Loss of life expectancy attributed to long-term ozone exposure in Chinese older adults: Cross-cohort analysis from 3 national cohorts

Graphical abstract

Highlights

• •A cross-cohort analysis of 3 national cohorts to investigate the ozone-mortality association in Chinese older adults.
• •Per 10-ppb increase in warm-season ozone was associated with a hazard ratio of 1.076 (95% CI: 1.050–1.102) for mortality.
• •Ozone exposure contributed to 0.88 million deaths and 9.05 million years of life lost, equivalent to a loss of life expectancy of 0.93 years in 2019.

Abstract

Background

Cohort evidence linking ozone (O₃) exposure with mortality was sparsely investigated among the elderly in low- and middle-income countries. This study aims to quantify mortality risk and burden attributed to chronic O₃ exposure in Chinese older adults.

Methods

A total of 30,874 older adults aged ≥65 years were recruited from 3 national dynamic cohorts across 29 provincial regions in China, 2005–2018. Annual warm-season (April–September) O₃ and year-round PM_2.5 concentrations were estimated through well-validated satellite-based spatiotemporal models and were assigned to participants for each survey year. Time-dependent Fragility Cox models with random intercept for study cohort were employed to quantify O₃-mortality association, adjusting for demographic, behavioral, health, and environmental covariates. A counterfactual causal framework was used for assessment of O₃-attributable premature deaths in older adults based on exposure–response relationship derived from multi-cohort two-pollutant analysis (+PM_2.5). Years of life lost and loss of life expectancy were subsequently evaluated based on the burden estimation model by incorporating the comparative risk assessment method and reference life tables.

Results

16,939 death events occurred during 0.16 million person-years of follow-up surveys. Each 10-ppb increase in O₃ exposure was linked with a hazard ratio of 1.076 (95% confidence interval [CI]: 1.050, 1.102) for all-cause mortality. By achieving the counterfactual target (WHO AQG 2021) of 60 μg/m³ for warm-season O₃, 0.88 (95% CI: 0.60, 1.14) million premature deaths could be avoidable among Chinese older population in 2019, yielding an inconspicuous reduction of 0.11 million compared to the estimate in 2011 (0.99 million, 95% CI: 0.68, 1.28). O₃-attributable deaths amounted to 9.05 (95% CI: 6.19, 11.70) million years of life lost in 2019, equivalent to a loss of life expectancy of 0.93 (95% CI: 0.63, 1.20) years for older population in China.

Conclusions

Our multi-cohort analysis suggested that reducing ambient O₃ exposure could increase the life expectancy of Chinese older adults, which may contribute to the development of healthy aging strategies and national cleaning air policies.

Introduction

As a highly reactive and potently oxidative pollutant, ambient ozone (O₃) has exhibited an upward trajectory in recent years amidst the backdrop of global warming, emerging as a significant global public health concern [38]. Inhalation of O₃ can induce the production of inflammatory cytokines and elicit an oxidative stress response within the respiratory tract [22], potentially leading to more severe damage to the respiratory system [27]. Also, the inhaled O₃ and generated inflammatory cytokines possess the potential to permeate into the blood vessels [6], causing a systemic inflammatory response and triggering hypercoagulability with differentially expressed proteins and enriched pathways [24]. The temporal variation of O₃ pollution exhibited an upward trend, peaking in 2019 with a population-weighted concentration of 119 μg/m³ [45], which was twice higher than the standard recommended standard by the World Health Organization's Air Quality Guidelines in 2021 (WHO AQG 2021). Ambient O₃ levels are typically correlated with temperature, exhibiting significantly higher concentrations during warm seasons compared to other months of the year. Moreover, O₃ concentrations tend to be markedly higher in rural locations in China [31], showing a distinct regional disparity in the health threat to Chinese older populations.

Accumulated evidence from developed countries indicates a link between chronic O₃ exposure and elevated mortality risk from cardiopulmonary diseases [32], and significantly higher susceptibility was revealed among the elderly [14], [28]. Existing high-quality cohort studies from Europe [5], [44] and North America [7], [19] revealed significant associations between O₃ exposure and elevated risks of non-accidental and cause-specific (cardiovascular and respiratory diseases) mortality. In low- and middle-income countries (LMICs), however, such cohort nexus remains insufficiently investigated. The growing air pollution-related burden of disease is anticipated to further exacerbate health inequalities among the elderly as the population ages [51], especially in LMICs such as China experiencing rapid population aging [11] and pervasive high-level O₃ pollution [42]. It is projected that the old population in China will surge to 487 million, making up at least one-third of the national population by 2050 [39]. Given the increasing growth of the elderly population in the coming decades, discerning modifiable risks that threaten their well-being is of great importance to achieve healthy longevity and reduce economic loss in a population-ageing scenario. Revealing the burden of O₃-related mortality in Chinese older adults is a crucial foundation for developing effective O₃-control strategies and embracing healthy ageing [4].

The most recent empirical research suggests that the burden of O₃-induced cardiopulmonary disease in China is expected to be exacerbated by global warming [40], [49], while the assessment of O₃-related mortality burden remains limited due to the absence of national concentration–response (C-R) functions [33]. Several studies have evaluated mortality burden of chronic obstructive pulmonary disease [45] or cardiopulmonary disease [21], [41] attributable to O₃ exposure. However, it is noteworthy that these estimations fail to incorporate Chinese population-based evidence, and there remains a dearth of information regarding spatial and temporal trends in O₃-attributed mortality burden among older adults. These gaps in knowledge pose a great obstacle against the advancement of targeted public health interventions aimed at safeguarding this vulnerable group [21].

Here, we carried out a national multi-cohort study during 2005–2018 in the Chinese mainland, aiming to depict the nationally representative C-R relationship between O₃ exposure and mortality among older adults (≥65 years). Based on this cross-cohort analysis, an indigenous estimation model was developed to assess the burden of premature deaths and loss of life expectancy (LLE) attributable to O₃ exposure among the Chinese older population from 2011 to 2019.

Methods

Study population

Study population was derived from 3 dynamic national cohorts including the Chinese Longitudinal Healthy Longevity Study (CLHLS) [50], the China Family Panel Studies (CFPS) [46], and the China Health and Retirement Longitudinal Study (CHARLS) [56]. A multi-cohort study covering 29 provincial regions of China was designed via selecting participants aged 65+ years from these cohorts. In total, 30,874 older adults were included in the multi-cohort analyses, with a follow-up duration of 2005–2018 (Fig. S1). Death information of the deceased was confirmed from community doctors or their family members in subsequent follow-up interviews. We excluded participants with missing data for covariates in multiple waves of surveys, and description for inclusion of the participants was detailed in Supplementary Material (Section 1). Specifically, considering the relatively low proportion of persons who were lost to follow-up (9.8%), we deleted these records in the cross-cohort analysis. All these participants or their legal representatives signed informed consent forms, and the Peking University Biomedical Ethics Review Committee reviewed and approved the CLHLS (No. IRB00001052–13074), CFPS (No. IRB00001052-14010), and CHARLS (No. IRB00001052-11015) surveys.

Exposure assessment

For each participant, annual warm-season (April–September) maximum daily 8-hour (MDA8) O₃ and year-round fine particulate matter (PM_2.5) concentration was calculated for each follow-up calendar year as their time-dependent exposures. For individuals from CLHLS, O₃ and PM_2.5 exposures were assigned by extracting gridded concentrations at their geocoded residential address. Residential addresses of participants from CFPS and CHARLS cohorts were concealed on account of personal privacy and security. We thus carried out exposure assessment of air pollutants at county level for CFPS participants and at city level for CHARLS participants. Gridded estimates of O₃ and PM_2.5 were aggregated into the county-level and city-level averages and assigned to participants by survey years. Annua warm-season MDA8 O₃ concentration (0.1° × 0.1° resolution) was derived from the Global Burden of Disease Study 2019 (GBD 2019, https://ghdx.healthdata.org/gbd-2019, accessed on May 7, 2024) and annual PM_2.5 concentration (0.01° × 0.01° resolution) was obtained from Tracking Air Pollution in China (TAP, http://tapdata.org.cn, accessed on May 7, 2024). Modeling details for gridded O₃ and PM_2.5 estimates could be found in Supplementary Material (Section 2).

Covariates

Individual-level information for participants was collected through computer-assisted face-to-face interviews by well-trained staff. Based on previous literatures [25], [53], we considered an extensive range of covariates including 1) demographic characteristics: sex, age, current marital status (yes or no), and education attainment (education duration less than 1 year, or between 1 and 9 years, or exceeding 9 years); 2) regional factors: geographical region (North, Midwest, or Southeast of China), residence (rural or urban), and annual average temperature; 3) behavioral factors: alcohol drinking (never, former, or current), cigarette smoking (never, former, or current), and regular exercise (no or yes); and 4) heath status: self-reported diagnoses of diabetes (absent or present), heart diseases (absent or present), hypertension (absent or present), and respiratory diseases (absent or present). Regular exercise was defined as one or more bouts of outdoor physical exercise with a moderate intensity level, lasting for at least 20 min per session in the past week. Prevalent status of hypertension, diabetes, heart diseases, and respiratory diseases was defined as either “present” or “absent”, depending on whether the participants had received a diagnosis for these conditions from a medical professional.

Statistical analyses

To provide a perspicuous description of the data, we utilized frequencies (percentages, %) for categorical variables and mean (standard deviation, SD) for continuous variables. Time-dependent Fragility Cox models were first employed to quantify chronic O₃-mortality association. Briefly, to account for spatiotemporal variations of ozone and other pollutants [18], we initially incorporated time-varying exposure at annual time scale into the Fragility Cox models in our study rather than static exposure at baseline. One observation was created for each person for each year of mortality follow-up, and the corresponding annual average exposures were subsequently assigned to each observation [1], [54]. A counterfactual causal framework was then utilized to assess O₃-attributable premature deaths in older adults, based on the C-R function derived from multi-cohort bi-pollutant analysis (+PM_2.5). Subsequently, O₃-attributable years of life lost (YLL) and loss of life expectancy (LLE) were evaluated based on the burden estimation model by incorporating the comparative risk assessment method and reference life tables.

We performed all data analyses using R software 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria). The analysis of CFPS county-level restricted data was carried out in the restrictive computer room of the Institute of Social Science Survey at Peking University.

Association between O₃ and mortality

To quantify the effect of chronic ambient O₃ exposure on all-cause mortality, Fragility Cox models with time-varying exposures were adopted in this multi-cohort analyses, where cohort source (CLHLS, CFPS, or CHARLS) was fitted as a random intercept to eliminate its cluster effect. The duration of follow-up for each participant was computed by determining their time interval from the date of enrollment in the multi-cohort to the date of the last follow-up surveys, the date of death, or the termination for participants, whichever came first. In our survival analyses, sex (female and male) and age (e.g., 65–69, 70–74, 75–79, 80–84, 85–89, 90–94, 95–99 and ≥100) were treated as the strata term in the Fragility Cox model to allow for flexible stratum-specific hazard functions [29]. We adopted several models using an approach of sequential adjustment for covariates to examine the stability of estimated association in the pooled and cohort-specific analyses. Subgroup analyses stratified by demographic and clinical variables were conducted to identify the vulnerable subpopulations. To depict the C-R relationships between mortality risk and warm-season O₃ exposure, we utilized RCS smoothing spline with three knots at the 10th, 50th, and 90th percentiles [9] both in single- and bi-pollutant models (+PM_2.5). Possible nonlinearity of C-R relationship was checked by likelihood ratio test.

Several sensitivity analyses were conducted to examine the robustness of the O₃-mortality association. First, to reduce the influence of high-risk participants, we re-ran the association analyses by excluding persons who died within the first year after enrollment of the cohort. Second, we used 2-year average estimates of O₃ and PM_2.5 concentration to account for the interannual variance in O₃ exposure. Third, to eliminate over adjustment of variables, directed acyclic graph (DAG, Fig. S2) method was employed to identify the minimal the minimal adequately adjusted set of variables in single- and bi-pollutant models. In addition, the E-values were calculated to account for the effect of unmeasured confounders on the association [37].

Estimation of O₃-attributable deaths

In coincidence with the methodological strategies widely utilized in prior modeling studies [57], [58], we quantified O₃-attributed deaths among older adults (≥65 years) in China for the years between 2011 and 2019 at 0.1° × 0.1° grids to depict the spatiotemporal trend over the recent decade (in the 2020 s). This interval (2011–2019) was highly coherent with the overlapped period of the 3 cohorts (2011–2018) in our association analysis, which should make our estimation of mortality burden more parallel with the observed data. For a given year, ozone-related premature mortality amounted to the number of deaths that could be avoided by achieving the predefined target concentration of ambient warm-season O₃ in all settings across China. Briefly, a counterfactual O₃ concentration (60 μg/m³), recommended by air quality guidelines (AQG) of World Health Organization (WHO) in 2021, was used as theorical minimum referent exposure level (TMREL), indicating no excess risks (HR = 1) for grids with warm-season O₃ <60 μg/m³ in a given year. This hypothetical O₃ level is evaluated through an integration of population-based cohort evidence from multiple regions globally, ensuring the broad demographic representation and high comparability between studies. The total attributable premature deaths (ADs) related to chronic O₃ exposure were calculated through summing up grid-specific estimates via the following equation [45]:

$${\mathit{ADs}}_{i,t,k}={\mathit{Pop}}_{i,t,k}\times {\mathit{Mort}}_{t,k}\times (1-\frac{1}{e^{\beta \times \left(C_{i,t}-TMREL\right)}})$$

where the subscripts of $i$, $t$, and $k$ represent the indexes of spatial grids, calendar year, and age- and sex-specific population group (i.e., 65–69, 70–74, 75–79, and 80+ years), respectively. Pop denotes the number of populations. Mort refers to the all-cause mortality rates. $\beta$ represents the log-linear HR derived from the bi-pollutant analysis in multivariable-adjusted model (model 4). C denotes the warm-season O₃ concentration. City-, provincial-, and national-level ADs could be yielded by summing up grid-specific estimates of $\mathit{ADs}$. The attributable premature deaths refer to deaths that occur before life expectancy due to exposure to ambient O₃. Age- and sex-specific gridded estimates of older population (≥65 years) in China at a resolution of 0.1° × 0.1° was resampled from the global population database at 0.0083° × 0.0083° resolution released by the WorldPop (https://www.worldpop.org/, accessed on April 24, 2024). Age- and sex-specific all-cause mortality rates among Chinese older adults in each year were estimated by the GBD 2019 study (Table S1). Several key regions (the BTH, the Beijing-Tianjin-Hebei and surrounding areas; the YRD, the Yangtze River Delta; the JASH, the juncture of Jiangsu-Anhui-Shandong-Henan; the CYD, the Cheng-Yu District; the TCC, the Triangle of Central China; the FWP, the Fen-Wei Plain; and the GBA, the Greater Bay Area) were geographically emphasized in the spatial analysis owing to its critical importance in China's air pollution control policies.

Estimation of O₃-attributable LLE

Based on the analysis of O₃-attributable deaths, we estimated the LLE in 2011 and 2019, which should serve as a more informative metric in quantifying the mortality burden at the individual level. By combining ${\mathit{ADs}}_{i,t,k}$ and the reference life table derived from GBD 2019 study, we firstly calculated the YLL through the equations as follows [48]:

$${\mathit{YLL}}_{i,t,k}={\mathit{ADs}}_{i,t,k}\times {\mathit{LE}}_k,{\mathit{YLL}}_t={\sum }_i\sum _k{\mathit{YLL}}_{i,t,k}$$

where ${\mathit{YLL}}_{i,t,k}$ and ${\mathit{ADs}}_{i,t,k}$ refers to the estimate of O₃-attributable YLL and deaths for $i^{\mathit{th}}$ grids, $t^{\mathit{th}}$ year, and $k^{\mathit{th}}$ subpopulation, respectively. ${\mathit{LE}}_k$ denotes the theorical life expectancy at a specific age for the $k^{\mathit{th}}$ subpopulation (Table S2). ${\mathit{YLL}}_t$ signifies the total years of life lost attributable to O₃ exposure for $t^{\mathit{th}}$ year.

For a given year, LLE for an individual aged 65 years and older is equivalent to a rescaled average of YLL and was evaluated via the equation $\mathit{LLE}=YLLs/Pop\times {\mathit{LE}}_0$, where $\mathit{Pop}$ represents the total population aged ≥65 years, and ${\mathit{LE}}_0$ denotes the theorical life expectancy at 65 years [15]. The estimation of LLE could provide valuable insights into the chronic impact of O₃ exposure on population longevity, and could help policymakers prioritize interventions for effectively improving overall life expectancy.

Results

The pooled analysis integrated three national cohorts across the Chinese mainland. Within this collective analysis, we investigated nearly 160 thousand person-years (median 4.9 years), and 16,939 from 30,874 older adults aged ≥65 reported death events (Table 1). Summary descriptions of specific cohorts were presented in the supplementary materials (Tables S3–S5). Participants were averagely aged 82.1 (SD: 12.2) years and around a half (55.1%) were female. The average air pollution exposures for participants during the study period were 53.2 ppb (SD: 8.1 ppb) for warm-season O₃, and 54.4 μg/m³ (SD: 17.2 μg/m³) for year-round PM_2.5 (Table S6).

Table 1.

Descriptive characteristics of study population in the multi-cohort at baseline.

Variables	All participants	All-cause death
		Yes	No
Population
Persons, n	30,874	16,939	13,935
Total person-years	159,947	62,847	97,090
Median year of follow-up	4.9	2.9	7.1
Demographic characteristics, n (%)
Age (years), mean ± SD	82.1 ± 12.2	88.3 ± 11.0	74.4 ± 8.8
Male sex	13,856 (44.9)	7327 (43.3)	6529 (46.9)
Married	13,629 (44.1)	4820 (28.5)	8809 (63.2)
Education attainment (years)
0	18,144 (58.8)	11,449 (67.6)	6695 (48.0)
1–9	10,993 (35.6)	4901 (28.9)	6092 (43.7)
9+	1737 (5.6)	589 (3.5)	1148 (8.2)
Regional factors, n (%)
Geographical region
North	5900 (19.1)	2727 (16.1)	3173 (22.8)
Midwest	9386 (30.4)	5162 (30.5)	4224 (30.3)
Southeast	15,588 (50.5)	9050 (53.4)	6538 (46.9)
Rurality residence	18,309 (59.3)	10,517 (62.1)	7792 (55.9)
Behavioral factors, n (%)
Cigarette smoking
Never	16,499 (53.4)	10,331 (61.0)	6168 (44.3)
Former	4264 (13.8)	2646 (15.6)	1618 (11.6)
Current	10,111 (32.7)	3962 (23.4)	6149 (44.1)
Alcohol drinking
Never	15,760 (51.0)	10,124 (59.8)	5636 (40.4)
Former	3341 (10.8)	2128 (12.6)	1213 (8.7)
Current	11,773 (38.1)	4687 (27.7)	7086 (50.9)
Regular exercise	11,329 (36.7)	4626 (27.3)	6703 (48.1)
Health factors, n (%)
Diabetes	2727 (8.8)	1665 (9.8)	1062 (7.6)
Hypertension	7810 (25.3)	4372 (25.8)	3438 (24.7)
Heart diseases	4351 (14.1)	2530 (14.9)	1821 (13.1)
Respiratory diseases	4253 (13.8)	2654 (15.7)	1599 (11.5)

Notes: The sum of proportions may not equal 100% exactly due to the use of rounding-off method. SD, standard deviation.

Table 2 estimates the associations of warm-season O₃ with all-cause mortality across multiple models. Sequential adjustment for variables did not substantially alter the estimated effect magnitude, suggesting robust evidence for heightened mortality risk associated with chronic O₃ exposure. Specifically, in the multivariable-adjusted model (model 4), we associated an HR of 1.081 (95% CI: 1.061, 1.102) with a 10-ppb increase in warm-season O₃ exposure through single-pollutant analyses. When adjusting for annual PM_2.5 exposure in model 4, a slightly lower risk of 1.076 (95% CI: 1.050, 1.102) was observed. Stratified analyses based on bi-pollutant models estimated consistently positive associations in three cohorts, with an HR of 1.115 (95% CI: 1.030, 1.208) for CFPS, 1.456 (95% CI: 1.328, 1.597) for CHARLS, and 1.055 (95% CI: 1.026, 1.085) for CLHLS (Table S7). These results indicated that the inherent variability among cohorts did not substantially affect the observed associations. Subgroup analysis revealed that elderly individuals residing in rural or high-temperature areas, as well as those with unhealthy lifestyle habits (cigarette smoking and alcohol drinking), were possibly at a higher risk of ozone-related mortality (Table S8). The O₃-mortality association kept robust when excluding participants dying within the initial 12 months since baseline, extending the exposure window from 1- to 2-year interval, and performing DAG-adjusted analysis (Table S9).

Table 2.

Hazard ratio of all-cause mortality associated with per 10-ppb increase in long-term ozone exposure.

Model	Single-pollutant		Bi-pollutant (+PM2.5)
	HR (95% CI)	E-value	HR (95% CI)	E-value
Model 1	1.088 (1.068, 1.109) ***	1.313	1.083 (1.058, 1.109) ***	1.302
Model 2	1.084 (1.064, 1.105) ***	1.304	1.077 (1.052, 1.103) ***	1.228
Model 3	1.080 (1.060, 1.101) ***	1.295	1.070 (1.045, 1.096) ***	1.210
Model 4	1.081 (1.061, 1.102) ***	1.297	1.076 (1.050, 1.102) ***	1.223

Notes: Model 1: stratified by sex and age, and adjusted for cohort as a random term; Model 2: model 1 plus demographic characteristics (educational attainment and marital status), geographical factors (geographical region and residence), and annual average temperature; Model 3: model 2 plus behavioral factors (regular exercise, alcohol consumption, and cigarette smoking); Model 4: model 3 plus health factors (hypertension, diabetes, heart diseases, and respiratory diseases). *** P < 0.001.

Abbreviations: HR, hazard ratio; CI, confidence interval; PM_2.5, fine particulate matter.

On the basis of model 4, we observed highly comparable O₃-mortality associations in single- and bi-pollutant analyses (Fig. 1), both highlighting an approximately linear (P >0.4 for nonlinearity) and threshold-free C-R function across a broad concentration range of 22.7–81.6 ppb. Similar patterns in C-R relationships were also observed in both single- and bi-pollutant analyses utilizing sequentially adjusted models (model 1–3, Fig. S3).

Fig. 1.

Concentration-response relationship between warm-season ozone exposure and all-cause mortality in multi-variable adjusted models estimated by single-pollutant and bi-pollutant analysis. Notes: Dash area represents the 95% CI; the box plot is used to describe the distribution of ozone exposure for participants during study period, with 47.4 ppb, 52.9 ppb, and 58.9 ppb for 25th, 50th, and 75th percentiles ozone concentration, respectively. Abbreviations: HR, hazard ratio; CI, confidence interval.

Fig. 2 displays the spatial variations in O₃-related mortality burden among Chinese older adults (≥65 years) at resolution 0.1° × 0.1° across China in 2011 and 2019. Despite a sharp growth in the elderly population in China during the study period (+50.8 million, Table S10 & Fig. S4), the overall burden of O₃-related mortality decreased by 0.11 million, indicating that epidemic decline in mortality rate and fluctuation in O₃ pollution level may have greatly dominated this trend. We estimated totally 0.88 (95% CI: 0.60, 1.14) million premature deaths attributable to O₃ exposure in 2019, yielding an inconspicuous reduction of 0.11 million compared to the estimate in 2011 (0.99 million, 95% CI: 0.68, 1.28). During 2011–2019, O₃-attributed deaths increased by 28.6% (+0.04 million) in BTH and 14.3% (+0.02 million) in YRD but declined by 7.7% (−0.01 million) in JASH. The attributable premature deaths led to 9.05 (95% CI: 6.19, 11.70) million YLL in 2019, representing a negligible decline of 1.3% compared to 2011 (9.17 million, 95% CI: 6.30, 11.80). Nevertheless, the burden of YLL in the top three regions exhibited a significant upward trend from 2011 to 2019, with a remarkable ascent of approximately 40.9% (+0.54 million) for BTH, 23.4% (0.30 million) for YRD, and 7.8% (+0.09 million) for JASH, respectively (Table S11).

Fig. 2.

Estimates of ozone-attributable premature deaths (A) and years of life lost (B) among older adults in China at 0.1° × 0.1° grid level. Abbreviations: ADs, attributable deaths; YLL, years of life lost; BTH, the Beijing-Tianjin-Hebei and surrounding areas; YRD, the Yangtze River Delta; JASH, the juncture of Jiangsu-Anhui-Shandong-Henan; CYD, the Cheng-Yu District; TCC, the Triangle of Central China; FWP, the Fen-Wei Plain; GBA, the Greater Bay Area.

Fig. 3 lists the top-leading provinces in O₃-attributable premature deaths and LLE. For the attributable deaths (Fig. 3A), the top 5 provinces were Shandong (0.09 and 0.11 million), Jiangsu (0.09 and 0.08 million), Henan (0.08 and 0.07 million), Hebei (0.05 and 0.06 million), and Sichuan (0.09 and 0.05 million) in 2011 and 2019, due to the high population density in these areas. Comparatively, ranking of LLE attributable to ambient O₃ exposure showed a shuffled pattern from 2011 to 2019 (Fig. 3B). The top five provinces for O₃-attributable LLE in 2011 were Sichuan (1.59 years), Henan (1.56 years), Shandong (1.52 years), Chongqing (1.50 years), and Shaanxi (1.46 years), however, four of these provinces experienced an appreciable decline of LLE in 2019, with the exception of Shandong. Assuming causality, by achieving the AQG 2021 of WHO for warm-season O₃ (60 μg/m³), the average life expectancy for the Chinese elderly (age ≥65 years) could increase by 0.93 (95% CI: 0.63, 1.20) years in 2019. The greatest lifespan gains in 2019 were seen in Beijing (1.72 years, 95% CI: 1.20, 2.20), followed by Shandong (1.53 years, 95% CI: 1.06, 1.96), and Hebei (1.39 years, 95% CI: 0.96, 1.79).

Fig. 3.

Leading 10 provinces and ranking changes of attributable deaths and loss of life expectancy among Chinese older adults aged 65 + years from 2011 to 2019. Notes: Provinces are ranked in descending order according to the numbers of excess deaths and LLE (with 95% CIs estimated by Monte Carlo bootstrap simulation) attributable to long-term ambient ozone exposure. Abbreviations: ADs, attributable deaths; LLE, loss of life expectancy.

Discussion

This pooled analysis involved 30,874 individual participants aged ≥65 years from three national cohorts across China, encompassing a wide range of O₃ concentrations from 22.7 to 81.6 ppb. This nationwide cross-cohort analysis provided convincible novel evidence from LMICs, demonstrating a nexus between increased mortality and chronic O₃ exposure among Chinese older adults. We observed no departure from linear C-R curves for the O₃-mortality relationship both in single- and bi-pollutant models. On the basis of the analysis of mortality burden, we identified great between-province disparities in O₃-attributable premature deaths and LLE.

Prior population-based studies investigating chronic O₃ exposure and mortality were dominantly carried out in high-income countries, revealing regional heterogeneities with mixing evidence [12], [26], [32]. Research conducted in North America generally suggested an increased mortality risk associated with O₃ exposure, supported by evidence from several large single cohorts including the American Cancer Society Cancer Prevention Study-II [36], the US Medicare beneficiary Cohort [7], [28], and the Canadian Census Health and Environment Cohort [3]. However, contrasting findings of inverse [13], [30] or null [35] O₃-mortality associations were widely reported in European cohorts, while relevant studies were still incomprehensively conducted in LMICs [32]. In line with the results (HR: 1.108, 95% CI: 1.099, 1.117) from a longitudinal study involving 68.7 million American older adults [28], our multi-cohort analyses found an increased mortality risk of 1.076 (95% CI: 1.050, 1.102) associated with a 10-ppb increase in O₃. This finding was also largely sustained by several recent nationwide single cohort studies [25], [53], [55] and regional research [20] in the Chinese mainland. The disparity in associations between O₃ and mortality across the globe may possibly be attributed to factors such as between-study differences in follow-up duration, population characteristics, exposure levels and assessment methods [32]. Therefore, it is crucial to conduct further population-based studies encompassing diverse regions to comprehensively understand the significant variations in mortality risk associated with global O₃ pollution. Our study identified greater O₃-related mortality risks among Chinese older populations in warmer locations, while individuals living in lower-temperature settings may experience heightened mortality risks from ambient PM_2.5 exposure [16], [17]. Besides the distinct seasonal exposure patterns of ambient O₃ and PM_2.5 across diverse climate zones, the complex interplay between extreme temperature events and air pollution might partially contribute to intensified biological susceptibility through atmospheric chemical kinetics and thermoregulatory stress pathways [8], [47].

Evaluating C-R function is essential for estimating the disease burden [52], while few research had explored the C-R association of chronic O₃ exposure with mortality [32]. In this cross-cohort analysis, we observed approximately linear C-R relationship of warm-season O₃ exposure with all-cause mortality in both single-pollutant (P_nonlinear = 0.58) and bi-pollutant (P_nonlinear = 0.46) analyses, revealing stabilized rise in the risk of mortality among older adults in China as O₃ exposure increased from 22.7 to 81.6 ppb. This threshold-free evidence was consistently seen from two large older (age ≥65) cohort studies involving nearly 60 million participants in America [7] and 20 thousand subjects across the Chinese mainland [55]. In contrary, pre-existing studies based on the low-exposure (<50 ppb) US Medicare cohort [28] and the middle-aged and elderly cohort in China [53] reported potential thresholds of O₃ ranging from 35 to 56 ppb, below which there was no apparent increase in excess mortality. These epidemiological evidence suggested that there may be variations in the nonlinearity of the C-R relationship between chronic O₃ exposure and mortality across different age groups. Specifically, a linear association was observed in the elderly population, whereas a nonlinear association with a threshold was detected in the young and middle-aged cohorts. This interesting phenomenon could be partly attributed to the inherent vulnerability of the elderly population [2]. Older adults may have weakened respiratory and cardiovascular systems, making them more susceptible to the oxidative stress and inflammation induced by O₃ exposure [2]. In contrast, younger populations have greater physiological reserves and may not be at risk until exposure levels exceed a certain threshold. Additionally, due to the absence of the necessity to commute for work, elderly individuals typically exhibit a more fixed pattern of activity and lifestyles. Such differences in life patterns may lead to individual differences in O₃ exposure and thus influence the C-R relationship. Further in-depth mechanism- and population-based investigations are imperative for understanding the C-R relationship between long-term survival and O₃ exposure.

Global O₃-attributable chronic respiratory deaths have been previously assessed through assuming a log-linear C-R relationship [21], [23], where attributable deaths (0.36 to 0.42 million in 2019) were estimated merely based on cohort analysis from the US, the UK, and Canada, without including any data from LMICs. A recent epidemiological study estimated approximately 0.48 million cardiopulmonary deaths related to O₃ in 2019 across China [40]. However, the mortality risk employed in the analysis was derived from a single cohort study conducted in America [14]. Given the substantial knowledge gaps in assessing the mortality burden associated with O₃ [33], we established a more locally appropriate estimation model by incorporating three national investigations in China into our cross-cohort analysis. By achieving a counterfactual concentration of AQG 2021 for O₃ (60 µg/m³), this study estimated that there could be an avoidance of 0.88 (95% CI: 0.60, 1.14) million deaths attributable to O₃ exposure among Chinese older adults (age ≥65 years) in 2019, simultaneously leading to a potential increase of 0.93 (95% CI: 0.63, 1.20) years in average life expectancy for the elderly in China. These findings provided a preliminary insight into the challenge posed by O₃ pollution to the health of older adults in China, underscoring the urgent necessity for developing the localized chronic standards for ambient O₃ concentration.

Despite the overall decline in the estimates of deaths and LLE attributed to ozone from 2011 to 2019, we observed a year-by-year upward trend in the period between 2013 and 2017 (Fig. S5). Contrarily, in this period, PM_2.5 air pollution and related health burden have been substantially reduced owing to the national efforts of the Action Plan for Air Pollution Prevention and Control [10], [45]. In the context of compound air pollution in China, effective collaborative control of major pollutants (e.g., PM_2.5 and O₃) should be of great significance for further improving public and environmental health in the coming decades [45]. We observed an upward trend for O₃-related mortality burden during 2011–2019 in several local areas, including the Beijing-Tianjin-Hebei and surrounding areas (BTH), the Yangtze River Delta (YRD) and the juncture of Jiangsu-Anhui-Shandong-Henan (JASH). These regions have experienced notable economic growth accompanied by the considerable rise in energy consumption and industrial emissions, which could greatly aggravate the urban air pollution [45]. In addition, the progress of urbanization and population expansion may potentially lead to a decline in the availability of high-quality medical resources per capita.

The current study possesses several major strengths. Primarily, by performing a pooled analysis of multiple national cohorts, this study unveiled a well-established C-R function for O₃-mortality association among Chinese older adults. Additionally, it firstly depicted the spatiotemporal variations of the mortality burden attributed to O₃, on the basis of the C-R function specialized for Chinese older populations. Lastly, this national multi-cohort study provides novel evidence linking improved air quality for O₃ with potential gain in life expectancy of older adults, which can inform policy development related to air pollution control and support a healthy aging strategy in the context of global warming. The burden assessment results reveal substantial regional variations in O₃-related mortality across China, underscoring the urgent need to establish localized ambient O₃ concentration standards.

The current analyses also had several limitations that should be discussed. First, the exposure assessment for participants in the CFPS and CHARLS cohorts was conducted at the administrative district level rather than based on residential addresses, which might lead to inevitable exposure misclassification. This approach of utilizing average-based exposure assessment tends to attenuate O₃-mortality association towards the null [43]. Nevertheless, we observed a consistently positive association between O₃ and mortality among participants in the CFPS and CHARLS cohorts, evidencing the robustness of the O₃-mortality association in our multi-cohort analyses. Second, the absence of information on individual time-activity patterns (e.g., home-based and full-time worker activities) may introduce to-some-extent bias in the evaluation of actual exposure; however, this bias was likely to be universal and non-differential, which may have a minimal impact on our association analysis [34]. Third, due to the constraints of the current database, the participants lacked a unique identifier that could be utilized for matching with the Chinese national cause of death registry database, thereby hindering our ability to analyze the effect of O₃ exposure on cause-specific mortality. We are therefore only able to estimate O₃-attributable mortality burden due to all causes rather than specific causes. Additionally, despite considering a rich set of variables in our analyses, there may still exist unmeasured factors (e.g., residential greenness and road traffic noise) that could potentially influence the O₃-mortality association. However, the relatively large E-value for risk estimates implied limited unmeasured confounding when analyzing O₃-mortality associations [37].

Conclusions

In summary, this multi-cohort analysis provided compelling evidence of elevated risks and attributable mortality burden associated with chronic O₃ exposure among Chinese older adults. Under causality and counterfactual scenarios, our analysis revealed that reducing ambient O₃ exposure could effectively prolong life expectancy of the elderly in China. These findings may have potential decision-making implications for fostering the development of national cleaning air policies and promoting healthy ageing in the future.

Compliance with Ethics Requirements

All procedures were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2008 (5). Informed consent was obtained from all patients for being included in the study. All these participants or their legal representatives signed informed consent forms, and the Peking University Biomedical Ethics Review Committee reviewed and approved the CLHLS (No. IRB00001052–13074), CFPS (No. IRB00001052-14010), and CHARLS (No. IRB00001052-11015) surveys.

CRediT authorship contribution statement

Minjin Peng: Methodology, Formal analysis, Writing – original draft, Visualization, Funding acquisition. Yang Yuan: Methodology, Formal analysis, Writing – original draft, Visualization. Haitong Zhe Sun: Conceptualization, Supervision, Funding acquisition. Jing Wu: Methodology, Formal analysis. Lifeng Zhu: Methodology, Formal analysis, Visualization. Yi Zeng: Resources, Supervision. Yunquan Zhang: Conceptualization, Methodology, Software, Supervision, Funding acquisition. Yao Yao: Conceptualization, Software, Methodology, Supervision.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: