Maternal Exposure to Local, Source-specic Ambient Air Pollution During Pregnancy and Autism in Children: a Cohort Study From Southern Sweden

Evidence of air pollution exposure, namely, ambient particulate matter (PM), during pregnancy and an increased risk of autism spectrum disorders (ASD) in children is growing. Which PM sources contribute to this association is currently unknown, however. The aim of the present study was to investigate local source-specic ambient PM exposure and its association with ASD. A cohort of 48,571 singleton births from 1999 to 2009 in Scania, Sweden, was combined with data on locally emitted PM with an aerodynamic diameter <2.5 µm (PM 2.5 ). A at, two-dimensional dispersion model was used to model PM 2.5 exposure (all-source PM 2.5 , tailpipe exhaust, vehicle wear-and-tear, and small-scale residential heating- mainly wood burning) at the residential address. Associations were analyzed using binary logistic regression in partially and fully adjusted models. Clear associations were observed between PM 2.5 and ASD, with statistical signicance for all investigated sources in the partially adjusted model. In the fully adjusted model, only all-source PM 2.5 was statistically signicant. The results add to existing evidence that exposure to air pollution during pregnancy may be associated with an increased risk for ASD among children. Further, these ndings suggest that locally emitted tailpipe exhaust, vehicle wear-and-tear, and small-scale residential heating all contribute to this association.


Introduction
Air pollution is a major cause of morbidity and mortality worldwide 1 . For example, particulate matter (PM) with an aerodynamic diameter less than 2.5 µm (PM 2.5 ) derived from fossil fuels alone has recently been estimated to contribute to 10.2 million premature deaths around the world each year, which corresponds to one in ve deaths 2 . The relationship between tra c-related air pollution, often tailpipe emissions stemming from diesel fuel and petrol combustion, and adverse health effects are well established 3,4 .
Aside from tra c and fossil fuels, wood burning also produces PM as well as other toxic compounds [5][6][7] , yet little is known about the health effects of ambient wood smoke. This is generally because prior research on the health impacts of PM 2.5 has typically been based on exposure from aggregate particle mass rather than speci c sources. However, there is evolving evidence indicating that health effects may vary in type and degree depending on the source of PM 2.5 exposure 8 . Source-speci c epidemiological studies are still scarce, though, as such separation requires high quality exposure data that has not been available previously. Because air pollution is far from static but varies substantially in time and space, advanced exposure assessment can better capture local exposures' unique composition. Indeed, an advantage of using such methods is the ability to distinguish the diverse health effects of distinct, speci ed sources, rather than from the total concentration. For instance, associations between vehicle exhaust particles and low birth weight have been observed in a Swedish study using advanced sourcespeci c exposure data 9 . Associations between ambient wood smoke, and their related markers, and both dementia and childhood asthma 10,11 have also been found in other studies using advanced exposure assessment data.
While all people are susceptible to the adverse health effects of air pollution, pregnant women and their developing fetuses may be particularly vulnerable. For instance, exposure to various air pollutants has been associated with a range of adverse pregnancy 12 and birth outcomes 13 . An experimental cell study reported that intracellular accumulation of PM may contribute to the placental dysfunction associated with unfavorable pregnancy outcomes, such as preeclampsia and intrauterine growth restriction, through their direct and indirect effects on trophoblast protein secretion, hormone regulation, in ammatory response, and mitochondrial interference 14 . Considering source-apportionment studies, associations have been identi ed between PM 10 from brake dust and combined tra c-related sources with preeclampsia 15 .
Preterm birth was found to be associated with total PM 2.5 and PM 2.5 derived from tra c, oil combustion, as well as secondary sulfates and organics (derived from a reaction of ammonia from livestock and sulfuric acid from fuel) 16 . Moreover, PM 2.5 total mass, secondary ammonium sulfate, secondary ammonium nitrate (derived from a reaction of ammonia and nitric acid from vehicular sources), and resuspended soil (i.e., road dust) were associated with low birth weight 17 . A further source-apportionment study found associations between stillbirth and vehicular emissions, re-suspended soil, and secondary ammonium sulfate 18 . Beyond these immediate complications, exposure to air pollution during pregnancy can lead to adverse health outcomes later in the child's life. Indeed, the in-utero period can affect immunologic, respiratory, gastrointestinal, and central nervous system development; delay motor, hearing, visual, cognitive, behavioral, and social-emotional function; and impact their overall health and wellbeing 19 . This has been termed "fetal programming" 20 .
With this, an additional health outcome has been considered in connection to air pollution exposure during pregnancy and in early life: autism spectrum disorders (ASD). Evidence of this association has been developing, with a 2016 systematic review and meta-analysis supporting the association but emphasizing cautious interpretation due to the relatively small sample of available studies 21 . A 2019 systematic review and meta-analysis including both adults and children has demonstrated clearer associations, naming ASD among other neurological and developmental disorders associated with PM 2.5 exposure 22 . Two, more current, systematic reviews and meta-analyses focusing on children only and prenatal exposure speci cally identi ed the strongest evidence for PM 2.5 exposure and ASD 23,24 , with one reporting each 5 µg/m 3 incremental increase in PM 2.5 leading to a 5%, 7%, or 15% increased risk of ASD in newborns depending on the model used 24 . Prenatal exposure to nitrogen oxides (NO X ), mainly from tra c, has also been associated with an increased risk of ASD among children in a low-exposure setting 25 .
ASD development is poorly understood, but mounting evidence has emphasized the importance of genetic and environmental factors 26 . With the latter being modi able, research on the role of the environment, particularly air pollution, has increased. Suggested biological mechanisms include oxidative stress, in ammation (neuro-in ammation and systemic in ammation), neurotoxicity, and endocrine disruption [27][28][29][30] . Animal models also support developmental neurotoxicity as a plausible pathway 31,32 , and one such study has demonstrated ASD traits in mice following prenatal exposure to diesel exhaust 33 .
Regarding the underlying pathways of prenatal air pollution exposure in particular, a recent study of maternal serum samples indicated that high exposure to tra c-related air pollution during pregnancy disturbed metabolic pathways and mitochondrial function among mothers with children that developed ASD 34 . Moreover, short-and long-term species of PM 2.5 that can be traced to wood burning, including ultra-ne particulate matter (UFP), black carbon (BC), potassium, and copper, have been shown to negatively affect several metabolic pathways involved in oxidative stress and in ammation 35 , but this was not speci c to ASD development. Additionally, observational and controlled human exposure studies have found positive associations between Delta-C exposure, a marker for wood smoke, and in ammation 36 and systemic oxidative stress 37 , respectively. Others have not found associations, however 36,38,39 .
Although air pollution has been demonstrated to signi cantly affect neurodevelopment and contribute to ASD according to the latest systematic reviews and meta-analyses 23,24 , no study could be identi ed considering these associations with respect to source-speci c exposure. As the sources of locally produced air pollution may vary from setting to setting and each source can have unique effects on human health, previous research has encouraged more source-apportionment studies of PM 40 . Road tra c is often one of the largest contributors to PM emissions and is predicted to increase 24 , thus, exploring the various sources of tra c-related PM can provide a better understanding of their individual impacts on public health. Additionally, too few epidemiological studies investigating the association between ambient wood smoke from residential wood combustion and health exist to date 41 , and none consider ASD. Indeed, a review on the health impacts of wood burning emissions cited the need for more studies on non-respiratory pediatric outcomes, with better exposure assessment and proper adjustment for confounding 42 . The knowledge gleaned from source-speci c ndings can be used to inform air quality policy on all potentially detrimental emission sources.
The aim of the present study was to investigate source-speci c ambient PM 2.5 in association with autism spectrum disorders in a low-exposure setting where air pollution concentrations generally comply with the current European air quality guidelines 43 .

Study setting and study population
This study was undertaken in Scania (Skåne), the southernmost county in Sweden, which had a total Using each woman's unique personal identi cation number, PRS was linked to air pollution levels at each residential address as well as demographic information and socioeconomic characteristics obtained from Statistics Sweden. This data combined constitutes MAPSS. See Figure 2 for a detailed owchart of the study population for the present paper.

Exposure assessment
A Gaussian plume air dispersion model was developed to extrapolate the monthly mean air pollution concentrations in Scania, Sweden, during the study period 45  concentrations of all local PM 2.5 were available from a previous study 45 . The interpolation of the years and months between them was based on an atmospheric ventilation index using year-and monthspeci c meteorological parameters 45 .
Emission sources included aviation, industries, major energy and heat producers, railroads, road tra c, shipping, small-scale residential heating, non-road vehicles, and local emissions in Zealand, Denmark. Aviation emission data were obtained from Scandinavian airports' annual environmental reports. Industry and energy production emissions were reported by relevant entities within the study area. Railroads in Sweden are mainly electric; therefore, railroad emissions were estimated using the fuel consumption of the few operational diesel engine freight trains both while in transit on railway lines and while static at railway stations 44 . Detailed data on fuel sources, types of vehicles, and speed limits was gathered from the Swedish Road Administration to account for emissions from road tra c. Furthermore, the emission factors developed by the Handbook of Emission Factors for Road Transport (HBEFA), version 3.2 47 were applied to the vehicle types when estimating the emissions of PM 2.5 from the local tra c emission sources (road segments). Shipping emissions for the year 2000 were estimated by Gustafsson 46 , and 2011 emissions were described by Project Shipair 48 . Regarding small-scale residential heating, the frequency of for example stove-use was estimated using chimney sweeping records from the National Rescue Agency 46 . Emissions from non-road vehicles were derived from a report by the Swedish Environmental Research Institute, IVL. Scania's proximity to the industrial island Zealand and the area's prevailing westerly winds warrants the inclusion of their local emissions in the dispersion model 45 .
Temporal and meteorological variations of air pollution, were, furthermore, addressed through the atmospheric ventilation index using a complex method developed by the SCAC project 45 . The modeled source speci c annual means (for 2000 and 2011) were transformed to monthly means based on monthly pro les from the developed model (emission database). For all-source local PM 2.5 monthly means were already available. By applying the atmospheric ventilation index to the start and end values, monthly means were interpolated for each 100 m by 100 m grid cell. Those monthly mean levels approximately corresponded to the calendar months of the women's pregnancies.
The geographical coordinates of each woman's residential address were obtained from Statistics Sweden, linked to MAPSS, and used to calculate individual exposure. Changes in residency were only updated at the end of the calendar year. Exposure estimates for every gestational month were, therefore, based on the nearest available time: January-June coordinates from the end of the previous year and July-December coordinates from the end of the current year. Moreover, if 67% or more of monthly exposure data was non-missing, the pregnancy exposure overall was designated as non-missing.
Because this study examined locally emitted PM 2.5 , regional background levels (i.e., those generated elsewhere that have traveled into the study area) were not considered. As regional background emissions typically comprise the majority of total PM, investigating only local PM concentrations results in seemingly low levels. However, the spatial contrast for regional background levels in this study area is low 45 . Because of this, effect estimates would still describe contrasts in local exposures even if regional levels were included. The interpretation of the estimates becomes somewhat different though, illustrating the effect of local contrasts rather than the aggregate effect of total PM, which is more traditionally studied.
Four measures of locally emitted PM 2.5 were investigated: all-source PM 2.5 , tailpipe exhaust, vehicle wearand-tear, and small-scale residential heating. Wear-and-tear comprises the air pollution generated from vehicles' brakes and tires, for example, and small-scale residential heating mainly consists of wood burning. All other source contributions (aviation, industries, major energy and heat producers, railroads, shipping, non-road vehicles, etc.) were very small. Therefore, these were not examined separately but were incorporated into the all-source PM 2.5 category.

Outcome assessment
The outcome of interest in this study was autism spectrum disorders (ASD). When a child is suspected of having autism in Scania, they are referred to the Departments of Child and Adolescent Psychiatry and are examined by a multidisciplinary team 25 . These evaluations utilize both the Autism Diagnostic Observation Schedule-Generic (ADOS-G) 49 and the Autism Diagnostic Interview-Revised (ADI-R) 50 for most (75%) cases. Finally, an ASD diagnosis is assigned according to the International Classi cation of Mental and Behavioral Disorders version 10 (ICD-10) and entered into the Skåne Healthcare Database (SHR). The outcome data used in this study was extracted from SHR. ASD was identi ed by ICD-10 diagnosis codes starting with F84, which comprise all pervasive developmental disorders. This is characterized by one or more of the following areas of neurodivergence: qualitative variations in patterns of communication; di culties with reciprocal social interactions; and a restricted, repetitive collection of behaviors and interests. Children usually receive a diagnosis around 7 years of age 51 ; however, symptoms do not have to be present by a speci ed age.
A further restriction of autism cases to childhood autism (ICD-10 code F84.0), where symptoms within all three areas of neurodivergence must be present before the age of 3 years, was also applied.
Socioeconomic status (SES) predictors were also incorporated to control for potential confounding; these included maternal birth country (Sweden, Europe, other), maternal education level (pre-secondary, secondary, and post-secondary), annual household disposable income (quartiles), and neighborhoodlevel SES. Neighborhood-level SES was used a continuous measure, and captures the proportion of inhabitants in the neighborhood with "low economic standard", which is de ned by Statistics Sweden as the number of people living in a household with an economic standard that is less than 60% of the national median value 52 . Other covariates considered in sensitivity analyses include birth month, birth weight, low birth weight.

Statistical methods
The main analysis applied binary logistic regression models with ASD as the outcome in a univariate analysis and two multivariate analyses. The rst was a partially adjusted model which included only variables that were statistically signi cant in the models where total (local) PM 2.5 was investigated, namely, parity, pre-pregnancy body mass index, sex of the child, and birth year. In addition to these variables, the second multivariate model, hereafter referred to as the fully adjusted model, also included maternal age, smoking status at rst antenatal visit, maternal birth country, maternal education level, annual household disposable income, and neighborhood-level SES. As correlation coe cients between the different sources of PM 2.5 (Supplementary Table S1, Supplementary Information) were very high, a single pollutant model was used. Linear exposure trends were assessed for atmospheric particles, using a continuous increment of a 1 µg/m 3 increase in local PM 2.5 levels.
Several secondary analyses were conducted. A subgroup analysis restricted analyses to childhood autism (ICD-10 diagnosis code F84.0) only. A subpopulation analysis including only Swedish-born women was also performed, with ASD as the outcome. Finally, exposure during the rst year of life, instead of the pregnancy period, was explored for ASD. All utilized all-source PM 2.5 as the exposure.
To address uncertainties, three sensitivity analyses were conducted. One adjusted for birth month and excluded children with a low birth weight and children born to mothers born outside Sweden. Finally, the main analysis was rerun with imputed exposure data, using the Expectation-Maximization algorithm. These sensitivity analyses were conducted for all-source PM 2.5 with ASD as the outcome.
All statistical analyses were carried out using SPSS version 27. Odds ratios (OR) and their corresponding 95% con dence intervals (CI) were reported for all analyses.

Ethical approval
The Lund University Ethical Committee approved this study prior to its realization (permission number 2014/696 and amendment 2016/211).

Results
The relationship between the exposures, covariates and ASD are documented in Table 1. There were clear associations between ASD and various markers of socioeconomic status, smoking intensity, BMI and PM 2.5 exposure. Additionally, when excluding LBW babies, the OR was 1.27 (95% CI: 1.05-1.53). The sensitivity analysis using imputed exposure data resulted in higher precision in the estimates, but the overall effect estimates were only marginally altered (data not shown).

Discussion
In this population-based study from southern Sweden, associations were observed between exposure to all investigated local sources of PM 2.5 (all-source PM 2.5 , tailpipe exhaust, vehicle wear-and-tear, and small-scale residential heating) during pregnancy and ASD.
Although only statistically signi cant in the partially adjusted models and not in the fully adjusted models, the results suggest that both local PM 2.5 from tailpipe exhaust and vehicle wear-and-tear contribute to the observed associations with ASD. These results are in line with our previous study on prenatal exposure to ambient NO X concentrations and autism diagnosis using the same cohort (MAPSS), where children in the highest exposure quartile during the entire pregnancy period had a 40% greater risk of developing ASD compared to those in the lowest 25 . Under an assumption of a causal association between PM 2.5 and childhood autism, a health impact assessment conducted for Scania identi ed 3% of autism cases to be attributable to locally emitted PM 2.5 , of which ~30% is derived from tra c 53 . Outside our study setting, two case-control studies from California found 15% increased odds 54 and just over double the risk 55 of autism development due to tra c-related PM 2.5 exposure during pregnancy.
However, previous studies conducted in Stockholm, Sweden, did not nd associations between exposure to tra c-related air pollution during pregnancy and ASD 56,57 . Reasons for this con ict could include that these studies considered PM 10 and NO X , while ours investigated PM 2.5 , which has been the pollutant with the most evidence for autism development in systematic reviews 23,24 . One register-based study also used "symptoms of neurodevelopmental disorders" as its outcome as opposed to physician-diagnosed ASD 56 . Similarly, a study using four European cohorts, including a Swedish one, explored autistic traits, both symptom within the borderline/clinical range and within the clinical range using validated cut-offs, but did not nd an association with air pollution, even for PM 2.5 , using land-use regression models, with main predictor variables being tra c, space heating, and population/household density 58 . Interestingly, in neighboring Denmark, researchers found that exposure to tra c-related NO 2 , PM 10 , and PM 2.5 in early infancy, not during pregnancy, was associated with autism 59 . While our study's emphasis is on the pregnancy period, exposure during the rst year of life was also considered, and statistically signi cant associations with ASD and PM 2.5 were found. However, associations for exposure during fetal life appeared stronger in our data (due to high correlation we could not investigate both fetal life and rst year of life in the same statistical models). Systematic reviews have noted both the pregnancy period and postnatal periods as decisive exposure windows 24 .
To the best of our knowledge, no previous study has further separated PM 2.5 from road tra c into tailpipe exhaust versus vehicle wear-and-tear and explored their associations with ASD.
No previous epidemiological studies investigating the effects of source-speci c ambient wood smoke exposure from small-scale residential heating on autism development could be identi ed. Concerning neurodegenerative conditions in general, a longitudinal study in northern Sweden has indicated that PM 2.5 from residential wood burning is associated with dementia incidence 10 . In an experimental study using a placental rst trimester trophoblast cell line, exposure to wood smoke particles caused cytotoxicity and disrupted proliferation in exposed placenta cells, and particles detected inside cells caused structural damage to mitochondria and endoplasmic reticulum 60 . In line with this, sourceapportionment studies on pregnancy complications found increased concentrations of Delta-C (a marker for wood smoke) and BC during wintertime to be associated with greater odds of developing early-onset preeclampsia 61 . Similar studies exploring birth outcomes, however, reported that PM 2.5 from ambient biomass burning (i.e., wood smoke) was associated with a lower risk of preterm birth 16 , low birth weight 17 , and stillbirth 18 . Authors attributed these ndings to the high winter seasonality of biomass burning 16 ; it being inversely associated with most of the other PM 2.5 sources explored 17 ; and its negative correlation with re-suspended soil in particular 18 . In our study, however, no inverse correlations were seen between small-scale residential heating and the other PM 2.5 sources.
Due to its unique chemical composition, PM 2.5 derived from wood smoke may have varying toxicity compared to other sources of ambient PM 2.5 . Indeed, a 2003 review stated that studies including residential wood combustion as a major source of PM reported higher relative risks for adverse health outcomes compared to general ambient PM estimations 62 . Conversely, a 2018 review found that most source-apportionment studies reported PM from ambient biomass to be less detrimental to health than other sources; however, the wood smoke assessed was not necessarily derived from residential wood burning only 40 . A noted exception from Copenhagen observed stronger point estimates for PM 10 apportioned to biomass (mainly wood burning) than PM 10 derived from tra c 63 . Examining the impact of short-term exposure to PM 2.5 from tra c and wood smoke on mortality, a study also found higher statistically signi cant risk-increases when 24-hour average concentrations were used and stronger increased risks for deaths occurring in the cold season, both of which better represent exposure from wood burning 64 . In the present study, emphasis is placed on the positive associations observed for all PM 2.5 sources as opposed to comparing individual point estimates.
Despite the varying results in current literature, the continued inclusion of residential wood burning in source-speci c air pollution epidemiology is pertinent, as it has been shown to be a signi cant source of ambient PM, particularly in the wintertime. According to a recent review, residential wood combustion in developed countries may even dominate PM 2.5 and PM 10 concentrations during colder seasons, contribute to more than 40% of organic carbon attached to PM 2.5 and PM 10 , and in uence ambient elemental carbon (EC) concentrations, which is correlated with BC 65 . In Sweden, speci cally, recreational wood burning accounts for approximately 34% of small houses' heat demand 66 . In Scania, emissions from small-scale residential heating (mainly wood burning) have been calculated to comprise 12% of the total PM 10 , and 21% of the total soot in 2011; this source was even estimated to account for as much as 89% of the local PM 2.5 concentrations in certain locations 67 . Moreover, wood burning's relative contribution to total EC was more than three times higher at a rural site outside of Gothenburg, Sweden, than an urban one 68 . Interestingly, authors found that the spatial variability of wood burning aerosols between the two sites was moderate, and that high fossil fuel emissions at the urban site decreased wood smoke's relative contribution 68 . This demonstrates that small-scale residential heating is also prominent in urban areas.
No studies could be identi ed that apportioned PM 2.5 tra c emissions into its tailpipe exhaust versus vehicle wear-and-tear components and included autism spectrum disorder as an outcome, thus future ASD research should consider further source-speci c separation. Furthermore, health care systems in Sweden are tax-subsidized and used by virtually all residents, which increases the ability to identify physician-diagnosed cases of ASD and childhood autism (ICD-10 codes F84 and F84.0, respectively) and record them in high quality healthcare databases. With this, outcome misclassi cation, response-bias, recall-bias, and selection bias were avoided. Information on covariates incorporated into the adjusted models were similarly collected from well-managed, precise registers.
These results are also considered generalizable to pregnant women in other similar study areas, where the sources of air pollution are comparable. The ndings are relevant to both public health in general and the clinical setting speci cally because they indicate that even small changes in locally produced PM 2.5 concentrations can affect the risk of ASD among children. Finally, this study contributes evidence to an emerging research area investigating the health effects of local, source-speci c air pollution exposure. Interestingly, accumulating evidence suggests that locally produced PM may be more hazardous to human health than regional, background concentrations 80 .
This study also has several limitations. For example, quite a large proportion of the study population had missing data on exposure, outcome, and/or covariates. Missing observations were able to be imputed in a sensitivity analysis, which resulted in increased precision of the corresponding estimates, as illustrated by narrower CIs, and relatively unchanged point estimates. Moreover, data on parental diagnoses was not available. Given that genetic factors account for a considerable part of the variation in autism development and emergence 81 , our results could partly be explained by heredity, if parents with autism were more likely to reside in areas characterized by higher levels of air pollution than parents without autism. Residual confounding due to other risk factors for autism, which are also associated with the exposure and not accounted for in our statistical models, may also be present. However, relevant risk factors have been included, which was based on current literature as well as a directed acyclic graph. Exposure misclassi cation may exist, as exposure was assessed at each woman's home residence, and participants' total exposure, including indoor, behaviour-related, transport-related and occupational, was not considered. This is deemed standard practice in air pollution epidemiology research, with the assumption that the resulting misclassi cation is non-differential. Results pertaining to the sourcespeci c fractions of PM 2.5 derived from tailpipe exhaust produced high point estimates and notably wide con dence intervals. The levels of locally produced PM 2.5 in the study area are quite low, which makes the 1 µg/m 3 increase in PM 2.5 a sizeable increment and can contribute to high point estimates; nevertheless, the study results emphasize that despite being low, the local contribution to PM 2.5 may still have detrimental health effects. As the differences in ORs between the various sources were not formally tested and some sources are highly correlated with one another, direct comparisons between our sourcespeci c risk estimates cannot be made. Instead, more studies, preferably in a multi-cohort setting, are needed to increase statistical power. When considering only mothers born in Sweden, the associations tended to be lower than for that of the entire study population. Findings from previous research on environmental injustice in Scania, showed that non-Swedish-born persons had higher odds of being exposed to greater concentrations of air pollution 82 . In Sweden, children born to women who emigrated from Sub-Saharan Africa and the Middle East are, furthermore, more commonly diagnosed with ASD 51 .
Our results may also suggest that there is some residual confounding with respect to SES in the statistical models.

Conclusion
These ndings add to current evidence that prenatal exposure to air pollution is associated with the risk of developing autism spectrum disorders and offers insight on these associations in a relatively low exposure setting. Additionally, this source-speci c study indicates that each local PM 2.5 source investigated (all-source, tailpipe exhaust, vehicle wear-and-tear, and small-scale residential heatingmainly wood burning) may be associated with the increased risk of ASD. This supports existing literature documenting the substantial health effects of locally produced particles, despite long-range, intransported contributions often comprising the majority of PM's total concentration.

Declarations
Data Availability: The datasets generated during and/or analyzed during the current study are stored on a secure server and are not publicly available because they contain sensitive information (on health data, demographic characteristics, socioeconomic status) and, therefore, cannot be shared openly. However, they are available from Anna Oudin (anna.oudin@med.lu.se) on reasonable request.