Flooding exposure accelerated biological aging: a population-based study in the UK

Floods have been the most common type of disaster and are expected to increase in frequency and intensity due to climate change. Although there is growing evidence on the impacts of floods on human health, none has so far investigated the association between flooding exposure and biological aging acceleration. We collected data from 364 841 participants from the UK Biobank project. Flooding data before baseline were retrieved from the Dartmouth Flood Observatory and linked to each participant. Cumulative flooding exposure within six years before the baseline was calculated. We calculated the two biological aging measures at baseline: PhenoAge and Klemera-Doubal method biological age (KDM-BA) and assessed their associations with flooding exposure using mixed-effects linear regression models. We observed that participants exposed to higher levels of floods were more likely to have accelerated biological aging. The risks associated with flooding exposure could last for several years, with the highest cumulative effect observed over 0–4 years. In the fully adjusted model, per interquartile increase in cumulative flood exposure was associated with an increase of 0.24 years (95% CI: 0.14, 0.34) in PhenoAge acceleration and 0.14 years (95% CI: 0.07, 0.21) in KDM-BA acceleration over lag 0–4 years. The associations were consistent regardless of lifestyles, demographics, and socio-economic status. Our findings suggest that exposure to floods may lead to accelerated biological aging. Our work provides the basis for further understanding of the flood-related health impacts and suggests that public health policies and adaptation measures should be initiated in the short-, medium- and even long-term after flooding.


Introduction
Floods are one of the most widespread natural disasters globally.According to the World Meteorological Organization Atlas of Mortality and Economic Losses from Weather, Climate, andWater Extremes (1970-2019), 44% of recorded natural disasters have been associated with floods.Approximately 1.81 billion people are directly exposed to flood depths > 0.15 meters in a 1-in-100 year flood (i.e. a 100 year recurrence interval) (Rentschler et al 2022).
Floods present a multifaceted array of direct and indirect health hazards (Du et al 2010).Drowning, injuries, hypothermia, and zoonotic infections are the immediate threats, posing a significant risk to life (Turgut and Tevfik 2012, Doocy et al 2013, Paterson et al 2018, Oshiro et al 2022).Evacuation and displacement, loss of healthcare personnel, and damage to health infrastructure, including essential medications and supplies, further compound the crisis, disrupting access to healthcare or treatment (Loehn et al 2011, Mckinney et al 2011, Rath et al 2011, Buajaroen 2013, Ryan et al 2015, Man et al 2018).In the aftermath, flood survivors may experience infected wounds, complications from injuries, carbon monoxide poisoning from incorrect use of power generators, mental health disorders, outbreaks of communicable diseases, and even starvation (Platz et al 2007, Saulnier et al 2017, Paterson et al 2018, Lieber et al 2022).The long-term health impacts of floods extend even further, with potential consequences including persistent mental health challenges, exacerbation of chronic illnesses, reduction in health-related quality of life, and premature death (Fernandez et al 2015, Zhong et al 2018, Lieber et al 2022, Wu et al 2024).During these processes, the psychological stress begins immediately following the flood event and can persist for extended periods, leading to deteriorating physical and social functioning (Alderman et al 2012, Morganstein andUrsano 2020).
Aging is a multifaceted biological process that progressively undermines the integrity and resilience capacity of cells, tissues, and organs (Gao et al 2023).While aging is an inevitable process, it does not progress uniformly across individuals (Martens and Nawrot 2016).It has long been recognized that individuals of the same chronological age can exhibit considerable variation in their rate of aging, which likely reflects disparities in their underlying biological aging processes, measured as biological age (also called physiological age) (Levine et al 2018, Wu et al 2021).Biological age incorporates information from biological markers and may serve as a more accurate reflection of an individual's physiological state.Moreover, it demonstrates robust predictive capabilities regarding the risks of all-cause and cause-specific mortality, physical functioning, cognitive performance, and even facial aging (Levine et al 2018).
Previous evidence suggests external factors can influence the process of aging, among which, psychosocial stress could act as an accelerator of biological aging (Polsky et al 2022, Watowich et al 2022).This process involves the promotion of oxidative stress, DNA damage, inflammation, telomere shortening, mitochondrial dysfunction, and ultimately, cellular senescence (Polsky et al 2022, Watowich et al 2022).Therefore, a hypothesis might be derived that psychological stress arising from experiencing flooding disasters may accelerate biological aging, potentially contributing to the observed associations between floods and long-term health consequences (Reacher et al 2004, Gautam et al 2009, Jiao et al 2012, Zhong et al 2018, Fitzgerald et al 2019).However, to our knowledge, no study has investigated the longterm effect of floods on biological aging.To fill this gap, we conducted a investigation utilizing data from the UK Biobank project, a population-based cohort study.Our study aimed to examine the association between long-term flooding exposure and biological aging and to further assess the modifying effects of lifestyle, demographics, and socioeconomic status on this association.

Study design and participants
In this population-based study, we used crosssectional baseline data from the UK Biobank cohort which recruited about 0.5 million residents aged between 37 and 73 years from 2006 to 2010 across the UK.The study protocol has been described elsewhere (Biobank 2020).Briefly, 9.8 million UK residents registered with the UK's National Health Service (NHS) were sent postal invitations to attend one of 22 assessment centres within a reasonable travelling distance of their home addresses.Upon arrival at the assessment centres, participants were asked for consent, and underwent a series of baseline assessments involving questionnaires collecting sociodemographic, lifestyle, and health-related information, as well as physical measurements, and blood/urine sampling.Participant data and samples were transferred to the UK Biobank coordinating centre.All participants provided consent for their health to be followed-up through linkage to healthrelated records.
Among the 502 414 adult participants with available data, we excluded participants lacking data on longitude and latitude of residence, recruitment date, with any missing data of biomarkers required to calculate biological age.A total of 364 841 participants were included in the final analysis.The baseline characteristics of participants who were excluded from this study are shown in table S1.The included participants exhibit higher proportions of males (46.5% versus 43.1%) and individuals of White European ancestry (94.9% versus 93.7%), coupled with a lower proportion of individuals living in highly deprived areas (49.5% versus 51.3%).The flooding exposure was slightly higher among the included participants than those excluded (table S1).No substantial differences were found for other socio-demographic characteristics.UK Biobank has ethical approval from the North West Multi-centre Research Ethics Committee (reference 16/NW/0274).All participants gave written informed consent for data collection, analysis, record linkage, and publication of research.The generation and use of the data presented in this paper were approved by the UK Biobank access committee under UK Biobank application number 55257.

Assessment of biological age and age acceleration
We quantified biological age based on two composite measures of blood-chemistry-derived markers and other clinical data collected at the baseline: Klemera-Doubal method biological age (KDM-BA) (Klemera and Doubal 2006)  Both measures were computed using the R package BioAge, with NHANES III as the reference population.Briefly, the KDM-BA algorithm included forced expiratory volume in one second, systolic blood pressure, and seven blood chemistry parameters (albumin, alkaline phosphatase, blood urea nitrogen, creatinine, C-reactive protein, glycated haemoglobin, and total cholesterol).As recommended, we used a reference population consisting of nonpregnant individuals aged 30-75 years, with complete biomarker data, for the calculation of KDM-BA (Kwon andBelsky 2021, Mak et al 2023).The PhenoAge algorithm incorporated various biomarkers including albumin, alkaline phosphatase, creatinine, C-reactive protein, glucose, mean cell volume, red cell distribution width, white blood cell count, and lymphocyte proportion.The reference population for PhenoAge included individuals aged 20-84 years with complete biomarker data.To address any skewness in the distribution of biomarker values, we adjusted the top and bottom 1% of all biomarkers by setting them to the 99th and 1st percentiles, respectively.
To facilitate comparisons between participants in biological aging, we conducted a linear regression model where we regressed the computed biological age measures onto the corresponding chronological ages.The resulting residual values, which we refer to as PhenoAge acceleration and KDM-BA acceleration, were used to quantify biological aging.As such, biological age acceleration represents the extent to which an individual's biological age diverges from the chronological age.A positive value, such as 2, indicates that an individual is biologically older by 2 years compared to what is anticipated (i.e.faster aging), whereas a negative value suggests that the individual is biologically younger than anticipated (i.e.slower aging) (Kuo et al 2020).

Flooding exposure
Flooding exposure was obtained from a flood database for the period 1985-2020 provided by the Dartmouth Flood Observatory (DFO) (Brakenridge 2016), which is a global catalogue of all flood events that have been documented by the news, government, and the FloodList (http://floodlist.com/).The DFO provides detailed information on each flood event, including start date, end date, centroids, impacted geographic areas, and severities.The information has been validated by news or official reports from governmental sources (Carozza and Boudreault 2021).Participants whose home addresses fall within floodaffected areas were considered as having been exposed to a flood event.We calculated yearly cumulative exposure to floods for each participant by multiplying the duration and severity of each flood event and summing these values for each year (White et al 1998, Steenland et al 2011), with equation ( 1): where Flood index i,year=m stands for the cumulative flood exposure in year m for participants i. Duration ij represents the duration of the jth flood event in year m, defined as the time interval in days between its onset and cessation.Severity ij represents the level of severity of the jth flood event in year m.The severity of each flood event documented in the DFO was categorized according to a pre-defined scale.Specific numerical values were designated for each severity classification.For example, a value of 1 corresponds to large flood events with a return period of 1-2 decades, and a value of 2 denotes extreme flood events with a return period exceeding 100 years.Detailed information on the severity of flood events is shown in table S2.If there were no flood events within a given year, a flood index of 0 was recorded.Given our focus on flood exposure preceding the date of blood draw, the flood index in the year of blood draw was denoted as lag 0, and the flood index in preceding years was designated as lag 1, lag 2, lag 3, and lag 6 for exposures occurring one, two, three, and six years before the date of blood draw, respectively.Additionally, we calculated the average value of the annual flood index to reflect the long-term level of flooding exposure.

Meteorological data
We extracted hourly temperature data from the European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ERA-5) reanalysis data set with a spatial resolution of 0.1 • × 0.1 • (ECMWF).We mapped meteorological data to the participant's geocoded residential address at baseline.Daily meteorological data were calculated by averaging hourly data within each day.Daily relative humidity was calculated from the daily mean temperature and daily mean dew point temperature.Daily temperature and relative humidity were then aggregated into yearly averages.We included both temperature and relative humidity in the model due to their established association with both floods and adverse health outcomes

Covariates
Covariates adjusted for in the model were: age, sex, body mass index (BMI), alcohol consumption status, cigarette smoking status, physical activity, healthy diet score, ethnicity, education attainment, annual household income, and area deprivation level.These covariates were selected because they were potential risk factors of biological age and might confound the association between floods and biological aging (Horvath et ], or none of the above).Smoking status was categorized as current, former, and never.Low-risk alcohol consumption was defined as moderate drinking (no more than one drink/day for women and two drinks/day for men; one drink is measured as 8 g ethanol in the UK) on a relatively regular frequency (Anstey et al 2009, Chen et al 2023).Annual household income was classified into five groups (<£18 000, £18 000-£30 999, £31 000-£51 999, £52 000-£100 000, and >£100 000).The diet score was computed by assigning one point to each of the following favourable dietary factors: vegetable intake ⩾ 3 servings/day; fruit intake ⩾ 3 servings/day; whole grains ⩾ 3 servings/day; refined grains ⩽ 1.5 servings/day; fish intake ⩾ 2 servings/day; unprocessed red meat intake ⩽ 2 servings/week; and processed meat intake ⩽ 2 servings/week.A healthy diet was defined as having a diet score of 4 or higher.BMI was calculated from objectively measured weight and height as weight over height squared.Townsend deprivation index (TDI) was used to evaluate the level of deprivation in each area.Participants were classified as high or low deprivation depending on whether their TDI score exceeded the median value.

Statistical analysis
We examined the association between flooding exposure and each of the biological age measures (PhenoAge acceleration and KDM-BA acceleration) using a mixed-effects linear regression model with a random intercept to control the effect of participants nested within regions (England, Scotland, and Wales).We modelled the association with flood using a distributed lag non-linear model, a flexible methodological framework widely used to investigate the health effects of environmental exposures (Gasparrini et al 2010).This model defines a cross-basis function, featuring a non-linear exposure-response association and an additional lag-response association.For the exposure dimension, we pre-specified a natural cubic spline with one internal knot placed at the 50th percentile of the flood index distribution.As the nonlinear analysis suggested an approximately linear relation (figure S1), we applied a linear exposure-response relationship in the formal analysis.The lag-response curve was modelled with a natural spline function with two internal knots at equally spaced values in the log scale over 6 years of lag for the lag-response dimension.The choice of a maximum lag period of 6 years was informed by a literature review, where the majority of studies reported health impacts within two years following floods, with one study indicating that the health impacts of floods could persist for up to six years (Jermacane et al 2018, Zhong et al 2018, Fitzgerald et al 2019).
A stepwise approach was used to adjust for covariates in the model.Models were estimated unadjusted (model 1); adjusted for age, sex, and ethnicity (model 2); additionally adjusted for education attainment, annual household income, and area deprivation level (model 3); additionally adjusted for potential mediating biomarkers and lifestyles, including BMI, alcohol consumption status, cigarette smoking status, physical activity, and healthy diet score (model 4); and additionally adjusted for average daily mean temperature and relative humidity in the year of recruitment (model 5).Temperature and relative humidity were modelled through a natural cubic spline with three degrees of freedom.Results are presented as the change in biological age acceleration (years) and their 95% confidence intervals (95% CIs) per interquartile range (IQR) increase in the flood index.We used the IQR measure because of its capacity to capture a more comprehensive range of exposure variability compared to per unit increases.
We also performed analysis stratified by age group, sex, ethnicity, weight status, education attainment, annual household income (⩽£31 000 and >£31 000), alcohol consumption status, drinking status, healthy diet status, and area deprivation level.Weight status was defined based on BMI as underweight (below 18.5 kg m −2 ), normal weight (18.5-24.9kg m −2 ), overweight (25.0-29.9kg m −2 ), and obese (above 30 kg m −2 ).We excluded underweight as less than 1% of participants were classified as underweight.We tested the statistical significance of the difference in the effect estimates across subgroups using a random-effects meta-regression model.
We conducted several sensitivity analyses to test the robustness of our estimates.We used alternative degrees of freedom for temperature and relative humidity.As there are missing values in some covariates, models 2-5, and subgroup analyses were performed only among participants with complete data with the sample size decided by covariates included in these models.Therefore, we performed sensitivity analyses by using complete data after multivariate imputation by chained equations.In addition, we performed sensitivity analyses among participants who have been living in the current address for at least 5 years.R (version 3.6.2) was used for all analyses.Twosided P-values < 0.05 were considered statistically significant.

Descriptive statistics
Descriptive statistics across regions at baseline are presented in table 1.Among 364 841 participants, 91.4% (333 350) of participants were from England.46.5% (169 758) were male and 94.9% (344 738) selfidentified as White.The mean chronological age at baseline was 56.6 years (SD: 8.1), 11 years older than the mean PhenoAge (45.7,SD: 9.9).The mean PhenoAge acceleration was 0.0 (SD: 5.2), similar to the mean KDM-BA acceleration (0.0, SD: 3.6).Baseline characteristics of participants with any missing values in covariates are shown in table S3.The average value of the annual flood index over the six years prior to the date of blood draw was 6.5 (SD: 5.8), with a range from 0.6 in the lag 5 to 19.0 in the lag 0. The detailed descriptive statistics of the annual flood index are shown in table S4.We observed small variations in meteorological variables among participants, with an annual mean temperature of 9.3 • C and an annual mean relative humidity of 80.6% (table S4).The geographical distributions of PhenoAge, KDM-BA, and annual average flood index during the study period are displayed in figure 1. Flood levels were higher in Birmingham, Sheffield, and Nottingham.In contrast, biological age levels are more evenly distributed across the UK compared to floods.

Regression results
Figure 2 shows the associations between flood index and biological age acceleration measures on different lag years.The magnitude of associations increased from the current year (lag 0) to the lag year 3 and then decreased on the lag year 5 for both PhenoAge and KDM-BA.As a result, the associations reached the highest over lag 0-4 years, which was used as the main lag value for the flood index in subsequent analysis.
Figure 3 shows the estimates for the associations of flood index with age acceleration measures.Results of the crude model suggested that per IQR increase in flood index over lag 0-4 years was associated with an increase of 0.28 years (95% CI: 0.21, 0.36) in PhenoAge and 0.15 years (95% CI: 0.15, 0.20) in KDM-BA.The magnitude of associations decreased slightly after adjustment for demographics and socioeconomic status.After additional adjustment for temperature and relative humidity, the corresponding increases were 0.24 years (95% CI: 0.14, 0.34) in PhenoAge acceleration and 0.14 years (95% CI: 0.07, 0.21) in KDM-BA acceleration.The exact values are shown in table S5.
The associations of flood index with biological age measures were observed for most subgroups.For PhenoAge acceleration, the associations were marginally more pronounced among individuals with lower household income, higher educational attainment, moderate alcohol consumption, and residing in areas characterized by lower deprivation levels, in comparison to those with higher household income, lower education levels, non-moderate alcohol consumption, and residing in areas with higher deprivation levels, respectively (figure 4).For KDM-BA, effect sizes were larger for older as compared with younger participants (figure 4).
In the sensitivity analyses, the estimates for the associations of floods with biological aging acceleration were robust when using multiple imputed data (table S6).Compared with the main models, the results were generally similar when only participants living in the current address for at least five years were included (table S6).The results remained similar after using alternative degrees of freedom for both temperature and relative humidity (figure S2).

Discussion
To our knowledge, this study is the first epidemiological study investigating flooding exposure as a risk factor for biological aging acceleration.We observed that participants exposed to higher levels of floods were more likely to have advanced biological age.The risks associated with flooding exposure lasted     to the flood events, but always persist for longer than the events.Typically, they can include financial losses, displacement, concerns about health due to poor access to medical and health facilities, or difficulties with insurance and compensation (Tempest et al 2017).Transient stress exposure has been shown to augment cellular stress responses (referred to as 'hormetic stress'), leading to lengthened lifespan.In contrast, exposure to stressful life experiences, when occurring repeatedly or over a prolonged period, can overwhelm compensatory responses (referred to as 'toxic stress') and accelerate the rate at which the body ages (Epel and Lithgow 2014).Some physiological mechanisms have been proposed through which experiences of psychosocial stress and adversity contribute to accelerated aging (Polsky et al 2022).In response to stress, norepinephrine and epinephrine are released to stimulate the metabolic activity of mitochondria, resulting in elevated production of reactive oxygen species (ROS).Excessive ROS can cause cellular senescence or necrosis through telomere shortening and DNA damage.The accumulation of senescence or necrosis further promotes systemic inflammation, which leads to increased tissue damage.Meanwhile, norepinephrine and epinephrine can inhibit the clearance of damaged mitochondria, which exacerbates cellular dysfunction and the activation of cellular stress (Polsky et al 2022).
The association between biological aging acceleration and floods is comparable between PhenoAge and KDM-BA over a lag period of 0-4 years.However, there is heterogeneity in the lag pattern of the association observed in the two biological aging measures.This finding aligns with previous research indicating varying associations among different measures of biological age (Kresovich et al 2019, Mak et al 2023).The inconsistent results may be attributed to the distinct algorithms used for generating the biological age measures.Different from KDM-BA, which primarily captures chronological age, PhenoAge incorporates additional information on mortality risk alongside chronological age.In addition, we observed a significantly negative association for PhenoAge in the current year.The negative association may partially arise from a relatively short duration of exposure, which may not fully capture the delayed effects of floods.Additionally, as mentioned earlier, transient stress exposure has been associated with enhanced cellular stress responses, resulting in a deceleration of biological aging and an extended lifespan (Epel and Lithgow 2014).Nonetheless, further analyses in large cohorts are required to validate our results and establish more robust associations between biological age acceleration and floods.
In the context of climate change, flooding is one of the most recurrent and stressful disasters to deal with.The stress it induces persists long after the flood water has subsided (Stanke et al 2012).Although the majority of individuals affected by flooding exhibit remarkably resilience, there remains a potential for them to develop advanced biological age, which is associated with increased risk of cardiovascular diseases, mental health disorders, cancer, and all-cause mortality (Chen et al 2016, Gao et al 2023, Mak et al 2023).Our findings suggest that substantial public health support should be put in place after floods in the short-, medium-and longer-terms, to support psychosocial resilience and maintain the well-being of people who are affected.In addition to financial support and reconstruction during recovery, such support should also include: recognizing and responding to their distress; taking actions to prevent the onset of additional mental health problems or disorders; and providing primary care services or specialized mental healthcare if needed.As biological age is associated with a wide range of health consequences (Chen et al 2016, Gao et al 2023, Mak et al 2023), future research could further explore the role of biological aging in floodrelated diseases.Some limitations must be acknowledged.Similar to other observational studies, although we used exposure preceding the date of blood draw, causality remains uncertain.Despite extensive adjustment for covariates, the presence of residual and unmeasured confounding factors cannot be completely ruled out, leaving room for potential influence on the observed associations.There is no gold standard for the algorithms of biological aging and different algorithms might have different implications for biological aging.Therefore, future studies are needed to replicate our findings using different biological aging measures (e.g.telomere length, deficit-accumulation frailty indices, and epigenetic clocks).Our participants were residents in the UK who were more likely to be healthier and wealthier, therefore, our results may not be generalizable to a whole population, especially people in low-and middle-income countries.Despite the robustness of our effect estimates, which were based on participants residing at their current address for at least five years, there remains the possibility that some individuals may relocate before a flood and return to their homes once the waters recede.We assumed that participants did not move, which may have underestimated the effect of floods if an individual moved from an area with a high risk of flooding to an area with a lower risk of flooding.

Conclusion
In conclusion, our findings suggest that long-term exposure to floods may lead to advanced biological aging.The association was consistent regardless of demographics, socio-economic status, and lifestyles.Our work provides the basis for further understanding of the flood-related health impacts and suggests that public health policies and adaptation measures should be initiated in the short-, medium-and longer-terms after floods.

a
Categorical variables were expressed as count (percentage) and continuous variables were presented as mean ± SD; Baseline characteristics of participants with any missing values in covariates are shown in tableS3.Abbreviation: BMI: body mass index; KDM: Klemera-Doubal method.

Figure 2 .
Figure 2. Change in biological age acceleration measures associated with per interquartile range (IQR) increase in flood index on different lag years for biological age measures from the fully adjusted model.Abbreviation: KDM, Klemera-Doubal method; AC, acceleration.

Figure 3 .
Figure 3. Change in biological age acceleration measures associated with per interquartile range (IQR) increase in flood index over lag 0-4 years.Model 1 was the crude model; Model 2 adjusted for age, sex, and ethnicity; Model 3 additionally adjusted for socio-economic factors (education attainment, annual household income, and area deprivation level); Model 4 additionally adjusted for BMI, alcohol consumption status, cigarette smoking status, physical activity, and healthy diet score; Model 5 was additionally adjusted for temperature and relative humidity.Abbreviation: KDM, Klemera-Doubal method; AC, acceleration.

Table 1 .
Baseline characteristics of the study population.