Early life PM2.5 exposure, childhood cognitive ability and mortality between age 11 and 86: A record-linkage life-course study from Scotland

Gergő Baranyi, Lee Williamson, Zhiqiang Feng, Sam Tomlinson, Massimo Vieno, Chris Dibben
doi: https://doi.org/10.1101/2023.05.18.23289845

ABSTRACT

Background Living in areas with high air pollution concentrations is associated with all-cause and cause-specific mortality. Exposure in sensitive developmental periods might be long-lasting but studies with very long follow-up are rare, and mediating pathways between early life exposure and life-course mortality are not fully understood.

Methods Data were drawn from the Scottish Longitudinal Study Birth Cohort of 1936, a representative record-linkage study comprising 5% of the Scottish population born in 1936. Participants had valid age 11 cognitive ability test scores along with linked mortality data until age 86. Fine particle (PM2.5) concentrations estimated with the EMEP4UK atmospheric chemistry transport model were linked to participants’ residential address from the National Identity Register in 1939 (age 3). Confounder-adjusted Cox regression estimated associations between PM2.5 and mortality; regression-based causal mediation analysis explored mediation through childhood cognitive ability.

Results The final sample consisted of 2734 individuals with 1608 deaths registered during the 1,833,517 person-months at risk follow-up time. Higher early life PM2.5 exposure increased the risk of all-cause mortality (HR=1.03, 95% CI: 1.01-1.04 per 10μg m-3 increment), associations were stronger for mortality between age 65 and 86. PM2.5 increased the risk of cancer-related mortality (HR=1.05, 95% CI: 1.02-1.08), especially for lung cancer among females (HR=1.11, 95% CI: 1.02-1.21), but not for cardiovascular and respiratory diseases. Higher PM2.5 in early life (≥50μg m-3) was associated with lower childhood cognitive ability, which, in turn, increased the risk of all-cause mortality and mediated 25% of the total associations.

Conclusions In our life-course study with 75-year of continuous mortality records, we found that exposure to air pollution in early life was associated with higher mortality in late adulthood, and that childhood cognitive ability partly mediated this relationship. Findings suggest that past air pollution concentrations will likely impact health and longevity for decades to come.

  • We explored PM2.5 at age 3 and mortality between age 11 and 86 in a Scottish cohort

  • PM2.5 increased the risk of all-cause mortality, especially between the age of 65 and 86

  • Childhood cognitive ability mediated 25% of the total association

  • Associations were prominent for (lung) cancer mortality, especially among females

  • Air pollution in early life may affect health and longevity across the life course

INTRODUCTION

Ambient air pollution poses one of the greatest environmental threats to human health. Fine particulate matter with the aerodynamic diameter of less than 2.5µm (PM2.5) is the fifth largest mortality risk factor, with an estimated 4.2 million attributable death and 103 million disability adjusted life years in 2015 (1). Consistent and of high certainty evidence links PM2.5 exposure to increased risk of all-cause and cause-specific (e.g., cardiovascular, lung cancer, respiratory) mortality in the general population (2–5). Large number of studies focussed on the short (from one hour to 7 days) (5), medium (<10 years) and long-term impacts (<25 years) (2) or poor air quality, however, the number of investigations drops when it comes to very-long term (≥25 years) associations (6–11).

Exposure to high levels of air pollution may be particularly detrimental and long-lasting when it intersects with sensitive periods of human development; still, few studies followed up narrow-age cohorts (9) which is crucial to assess the developmental timing of exposure. Evidence suggests that the time before birth and the first years of life are critical, providing the foundation of later health and lifespan development (12, 13). Air pollution exposure around birth is associated with adverse birth outcomes (14, 15), altered brain development (16), and modifications in DNA methylation patterns (17), key predictors of healthy ageing and longevity (18, 19). It is plausible that air pollution during this early sensitive/critical period has very long-term impact on health (and subsequent mortality), likely lasting throughout the entire life course. Still, due to lack of historical air pollution data and cohort studies with life-course follow-up, previous investigations were not able to explore this hypothesis.

Using a representative cohort of the Scottish population born in 1936, we examined whether and how exposure to PM2.5 in early life contributed to a) all-cause mortality across almost the entire life course (i.e., between age 11 and 86), to b) cause-specific mortality during late adulthood, and c) whether associations with all-cause mortality were mediated by childhood general cognitive ability.

METHODS

Data were taken from the Scottish Longitudinal Study Birth Cohort of 1936 (SLSBC1936), a retrospective cohort study assembled from routinely collected administrative data, representative of the Scottish population born in 1936. SLSBC1936 is based on very high-quality data linkages between four sources, a) the 1939 National Identity Register, a census-like registry of the entire UK population carried out on the 29th September 1939 at the start of World War Two; b) the Scottish Mental Survey 1947 (SMS1947), a general cognitive ability test undertaken among almost all 11-year old Scottish schoolchildren in 4th June 1947; c) Scotland’s National Health Services Central Register (NHSCR); and d) the Scottish Longitudinal Study, a 5.3% administrative sample of the Scottish population selected based on 20 semirandom birthdays and followed up since 1991 (20). In addition to the core SLSBC1936 sample, we also included individuals who would have entered the Scottish Longitudinal Study if they did not die or leave the country before in 1991; detailed information on matching procedure are available elsewhere (20). Therefore, our sample comprised individuals born in 1936 on one of the 20 Scottish Longitudinal Study birth dates, enumerated in the 1939 National Identity Register, participated in the SMS1947, and captured by the NHSCR system, totalling to a maximum sample size of 3626 individuals.

Exposure to PM2.5 pollution

Annual level of PM2.5 concentrations for the modelling year of 1935 was estimated using the EMEP4UK version 4.17 derived from the EMEP MSC-W version 4.17 (21) atmospheric chemistry transport model (22–24). The meteorology year used here was 2015. The meteorological driver used was the Weather Research and Forecast (WRF) model version 3.9.1.1 (25) based on the Global forecast system final re-analysis (GFS-FNL) as initial condition and 6-h nudging (26). Gridded anthropogenic emission of pollutants (nitrogen oxides, sulphur oxides, ammonia, non-methane volatile organic compounds, carbon monoxide, course and fine particulate matter) were estimated for the year 1950 (27), and a scaled version of these data distributions, based upon activity data research, were used to make emission estimates in 1935 (28). Non-NH3 activity data in that era are largely a reflection of the use of fossil fuels such as coal and of some oil-derived products while agricultural data, the majority source of NH3 emissions, was derived from the Vision of Britain database (29). The model employs a horizontal resolution of 0.5 x 0.5 degrees to a greater European domain providing the boundary condition for a nested UK domain with a horizontal resolution of 0.037 x 0.037 degrees (∼ 3km × 4km). Estimated PM2.5 concentrations ranged between 4.2μg m-3 to 116.9μg m-3 for Scotland (Figure 1). The EMEP4UK model has been evaluated in the UK and internationally showing good agreement between modelled and observed concentrations (30, 31); historical concentrations of air pollution exposure derived from this model have been used in previous epidemiological studies (12, 28).

Figure 1:
Figure 1:

Mean 1935 PM2.5 concentration (in μg m-3) in Scotland, and in the Central Belt of Scotland, estimated using the EMEP4UK atmospheric chemistry transport model with a horizontal resolution of ∼3 × 4 km. Index map in the right bottom corner presents Scotland’s location within the United Kingdom; black lines are council areas based on current boundaries.

Participants’ residential addresses were transcribed from the 1939 National Register and were geocoded using the Historical Address Geocoding – GIS software (32), which links historical records to contemporary addresses. Grid reference was linked to the sample using the 1939 street index (records with house numbers, street names, enumeration district and the registration district codes) and centroid of the 1939 enumeration districts, with very high linkage rate (33). Geocoded 1939 addresses were intersected with PM2.5 concentrations modelled for the year 1935.

Mortality records

All-cause mortality was derived from NHSCR data indicating the month and year of death between June 1947 (i.e., date of SMS1946) and June 2022 (i.e., 901 months). The NHSCR data also includes information for those deaths occurring in other parts of the United Kingdom or overseas, which are not part of the routine Vital Event deaths (i.e., civil registration deaths). Cause-specific mortality was obtained from the routine Vital Events deaths database, available only for those who became part of the Scottish Longitudinal Study after 1991 and died within Scotland between April 1991 (i.e., date of 1991 Census) and December 2017 (i.e., 321 months). Cause of death was reported according to the International Classification of Diseases 9th Revision (ICD-9) or 10th Revision (ICD-10) which we classified into natural causes (ICD-9: 001 to 799; ICD-10: A00 to R99), cardiovascular (ICD-9: 390 to 459; ICD-10: I00 to I99), respiratory (ICD-9: 460 to 519; ICD-10: J00 to J99), and cancer-related deaths (ICD-9 140 to 239; ICD-10: C00 to D48).

General cognitive ability in childhood

On the Wednesday 4th June 1947, 94% of all Scottish children born in 1936 (n=70805) participated on a cognitive assessment as part of the SMS1947. The survey utilised the Moray House No.12 test, a 71 item general cognitive ability test, including mental tasks such as same-opposites, word classification, analogies, reasoning, cipher decoding and spatial items (18, 34). Tests were administered in school classes across the entire country on the same day, with the same instructions and within a 45-minute test interval (18). The test score had a maximum value of 76 and it correlated highly with the Stanford-Binet test providing good criterion validity (34).

Covariates

Potential confounders of the exposure, outcome and mediator relationship were identified from the literature and are presented in a directed acyclic graph (Figure 2). Sex (male, female) and age (in years) were derived from the SMS1946. Mother’s age (in years), mother’s marital status (married, not married) and parental occupational social position was derived from the 1939 National Identity Register. For the latter, occupational titles were first manually coded for each member of the household using the Historical International Classification of Occupations coding system (35). Occupational codes were assigned a value of occupational social position based on the HISCAM scale (36), which is a historic measure of occupational social stratification during the 19th and early 20th centuries. The continuous HISCAM score is centred around the mean of 50 with standard deviation of 10 in the underlying population; values may range between 1 and 99 with greater numbers indicating higher social position (36). To each participant, we assigned the father’s social position (or the mother’s if fathers’ not available); otherwise, we calculated the average social position of all other household members.

Figure 2:
Figure 2:

Directed acyclic graph presenting associations between exposure, mediator, outcome and confounders, and the main source of data. Associations between confounders are not shown for simplicity, solid black lines are associations of interest. Geocoded 1939 addresses were linked to PM2.5 concentrations estimated with the EMEP4UK model for the year 1935.

Statistical analysis

Cox proportional hazards regressions estimated the associations between PM2.5 and mortality using months of follow-up as the time scale. Effect estimates were presented as Hazard Ratios (HR) with 95% confidence intervals (CI) for each 10μg m-3 incremental increase. Subjects were censored at the time of their death, at the time they last left Scotland (from NHSCR data), or at the end of the study period if no events were registered during the follow-up period. Based on the Schoenfeld residuals, the proportional hazards assumption was met for all covariates but for sex; therefore, we incorporated strata for sex and presented stratified findings for males and females. To determine the representativeness of the analytical sample, we compared the sex distribution, average age and average Moray House test scores with the population values coming from the SMS1947 (37).

The main analysis focussed on early life PM2.5 exposure and all-cause mortality. Two confounder models were set a priori: Model 1 controlled for age and sex (strata); Model 2 further adjusted for parental occupational position, mother’s age, and mother’s marital status. Continuous variables were scaled to support interpretation. In addition to treating the associations as time-independent, we also tested whether the relationship between PM2.5 exposure and all-cause mortality changed during follow-up by splitting the follow-up time into four intervals: first 527 months (i.e., average age of 11-55 years), from 528 to 647 months (i.e., average age of 55-65 years), from 648 to 767 months (i.e., average age of 65-75 years), and from 768 to 901 months (i.e., average age of 75-86 years). Finally, based on Model 2 adjustments we explored concentration-response relationship by adding penalized splines with 2, 3 and 4 degrees of freedom. The best model was determined based on the Bayesian Information Criterion (BIC) and linearity of relationship was examined with χ2 test.

Associations between PM2.5 in early life and mortality due to natural causes, as well as cardiovascular, respiratory and cancer mortality were estimated in competing risks analysis. Models were operationalised in Cox proportional hazards regression by fitting two endpoints at the same time (i.e., outcome of interest versus other deaths) with distinct baseline hazards for PM2.5, age, and sex (strata), and shared coefficients for parental occupational social position, mother’s age, and mother’s marital status. In contrary to calculating cause-specific hazards regression, this approach takes into consideration that competing events are mutually exclusive (38). Model 1 and Model 2 adjustments, as well as sex-stratified findings were presented.

Total association between PM2.5 exposure and all-cause mortality was decomposed into direct and indirect (through childhood general cognitive ability) effects applying regression-based causal mediation analysis within the counterfactual framework (39). For the mediator model, we fitted a linear regression, for the outcome model a Cox regression; the same Model 2 confounders were included in both mediator and outcome models. Percentage mediated was expressed as the percentage of total association explained by the mediator; 95% CIs for total, direct and indirect effects were estimated based on 10000 bootstrap replications. For the mediation analysis, PM2.5 exposure was dichotomised as <50 versus ≥50μg m-3, a cut off separating the highest ∼20% from the rest of the sample. We presented findings for the non-stratified sample only.

Six sets of sensitivity analyses were carried out. First, we presented main results (i.e., all-cause mortality) for alternative exposure operationalisations (i.e., interquartile range [IQR] increase, threshold [<50, ≥50μg m-3]). Second, we provided main models after adjusting for urban versus rural area of residence, which was determined based on living in Aberdeen, Dundee, Edinburgh, or Glasgow (i.e., urban) versus in other areas (i.e., rural), by intersecting 1939 addresses with current council boundaries. Third, proportion of missing data was significant for some of the covariates, therefore, we reran main models with multiple imputation by chained equation. Number of imputations was determined automatically based on the proportion of incomplete cases (40). Four, we dropped individuals with extremely high PM2.5 values (≥100μg m-3) residing in the centre of Glasgow City and reran main models. Five, instead of fitting two endpoints in Cox regression, competing risks were calculated with the Fine-Gray proportional subdistribution hazards regression. Last, we presented the mediation analysis with an alternative cut-off for ‘low’ and ‘high’ air pollution levels (<40 versus ≥40μg m-3) capturing ∼75% versus ∼25% of the sample.

Analyses were carried out using the survival (41), regmedint (42), mice (43) and tidycmprsk (44) packages in R 4.2.2 (45).

RESULTS

The final SLSBC1936 sample without missing data included 2734 individuals (48.4% females) contributing to a total of 1,833,517 person-months (152,793.1 person-years) at risk with a median follow-up time of 771 months (64.3 years). During follow-up, 1608 individuals died, 475 left the study and 651 were still alive at the end of the study. The average exposure to PM2.5 was 31.3μg m-3 (SD=32.6; IQR=9.3-39.0). Comparing the SLSBC1936 analytical sample with the population of individuals born in 1936 (i.e., the main SMS1947 sample) suggested that neither the percentage of males and females, nor the average Moray House test scores differed between the SLSBC1936 sample and the population; however, our sample was approximately one month older than the population average likely due to the distribution of 20 semirandom birth days across the year (Table 1).

View this table:
Table 1:

Characteristics of participants in the analytical sample and in the respective population.

After adjusting for age and sex (strata), we found 3% higher hazard of all-cause mortality per 10μg m-3 increment increase in early life PM2.5 exposure (HR=1.03, 95% CI: 1.01-1.04). The association remained the same after further adjustments for parental occupation social position, mother’s marital status, and mother’s age (HR=1.03, 95% CI: 1.01-1.04) (Table 2). Associations were varying across the life course: while between age 11 and 65 there was no association, 10μg m-3 higher early life PM2.5 exposure increased the hazard of all-cause mortality between age 65 and 75 years by 4% (95% CI: 1.02-1.07), and between age 75 and 86 by 3% (95% CI: 1.00-1.05) (Model 2). Stratified analyses suggested that among females the association was strongest between age 75-86, while among males it was strongest between age 65-75 (Table 2). Furthermore, in the total sample, concentration-response relationship was monotonic, linear and best-fitting with two degrees of freedom (likelihood ratio test for non-linear component: χ2=0.43; p=0.530; see Bayesian Information Criterion values in Supplementary Table 1). HRs constantly increased with increasing PM2.5 concentrations but first crossed the value of HR=1.0 at the value of ∼38μg m-3. Although the shape of the association visually differed between females and males, confidence intervals were wide and overlapping (Figure 3).

Figure 3:
Figure 3:

Concentration-response relationship between PM2.5 in early life and all-cause mortality in the Scottish Longitudinal Study Birth Cohort of 1936 for the (a) total sample; and among (b) females and (c) males. Curves were obtained by penalised splines with two degrees of freedom; observed relationships were linear. All-cause mortality was available between June 1947 and June 2022. Model was adjusted for age, (sex (strata),) parental occupational social position, mother’s marital status, and mother’s age. The total sample size was n=2734. Source: Scottish Longitudinal Study.

View this table:
Table 2:

Associations between early life PM2.5 exposure and all-cause mortality in the Scottish Longitudinal Study Birth Cohort of 1936 a) during the total follow up and b) by follow-up intervals.

The association with cause-specific mortality was assessed in a restricted sample (n=1947) as information on cause of death was only available between April 1991 and December 2017. During this period 939 deaths were registered due to natural causes, 321 cardiovascular, 109 respiratory, and 341 cancer-related deaths. Competing risks analysis with adjustment for all covariates suggested that early life PM2.5 exposure increased the hazard of cancer-related mortality by 5% (95% CI: 1.02-1.08) but there was no association with cardiovascular (HR=1.00, 95% CI: 0.97-1.04) or respiratory mortality (HR=1.02, 95% CI: 0.96-1.07) (Table 3). Sex-specific analyses showed that the association with cancer-related mortality was prominent in the female subsample. As a post-hoc analysis, we further explored whether lung cancer (ICD-9: 162; ICD-10: C34) was the main driver of cancer-related mortality among females, which was confirmed (HR=1.11, 95% CI: 1.02-1.21). Among males, mortality due to natural cases other than cancer, cardiovascular or respiratory was significantly associated with early life exposure to PM2.5. Although not statistically significant, post-hoc we found increased risk of death (HR=1.08, 95% CI: 1.00-1.17; p=0.055) due to neurodegenerative disorders (Alzheimer’s disease and related disorders [ICD-9: 046.1, 290.0–290.4, 294, 331; ICD-10: F00-F03, G30-31], Parkinson’s disease [ICD-9: 332; ICD-10: G20-22], amyotrophic lateral sclerosis [ICD-9: 335.2; ICD-10: G12.2]).

View this table:
Table 3:

Associations between early life PM2.5 exposure and cause-specific mortality in Scottish Longitudinal Study Birth Cohort of 1936.

The total effect of early life PM2.5 exposure on all-cause mortality was decomposed into direct association and indirect association mediated by childhood general cognitive ability. Individuals with ‘high’ early life PM2.5 exposure (≥50μg m-3) had a 39-month shorter median survival time in comparison to those from relatively ‘low’ concentration areas (<50μg m-3). Regression-based causal mediation analysis suggested that – in the fully adjusted model – being exposed to high concentrations of PM2.5 was associated with a 22% higher risk of all-cause mortality (95% CI: 1.08-1.38) (i.e., total effects). Approximately 25% of this association was mediated by general cognitive ability (HR=1.05; 95% CI: 1.03-1.07) (i.e., indirect effects): higher PM2.5 levels contributed to lower test scores, and lower test scores were associated with higher risk of all-cause mortality. The direct effect (i.e., not explained by general cognitive ability) of PM2.5 exposure on all-cause mortality was HR=1.17 (95% CI: 1.03-1.32) (Figure 4; Supplementary Table 2).

Figure 4:
Figure 4:

Total, direct, and indirect associations between ‘high’ early life PM2.5 exposure (≥50μg m-3) and all-cause mortality, with mediation through childhood general cognitive ability. All-cause mortality was available between June 1947 and June 2022. Model was adjusted for age, sex (strata), parental occupational social position, mother’s marital status, and mother’s age. The total sample size was n=2734. Source: Scottish Longitudinal Study.

Sensitivity analyses determined that 1-IQR increase in PM2.5 exposure was associated with 8% (95% CI: 1.03-1.13) increased risk of all-cause mortality in the fully adjusted model (Supplementary Table 3); findings for dichotomised air pollution levels (<50μg m-3, ≥50μg m- 3) can be found in Supplementary Table 4. Adjusting for urban-rural differences in 1939 slightly attenuated the findings which did not remain significant in the female subsample (Supplementary Table 5). Imputing missing data through multiple imputation did not change the results (Supplementary Table 6); excluding participants with very high exposure (≥100μg m-3) only led to marginal changes (Supplementary Table 7). Exploring competing risks with Fine-Gray proportional subdistribution hazards regression produced the same cause-specific associations as presented earlier (Supplementary Table 8). Finally, after applying the 40μg m- 3 cut-off for the mediation analysis, 21% of the total association was mediated by childhood cognitive ability (Supplementary Table 9; Supplementary Figure 1).

DISCUSSION

Using a representative administrative cohort, we explored the direct and indirect associations between PM2.5 concentrations in early life and mortality across 75 years in 2734 participants. Exposure to higher levels of PM2.5 were associated with increased hazard of all-cause mortality, especially between the ages of 65 and 86. Competing risk analyses highlighted that associations were only present for cancer-related deaths, especially for lung-cancer mortality among females. One potential pathway connecting early life air pollution exposure to mortality is via lower childhood cognitive ability, which mediated 25% of the total associations in our sample.

This study strengthens the evidence base linking long-term PM2.5 exposure to increased risk of mortality in the UK and worldwide (2, 6–11). The estimated HR was 1.03 per 10μg m-3 incremental increase for all-cause mortality, which is much lower than the risk (Risk Ratio[RR]=1.08) estimated in a recent meta-analysis (2); however, this is not unexpected as previous findings with long-term follow-ups pointed towards decreasing effect sizes by more distal air pollution exposures in comparison to proximal ones (8, 46). Given the wide range of PM2.5 exposures, and the gentle, but steadily increasing slope of association, PM2.5 first crossed the HR=1 at the relatively higher concentration of 38μg m-3, which is much higher than thresholds of unhealthy exposure suggested by the literature (2, 47), including the WHO guidelines (48). We are not aware of previous investigations with a follow-up time of longer than 40 years (e.g., (8)) and making direct comparison to the literature is challenging; still, it is remarkable that toxic levels of air pollution might contribute to mortality 60-80 years after exposure.

Early life PM2.5 exposure was more strongly associated with all-cause mortality between age 65 and 86 and with cancer-related mortality, especially lung cancer. In contrary to all-cause mortality, HR for lung cancer in our female subsample (HR=1.11) was close to identical to risks reported in the literature based on shorter follow-up studies (RR=1.12) (2). Cancer-related mortality is the second leading cause of death according to the Global Burden of Disease Study, with tracheal, bronchus, and lung cancer being the leading cause of cancer deaths (49). Their relative proportion in comparison to other cancers is particularly high between the ages of 60 and 80 years (49), which overlaps with our findings. In contrary to the literature suggesting consistent associations between PM2.5 pollution and cardiovascular mortality (2), there was null association in our study. It might be plausible that PM2.5 in early life, or its chemical composition in the late 1930s, did not have such a long-lasting association with cardiovascular diseases or that early life is not a sensitive period for late adulthood cardiovascular events. Air pollution exposure has been also associated with neurodegenerative disorders in the literature (50), with stronger associations in males (51), which align with our preliminary findings. Still, these hypotheses cannot be answered from this study and need to be explored in further investigations.

During the prenatal period, air pollution-induced oxidative stress and systematic inflammation in the mother’s body may affect in utero development (52), and there is some evidence indicating that very tiny particles might even translocate into the placenta causing oxidative stress in the foetus (53). In early life, inhaled air pollutants can affect the children’s development more directly (17, 54). Toxic environmental exposures – through oxidative stress and inflammation – are associated with molecular changes in the somatic cells, and these alterations might survive subsequent cell replications and affect healthy development and longevity. For example, modified DNA methylation patterns, responsible for gene expression and genome stability, as well as shorter telomeres, a marker of cellular ageing, have been associated with air pollution exposure in early and later life (17, 55). A recent Scottish study, based on a sample (Lothian Birth Cohort 1936) selected from the same 1936-born population as in our study, showed that higher air pollution exposure around birth was associated with shorter DNA methylation-based telomere length in late adulthood, especially among females (12), even after adjusting for a variety of life-course social and behavioural (e.g. smoking) risk factors. Epigenetic markers increase the risk of developing cancer (56) and mortality (57), presenting a plausible explanation for our (sex-specific) findings.

Air pollution in early life may also affect brain development: maternal air pollution exposure from the 25th gestational week onwards is associated with head growth (58) and being exposed to higher prenatal and postnatal PM2.5 concentration is linked to poorer cognitive performance in childhood (52, 59, 60). The human brain undergoes rapid developments in the early years which continues well into childhood with the differentiation and maturation of neurons and myelination of axons (61, 62). A whole population follow up on the SMS1947 with more than 65000 participants showed that higher general cognitive ability scores in childhood decreased all-cause mortality and cause-specific mortality for almost all major causes of death, including smoking-related cancers (18). Higher cognitive ability in childhood is a powerful predictor of longevity with pathways including higher educational attainment and adult socioeconomic status, which can impact among others health literacy, health behaviour, and diseases management (19, 63). Our findings adds that higher early life PM2.5 exposure modestly decreases general cognitive ability in childhood (even after adjusting for parental occupational social position), thus indirectly contributing to all-cause mortality; approximately one-quarter of the total association was mediated through childhood cognitive ability presenting an important pathway between early life toxic exposure and mortality.

Strengths and limitations

This study demonstrated the feasibility of exploring the relationship between early life PM2.5 exposure and mortality over almost the entire life course. The study benefitted from high quality data linkages between different historical and contemporary administrative sources, included key family and household confounders in early childhood, a validated measure of general cognitive ability at age 11, and an exceptionally long follow-up of mortality stretching over 75 years. Residential information at age 3 was derived from administrative data, thus not prone to recall bias. However, several key limitations should be considered when interpreting the results. First, cause-specific mortality was only available for a relatively shorter follow-up time (∼26 years). Although we do not assume that longer follow-up would have change the reported associations with cancer-related deaths, it is plausible that adding further mortality records across the life course may reveal other cause-specific mortality associations in future studies. Second, information on mortality before age 11 was not available in the SLSBC1936. Given the high mortality rates among 1-year olds in England during the 1930s (64), and the strong association between PM2.5 concentrations and infant mortality in high pollution contexts (e.g. India (65)), it is plausible that air pollution had a strong effect on mortality in the earlier years which we were not able to capture. Third, early life residential location was only available in the 1939 National Identity Register. Although the probability of residential mobility for this birth cohort is relatively low in their earliest years (66), we cannot ascertain whether air pollution exposure during prenatal or during postnatal periods contributed to higher risk of mortality. Moreover, missing residential history during childhood introduces further uncertainty to identify sensitive (or critical) exposure windows. Fourth, there are inherent uncertainties at all stages of the emissions estimation process, even for the present day, and many of these problems are magnified when trying to recreate a historical context. The same atmospheric chemistry transport models for newer modelling years (2009-2010) have been evaluated against concurrent PM2.5 concentrations, which indicated good agreement between estimated and measured concentrations, but also suggested that PM2.5 concentrations might be underestimated in the EMEP4UK models (30). While there are increasing uncertainties when pollutants are estimated further back in time, their relative spatial concentrations can be considered as sufficiently accurate. Last, key air pollution regulatory policies throughout the 20th century (e.g., Clean Air Acts 1956, 1968, 1993) reduced PM2.5 concentrations in the UK (67), and probably changed its chemical components. Pollutant concentrations experienced by sample participants in their early lives are rarely present in high-income countries anymore. Still, with increasing global inequities in pollution though raising concentration in low– and middle-income countries (48), our findings have relevant global implications.

Future studies could capitalise on routinely collected data utilising census responses, healthcare service use data and Vital Events on marriages and children’s birth to extract information on residential locations across the entire life course and provide a more comprehensive assessment on how air pollution contributes to morbidity and mortality from foetal period onwards. Mechanistic pathways, including social, cognitive, and biological processes, should be investigated which have the potential to connect childhood exposure to late adulthood health.

Conclusions

Exposure to higher PM2.5 in early life was associated with higher all-cause and cancer-related mortality, especially between the ages of 65 and 75, with one-quarter of the association being mediated through lower childhood general cognitive ability. Future investigation should explore mechanistic pathways and assess the relative importance of air pollution exposure at different time-points in our lives. The findings suggest that air pollution may have very long-term impact on health and longevity, with its detrimental impact lasting long after mitigation policies reduced concentration levels.

Data Availability

Due to the sensitive nature of the data produced for this study, it is only accessible for SLS Approved Researchers from a Safe Setting. Access to data is granted after the SLS Research Board's approval.

Supplementary Materials

View this table:
Supplementary Table 1:

Model fit with penalised splines

View this table:
Supplementary Table 2:

Outcome and mediator models in the regression-based causal mediation analysis (50μg m-3 cut-off)

View this table:
Supplementary Table 3:

Associations between early life PM2.5 exposure and all-cause mortality expressed as interquartile range change.

View this table:
Supplementary Table 4:

Associations between low versus high early life PM2.5 exposure and all-cause mortality.

View this table:
Supplementary Table 5:

Associations between early life PM2.5 exposure and all-cause mortality a) during the total follow up and b) by follow-up intervals after adjusting for residential location.

View this table:
Supplementary Table 6:

Associations between early life PM2.5 exposure and all-cause mortality a) during the total follow up and b) by follow-up intervals after multiple imputation.

View this table:
Supplementary Table 7:

Associations between early life PM2.5 exposure and all-cause mortality a) during the total follow up and b) by follow-up intervals after dropping individuals with very high exposure (≥100μg m-3).

View this table:
Supplementary Table 8:

Associations between early life PM2.5 exposure and cause-specific mortality calculated with Fine-Gray regression.

View this table:
Supplementary Table 9:

Outcome and mediator models in the regression-based causal mediation analysis (40μg m-3 cut-off)

Supplementary Figure 1:
Supplementary Figure 1:

Total, direct, and indirect associations between high PM2.5 exposure in early life (≥40μg m-3) and all-cause mortality, with mediation through childhood general cognitive ability. All-cause mortality was available between June 1947 and June 2022. Model was adjusted for age, sex (strata), parental occupational position, mother’s marital status, and mother’s age. The sample size was n=2734. Source: Scottish Longitudinal Study.

ACKNOWLEDGEMENT

The help provided by staff of the Longitudinal Studies Centre – Scotland (LSCS) is acknowledged. The LSCS is supported by the ESRC/JISC, the Scottish Funding Council, the Chief Scientist’s Office and the Scottish Government. The authors alone are responsible for the interpretation of the data. Census output is Crown copyright and is reproduced with the permission of the Controller of HMSO and the King’s Printer for Scotland.

Footnotes

  • FUNDING This work was supported by Health Data Research UK (HDR-5012). This work was also in part supported by the Natural Environment Research Council award number NE/R016429/1 as part of the UK-SCAPE programme delivering National Capability.

  • Competing Interest: No competing interest declared.

REFERENCES

  1. 1.↵
    Cohen AJ, Brauer M, Burnett R, Anderson HR, Frostad J, Estep K, et al. Estimates and 25-year trends of the global burden of disease attributable to ambient air pollution: an analysis of data from the Global Burden of Diseases Study 2015. The Lancet. 2017;389:1907–18.
    OpenUrl
  2. 2.↵
    Chen J, Hoek G. Long-term exposure to PM and all-cause and cause-specific mortality: A systematic review and meta-analysis. Environ Int. 2020;143:105974.
    OpenUrlCrossRefPubMed
  3. 3.
    Pope CA, 3rd., Coleman N, Pond ZA, Burnett RT. Fine particulate air pollution and human mortality: 25+ years of cohort studies. Environ Res. 2020;183:108924.
    OpenUrlCrossRefPubMed
  4. 4.
    Liu C, Chen R, Sera F, Vicedo-Cabrera AM, Guo Y, Tong S, et al. Ambient Particulate Air Pollution and Daily Mortality in 652 Cities. New England Journal of Medicine. 2019;381:705–15.
    OpenUrlCrossRefPubMed
  5. 5.↵
    Orellano P, Reynoso J, Quaranta N, Bardach A, Ciapponi A. Short-term exposure to particulate matter (PM(10) and PM(2.5)), nitrogen dioxide (NO(2)), and ozone (O(3)) and all-cause and cause-specific mortality: Systematic review and meta-analysis. Environ Int. 2020;142:105876.
    OpenUrl
  6. 6.↵
    Filleul L, Rondeau V, Vandentorren S, Le Moual N, Cantagrel A, Annesi-Maesano I, et al. Twenty five year mortality and air pollution: results from the French PAARC survey. Occup Environ Med. 2005;62:453–60.
    OpenUrlAbstract/FREE Full Text
  7. 7.
    Rosenlund M, Berglind N, Pershagen G, Hallqvist J, Jonson T, Bellander T. Long-term exposure to urban air pollution and myocardial infarction. Epidemiology. 2006;17:383–90.
    OpenUrlCrossRefPubMedWeb of Science
  8. 8.↵
    Hansell A, Ghosh RE, Blangiardo M, Perkins C, Vienneau D, Goffe K, et al. Historic air pollution exposure and long-term mortality risks in England and Wales: prospective longitudinal cohort study. Thorax. 2016;71:330–8.
    OpenUrlAbstract/FREE Full Text
  9. 9.↵
    Dehbi H-M, Blangiardo M, Gulliver J, Fecht D, de Hoogh K, Al-Kanaani Z, et al. Air pollution and cardiovascular mortality with over 25years follow-up: A combined analysis of two British cohorts. Environment International. 2017;99:275–81.
    OpenUrl
  10. 10.
    Yang J, Sakhvidi MJZ, de Hoogh K, Vienneau D, Siemiatyck J, Zins M, et al. Long-term exposure to black carbon and mortality: A 28-year follow-up of the GAZEL cohort. Environ Int. 2021;157:106805.
    OpenUrl
  11. 11.↵
    Yap C, Beverland IJ, Heal MR, Cohen GR, Robertson C, Henderson DE, et al. Association between long-term exposure to air pollution and specific causes of mortality in Scotland. Occup Environ Med. 2012;69:916–24.
    OpenUrlAbstract/FREE Full Text
  12. 12.↵
    Baranyi G, Deary IJ, McCartney DL, Harris SE, Shortt N, Reis S, et al. Life-course exposure to air pollution and biological ageing in the Lothian Birth Cohort 1936. Environ Int. 2022;169:107501.
    OpenUrl
  13. 13.↵
    Darling JC, Bamidis PD, Burberry J, Rudolf MCJ. The First Thousand Days: early, integrated and evidence-based approaches to improving child health: coming to a population near you? Arch Dis Child. 2020;105:837–41.
    OpenUrlAbstract/FREE Full Text
  14. 14.↵
    Bekkar B, Pacheco S, Basu R, DeNicola N. Association of Air Pollution and Heat Exposure With Preterm Birth, Low Birth Weight, and Stillbirth in the US: A Systematic Review. JAMA Netw Open. 2020;3:e208243.
    OpenUrl
  15. 15.↵
    Ju L, Li C, Yang M, Sun S, Zhang Q, Cao J, et al. Maternal air pollution exposure increases the risk of preterm birth: Evidence from the meta-analysis of cohort studies. Environ Res. 2021;202:111654.
    OpenUrl
  16. 16.↵
    Lubczyńska MJ, Muetzel RL, El Marroun H, Hoek G, Kooter IM, Thomson EM, et al. Air pollution exposure during pregnancy and childhood and brain morphology in preadolescents. Environ Res. 2021;198:110446.
    OpenUrl
  17. 17.↵
    Isaevska E, Moccia C, Asta F, Cibella F, Gagliardi L, Ronfani L, et al. Exposure to ambient air pollution in the first 1000 days of life and alterations in the DNA methylome and telomere length in children: A systematic review. Environ Res. 2021;193:110504.
    OpenUrlPubMed
  18. 18.↵
    Calvin CM, Batty GD, Der G, Brett CE, Taylor A, Pattie A, et al. Childhood intelligence in relation to major causes of death in 68 year follow-up: prospective population study. BMJ. 2017;357:j2708.
    OpenUrlAbstract/FREE Full Text
  19. 19.↵
    Calvin CM, Deary IJ, Fenton C, Roberts BA, Der G, Leckenby N, et al. Intelligence in youth and all-cause-mortality: systematic review with meta-analysis. Int J Epidemiol. 2011;40:626–44.
    OpenUrlCrossRefPubMedWeb of Science
  20. 20.↵
    Huang Z, Feng Z, Dibben C, Brett C, Deary IJ. The Scottish Longitudinal Study 1936 Birth Cohort. Scottish Longitudinal Study Development & Support Unit; 2017.
  21. 21.↵
    Simpson D, Benedictow A, Berge H, Bergström R, Emberson LD, Fagerli H, et al. The EMEP MSC-W chemical transport model – technical description. Atmos Chem Phys. 2012;12:7825–65.
    OpenUrl
  22. 22.↵
    Vieno M, Dore AJ, Stevenson DS, Doherty R, Heal MR, Reis S, et al. Modelling surface ozone during the 2003 heat-wave in the UK. Atmos Chem Phys. 2010;10:7963–78.
    OpenUrl
  23. 23.
    Vieno M, Heal MR, Hallsworth S, Famulari D, Doherty RM, Dore AJ, et al. The role of long-range transport and domestic emissions in determining atmospheric secondary inorganic particle concentrations across the UK. Atmos. Atmos Chem Phys. 2014;14:8435–47.
    OpenUrl
  24. 24.↵
    Vieno M, Heal MR, Williams ML, Carnell EJ, Nemitz E, Stedman JR, et al. The sensitivities of emissions reductions for the mitigation of UK PM2.5. Atmos Chem Phys. 2016;16:265–76.
    OpenUrl
  25. 25.↵
    Skamarock WC, Klemp JB, Dudhia J, Gill DO, Barker DM, Duda MG, et al. A description of the advanced research WRF version 3. Boulder, Colorado: U.S. National Center for Atmospheric Research; 2008. Report No.: NCAR/TN–475+STR.
  26. 26.↵
    National Centers for Environmental Prediction/National Weather Service/NOAA/U.S. Department of Commerce. NCEP FNL Operational Model Global Tropospheric Analyses, continuing from July 1999. Boulder, CO: Research Data Archive at the National Center for Atmospheric Research, Computational and Information Systems Laboratory; 2000.
  27. 27.↵
    Tipping E, Davies JAC, Henrys PA, Kirk GJD, Lilly A, Dragosits U, et al. Long-term increases in soil carbon due to ecosystem fertilization by atmospheric nitrogen deposition demonstrated by regional-scale modelling and observations. Scientific Reports. 2017;7:1890.
    OpenUrl
  28. 28.↵
    Russ TC, Cherrie MPC, Dibben C, Tomlinson S, Reis S, Dragosits U, et al. Life Course Air Pollution Exposure and Cognitive Decline: Modelled Historical Air Pollution Data and the Lothian Birth Cohort 1936. J Alzheimers Dis. 2021;79:1063–74.
    OpenUrl
  29. 29.↵
    A Vision of Britain through Time. Agricultural data. [Available from: https://www.visionofbritain.org.uk/.
  30. 30.↵
    Lin C, Heal MR, Vieno M, MacKenzie IA, Armstrong BG, Butland BK, et al. Spatiotemporal evaluation of EMEP4UK-WRF v4.3 atmospheric chemistry transport simulations of health-related metrics for NO2, O3, PM10, and PM2.5 for 2001–2010. Geosci Model Dev. 2017;10:1767-87.
    OpenUrl
  31. 31.↵
    Ge Y, Heal MR, Stevenson DS, Wind P, Vieno M. Evaluation of global EMEP MSC-W (rv4.34) WRF (v3.9.1.1) model surface concentrations and wet deposition of reactive N and S with measurements. Geoscientific Model Development. 2021;14:7021-46.
    OpenUrl
  32. 32.↵
    Daras K. Historical Address Geocoder (HAG-GIS) 1.0.0. 2015. http://lscs-projects.github.io/HAGGIS/
  33. 33.↵
    Huang Z. Geocoding 1939 address data for the SLSBC1936. 2020. Unpublished report.
  34. 34.↵
    Deary IJ, Whiteman MC, Starr JM, Whalley LJ, Fox HC. The Impact of Childhood Intelligence on Later Life: Following Up the Scottish Mental Surveys of 1932 and 1947. J Pers Soc Psychol. 2004;86:130–47.
    OpenUrlCrossRefPubMedWeb of Science
  35. 35.↵
    van Leeuwen M, Maas I, Miles A. HISCO: Historical International Standard Classification of Occupations. Leuven, Belgium: Leuven University Press; 2002.
  36. 36.↵
    Lambert PS, Zijdeman RL, Van Leeuwen MHD, Maas I, Prandy K. The Construction of HISCAM: A Stratification Scale Based on Social Interactions for Historical Comparative Research. Historical Methods: A Journal of Quantitative and Interdisciplinary History. 2013;46:77–89.
    OpenUrl
  37. 37.↵
    Scottish Council for Research in Education. The trend of Scottish intelligence: a comparison of the 1947 and 1932 surveys of the intelligence of eleven-year-old pupils. London, England: University of London Press; 1949.
  38. 37.↵
    Therneau T, Crowson C, E. A. Multi-state models and competing risks. 2022.
  39. 39.↵
    VanderWeele TJ, Vansteelandt S. Conceptual issues concerning mediation, interventions and composition. Stat Interface. 2009;2:457–68.
    OpenUrlCrossRefWeb of Science
  40. 39.↵
    Nahhas R. Introduction to Regression Methods for Public Health Using R2023.
  41. 41.↵
    Therneau T. A Package for Survival Analysis in R. R package version 3.5–5. 2023. https://CRAN.R-project.org/package=survival.
    OpenUrl
  42. 41.↵
    Yoshida K., Li Y., Mathur M. regmedint: Regression-Based Causal Mediation Analysis with Interaction and Effect Modification Terms. 2022. https://kaz-yos.github.io/regmedint/
  43. 43.↵
    van Buuren S, Groothuis-Oudshoorn K. mice: Multivariate Imputation by Chained Equations in R. Journal of Statistical Software. 2011;45:1–67.
    OpenUrl
  44. 43.↵
    Sjoberg DD., Fei T. tidycmprsk: Competing Risks Estimation. 2022. https://mskcc-epi-bio.github.io/tidycmprsk/
  45. 45.↵
    R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2022.
  46. 46.↵
    Elliott P, Shaddick G, Wakefield JC, de Hoogh C, Briggs DJ. Long-term associations of outdoor air pollution with mortality in Great Britain. Thorax. 2007;62:1088–94.
    OpenUrlAbstract/FREE Full Text
  47. 47.↵
    Strak M, Weinmayr G, Rodopoulou S, Chen J, de Hoogh K, Andersen ZJ, et al. Long term exposure to low level air pollution and mortality in eight European cohorts within the ELAPSE project: pooled analysis. BMJ. 2021;374:n1904.
  48. 48.↵
    World Health Organization. WHO global air quality guidelines: particulate matter (PM2.5 and PM10), ozone, nitrogen dioxide, sulfur dioxide and carbon monoxide. Geneva: World Health Organization; 2021.
  49. 49.↵
    Fitzmaurice C, Allen C, Barber RM, Barregard L, Bhutta ZA, Brenner H, et al. Global, Regional, and National Cancer Incidence, Mortality, Years of Life Lost, Years Lived With Disability, and Disability-Adjusted Life-years for 32 Cancer Groups, 1990 to 2015: A Systematic Analysis for the Global Burden of Disease Study. JAMA Oncol. 2017;3:524-48.
    OpenUrl
  50. 50.↵
    Shi L, Wu X, Danesh Yazdi M, Braun D, Abu Awad Y, Wei Y, et al. Long-term effects of PM2.5 on neurological disorders in the American Medicare population: a longitudinal cohort study. Lancet Planet Health. 2020;4:e557–e65.
    OpenUrlCrossRefPubMed
  51. 51.↵
    Cole-Hunter T, Zhang J, So R, Samoli E, Liu S, Chen J, et al. Long-term air pollution exposure and Parkinson’s disease mortality in a large pooled European cohort: An ELAPSE study. Environ Int. 2023;171:107667.
  52. 52.↵
    Ni Y, Loftus CT, Szpiro AA, Young MT, Hazlehurst MF, Murphy LE, et al. Associations of Pre– and Postnatal Air Pollution Exposures with Child Behavioral Problems and Cognitive Performance: A U.S. Multi-Cohort Study. Environ Health Perspect. 2022;130:067008.
  53. 53.↵
    Saenen ND, Martens DS, Neven KY, Alfano R, Bové H, Janssen BG, et al. Air pollution-induced placental alterations: an interplay of oxidative stress, epigenetics, and the aging phenotype? Clin Epigenetics. 2019;11:124.
  54. 54.↵
    Bettiol A, Gelain E, Milanesio E, Asta F, Rusconi F. The first 1000 days of life: traffic-related air pollution and development of wheezing and asthma in childhood. A systematic review of birth cohort studies. Environ Health. 2021;20:46.
    OpenUrl
  55. 55.↵
    Miri M, Nazarzadeh M, Alahabadi A, Ehrampoush MH, Rad A, Lotfi MH, et al. Air pollution and telomere length in adults: A systematic review and meta-analysis of observational studies. Environ Pollu. 2019;244:636–47.
    OpenUrl
  56. 56.↵
    1. Herceg Z,
    2. Ushijima T
    Kulis M, Esteller M. 2 – DNA Methylation and Cancer. In: Herceg Z, Ushijima T, editors. Advances in Genetics. 70: Academic Press; 2010. p. 27–56.
  57. 57.↵
    Marioni RE, Shah S, McRae AF, Chen BH, Colicino E, Harris SE, et al. DNA methylation age of blood predicts all-cause mortality in later life. Genome Biology. 2015;16:25.
  58. 58.↵
    Clemens T, Turner S, Dibben C. Maternal exposure to ambient air pollution and fetal growth in North-East Scotland: A population-based study using routine ultrasound scans. Environ Int. 2017;107:216–26.
    OpenUrl
  59. 59.↵
    Chiu Y-HM, Hsu H-HL, Coull BA, Bellinger DC, Kloog I, Schwartz J, et al. Prenatal particulate air pollution and neurodevelopment in urban children: Examining sensitive windows and sex-specific associations. Environ Int. 2016;87:56–65.
    OpenUrlCrossRefPubMed
  60. 60.↵
    Lopuszanska U, Samardakiewicz M. The Relationship Between Air Pollution and Cognitive Functions in Children and Adolescents: A Systematic Review. Cogn Behav Neurol. 2020;33:157–178.
    OpenUrl
  61. 61.↵
    Hedman AM, van Haren NE, Schnack HG, Kahn RS, Hulshoff Pol HE. Human brain changes across the life span: a review of 56 longitudinal magnetic resonance imaging studies. Hum Brain Mapp. 2012;33:1987–2002.
    OpenUrlCrossRefPubMedWeb of Science
  62. 62.↵
    Stiles J, Jernigan TL. The Basics of Brain Development. Neuropsychology Review. 2010;20:327–48.
    OpenUrlCrossRefPubMedWeb of Science
  63. 63.↵
    Sörberg Wallin A, Allebeck P, Gustafsson J-E, Hemmingsson T. Childhood IQ and mortality during 53 years’ follow-up of Swedish men and women. J Epidemiology Community Health. 2018;72:926–32.
    OpenUrlAbstract/FREE Full Text
  64. 64.↵
    Birth– and Death-Rates for 1936 in England. Nature. 1937;139:189–189.
    OpenUrl
  65. 65.↵
    deSouza PN, Dey S, Mwenda KM, Kim R, Subramanian SV, Kinney PL. Robust relationship between ambient air pollution and infant mortality in India. Sci Total Environ. 2022;815:152755.
  66. 66.↵
    Falkingham J, Sage J, Stone J, Vlachantoni A. Residential mobility across the life course: Continuity and change across three cohorts in Britain. Adv Life Course Res. 2016;30:111–23.
    OpenUrl
  67. 67.↵
    Carnell E, Vieno M, Vardoulakis S, Beck R, Heaviside C, Tomlinson S, et al. Modelling public health improvements as a result of air pollution control policies in the UK over four decades—1970 to 2010. Environ Res Lett. 2019;14:074001.
Back to top
Posted May 21, 2023.
Download PDF
Data/Code
Email
Early life PM2.5 exposure, childhood cognitive ability and mortality between age 11 and 86: A record-linkage life-course study from Scotland
Gergő Baranyi, Lee Williamson, Zhiqiang Feng, Sam Tomlinson, Massimo Vieno, Chris Dibben
medRxiv 2023.05.18.23289845; doi: https://doi.org/10.1101/2023.05.18.23289845
Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools

Subject Area