The BEAR study is a prospective cohort study on children, using a natural experiment of the relocation of air traffic in the city of Berlin.

Prior to October 2020, Berlin had two operating airports, TLX and Berlin Schönefeld Airport (SFX) with about 2 million (TXL) and 1 million (SFX) passengers per month. A third airport was under construction directly south of SFX (Berlin-Brandenburg Airport, BER) (Supplementary Figure S1). SFX operated as an independent airport until the end of October 2020. The airport’s check-in building was integrated into BER as an additional terminal, but has not been used since February 2021. The relocation of air traffic from TLX to BER in November 2020 implied a total loss of air traffic at TLX and thus a drastic reduction of airport-related emissions for children living near TLX. Simultaneously, through the commissioning of BER, a sudden and strong increase of air traffic and of airport-related emissions at SFX and BER, was expected, subjecting the children living near SFX/BER to substantial increases in aircraft-related exposures. Children living in areas away from all airports [control areas (CA)] are assumed to be not affected by the air traffic relocation.

We recruited children attending elementary schools and their affiliated after-school centers in the city of Berlin and in the federal state of Brandenburg: near TXL, near BER, and from CA in Berlin (Supplementary Figure S1). Parallel to the repeated health examinations of the children, we measure the outdoor UFP concentration at the participating schools. AP is measured continuously at several measuring stations in Berlin and Brandenburg near TLX, SFX and BER as well as in CA. Modelled source-specific UFP and other air-pollutants at the schools and home addresses of participating children will be available for long-term exposure pre- and post-relocation of Berlin AC traffic (Figure 1).

Health examinations started in January 2020. Follow-up examinations are ongoing and are expected to end in May 2024. The BEAR study has been approved by the local ethics committee at Heinrich-Heine-University of Düsseldorf (protocol number: 2019-695). For all children, a parent or legal guardian provided informed consent. From a list of all primary schools in Berlin, we prioritized eligible schools according to location (Supplementary Figure S1), comparable SES of the neighborhoods, local road traffic exposure and exposure to industrial sources of AP and noise.

For prioritization, we used the sociodemographic atlas of Berlin [30] and the existing noise and road traffic maps of Berlin [31]. Initially, we planned to examine altogether 550 schoolchildren aged 7–12 years.

Health examinations started in January 2020 in Berlin and December 2020 in Brandenburg, with baseline examinations completed by December 2022. Parallel to baseline examinations, the follow-up examinations started in August 2021 and are still ongoing. These examinations were integrated into the regular school day in Berlin and after-school day-care in Brandenburg, accompanied by an educational module on healthy living performed by the study staff. To ensure consistency, study staff underwent regular training, certification, recertification, and process monitoring. All tests adhered to standard operating procedures in line with international and national standards. Parents who consented to participate completed a comprehensive questionnaire covering their child’s health, COVID-19 history, parental demographics, and family lifestyle factors (in German Supplementary Baseline Questionnaire parents).

At each examination, acute physical illnesses and medication use were assessed with an examiner-conducted questionnaire (in German Supplementary Acute Health Questionnaire).

Quality of life (QoL) was evaluated using the KIND-L (“Revidierter Fragebogen für KINDer und Jugendliche zur Erfassung der gesundheitsbezogenen Lebensqualität”) questionnaire for children through interviews (in German Supplementary KINDL Questionnaire).

We measured height and weight without shoes and in light clothing with a fixed ruler and a digital scale, respectively. For blood pressure measurements, children were placed in an upright sitting position with their arms placed on the table in front of them. The blood pressure cuff was placed on the arm used for writing. Care was taken to ensure that clothing did not constrict the arm. After resting for 5 minutes, systolic and diastolic blood pressure measurements were taken in triplicate with 2 minutes rest in between measurements using a digital oscillometric device (ProBP 2000, Welch Allyn).

Exhaled nitric oxide (FeNO) tests (NIOX VERO, Circassia) were conducted, with children attempting the 10 s version first and switching to the 6 s version if necessary. High FeNO values (>35 ppb) were confirmed in a second test for those children. FeNO tests were conducted before spirometry in order to avoid biasing nitric oxide values towards the null.

Spirometry tests were performed in accordance with American Thorax Society (ATS) and European Respiratory Society (ERS) guidelines [32] using the EasyOne spirometer (NDD Medical Technologies). Children underwent an initial eligibility screening with exclusion criteria such as current acute infections, recent chest or torso surgery within the past 3 months, recent eye surgery within the past month, or the use of tuberculosis medication. Children with asthma taking medication were allowed to participate if their last use was sufficiently long before the lung function test, ranging from up to 4 h (short-acting ß2 mimetics like salbutamol) to 5 days (leukotriene antagonist like montelukast).

During the lung function test, children were seated upright with their feet on the ground, maintaining a proper distance from the table with the computer screen. They watched an instructional video demonstrating full spirometry with forced maximal exhalation and received verbal instructions for the maneuver. They were required to perform two consecutive maneuvers, with only the second one analyzed. Valid maneuvers met the following criteria: back extrapolated volume (BEV) <0.15 L, time to peak flow (PEFT) <0.12 s, and either forced expiratory time (FET) >3 s or end of test volume (EOTV) <0.04 L. Valid maneuvers were assessed for reproducibility, and the lung function test ended after obtaining three valid and reproducible maneuvers or reaching a maximum of eight maneuvers. If fewer than two valid and reproducible maneuvers were achieved, a repeat examination was scheduled.

To assess cognitive function, children completed the standardized N-back and Attentional Network Task (ANT) Test. The N-back test is a well-established working memory assessment where participants determine if the current color matches the one shown n colors back [33–35]. It was conducted for 1-, 2-, and 3-back conditions, generating scores based on correct and incorrect answers. The ANT test evaluates various forms of attention (alerting, orienting, and conflict) by presenting an image of five fishes in a line [36]. Children click the mouse button corresponding to the middle fish’s direction, with reaction times and accuracy recorded and calculated into scores [36]. Both tests were programmed in e-prime [37] and administered at separate tables with headphones and privacy screens to ensure a quiet environment. Children received instructions and a test run to ensure task comprehension.

We plan to examine each child at least three times until spring 2024. Should children move schools or no longer choose to participate, we will recruit additional participants within the schools already participating in the study.

Air pollutant concentrations near Berlin airports and in the CA have been assessed at several monitoring sites mostly since 2019. A detailed description of the measurement stations is given in Table 1 and the location in the Supplementary Figure S1. All data collected at the measurement stations will be made available for the BEAR project.

The monitoring of air quality began in July 2020 at a fixed-site monitoring station in the downwind area of the old airport TXL in order to assess the change in pollutant concentrations in the direct vicinity of the TXL school sites Berlin-Reinickendorf (RKD). In June 2021, this fixed monitoring station was moved to Blankenfelde-Mahlow, near the new airport BER. This measurement station was operated by the Leibniz Institute for Tropospheric Research (TROPOS) by October 2022 in the framework of the research project “Ultrafeinstaubbelastung durch Flughäfen in Berlin” (ULTRAFLEB), a BEAR-associated 3 years research project funded through the German Federal Environment Agency. Near the new airport BER, two additional fixed measurement stations including highly time-resolved PNC measurements (GRIMM 5420) were already present. The airport company “Flughafen Berlin Brandenburg” (FBB) operates both monitoring stations. The first one is located directly adjacent to the airfield (site SFX) and a second one in Berlin-Bohnsdorf (BDF) about 5 km downwind of SFX/BER since October 2016.

In order to estimate UFP concentrations at the recruited schools, a condensation particle counter (CPC; EDM465, Grimm Aerosol Technik, model EDM465) was placed to school sites concurrently with the baseline health examinations. This device measures total PNC in the size range from 7 nm to 1 µm utilizing pre-impactor. This measurement campaign started in January 2020 and is still ongoing.

To distinguish between UFP emissions from aircraft-related sources and other PNC sources like road traffic and wood combustion, we employ various strategies. Initially, we utilize backward air mass trajectories to assess airport influence at school locations. By defining sectors where the airport is downwind, we can estimate airport-related particle signals in the CPC measurements, accounting for nearby roads and highways. Further analysis relies on particle size distribution (PDS) data and additional air pollutant measurements gathered from monitoring stations in the vicinity of the airports (Table 1). We conduct correlation analyses between PNC at schools and monitoring sites. After determining how those stations represent the particle number concentrations at schools (work ongoing in separate publication), long-term high quality data will be used to run source apportionment by use of positive matrix factorization (PMF) method (following methodology by, e.g., Tremper et al., 2022) [38], segregating between different pollution sources, including airport and traffic. Finally, the results of the source apportionment will be transferred to the school locations.

It’s worth noting that our study quantifies airport activity-induced concentration increases in total PNC, which we convert to UFP PNC based on PDS measurements, Although several studies do refer to total PNC as UFP [12, 39–41]. To validate our approach, we examine particles with a mobility diameter of 30 nm, indicative of airport-related activities [38].

Additionally, ULTRAFLEB provides BEAR with modeled source-specific UFP and air pollutant data for long-term exposure assessments pre- and post-relocation of Berlin AC traffic (see Supplementary Figure S1). This modeling covers Berlin and Brandenburg, including residential addresses of participating children and schools, at a spatial resolution of 50 m*50 m. We generate exposure estimates for PM₁₀, PM_2.5, NO₂, black carbon, and UFP from 2019 to 2023, validating them against school site, monitoring station, and mobile campaign data.

We plan to analyze short-, medium- and long-term effects of total and source-specific UFP. Short-term periods refer to hourly and daily, as medium- and long term periods refer to monthly and yearly effects of air pollution measured and modelled over the appropriate time at children`s schools and residential addresses.

To identify the appropriate confounders for each analysis, we will employ a directed acyclic graph (DAG) approach. This method allows us to identify the minimal sufficient adjustment sets needed to establish the association between UFP exposure and the outcome parameters. Possible confounders may include socio-economic status (SES) of the area or parents, history of chronic diseases (e.g., asthma), history of attention deficit disorder/attention deficit hyperactivity disorder, season, and other relevant factors.

We will initially assess short-term changes in various health factors, including respiratory symptoms, lung function, cognitive function, blood pressure, and quality of life. We’ll use generalized linear mixed models, considering exposure to air traffic, location (TXL, BER, CA), and examination time as variables. This analysis will involve all collected data across all groups. For source-specific (AC-UFP) exposure, we will include other airport-related factors like road traffic noise, air pollution (AP), and changes in socioeconomic-status (SES) in sub-analyses. We will also explore different exposure time windows based on existing literature. To ensure accuracy, we will adjust our models for factors such as age, gender, weather conditions, background air pollution, NDVI (normalized difference vegetation index), and relevant covariates that could influence both exposure and outcomes (confounding variables).

The primary outcomes of long-term health effects are lung growth and cognitive development. We will consider background air pollution (AP) exposure through a designated control region. The central hypothesis of the BEAR Study is that reducing AC-UFP and other airport-related exposures near TXL will improve lung and cognitive function, while an increase in these exposures near SFX/BER will reduce lung and cognitive function. To thoroughly investigate these effects, we will analyze changes in pollutant levels downwind of the airports and in CA, changes in exposure for schoolchildren near TXL, SFX/BER, and in CA, and changes in the children’s health outcomes over time.

Our analysis will involve linear mixed-effects models, assessing lung capacity (FVC and FEV₁) from up to three visits, as well as working memory and attention. We will include a random participant intercept for each child and account for temporal trends in exposure and outcomes over the 4 years study period, using the examination date as a variable. We will also consider seasonal, meteorological, and other variables that could predict the outcomes. Similar to the short-term analysis, the models will be adjusted for age, gender, weather conditions, background air pollution, NDVI, and potential confounding factors, which may vary depending on the specific outcome under study. We will particularly address noise-related confounding by incorporating changes in exposure to different noise sources based on modeled noise levels at residential addresses.

In addition to these analyses we will conduct a sub-analysis focusing solely on individuals exposed to airport-related UFP to assess the source-specific short-term effects more precisely. This sub-analysis will allow us to gain deeper insights into the specific impacts of airport-related UFP exposure on the outcomes under study. In contrast, for the long-term analysis, we will utilize a binary exposure variable that distinguishes between two groups: those living near the airport and the control group. This approach will allow us to examine and compare the differences in lung growth and cognitive development over time between these two distinct groups.

Potential mediators of the total effect of the intervention on health outcomes in our analysis are changes in AC-related UFP or other AC-related air pollutants, road traffic-related UFP or other road traffic-related air pollutants, AC-noise, road traffic noise, and area-level SES. Outcomes will be short-term changes in lung function, lung inflammation, and cardiovascular health, as well as medium- and long-term changes in lung growth and cognitive development. We will conduct mediation analyses according to the principles and methods as outlined by VanderWeele [42]. We will also conduct sensitivity analyses evaluating whether the exposure and mediators may interact in their effect on the outcome.

We calculated the detectable effect size for our study based on dynamic lung function volumes (FEV₁, FVC) as the outcome. We will assess 1,070 children, resulting in a total of 3,210 observations (three observations per child). With an average FEV₁ of 1.6 L (SD 0.27 L) in elementary school children in Germany [43], a 5% alpha-level, and a sample size of 503 children, we have at least 80% power to detect a 0.030 L reduction in FEV₁. This change is similar to what’s observed after a 2 h exposure of participants with mild to moderate asthma per 10,000 particles/cm³ [44] and smaller than daily changes in ambient PM AP on lung function in children [45, 46]. With a sample size of 856 children, we can detect a 0.015 L reduction in FEV₁ (alpha-level of 5% and 80% power), allowing for a 20% loss of participants.

For the analysis of short-term effects on FeNO, we estimate that with a mean FeNO level of 11 ppb (SD 1.1 ppb), we can detect a moderate effect (0.5 ppb decrease) with 232 measurements, using a 5% alpha level and 90% power, with an allocation ratio of 2:1.

Regarding the analysis of lung growth trajectories and cognitive development, we can detect small to moderate differences with an alpha-level of 5% and 80% power in a sample of 282 children.

As in many studies, the appearance of COVID-19 in late 2019 and the subsequent global pandemic have significantly impacted several aspects of this study. The natural experiment involving an abrupt closing of TXL Airport and a quick takeover of all air traffic by BER did not occur due to the large-scale cessation of air traffic starting in March 2020 at TLX and SFX. In summer 2020 reduced air traffic resumed TLX, and in November 2020 air traffic relocated from TLX to BER but still with a reduced volume (Figure 2).

With the gradual increase in air traffic at BER we conducted more ‘low exposure’ measurements in children attending schools in that area and had fewer measurements available under high exposure conditions. Additionally, because of intermittent school closures due to high local COVID-19 incidence rates (Figure 3), recruitment and examination of children were more difficult than expected. As a contingency measure, we decided to include additional schools near BER Airport in Brandenburg and to extend our observation period to at least 4 years. With the addition of new study areas in Brandenburg, we increased the number of participants to more than 1,000 schoolchildren (Table 2).