Research and analysis

COMEAP statement on possible approaches to estimating the health effects attributable to exposure to PM2.5 and NO2 indoors, using epidemiological studies on outdoor air pollution

Published 18 December 2025

Summary

Indoor air quality can be affected by infiltration of pollutants from outdoors and also by pollutants emitted or generated indoors. Because most people spend most of their time indoors, exposure indoors makes an important contribution to personal exposure to air pollutants. Estimating the effects of indoor exposure to air pollutants on health would allow their public health importance to be understood. However, epidemiological evidence from indoor environments, suitable for use as the basis for such estimates in the UK, is lacking for many pollutants and health effects.

Until relevant epidemiological studies on indoor air pollution are available, it may be possible to develop methods for quantification of effects attributable to fine particulate matter (PM_2.5) and nitrogen dioxide (NO₂) indoors, using the large evidence base linking outdoor concentrations of these pollutants with adverse effects on health. This statement is aimed at researchers and analysts who might wish to explore the use of these methods. In it, we discuss approaches that could be used to modify concentration-response functions (CRFs) reported in epidemiological studies of outdoor concentrations, in order to make them more appropriate for quantifying effects attributable to exposure in indoor microenvironments such as at home, at school, during transport and in workplaces such as offices.

We think that these approaches are promising, but acknowledge the uncertainties and limitations involved. For example, the existing approaches ignore possible differences in the toxicity of particles from indoor and outdoor sources and the effect of modification by chemical reactions indoors; and do not take into account that concentrations of pollutants do not act as markers of the same sources or mixtures of pollutants indoors as outdoors. We consider these approaches most appropriate to estimate effects attributable to exposure occurring indoors, to PM_2.5 and NO₂ infiltrating from outdoor sources. There would be additional uncertainties when applying them to estimate health effects attributable to exposure to PM_2.5 and NO₂ arising from indoor sources.

COMEAP’s recommended methods for quantification of effects attributable to exposure to PM_2.5 and NO₂ use summary effects estimates (CRFs) from meta-analyses, rather than adopting a CRF from a single individual study. Appropriate modification of a CRF from a meta-analysis might not be straightforward.

We make suggestions for further work that would help improve these approaches.

Introduction

Exposure indoors makes an important contribution to overall personal exposure to air pollution. Recent reports (for example CMO, 2022; AQEG, 2022) have highlighted this, and noted the need for further research on indoor air quality and its effects on health. COMEAP has discussed the report by the Air Quality Expert Group (AQEG, 2022) and made some observations in response (COMEAP, 2025).

There is rather limited epidemiological evidence, relevant to the UK, which directly investigates health effects associated with non-occupational indoor concentrations of air pollutants. In addition, there are practical difficulties in undertaking epidemiological studies to investigate associations of indoor air pollution with infrequent health outcomes, such as mortality. These, and other, issues are mentioned in Annexe B which comprises a short discussion of the availability of epidemiological studies on indoor and outdoor air pollution, and the implications for quantification of attributable effects.

Being able to quantify the effects of indoor air pollutants on health would help understanding and communication of their importance to public health, and allow the benefits of interventions to be evaluated. COMEAP’s quantification sub-group (QUARK) was, therefore, asked to consider whether the epidemiological literature linking health endpoints with concentrations of pollutants outdoors might be used to quantify effects attributable to exposure indoors. QUARK’s consideration has focused on PM_2.5 and NO₂, which are important pollutants both outdoors and indoors. Outdoor ozone (O₃) has also been well studied, but concentrations indoors are much lower due to reactions with other chemicals and interactions with indoor surfaces.

QUARK examined some possible approaches to modifying the concentration-response functions (CRFs) for PM_2.5 or NO₂ reported in the outdoor air pollution literature, in order to make them suitable for use to quantify effects attributable to exposure indoors. This statement includes a description of these approaches, along with a discussion of their limitations. The statement does not review epidemiological studies which have examined associations of health effects with exposure to air pollution indoors.  

In the statement, we discuss quantification of effects attributable to 3 different exposures:

a) exposure indoors to pollution that has infiltrated from outdoors

b) exposure indoors to pollution that has both infiltrated from outdoors and been emitted or generated indoors (combined)

c) exposure indoors to pollution arising from indoor sources only

We describe some of the different uncertainties that need to be considered when using these approaches to quantify effects attributable to these different exposures.

Modifying associations from outdoor air pollution epidemiology for use to quantify health effects attributable to PM_2.5 and NO₂ exposures in other microenvironments

Transferability and modification of epidemiological CRFs

The quantitative transferability of findings from a study to another location – such as using the CRF from a study in one location (for instance, a city or country) to estimate the effects of pollution in another – requires a number of assumptions. These include, for example, assumptions that the toxicity and size distribution of particulate matter are similar in the different locations, regardless of differences in local sources, or that NO₂ acts as a marker for a similar mixture of co-emitted pollutants in the same way in both places. Population differences in physiological responses to exposure to pollutants, perhaps due to genetics or differences in underlying health status, are also possible but are not taken into account.

The transferability of CRFs from one location to another also involves the implicit assumption that the relationship between the pollutant concentration and personal exposure is sufficiently similar in the 2 locations for this to be a reasonable approach. The exposure metrics used in epidemiological studies of outdoor pollutant concentrations vary, ranging from concentrations measured at the nearest fixed central monitoring site, to high resolution modelling of pollutant concentrations at the residential addresses of study participants. These concentrations are regarded as surrogates for the exposure which individuals experience, in both indoor and outdoor environments, to pollutants from outdoor sources [1] as they move around a city, or area, during their daily lives.

The relationship between indoor pollutant concentrations and personal exposure will be different from that for outdoor pollutant concentrations. This means that CRFs reported in the epidemiological literature on outdoor air pollution are not directly transferable to assessing the effects of concentrations experienced indoors or in other microenvironments, such as during transport.

Pollutants in outdoor air can infiltrate into indoor environments such as homes. Therefore, the associations reported in epidemiological studies of outdoor concentrations are understood to represent not just the effects of exposure to pollution outdoors: they also include the effects of exposure, while indoors, to pollution infiltrating from outdoors. This exposure indoors needs to be included when considering the personal exposure represented by the outdoor air pollution epidemiological literature. In contrast, exposure indoors to pollutants arising from indoor sources is unlikely to be reflected in the epidemiological associations with outdoor pollutant concentrations. Therefore, exposure to pollutants from indoor sources does not need to be considered when assessing personal exposure relevant to the outdoor air pollution literature.

By taking factors such as this into account, some authors have attempted to modify the CRF for pollutants (notably PM_2.5) in outdoor air, so that the CRF is expressed in relation to integrated personal exposure to pollution from outdoor sources, rather than in relation to outdoor concentrations. The modified CRF can then be applied, in quantification, to estimates of the personal exposure of interest, for example to estimate effects attributable to exposure to pollution indoors. The approaches that we have examined express personal exposure as a concentration: the integrated exposure (µg/m³) taking into account the proportion of time spent, and the concentration of pollutant, in each microenvironment [2] (see Annexe C). These approaches require information, relevant to the population in the study from which the CRF was obtained, on the pollutant concentrations arising from outdoor sources in various microenvironments; and also on the time that the study’s population spends in each of those microenvironments. Depending on the aim of the assessment in which the CRF is to be applied, similar information is also required for the target population for which attributable effects are to be estimated.

Both the modification of the CRF, and the estimation of exposures of the target population in different microenvironments, require a number of assumptions. In particular, these extrapolations rely on the use of summary data for parameters which are likely to vary considerably between individuals. Thus, this modification introduces another source of uncertainty, in addition to the uncertainty arising from the potential geographical heterogeneity in the CRFs linking ambient air pollution with health outcomes in possibly heterogenous populations. This additional uncertainty is not reflected in the uncertainty interval around the CRF.

In this statement, we use 2 published examples (Milner and co-authors, 2017 and Azimi and Stephens, 2020) to illustrate approaches that have been used to modify CRFs from the outdoor air pollution literature and use them to quantify attributable effects indoors and/or in other microenvironments. Both papers focus on assessing mortality associated with long-term average PM_2.5 concentrations. In principle, similar approaches could be proposed to modify CRFs for other pollutants or health endpoints.

We think that approaches such as those used by Milner and co-authors (2017) and Azimi and Stephens (2020) to modify CRFs from the epidemiological literature on outdoor air pollution, are promising. They may provide a useful method to allow burden estimates and impact assessments for indoor pollutants, in the absence of sufficient empirical epidemiological evidence on indoor air pollution.

However, these types of approaches are very dependent on both the underlying assumptions and the data used to parameterise the calculations. Even if robust data was available for the various parameters, there are a number of inherent uncertainties in the methods. Some of these are discussed in this statement. Careful consideration would be needed to ensure that the methods developed, data selected for parameterisation, and the uses to which they were applied, were appropriate. The uncertainties and limitations would also need to be acknowledged.

Scenarios

Many different scenarios could be envisaged in which researchers or policy-makers might wish to quantify the health effects attributable to indoor exposure to air pollutants. The level of uncertainty, when applying a CRF modified from the outdoor epidemiological literature, will vary between scenarios depending on the source of the pollution being evaluated. For example, some policies or interventions (such as installing new ventilation systems or increasing the airtightness of buildings) change the extent to which outdoor air pollutants infiltrate indoors. In this scenario, a health impact assessment of the resulting changes in exposure, indoors, to pollution from outdoor sources could be undertaken. However, this is unlikely to represent the full health impact of such interventions: indoor concentrations of pollutants arising from indoor sources would also be affected by changes in the ventilation characteristics of buildings. Therefore, estimating the combined health impact of changes to exposure, indoors, to pollution that has both infiltrated from outdoors and been emitted or generated indoors would be desirable. Other interventions might affect emissions from indoor sources (for example, replacing gas cookers by electric hobs). In this scenario, an estimate of the effects attributable to changes in exposure, indoors, to pollution arising from indoor sources would be needed.

Conceptually, the approaches discussed in this statement could be used to estimate effects attributable to exposure to pollutants from outdoors and from indoor sources, in a variety of different microenvironments. Exposure within each microenvironment will differ, depending upon ingress of outdoor pollution and sources specific to the microenvironment, as well as the amount of time spent there.

As well as considering which type of exposure is most relevant to the scenario being evaluated, an appreciation of the relationship between the different components of exposure is also needed. For example, an estimate of effects attributable to outdoor air pollution (calculated by applying a CRF from outdoor air pollution epidemiology to outdoor concentrations) will include the effects of exposure, indoors, to pollution that has infiltrated from outdoors. Similarly, “exposure indoors to pollution that has infiltrated from outdoors” and “exposure indoors to pollution arising from indoor sources” are both subsets of “combined exposure indoors to both pollution that has infiltrated from outdoors and pollution that has been emitted or generated indoors”.

The type and level of uncertainty when using these methods will depend upon the exposure being assessed. We therefore discuss 3 possible exposures separately:

a) exposure indoors to pollution that has infiltrated from outdoors

b) exposure indoors to pollution that has both infiltrated from outdoors and been emitted or generated indoors (combined)

c) exposure indoors to pollution arising from indoor sources only

Some indoor microenvironments – such as the home, school or office – make an important contribution to overall personal exposure because of the amount of time which individuals spend in them. Others, such as transport microenvironments, might be of interest because of their high concentrations of pollutants, even though most people would be expected to spend only relatively short amounts of time there; there will be more uncertainty when attempting to estimate the health effects attributable to exposure in these types of microenvironments than for microenvironments where people spend a larger proportion of their time, or for total personal exposure. Random error in the small time periods spent in these microenvironments may make a large difference in estimated exposures, leading to larger measurement error.

In addition, assumptions would need to be made about the amount of time spent, and the concentrations experienced, in such microenvironments. Taking bus travel as an example, assumptions would need to be made about the amount of time spent in locations such as bus garages or bus stops, as well as on the bus itself, and also about the concentrations in each of these microenvironments which can vary greatly depending on factors such as the age of the vehicles, emission standards, in-vehicle ventilation and idling status. In some cases, the pollutants experienced might be very different from those in ambient air, making the application of a CRF modified from the outdoor air pollution literature inappropriate. For example, there are high concentrations of particulates in the air within the London Underground. However, COMEAP (2019) concluded that the outdoor air pollution epidemiology did not provide an appropriate basis on which to base quantitative comment on the health risk posed to the travelling public, because of the differences between the properties of particles in the air of the London Underground and those in ambient air.

(a) Quantifying effects of exposure indoors to pollution that has infiltrated from outdoors

Milner and co-authors (2017) developed a method to evaluate the health impact of interventions (such as home energy efficiency interventions, or ventilation control) intended to mitigate climate change, that also change the infiltration of PM_2.5 outdoors into homes. The CRF linking increases in outdoor PM_2.5 concentrations (measured at fixed monitoring sites) with increased mortality risk from the American Cancer Society (ACS) study (Pope and co-authors, 2002) was used as the starting point. The CRF was modified, in an attempt to make it appropriate to assess the mortality effect of the changes to residential indoor exposure to PM_2.5 from outdoor sources that would arise from the types of interventions under evaluation. Consideration of changes in exposure to pollutants arising from indoor sources was not included.

Because the study from which the CRF was derived was undertaken in the US, information from the US was used to modify the CRF so that it could be expressed in relation to integrated personal exposure (µg/m³) to pollution from outdoor sources. This modification involved a number of steps. First, it took into account the relationship between outdoor concentrations measured at fixed monitoring sites and the concentrations directly outside homes; concentration ratios were based on analysis of monitoring sites in New York. The extent of infiltration of PM_2.5 indoors was assessed, based on infiltration ratios from a study in US homes. Then, the proportion of time spent indoors was accounted for using human activity data from the US national human activity pattern survey (NHAPS).

Milner and colleagues wanted to apply the modified CRF to undertake health impact assessments in the UK, as they were interested in the impacts of interventions in the UK housing stock. They therefore used time-activity information on the amount of time spent indoors by different age groups in the UK population when undertaking the health impact assessment. This allowed them to estimate the changes in integrated personal exposure (µg/m³) that would result from the expected changes in indoor concentrations arising from the interventions affecting infiltration of outdoor air.

Some of the uncertainties in applying the modified CRF to this type of exposure include:

• the extent to which different pollutants infiltrate into the indoor environment varies – this means that concentrations of PM_2.5 and NO₂ indoors, that have arisen from infiltration from outdoors, would not act as a marker for the whole outdoor pollutant mixture indoors in the same way that they did outdoors

• it is possible that chemical reactions might occur indoors and affect the properties of PM which has infiltrated from outdoors

(b) Quantifying effects of exposure indoors to pollution that has both infiltrated from outdoors and been emitted or generated indoors (combined)

Azimi and Stephens (2020) used a conceptually similar approach to that used by Milner and co-authors (2017), to develop a framework which they then used to estimate the US mortality burden of total exposure to PM_2.5, from both indoor and outdoor sources, taking into account exposure in different microenvironments.

The CRF used as the starting point was from a meta-analysis of cohort studies in the US (Fann and co-authors, 2016). Therefore, US data (human activity surveys, and infiltration factors) was used to estimate integrated exposure (µg/m³) of the US population to PM_2.5 from outdoor sources in a number different microenvironments (residences, vehicles and other indoor locations) as well as outdoors, in order to modify the CRF.

The aim was to estimate the mortality burden in the US. Therefore, data from the US, as far as possible, was also needed as inputs when applying the modified CRF in a burden calculation. Information on time-activity patterns, and concentrations of PM_2.5 in different microenvironments, was used to estimate total integrated exposure (µg/m³) to PM_2.5 from both indoor and outdoor sources. The burden estimate was calculated by applying the modified CRF to these integrated exposures. In the absence of quantitative information to take an alternative approach, equal toxicity of PM from indoor and outdoor sources was assumed, with no further modification made to the CRF.

Some of the uncertainties in applying the modified CRF to this type of exposure include:

• the extent to which different pollutants infiltrate into the indoor environment varies – this means that concentrations of PM_2.5 and NO₂ indoors, that have arisen from infiltration from outdoors, would not act as a marker for the whole outdoor pollutant mixture indoors in the same way that they do outdoors

• concentrations of PM_2.5 and NO₂ do not act as markers of the same sources or mixtures of pollutants indoors as in outdoor air

• for particulate matter, using a modified CRF based on outdoor epidemiology requires an assumption of equal toxicity of PM indoors and outdoors, which might not be the case – it is possible that physico-chemical interactions might occur indoors and affect the properties of PM which has infiltrated from outdoors; applying the modified CRF to estimate the effects of PM from indoor sources requires an assumption that PM from indoor sources is equally harmful to health as that from outdoor sources

• emissions from occupant activities often lead to short-term exposure to high concentrations; this is different from typical time profiles of concentrations arising from outdoor sources – it is possible that intermittent high peaks of pollution are more damaging to health than a more constant exposure to lower concentrations although it is not clear, from the available evidence, whether this is the case (COMEAP, 2021)

(c) Quantifying effects of exposure indoors to pollution arising from indoor sources

Conceptually, it would be possible to apply the same CRF, modified from the outdoor air pollution literature, to estimate effects associated with exposure indoors only to pollution arising from indoor sources. However, in this case the uncertainties would be even greater than for the other 2 types of exposures discussed, and would include:

• uncertainty in the apportionment of pollutant concentrations to indoor and outdoor sources

• concentrations, indoors, of PM_2.5 and NO₂ from indoor sources do not act as markers of the same sources or mixtures of pollutants as outdoor concentrations [3]

• for particulate matter, applying the modified CRF to estimate the effects of PM from indoor sources requires an assumption that PM from indoor sources is equally harmful to health as that from outdoor sources – this is, therefore, an important additional source of uncertainty

• emissions from occupant activities often lead to short-term exposure to high concentrations – this is different from typical time profiles of concentrations arising from outdoor sources

In our view, applying CRFs modified from the outdoor air pollution literature to quantify effects attributable to exposure indoors to pollution that has infiltrated from outdoors (a), has the least uncertainty of these 3 possibilities. Applying modified CRFs to quantify effects attributable to exposure indoors to pollution arising from indoor sources only (c) has the most uncertainty.

All scenarios

Some uncertainties apply to all 3 of the exposures discussed above; a number of these are listed below:

• time-activity patterns differ depending on the time of year and between different groups within the population, varying according to factors such as age, gender and socio-economic status – this adds to the complexity of parameterising both the modification of the CRF and the application of the modified CRF in health impact assessments, or to estimate burdens of disease attributable to exposure to air pollution indoors

• activities undertaken indoors might be different from those undertaken outdoors – in some indoor environments, people may be likely to be more sedentary than when they are outdoors, while in others they may be more active; the level of physical activity will affect the volume of air inhaled, and therefore change the relationship between integrated personal exposure expressed as a concentration and integrated personal exposure calculated as an inhaled dose. Methods currently used to quantify effects attributable to outdoor concentrations of PM_2.5 or NO₂ do not involve estimating inhaled dose. The extent to which this is an additional uncertainty, when extrapolating from the outdoor epidemiology to indoor exposures, is not clear

• estimates produced by applying these approaches should be regarded as population-level summaries – summary data, applicable at a population level, is used to modify the CRFs; the additional uncertainty arising from using population-level data to modify a CRF is not reflected in the uncertainty interval around a CRF

• data on characteristics (time-activity patterns and infiltration factors) relevant to the population in the (outdoor air pollution) epidemiological study from which a CRF is derived is most appropriate to modify the CRF for use to quantify effects attributable to pollutants indoors – however, where possible, COMEAP’s recommended methods for quantification use summary effects estimates (CRFs) from meta-analyses rather than adopting a CRF from a single individual study; these meta-analyses often include studies from a number of different cities, countries or continents. This means that identifying appropriate data (such as infiltration factors or time-activity patterns) to modify a CRF from a meta-analysis might not be straightforward

• another point for consideration, related to the interface between outdoor and indoor air, is that outdoor pollutant concentrations at different heights above the ground vary (Eeftens and co-authors, 2019; Wong and co-authors, 2019) – for some indoor environments, notably those on upper floors of high buildings, this will affect the relationship between the monitored (or modelled) outdoor concentration at ground level and the infiltration of pollutants indoors

Discussion

We are aware of wider interest in the possibility of using modified CRFs from the outdoor air pollution epidemiology to estimate effects indoors (for example, Clark and co-authors, 2025) and that alternative approaches may be considered. In our view, appropriate approaches should account for time-activity patterns and concentrations of pollutants in different microenvironments. However, information on parameters such as time-activity patterns and infiltration factors, which are needed to modify the CRFs using our preferred approaches, may not be available for the subjects of the epidemiological studies from which the CRFs are obtained. Nonetheless, there is a scientific literature on research investigating these factors which can be drawn upon. For example, a number of methods have been used to estimate the infiltration factor for PM_2.5; the most commonly used involves measuring concentrations of sulphate, as a specific tracer of PM_2.5 arising from outdoor sources (Diapouli and co-authors, 2013; Evangelopoulos and co-authors, 2020). Different approaches are required to estimate infiltration of NO₂ (Evangelopoulos and co-authors, 2020).

We have used data from a review of studies investigating the relationship between ambient concentrations and integrated personal exposure concentrations (indoors and outdoors) to pollution from outdoor sources (Evangelopoulos and co-authors, 2020) to illustrate the possible modification required when expressing a CRF in terms of an association with integrated (time- and concentration-weighted) personal exposure (µg/m³) rather than with outdoor concentrations. This suggested that the CRF describing the mortality risk associated with PM_2.5 could be approximately 50% higher [4] when expressed in terms of an association with integrated personal exposure (µg/m³) to pollution from outdoor sources, rather than with concentrations in outdoor air (see Annexe A). However, there is considerable variation in the exposure data from different studies, and the size of the modification needed might be rather more or less than this.

Use to inform guidelines

As for previous editions, the World Health Organization (WHO) recommends that its Air Quality Guidelines (WHO, 2021) for outdoor air are applicable to both outdoor and indoor environments and, therefore, cover all settings in which people spend a significant portion of their time. This draws attention to the fact that pollutants found in outdoor air can also occur indoors, and acts as a reminder that reducing exposure indoors is also important to ensure that pollutants do not adversely affect health.

Most people spend a large proportion of their time indoors. This means that, if pollutant concentrations were the same (for example, equal to the Air Quality Guideline) indoors and outdoors, for most people the integrated personal exposure (µg/m³) from time spent indoors would be higher than the integrated personal exposure (µg/m³) (both indoors and outdoors) arising solely from outdoor sources. This scenario is unlikely to occur in practice. Nonetheless, it suggests that applying the WHO Air Quality Guideline for ambient air to indoor settings would not be over-protective, and might be under-protective for people who spend much or all of their time indoors. We note that the same uncertainties apply to this comparison as for modifying CRFs from the outdoor air pollution literature for use indoors. These include uncertainties relating to differences in the composition and sizes of particles indoors and outdoors, and because pollutants indoors act as proxies of different sources and so represent a different pollutant mix with different toxicity.

We have noted elsewhere (COMEAP, 2025) that it is important that an integrated approach is used to assess proposed policies and interventions intended to mitigate climate change or improve indoor or outdoor air quality, in order to avoid unintended consequences and to maximise co-benefits to health. Interventions which reduce sources of pollution are the most effective for improving indoor air quality (CMO, 2022), and are highest in the air pollution intervention hierarchy (PHE, 2019). However, which type of interventions to improve indoor air are most appropriate will depend upon the setting and the specific circumstances (NASEM, 2024).

Conclusions

Drawing on the points discussed above, we propose the following conclusions and suggestions:

i) If CRFs for indoor pollution are not available, it may be possible to modify a CRF from the outdoor air pollution literature (particularly for PM_2.5 or NO₂) to make it suitable for use to quantify effects attributable to exposure indoors. This involves modifying the CRF so that it is expressed in terms of integrated personal exposure (indoors and outdoors; µg/m³) to pollution from outdoor sources. This modified CRF can then be applied to time-adjusted personal exposure (µg/m³) indoors to quantify attributable health effects

ii) When modifying a CRF, use of time-activity and infiltration data relevant to the population in the study (or studies) from which the CRF was derived would be most appropriate

iii) When applying the modified CRF in quantification, use of time-activity and concentration data relevant to the target population of interest would be most appropriate

iv) The results of quantification using modified CRFs should be regarded as population-level estimates, and the various assumptions and uncertainties acknowledged. The additional complexity in the indoor pollution environment – in which there are many different chemical species, and different physico-chemical interactions than outdoors – is not taken into account in the modification

v) Conceptually, the modified CRF could be applied in a range of assessments (such as evaluating the impact of a change in exposure from a specific source, or in a specific microenvironment, or to estimate the burden attributable to total exposure in all microenvironments) as long as the exposure being evaluated is expressed in terms of integrated personal exposure (µg/m³) taking into account the proportion of time spent, and the pollutant concentrations, in the microenvironment(s) concerned

vi) However, we draw attention to the increased uncertainty when applying the modified CRF to quantify the effects attributable to pollution from indoor sources, rather than to pollution which has infiltrated indoors from outdoors. Uncertainty arises because of differences in the temporal pattern and co-pollutant mixtures of emissions indoors and outdoors, and because particulate pollution arising from indoor sources may have a very different composition and size distribution to particles in outdoor air

We note that whether it is possible, and appropriate, to apply the approaches discussed will depend upon the availability of suitable data and the purpose (and, hence, level of uncertainty that can be tolerated) of the quantification being undertaken.

Recommendations for future work

We make some suggestions for future work that would help develop methods to estimate health effects attributable to exposure to air pollutants indoors:

Monitoring and time-activity data

i) We would encourage researchers undertaking epidemiological studies of health effects associated with concentrations of pollutants in outdoor air to include nested studies which examine the relationship between indoor and outdoor concentrations, and the time-activity patterns of participants. This would allow appropriate study-specific modifications of CRFs to be made

ii) Information on how pollutants act as markers for mixtures in indoor air (in comparison with outdoor air) would help address some of the uncertainties in applying a modified CRF to quantify effects attributable to exposures indoors

Modelling

iii) Continued improvement of exposure modelling with the use of time-activity patterns and monitoring data, and also development or improvement of modelling assessments that differentiate personal exposures from indoor or outdoor sources, are recommended

iv) Modelling air quality indoors could be improved by the availability of more concurrent indoor and outdoor measurements, for example for residential buildings, to estimate infiltration

Health effects

v) Approaches to modify the evidence from outdoor epidemiological studies to make it applicable for use to assess the effects attributable to exposures indoors, such as those discussed in this statement, would be worth exploring. These may provide a useful method to allow burden estimates and impact assessments, in the absence of suitable empirical epidemiological evidence based on exposures indoors

vi) Epidemiological studies investigating associations of health effects with indoor exposures to pollutants would increase the evidence base and allow more direct estimates of the public health importance of exposures indoors to be made. Evaluation of the contribution of exposure from different sources (indoor and outdoor) would be valuable to inform policy. Understanding the importance of short-term exposures to elevated concentrations arising from occupant activities would also be valuable

vii) Research to investigate the toxicity of particles produced by various indoor sources, and to understand the differential toxicity of particles indoors compared to those outdoors, would help address the uncertainties in applying a modified CRF from the outdoor literature to other microenvironments

Summary recommendations

Our main recommendations can be summarised as:

Being able to quantify the effects of indoor air pollutants on health would help understanding and communication of their importance to public health, and allow the benefits of interventions to be evaluated.

Where CRFs for indoor pollution are not available, it may be possible (particularly for PM_2.5 and NO₂) to modify a CRF from the outdoor air pollution literature to make it suitable for use to estimate the health effects associated with exposures indoors and in other microenvironments. However, this approach introduces additional uncertainties. We consider it most appropriate for estimating effects attributable to exposure (indoors) to PM_2.5 and NO₂ which has infiltrated from outdoors, and least appropriate for estimating effects attributable to exposure to pollution arising from indoor sources.

References

AQEG. Indoor Air Quality This is a report to the Department for Environment, Food and Rural Affairs; Scottish Government; Welsh Government; and Department of Agriculture, Environment and Rural Affairs in Northern Ireland, on indoor air quality in the UK. Air Quality Expert Group 2022 [viewed on 27 June 2023]

Azimi P, Stephens B. A framework for estimating the US mortality burden of fine particulate matter exposure attributable to indoor and outdoor microenvironments Journal of Exposure Science and Environmental Epidemiology 2020: volume 30, pages 271-284

Clark SN, Lam HCY, Goode E-J, Marczylo EL, Exley KS, Dimitroulopoulou S. The Burden of Respiratory Disease from Formaldehyde, Damp and Mould in English Housing Environments 2023: volume 10, article 136

Clark SN, Anenberg SC, Brauer M. Global Burden of Disease from Environmental Factors Annual Review of Public Health 2025: volume 46, pages 233-251

CMO. Chief Medical Officer’s Annual Report 2022: air pollution Department of Health and Social Care 2022 [viewed on 15 November 2023]

COMEAP. Statement on the evidence for health effects in the travelling public associated with exposure to particulate matter in the London Underground Committee on the Medical Effects of Air Pollutants: 2019

COMEAP. Advice on health evidence relevant to setting PM2.5 targets Response to question B1(i). Committee on the Medical Effects of Air Pollutants: 2021 [viewed on 31 October 2025]

COMEAP. Summary of COMEAP recommendations for the quantification of health effects associated with air pollutants Committee on the Medical Effects of Air Pollutants 2023 [viewed on 1 March 2024]

COMEAP. Statement in response to the Air Quality Expert Group (AQEG) report on indoor air quality Committee on the Medical Effects of Air Pollutants 2025

Diapouli E, Chaloulakou A, Koutrakis P. Estimating the concentration of indoor particles of outdoor origin: A review Journal of the Air and Waste Management Association 2013: volume 63 pages 1,113-1,129

Eeftens M, Odabasi D, Flückiger B, Davey B, Ineichen A, Feigenwinter C, Tsai M-Y. Modelling the vertical gradient of nitrogen dioxide in an urban area Science of the Total Environment 2019: volume 650, pages 452-458

Evangelopoulos D, Katsouyanni K, Keogh RH, Samoli E, Schwartz J, Barratt B, Zhang H, Walton H. PM2.5 and NO2 exposure errors using proxy measures, including derived personal exposure from outdoor sources: A systematic review and meta-analysis Environment International 2020: volume 137, article 105,500

Evangelopoulos D, Zhang H, Chatzidiakou L, Walton H, Jones RL, Quint JK, Barratt B. Air pollution and respiratory health in patients with COPD: should we focus on indoor or outdoor sources? Thorax 2024: volume 79, pages 1,116-1,123 

Fann N, Gilmore EA, Walker K. Characterizing the long-term PM2.5 concentration-response function: comparing the strengths and weaknesses of research synthesis approaches Risk Analysis 2016: volume 36, pages 1,693-1,707

Lin W, Brunekreef B, Gehring U. Meta-analysis of the effects of indoor nitrogen dioxide and gas cooking on asthma and wheeze in children International Journal of Epidemiology 2013: volume 42, pages 1,724-1,737

Li W, Long C, Fan T, Anneser E, Chien J, Goodman JE. Gas cooking and respiratory outcomes in children: A systematic review Global Epidemiology 2023: volume 5, article 100,107

Milner J, Armstrong B, Davies M, Ridley I, Chalabi Z, Shrubsole C, Vardoulakis S, Wilkinson P. An Exposure-Mortality Relationship for Residential Indoor PM2.5 Exposure from Outdoor Sources Climate 2017: volume 5, article 66

PHE. Review of interventions to improve outdoor air quality and public health Public Health England 2019

Pope CA III, Burnett RT, Thun MJ, Calle EE, Krewski D, Ito K, Thurston GD. Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution JAMA 2002: volume 287, pages 1,132-1,141

WHO. WHO guidelines for indoor air quality: selected pollutants World Health Organization 2010 [viewed on 27 June 2023]

WHO. WHO global air quality guidelines: particulate matter (‎PM2.5 and PM10)‎, ozone, nitrogen dioxide, sulfur dioxide and carbon monoxide World Health Organization 2021 [viewed on 27 June 2023]

Wong PPY, Lai P-C, Allen R, Cheng W, Lee M, Tsui A, Tang R, Thach T-Q, Tian L, Brauer M, Barratt Ben. Vertical monitoring of traffic-related air pollution (TRAP) in urban street canyons of Hong Kong Science of the Total Environment 2019: volume 670, pages 696-703

Annexe A

Exploration of modifying a CRF for mortality associated with PM_2.5 in outdoor air for application to assessments based on integrated personal exposure concentrations

In our statement, we refer to the modification of CRFs derived from outdoor air measurements for use in assessments based upon integrated personal exposures (µg/m³).

COMEAP (2022) recommends using the CRF from a meta-analysis by Chen and Hoek (2020) to quantify mortality attributable to particulate air pollution:
RR 1.08 (95% confidence interval (CI) 1.06, 1.09) for mortality associated with each 10 µg/m³ change in outdoor concentrations of PM_2.5. This is the starting point for our exploratory calculation.

We have used data collated by Evangelopoulos and co-authors (2020) on the relationship between pollutant concentrations outdoors (C) and mean integrated personal exposure (indoors and outdoors; µg/m³) from pollution from outdoor sources (A) to modify this CRF so that it can be applied to changes in integrated personal exposure (µg/m³) (Table 1). We have used the median, and also 5th and 95th percentiles, of the data from all of the studies reviewed to calculate a range of modified CRFs and 95% confidence intervals (see Table 1).

Table 1. Illustration of modification, for application to personal exposure, of a CRF of RR 1.08 (95% CI 1.06, 1.09) based on outdoor concentrations

PM2.5	A÷C	Modified RR	LCI	UCI
Median	0.64	1.13	1.09	1.14
5th percentile	0.42	1.20	1.15	1.23
95th percentile	0.74	1.11	1.08	1.12

Notes:

A = mean integrated personal exposure (indoors and outdoors, µg/m³) from pollution from outdoor sources in a study

C = outdoor pollutant concentration in a study

A÷C = the ratio in a study between mean personal exposure (indoors and outdoors; µg/m³) from pollution from outdoor sources and outdoor concentrations

LCI = lower 95% confidence interval

UCI = upper 95% confidence interval

This modification is intended as an illustration only and is subject to uncertainties and assumptions. For example, we note that all of the studies in the review by Evangelopoulos and co-authors (2020) were undertaken in North America and Europe, while some of the studies included in the meta-analysis by Chen and Hoek report associations found in studies in Asia.

References

Chen J and Hoek G. Long-term exposure to PM and all-cause and cause-specific mortality: A systematic review and meta-analysis Environment International 2020: volume 143, article 105,974

COMEAP. Statement on quantifying mortality associated with long-term exposure to PM2.5 Committee on the Medical Effects of Air Pollutants 2022

Evangelopoulos D, Katsouyanni K, Keogh RH, Samoli E, Schwartz J, Barratt B, Zhang H, Walton H. PM2.5 and NO2 exposure errors using proxy measures, including derived personal exposure from outdoor sources: A systematic review and meta-analysis Environment International 2020: volume 137, article 105,500

Annexe B

Indoor and outdoor air pollution epidemiology – implications for quantification

In both developing and developed countries, most people spend most of their time indoors, with much of that time being in the home. This makes exposure while indoors a very important contributor to overall pollutant exposure. Epidemiological studies (mostly cross-sectional or case-control studies) examining health effects of exposure indoors are available for some pollutants, particularly those – such as formaldehyde, other VOCs and mould – which mainly arise from sources indoors.

However, the number of epidemiological studies which have directly investigated the effects of exposure to pollutants such as PM and NO₂ indoors is limited, compared with the large epidemiological literature on outdoor concentrations. For indoor environments, there is a body of evidence from lower income countries which focuses on the health effects associated with the high concentrations of particles experienced when using combustion of solid fuels as the heat source for heating or cooking indoors. This is of limited relevance to indoor environments in the UK. There is also a sizeable literature, from higher income countries, investigating the health effects associated with concentrations of NO₂ that arise from burning gas indoors for heating or cooking. These studies provided the evidence base for the derivation of the WHO Air Quality Guidelines for Indoor Air Quality (2010) [5] for NO₂. The majority of these studies have investigated respiratory effects. For example, systematic reviews on studies investigating possible links of increased risks of incidence, prevalence or exacerbation of asthma and wheeze with indoor NO₂ and gas cooking have been published (including Lin and co-authors, 2013; Li and co-authors, 2023).

Because of the need for building- or individual-level exposure data (such as indoor measurements or modelling of indoor concentrations), epidemiological studies on indoor air quality are necessarily limited in size, compared with those that can be undertaken on outdoor pollution. This means that direct evidence linking indoor air pollution with infrequent health outcomes, such as mortality, are lacking. Much of the current epidemiological evidence demonstrating the adverse health effects of classical air pollutants such as PM and NO₂ is, therefore, from studies reporting associations with levels of these pollutants outdoors.

Concentration-response functions (CRFs) from epidemiological studies, or summary CRFs from meta-analyses of effect estimates from many studies, are used to quantify health effects attributable to outdoor air pollutants such as PM_2.5 and NO₂ (for example COMEAP, 2023). Estimates of the health burden attributable to exposure to current levels of air pollution have been useful in informing the public, and policy makers, about the importance of air quality for public health. Health impact assessments (HIA) of the benefits of improving air quality are used in cost-benefit analyses (CBA), which inform policy and operational decision-making.

Methods for quantifying the effects on health attributable to indoor air pollution are also needed. These would allow estimates to be made of the public health burden attributable to exposure to pollutants indoors, and enable assessment of the health impacts (benefits or adverse consequences) of interventions which affect indoor air quality. Such methods are already available for some pollutant-outcome pairs, such as asthma associated with damp and mould, or with formaldehyde (Clark and co-authors, 2023). However, where there is an absence of sufficient epidemiological evidence directly linking indoor air pollution with health outcomes, there might be a need to draw on the outdoor epidemiological evidence to inform quantification of effects of exposures to pollutants indoors and in other microenvironments.

Annexe C

Integrated personal exposure

The exposure of an individual over a 24-hour period is a time-weighted sum of all of the micro-environments in which they have spent time. In this document, we refer to this time- and microenvironment-weighted exposure as the “integrated personal exposure”, which is expressed as a concentration (for example, µg/m³).

Integrated personal exposure, C_p = ΣC_i.t_i/Σt_i, in which C_i is the concentration and t_i the time spent in microenvironment, i.

Supposing that exposure is made up solely of time spent indoors, t_in at a concentration of C_in, and time spent outdoors for time t_out and concentration C_out, then:

C_p = (C_in.t_in + C_out.t_out)/(t_in + t_out)

If, for example C_in/C_out = 0.7 and t_in 16h, and t_out = 8h, then
C_p = Σ(0.7C_out.16 + C_out.8)/24 = 0.8C_out

COMEAP Sub-group on the quantification of air pollution risks in the UK (QUARK)

Chair

Dr Mike Holland (EMRC and Imperial College London) from December 2023 (COMEAP Member until April 2024, co-opted Member from May 2024)

Dr Heather Walton (Imperial College London) until November 2023 

Members

Dr Suzanne Bartington (University of Birmingham) (from April 2024)

David Birchby (Logika Group) (from April 2024)

Dr Dimitris Evangelopoulos (Imperial College London) (co-opted until May 2024, Associate Member from June 2024)

Professor Francesco Forastiere (Imperial College London) (from April 2024)

Professor Klea Katsouyanni (University of Athens, Greece and Imperial College London)

Dr Haneen Khreis (Cambridge University and Texas A&M Transportation Institute) (from April 2024)

Professor Duncan Lee (University of Glasgow)

Dr James Milner (London School of Hygiene and Tropical Medicine) Associate Member (from June 2024)

Professor Gavin Shaddick (Cardiff University)

John Stedman (Ricardo Energy and Environment) Co-opted

Secretariat

Dr Christina Mitsakou (UKHSA)

Alison Gowers (UKHSA)

Dr Artemis Doutsi (UKHSA)

COMEAP Chair

Professor Anna Hansell (University of Leicester)

Acknowledgements

We thank Dr James Milner (London School of Hygiene and Tropical Medicine) for presenting his work on this topic to QUARK, before he became an Associate Member of COMEAP.

Valuable comments from Dr Sierra Clark (City St Georges, University of London, previously UKHSA) are also acknowledged.

---

[1] In this statement we have adopted the convention (widely used elsewhere) to refer to pollution which is experienced outdoors, or has infiltrated indoors from outdoors, as being ‘from outdoor sources’, although it is possible that some may have originated from emissions indoors

[2] The integrated personal exposure of an individual over a whole 24-hour period is a time-weighted sum of the concentrations of all of the micro-environments in which they have spent time – see Annexe C

[3] We note that Evangelopoulos and co-authors (2024) found a slightly (but not statistically significantly) higher Odds Ratio for exacerbations in individuals with chronic obstructive pulmonary disease (COPD) per unit concentration of NO₂ arising from indoor sources than for NO₂ from outdoor sources. This difference may be due to differences in concentrations of co-emitted pollutants arising from indoor and outdoor sources

[4] Note that this represents the same level of risk, expressed using a different exposure metric. It does not suggest that attributable effects are 50% higher

[5] WHO (2010) did not develop specific guidelines for PM_2.5 and PM₁₀ in indoor air, because the Air Quality Guidelines for outdoor air were regarded as also being applicable to indoor spaces