Critical mechanisms

PM_2.5 can enter the lungs and bloodstream, reaching brain tissue to produce immune and inflammatory responses locally^{Reference Qi, Wei, Xin, Huang, Pu and Ma37}, and may lead to cellular damage, depending on size and chemical or biological properties. We focus here on inflammation and immune mechanisms, as these are increasingly linked to poor mental and physical health; indeed, reverse causality between mental and physical health and shared aetiological mechanisms suggest we should look at the two together and not adopt the Cartesian dichotomy which has bedevilled health research. People with pre-existing mental illnesses are vulnerable and already have a life expectancy shortened by 15–20 years owing to long-term physical illnesses.^{Reference Baxter, Harris, Khatib, Brugha, Bien and Bhui38}

Endotoxins are pro-inflammatory, and metabolites from BioPM may be toxic. Mechanisms include oxidative stress and immune dysregulation, for example, by release of cytokines, with an impact on T and B cells that form part of the immune response, and disruption in the balance of immunoglobulins that fight infection and invasion by contaminating substances. There may be direct carcinogenic risks or toxicity to cells, from lead and cadmium, for example.^{Reference King, Zhang and Cohen9} These processes can harm brain function and also lead to depression and cognitive decline. Particulate matter may damage respiratory cells, provoking a local immune response, and changes in the lung microbiome, in turn, can affect the gut microbiome.^{Reference Xue, Chu, Li and Kong39} This can worsen existing conditions or precipitate new episodes of physical and mental illness.

Adversity and poverty can produce inflammation in mothers, the fetus in utero, young children and young adults, as well throughout the life course.^{Reference Giurgescu, Engeland, Templin, Zenk, Koenig and Garfield40} Inflammation can lead to depressive experiences, mental illnesses and physical illnesses. Depression, both as a clinical diagnosis and as less severe symptoms, includes pessimism, poor self-care and a lack of motivation and can thus compound the adverse effects of poverty by reducing social support and risking unemployment. Indeed, poverty and food scarcity are known to directly influence cognitive decisions,^{Reference Mani, Mullainathan, Shafir and Zhao41} and this makes brain health vulnerable to additional risks such as poor air quality.

There are direct effects of particulate matter and/or BioPM and inflammation on health; however, epigenetic mechanisms can also be activated by adversity to alter immune responses, leading to poor mental and physical health. For example, early life adversity, trauma and poverty can lead to premature ageing and a greater risk of multiple long-term conditions.^{Reference Cunliffe42–Reference Franke45} Studies show that even exposure to particulate matter in the prenatal and perinatal period can affect later health.

These webs of causation and social determinants of poor health are more common in urban environments.⁴⁶ Among deprived inner-city areas, health risk behaviours are more common and can add additional harms, for example, through smoking and excessive use of alcohol, a lack of physical activity, greater risk of adverse childhood experiences, poor early-life care, neglect, parental mental illness, poor educational achievement and school exclusion, and incarceration.^{Reference Bhui, Shakoor, Mankee-Williams and Otis47} These all promote ‘inflammogenic’ environments that are vulnerabilities for additional risk exposures. Indeed, these are the mechanisms by which COVID-19 was proposed to more specifically affect people facing multiple social inequalities and marginalisation.^{Reference Bhui48} Therefore, among people with established physical or mental illnesses, particulate matter can exacerbate and precipitate additional episodes of ill health requiring more specialist and intensive interventions.^{Reference Gu, Guo, Si, Wang, Zhang and Deng49}

Quantification of BioPM exposures

The existing evidence base on BioPM stems from disconnected scientific disciplines and sectoral foci, each with their own perspectives and methods. Currently, detection and characterisation of BioPM is based largely on culture-based microbiology, microscopy (spores, pollen), bioassays (endotoxin, 1,3-β-d-glucan), chromatography and mass spectroscopy for specific biochemicals (ergosterol and mannitol/arbitol, 3-hydroxy fatty acids, microbial volatile organic compounds, phospholipid-derived fatty acids, and molecular microbiology (DNA, RNA). Each sampling and analysis method has various advantages and disadvantages, which have been thoroughly discussed previously.^{Reference Whitby, Ferguson, Colbeck, Dumbrell, Nasir, Marczylo, Bohan and Dumbrell50} Moreover, there are currently no standardised protocols for bioaerosol sampling and analysis, making it difficult to compare studies and assess dose–effect relationships.^{Reference Whitby, Ferguson, Colbeck, Dumbrell, Nasir, Marczylo, Bohan and Dumbrell50} To better understand the impact of bioaerosol exposure on human health, comprehensive collection and analysis methods are needed to detect, characterise and quantify BioPM and its interactions with other pollutants.^{Reference Whitby, Ferguson, Colbeck, Dumbrell, Nasir, Marczylo, Bohan and Dumbrell50} The most frequently used collection methods include gravitation, impaction, impingement, cyclone and filtration, allowing a range of analysis options. Methods to characterise BioPM in real time are also emerging (Fig. 2).^{Reference Whitby, Ferguson, Colbeck, Dumbrell, Nasir, Marczylo, Bohan and Dumbrell50} In general, the most appropriate method must obtain a representative sample of the environment being investigated and will depend upon the taxa of interest, preservation of sample integrity and the purpose of the study.^{Reference Whitby, Ferguson, Colbeck, Dumbrell, Nasir, Marczylo, Bohan and Dumbrell50} To achieve this, bespoke combinations of methods are needed that are tailored to the environment, context, downstream analysis and research questions under investigation.^{Reference Whitby, Ferguson, Colbeck, Dumbrell, Nasir, Marczylo, Bohan and Dumbrell50}

Fig. 2

Overview of BioPM (particulate matter of biological origin) collection and analysis methods (adapted from ref. ^{Reference Whitby, Ferguson, Colbeck, Dumbrell, Nasir, Marczylo, Bohan and Dumbrell50}). MALDI-TOF, matrix-assisted laser desorption/ionisation; FISH, fluorescence in situ hybridisation; DGGE, denaturing gradient gel electrophoresis; RT-qPCR, real-time quantitative polymerase chain reaction; ATP, adenosine triphosphate; GC-MS, gas chromatography-mass spectrometry; LC-MS, liquid chromatography-mass spectrometry.

Each BioPM collection and analysis method has advantages and disadvantages. Although significant progress in bioaerosol collection and analysis has been achieved over the past two decades, there is still no consensus on collection or any standardised analysis method for a particular context or environment. This makes it difficult for researchers to compare data across studies and for regulators to set meaningful exposure limits for BioPM for a particular environment. Although some headway has been made towards addressing this issue,^{Reference Ferguson, Garcia-Alcega, Coulon, Dumbrell, Whitby and Colbeck51} more research is needed in this area. Analytical methods to provide real-time detection of BioPM need further development to improve their capabilities for rapid identification of BioPM.

Innovation in studies of mental health impacts of air pollution

We should improve measurement and evidence of harms for different exposures to define thresholds by which one might define ‘clean’ or ‘healthy’ air. Even this terminology attracts ethical and philosophical questions about what the public might understand and how we apply criteria to define ‘clean’ or ‘healthy’. We need to develop ways of reliably measuring air particles, including bioaerosols, with sufficient precision to build evidence on what is healthy. Given the lack of knowledge on BioPM and the widespread exposure, we need better methods of measurement. More specifically, we need to establish a consensus on BioPM sampling and analysis methods. This includes quantification and characterisation of BioPM in indoor and outdoor locations.^{Reference Ferguson, Neath, Nasir, Garcia-Alcega, Tyrrel and Coulon52,Reference Grydaki, Colbeck, Mendes, Eleftheriadis and Whitby53} Separate studies might consider measurement of exposures in the most vulnerable groups, for example, those living in areas of high exposure near to farmland or waste disposal,^{Reference Ferguson, Neath, Nasir, Garcia-Alcega, Tyrrel and Coulon52} and where populations already carry a higher burden of health problems.

In our introduction and our previous paper, we set out the evidence on how air pollution in general might drive poor health, including the onset of new mental illnesses and exacerbation of existing mental illnesses.^{Reference Bhui, Newbury, Latham, Ucci, Nasir and Turner1} These studies may be epidemiological, ecological or observational clinical studies and trials of interventions (medical, social or policy). Ecological studies are weaker owing to high residual confounding; however, they are an important step towards identifying mechanisms of action and prevention. Thus, epidemiological studies in populations and new cohorts collected specifically for testing the effects of air pollution are the most likely to yield valid data in which individual and area confounders can be considered and immunological mechanisms can also be investigated. Studies of clinical populations with poor mental health and chronic medical conditions could reveal how air pollution affects ongoing health and might lead to relapse and greater service use.

Mental illnesses can be measured using several valid methods, and this consists of measuring distinct diagnosed conditions (including psychoses, depression, anxiety and personality difficulties, as well as degenerative brain conditions such as dementia or cognitive impairment associated with neurological conditions, e.g. epilepsy, Parkinson's disease). Studies assessing associations with mental illness will need to better define which types of mental illness are considered. Psychoses are rarer, and neurological conditions specifically carry additional risks for poor mental health, and there may be shared mechanisms and pathway to several mental illnesses.^{Reference Baker, Kale and Menken54}

Many non-specialists are unaware of how to assess mental illnesses. Although some biomarker research is underway, there are not yet any biomarkers that have sufficient specificity and sensitivity for diagnosis. Most diagnoses are based on assessment of experiences and symptoms using the following methods.

1. (a) Self-report screening and outcome measures, for example, Generalized Anxiety Disorder 7 to measure anxiety and nine-item Patient Health Questionnaire to measure depression.^{Reference Kroenke, Wu, Yu, Bair, Kean and Stump55}

2. (b) Interviews using structured questionnaires that generate valid ICD-10 diagnoses, for example; the clinical interview schedule is one such measure.^{Reference Das-Munshi, Castro-Costa, Dewey, Nazroo and Prince56}

3. (c) Primary care and health service records including verbatim accounts can be mined using artificial intelligence software, but there is variation in data quality.^{Reference Davis, Sudlow and Hotopf57,Reference Frayne, Miller, Sharkansky, Jackson, Wang and Halanych58}

4. (d) Social care and criminal justice systems have their own administrative data systems, and these can be linked with other data systems.^{Reference DeHart and Shapiro59}

5. (e) Prescription of psychotropic medication is similarly noted in several national data-sets, as well as local cohorts.^{Reference Lemanska, Hoang, Jeffreys, Bankhead, Bhui and Ferreira60}

Pathways into care and help-seeking are known to vary by age, gender and ethnicity.^{Reference Bhui and Bhugra61} In terms of estimating levels of mental health problems, there are discrepancies between hospital and primary care and population levels of poor mental health, given that not everyone seeks help from primary care. Even fewer people seek help from or are referred to hospital care; rather, most will seek help in the community or from primary care.^{Reference Moffat, Sass, McKenzie and Bhui62} Indeed, many people with poor mental health may not be recognised as having poor mental health, especially in low-income countries.^{Reference Kisely and Siskind63} Therefore, population research is necessary to provide a true picture of levels of mental illness and the impact of poor air quality on mental health. We need better methods to engage underrepresented groups in research, for example, ethnic minorities and those living in poorer areas, the very people who are likely to be most affected.

Two areas of technology development are likely to have a major impact on exposure and outcome measurement, wearable technology and artificial intelligence as a research tool.^{Reference Bernasconi, Angelucci and Aliverti64} Recent advances in technologies make it possible to wear devices that measure respiration, heart rates and blood oxygen and even take a basic electrocardiogram; some apps can pose questions about mental health on a hour-by-hour or daily basis. These ‘wearable’ and ‘experience sampling’ approaches permit real-time collection of data on emotional states, behaviours, eating patterns (including use of alcohol, medication and smoking) and physiological measurements that might indicate anxiety and distress.^{Reference Pardamean, Soeparno, Budiarto, Mahesworo and Baurley65} Experience sampling can also be used to record self-reported hallucinations, depressive thinking or paranoid beliefs.

Wearables offer advantages in that real-time and spatial data can be gathered that provide a better approximation of real-world exposures to air particles, as well as mental health and physiological parameters. For example, personal exposure to air pollution can be measured, and people can rate their mood or levels of anxiety. Wearables also offer opportunities for health-protective and health-promoting actions, for example, avoiding high-pollution areas if there are pre-existing conditions. For example, a person with asthma could choose to do sport, for example, only when pollution levels are not high. Artificial intelligence methods might also provide methods for exposure measurement and prediction.^{Reference Guo, Ren, Wu, Sun, Wang and Wang66} Such techniques have been used to predict hospital admission for cardiorespiratory conditions^{Reference Usmani, Pillai, Hashem, Marjani, Shaharudin and Latif67} and are being used to better refine classification of mental illnesses,^{Reference Lee, Torous, De Choudhury, Depp, Graham and Kim68} with aspirations to generate better evidence on the links between air pollution and mental illness.^{Reference Sun, Han and Zhu69}