Quantifying the short-term mortality effects of wildfire smoke in Europe: a multicountry epidemiological study in 654 contiguous regions

Quantifying the short-term mortality effects of wildfire smoke in Europe: a multicountry epidemiological study in 654 contiguous regions Anna Alari, Joan Ballester, Carles Milà, Tarik Benmarhnia, Mikhail Sofiev, Andreas Uppstu, Risto Hänninen, Cathryn Tonne Summary Background Fine particulate matter (PM 2⋅5 ) from wildfire smoke could be more harmful to human health than that from other sources. Evidence of the short-term association between wildfire-related PM 2⋅5 and mortality in Europe remains sparse, leading to uncertainties in the fire-related PM 2⋅5 mortality burden. Methods In this retrospective, multicountry epidemiological study, we used the EARLY-ADAPT database to obtain daily mortality records in 654 contiguous subnational regions from 32 European countries, representing a population of 541 million individuals. We combined these data with daily estimates of fire-related and non-fire-related PM 2⋅5 from the System for Integrated Modelling of Atmospheric Composition model during 2004–22. Regional and pooled associations between daily fire-related PM 2⋅5 and all-cause and cause-specific mortality were quantified using quasi-Poisson regression. We compared deaths attributable to fire-related PM 2⋅5 using relative risks (RRs) specific to fire-related PM 2⋅5 versus total PM 2⋅5 . Findings Our study data included 95⸱3 million daily deaths for all-cause mortality, 19⸱5 million daily deaths for cardiovascular mortality, and 3⸱9 million daily deaths for respiratory mortality from Jan 1, 2004, until the latest available year in each country. Pooled cumulative (lags 0–7) RRs associated with a 1 μμg per m 3 increase in fire-related PM 2⋅5 were 1⸱007 (95% CI 1⸱004–1⸱010) for all-cause mortality, 1⸱009 (1⸱006–1⸱013) for cardiovascular mortality, and 1⸱013 (1⸱008–1⸱019) for respiratory mortality. RRs were larger for fire-related PM 2⋅5 than for non-fire-related PM 2⋅5 . Using RRs for total (fire-related and non-fire-related) PM 2⋅5 underestimated the fire-related PM 2⋅5 attributable mortality by 93%. Interpretation Associations with all-cause and cause-specific mortality were larger for fire-related compared with non-fire-related PM 2⋅5 . Assuming wildfire PM 2⋅5 has the same effect as total PM 2⋅5 substantially underestimates the mortality burden of wildfire smoke. Funding The European Commission, EU Horizon Europe, the European Research Council, and the Research Council of Finland. Copyright © 2025 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/). Introduction A feedback loop links climate change and wildfires. Anthropogenically induced climate change is among the leading causes of increasing trends in the frequency and intensity of wildfires, 1,2 creating favourable conditions for wildfire spread and increasing the number of days in which there is very high or extremely high fire danger. Both the size of the population at risk from wildfire smoke 3 and the number of days with very high or extremely high fire danger have increased 4 since 2003 in many regions globally. In Europe, 96% of all ignitions are caused by human activity, either intentionally or accidentally. 5 Regardless of the initial source of ignition, the spread and intensity of wildfires are predominantly influenced by weather and fuel conditions, vegetation, and topography. Wildfires can also release large quantities of CO 2 , 6–8 contributing to climate change and further increasing the likelihood of wildfires. Estimates suggest that in Europe the burned area under no adaptation will increase by 200% during the 21st century. 9 Even under a moderate climate change scenario, southern Europe could experience a tenfold increase in the probability of catastrophic fire, and central and northern Europe could also become more susceptible to wildfires during droughts. 10 Beyond the loss of essential and health-supporting physical infrastructure, economic consequences, 11–13 and direct health effects through injury and mental health consequences of displacement and housing loss, wildfire events indirectly affect population health through smoke exposure over long distances. Wildfires emit a complex mixture of health-damaging components, such as fine particulate matter (PM 2⋅5 ) and a variety of toxic gases. Like ambient air pollution or tobacco smoke, wildfire smoke PM 2⋅5 can lead to inflammation and oxidative stress as well as impairing the immune system response by reducing macrophage capacity. 14 Epidemiological studies provide Lancet Planet Health2025 Published Online https://doi.org/10.1016/ j.lanplh.2025.101296 ISGlobal, Barcelona, Spain (A Alari PhD, J Ballester PhD, C Milà PhD, Prof C Tonne ScD); Pompeu Fabra University, Barcelona, Spain (A Alari, C Milà, Prof C Tonne); Scripps Institution of Oceanography, University of California San Diego, San Diego, CA, USA (Prof T Benmarhnia PhD); Institute for Research in Health, Environment and Work, UMR-S 1085, Inserm, University of Rennes, Rennes, France (Prof T Benmarhnia); Finnish Meteorological Institute, Helsinki, Finland (M Sofiev PhD, A Uppstu PhD, R Hänninen DSci); CIBER Epidemiology and Public Health, Madrid, Spain (Prof C Tonne) Correspondence to: Prof Cathryn Tonne, ISGlobal, Barcelona 08003, Spain [email protected] Articles www.thelancet.com/planetary-health Vol ▪ ▪ 2025 1 evidence that short-term exposure to wildfire PM 2⋅5 is significantly associated with increased risk of mortality and morbidity, particularly respiratory morbidity. 15–17 Most previous studies estimating exposure–response functions (ERFs) for the short-term association between fire-related PM 2⋅5 and mortality used data from specific states within the USA, 18–24 or in the southern hemisphere, 25–27 where patterns of exposure to fire-related PM 2⋅5 differ from those observed in Europe. 28 Only two large multilocation studies reported European-specific associations between fire-related PM 2⋅5 and mortality, estimated for a large set of European cities and pooled at country and European subregion levels. 29,30 In these studies, short-term exposure to fire-related PM 2⋅5 was significantly associated with increased all-cause, cardiovascular, and respiratory mortality in Europe and associations were higher than those observed in the non-European locations. However, these two multicity studies covered only a small proportion of European countries, with low representation of Eastern Europe, which is heavily affected by wildfire smoke. Moreover, results based on data covering only cities might not adequately represent fire-affected populations in rural, often inland, areas. The epidemiological literature linking wildfire smoke and health effects remains sparse in Europe. Estimates of the ERF for the short-term association between wildfire PM 2⋅5 and mortality, as well as the related attributable health burden, are inadequate due to insufficiently representative data. Research in context Evidence before this study Wildfire smoke contains hazardous pollutants, including fine particulate matter (PM 2⋅5 ), which have been linked to a range of health issues, including increased mortality and morbidity. Previous research indicates that PM 2⋅5 from wildfire smoke can be up to ten times more harmful than PM 2⋅5 from other sources, such as traffic emissions. Not accounting for exposure–response functions (ERFs) specific to wildfire smoke in health impact assessment may substantially underestimate its health burden. We searched PubMed and Google Scholar on Dec 15, 2024, for studies published between Jan 1, 2000, and Dec 15, 2024, using combinations of the terms “wildfires” or “landfires”, “daily” or “short-term”, and “mortality”, without language or time restrictions. Many identified studies estimated the mortality burden attributable to fire-related PM 2⋅5 using ERFs for total PM 2⸱5 derived from systematic reviews and meta-analyses of studies of total PM 2⋅5 –mortality association. Among studies that estimated ERFs for the short-term association between fire-related PM 2⋅5 and mortality, most used data from specific states within the USA. Studies including only European data reported mixed results for fire-related PM 2⋅5 –mortality association. In two large multilocation studies including numerous European cities, shortterm exposure to fire-related PM 2⋅5 was significantly associated with increased all-cause, cardiovascular, and respiratory mortality. The strongest associations were observed in Europe for both allcause and cause-specific mortality. In a study estimating the mortality burden attributable to both short-term and long-term exposure to wildfire smoke (also including fire-related ozone), an increasing trend of all-cause, cardiovascular, and respiratory attributable mortality was reported for the Mediterranean region, and the highest cardiovascular attributable mortality burden was reported for Eastern Europe. However, these two multicity studies covered only a small proportion of European countries, with low representation of Eastern Europe, which is heavily affected by wildfire smoke. Moreover, by including data only on cities, their results might not represent fire-affected populations in rural, often inland, areas. Added value of this study Available evidence for Europe is based on subnational studies or multicountry studies including data only from cities. We used daily mortality records in 654 contiguous regions from 32 European countries, representing entire national populations (541 million individuals in total). Our dataset includes contiguous urban and rural subnational regions, providing more representative estimates compared with those based on urban areas alone. We also compared fire-related PM 2⋅5 -attributable deaths estimated using relative risks (RRs) derived specifically from fire-related PM 2⋅5 exposure versus those based on total PM 2⋅5 exposure (including both fire-related and non-fire-related sources). Pooled cumulative (lags 0–7) RRs associated with a 1 μg per m 3 increase in fire-related PM 2⋅5 were 1⸱007 (95% CI 1⸱004–1⸱010) for all-cause mortality, 1⸱013 (1⸱008–1⸱019) for respiratory mortality, and 1⸱009 (1⸱006–1⸱013) for cardiovascular mortality. RRs per 1 μg per m 3 increase were higher for fire-related compared with non-fire-related PM 2⋅5 for all-cause and causespecific mortality. Consequently, using RRs for total PM 2⋅5 underestimated the fire-related PM 2⋅5 -attributable mortality by 93%. Implications of all the available evidence Drawing on data encompassing the entire population across 654 continuous regions, our findings provide strong evidence that exposure to fire-related PM 2⋅5 is significantly associated with increased all-cause, cardiovascular, and respiratory mortality in Europe. Additionally, assuming that fire-related PM 2⋅5 has the same RR as total PM 2⋅5 substantially underestimates the number of deaths attributable to fire-related PM 2⋅5 . With climate change expanding wildfire-prone areas, intensifying individual fires, extending fire seasons beyond summer, and sparking fires in previously unaffected regions, improved estimates of fire-related PM 2⋅5 -attributable mortality are needed to track the public health burden of this climate hazard and effectiveness of adaptation strategies. Articles 2 www.thelancet.com/planetary-health Vol ▪ ▪ 2025 Previous estimates of attributable numbers of deaths due to wildfire smoke in Europe 31 have used ERFs for shortterm exposure to total PM 2⋅5 (not distinguishing wildfire smoke from other sources) based on available systematic reviews and meta-analyses. 32 Due to differences in chemical composition, particle size, and oxidative potential, the toxicity of wildfire smoke PM 2⋅5 could be higher than that of PM 2⋅5 from other sources. 33 PM 2⋅5 from wildfire smoke was reported to be up to ten times more dangerous than PM 2⋅5 emitted from other sources among both adults and children in California, USA. 34,35 Assuming that PM 2⋅5 from wildfire smoke has the same toxicity as total PM 2⋅5 can substantially underestimate attributable health burdens. 36 As wildfire smoke becomes more prevalent in a changing climate, there is a need for robust Europe-specific estimates of the ERF between wildfire smoke and mortality capturing potential heterogeneity across European countries. Such ERFs specific to wildfire smoke will have direct relevance in health impact estimates, such as those used in the Lancet Countdown for Health and Climate Change in Europe, 31 and for monitoring the effectiveness of fire control strategies to protect public health. We addressed these gaps using a multicountry dataset including daily mortality data from 32 countries and daily estimates of fire-related and non-fire-related PM 2⋅5 . First, we quantified associations between fire-related and nonfire-related PM 2⋅5 and all-cause and cause-specific mortality. Second, we compared the number of deaths attributable to wildfire PM 2⋅5 using ERFs specific to wildfire smoke versus those attributable to total PM 2⋅5 . Methods Mortality data In this retrospective, multicountry epidemiological study, we used daily subnational regional mortality data assembled through the European Research Council project EARLY-ADAPT. 37 This spatiotemporally homogeneous daily regional mortality database contains more than 164 million counts of deaths from 654 contiguous regions in 32 European countries, representing their entire urban and rural population of more than 541 million people. For the purpose of this study, data from 2004 until the latest available year in each country were selected to match the period with available exposure data in 32 European countries. Data were available at different regional levels based on the European Nomenclature of Territorial Units for Statistics (NUTS) classification: the finest-level (NUTS3) resolution was available for 16 countries, NUTS2 for nine countries, NUTS1 for four countries, and NUTS0 for two countries (appendix p 6). For the UK, data were available at the NUTS3 level for Scotland, at NUTS2 for Northern Ireland, and at NUTS1 for England and Wales. Daily counts for all-cause mortality were available for all 32 countries and daily numbers of deaths for cardiovascular causes (ICD-10 codes I00–I99) and for respiratory causes (ICD-10 codes J00–J99) were available for a subset of 22 countries. Fire-related PM 2⋅5 and other covariates We computed daily averages for fire-related PM 2⋅5 , other non-fire-related PM 2⋅5 , and ozone (O 3 ) concentrations (μg per m 3 ), along with daily averages for temperature (◦C) and relative humidity (%), from Jan 1, 2004, until Dec 31, 2022. Estimates were obtained from the System for Integrated Modelling of Atmospheric Composition model (SILAM, version 5.9.1) developed by the Finnish Meteorological Institute. SILAM is an offline global-to-local chemistry transport model equipped with a variety of source modules, including anthropogenic and natural emissions, as well as a dedicated fire source named IS4FIRES. The spatial resolution of the product is 0⋅1◦(approximately 10 km) and the temporal resolution is 1 h. 38 Briefly, IS4FIRES processes the fire radiative power products of Moderate Resolution Imaging Spectroradiometer spaceborne instruments to multipollutant smoke emissions using empirical emission factors, diurnal fire intensity profiles, a smoke injection model, etc. Emissions are then used by SILAM for computing plume dispersion, chemical transformations, and deposition. The meteorological data driving the SILAM dispersion simulations are taken from the European reanalysis ERA5 at 0⸱25◦resolution (approximately 25 km). SILAM has been used for health impact assessments and extensively evaluated in many regional and global studies, showing robust performance. 39–43 Real-time operational applications of SILAM are also evaluated within the scope of corresponding application projects and operational services (eg, the Copernicus Atmosphere Monitoring Service and the Barcelona Dust Regional Center). Outputs from this model have previously been used to estimate the health burden of wildfire smoke in Europe (appendix pp 2–5). 4,13,44 We used GEOSTAT population grids (1 km resolution) for Europe from 2006, 2011, and 2018 to estimate population-weighted exposures at the corresponding NUTS level. We used the 2021 edition of the Eurostat Geographic Information System of the Commission NUTS boundaries for Europe at a scale of 1:1 000 000. Statistical analysis To evaluate the short-term association between daily firerelated PM 2⋅5 and counts of mortality, we used distributed lag non-linear models to allow for autocorrelated lagged exposure and non-linearity in the ERF and lag structure. For each mortality outcome (all-cause, respiratory, and cardiovascular mortality), we ran a two-stage analysis to obtain European, regional (ie, NUTS-level), and national relative risks (RRs). We used the estimated regional RRs, as well as the exposure and mortality data until the most recent year with available data (appendix p 6), to obtain the number of deaths attributable to fire-related PM 2⋅5 . In the first stage, quasi-Poisson regressions were fitted to estimate the regional-specific association between daily concentrations of fire-related PM 2⋅5 and mortality at the See Online for appendix Articles www.thelancet.com/planetary-health Vol ▪ ▪ 2025 3 smallest spatial scale available (654 contiguous NUTS regions), as follows: E[mort]=exp(θ0+θ1dow + + ns(time,8 df /year) +ns(at07,3df )+cb(firePM2⸱5) +cb(nonfirePM2⸱5)) We a priori retained lagged effects up to 7 days based on previous studies 29,30 and evaluated different lag and ERFs according to the quasi-likelihood version of the Akaike information criterion. Guided by preliminary model diagnostics, fire-related PM 2⋅5 concentrations were included in the final model using a cross-basis function (cb), with a linear function for the exposure–response dimension and a natural cubic spline with intercept and two df equally spaced in the log scale for 7 lag days. We accounted for the day of the week (dow) and used a natural cubic spline of time with eight df per year to account for seasonal and longer-term time trends in the model, on the basis of preliminary analyses. We controlled for apparent temperature (at, surface air temperature combined with the level of humidity to represent temperature perceived by a typical human) 45,46 using the moving average of the current day and the previous 7 days (according to previous studies 29,30 ) through a natural cubic spline with three df. In order to account for potential confounding by PM 2⋅5 from other sources, daily non-fire-related PM 2⋅5 was included in the model using the same functional form as for fire-related PM 2⋅5 to facilitate comparisons of the associations. In the second stage, we ran a meta-analysis to obtain overall pooled estimates at European level from the regional-specific estimates, including two levels of random effects by regions nested within countries. We finally used the fitted meta-analysed models to obtain the best linear unbiased predictions of the overall cumulative and lagspecific exposure–response associations in each NUTS region. We also obtained country-level and macroregional estimates based on the four geographical European subregions of the UN geoscheme (Southern, Eastern, Western, and Northern Europe), by directly pooling regional results within each country or macroregion. As additional analyses, we also pooled regional estimates according to rural– urban typologies, based on the Geographic Information System classifications from Eurostat. 47 We presented associations with mortality as cumulative (ie, summing all the contributions up to the indicated maximum lag) and lag-specific RRs with their corresponding 95% CIs. RRs were expressed for an increase from 0 to 1 μg per m 3 . The proportion of deaths attributable to short-term exposure to wildfire-related PM 2⋅5 was calculated using the population attributable fraction (PAF) based on the regional-level best linear unbiased prediction RRs (appendix p 7). We estimated the number of annual deaths attributable to fire-related PM 2⋅5 for each region and aggregated them to obtain the number of annual deaths attributable to fire-related PM 2⋅5 at the country level. We computed 1000 Monte Carlo simulations of the regional numbers of death attributable to fire-related PM 2⋅5 and we separately aggregated them in each simulation to calculate the 95% CIs at the national and European levels. 48–50 We compared wildfire attributable deaths estimated using RRs based on fire-related and total PM 2⋅5 (fire-related plus non-fire-related). To make this comparison, we replicated the main analysis with total PM 2⋅5 as the main exposure. We used regional-level estimates for the association between total PM 2⋅5 and mortality to compute alternative PAFs and the number of annual deaths attributable to fire-related PM 2⋅5 . We ran several sensitivity analyses. First, we assessed the robustness of our estimates to different covariate adjustments by comparing estimates from our main model with estimates from a model that did not adjust for non-firerelated PM 2⋅5 , and with estimates obtained by adjusting for daily ozone concentrations. Second, we restricted the estimation of regional-level and Europe-level RRs to data from 2004 to 2019 in order to check the influence of the COVID-19 pandemic. Third, we assessed non-linearity in the exposure–response dimension of our effect estimates, using a natural spline with knots located at different values of the overall fire-related PM 2⋅5 distribution (at the 75th percentile, the 99th percentile, and both). Finally, we extended the lag–response function up to lag 10. Analyses were conducted with R and RStudio software (version 4.3.0) using the packages dlnm and mixmeta for meta-analysis. Syntax to replicate the analysis are available online. Role of the funding source The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. Results During the study period, 95⋅3 million deaths for all causes, 19⋅5 million deaths for cardiovascular causes, and 3⋅9 million deaths for respiratory causes were recorded in the dataset (appendix p 6). All countries had data until at least the end of 2020 except for Greece (until 2018) and Italy (until 2019). Daily regional fire-related PM 2⋅5 concentrations were highly skewed; regional median values over the whole study period ranged from 0⸱003 μg per m 3 to 0⸱13 μg per m 3 and 99% of the overall observations were below 2⸱8 μg per m 3 . The 99th percentile of daily regional fire-related PM 2⋅5 concentrations was generally higher in countries in Southern and Eastern Europe—reaching 21 μg per m 3 in northern Portugal (figure 1)—and was also slightly higher in rural and intermediate regions (2⸱9 μg per m 3 ) compared with predominantly urban areas (2⸱4 μg per m 3 ). Within-region variability was largest in Eastern Europe; a NUTS region in Romania (RO317) had the largest IQR (0⸱6). Distributions of daily fire-related and nonfire-related PM 2⋅5 differed considerably (appendix p 13). Regional median concentrations of daily non-fire-related For syntax to replicate the sensitivity analyses see https:// github.com/aalari/wildfirePM2.5 Articles 4 www.thelancet.com/planetary-health Vol ▪ ▪ 2025 PM 2⋅5 ranged from 1 μg per m 3 to 29 μg per m 3 , with regional IQRs ranging from 1⸱4 μg per m 3 to 44⸱9 μg per m 3 (European median 10⸱7 μg per m 3 [IQR 13⸱6]). Exposure to fire-related PM 2⋅5 was associated with higher mortality risks than non-fire-related PM 2⋅5 , for both allcause and cause-specific mortality. A 1 μg per m 3 increase in fire-related PM 2⋅5 was associated with a 0⸱7% (95% CI 0⸱4–1⸱0) increase in all-cause mortality, a 1⸱3% (0⸱8–1⸱9) increase in respiratory mortality, and a 0⸱9% (0⸱6–1⸱3) increase in cardiovascular mortality in cumulative (lags 0–7) Europe-wide pooled estimates (table 1). A 1 μg per m 3 increase in the same-day (lag 0) exposure to fire-related PM 2⋅5 was associated with an increase of 0⸱7% (0⸱5–0⸱9) in all-cause mortality, 0⸱8% (0⸱5–1⸱2) in respiratory mortality, and 0⸱7% (0⸱5–0⸱9) in cardiovascular mortality. Lag 1 and lag 2 exposure were also associated with increased allcause and cause-specific mortality. Beyond lag 2, exposure to fire-related PM 2⋅5 was associated with increased respiratory mortality, whereas negative or null associations were observed at longer lags for all-cause mortality and cardiovascular mortality (table 1, appendix p 14). When considering cumulative associations until lag 2, a 1 μg per m 3 increase in fire-related PM 2⋅5 was associated with an increase of 1⸱1% (0⸱8–1⸱4) in all-cause mortality, 1⸱3% (0⸱7–1⸱8) in respiratory mortality, and 1⸱2% (0⸱8–1⸱6) in cardiovascular mortality (table 1). Regional estimates for the cumulative (lag 0–7) associations with all-cause mortality were positive in most countries and statistically significant in 457 (70%) NUTS regions, most located in Western and Eastern Europe (figure 2). Regions in Portugal, Bosnia and Herzegovina, Latvia, Estonia, and Spain showed non-statistically significant negative RRs (figure 2). The largest regional and national RRs were observed in Eastern and Western Europe (figures 2, 3, appendix pp 9–10), with regional RRs for a 1 μg per m 3 increase reaching 1⸱019 (95% CI 1⸱011–1⸱026) in a Serbian region (NUTS RS221) and 1⸱019 (1⸱008–1⸱030) in a Belgian region (NUTS BE211). Regional RRs for respiratory mortality were generally higher than RRs for allcause mortality (except in Belgium) and were largest in Eastern Europe (Serbian, Romanian, and Bulgarian regions, and Hungary; figure 4, appendix p 15). RRs for allcause and cardiovascular mortality were similar in magnitude (figure 4, appendix p 15) and were also largest in Eastern Europe (especially in Hungary and Serbia). RRs showed significant heterogeneity across regions for allcause and cardiovascular mortality but not respiratory mortality (appendix p 8). Magnitude of pooled RRs did not vary by degree of urbanisation (appendix p 16). For a 1 μg per m 3 increase, regional RRs were considerably larger for A B C 0–3 μg/m3 3–6 6–9 9–12 12–15 15–18 18–21 0·0–0·1 μg/m3 0·1–0·2 0·2–0·3 0·3–0·4 0·4–0·5 0·5–0·6 0·0–0·2 % 0·2–0·4 0·4–0·6 0·6–0·8 0·8–1·0 1·0–1·2 Figure 1: Regional 99th percentile (A) and IQR (B) of daily fire-related PM 2⋅5 concentrations, and average ratio of fire-related to total PM 2⋅5 concentrations (C) Light grey areas indicate regions with no data. PM 2⋅5 =fine particulate matter. All-cause mortality Respiratory mortality Cardiovascular mortality 0 1⸱007 (1⸱005–1⸱009) 1⸱008 (1⸱005–1⸱012) 1⸱007 (1⸱005–1⸱009) 1 1⸱004 (1⸱003–1⸱005) 1⸱005 (1⸱003–1⸱007) 1⸱004 (1⸱003–1⸱005) 2 1⸱001 (1⸱001–1⸱002) 1⸱003 (1⸱001–1⸱004) 1⸱002 (1⸱001–1⸱002) 3 1⸱000 (0⸱999–1⸱000) 1⸱001 (1⸱000–1⸱002) 1⸱000 (1⸱000–1⸱001) 4 0⸱999 (0⸱998–0⸱999) 1⸱000 (1⸱000–1⸱001) 0⸱999 (0⸱999–1⸱000) 5 0⸱999 (0⸱998–0⸱999) 1⸱000 (1⸱000–1⸱001) 0⸱999 (0⸱999–1⸱000) 6 0⸱999 (0⸱999–0⸱999) 1⸱001 (0⸱999–1⸱002) 0⸱999 (0⸱999–1⸱000) 7 0⸱999 (0⸱999–1⸱000) 1⸱001 (0⸱999–1⸱003) 1⸱000 (0⸱999–1⸱000) 0–2 1⸱011 (1⸱008–1⸱014) 1⸱013 (1⸱007–1⸱018) 1⸱012 (1⸱008–1⸱016) 0–7 1⸱007 (1⸱004–1⸱010) 1⸱013 (1⸱008–1⸱019) 1⸱009 (1⸱006–1⸱013) Data are relative risk (95% CI). Table 1: Lag-specific and cumulative (lag 0–2 and lag 0–7) pooled relative risk for 1 μμg per m 3 between firerelated fine particulate matter and mortality Articles www.thelancet.com/planetary-health Vol ▪ ▪ 2025 5 A B 0·993–1·000 RR 1·000–1·002 1·002–1·006 1·006–1·010 1·010–1·015 1·015–1·019 1·0000–1·0002 RR 1·0002–1·0004 1·0004–1·0006 1·0006–1·0008 1·0008–1·0010 Bosnia and Herzegovina Croatia Cyprus Greece Spain Italy Montenegro Portugal Serbia Slovenia Bulgaria Czechia Hungary Poland Romania Slovakia Austria Belgium Switzerland Germany Luxembourg Netherlands France Denmark Estonia Finland Ireland Lithuania Latvia Norway Sweden UK C 1·01 1·021·00 RR for fire-related PM 2·5 1·0000 1·0025 1·0050 1·0075 1·0100 RR for other PM 2·5 Articles 6 www.thelancet.com/planetary-health Vol ▪ ▪ 2025 fire-related compared with non-fire-related PM 2⋅5 (figure 2, appendix p 11), particularly for cause-specific mortality (appendix p 17). We estimated that, on average, 535 (95% CI 151 to 928) annual deaths from all causes, 31 (–7 to 68) from respiratory causes, and 184 (69 to 303) from cardiovascular causes were attributable to short-term exposure to fire-related PM 2⋅5 during the study period (table 2), corresponding to PAFs of 0⋅16% (95% CI 0⋅05 to 0⋅28) for all-cause mortality, 0⋅21% (–0⋅04 to 0⋅46) for respiratory mortality, and 0⋅26% (0⋅10 to 0⋅43) for cardiovascular mortality (appendix p 12). The highest PAFs were observed in Eastern Europe countries (Bulgaria, Romania, and Hungary) and Serbia. Predominantly rural regions showed higher PAFs compared with urban regions, especially for cause-specific mortality (appendix p 12). When using RRs for total (fire-related plus non-fire-related) PM 2⋅5 , the average number of deaths per year from all causes attributable to fire-related PM 2⋅5 was around 14 times smaller, ie, 38 (8 to 69), indicating the higher toxicity of fire-related PM 2⋅5. Cumulative (lag 0–7) associations between fire-related PM 2⋅5 and mortality were robust to different confounder adjustment and the exclusion of data from more recent years (appendix p 18). When using a non-linear model with a knot on the 75th percentile of the exposure distribution (corresponding to a concentration level of 0⋅15 μg per m 3 ), estimated cumulative (lag 0–7) associations were closer to null and lag-specific effects were larger for more immediate lag periods. However, in most NUTS regions analysed— 60% for all-cause, 81% for respiratory, and 74% for cardiovascular mortality—the quasi-likelihood Akaike information criterion values suggested a better fit of the linear model used in the main analysis, consistent with findings from previous studies. 26,29,30 The pattern of the lag–response curve was broadly consistent across sensitivity analyses. When considering different cumulative lag span, cumulative RRs were consistent with RRs in the main analysis (table 1, appendix p 11), with cumulative lag 0–2 RRs generally higher than lag 0–7 RRs, especially in Western Europe (appendix pp 9–10). Discussion Our analysis resulted in several key findings. First, our results provide robust evidence that for the same change in mass concentration, the RR is larger for fire-related compared with non-fire-related PM 2⋅5 for all-cause and cause-specific mortality. Second, the association between fire-related PM 2⋅5 and mortality was stronger for respiratory compared with all-cause or cardiovascular mortality. Third, assuming that fire-related PM 2⋅5 has the same RR as total PM 2⋅5 underestimates the number of deaths attributable to fire-related PM 2⋅5 by 93%. Although our European-pooled results are not directly comparable due to differences in populations covered, our rescaled cumulative RR estimates for all-cause mortality in several individual countries align closely with those reported in previous studies. For example, after rescaling to a 10 μg per m 3 increase in exposure, our lag 0–2 RR estimates for France (1⋅255 [95% CI 1⋅207–1⋅293]), Italy (1⋅138 [1⋅116–1⋅172]), and Romania (1⋅149 [1⋅127–1⋅172]) were very similar in magnitude to those reported in a global, multicity study by Chen and colleagues (2021), 29 who found corresponding values of 1⋅253 (1⋅079–1⋅455) for France, 1⋅139 (1⋅081–1⋅199) for Italy, and 1⋅116 (1⋅075–1⋅158) for Romania. Moreover, our results are consistent with previous literature indicating larger associations between firerelated and total PM 2⋅5 . In the same study, 29 the RR between wildfire PM 2⋅5 and all-cause mortality per 10 μg per m 3 PM 2⋅5 was 1⋅019 (1⋅016–1⋅022) for the moving average of lag 0–2 days, whereas in a separate analysis 51 using similar data and modelling approaches, the RR for total PM 2⋅5 was 1⋅0068 (1⋅0059–1⋅0077) for the moving average of lag 0–1 days. A study based on data from southern California used three different techniques to estimate associations between wildfire and non-wildfire PM 2⋅5 with respiratory hospital admissions. Per 10 μg per m 3 increase, the increase in respiratory admissions ranged from 1⋅3% to 10% for wildfire PM 2⋅5 compared with 0⸱67% to 1⋅3% for nonwildfire PM 2⋅5 . Possible explanations for larger associations between wildfire compared with non-wildfire or total PM 2⋅5 include differences in toxicity, reflected through a higher oxidative potential, 52–54 and differences in the duration and level of exposure. For example, exposure to wildfire PM 2⋅5 is frequently close to zero in most locations, although can be very high during fire episodes spanning days or weeks. This pattern is distinct from that of total PM 2⋅5 , which represents a continuous, non-zero exposure. Differences in these exposure patterns might induce different biological mechanisms. 55 A growing body of literature indicates that wildfire PM 2⋅5 exposure is particularly relevant for respiratory health outcomes. 56 A 2024 systematic review 55 and meta-analysis reported stronger increases in respiratory hospital admissions (0⋅25% [95% CI 0⋅09–0⋅42]) and emergency visits (0⋅36% [0⸱19–0⋅53]) compared with all-cause mortality (0⋅15% [0⋅01–0⋅28]) and cardiovascular hospital admissions (0⋅06% [0⋅00–0⋅12]) per 1 μg per m 3 wildfire PM 2⋅5 . These findings are consistent with our results, which indicate a stronger association between respiratory mortality compared with all-cause and cardiovascular mortality. The biological mechanisms underpinning larger effects for respiratory outcomes are not fully understood. Possible factors include lung injury via oxidative stress, local and systemic inflammation, airway epithelium compromise, Figure 2: Cumulative RR of all-cause mortality in relation to fire-related and non-fire-related PM 2⋅5 Cumulative (lag 0–7) RR of all-cause mortality for a 1 μg per m 3 increase in fire-related PM 2⋅5 (A) and non-fire-related PM 2⋅5 (B), and comparison of their values across European regions (C). European Nomenclature of Territorial Units for Statistics regions marked with a star have RR estimates that are significantly different from 1. Symbol colours in (C) represent areas of Europe based on the M49 code of the UN Statistics Division geoscheme: red for Southern Europe countries, yellow for Eastern Europe countries, blue for Western Europe countries, and green for Northern Europe countries. PM 2⋅5 =fine particulate matter. RR=relative risk. Articles www.thelancet.com/planetary-health Vol ▪ ▪ 2025 7 and increased vulnerability to infection. 57,58 Exposure to wildfire particulate matter has been consistently associated with asthma and chronic obstructive pulmonary disease exacerbations, 12,56,57,59,60 with larger associations compared with other particulate matter, probably due to more abundant oxidative and pro-inflammatory components. 61,62 Compared with results in the most recent meta-analysis, 55 our cumulative association based on a 1 μg per m 3 increase in fire-related PM 2⋅5 was considerably larger for all-cause mortality and very similar for cardiovascular mortality. We observed some heterogeneity across regions in the associations between exposure to fire-related PM 2⋅5 Southern Europe Eastern Europe Western Europe Northern Europe Europe Pooled Spain Slovenia Serbia Montenegro Portugal Italy Greece Cyprus Croatia Bosnia and Herzegovina Pooled Slovakia Romania Poland Hungary Czechia Bulgaria Pooled Switzerland Netherlands Luxembourg Germany France Belgium Austria Pooled UK Sweden Norway Lithuania Latvia Ireland Finland Estonia Denmark Pooled 1·00 1·02 1·040·98 RR for 1 µg/m 3 increase in fire-related PM 2.5 Figure 3: Pooled estimates for cumulative RR of all-cause mortality Country-level and macro-regional (blue dashed line) and Europe-wide (red dashed line) pooled estimates for cumulative (lag 0–7) RR of all-cause mortality for a 1 μg per m 3 increase in fire-related PM 2⋅5 . PM 2⋅5 =fine particulate matter. RR=relative risk. Articles 8 www.thelancet.com/planetary-health Vol ▪ ▪ 2025 and all-cause and cardiovascular mortality. In particular, negative and imprecise associations were observed in highly exposed regions in Portugal and Spain. Possible explanations can be related to regional and national adaptation and fire management strategies. More research is needed on the factors driving spatiotemporal variability in fire-related PM 2⋅5 –mortality associations. Wildfire events can also elevate tropospheric ozone levels, a potent oxidant linked to increased mortality and morbidity. Consistent with observations from a study on adverse respiratory health effects during the 2008 northern California wildfires, 63 our analysis observed significant associations between fire-related PM 2⋅5 and mortality after adjusting for ozone, although there was some attenuation of the associations. This finding suggests that fire-related Correlation=0·69 A Correlation=0·81 B Croatia Cyprus Greece Serbia Slovenia Spain Bulgaria Czechia Poland Romania Slovakia Hungary Austria Belgium Switzerland Luxembourg Estonia Finland Lithuania Latvia Sweden UK 1·00 1·01 1·02 1·03 RR for all-cause mortality 1·00 1·01 1·02 1·03 RR for all-cause mortality 1·031·021·011·00 RR for cardiovascular mortality 1·01 1·02 1·031·00 RR for respiratory mortality Figure 4: Regional cumulative RRs of all-cause versus cause-specific mortality Regional cumulative (lags 0–7) RRs of all-cause mortality versus cause-specific mortality: respiratory causes (A) and cardiovascular causes (B). RRs are expressed for an increase in fire-related PM 2⋅5 from zero up to 1 μg per m 3 . Symbol colours represent areas of Europe based on the M49 code of the UN Statistics Division geoscheme: red for Southern Europe countries, yellow for Eastern Europe countries, blue for Western Europe countries, and green for Northern Europe countries. PM 2⋅5 =fine particulate matter. RR=relative risk. Articles www.thelancet.com/planetary-health Vol ▪ ▪ 2025 9