Schizophrenia is considered to be one of the most severe mental disorders. It is a complicated and clinically heterogeneous behavioural and cognitive syndrome, with an estimated lifetime risk of approximately 1%.Reference Owen, Sawa and Mortensen1 Individuals affected by this disease often struggle to achieve complete recovery, and even those with a good prognosis may experience negative effects on their lives, such as poor social and occupational functioning.Reference Marwaha, Johnson, Bebbington, Stafford, Angermeyer and Brugha2,Reference Foussias, Agid, Fervaha and Remington3 Moreover, the annual economic burden of schizophrenia is huge, ranging from an estimated US$94 million in Puerto Rico to US$102 billion in the USA.Reference Chong, Teoh, Wu, Kotirum, Chiou and Chaiyakunapruk4
Environmental risk factors
Evidence indicates that the environment and genes are both risk factors for schizophrenia. As an important environmental factor, air pollution, has been proven to strongly affect respiratory diseasesReference Liu, Lim, Pedersen, Jørgensen, Amini and Cole-Hunter5 and other health problems. In recent years, an increasing amount of research has focused on the influence of air pollution on mental health. Population studies have indicated that air pollution may raise schizophrenia risk through cytokine mediation.Reference Gao, Wei, Pan, Yi, Xu and Duan6 Experimental studies have reported that exposure to air pollution induces neuro-inflammation, endothelial dysfunction and microglia activation, all of which contribute to the pathogenesis of schizophrenia.Reference Li, Han, Guo, Li and Sang7 However, epidemiological research on this association is quite scarce and still ambiguous. For instance, some studies showed that short-term exposure to air pollution with fine particulate matter (particle diameter <2.5 μm, PM2.5) was positively related to schizophrenia,Reference Gao, Wei, Pan, Yi, Xu and Duan6,Reference Zhang and Zhou8 whereas others on short-term exposure reported no significant relationship between PM2.5 pollution and schizophrenia.Reference Nguyen, Malig and Basu9,Reference Li, Zhang, Qian, Xie, Luo and Han10 It should be noted that all these studies have focused on the impact of short-term exposure to air pollution. Furthermore, because of limitations in the design of time-series studies, these studies cannot be used to infer causal relationships well. To our knowledge, one study has explored the relationship between long-term air pollution exposure during childhood and schizophrenia risk,Reference Antonsen, Mok, Webb, Mortensen, McGrath and Agerbo11 but the effects on schizophrenia of long-term exposure to air pollutants during adulthood remain unknown. Therefore, cohort studies involving large populations are essential to investigate the possible link between long-term exposure to air pollution and schizophrenia risk in adults.
Genetic risk factors
Evidence indicates that the risk of schizophrenia can also be influenced by genetic factors. Comprehensive genome-wide association studies (GWAS) have pinpointed certain genetic variants associated with the risk of schizophrenia.12 Through these susceptible regions, researchers can calculate polygenic risk scores (PRS) to assess genetic susceptibility and identify high-risk populations.Reference Lewis and Vassos13 Emerging evidence has indicated that the interplay between genetic factors and the environment might play a crucial role in the aetiology of the disease.Reference Li, Ma, Cui, Yang, Liu and Tang14 Gene–environment interaction is currently explored as a facet of schizophrenia aetiology. A previous study suggested that the YWHA gene family and the TPH1 gene potentially exert a cumulative effect in schizophrenia.Reference Jacobsen, Kleppe, Johansson, Zayats and Haavik15 Lei et al discovered that PM2.5 and the YWHAB gene polymorphism locus rs6031849 together affected the relapse of schizophrenia.Reference Lei, Huang, Li, Zhong, Chen and Pan16 The YWHAB gene is associated with schizophrenia through inflammatory pathways, aligning with a pathway linked to air pollution.Reference Rodriguez-Muguruza, Altuna-Coy, Castro-Oreiro, Poveda-Elices, Fontova-Garrofe and Chacon17 Therefore, it is reasonable to assume that interaction of exposure to air pollution with genetic risk of schizophrenia significantly contributes to the development of schizophrenia. However, no research has been conducted so far to investigate how genetic susceptibility modifies the association between air pollutants and schizophrenia risk.
The current study
Using data derived from the UK Biobank, the current study aimed to explore the association between long-term exposure to air pollution and the development of schizophrenia in adults. We also assessed how genetic susceptibility modified this relationship.
Method
Study design and population
The UK Biobank recruited approximately 500 000 individuals between the ages of 37 and 73 from 22 different centres across the UK during the period 2006–2010.Reference Sudlow, Gallacher, Allen, Beral, Burton and Danesh18 It gathered biological and medical data from participants through touch-screen questionnaires, computer-assisted interviews and biological specimens.
The current study was conducted using data from the UK Biobank, under application number 69741. Ethical approval for the study was granted by the North West Multi-Centre Research Ethics Committee (reference no. 16/NW/0274), and all participants had previously provided written informed consent to the UK Biobank.
Air pollutants
We obtained data on annual average concentrations of the air pollutants PM2.5, PM10, nitrogen dioxide (NO2) and nitrogen oxides (NOx) for each year during the study period from UK AIR, which is developed by the Department for Environment, Food and Rural Affairs (DEFRA) (uk-air.defra.gov.uk). The platform offered UK near-surface data on air pollution for the years 2001 through 2021. Adopting an air diffusion model, UK AIR models the concentrations of various air pollutants at a spatial resolution of 1 × 1 km. We associated the residential addresses gathered at baseline in time and space with ambient air pollution.Reference Li, Ma, Cui, Yang, Liu and Tang14 To ensure the reliability of the models, DEFRA compared the modelled and measured values of air pollutants, and the results demonstrated a satisfactory level of agreement. Detailed information regarding the performance of the model can be obtained at https://uk-air.defra.gov.uk/data/pcm-data.
Genetic data and PRS calculation
We derived the PRS for schizophrenia from the established PRS set in the UK Biobank PRS Release. Previous publications provide more information on the methods employed for calculating the PRS.Reference Lewis and Vassos13,Reference Thompson, Wells, Selzam, Peneva, Moore and Sharp19 In summary, the UK Biobank used a standardised test subgroup and a predefined set of disease and trait definitions and applied Bayesian methods to derive the PRS algorithm, appropriately combining data from various ancestral and relevant characteristics. The UK Biobank determined the PRS value for each individual by summing the posterior effect size of each genetic variant multiplied by the allele gene dosage for the whole genome. We then categorised participants into three groups: low (lowest tertile), medium (middle tertile) or high (highest tertile).
Assessment of schizophrenia
The first occurrence of schizophrenia was determined by using data from medical history, linked death register data, hospital admissions data and primary care records. The related algorithms were provided by the UK Biobank. The diagnosis of schizophrenia was based on ICD-10 (F20).Reference Jongsma, Turner, Kirkbride and Jones20 Follow-up of all participants was conducted from enrolment to the occurrence of schizophrenia, death or 12 December 2020, whichever came first.
Covariates
To identify covariates that required adjustment in our multivariate analyses, we used an online directed acyclic graph (DAG) tool DAGitty (www.dagitty.net) to construct a DAG.Reference Greenland, Pearl and Robins21 Considering prior knowledge and literature,Reference Antonsen, Mok, Webb, Mortensen, McGrath and Agerbo11,Reference Song, Liu, Wei, Li, Liu and Yuan22 we incorporated a comprehensive range of covariates into the DAG for analysis, which included age, gender, ethnicity, employment, education, income, residential area, migration, social isolation, substance use, pregnancy and birth complications, Townsend deprivation index (TDI), cardiovascular disease, diabetes and lifestyle factors (alcohol consumption, smoking, healthy diet score and physical activity). By referring to the DAG (Supplementary Fig. 1, available at https://doi.org/10.1192/bjp.2024.118), we retained a minimal set of essential variables for adjustment, including age, gender (male or female), ethnicity (Black and minority ethnic or White), education background (degree level education; non-college level, i.e. below degree level; or none of above), employment (employed, retired, unemployed, homemaker or others), annual income (<£31 000 or ≥£31 000), TDI, social isolation (least isolated, moderately isolated or most isolated: see Supplementary material, Method 1) and residential area (urban or rural). Data on covariates were obtained from baseline assessments.
Statistical analysis
Baseline characteristics were computed and displayed as mean (s.d.) for continuous data, or frequency and percentages for categorical data. Missing data in covariates were addressed using the fully conditional specification (FCS) technique through multiple imputation. Student t-test, Mann–Whitney U-test, or χ 2-tests were used to assess the differences in baseline features between individuals with and without schizophrenia.
We used the Cox proportional hazards model, which included time-varying annual air pollution exposure data, to separately investigate the associations between PM2.5, PM10, NO2, NOX and the risk of developing schizophrenia. In the multivariate-adjusted models, the annual average concentrations of air pollutants for each participant were classified into tertiles, and the categorised air pollutant measure was treated as a time-varying variable. Using the first tertiles as the reference, the hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) were calculated after adjusting for confounders including age, gender, ethnicity, employment, education, income, TDI, social isolation and residential area. In genetics-related studies, additional adjustments were made to include genotyping batch and the first ten genetic principal components (Supplementary Method 2). We further conducted a trend analysis by assigning tertiles as continuous variables in a Cox regression model. Additionally, restricted cubic spline (RCS) analysis was employed to explore the dose–response relationships between air pollution and schizophrenia risk.
We conducted a stratified analysis to examine the relationship of air pollution with the schizophrenia risk in three different genetic risk (low, intermediate and high) groups by adding a product term of air pollution and genetic risk to the model. We then calculated P-values for interaction (Supplementary Method 3). To assess the joint associations, we further categorised participants into nine groups according to air pollution exposure (tertiles) and genetic risk (tertiles) and evaluated the schizophrenia risk in different groups compared with those having low air pollution exposure and low genetic risk.
To ensure the robustness of our findings, several sensitivity analyses were conducted: (a) excluding schizophrenia events occurring within the initial 2 years of follow-up; (b) limiting our analysis to participants who have lived at their baseline address for more than 5 years; (c) analysing the completed data-set after removing participants with missing covariate data; (d) performing an analysis after further adjusting for lifestyle; (e) fitting a two-pollutant model for each air pollutant by incorporating various types of pollutant into the model; (f) restricting analyses to non-movers (those who did not move to a different residential area) during the follow-up period. Software R (version 4.2.0 for Windows) was used for statistical analyses.
Results
Among the 502 480 individuals initially included in the study, 917 individuals diagnosed with schizophrenia at baseline and 16 275 with incomplete genetic data were excluded. Eventually, 485 288 individuals were included in main analysis (Supplementary Fig. 2). At baseline, the average age of all individuals was 56.55 years (s.d. = 8.09) and 263 336 (54.3%) were female (Table 1). During a median follow-up period of 11.9 years, a total of 417 schizophrenia events were recorded. In comparison with those without schizophrenia, individuals with schizophrenia tended to be younger, male, live in urban areas, have lower employment rates, education levels and income levels, and have higher levels of social isolation. The mean (s.d.) levels of PM2.5, PM10, NO2 and NOx for participants during study period were 10.20 (s.d. = 2.16), 15.10 (s.d. = 2.97), 18.70 (s.d. = 6.80) and 28.20 (s.d. = 12.60) μg/m3 respectively (Supplementary Table 1).
Table 1 Baseline characteristics of participants included in study
a. Below degree level.
b. Positive values of the Townsend deprivation index indicate higher levels of deprivation, whereas negative values indicate lower levels of deprivation.
In the multivariable-adjusted model, significant links were identified between long-term exposure to air pollutants and increased schizophrenia risk (Table 2). The adjusted hazard ratios (95% CI) for the highest tertile compared with the lowest were 1.98 (95% CI 1.80–2.19) for PM2.5, 2.30 (95% CI 2.08–2.55) for PM10, 2.30 (95% CI 2.05–2.58) for NO2 and 2.35 (95% CI 2.09–2.64) for NOx (P for trend of all pollutants <0.001). The sensitivity analyses also yielded consistent results (Supplementary Tables 2–7). For instance, even after excluding missing covariates and further restricting analyses to non-movers during the follow-up period, the correlation between the four pollutants and schizophrenia risk remained robust (Supplementary Tables 4, 7). Additionally, RCS analyses were utilised to evaluate the dose–response curve linking air pollution and schizophrenia (Fig. 1), revealing evidence of non-linear associations (P for non-linearity <0.05). Supplementary Table 8 shows a significant relationship between PRS and the schizophrenia risk (hazard ratio = 1.65 (95% CI 1.50–1.81)).
Table 2 Associations between air pollutants and the risk of incident schizophrenia among participants in the UK Biobanka
PM2.5, fine particulate matter with diameter <2.5 μm; PM10, particulate matter with diameter <10 μm; NO2, nitrogen dioxide; NOx, nitrogen oxides.
a. Cox regression models adjusted for age, gender, ethnicity, education, employment, household income, Townsend deprivation index, residential area and social isolation.
Fig. 1 Associations between long-term exposure to air pollutants and the risk of schizophrenia among participants in the UK Biobank.
A restricted cubic spline regression model with four knots (at the 5th, 35th, 65th and 95th percentiles) was used to estimate the dose–response relations between air pollutants and the risk of schizophrenia among participants. Hazard ratios ((HRs) solid lines) and 95% CIs (shaded areas) were adjusted for age, sex, ethnicity, education, employment, household income, Townsend deprivation index, residential area, social isolation. PM2.5, fine particulate matter with diameter <2.5 μm; PM10, particulate matter with diameter <10 μm; NO2, nitrogen dioxide; NOx, nitrogen oxides.
As shown in Fig. 2, significant correlations were observed between the four air pollutants and schizophrenia risk in each genetic risk group. We observed significant interaction effects of genetic risk and NO2 and NOx exposure on schizophrenia risk (P for interaction <0.001), whereas no significant interaction was found for PM2.5 and PM10. Additionally, we fitted a model excluding the interaction terms between air pollution and PRS, and the analysis revealed that both air pollution and PRS still exhibited positive associations with schizophrenia risk (Supplementary Table 9). We also conducted an analysis to explore the joint effects of air pollutant exposure and genetic factors on schizophrenia risk (Fig. 3). The results revealed that individuals with both high genetic risk and high air pollution exposure had the highest risk of developing schizophrenia, surpassing those with low air pollution exposure and low genetic risk (PM2.5: hazard ratio = 6.25 (95% CI 5.03–7.76); PM10: hazard ratio = 7.38 (95% CI 5.86–9.29); NO2: hazard ratio = 6.31 (95% CI 5.02–7.93); NOx: hazard ratio = 6.62 (95% CI 5.24–8.37)). The effects of NO2 and NOx on schizophrenia risk were greater in the higher genetic risk compared with lower genetic risk group and the joint effect was larger than the additive effect, which indicates a synergistic relationship. Additive effects were observed between the four air pollutants and the PRS (Supplementary Table 10 and Method 4).
Fig. 2 Associations between air pollutants and the risk of incident schizophrenia (SCZ) stratified by genetic risk.
The P-interaction was evaluated using hazard ratios for the product term between air pollutants and effect modifiers. The genetic risk subgroup was defined according to the polygenic risk score, as low (lowest tertile), intermediate (middle tertile) and high (highest tertile). Hazard ratios ((HRs) data points) and 95% CIs (solid lines) were adjusted for age, gender, ethnicity, education, employment, household income, Townsend deprivation index, residential area and social isolation. PM2.5, fine particulate matter with diameter <2.5 μm; PM10, particulate matter with diameter <10 μm; NO2, nitrogen dioxide; NOx, nitrogen oxides.
Fig. 3 The joint associations of long-term exposure to air pollutants and polygenic risk score with the risk of incident schizophrenia among participants in the UK Biobank.
Cox regression models adjusted for age, gender, ethnicity, education, employment, household income, Townsend deprivation index, residential area, social isolation, genotyping batch and the first ten genetic principal components. PM2.5, fine particulate matter with diameter <2.5 μm; PM10, particulate matter with diameter <10 μm; NO2, nitrogen dioxide; NOx, nitrogen oxides; HRs, hazard ratios; T1–T3, first, second and third tertile.
Discussion
To the best of our knowledge, this is the first large-scale population study to comprehensively summarise the relationship between long-term exposure to air pollution and schizophrenia risk in adults. We found that long-term exposure to PM10, PM2.5, NOx and NO2 was related to increased schizophrenia risk. We also identified a positive correlation of PRS with the schizophrenia risk. The highest risk of developing schizophrenia was observed in participants with high PRS and exposure to high air pollution levels. Moreover, interactions were observed between genetic risk and two pollutants: NO2 and NOx.
Comparison with existing literature
Previous studies have explored the potential relationship between short-term exposure to air pollution and schizophrenia. However, the results were inconsistent. A systematic review conducted in 2022, which summarised 13 papers, found a positive correlation between short-term exposure to PM2.5, PMC (coarse fraction, diameter 2.5–10 μm) and PM10 and the risk of schizophrenia.Reference Song, Liu, Wei, Li, Liu and Yuan22 In the current study, we found similar results that long-term exposure to air pollutants, including particulate matter and nitrogen oxides, is linked to increased risk of schizophrenia. However, completely contrary results have been found in other studies. For instance, a study carried out in three subtropical Chinese cities reported that short-term PM2.5 exposure did not exhibit any association with schizophrenia risk.Reference Li, Zhang, Qian, Xie, Luo and Han10 Additionally, although some studies have identified a positive effect of NO2 on the increased risk of schizophrenia,Reference Song, Liu, Wei, Li, Liu and Yuan22 others have observed no significant results regarding the association of short-term exposure to NO2 and NOx and the risk of schizophrenia.Reference Ji, Liu, Song, Pan, Cheng and Wang23 Unlike these studies on short-term pollution exposure, only one national study has used long-term air pollution exposure data and suggested positive associations between long-term exposure and schizophrenia in adults.Reference Antonsen, Mok, Webb, Mortensen, McGrath and Agerbo11
In our study, we assessed air pollution exposure by matching it to residential addresses with a spatial resolution of 1 × 1 km. This method has been widely adopted in previous research to investigate the relationship between air pollution and various diseases, including mental disorders.Reference Liu, Lim, Pedersen, Jørgensen, Amini and Cole-Hunter5,Reference Li, Ma, Cui, Yang, Liu and Tang14 A Danish study matched children's addresses with air pollution information, confirming a positive relationship between air pollution and schizophrenia risk.Reference Antonsen, Mok, Webb, Mortensen, McGrath and Agerbo11 Similar results were observed in our study of adults. Moreover, the consistency in methodologies across these studies, including our own, reinforces the reliability and relevance of using address-matched air pollution data to probe the intricate association of air pollution and mental illness. Our findings for the four air pollutants and schizophrenia risk provide new evidence for future insights.
Potential confounders
There are other confounders that can influence the risk of schizophrenia. For instance, evidence shows that metropolitan indoor environments may contribute significantly to overall exposure to adverse environmental conditions, particularly for people in industrialised nations.Reference O'Lenick, Wilhelmi, Michael, Hayden, Baniassadi and Wiedinmyer24 Despite this, most population-based epidemiological studies have primarily focused on outdoor exposure, owing to data availability, neglecting indoor conditions. Quantifying relationships between indoor environments and health remains a substantial challenge. Furthermore, although the current study indicates a positive association between long-term air pollution exposure and increased schizophrenia risk, observational research design does not support the view that air pollutants causally affect the schizophrenia risk. Considering the facts that unmeasured and unknown potential covariates may exist and associations do not mean causality, the results should be interpreted with caution. Further research is warranted to examine the causality of air pollution in relation to schizophrenia risk using methods of causal inference such as intervention studies.
Derivation of the schizophrenia PRS
The PRS calculated by aggregating multiple variants has the capacity to evaluate genetic susceptibility and identify people with high genetic risk.Reference Lewis and Vassos13 The current study used a standard set of PRS for schizophrenia generated by the UK Biobank. The UK Biobank PRS Release provides well-validated PRS across multiple diseases. The UK Biobank's PRS algorithm was developed using trait-specific meta-analyses based on a Bayesian approach, appropriately pooling data across various ancestries and related traits; these tools are especially valuable as PRS algorithms move beyond simple linear combinations of variant weights.Reference Thompson, Wells, Selzam, Peneva, Moore and Sharp19 Furthermore, although validating the performance of PRS or comparing different PRS algorithms presents considerable challenges, the UK Biobank has developed and released a robust PRS evaluation tool to facilitate comprehensive and comparable assessments of predictive performance across various PRS in the UK Biobank data-set. Broad benchmark tests have shown that the PRS in the UK Biobank Release outperforms a series of published PRS. Additionally, the UK Biobank is set to routinely update its PRS Release, with goals to improve performance and enlarge the range of traits included. This approach anticipates the ongoing development of improved PRS scores as both data and methodologies progress, facilitating continuous research and advancements in clinical model development.
Biological mechanisms underlying the link between air pollution and schizophrenia
Suggested mechanisms underlying the increased schizophrenia risk due to air pollutants are currently limited and still being explored. Research indicates that air pollution can reach and affect the brain by various pathways, including through the nasal pathway and olfactory bulbs or through respiration, systemic circulation and the blood–brain barrier.Reference Genc, Zadeoglulari, Fuss and Genc25 Air pollutants have the potential to induce neuroinflammation, endothelial dysfunction and microglia activation and cause cerebrovascular injury,Reference Li, Han, Guo, Li and Sang7,Reference Genc, Zadeoglulari, Fuss and Genc25 thereby promoting mental disorders, including schizophrenia. Experimental studies conducted on mice have demonstrated that PM10 can trigger inflammation and endothelial dysfunction in the brain.Reference Guo, Zhu, Guo, Li, Chen and Sang26 Air pollution has also been shown to affect central nervous system (CNS) function by activating microglia, with ensuing oxidative stress and neuroinflammation,Reference Genc, Zadeoglulari, Fuss and Genc25 which may increase the schizophrenia risk. Mice experiments have also found that NO2 can generate reactive nitrogen species (RNS) and reactive oxygen species (ROS), damaging mitochondria in the brain.Reference Yan, Ji, Shi, Li and Sang27 Population studies have further revealed that environmental PM2.5 can increase the risk of relapse in schizophrenia through the mediation of cytokines, including IL-17 and IL-13.Reference Gao, Wei, Pan, Yi, Xu and Duan6 The above evidence indicates that each pollutant may act through distinct mechanisms and have varying associations with schizophrenia. Future exploration is required to understand the specific biological mechanisms by which various air pollutants contribute to the pathogenesis of schizophrenia.
Genetic susceptibility to schizophrenia
Existing studies have established that schizophrenia's aetiology is multifactorial, with a substantial genetic component. GWAS have reported that numerous common variants, each with a minor impact, are associated with schizophrenia. Over 100 loci have been found to be significantly linked to schizophrenia.12 PRS has been demonstrated to be associated with the risk of schizophrenia and it accounted for approximately 7.7% of the variability in schizophrenia case–control status.Reference Trubetskoy, Pardiñas, Qi, Panagiotaropoulou, Awasthi and Bigdeli28 In the current study, we examined the role of genetic factors in the connection between air pollution and schizophrenia, and observed that genetic susceptibility exacerbates the increased schizophrenia risk in relation to air pollution, especially in participants with high genetic susceptibility. The cellular mechanisms underlying the joint effect of air pollution and genetic factors on schizophrenia are not elucidated. However, previous studies have summarised the association of genetic factors and air pollutants in relation to other mental diseases,Reference Li, Ma, Cui, Yang, Liu and Tang14 and a few studies also indicated that genetic factors may affect schizophrenia through a common mechanism in air pollution such as the YWHA gene family,Reference Lei, Huang, Li, Zhong, Chen and Pan16 implying that air pollutants and genetic variations might lead to schizophrenia by means of shared mechanistic pathways such as oxidative stress, neuroinflammation and endothelial dysfunction.
Strengths and limitations
The study included some notable strengths that enhance its validity and reliability. A large sample and a prospective study design were employed to enhance the statistical power. Furthermore, reliable nationwide data on time-varying exposure to air pollutants were used. Finally, a novel aspect of the study was the investigation of the interaction and joint impact of air pollutants, genetic factors and schizophrenia risk, which has never been conducted previously. Based on these strengths, we discovered that air pollution has a more extensive impact than previously believed. It affects not only physical health but also significantly affects mental health. The findings emphasise the need for disease prevention and mental health improvement through reductions in air pollution. Moreover, our study reaffirms the positive correlation between air pollution exposure and schizophrenia risk. These findings offer new insights into managing schizophrenia, which is a hidden yet significant public health challenge causing daily difficulties and considerable societal expense. These findings enhance our understanding of the environmental factors linked to this condition and underscore the pressing need for better air pollution control measures.
Nevertheless, some limitations of our study need to be acknowledged. First, in the UK Biobank, a swift and efficient recruitment process yielded a sample of 500 000 participants. However, this efficiency was accompanied by a response rate of 5.5%, raising the possibility of selection bias. However, the actual difference in these estimates is small, and consequences of potentially underestimating such risks are expected to be minimal.Reference Stamatakis, Owen, Shepherd, Drayton, Hamer and Bauman29 Second, since we did not directly measure personal exposure to air pollution, there may be methodological issues regarding ecological fallacy and potential misclassification errors in exposure. One of the methods to reduce the ecological fallacy is the use of the smaller spatial units. In this study, we estimated the concentrations of ambient air pollution at a spatial resolution of 1 × 1 km, which is a reasonably fine precision of exposure modelling. However, it should be noted that the existence of an ecological fallacy cannot be completely avoided with this measure. Third, the UK Biobank data-set has only limited information on the composition of air pollution, and consequently, uncertainty remains regarding the specific components of air pollution that may be most harmful. Fourth, the identification of incident schizophrenia cases based only on registered medical information and healthcare records may not capture all cases accurately. It is possible that individuals with milder forms of schizophrenia may not seek medical attention, leading to underreporting of schizophrenia cases. Fifth, the identification of incident schizophrenia cases in the UK Biobank cohort relied on the ICD-10 coding system. The use of ICD-10 codes in administrative databases presents an opportunity for studying medical conditions and a variety of diseases in a large ‘real-world’ setting.Reference Pu, Ramani, Chen, Perry and Hong30 However, it is important to acknowledge inherent limitations in population studies using ICD-10 codes, which can include the possibility of misclassification or underdiagnosis. These inaccuracies are generally unrelated to air pollutant levels and often result in less precise estimates while also potentially biasing the risk estimates downwards.Reference Wellenius, Schwartz and Mittleman31 Sixth, despite adjusting for numerous potential confounding variables in our analysis, there is still a possibility of residual confounding from unmeasured or unknown factors. Seventh, although we used a time-varying analysis, which is an effective method of avoiding introducing immortal time bias, it remains possible that immortal time bias may have influenced the quantitative results, and caution should be exercised in extrapolating these findings. Eighth, all individuals included in the study were drawn from the UK Biobank, with the predominant proportion being White, which could potentially reduce the generalisation to diverse populations. Therefore, it is imperative to validate the existing results in other ethnicities. Finally, we did not account for the impact of indoor air pollution, owing to the lack of relevant data in the UK Biobank. Although ambient sources contribute the most to indoor air pollution levels and the effect of indoor sources may not be as significant as that of outdoor sources,Reference Liu, Zhou, Wang and Zhao32 we should be aware that the time spent indoors has continuously increased with the scale-up of cities, a better understanding of the contribution of the indoor environment on mental health is necessary to protect human health now and in the future.
Conclusions
To conclude, long-term exposure to air pollution was positively associated with an increased schizophrenia risk. Additionally, interactions between NO2 and NOx and genetic susceptibility were observed, and the joint effect is larger than the additive effect, indicating a synergistic relationship. In the future, more exploration is needed to provide new evidence that may contribute to changes in policy-making and individual behaviour aimed at reducing the risk of schizophrenia in the population.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjp.2024.118.
Data availability
The data-set can be accessed from the UK Biobank (https://www.ukbiobank.ac.uk/enable-your-research/apply-for-access). The data on air pollution can be obtained through the DEFRA (https://uk-air.defra.gov.uk/data/pcm-data).
Acknowledgements
We thank all participants from the UK Biobank who have shown their continued commitment, as well as the UK Biobank team for collecting and providing the data.
Author contributions
R.L., R.C. and Y.T. conceived the study. R.L., D.L. and Y.T. contributed to the study design. Y.T. prepared and cleaned the data. R.L. and Y.M. conducted the data analysis. R.L. and Y.T. drafted the manuscript. R.L., R.C., D.L., Y.M., L.T. and Y.T. critically revised the manuscript for intellectual content. All authors approved the final version of the manuscript. R.C. and Y.T. act as guarantors.
Funding
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Declaration of interest
None.
eLetters
No eLetters have been published for this article.