1. ${\mathrm{PM}}_{2.5}$, ${\mathrm{SO}}_2$, ${\mathrm{NO}}_2$, and CO were associated with BTIs in Model 4. Specifically, the adjusted OR per IQR was 1.65 (95% CI: 1.30, 2.08) for ${\mathrm{PM}}_{2.5}$, 1.30 (95% CI: 1.12, 1.50) for ${\mathrm{SO}}_2$, 1.63 (95% CI: 1.20, 2.20) for ${\mathrm{NO}}_2$, and 1.24 (95% CI: 1.06, 1.45) for CO. Among the 5,497 infected individuals, ${\mathrm{PM}}_{2.5}$ was positively associated with Long COVID in Models 1–4. ${\mathrm{PM}}_{2.5}$ had an OR of 1.78 (95% CI: 1.03, 3.03) in Model 4. ${\mathrm{PM}}_{2.5}$ and ${\mathrm{PM}}_{10}$ were associated with hospitalization in Model 4, with ORs of 2.20 (95% CI: 1.18, 3.54) and 2.02 (95% CI: 1.20, 3.42). A notable finding was that ${\mathrm{O}}_3$ was negatively associated with hospitalization, with an OR of 0.65 (95% CI: 0.46, 0.90). There was no clear evidence showing that the relationship between air pollutants and BTIs deviates from linearity, especially in the most common exposure range (the central 50% of measured concentrations: ${\mathrm{PM}}_{2.5}$, $36.54–43.18\mathrm{\mu }\mathrm{g}/\mathrm{m}³$; ${\mathrm{NO}}_2$, $22.44–30.16\mathrm{\mu }\mathrm{g}/\mathrm{m}³$; ${\mathrm{SO}}_2$, $9.16–10.16\mathrm{\mu }\mathrm{g}/\mathrm{m}³$; CO, $0.83–0.88\mathrm{mg}/\mathrm{m}³$) (Figure S6).

Figure 2.

Associations of air pollutants exposure with SARS-CoV-2 BTIs, Long COVID, and hospitalization in YC-CALS study based on the single-pollutant model. YC-CALS refers to Yichang COVID-19 Antibody Longitudinal Survey. The numeric summary results supporting all the results in Figure 2 are presented in Tables S3–S5. The characteristics of participants included in the main analysis are presented in Table S2. Estimates are presented as ORs and 95% CIs of SARS-CoV-2 BTIs based on IQR increases in ${\mathrm{PM}}_{2.5}$ ($6.64{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{PM}}_{10}$ ($10.06{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{PM}}_1$ ($5.96{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{NO}}_2$ ($7.72{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{SO}}_2$ ($1.00{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), CO ($0.05{\mathrm{mg}/\mathrm{m}}^3$), and ${\mathrm{O}}_3$ ($4.76{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$). Estimates are presented as ORs and 95% CIs of Long COVID and hospitalization based on IQR increases in ${\mathrm{PM}}_{2.5}$ ($5.54{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{PM}}_{10}$ ($8.26{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{PM}}_1$ ($5.76{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{NO}}_2$ ($7.46{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{SO}}_2$ ($1.00{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), CO ($0.04{\mathrm{mg}/\mathrm{m}}^3$), and ${\mathrm{O}}_3$ ($4.68{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$). Analyzed using logistic regression models. Dots represent the point estimates of OR values, and short vertical lines indicate the 95% CI. Different colors and line types represent different models, labeled as Model 1 through Model 4. Model 1, adjusted for age, gender, residence, region, and risk population. Model 2, further adjusted for other environmental exposures, including temperature, relative humidity, NDVI, and population density based on Model 1. Model 3, further adjusted for BMI, comorbidities, and smoking based on model 2. Model 4, further adjusted for vaccination doses, vaccine type, and the interval between the last vaccination and current infection or enrollment investigation based on Model 3. Note: BMI, body mass index; BTIs, breakthrough infections; CI, confidence interval; CO, carbon monoxide; IQR, interquartile range; NDVI, Normalized Difference Vegetation Index; ${\mathrm{NO}}_2$, nitrogen dioxide; ${\mathrm{O}}_3$, ozone; OR, odds ratio; PM, particulate matter; ${\mathrm{SO}}_2$, sulfur dioxide.

The findings of the subgroup analyses on the associations between air pollutants and BTIs, Long COVID, and hospitalization are presented in Figures S7–S9 and Excel Tables S1–S8. In comparison with participants $<20$ y of age, exposure to air pollutants was associated with a higher risk of BTIs in those $\ge 60$ y of age ($p<0.05$), with ORs for ${\mathrm{PM}}_{2.5}$ of 1.52 (95% CI: 0.99, 2.35) vs. 3.51 (95% CI: 2.07, 6.04), for ${\mathrm{SO}}_2$ of 1.26 (95% CI: 0.94, 1.71) vs. 2.29 (95% CI: 1.63, 3.26), and for CO of 1.35 (95% CI: 1.02, 1.79) vs. 1.98 (95% CI: 1.37, 2.91). Exposure to ${\mathrm{PM}}_{2.5}$, ${\mathrm{NO}}_2$, and CO was associated with an increased risk of BTIs among individuals with $\le 2$ vaccine doses in comparison with those with $\ge 3$ doses ($p<0.05$). Exposure to ${\mathrm{PM}}_{2.5}$ and ${\mathrm{PM}}_{10}$ was associated with a higher risk of hospitalization in individuals with comorbidities ($p<0.05$).

The associations between dual pollutants and SARS-CoV-2 infection outcomes are shown in Figures S10–S12 and Excel Tables S9–S16. In the two-pollutant models for ${\mathrm{PM}}_{2.5}$ and ${\mathrm{SO}}_2$, the associations with BTIs remained significantly positive after adjusting for all pollutants. The associations between ${\mathrm{PM}}_{2.5}$ and Long COVID remained significantly positive after adjusting for all pollutants. The same finding was also observed with hospitalization as an outcome. Tables S6–S7 present the results of the multipollutant models for BTIs and hospitalization. A one-quarter increase in the mixture of ${\mathrm{PM}}_{2.5}$, ${\mathrm{SO}}_2$, ${\mathrm{NO}}_2$, and CO was significantly associated with an OR of 1.87 (95% CI: 1.53, 2.29) in the risk of BTIs. The combined effect of mixed pollutants was predominantly positive, with a sum of positive coefficients of 0.63. The positive effect was mainly due to ${\mathrm{PM}}_{2.5}$ (59.4%), followed by ${\mathrm{SO}}_2$ (24.8%). Multipollutant analysis of ${\mathrm{PM}}_{2.5}$, ${\mathrm{PM}}_{10}$, and ${\mathrm{O}}_3$ for hospitalization outcomes indicated predominantly positive effects, mainly driven by ${\mathrm{PM}}_{10}$ (73.1%), with an OR of 2.16 (95% CI: 1.49, 3.13).

Air Pollution and IgG Antibody Response in SARS-CoV-2 Breakthrough Infections

The effects of exposure to air pollutants on IgG antibody response among 5,497 infected participants are shown in Table 3. In Model 4, the percentage changes per IQR were $-2.31\%$ (95% CI: $-4.49\%$, $-0.13\%$) for ${\mathrm{PM}}_1$, and $-2.69\%$ (95% CI: $-5.35\%$, $-0.03\%$) for ${\mathrm{NO}}_2$. There was no clear evidence showing that the relationship between pollutants and IgG antibody response deviates from linearity (Figure S13). The subgroup analysis showed that ${\mathrm{PM}}_1$ and ${\mathrm{NO}}_2$ exposure significantly reduced IgG antibody levels in individuals $\ge 60$ y of age compared with those $<20$ ($p=0.028$) and in smokers in comparison with nonsmokers ($p=0.034$) (Figures S14 and Excel Tables S17–S18). In the dual-pollutant analysis, the correlations for ${\mathrm{PM}}_1$ (adjusted for other pollutants, excluding ${\mathrm{NO}}_2$ and CO) and for ${\mathrm{NO}}_2$ (adjusted for other pollutants, excluding ${\mathrm{PM}}_1$, ${\mathrm{SO}}_2$, and CO) remained significant (Figure S15 and Excel Tables S19–S20). A one-quarter increase in the mixture of ${\mathrm{PM}}_1$ and ${\mathrm{NO}}_2$ was significantly associated with a $-1.97\%$ (95% CI: $-3.29\%$, $-0.69\%$) decrease in mean IgG level, with ${\mathrm{NO}}_2$ contributing 63.7% of the effect (Table S8).

Table 3.

Association between air pollutant concentrations and IgG antibody levels in participants reported having SARS-CoV-2 BTIs in YC-CALS study based on the single-pollutant model ($n=\mathrm{5,497}$).

Air pollutants	% change (95% CI)a	$p$-Value*	% change (95% CI)b	$p$-Value*	% change (95% CI)c	$p$-Value*	% change (95% CI)d	$p$-Value*
${\mathrm{PM}}_{2.5}$	$-0.89$ ($-1.73$, $-0.05$)	0.038	$-0.34$ ($-2.29$, 1.61)	0.730	$-0.56$ ($-2.49$, 1.37)	0.572	$-0.39$ ($-2.31$, 1.54)	0.693
${\mathrm{PM}}_{10}$	$-0.94$ ($-1.79$, $-0.10$)	0.029	$-0.45$ ($-2.37$, 1.48)	0.650	$-0.67$ ($-2.58$, 1.24)	0.490	$-0.57$ ($-2.46$, 1.33)	0.557
${\mathrm{PM}}_1$	$-2.05$ ($-3.34$, $-0.76$)	0.002	$-2.22$ ($-4.42$, $-0.01$)	0.049	$-2.42$ ($-4.61$, $-0.24$)	0.030	$-2.31$ ($-4.49$, $-0.13$)	0.037
${\mathrm{SO}}_2$	0.60 ($-0.67$, 1.86)	0.354	0.44 ($-0.89$, 1.76)	0.519	0.35 ($-0.97$, 1.66)	0.605	0.34 ($-0.96$, 1.65)	0.608
${\mathrm{NO}}_2$	$-2.38$ ($-3.68$, $-1.09$)	$<0.001$	$-3.07$ ($-5.76$, $-0.38$)	0.025	$-3.32$ ($-5.98$, $-0.66$)	0.014	$-2.69$ ($-5.35$, $-0.03$)	0.047
CO	$-0.99$ ($-1.66$, $-0.32$)	0.004	$-0.95$ ($-2.28$, 0.37)	0.160	$-1.01$ ($-2.32$, 0.30)	0.132	$-0.97$ ($-2.28$, 0.34)	0.146
${\mathrm{O}}_3$	0.16 ($-0.78$, 1.10)	0.737	0.19 ($-1.03$, 1.41)	0.761	0.17 ($-1.04$, 1.38)	0.788	0.16 ($-1.04$, 1.37)	0.791

Note: YC-CALS refers to Yichang COVID-19 Antibody Longitudinal Survey. The characteristics of participants included in the main analysis are presented in Table 1. Analyzed using linear regression models. Estimates are presented as percentage change and 95% CIs based on IQR increases in ${\mathrm{PM}}_{2.5}$ ($5.54{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{PM}}_{10}$ ($8.26{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{PM}}_1$ ($5.76{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{NO}}_2$ ($7.46{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), ${\mathrm{SO}}_2$ ($1.00{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$), CO ($0.04{\mathrm{mg}/\mathrm{m}}^3$), and ${\mathrm{O}}_3$ ($4.68{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$). % change, percentage change in IgG antibody levels; BMI, body mass index; BTIs, breakthrough infections; CI, confidence interval; CO, carbon monoxide; IgG, immunoglobulin G; IQR, interquartile range; NDVI, Normalized Difference Vegetation Index; ${\mathrm{NO}}_2$, nitrogen dioxide; ${\mathrm{O}}_3$, ozone; PM, particulate matter; ${\mathrm{SO}}_2$, sulfur dioxide. *The $p$-values represent the statistical significance of the association between a one IQR increase in air pollutant concentration and IgG antibody levels and were derived from two-sided tests of the regression coefficients in the corresponding linear regression models.

^a Model 1, adjusted for age, gender, residence, region, and risk population.

^b Model 2, further adjusted for other environmental exposures, including temperature, relative humidity, NDVI and population density based on Model 1.

^c Model 3, further adjusted for BMI, comorbidities and smoking based on Model 2.

^d Model 4, further adjusted for vaccination doses, vaccine type, the interval between the last vaccination and current infection or enrollment investigation, time from infection to sampling, Long COVID, and hospitalization based on Model 3.

Air Pollution and IgG Antibody Dynamics in SARS-CoV-2 BTIs

Figure 3 and Table S9 show the effect of pollutants on the dynamics of IgG antibodies in 980 participants with BTIs. IgG levels showed a gradual decrease over time. In comparison with the low-exposure (Q1) groups, participants with high exposure (Q4) to ${\mathrm{PM}}_{2.5}$ ($\text{coefficient}=-0.086$, $p=0.002$), ${\mathrm{PM}}_1$ ($\text{coefficient}=-0.138$, $p<0.001$), and ${\mathrm{NO}}_2$ ($\text{coefficient}=-0.086$, $p=0.002$) showed a significantly decreased pattern of IgG levels over time. However, participants with high ${\mathrm{SO}}_2$ exposure (Q4) exhibited opposite results ($\text{coefficient}=0.125$, $p<0.001$).

Figure 3.

Dynamics of IgG antibody in participants reported having SARS-CoV-2 BTIs in different exposure groups to air pollutants in YC-CALS study based on the single-pollutant model ($n=980$). YC-CALS refers to Yichang COVID-19 Antibody Longitudinal Survey. The numeric results from the linear mixed models concerning the effect of air pollutant exposure on the dynamics in IgG levels shown in Figure 3 are presented in Table S9. The characteristics of participants included in the main analysis are presented in Table 1. The different exposure groups were based on the four quartiles of the pollutants. Analyzed using linear mixed-effects models. Lines of varying types and color combinations represent the predicted levels of IgG antibodies calculated by the LMM over time. Shadowed areas represent the 95% CI. Excluding 211 cases who had reinfections during the follow-up period. Models were adjusted for age, gender, residence, region, risk population, temperature, relative humidity, NDVI, population density, BMI, comorbidities, smoking, vaccination doses, vaccine type, the interval between the last vaccination and current infection or enrollment investigation, time from infection to sampling, Long COVID, hospitalization and interaction between time and pollutants. Note: BMI, body mass index; BTIs, breakthrough infections; CI, confidence interval; CO, carbon monoxide; IgG, Immunoglobulin G; LMM, linear mixed model; NDVI, Normalized Difference Vegetation Index; ${\mathrm{NO}}_2$, nitrogen dioxide; ${\mathrm{O}}_3$, ozone; PM, particulate matter; ${\mathrm{SO}}_2$, sulfur dioxide.

Sensitivity Analysis of the Association of Air Pollution with Outcomes and IgG Antibody Response in SARS-CoV-2 BTIs

After excluding participants who did not undergo nucleic acid and/or antigen testing or tested negative, sensitivity analysis in Model 4 showed that exposure to ${\mathrm{PM}}_{2.5}$, ${\mathrm{SO}}_2$, ${\mathrm{NO}}_2$, and CO remained associated with BTIs, with ORs per IQR of 1.60 (95% CI: 1.24, 2.07), 1.24 (95% CI: 1.07, 1.43), 1.89 (95% CI: 1.36, 2.64), and 1.24 (95% CI: 1.05, 1.47) (Table S10). Among participants who tested positive, IgG percentage changes per IQR of $-2.89\%$ (95% CI: $-5.28\%$, $-0.50\%$) for ${\mathrm{PM}}_1$ and $-3.14\%$ (95% CI: $-5.88\%$, $-0.40\%$) for ${\mathrm{NO}}_2$ (Table S10). Similarly, in self-reported symptomatic participants, ${\mathrm{PM}}_1$ and ${\mathrm{NO}}_2$ exposure were linked to decreased IgG levels, with percentage changes per IQR of $-2.37\%$ (95% CI: $-4.41\%$, $-0.33\%$) and $-2.34\%$ (95% CI: $-4.68\%$, $-0.01\%$) (Table S12). For seasonal ${\mathrm{O}}_3$ exposure, significant inverse associations with hospitalization were observed in both warm and cool seasons, with ORs of 0.70 (95% CI: 0.53, 0.93) and 0.58 (95% CI: 0.35, 0.94) (Table S14), consistent with the main analysis. In addition, 3-y average air pollutant exposure (${\mathrm{PM}}_1$ for 2020–2021) remained associated with BTIs, with ORs of 1.63 (95% CI: 1.33, 2.01) for ${\mathrm{PM}}_{2.5}$, 1.23 (95% CI: 1.08, 1.39) for ${\mathrm{SO}}_2$, 1.76 (95% CI: 1.28, 2.42) for ${\mathrm{NO}}_2$, and 1.28 (95% CI: 1.08, 1.51) for CO (Table S15). The IgG percentage changes per IQR were $-2.09\%$ (95% CI: $-4.22\%$, $-0.01\%$) for ${\mathrm{PM}}_1$ and $-2.14\%$ (95% CI: $-4.94\%$, 0.65%) for ${\mathrm{NO}}_2$ (Table S15).

Discussion

To our knowledge, we are the first to assess whether long-term exposure to air pollution is associated with disease outcomes and immune responses of SARS-CoV-2 BTIs in a comprehensive population-based cohort. We found that the average exposure levels of ${\mathrm{PM}}_{2.5}$, ${\mathrm{PM}}_{10}$, ${\mathrm{NO}}_2$, and ${\mathrm{O}}_3$ in this region exceeded the WHO standards. Long-term exposure to ${\mathrm{PM}}_{2.5}$, ${\mathrm{SO}}_2$, ${\mathrm{NO}}_2$, and CO significantly increased the risk of SARS-CoV-2 BTIs. ${\mathrm{PM}}_{2.5}$ exposure was linked to higher risks of Long COVID and hospitalization. Higher levels of ${\mathrm{PM}}_1$ and ${\mathrm{NO}}_2$ were associated with lower IgG antibody levels. These effects were stronger among older adults, smokers, individuals with comorbidities, and those with fewer vaccine doses. The combined effect of air pollutants on SARS-CoV-2 BTIs and IgG antibody response was significant, with ${\mathrm{PM}}_{2.5}$ and ${\mathrm{NO}}_2$ being major contributors, respectively. Participants exposed to higher levels of specific pollutants possibly experienced more pronounced declines in IgG antibody levels over time.

Long-term air pollution significantly increased the risk of SARS-CoV-2 infection and contributed to persistent symptoms and worsening disease severity after infection. A systematic review that included studies of the early COVID-19 pandemic found an increased probability of a positive association between long-term exposure to ${\mathrm{PM}}_{2.5}$, ${\mathrm{PM}}_{10}$, ${\mathrm{O}}_3$, ${\mathrm{NO}}_2$, and CO and the incidence of SARS-CoV-2 infection (63.8%).³³ A cohort of health care workers showed that each $1\text{-}\mathrm{\mu }\mathrm{g}/\mathrm{m}³$ increase in ${\mathrm{NO}}_2$ increased SARS-CoV-2 infection probability by 2%, with no significant association for ${\mathrm{PM}}_{10}$ and ${\mathrm{O}}_3$.¹⁷ A nationwide cohort study in Denmark demonstrated a significant positive correlation between infection and exposure to ${\mathrm{PM}}_{2.5}$, ${\mathrm{NO}}_2$, and ${\mathrm{PM}}_{10}$.³⁴ Our findings are consistent with previous studies. However, previous studies have primarily focused on the early stages of the COVID-19 pandemic, rarely considering the impact of vaccination. A study from a Swedish adolescent birth cohort reported that each IQR increase in ${\mathrm{PM}}_{2.5}$ exposure raised the likelihood of Long COVID by 30%.²⁰ Research from Saudi Arabia also found a significant association between ${\mathrm{PM}}_{2.5}$ exposure and Long COVID in vaccinated individuals, with each increase in exposure linked to a 32% higher risk of developing Long COVID.³⁵ Similarly, our study observed a 78% increased risk of Long COVID per IQR increase in ${\mathrm{PM}}_{2.5}$ exposure. Our study, as well as those by Yu et al.²⁰ and Saleh et al.,³⁵ employed the WHO definition of Long COVID (symptoms lasting $\ge 2$ months).²³ However, the updated criteria proposed by the US National Academies of Sciences, Engineering, and Medicine (NASEM) define Long COVID as an infection-associated chronic condition with symptoms persisting for at least 3 months.³⁶ This variation in diagnostic criteria may lead to differences in the disease stages of included cases, thereby affecting exposure–effect estimates. Future studies should consider multiple definitions of Long COVID and perform sensitivity analyses. Furthermore, estimates of long-term ${\mathrm{PM}}_{2.5}$ and ${\mathrm{PM}}_{10}$ exposure related to COVID-19 hospitalizations align with other large-scale SARS-CoV-2-infected cohorts, although with notable inconsistencies between studies.^{19,34,37–40} Our study found that per IQR increase in ${\mathrm{PM}}_{2.5}$ and ${\mathrm{PM}}_{10}$ exposure was associated with an OR of 2.02 for hospitalization. In comparison, Hyman et al. reported OR values of 1.15 and 1.12 for ${\mathrm{PM}}_{2.5}$ and ${\mathrm{PM}}_{10}$, respectively, in a UK cohort study.¹⁹ Chen et al. found that per IQR increase in ${\mathrm{PM}}_{2.5}$, the hospitalization risk was 1.06 in a large cohort in Ontario, Canada.³⁷ Zhang et al. reported that per $0.53{\text{-}\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase in ${\mathrm{PM}}_{2.5}$, the hospitalization risk was 1.10 in a nationwide cohort in Danish.³⁴ Several factors may explain the higher results in our study. First, our study had a small sample size, with only 113 hospitalized cases, which may lead to larger effect estimates. In addition, we adjusted for more confounding variables and considered other potential environmental factors, which may have further amplified the association. A unique finding in our study was the significant negative (protective) association between ${\mathrm{O}}_3$ and hospitalization. The role of ${\mathrm{O}}_3$ in COVID-19 severity remains unclear, with several studies assessing long-term exposure showing either no association or a negative correlation.^19,33,38,41 This effect may be due to ground-level ozone concentrations depending on other pollutants [e.g., nitrogen oxides (${\mathrm{NO}}_{\mathrm{X}}$)] and solar energy.⁴² However, our findings did not observe a strong correlation between ${\mathrm{O}}_3$ and ${\mathrm{NO}}_2$ ($r=0.03$). Another possible explanation is that unmeasured confounders related to urban environments (e.g., better health care access in urban areas with higher ${\mathrm{O}}_3$ levels) might contribute to this observed association. An ecological study focusing on the distribution of ${\mathrm{O}}_3$ and COVID-19 in China also found a protective effect, suggesting that the antiviral activity of ${\mathrm{O}}_3$ and possible stimulatory effects on the hosts’ innate defenses.⁴³ In addition, the limited sample size for COVID-19 hospitalizations in our study ($n=113$) may have also influenced the stability of the results. Future research needs to expand the sample size and consider potential confounding factors such as urbanization (e.g., health care infrastructure) and socioeconomic status to further investigate the potential relationship between ${\mathrm{O}}_3$ and COVID-19 hospitalizations.

Our study found that long-term exposure to air pollutants reduced IgG levels in SARS-CoV-2 BTIs, indicating the protective effect of the antibodies may be weakened. Two studies showed that exposure to ${\mathrm{NO}}_2$ or ${\mathrm{PM}}_{2.5}$ was associated with increased IgG levels in unvaccinated individuals with SARS-CoV-2 infection.^17,18 However, exposure to pollutants such as ${\mathrm{PM}}_{2.5}$, ${\mathrm{PM}}_{10}$, ${\mathrm{SO}}_2$, ${\mathrm{NO}}_2$, or CO in vaccinated individuals without SARS-CoV-2 infection can lead to a decrease in specific antibody or neutralizing antibody responses.^16,44 We found that ${\mathrm{PM}}_1$ and ${\mathrm{NO}}_2$ significantly reduce IgG antibody levels in participants with SARS-CoV-2 BTIs. This difference may be due to the hybrid immune response that we observed induced by infection after vaccination. It is worth noting that the proportion of individuals who completed the full vaccination (two doses) was as high as 99.5%, and even 81.1% of individuals had received more than three doses. In the context of hybrid immunity, the vaccine-induced immune response may constitute a larger proportion in comparison with the response elicited by natural infection, which may be a key factor leading to differences from previous studies. In addition, we found that high levels of ${\mathrm{PM}}_{2.5}$, ${\mathrm{PM}}_1$ and ${\mathrm{NO}}_2$ exposure were associated with a rapid decline in IgG levels over time. A similar finding was reported in a cohort study in which IgG levels were found to decrease significantly over time in uninfected individuals vaccinated with two doses in a high-exposure group.¹⁶ It should be noted that this report was based on different individual-level antibody measurements at different times, rather than repeated measurements at multiple time points from the same individuals. In contrast, our study collected antibody data through long-term follow-up and used a linear mixed model to account for individual variation, the interaction between pollutants and time, and other covariates. Thus, our results may be more robust. An interesting finding in our study was that high exposure to ${\mathrm{SO}}_2$ concentrations slowed the rate of decline in antibody levels. This result may be due to the unique biological mechanisms of this pollutant, such as the potential protective role of ${\mathrm{SO}}_2$ in lung immune responses, which is supported by an animal study.⁴⁵ Further studies with larger sample sizes and more refined analytical models are needed to fully understand the impact of ${\mathrm{SO}}_2$ on immune responses.

Air pollution exacerbates the impacts of SARS-CoV-2 infection on health, particularly for vulnerable populations, highlighting the complex interplay of environment and disease. We found that the effects of air pollutants on disease outcomes and immune responses to SARS-CoV-2 BTIs are more pronounced in the older individuals, those with comorbidities, and undervaccinated people. Previous studies also highlighted the increased susceptibility of vulnerable groups to adverse respiratory health effects of air pollution.^19,34 These results emphasized the need for regular booster doses, particularly in older people who are undervaccinated or who have had a long interval since their last vaccination. This segment of society is particularly vulnerable to detrimental respiratory health implications stemming from the synergistic effects of air pollutants and COVID-19. Exposure to multiple pollutants increases the risk of SARS-CoV-2 infection, with ${\mathrm{PM}}_{2.5}$ identified as a predominant contributor. Similar findings have been reported in studies examining the synergistic effects of air pollutants on broader adverse events.⁴⁶ We also identified ${\mathrm{NO}}_2$ as a significant contributor to reduced IgG levels. Based on evidence from an animal study, high concentrations of ${\mathrm{NO}}_{\mathrm{x}}$ exposure can enhance oxidative stress and inflammatory responses, thereby impairing immune function.⁴⁷ Associations have been observed between ${\mathrm{PM}}_{2.5}$ exposure and increased inflammation, oxidative stress, and impaired lung function, which may influence the severity and duration of COVID-19 symptoms.⁴⁸ In addition, air pollution leads to overexpression of ACE2, the receptor for SARS-CoV-2 cell entry, increasing viral load during invasion and thereby increasing the risk of severe disease.⁴⁹

We used comprehensive cohort data from late-stage pandemic surveys to capture a series of outcomes regarding the impact of long-term air pollution on SARS-CoV-2 BTIs. To minimize exposure misclassification potentially caused by population mobility, we used historical residential addresses as the basis for exposure assessment. We systematically accounted for individual-level and environmental confounders, and estimates from our sequential models supported our covariate adjustment strategy. However, we acknowledge some limitations. First, after discontinuation of mass nucleic acid testing measures in December 2022 in China, complete sample population testing results were unavailable. We defined SARS-CoV-2 infection based on participants’ self-reported symptoms or self-nucleic acid or antigen test results, which may introduce diagnostic misclassification. However, sensitivity analyses indicated results from samples testing positive were similar to primary outcomes. Second, the limited sample size of Long COVID and hospitalized cases could impact result stability. We observed Long COVID outcomes for only approximately 4 months. Vaccinated participants gained immune protection, and during the Omicron variant surge, fewer severe illnesses were observed in comparison with earlier phases of the pandemic.⁵⁰ Future research should expand sample sizes and conduct long-term follow-ups to further explore the impact of environmental factors on Long COVID. Third, we excluded individuals infected early in 2020 and those reinfected during the follow-up period (April to October 2023), which potentially introduced some selection bias. If air pollution is associated with the high risk of infection, individuals who are more sensitive to air pollutants might experience infection earlier. Although excluding these subjects could avoid the impact of reinfection effects from air pollutants exposure, the subjects without early infection may represent a population with higher tolerance to pollution. It potentially underestimated the true association between air pollution and outcomes. Moreover, although excluding reinfected individuals in the antibody dynamics analysis reduces confounding from repeated immune activation, it may also omit those prone to reinfection under pollution exposure, thereby distorting the assessment of dynamic antibody responses. Future studies should consider retaining these susceptible individuals or employing appropriate statistical models to adjust for bias arising from the “loss of susceptible individuals.” Fourth, antibody dynamics may be also influenced by short-term exposure to air pollutants. However, we currently lack air pollution exposure data for the period from March to October 2023, corresponding to the time of antibody measurements. Future studies should address this concern. Fifth, we did not assess indoor exposure to air pollutants (e.g., in transportation, homes, or offices), and emerging studies suggest complex combinations of mitigating factors such as building ventilation.⁵¹ This factor should be considered in future studies. Sixth, because of the widespread use of masks during the COVID-19 pandemic from 2020 to 2022, there may have been an overestimation of the population’s exposure levels to air pollutants, potentially affecting the accuracy of the study results. Therefore, future research should take this factor into account and further explore the potential moderating effect of mask use on air pollution exposure. Finally, due to the limited sample size and geographical area in Yichang where the 14 subregions exhibit limited variation, we employed a coarse spatial adjustment method by grouping these 14 regions into south and north. Although this approach partially alleviated bias arising from large-scale spatial differences, it may not effectively capture the subtle heterogeneity at the local level. Considering that most participants resided in urban of Yichang and were exposed to similar levels of air pollution, future studies should cover more extensive geographical coverage and use more refined spatial adjustment methods to assess the relationship between air pollution and health outcomes. Nevertheless, our study provides important insights into understanding the role of environmental factors in SARS-CoV-2 disease outcomes and immune responses.

Conclusion and Potential Implications

Long-term exposure to air pollution increased the risk of SARS-CoV-2 BTIs and disease severity and weakened the immune response, particularly among older adults, people with comorbidities, and people who were undervaccinated. These findings not only emphasize the detrimental effects of air pollution on respiratory health but also highlight the increased susceptibility of vulnerable groups. Our study adds to the evidence on the adverse effects of air pollution, especially at levels exceeding the WHO recommended limits, and underscores the urgent need for stricter air quality control measures in central China.

Conclusions and opinions are those of the individual authors and do not necessarily reflect the policies or views of EHP Publishing or the National Institute of Environmental Health Sciences.