Acute effects of ambient air pollution on daily neurology clinic visits for vertigo: a time-series study in Wuhan, China

This study aimed to disclose the relationship between ambient air pollution and neurology clinic visits (NCVs) for vertigo. A time-series study was conducted to examine relationships between six different criteria air pollutants (SO2, NO2, PM2.5, PM10, CO, and O3) and daily NCVs for vertigo in Wuhan, China, from January 1st, 2017 to November 30th, 2019. Stratified analyses were computed according to gender, age, and season. A total of 14,749 records of NCVs for vertigo were enrolled in this study. Data showed that the increase in daily NCVs for vertigo corresponding to 10 μg/m3 increase of respective pollutants are: SO2 (− 7.60%; 95% CI: − 14.25 to − 0.44%), NO2 (3.14%; 95% CI: 0.23 to 6.13%), PM2.5 (0.53%; 95% CI: − 0.66 to 1.74%), PM10 (1.32%; 95% CI: − 0.36 to 3.06%), CO (0.00%; 95% CI: − 0.12 to 0.13%), and O3 (0.90%; 95% CI: − 0.01% to 1.83%). Males were more susceptible to acute exposure to SO2 and NO2, compared to females (SO2: − 11.91% vs. − 4.16%; NO2: 3.95% vs. 2.92%), whereas the acute effect of O3 exposure was more significantly obvious in females than males (0.94% vs. 0.87%). Moreover, correlations between daily NCVs for vertigo and acute exposure to SO2, NO2, and O3 were all stronger in individuals under 50 years old (SO2: − 12.75% vs. − 4.41%; NO2: 4.55% vs. 2.75%; O3: 1.27% vs. 0.70%). Short-term exposure to PM2.5 was more significantly associated with daily NCVs for vertigo in cool seasons (1.62% vs. − 0.68%), while the correlation between CO exposure and daily NCVs for vertigo was stronger in warm seasons (0.21% vs. − 0.03%). Our study demonstrated acute exposure to ambient NO2 and O3 positively associated with daily NCVs for vertigo. Acute effects of air pollution on daily NCVs for vertigo varied according to gender, age, and season.


Introduction
Air pollution is recognized as one of the top five risks for disease development, according to the Global Burden of Disease study and the World Health Organization (Collaborators 2018; Health Effects Institute 2017). Exposure to air pollutants led to more emergency room visits, outpatient service, hospitalizations, and decreased productivity . As a metropolitan city with a large population in central China, Wuhan also suffers from severe air pollution (Zeng et al. 2018). Thus, it is important to understand the potential health risks of ambient air pollution, serving to be helpful in health policy-making.
Vertigo is a common complaint in outpatient neurology clinic visits (NCVs) and could have various etiologies, according to Bárány Society's Committee for the Classification of Vestibular Disorders (Bisdorff et al. 2015). In the last decade, studies showed that exposure to some air pollutants contributed to the occurrence of various neurological Responsible Editor: Lotfi Aleya Jiachen Zheng and Min Xu are co-first authors.
consequences, leading to increased outpatient service of neurological clinics (Block and Calderon-Garciduenas 2009;Costa et al. 2020;Loane et al. 2013;Salvi and Salim 2019). These associated neurological disorders and symptoms include but are not limited to vertigo (Mariani et al. 2008;Yuan et al. 2021), depression (Braithwaite et al. 2019), migraines (Elser et al. 2021), and epilepsy (Chiang et al. 2021). However, the reported associations are mainly studied in developed countries or vulnerable populations such as the elderly and pregnant women. Data in this area conducted in the general population or in developing countries are scarce, and relations between ambient air pollution and daily NCVs for vertigo are still unclear.
Air pollution is a generalized term for its multiple components, including sulfides (SO 2 ), nitrogen oxides (NO 2 ), particulate matter (PM 2.5 and PM 10 ), carbon monoxide (CO), and ozone (O 3 ), and meteorological influence factors such as temperature, humidity, and barometric pressure. In this study, we conducted a time-series analysis to examine relationships between different ambient air pollutants (SO 2 , NO 2 , PM 2.5 , PM 10 , CO, and O 3 ) and daily NCVs for vertigo. Meanwhile, we incorporated a two-pollutant model to assess the robustness of the effect estimates after adjusting for the effects of co-pollutants. Furthermore, sensitivity analyses were also computed according to season, gender, and age. Our findings are important to health policy decisions in cities with similar emission conditions.

Materials and methods
Wuhan is the capital of Hubei Province, situated in Central China (latitude 30°35′N and longitude 114°17′E). It is a large metropolitan city with a population of over 10 million, and 13 districts covering about 8494.41 km 2 (http:// www. wuhan. gov. cn). As an important industrial base and transportation hub in China, the main sources of air pollution in Wuhan are automobile exhaust and industrial emissions (Liu et al. 2022). Wuhan has a subtropical monsoon climate with warm and rainy summers and cold and dry winters. The average ambient temperature of Wuhan is 30.1 °C in July and 4.1 °C in January. Due to the arrival of the monsoon from the north, where two heavily polluting provinces are located (Henan and Hebei), the air pollution in Wuhan is often more severe in the winter (Mao et al. 2018).

Neurology clinic visits data
Outpatient records of NCVs for vertigo were extracted from the hospital information system of Zhongnan Hospital, Wuhan University, for the study period between January 1st, 2017 and November 30th, 2019. NCVs for vertigo at Zhongnan Hospital are local outpatients, mainly from the Wuchang District of Wuhan, where the hospital is located (Fig. 1). Wuchang District contains 19% population of the whole urban area and 12% population of Wuhan (Liu et al. 2022). A total of 14,749 outpatient records of NCVs with a chief complaint of vertigo were enrolled in the study (N = 14,749). The dataset included dates of NCVs for vertigo and related demographic information including gender, age, and residential address of enrolled participants. Non-Wuhan residents and cases who recurrently visited for vertigo in Zhongnan hospital were excluded from the study. Disorders of vestibular function (H81), dizziness, and giddiness (R42) were classified and entered according to the 10th version of the International Classification of Diseases (ICD-10). This study protocol was approved by the Medical Ethics Committee of Zhongnan Hospital (IRB number: 2022142 K).

Environmental and meteorological data
Daily ambient air pollution data for the study period between January 1st, 2017 and November 30th, 2019 were obtained from the website of the Wuhan Ecological Environment Bureau (http:// hbj. wuhan. gov. cn/). The daily average concentrations of six criteria air pollutants (SO 2 , NO 2 , PM 2.5 , PM 10 , CO, and O 3 ) were calculated by averaging hourly values of each criteria air pollutant from ten fixed-site monitoring stations. All the stations are far from industrial, residential, and vehicular sources, ensuring the monitoring data reflects overall urban background air conditions without uncertain interference. Data of the following meteorological parameters (daily average ambient temperature [°C], relative humidity [%], and barometric pressure [KPa]) for the study period were obtained from the meteorological data sharing service of the China Meteorological Administration (Beijing, China). Seven days (0.66%) of environmental and meteorological data were missing, and data with missing information were eliminated from the study.

Statistical analysis
An over-dispersive generalized additive model (GAM) for time series analyses (Song et al. 2018a) was chosen in the exploration of the acute effects of six criteria ambient air pollutants on daily NCVs for vertigo. Data of daily NCVs for vertigo followed an over-dispersed Poisson distribution; therefore, quasi-Poisson regression was used in the above model. Meanwhile, distributional lag models were applied in the deduction of both cumulative exposure effects and displacement effects of air pollutants on daily NCVs for vertigo (Song et al. 2018b).
Several covariates were imported into the GAM model to control both time-invariant and time-varying confounding effects. First, a natural cubic regression smoothing function for calendar time with 7 degrees of freedom (df) per year excluded long-term and seasonal trends over 2 months. Second, natural smooth functions of daily average ambient temperature (6 df), relative humidity (3 df), and barometric pressure (3 df) were incorporated to control for the nonlinear confounding effects of meteorological factors. Third, other covariates, such as public holidays (Holidays) and days of the week (DOW), were adjusted as dummy variables in the GAM model. The core model is described as follows: where E(Yt) represents numbers of daily NCVs for vertigo on day t; denotes the log-relative rate of daily NCVs for vertigo associated with the unit increase of each criteria air pollutant; Z t denotes concentrations of each criteria air pollutant on day t; DOW is a dummy variable for days of the week; and ns refers to the natural cubic regression smoothing function. We derived the exposure-response (E-R) relationship curves between each criteria air pollutant and daily NCVs for vertigo by importing 3 df as a natural spline function into the above model. Three sensitivity analyses were subsequently performed to ensure the stability of this model. First, an alternative proxy of 4-10 df per year was selected for the smoothness of the temporal time trend. Second, a two-pollutant model was used to assess the robustness of effect estimates after adjusting for co-pollutants with correlation coefficients inferior to 0.7. Third, two different lag time constructions were examined for each air pollutant: the single-day lags algorithm (lag0 to lag7) and the multi-day algorithm of moving average lags (lag0-1 to lag0-7), and three statistics logE (Yt) = Z t + DOW + ns (time,df) + ns (temperature, 6) + ns (humidity, 3) + ns (pressure, 3)+ intercept methods of Akaike Information Criterion (AIC), generalized cross validation (GCV) and partial autocorrelation function (PACF) were applied to determine the optimal lag structure for this model.
In addition, stratification analyses were computed according to season (warm: April to September; cool: October to March), age (< 50 years; ≥ 50 years), and gender (females; males) to explore the potentially vulnerable factors related to short-term effects of ambient air pollution on daily NCVs for vertigo. The statistical significance of the differences between the strata effect estimates by calculating 95% confidence intervals as Q 1 −Q 2 ± 1.96 where Q 1 and Q 2 are the estimates for two categories, and SÊ 1 and SÊ 2 are their respective standard errors. All the statistical analyses were performed in R (version 4.1.0) using the MGCV package. A two-tailed p < 0.05 was used to determine the statistical significance. Effects were described as changes of percentage and 95% CI in numbers of daily NCVs for vertigo per 10 μg/m 3 increase of each criteria air pollutant (SO 2 , NO 2 , PM 2.5 , PM 10 , CO, and O 3 ).

Results
As shown in Table 1, a total of 14,749 records of NCVs for vertigo were enrolled in the present study for the period between January 1 st , 2017, and November 30th, 2019. Total NCVs for vertigo in warm seasons were slightly greater than in cool seasons (50.9% vs. 49.1%). The proportion of females was higher than males in total NCVs for vertigo (58.9% vs. 41.1%), while elders aged more than 50 years old presented more frequently than young individuals aged under 50 years old in NCVs for vertigo (65.0% vs. 35%). In addition, the average ambient temperature, relative humidity, and barometric pressure were 17.7 °C, 78.76%, and 101.51 kPa, respectively, during the study period in Wuhan. Annual average levels of air pollutants were 9.04 μg/m 3 for SO 2 , 46.22 μg/m 3 for NO 2 , 46.76 μg/m 3 for PM 2.5 , 76.89 μg/ m 3 for PM 10 , 996.56 μg/m 3 for CO, and 99.25 μg/m 3 for O 3 respectively. Figure 2 demonstrates moderate to strongly positive correlations among short-term exposure to five criteria air pollutants (SO 2 , NO 2 , PM 2.5 , PM 10 , and CO), with Spearman's correlation coefficients ranging from 0.52 to 0.87. However, the association between O 3 and other criteria air pollutants was weak, with correlation coefficients ranging merely from − 0.24 to 0.09. Furthermore, short-term exposures to SO 2 , NO 2 , PM 2.5 , and PM 10 were all negatively correlated with ambient temperature and relative humidity, whereas O 3 and CO exposures were positively associated with ambient temperature (Spearman's correlation coefficient: 0.66) and relative humidity (Spearman's correlation coefficient: 0.13). Meanwhile, short-term exposure to SO 2 , NO 2 , PM 2.5 , PM 10 , and O 3 all showed moderate positive correlations with atmospheric pressure, with the exception of CO, where exposure was negatively associated with atmospheric pressure with a correlation coefficient of − 0.55. Figure 3 further illustrates percentage changes of daily NCVs for vertigo (mean and 95% CI) associated with per 10 μg/m 3 increase of each criteria air pollutant using the single-day lags algorithm (lag0 to lag7) and multi-days algorithm of moving average lags (lag0-1 to lag0-7). As shown in Table 2, lag1 for SO 2 , lag07 for NO 2 , lag2 for PM 2.5 , lag4 for CO, lag01 for PM 10 , and lag0 for O 3 were chosen as the optimal lag structures to get the smallest AIC/GCV/PACF values according to the model fitting statistics. It exhibited percentage changes of daily NCVs for vertigo respectively corresponding to per 10 μg/m 3 increase of SO 2 (− 7.60%; 95% CI: − 14.25 to − 0.44%), NO 2 (3.14%; 95% CI: 0.23 to 6.13%), PM 2.5 (0.53%; 95% CI: − 0.66 to 1.74%), PM 10 (1.32%; 95% CI: − 0.36 to 3.06%), CO (0; 95% CI: − 0.12 to 0.13%) and O 3 (0.90%; 95% CI: − 0.01 to 1.83%). Furthermore, short-term NO 2 and O 3 exposures positively correlated to daily NCVs for vertigo, whereas the association between SO 2 exposure and daily NCVs for vertigo was conversely negative. Correlations remained significantly robust after adjusting for co-pollutants in the two-pollutant model (Table 2). Sensitivity analysis further demonstrated that an alternative proxy of 4-10 df did not significantly affect the real-world effects of six criteria air pollutants on daily NCVs for vertigo (Fig. S1). Table 3 demonstrates the different estimated effects of six criteria air pollutants on daily NCVs for vertigo stratified by season, gender, and age. It showed that short-term exposure to PM 2.5 was positively associated with daily NCVs for vertigo in cool seasons (1.62%; 95% CI: 0.15 to 3.12%); however, CO exposure and daily NCVs for vertigo showed a positive association in warm seasons (0.21%; 95% CI: 0.05 to 0.38%). Males were more susceptible to short-term NO 2 exposure, compared to females in daily NCVs for vertigo (3.95% vs. 2.92%), whereas a negative correlation was conversely observed between short-term SO 2 exposure and daily NCVs for vertigo in males (SO 2 : − 11.91%; 95% CI: − 20.42 to − 2.42%). Meanwhile, females presented more vulnerabilities than males when short-termly exposed to O 3 (0.94% vs. 0.87%). In addition, short-term exposure to NO 2 (4.55%; 95% CI: 0.15 to − 9.13%) and O 3 (1.27%; 95% CI: 0.07 to 2.48%) were positively correlated with daily NCVs for vertigo among individuals under 50 years of age. Conversely, the effects of short-term SO 2 exposure were significantly but negatively associated with daily NCVs for vertigo in the same subgroup (− 12.75%; 95% CI: − 22.05 to − 2.34%). Figure 4 graphically illustrates the exposure-response (E-R) relationship curves between short-term exposure to six criteria air pollutants and daily NCVs for vertigo. (1) SO 2 exposure: with the increasing concentration of ambient SO 2 , daily NCVs for vertigo showed a sustained declining trend firstly and then maintained in a steady state at more than 20 μg/m 3 . (2) NO 2 exposure: The E-R graph presented a flattened arcsine curve twisted in two opposite directions at a concentration of around 35 μg/m 3 . (3) PM 2.5 exposure: The E-R graph firstly maintained a flat steady trend within the interval of 0-40 μg/m 3 , followed by an inverted, flattened U-shaped curve, of which the highest point is at the concentration of 100 μg/m 3 . (4) PM 10 exposure: with the increasing concentration of PM 10 , daily NCVs for vertigo sharply rise within the interval of 0-70 μg/m 3 but then maintained in an approximately flat steady state at more than 70 μg/m 3 . (5) CO exposure: An approximately straight flattened curve was observed from the E-R relationship curve between short-term CO exposure and daily NCVs for vertigo. (6) O 3 exposure: The E-R graph presented a U-shaped curve symmetrically mirroring two opposite trends at the concentration of 70 μg/m 3 .

Discussion
During the study period, annual average concentrations of ambient NO 2 , PM 2.5 , and PM 10 in Wuhan exceeded the Chinese National Ambient Air Quality Standards (https:// www. mee. gov. cn). However, the annual average level of ambient Our study demonstrated that short-term exposure to ambient NO 2 presented statistically significant correlations with daily NCVs for vertigo. As we know, NO 2 is asphyxiating odorous gas characterized as one of the environmental irritants. Studies showed that airborne NO 2 could penetrate into the inner ear through the round window membrane (Aguilera et al. 2013;Sasa et al. 1989) and then dissolve in the perilymph and endolymph, and could result in elevated perilymphatic and endolymphatic acidity (Mun et al. 2021). It is reported that the acidic surrounding environment can facilitate the degeneration of otoconia, leading to the detachment of otoconia (Walther et al. 2014). The above findings underline plausible explanations for the onset of benign paroxysmal positional vertigo (BPPV) and Meniere's disease related to ambient NO 2 exposure. Furthermore, inhaled NO 2 exerts deleterious effects on the brain via olfactory epithelium triggering neuroinflammation and transferring inflammation to distal brain regions by damaging the permeability of the blood-brain barrier (Adams et al. 2016). From both cellular and anatomical levels, pathways of neuroinflammatory activation, impaired neurogenesis, and neurodegeneration are likely mechanisms of vertigo of central origin related to air pollution (Li and Xin 2013).
In our study, short-term exposure to O 3 also positively correlated with daily NCVs for vertigo, which is consistent with previous studies. One study conducted in Korea revealed that O 3 exposure was associated with a high incidence of Meniere's disease.  in Wuhan is mainly produced by photochemical reactions of nitrogen oxides and volatile organic compounds (Choi et al. 2021). As the powerful inhaled oxidizing agent, studies showed that oxidative stress induced by O 3 exposure is intimately linked with brain lipid peroxidation, neuroinflammation and subsequent neuron damage and impaired cerebral vascular endothelial injury (Cai et al. 2016;Yang et al. 2019). Moreover, the microglial activation in response to air pollutants and inflammatory cytokines or cells was also shown to be engaged in neuron and cerebral vasculature damage (Block and Calderon-Garciduenas 2009;Costa et al. 2020). The above evidence provided implications for the observed Fig. 3 Percentage change (%) of NCVs for vertigo (mean and 95% CI) associated with a 10μg/m3 increase in various air pollutant concentrations using different lag structures association between O 3 exposure and increased daily NCVs for vertigo in our study.
Unexpectedly, a significantly negative correlation was observed between acute exposure to SO 2 and daily NCVs for vertigo. In Wuhan, ambient SO 2 is mainly from burning coal and sulfur-containing fossil fuels in energy-intensive industries. Since exogenous SO 2 is known as toxic irritant gas with detrimental effects on human bodies, Chinese government has instituted a strict emission policy of SO 2 in the last decade (Jin et al. 2016). In our study, the annual average level of ambient SO 2 in Wuhan was well below the Chinese National Ambient Air Quality Standards and other criteria air pollutants (NO 2 , PM 2.5 , and PM 10 ), ascribing to the environmental policy tightening in China. Thus, it is difficult to determine the exact mechanism through which ambient O 3 exposure reduces the daily NCVs for vertigo. We speculated that human bodies might deal with inhaled SO 2 through conversion in a certain threshold range . It is reasonable that people will reduce outdoor activities when confronted with extreme air pollution. Animal studies showed that low levels of inhaled SO 2 could be converted to sulfur-containing amino acids ) and the endogenous derivatives through a series of physiological changes, resulting in significant anti-inflammatory, antioxidant, and maintenance of cerebrovascular normalcy (Du et al. 2008;Wang et al. 2017), neuroprotective effects (Ohtani and Nishimura 2020), and may have suppressed depression and anxiety ). On the other hand, it is possible that chronic high-level exposure to SO 2 in the past would lead to behavioral adaptations in the body and enhanced tolerance to SO 2 (Jun and Min 2019). To date, the potential roles of environmental sulfur dioxide in brain homeostasis remain elusive and deficient in concrete evidence and require further research.
In our study, correlations between daily NCVs for vertigo and acute exposure to NO 2 and SO 2 were stronger in males and young individuals aged under 50 years old. This may be linked to the fact that males and young people are more likely to undertake outdoor activities with more exposure to environmental pollution (Gu et al. 2020). The same reasoning also applies to the negative correlation between acute O 3 exposure and daily NCVs for vertigo with a stronger coefficient in young people. Conversely, females were found to be more susceptible to O 3 exposure in daily NCVs for vertigo. This phenomenon has also been shown in the respiratory and cardiovascular systems, which is attributed to greater airway reactivity and smaller airways in females (Mao et al. 2018). Moreover, the correlation between daily NCVs for vertigo and short-term exposure to PM 2.5 was stronger in cool seasons, which is consistent with previous studies (Song et al. 2018b). Due to the special geographical location and unfavorable meteorological conditions of Wuhan, it is difficult to disperse air pollutants in weak winter monsoons. And it is possible that the spreading velocities might vary among air pollutants of different molecular weights. However, the acute effect of CO on daily NCVs for vertigo was more pronounced in warm seasons. We speculated that the inconsistency of season differences among six different criteria air pollutants might attribute to the diversity of air pollutant mixtures in different seasons (Tsai et al. 2019). For example, the monsoon in the warm season led to less particulate matter but more opportunities for O 3 exposure, like from window openings that lead to increased exposure in the warm season (Tsai et al. 2019). Taken together, the stratification analysis of this study helped to figure out the vulnerable influence factors in daily NCVs for vertigo when confronted with extreme air pollution. The exposure-response (E-R) curve is critical for public health assessment. In the study, we observed that both NO 2 and O 3 exposure in a low range was not associated with daily NCVs for vertigo. This might be related to the fact that the E-R relationship is influenced by various factors, such as air pollution mixtures, climatic conditions,

Fig. 4
Exposure-response (E-R) relationship curves between short-term exposure to six criteria air pollutants and daily NCVs for vertigo and population sensitivity. However, the E-R curve of NO 2 exposure smoothed out at higher levels, which is referred to as the "harvest effect." Due to the "harvest effect," a vulnerable population will emerge before levels of air pollutants reach reasonably high levels (Chen et al. 2017). The absence of significant thresholds in the E-R curve of O 3 can be explained by the limited availability of single-city data (Song et al. 2018c) and the enhanced effects of high-level O 3 on other criteria air pollutants (Win-Shwe et al. 2013). Similarly, the E-R curves for SO 2 exposure smoothed out at higher levels, which may be related to limited outdoor activities with higher-level exposure to SO 2 . Thus, it is suggested to tailor emission policies of different air pollutants according to local conditions, of which scientific rationality of different criteria air pollutant thresholds warrants further exploration with larger samples. Our study has several limitations. First, we used average concentrations of six criteria air pollutants measured by stationary site monitoring to represent individual exposures, which may lead to inevitable exposure misclassification. Second, although we have considered some possible confounding effects of co-pollutants and meteorological factors (temperature, relative humidity, and pressure), there may be other factors that affect the onset of vertigo and impair a person's tolerance to air pollutants, such as pre-existing diseases and unhealthy factors. Third, we only collected data from one hospital in a highly polluted city, resulting in a possible selection bias. Therefore, further studies are needed to confirm our results, and molecular biology or animal experiments are necessary to explore the exact mechanisms between air pollution and the onset of vertigo.

Conclusions
Our study demonstrated that acute exposure to ambient NO 2 and O 3 positively correlated with daily NCVs for vertigo in Wuhan, China. The acute effects of ambient air pollution on daily NCVs for vertigo varied according to season, gender, and age. Our findings could be helpful in health policymaking when confronted with extreme air pollution in cities with similar emission conditions.