Urban Ambient Particle Metrics and Health: A Time-series Analysis : Epidemiology

Secondary Logo

Journal Logo

Air Pollution: Original Article

Urban Ambient Particle Metrics and Health

A Time-series Analysis

Atkinson, Richard W.a; Fuller, Gary W.b; Anderson, H. Rossa; Harrison, Roy M.c; Armstrong, Bend

Author Information
doi: 10.1097/EDE.0b013e3181debc88


There is convincing evidence that both short-term and long-term exposure to ambient particulate matter in the outdoor air is associated with mortality and morbidity.1 Much of this evidence relates to particle mass (PM) measurements such as black smoke, total suspended particles, and PM10—as well as fine particles that easily penetrate deep into the lung, remain suspended in the air for long periods and can be transported over long distances. The fine fraction of PM10, PM2.5, comes mainly from combustion sources. It is not clear to what extent health effects are caused by particles directly emitted from combustion processes (primary particles) or by those formed from complex chemical reactions in the atmosphere that may have occurred many miles away (secondary particles). Ultrafine particles (particles with a median diameter less than 0.1 μm) comprise the greatest number of particles but the least mass. It has been postulated that this fraction may also be important for health effects.2 However, the coarse fraction (particle mass with a median diameter between 2.5 and 10 μm [PM10–2.5], which includes particles from noncombustion sources such as wind-blown dust, cannot be discounted, as a number of studies report adverse associations.3 The physicochemical characteristics of particulate matter vary according to emission sources, secondary chemical reactions in the atmosphere, weather conditions and other factors. Other characteristics of PM that can affect toxicity include the metal content of the particles,4,5 their solubility6 and their reactivity.7

Current evidence suggests that a single component is unlikely to be responsible for particulate toxicity; different aspects of pollution may be relevant for different health outcomes and, furthermore, may affect human health over different time frames. In this study, we examined associations of various particle-size and composition metrics with a range of health outcomes and lags, to shed light on which aspects of particulate pollution might be relevant for different health outcomes and time domains of effect. We used time-series methods to investigate associations between daily deaths and hospital admissions in London and a range of particle metrics including PM10, PM2.5, individual chemical components (carbon, sulfate, nitrate, and chloride) and particle number concentrations. Recent PM measurements made using filter dynamic measurement system (FDMS) samplers were also available for comparison with both gravimetric and tapered-element-oscillating-microbalance (TEOM) measurements.


Details of all deaths in England and Wales were obtained from the Office for National Statistics. For the study period 1 January 2000 to 31 December 2005, we constructed daily counts of deaths for people who resided and died in London of any nonaccidental cause (ICD-10 Chapters A–R), respiratory (ICD-10 Chapter J), and cardiovascular causes (ICD-10 Chapter I). The Hospital Episode Statistics system routinely records details of all admissions to National Health Service hospitals in England. We used data extracted from this system to compile time series of daily counts of emergency admissions to hospital for respiratory causes (ICD-10 J) for the age groups 0–14 and 65+ years, and for all cardiovascular causes (ICD-10 I) for London for the same time period.

Daily concentrations for a range of particle metrics were obtained from a centrally located background monitoring station in London (North Kensington). Organic, elemental, and total carbon concentrations were measured using an R&P 5400 carbon monitor. Particle number concentrations were measured using a TSI 3022A condensation particle counter. PM10 24-hour filter samples (midnight to midnight), which were collected at 16.7 L per minute on membrane filters using Partisol 2025 instruments (Thermo Scientific, Waltham, MA) and analyzed by ion chromatography, provided daily chloride (Cl), nitrate (NO3), and sulfate (SO42−) concentrations. Gravimetric PM10 and PM2.5 were measured using a Partisol sampler according to the method described in EN12341.8 The gravimetric coarse fraction was calculated by subtraction.

TEOM particle mass measurements were obtained for 6 background and suburban locations in London for PM10 and 2 for PM2.5 from the Air Quality Archive9 and averaged to provide daily London-wide values. These data were adjusted for gravimetric equivalence by multiplying by a factor of 1.3.10 Missing values within the 2-pollutant time series were then imputed using a simple regression procedure.11

Daily mean PM10 concentrations were apportioned into primary and nonprimary sources using an apportionment model12–14 ie

PM10 = [primary PM10] + [nonprimary PM10]


and where [primary PM10] t is the primary PM10 concentration on day t, Ak is the ambient concentration ratio of primary PM10 to NOX derived annually for each year k from measurements throughout London,

is the annual mean NOX concentration at North Kensington during year k, NOXtn is the mean NOX concentration at site n on day t, and

is the annual mean NOX concentration at site n during year k (with n running from 0 to 3 covering 4 background NOX measurement sites in London).

Nonprimary PM10 is assumed to be a regional source with no spatial variation across London. The daily mean concentration of nonprimary PM10 is calculated from PM10 and NOX measurements at 7 long-term monitoring sites in London:

where [nonprimary PM10] t is the concentration of nonprimary PM10 in London on day t, PM10tm is the measured concentration of PM10 on day t at site m, Ak is the ambient concentration ratio of primary PM10 to NOX derived annually for each year k from measurements throughout London, and NOXtm is the mean NOX concentration at site m on day t where m runs from 0 to 6 covering 7 long-term monitoring sites in London.

[Nonprimary PM2.5] is calculated in the same way as [nonprimary PM10] but using Bk, the ambient concentration ratio of primary PM2.5 to NOX derived annually for each year k from measurements throughout London and daily mean measurements of PM2.5. The nonprimary PM10–2.5 was calculated by difference.

Daily measures of black smoke from 5 monitoring stations in Greater London for the period 2000–2005 were downloaded from the Air Quality Archive smoke and SO2 monitoring network.9

A PM10 measurement program using the FDMS has been operated by King's College London from 2003 in conjunction with London local authorities at several sites including North Kensington. The FDMS extends the TEOM method to produce a separate quantification of the mass concentration of semi-volatile PM (termed FDMS purge), nonvolatile PM (FDMS base), and total PM allowing for volatile loses.15

For each metric, daily 24-hour average concentrations were calculated subject to completeness criteria appropriate to the data recording frequency. Data recorded in 15-minute interval values (particle number concentration) were first aggregated into 1-hour average values subject to the criterion that 75% (ie, at least 3 of the 4 values within each hour) were available. Data recorded in hourly intervals were aggregated into daily 24-hour averages subject to the same 75% data availability criterion to ensure a representative spread of values across the day. Three-hour carbon data were averaged to obtain daily 24-hour averages subject to the availability of 6 or more values in each day. These summary procedures were in accordance with UK Local Air Quality Management guidance.16

Daily average temperature and dew point temperature (an indicator of humidity) for London (Holborn) were obtained from the British Atmospheric Data Centre website.17

Statistical Methods

We modeled the daily event counts using a Poisson log-linear time-series model, with month effects included as step functions (month strata), together with nonparametric temporal adjustment within month of year. This model formulation has been shown to be equivalent to the fixed-stratum form of the case-crossover time-stratified model, originally proposed by Maclure18 and developed by Lumley,19 when pollution exposures are derived from community monitors.20

The basic model specification for the regression analysis was therefore:

Yt ∼ Poisson(μt)

where Yt is the mortality/admission count on day t; mi is a binary indicator for each month in the time series (i= 2–72); sj is a penalized cubic regression spline of time within month of year (January–December), ie, c = 1–30, 1–31, or 1–28 (29) depending on month j; smooth functions gk(wt-k) and hk(dt-k) (cubic regression splines) of daily average temperature (wt-k) and dew point temperature (dt-k) measured on the same day and on the 3 previous days were included in the model; indicators for day of week (dow) and public holidays (bh) were also included; Pt-ln is a n-dimensional vector of pollutant concentrations on day t lagged by l days: n = 1 for single pollutant models and n = 2 for two-pollutant models.

The coefficients, γ, are interpreted as log relative risks (RR) per unit increase in the pollutant(s). For presentation and comparative purposes, the relative risks (RRs) and 95% confidence intervals (CIs) are expressed as percentage changes (100*RR-1) associated with interquartile range (IQR) increases in pollutant concentrations. Standard errors were scaled by the overdispersion parameter estimated from the model. Associations using concurrent exposure (lag 0) and lagged exposure up to 6 days were investigated. The R statistical package was used for all analyses.21


Particle-metric data (daily 24-hour averages) were not consistently available over the 6-year study period (1 January 2000 and 31 December 2005; 2192 days). Details are given in eTable 1 (https://links.lww.com/EDE/A386). Carbon data were available for only 618 days (39%) and were therefore not considered further. Particle number concentrations were available for 1194 days (69%) and concentrations of anions for 1160 days (78%). Other particle metrics were more complete.

Summary statistics for the health outcomes and particle measures are presented in Table 1. Time-series plots of these data are given in eFigures 1 and 2 (https://links.lww.com/EDE/A386). The median numbers of deaths per day from all causes, respiratory and cardiovascular diseases during the 6-year study period were 145, 22, and 54, respectively. Corresponding medians for hospital admissions were 153 for cardiovascular admissions and 63 and 35 for respiratory admissions (65+ and 0–14 years respectively). A number of the particle metrics exhibited a seasonal pattern, with increased concentrations in the colder months (most notably particle number concentration, black smoke, nitrate, chloride, and modeled primary PM10 (eFigure 2 [https://links.lww.com/EDE/A386]).

Summary Statistics for Daily Mortality, Hospital Admissions, and Particle Metrics for 1 January 2000 Through 31 December 2005

Seasonally adjusted correlation coefficients between the various parts of particle metrics are shown in Table 2. Particle number concentration was strongly correlated with modeled primary PM10 measurements. Nitrate and sulfate concentrations were correlated with both gravimetric and TEOM measures of PM10 and PM2.5, whereas chloride concentrations were only weakly correlated with other particle metrics.

Seasonally Adjusted Correlation Coefficients Between Pollutants (1 January 2000 Through 31 December 2005)

Regression Model Results

Results for particle number concentration, anions, black smoke, gravimetric PM10, PM2.5, and PM10–2.5 and modeled-source-apportioned PM concentrations are illustrated in Figures 1 (mortality) and 2 (admissions). A full set of results, including TEOM measurements at North Kensington and London-wide, are also given in tabular format in the eTables 2 and 3 (https://links.lww.com/EDE/A386).

Associations of particle metrics, lagged 0–6 days, with daily mortality from all causes (first panel), respiratory disease (second panel), and cardiovascular disease (third panel). All measurements are from North Kensington, except for black smoke, which was measured at a number of locations across London.
Associations of particle metrics, lagged 0–6 days, with daily hospital admissions for all causes (first panel), respiratory disease (second panel) and cardiovascular disease (third panel). All measurements are from North Kensington, except for black smoke, which was measured at a number of locations across London.

Particle number concentration lagged 1 day was positively associated with daily mortality from all causes and from cardiovascular and respiratory causes: increases of 1.4% (95% CI = 0.5% to 2.4%), 2.2% (0.6% to 3.8%), and 2.3% (−0.1% to 4.8%) per 10,166/cm3 increases in particle number concentration, respectively. Particle number concentration associations with hospital admissions were generally null with little consistency with mortality results other for respiratory admissions, for which positive associations were indicated for longer lag times.

We found some evidence of associations between concentrations of anions and respiratory hospital admissions, with less-convincing evidence for cardiovascular admissions and mortality. The strongest evidence for adverse associations was observed for the relationships of nitrate and sulfate with respiratory admissions in the 65+ age group. This finding was, however, inconsistent with the results for respiratory mortality. We found little evidence for associations of gravimetric measurements of PM10, PM2.5 or the coarse fraction (PM10–2.5) with all-cause mortality and cardiovascular diseases. However, increases in gravimetric PM10 and PM2.5 concentrations were associated with increases in a number of respiratory outcomes. For example, IQR increases in concentrations of PM10 and PM2.5 were associated with increases of 1.1% (−0.2% to 2.4%; lag 1 day) and 1.0% (−0.1% to 2.0%; lag 2 days) for respiratory deaths and 1.2% (0.4% to 2.0%; lag 1 day) and 1.3% (0.6% to 2.0%; lag 3 days) for respiratory admissions (65+), respectively. PM10 and PM2.5 were also associated with increases in respiratory admissions in children. We found little evidence for associations between these gravimetric measures and cardiovascular outcomes. Respiratory outcomes were associated with increases in the modeled nonprimary fraction of the particle mass and for admissions only, with the primary fraction at longer lag times.

Figure 3 illustrates results for all particle metrics grouped by disease outcome, rather than pollutant, for cardiovascular outcomes (lag 1) and respiratory outcomes (lag 2). These alternative views of the findings suggest a specific association between particle number concentration and cardiovascular endpoints and between PM10 and PM2.5 and respiratory endpoints. Figure 3 also illustrates the results for TEOM measurements of PM10 (made at North Kensington and averaged across London) and PM2.5 (average of 2 background locations).

Associations of particle metric lag 1 day with cardiovascular and respiratory mortality and admissions. A, Cardiovascular mortality; B, cardiovascular admissions; C, Respiratory mortality; and D, respiratory admissions. All measurements are from North Kensington, except black smoke, which was measured at a number of locations across London.

Sensitivity Analyses

Selected two-pollutant models were used to investigate further the associations observed in single-pollutant models. The interrelationships between the particle metrics and their correlations (Table 2) were used to guide model selection. Full results are presented in the on-line supplement (eTables 4–6 [https://links.lww.com/EDE/A386]). The main findings from these further analyses were: (1) associations between particle number concentration and cardiovascular admissions and deaths were independent of other particle metrics (black smoke, TEOM PM10, and PM2.5), (2) associations between particle metrics and respiratory mortality were driven by the nonprimary fine particulate component (TEOM PM2.5), (3) 2-pollutant models for respiratory admissions (0–14 and 65+ years) including nitrate and TEOM PM2.5 indicated that the associations with nitrate were independent of PM2.5 but not vice versa, and (4) for respiratory admissions in children, the observed particle number concentration was independent (during the summer months) of PM2.5 concentrations.

Comparison of PM Measurements From FDMS, TEOM, and Gravimetric Samplers

A separate set of analyses was conducted to compare the relative health effects (on respiratory outcomes) of PM measurements made using TEOM, FDMS, and gravimetric measurement methods—using only days on which all 3 PM metrics were measured (444 days). The results suggest comparable effect estimates for PM10 measured using filter-dynamic-measurement-system, TEOM and gravimetric methods with confidence intervals that overlap substantially (Table 3).

Comparison of Results for PM10 Using FDMS Measurements, TEOM and Gravimetric Mass Measurements (lag 2 days) Made at North Kensington and Respiratory Mortality and Admissions


We investigated associations of 15 particle metrics recorded on the same day as the health event (lag 0) and up to 2 days before (lag 1 and 2) with 6 health outcomes—daily mortality from all causes and from cardiovascular and respiratory causes, and hospital admissions for cardiovascular causes and for respiratory causes (ages 0–14 and 65+).

Particle number concentrations were associated with daily mortality and admissions, particularly for cardiovascular diseases, whereas several particle metrics were associated with adverse respiratory outcomes. The evidence suggests that the nonprimary fine (PM2.5) component is most important for respiratory outcomes, although other metrics may also be relevant. Nitrate and sulfate were positively associated with increases in numbers of respiratory admissions in both age groups studied.

The association between particle number concentration and both deaths and admissions for cardiovascular diseases were observed only at lag 1, with little evidence for associations at other lags. One must be cautious not to overinterpret this finding, because we had no a priori reason to suspect this very specific association. There are only a small number of published studies that have investigated particle number concentration2,22–26 and only the results reported by Stolzel et al,2 and by Wichmann26 are directly comparable with our own. Stolzel et al2 report associations between particle number concentration and total, cardiovascular and respiratory mortality of −0.3% (−2.8% to 2.3%) to 2.9% (0.3% to 5.5%) per 9748 n/cm3 increase, depending on the lag for total mortality; Wichman et al26 report 5.1% (−1.0% to 11.5%) and 4.8% (−4.4% to 14.9%) per 12,690 n/cm3 increments for cardiovascular and respiratory mortality. In London the main component contributing to particle number count is likely to be nucleation-mode particles from diesel traffic.27 These are too small to analyze other than by very sophisticated techniques and their behavior suggests that they are predominantly lubricating oil with a core of sulfate or trace metals.28 Consequently, in terms of the components considered in this study, these very small and numerous particles would be mainly organic carbon, a component linked with adverse health effects. Unfortunately, we were unable to assess the associations between organic and elemental carbon and cardiovascular outcomes due to the lack of valid carbon data.

We found that it was the nonprimary PM10 and PM2.5 metrics (those not associated with NOX sources), rather than primary PM10, that were associated with increases in respiratory mortality and admissions (for people age 65+). These associations tended to be stronger at longer rather than shorter lag periods (2–5 days). The question of whether the fine or coarse fraction of the nonprimary PM10 is driving the health effects is less clear, as the answer appears to depend upon the lag. Nonprimary particles in London comprise mainly sodium and ammonium nitrate, sulfate, chloride, and organic carbon. We were unable to investigate carbon due to poor data availability, and we did not find evidence for an association between anions and respiratory mortality. We did, however, find evidence for an association of concentrations of nitrate and sulfate with respiratory admissions. The respiratory toxicity of nitrate salts has not been well studied, although a small number of studies from the United States show associations with nitrate.29,30 In the United Kingdom, nitrate tends to be very seasonal, with high concentrations associated with continental air masses, low temperatures and high relative humidity. Nitrate episodes are particularly likely to occur between February and April. Nitrate may be acting as a surrogate for secondary organic carbon, as the 2 show a very similar seasonal cycle in the United Kingdom.31 In sensitivity analyses of respiratory admissions in children, a 2-pollutant model including nitrate and PM2.5 showed that the association with nitrate, rather than with PM2.5, retained its size and statistical significance.

Chloride was associated with increases in respiratory mortality but not respiratory admissions. This lack of consistency between the findings for the 2 outcomes leads us to be cautious in interpreting these associations as causal. The chloride association was found to be stronger during the summer months. One possible explanation is that these associations simply reflect an association between ozone and respiratory mortality. Results from 2-pollutant models for chloride and ozone (measured at North Kensington)—and using all days with chloride and ozone values—revealed that the chloride association in the summer months was weakened by the inclusion of ozone in the model.

Our findings suggest a degree of specificity may exist for the associations between the components of particle metrics and cause-specific health effects. They suggest the need for a regulatory framework based upon a range or mixture of pollutants targeting specific diseases. However, it is also interesting to note the continued relevance to human health of the PM10 and PM2.5 metrics that are not source-specific. These metrics may simply capture or encapsulate a range of effects from a number of specific pollutants. In the absence of further evidence to indicate specific sources and their respective health effects, the continued regulation and monitoring of these general mass metrics is of great importance.

In this study, PM10 concentrations at North Kensington were measured using 3 different methods: TEOM, FDMS, and gravimetric. Although all 3 methods measure mass concentration, they differ in their sensitivity to particle-bound water and volatile particulate matter (ammonium nitrate and volatile organics). We found that the differences in effect estimates of the 3 measures were well within the range expected due to random variation (Table 3). This is not surprising given the close temporal correlations among the 3 metrics. It therefore seems unlikely that one measure would be favored over the other as an indicator of the toxic component of particles.

Before considering causal interpretations of the associations observed, we should consider whether air-pollution effects are more likely than other explanations. Here, as generally in epidemiological studies, we need to consider chance, confounding, and information bias (exposure and outcome error).

The large number of analyses, although not independent, would lead us to expect some strong associations purely by chance. The patterns of associations were nonrandom, although this would be the case even with chance associations, given the dependence among the exposures. However, the preponderance of positive over negative associations suggests some mechanism other than chance. Despite considerable control for confounding, following the usual approach in time-series studies, residual confounding cannot be excluded. This applies in particular to confounding due to weather, which is a known determinant of pollution and of health beyond the extent reflected in the models. Finally, error in exposure measurement is likely to be a more important limitation here than error in outcome (especially for mortality). As shown by Zeger,32 what matters is how well the exposure series matches the mean daily exposures over the city as a whole. The availability of just one monitor for many of the metrics considered here makes this particularly an issue in this study. Strong temporal correlations between PM2.5 monitored at different geographical locations across London33 are reassuring in this respect and, to some extent, for particle number concentration also.34 It seems likely that the error would be independent of the outcome (not different on high- than on low-mortality days) and so would cause bias toward the null—thus not an explanation for positive associations.35 However, if error was associated with weather (ie, if weather determined the extent to which the North Kensington monitor represented London overall), this could exacerbate residual confounding by weather. Thus, overall, chance or confounding are possible explanations for the associations observed, but not so likely as to rule out possible causation.

The question of whether these relationships have biologic plausibility also needs consideration. Studies both in humans and in animal models have shown that particles have pathogenic effects through inflammation.36 PM pollution has long been recognized as a trigger for exacerbations of chronic obstructive pulmonary disease (COPD)—increasing oxidative stress and aggravating background inflammation in COPD, leading to acute exacerbations.37 Short-term-exposure studies have also observed PM-related exacerbation of respiratory symptoms and transient declines in lung function.38 For cardiovascular endpoints, pulmonary inflammation is thought to lead to cardiovascular effects via a number of pathways including systemic inflammation, imbalance in coagulation factors resulting in thrombogenesis and interference in the control of heart rhythm that could lead to fatal dysrhythmia.39

In February 2003 a congestion-charging policy, operating Monday–Friday, 7 am–6 pm, was introduced in central London. This resulted in an 18% reduction in traffic volumes and a 30% reduction in traffic congestion in the first year. Our study investigated associations between particles and health at the time of this phased intervention, and we must therefore consider the possible impact of the implementation of this policy on our study. The impact of this policy and the associated traffic management changes on pollution concentrations have been investigated.40 Atkinson et al observed reductions in concentrations of background levels of PM10 within the charging zone relative to areas of London where the policy was not in operation. If the changes in PM10 concentrations observed at the limited number of monitoring stations within the zone reflect a reduction in traffic-related PM, there may be corresponding changes in health effects. To date this has not been studied. However, for 2 reasons, the implications of the policy for our findings are likely to be relatively small. First, the seasonal and long-term trends in mortality and morbidity are adjusted for in the models and so long-term reductions in number of events are accounted for in the analysis. Second, the evidence for health effects from PM is thought to be linear.

This study has revealed some of the difficulties in attempting this type of investigation. First and foremost, it has illustrated the importance of obtaining complete time-series data for particle metrics. Lack of sufficient data in this study meant that it was not possible to analyze carbon measurements. Missing data also led to less statistical power (fewer days to analyze) and problems of interpretation, in that the analysis was restricted to days on which the particle metrics were available. Although there is no obvious reason to suspect that the missing data introduced bias, this assumption nevertheless limits our confidence in the results.

Much of the particle metric data were from a single, centrally located monitor. While the strong temporal correlations between some pollutants measured at different locations across a large urban area suggest that a time-series analysis is appropriate, this is not the case for all pollutants. Therefore additional time series from monitors spread geographically across the city would be extremely important in any further investigation. Until then, our results provide the best information available on the associations between a range of specific particle metrics and health outcomes and they, at least, suggest hypotheses that can be further evaluated as more complete and comprehensive data become available. For the meantime, PM mass measures remain important metrics for monitoring and regulation in relation to public health policy.


1.Pope CA III, Dockery DW. Health effects of fine particulate air pollution: Lines that connect. J Air Waste Manag Assoc. 2006;56:709–742.
2.Stolzel M, Breitner S, Cyrys J, et al. Daily mortality and particulate matter in different size classes in Erfurt, Germany. J Expo Sci Environ Epidemiol. 2007;17:458–467.
3.Brunekreef B, Forsberg B. Epidemiological evidence of effects of coarse airborne particles on health. Eur Respir J. 2005;26:309–318.
4.Zhang Q, Kusaka Y, Sato K, Nakakuki K, Kohyama N, Donaldson K. Differences in the extent of inflammation caused by intratracheal exposure to three ultrafine metals: role of free radicals. J Toxicol Environ Health A. 1998;53:423–438.
5.Lippmann M, Ito K, Hwang JS, Maciejczyk, Chen LC. Cardiovascular effects of nickel in ambient air. Environ Health Perspect. 2006;114:1662–1669.
6.Imrich A, Ning Y, Kobzik L. Insoluble components of concentrated air particles mediate alveolar macrophage responses in vitro. Toxicol Appl Pharmacol. 2000;167:140–150.
7.Risom L, Moller P, Loft S. Oxidative stress-induced DNA damage by particulate air pollution. Mutat Res. 2005;592:119–137.
8.Centre for European Normalization (CEN) Technical Committee CEN/TC264. EN 12341 Air quality-determination of the PM10 fraction of suspended particulate matter–reference method and field test procedure to demonstrate reference equivalence of measurement methods. Brussels, Belgium; CEN; 1998.
9.AEA Technology. Air Quality Archive. Available at: http://www.airquality.co.uk/archive/index.php. Accessed January 2007.
10.Department for the Environment, Transport and the Regions (DETR). Assistance with the review and assessment of PM10 concentration in relation to the proposed EU stage 1 limit values. London: HMSO; 1999.
11.Katsouyanni K, Schwartz J, Spix C, et al. Short term effects of air pollution on health: A European approach using epidemiologic time series data: the APHEA protocol. J Epidemiol Community Health. 1996;50(suppl 1):S12–S18.
12.Fuller GW, Carslaw DC, Lodge HW. An empirical approach for the prediction of daily mean PM10 concentrations. Atmos Environ. 2002;36:1431–1441.
13.Fuller GW, Green D. Evidence for increasing primary PM10 in London. Atmos Environ. 2006;40:6134–6145.
14.Fuller GW. Source apportionment of PM10 concentrations in south east England [PhD thesis]. London: King's College London; 2009.
15.Green D, Fuller GW, Baker T. The King's College London Volatile Correction Model for PM10–development, testing and application. London: King's College; 2007.
16.Department for the Environment, Food and Rural Affairs (Defra). Part IV of the Environment Act 1995, local air quality management. Technical guidance, LAQM. TG(03). London: Defra; 2003.
17.British Atmospheric Data Centre (BADC). Available at: http:www.badc.nerc.ac.uk/data. Accessed January 2007.
18.Maclure M. The case-crossover design: A method for studying transient effects on the risk of acute events. Am J Epidemiol. 1991;133:144–153.
19.Lumley L, Levy D. Bias in the case-crossover design: Implications for studies of air pollution. Environmetrics. 2000;11:689–704.
20.Farrington CP, Whitaker HJ. Semiparametric analysis of case series data. J R Stat Soc Ser C Appl Stat. 2006;55:553–594.
21.R Development Core Team (2007). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3–900051–07–0. Available at: http://www.R-project.org. Accessed January 2007.
22.Andersen ZJ, Wahlin P, Raaschou-Nielsen O, Ketzel M, Scheike T, Loft S. Size distribution and total number concentration of ultrafine and accumulation mode particles and hospital admissions in children and the elderly in Copenhagen, Denmark. Occup Environ Med. 2008;65:458–466.
23.Forastiere F, Stafoggia M, Picciotto S, et al. A case-crossover analysis of out-of-hospital coronary deaths and air pollution in Rome, Italy. Am J Resp Crit Care Med. 2005;172:1549–1555.
24.Lanki T, Pekkanen J, Aalto P, et al. Associations of traffic related air pollutants with hospitalization for first acute myocardial infarction: the HEAPSS study. Occup Environ Med. 2006;63:844–851.
25.von Klot S, Peters A, Aalto P, et al. Ambient air pollution is associated with increased risk of hospital cardiac readmissions of myocardial infarction survivors in five European cities. Circulation. 2005;112:3073–3079.
26.Wichmann HE, Spix C, Tuch T, et al. Daily mortality and fine and ultrafine particles in Erfurt, Germany. Part I, role of particle number and particle mass. Report 98. Cambridge, MA: Health Effects Institute; 2000.
27.Shi FJP, Evans DE, Khan AA, Harrison RM. Sources and concentrations of nanoparticles (<10 nm diameter) in the urban atmosphere. Atmos Environ. 2001;35:1193–1202.
28.Harrison RM, Yin J. Particulate matter in the atmosphere: which particle properties are important for its effects on health? Sci Total Environ. 2000;249:85–101.
29.Fairley D. 2003. Mortality and air pollution for Santa Clara County, California, 1989–1996. In: Revised Analyses of Time-Series Studies of Air Pollution and Health. Special Report. Boston: Health Effects Institute, 97–106. Available at: http://www.healtheffects.org/Pubs/TimeSeries.pdf. Accessed October 16, 2008.
30.Ostro B, Feng WY, Broadwin R, Green S, Lipsett M. The effects of components of fine particulate air pollution on mortality in California: results from CALFINE. Environ Health Perspect. 2007;115:13–19.
31.Harrison RM, Yin J. Sources and processes affecting carbonaceous aerosol in central England. Atmos Environ. 2008;42:1413–1423.
32.Zeger SL, Thomas D, Dominici F, et al. Exposure measurement error in time-series studies of air pollution: concepts and consequences. Environ Health Perspect. 2000;108:419–426.
33.Puustinen A, Hameri K, Pekkanen J, et al. Spatial variation of particle number and mass over four European cities. Atmos Environ. 2007;41:6622–6636.
34.Cyrys J, Pitz P, Heinrich J, Wichmann HE, Peters A. Spatial and temporal variation of particle number concentration in Ausburg, Germany. Sci Total Environ. 2008;401:168–175.
35.Armstrong BG. Effect of measurement error on epidemiological studies of environmental and occupational exposures. Occup Environ Med. 1998;55:651–656.
36.Donaldson K, Tran CL. Inflammation caused by particles and fibres. Inhal Toxicol. 2002;14:5–27.
37.MacNee W, Donaldson K. Mechanism of lung injury caused by PM10 and ultrafine particles with special reference to COPD. Eur Respir J Suppl. 2003;21:47s–51s.
38.Hoek G, Dockery DW, Pope CA III, Neas L, Roemer W, Brunekreef B. Association between PM10 and decrements in peak expiratory flow rates in children: Reanalysis of data from five panel studies. Eur Respir J. 1998;11:1307–1311.
39.Donaldson K, Mills N, MacNee W, Robinson S, Newby D. Role of inflammation in cardiopulmonary health effects of PM. Toxicol Appl Pharmacol. 2005;207(suppl 2):483–488.
40.Atkinson RW, Barratt B, Armstrong B, et al. The impact of the congestion charging scheme on ambient air pollution concentrations in London. Atmos Environ. 2009;43:5493–5500.

Supplemental Digital Content

© 2010 Lippincott Williams & Wilkins, Inc.