Abstract
Background
Short-term exposure to NO2 has been associated with adverse health effects and there is increasing concern that NO2 is causally related to health effects, not merely a marker of traffic-generated pollution. No comprehensive meta-analysis of the time-series evidence on NO2 has been published since 2007.
Objective
To quantitatively assess the evidence from epidemiological time-series studies published worldwide to determine whether and to what extent short-term exposure to NO2 is associated with increased numbers of daily deaths and hospital admissions.
Design
We conducted a quantitative systematic review of 204 time-series studies of NO2 and daily mortality and hospital admissions for several diagnoses and ages, which were indexed in three bibliographic databases up to May 2011. We calculated random-effects estimates by different geographic regions and globally, and also tested for heterogeneity and small study bias.
Results
Sufficient estimates for meta-analysis were available for 43 cause-specific and age-specific combinations of mortality or hospital admissions (25 for 24 h NO2 and 18 of the same combinations for 1 h measures). For the all-age group, a 10 µg/m3 increase in 24 h NO2 was associated with increases in all-cause, cardiovascular and respiratory mortality (0.71% (95% CI 0.43% to 1.00%), 0.88% (0.63% to 1.13%) and 1.09% (0.75% to 1.42%), respectively), and with hospital admissions for respiratory (0.57% (0.33% to 0.82%)) and cardiovascular (0.66% (0.32% to 1.01%)) diseases. Evidence of heterogeneity between geographical region-specific estimates was identified in more than half of the combinations analysed.
Conclusions
Our review provides clear evidence of health effects associated with short-term exposure to NO2 although further work is required to understand reasons for the regional heterogeneity observed. The growing literature, incorporating large multicentre studies and new evidence from less well-studied regions of the world, supports further quantitative review to assess the independence of NO2 health effects from other air pollutants.
Keywords: EPIDEMIOLOGY, PUBLIC HEALTH
Strengths and limitations of this study.
This is, to date, the most comprehensive, quantitative systematic review of the large time-series literature on NO2 published worldwide to provide evidence of associations between short-term exposure to NO2 and mortality and hospital admissions for a wide range of respiratory and cardiovascular diagnoses.
It reports meta-analytical estimates both globally and for different geographical regions, as well as an assessment of heterogeneity between the region-specific estimates.
The protocol-led approach to the identification of studies and estimates for use in meta-analysis minimised selection bias at each stage of the review.
Meta-analysis was limited to studies which provided effect estimates in numerical, rather than graphical, form along with sufficient quantitative data to enable standardisation of estimates.
Further work is needed to understand reasons for the regional heterogeneity observed and to quantitatively assess the independence of NO2 health effects from other air pollutants.
Introduction
Epidemiological time-series studies have found associations between daily concentrations of ambient nitrogen dioxide (NO2) and daily mortality and morbidity outcomes.1–7 Whether these associations reflect adverse health effects of NO2 has been a matter of debate.1 The uncertainty about causality arises largely because of the close correlations between NO2 and other air pollutants in the mixture of traffic-related pollutants, some of which are more plausible toxicants. This has led to a view that NO2 may be a surrogate for fine/ultrafine particulate matter (PM) produced by motor vehicles and other combustion sources. Despite this uncertainty, recent narrative reviews of the evidence linking NO2 with health effects concluded that the epidemiological and toxicological evidence regarding the pollutant has sufficiently strengthened the likelihood of a causal relationship.1 3 These reviews did not undertake any quantitative assessment of the recent time-series evidence on NO2, but instead relied on existing meta-analyses, with the last comprehensive one based on peer-reviewed papers published up to the beginning of 2006.6 Since 2006, the time-series evidence on NO2 has increased substantially, providing new data from single-city studies and from influential multicity studies from less well-studied locations, for example, Asia,8 9 and from regions previously examined, for example, Europe and Canada.10 11 The new data on NO2 will facilitate the calculation of more robust coefficients for a range of age-specific and disease-specific end points and the examination of regional variation and heterogeneity. Such quantitative analyses for NO2 are lacking but are needed for both hazard and health impact assessments.
We conducted a systematic, quantitative review of the time-series studies of daily NO2 and daily mortality and hospital admissions indexed in bibliographic databases up to May 2011. In this paper, we focus on single pollutant model results to answer the question of whether the evidence from epidemiological time-series studies suggests adverse associations between short-term exposure to NO2 and increased numbers of daily deaths and emergency admissions to hospital. We calculated meta-analytic estimates for mortality and hospital admissions for several diagnoses and age groups, by different geographic regions, as well as globally. Whether the associations identified in this review change on adjustment for the effects of co-pollutants is a different question which requires analysis of a subset of the studies presented in this review, that is, studies reporting estimates of NO2 from single-pollutant and multipollutant models. A similar review and meta-analysis has been completed for the subset of studies that reported two-pollutant model estimates of NO2 adjusted for PM. We intend to report those findings in a separate paper.
Methods
The full methodology and a priori protocols governing the identification of studies and effect estimates for our systematic review have been described previously, but a synopsis is provided below.6 12
Identification of studies for review
Bibliographic databases were searched to identify peer-reviewed time-series papers on NO2 and daily mortality and hospital admissions indexed up to May 2011. The search criteria used are shown in the online supplementary material. Papers suitable for inclusion in the review were identified by scanning titles and abstracts to remove irrelevant citations and then reviewing full papers of remaining citations. The following inclusion criteria were applied: papers must (1) have had a minimum of 1 year of data; (2) be based on the general population; (3) have controlled for important confounding factors, including season and meteorological factors; (4) have reported regression estimates and other quantitative data for analysis.
Data extraction and coding
Data from each relevant study were entered into a Microsoft Access database (Microsoft Office 2010, Microsoft Corporation). These included details of each publication (authors, journal name, paper title, year published, etc) and data required for calculating standardised effect estimates expressed as the percentage change (and 95% CI) in the mean number of daily events associated with a 10 µg/m3 increase in NO2 concentrations. We entered all outcomes, diagnoses (International Classification of Diseases (ICD) codes), ages and NO2 concentrations for each reported estimate. We subsequently classified all reported outcomes, diseases and ages into a smaller number of diagnostic categories and age groups to enable meta-analysis of estimates for specific outcome/disease/age combinations.6
Time-series studies often report results for several different time lags (in days) between exposure and health events and vary in the lag for the reported results for outcome/disease/age combinations. To avoid selection bias, we developed an a priori protocol for identifying the principal lag for each outcome/disease/age combination for use in our review. This was the lag highlighted by the author or stated a priori, and if this was not clear, because several lagged estimates were reported, we chose (1) the lag with the highest statistical significance, regardless of the estimate being positive or negative, or (2) the lag with the largest estimate, again, irrespective of its direction.6 If only results from cumulative or distributed lag models, that is, lags averaged over several days, were reported in a study, these were used. Further processing of data included the classification of each study into a geographical region, that is, the WHO region, in which the study was conducted (see online supplementary material).
Statistical analyses
We applied an a priori protocol for the selection of estimates for meta-analysis to avoid selection bias and duplication of studies from the same population. We gave priority to estimates from multicity studies over estimates from single-city studies and the results from a city appeared only once in a meta-analysis. The protocol also gave priority to the latest publications, the most recent study time periods, and did not take account of estimates being positive or negative (see online supplementary material). We only analysed all-year, single-pollutant model estimates.
Meta-analysis was conducted when ≥4 estimates were available for an outcome/disease/age combination—including where a multicity estimate was available—and summary estimates were calculated using a random-effects model.13 We used a staged approach to meta-analysis, with single-city estimates pooled within WHO region prior to the pooled single-city and selected multicity estimates being pooled to produce a global estimate and WHO region-specific summary estimates. Heterogeneity between WHO region summary estimates was assessed using the I2 statistic which describes the proportion of the variation in effect estimates attributable to between-study variation rather than chance.14 Where estimates permitted, meta-analyses were conducted for both 24 and 1 h average NO2. Finally, we investigated our selected single-city estimates and our pooled single-city and selected multicity estimates for evidence of small study bias using the methods of Begg and Berlin15 and Egger et al.16 We used the trim and fill technique to adjust the global random-effects estimate for small study bias where it was detected.17 This technique estimates and adjusts for the numbers and results of unreported studies. All analyses were conducted in STATA (STATA/SE V.11. StataCorp, Texas, USA).
Results
Two hundred and four time-series studies of NO2 and hospital admissions and mortality indexed in bibliographic databases up to May 2011 were available for review: a list of references is provided in the online supplementary material. Table 1 summarises the numbers of studies stratified by outcome, broad disease categories, WHO region, single-city and multicity study designs, and by averaging time (24 and 1 h). The majority of studies were conducted in Europe, North America and the WHO Western Pacific region and most used 24 h average concentrations of NO2. There were 123 and 94 studies of mortality and hospital admissions, respectively, and 42 studies used a multicity design.
Table 1.
Summary of time-series studies of NO2 and mortality and hospital admissions available for review
| Total |
Multicity study |
Single-city study |
||||
|---|---|---|---|---|---|---|
| Outcome | Mortality* | Hospital admission | Mortality | Hospital admission | Mortality | Hospital admission |
| Total | 123 | 94 | 25 | 21 | 98 | 73 |
| Disease† | ||||||
| Respiratory | 70 | 68 | 16 | 12 | 54 | 56 |
| Cardiovascular | 84 | 52 | 18 | 14 | 66 | 38 |
| All-cause | 101 | 4 | 21 | 0 | 80 | 4 |
| WHO region | ||||||
| American A | 24 | 23 | 6 | 5 | 18 | 18 |
| European A | 56 | 41 | 16 | 11 | 40 | 30 |
| Western Pacific B | 31 | 17 | 2 | 2 | 29 | 15 |
| American B | 7 | 5 | 0 | 0 | 7 | 5 |
| Western Pacific A | 5 | 7 | 1 | 3 | 4 | 4 |
| South East Asia B | 2 | 0 | 2 | 0 | 0 | 0 |
| Averaging time | ||||||
| 24 h | 107 | 82 | 21 | 17 | 86 | 65 |
| Maximum 1 h | 27 | 25 | 5 | 10 | 22 | 15 |
*One multicity study on mortality used multiple WHO regions, that is, South East Asia B and Western Pacific B.
†Respiratory includes all respiratory diseases, asthma, COPD (including asthma), lower respiratory infections and upper respiratory diseases; cardiovascular includes all cardiovascular diseases, cardiac disease, heart failure, ischaemic heart disease, dysrhythmia and stroke.
COPD, chronic obstructive pulmonary disease.
Mortality
Sufficient estimates for meta-analysis were available for seven combinations of all-age, all-cause and cause-specific mortality and 24 h NO2, four of which also provided sufficient estimates for 1 h NO2 for meta-analysis.
Summary estimates for all-cause and cause-specific mortality are given in figure 1, and for all-cause, all cardiovascular and all respiratory causes, stratified by WHO region in table 2. Estimates used in the meta-analyses for mortality are presented in forest plots in the online supplementary figures S1–S10. A 10 µg/m3 increase in 24 h NO2 was associated with a 0.71% (95% CI 0.43% to 1.00%) increase in the risk of death from all causes. This summary estimate is based on 31 single-city and multicity estimates selected for meta-analysis from the 84 identified in the review. Positive associations for all cardiovascular (0.88% (95% CI 0.63% to 1.13%)) and all respiratory (1.09% (95% CI 0.75% to 1.42%)) mortality were also observed, and though based on fewer estimates, we found positive summary estimates for mortality from other causes, that is, cardiac disease, ischaemic heart disease, stroke and chronic obstructive pulmonary disease (COPD) including asthma (figure 1).
Figure 1.
Random-effects summary estimates (95% CIs) for all-cause and cause-specific mortality in all ages per 10 µg/m3 increase in 24 h NO2 (COPD, chronic obstructive pulmonary disease).
Table 2.
Meta-analysis results for mortality in all ages associated with a 10 µg/m3 increase in NO2
| WHO region | NO2 24 h |
NO2 1 h |
||||||
|---|---|---|---|---|---|---|---|---|
| All SC/MC* | Selected SC/MC† | Random effects (95% CI)‡ | I2 (%)§ | All SC/MC* | Selected SC/MC† | Random effects (95% CI)‡ | I2 (%)§ | |
| All-cause, all ages | ||||||||
| Overall¶ | 72/12 | 5/6 | 0.71 (0.43 to 1.00) | 61.8 | 20/3 | 3/2 | 0.24 (0.00 to 0.49) | 88.2 |
| AMR A | 11/5 | 1/2 | 0.32 (−0.01 to 0.67) | 2/0 | 2/0 | 0.76 (−0.05 to 1.58) | ||
| AMR B | 1/0 | 1/0 | 0.59 (−0.26 to 1.45) | 2/0 | 2/0 | −0.06 (−0.13, 0.01) | ||
| EUR A | 40/6 | 14/3 | 0.90 (0.45 to 1.35) | 14/2 | 3/1 | 0.25 (0.10 to 0.40) | ||
| WPR A | 3/0 | 3/0 | 0.66 (−0.10 to 1.44) | 2/1 | 0/1 | 0.58 (0.21 to 0.94) | ||
| WPR B | 17/0 | 6/0 | 0.56 (0.18 to 0.94) | – | – | – | ||
| Multiregion** | 0/1 | 0/1 | 1.23 (0.84 to 1.62) | – | – | – | ||
| All cardiovascular, all ages | ||||||||
| Overall¶ | 54/8 | 5/5 | 0.88 (0.63 to 1.13) | 72.7 | 13/2 | 1/2 | 0.85 (0.20 to 1.50) | 0 |
| AMR A | 5/1 | 4/1 | 0.53 (0.38 to 0.67) | – | – | – | ||
| AMR B | 1/0 | 1/0 | 0.73 (−0.87 to 2.36) | – | – | – | ||
| EUR A | 30/6 | 10/3 | 1.03 (0.70 to 1.36) | 12/1 | 2/1 | 1.02 (−0.44 to 2.49) | ||
| WPR A | 1/0 | 1/0 | 0.67 (−0.24 to 1.60) | 1/1 | 0/1 | 0.94 (0.42 to 1.47) | ||
| WPR B | 17/0 | 7/0 | 0.91 (0.52 to 1.31) | – | – | – | ||
| Multiregion** | 0/1 | 0/1 | 1.36 (0.89 to 1.83) | – | – | – | ||
| All respiratory, all ages | ||||||||
| Overall¶ | 49/8 | 5/5 | 1.09 (0.75 to 1.42) | 14.7 | 16/3 | 1/2 | 1.02 (−0.37 to 2.44) | 87.4 |
| AMR A | 4/1 | 3/1 | 0.54 (−0.02 to 1.10) | – | – | – | ||
| AMR B | 1/0 | 1/0 | 1.21 (−1.43 to 3.91) | – | – | – | ||
| EUR A | 29/6 | 9/3 | 1.32 (0.74 to 1.91) | 14/2 | 2/1 | 0.38 (0.17 to 0.59) | ||
| WPR A | 3/0 | 3/0 | 1.27 (−0.86 to 3.45) | 2/1 | 0/1 | 2.00 (0.89 to 3.12) | ||
| WPR B | 12/0 | 5/0 | 1.49 (0.47 to 2.52) | – | – | – | ||
| Multiregion** | 0/1 | 0/1 | 1.48 (0.68 to 2.29) | – | – | – | ||
*Numbers of SC/MC estimates available from all studies.
†Numbers of SC/MC estimates selected for meta-analysis.
‡Random-effects summary estimate (95% CI) per 10 μg/m3 increase in NO2.
§I2 statistic for heterogeneity between WHO regions.
¶Estimate numbers for ‘overall’ refer to: (1) the number of SC/MC estimates available from all studies; (2) for selected estimates, it is the number of pooled (from SC estimates) and MC estimates used to calculate the overall summary estimate across WHO regions.
**Multiregion refers to more than one WHO region. In this table, only one MC study (RMID 313) provided estimates for WHO multiregion for mortality; the WHO regions covered in this study were SEAR B and WPR B.
AMR, Region of the Americas; EUR, European Region; MC, multicity; SC, single-city; SEAR, South East Asian Region; WPR, Western Pacific Region.
Evidence of high heterogeneity in WHO region-specific summary estimates was identified in more than half of the mortality combinations (table 2 and see online supplementary table S1). We regarded an I2 statistic of ≥50% as being evidence of high heterogeneity. The evidence regarding deaths from cardiac disease, ischaemic heart disease, stroke and COPD including asthma was limited to a small number of WHO regions (see online supplementary table S1). In general, larger summary estimates for mortality were found in Europe and Asia (as WHO regions Europe A, Western Pacific region and multiregion in table 2 and online supplementary table S1). Where available, summary estimates for WHO region America A, that is, USA and Canada, were consistently smaller than those from other WHO regions for all mortality combinations examined for 24 h NO2.
The available evidence for 1 h NO2 was smaller in comparison with that for 24 h NO2. The few available estimates were mainly limited to the broad mortality combinations and, with the exception of all-cause mortality, gave comparable summary estimates (table 2).
Hospital admissions
We analysed 18 combinations of cause-specific and age-specific hospital admissions and 24 h NO2, and 14 of the same combinations for 1 h NO2. Figure 2 presents summary estimates for hospital admissions and 24 h NO2. Estimates used in the meta-analyses for hospital admissions are presented in forest plots in the online supplementary figures S11–S39.
Figure 2.
Random-effects summary estimates (95% CIs) for cause-specific and age-specific hospital admissions per 10 µg/m3 increase in 24 h NO2 (COPD, chronic obstructive pulmonary disease).
We identified positive associations between 24 h NO2 and admissions to hospital for respiratory diseases. Based on 15 estimates from five WHO regions, we found a 0.57% (95% CI 0.33% to 0.82%) increase in hospital admissions for all respiratory diseases in all ages associated with a 10 µg/m3 increase in 24 h NO2 and no evidence of heterogeneity between WHO region-specific estimates (I2=0%). Positive associations with all respiratory admissions were also observed in children and the elderly; asthma in children (1.27% (95% CI 0.28% to 2.27%)); COPD including asthma in the elderly (1.42% (95% CI 1.07% to 1.76%)); and for COPD in all ages (1.24% (95% CI −0.01% to 2.50%))—all estimates per 10 µg/m3 increase in 24 h NO2.
For all-age all cardiovascular hospital admissions, the meta-analysis of seven single-city and multicity study estimates from three WHO regions produced a summary estimate of 0.66% (95% CI 0.32% to 1.01%). Overall estimates for hospital admissions for specific cardiovascular-related diseases in all ages ranged from 0.30% for stroke admissions to 1.41% for admissions for heart failure, and in all cases lower CIs were above, or very close to, 0% (figure 2). Summary estimates for all 24 h combinations of cardiovascular-related hospital admissions, except for stroke, in the elderly were larger than the corresponding all-age summary estimates, but there was some overlap in the CIs with those of the all-age combinations (see online supplementary table S3).
Evidence of high heterogeneity in WHO region-specific summary estimates was identified in more than half of the 18 combinations of cause-specific and age-specific hospital admissions and 24 h NO2 (see online supplementary tables S2 and S3).
The available evidence for 1 h NO2 and hospital admissions was smaller in comparison but showed positive associations for all combinations examined (see online supplementary tables S2 and S3).
Small study bias
Table 3 presents results of our assessment of small study bias in single-city studies within WHO regions and between the pooled single-city and multicity estimates for mortality. We restricted this assessment to the broad mortality combinations as sufficient estimates were available. With the exception of all respiratory mortality, there is little evidence of small study bias: adjustment for the trim and fill technique did not lead to a substantial reduction in the magnitude of the estimates, and there was no change in the summary estimate (or 95% CI) for all cardiovascular mortality. Evidence of small study bias was found in the multicity studies of respiratory mortality despite p values of 0.66 and 0.21 for the Begg and Egger tests, respectively.
Table 3.
Assessment of bias in single-city studies and in pooled estimates for all-cause, all cardiovascular and all respiratory mortality, in all ages, 24 h NO2
| All-cause |
All cardiovascular |
All respiratory |
||||
|---|---|---|---|---|---|---|
| Overall estimate no adjustment for bias* | 0.71 (0.43 to 1.00) | 0.88 (0.63 to 1.13) | 1.09 (0.75 to 1.42) | |||
| Single-city bias† | ||||||
| WHO region | EUR A | WPR B | EUR A | WPR B | EUR A | WPR B |
| p Value Begg | 0.87 | 0.57 | 0.18 | 0.65 | 0.68 | 0.05 |
| p Value Egger | 0.41 | 0.86 | 0.16 | 0.89 | 0.10 | 0.23 |
| # Estimates | 14 | 6 | 10 | 7 | 9 | 5 |
| # Trim and fill | 14 | 7 | 10 | 7 | 11 | 5 |
| Pooled single-city estimate after trim and fill‡ | 0.54 (0.31 to 0.77) | 0.60 (0.47 to 0.73) | 1.16 (0.56 to 1.77) | |||
| Multicity bias§ | ||||||
| p Value Begg | 0.82 | 0.93 | 0.66 | |||
| p Value Egger | 0.003 | 0.12 | 0.21 | |||
| # Estimates | 11 | 10 | 10 | |||
| # Trim and fill | 12 | 10 | 15 | |||
| Overall estimate with adjustment for bias¶ | 0.64 (0.36 to 0.92) | 0.88 (0.63 to 1.13) | 0.80 (0.48 to 1.12) | |||
*Random-effects overall estimate (95% CI) per 10 µg/m3 NO2 (24 h) without adjustment for publication bias.
†Bias in single-city estimates within WHO region: region, Begg's test p value, Egger's test p value, number of estimates prior to adjustment for trim and fill technique, number of estimates after adjustment for trim and fill technique. Assessment of bias in single-city estimates within WHO region was not conducted in regions providing <5 estimates.
‡Overall pooled singe-city estimate calculated after application of the trim and fill technique to the single-city estimates within WHO region.
§Bias between pooled single-city estimates and multicity estimates, Begg's test p value, Egger's test p value, number of estimates prior to adjustment for trim and fill technique, number of estimates after adjustment for trim and fill technique.
¶Overall pooled estimate calculated after application of the trim and fill technique to single-city estimates within WHO region and between pooled single-city estimates and multicity estimates.
EUR, European Region; WPR, Western Pacific Region.
Discussion
This systematic review identified 204 time-series papers on NO2 and daily mortality and hospital admissions indexed in bibliographic databases up to May 2011. We undertook a total of 25 meta-analyses for 24 h NO2 and combinations of all-age, all-cause and cause-specific mortality (7) and for disease-specific and age-specific hospital admissions (18). Eighteen of the 25 combinations were also analysed for 1 h NO2. Positive associations between short-term exposure to NO2 and mortality from all-cause and cause-specific diseases and with cardiovascular and respiratory hospital admissions were observed. Overall, there was evidence of high heterogeneity between estimates for WHO regions in more than half of the outcomes examined.
Our quantitative assessment of the literature supports the conclusions of recent narrative reviews1 3 of the epidemiological evidence on NO2 and is also broadly compatible with those from the last comprehensive quantitative review by Anderson et al6 based on time-series studies published up to the beginning of 2006. The meta-analyses by Anderson et al used only single-city estimates, reflecting the dominance of this study design at the time. Our summary estimates now include a number of multicity studies, which generally applied standardised methods, conducted in different geographical regions of the world, as well as newer results from single-city studies. Our review also reflects a growing literature from outside of North America and Europe, notably from Asia, with 47 studies identified by our literature search up to May 2011. This growth is also reflected in meta-analyses focusing on data solely from studies of short-term exposure to air pollution including NO2 conducted in China.18 The growth in the time-series literature since Anderson et al6 permitted an assessment of a broader range of outcomes in major depth, and also provides summary estimates not previously available for some outcomes, for example, cardiovascular admissions. It also enables the calculation of concentration response functions on a more substantial and broader evidence base increasing confidence in estimated concentration response functions which could be used in health impact assessment. The broader geographical representation in time-series studies also supports application of effect estimates for health impact assessment from long-term exposure cohort studies on NO2 based generally in North America and Western Europe.
We observed high heterogeneity between WHO region-specific estimates in more than half of the end points examined. We hypothesise that these geographical differences may be due to variations in population characteristics and in pollution sources, mixtures and ambient concentrations. However, additional work is required to investigate potential explanations for this heterogeneity. In the meantime, quantification of the health impact of short-term exposure to NO2 could therefore consider the use of local as well as global summary effect estimates, where a sufficient number of cities have been analysed.
Our review shows some coherence in terms of the direction, magnitude and statistical significance of the associations between mortality and hospital admissions for several diseases examined. There were however much higher estimates for stroke and respiratory mortality than hospitalisation for these diseases. In general, the interpretation of mortality versus admissions is problematic because deaths are coded to the underlying cause whereas admissions are coded to the cause of admission. In the present case, associations between NO2 and stroke mortality could be explained by effects on respiratory complications rather than on a mechanism related to stroke itself. It would require studies in the same population using associated ICD codes to further investigate this important question. Comparisons between results for mortality and admissions may also reflect systematic variation in the characteristics of the studies selected for each meta-analysis.
This review is, to date, the most comprehensive available for NO2, having worldwide coverage, examining major outcomes for several diseases and age groups, and providing meta-analytic estimates both globally and for specific geographical regions. A key strength of our review is the protocol-led approach to identifying studies and estimates for use in meta-analysis, which also sought to minimise selection bias in the different stages of the review. One feature of our approach, in common with other quantitative reviews, is the requirement that papers report numerical estimates but some fail to do so. In some instances, authors may only report statistically significant estimates. A further feature of our review is the inevitable time delay due to the necessity for extracting and processing large volumes of data from the studies. In addition, because our review has worldwide coverage and did not place any restriction on language, translation of some papers was necessary. We note that studies published since our literature cut-off date continue to provide evidence of positive associations between NO2 and daily mortality and hospital admissions. Recent multicity studies of NO2 and all-cause mortality reported increases (per 10 μg/m3 NO2) of 2.09% (95% CI 0.96% to 3.24%) in 10 cities in Italy;19 1.10% (95% CI 0.63% to 1.58%) in 25 cities in Italy;20 1.63% (95% posterior interval 1.09% to 2.17%) in 17 Chinese cities21 and 1.03% ((95% CI 0.91% to 1.18%) per 10 ppb NO2) in 72 US cities.22 Some of the cities investigated in these newer multicity studies had already contributed estimates to the present meta-analysis.
The causality of adverse health associations of NO2 is a matter of current debate since there are close correlations between NO2 and other air pollutants in the mixture of traffic-related pollutants, especially PM.1 The extent of these correlations varies from city-to-city and over time, due to variations in the emission sources (eg, dieselisation of cars and primary production of NO2 by certain particle traps). Multipollutant models have been used to attempt to identify effects of NO2 that may be independent of PM and other co-pollutants. A previous meta-analysis of multipollutant model estimates of NO2 indicated a reduction in the pooled estimate and loss of statistical significance following adjustment for co-pollutants.23 24 However, recent qualitative assessments suggest that the evidence for independent effects of NO2 is strengthening, but no recent quantitative analysis of multipollutant model estimates of NO2 has been undertaken.1 3 6 25 The current meta-analysis concentrated on results from single-pollutant models, since these provided the greatest number of cities, but we have completed a similar systematic review and meta-analysis of the subset of studies that reported two-pollutant model estimates of NO2 adjusted for PM. The purpose of this second review was to assess the extent to which the associations identified in this review change on adjustment for the effects of PM. We intend to report those findings in a separate paper.
Our review has established that there are positive and statistically significant associations between short-term exposure to NO2 and mortality and hospital admissions for a range of respiratory and cardiovascular diseases in different age groups. This comprehensive quantitative assessment of the time-series evidence on NO2 forms an important part of understanding whether NO2 contributes to the public health burden of air pollution. Further work is needed to understand reasons for the regional heterogeneity observed and to quantitatively assess the independence of NO2 health effects from other air pollutants.
Acknowledgments
This is independent research commissioned by the Policy Research Programme (PRP) in the Department of Health. The views expressed are not necessarily those of the Department.
Footnotes
Contributors: All authors contributed to the design of the study, to the drafting of the paper and have seen and approved the final version. ICM read all papers, checked data prior to meta-analysis and carried out all analyses. RWA produced the statistical code in STATA used by ICM in the analyses. ICM is responsible for the overall content as lead author of the paper.
Funding: Policy Research Programme (PRP) in the Department of Health (grant number PRP 002/037).
Competing interests: None declared.
Provenance and peer review: Not commissioned; externally peer reviewed.
Data sharing statement: No additional data are available.
References
- 1.World Health Organization (WHO) Regional Office for Europe. Review of Evidence on Health Aspects of Air Pollution—REVIHAAP Project: final technical report 2013. http://www.euro.who.int/en/what-we-do/health-topics/environment-and-health/air-quality/publications/2013/review-of-evidence-on-health-aspects-of-air-pollution-revihaap-project-final-technical-report (accessed Aug 2014).
- 2.World Health Organization (WHO) Regional Office for Europe. Air Quality Guidelines Global Update 2005: particulate matter, ozone, nitrogen dioxide and sulfur dioxide. Copenhagen: WHO Regional Office for Europe, 2006. http://www.euro.who.int/__data/assets/pdf_file/0005/78638/E90038.pdf (accessed Aug 2014). [Google Scholar]
- 3.U.S. EPA. Integrated Science Assessment for Oxides of Nitrogen—Health Criteria (first external review draft) Washington, DC: U.S. Environmental Protection Agency, EPA/600/R-13/202, 2013. http://epa.gov/ncea/isa/index.htm (accessed Aug 2014). [Google Scholar]
- 4.U.S. EPA. Integrated Science Assessment for Oxides of Nitrogen—Health Criteria (final report) Washington, DC: U.S. Environmental Protection Agency, EPA/600/R-08/071, 2008. http://cfpub.epa.gov/ncea/isa/recordisplay.cfm?deid=194645 (accessed Jun 2014). [Google Scholar]
- 5.HEI International Scientific Oversight Committee. Outdoor air pollution and health in the developing countries of Asia: a comprehensive review. Special Report 18 Boston, MA: Health Effects Institute, 2010. [Google Scholar]
- 6.Anderson HR, Atkinson RW, Bremner SA et al. Quantitative systematic review of short-term associations between ambient air pollution (particulate matter, ozone, nitrogen dioxide, sulphur dioxide and carbon monoxide), and mortality and morbidity. London: Department of Health, 2007. https://www.gov.uk/government/publications/quantitative-systematic-review-of-short-term-associations-between-ambient-air-pollution-particulate-matter-ozone-nitrogen-dioxide-sulphur-dioxide-and-carbon-monoxide-and-mortality-and-morbidity (accessed Aug 2014). [Google Scholar]
- 7.California Environmental Protection Agency. Review of the California ambient air quality standard for nitrogen dioxide. Sacramento: California Environmental Protection Agency Air Resources Board, 2007. http://www.arb.ca.gov/research/aaqs/no2-rs/no2-doc.htm#TechSuppDoc (accessed Aug 2014). [Google Scholar]
- 8.HEI Public Health and Air Pollution in Asia Program. Public Health and Air Pollution in Asia (PAPA): coordinated studies of short-term exposure to air pollution and daily mortality in four cities. HEI Research Report 154 Boston, MA: Health Effects Institute, 2010. [Google Scholar]
- 9.Wong CM, Vichit-Vadakan N, Kan H et al. Public health and air pollution in Asia (PAPA): a multicity study of short-term effects of air pollution on mortality. Environ Health Perspect 2008;116:1195–202. 10.1289/ehp.11257 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Samoli E, Aga E, Touloumi G et al. Short-term effects of nitrogen dioxide on mortality: an analysis within the APHEA project. Eur Respir J 2006;27:1129–38. 10.1183/09031936.06.00143905 [DOI] [PubMed] [Google Scholar]
- 11.Brook JR, Burnett RT, Dann TF et al. Further interpretation of the acute effect of nitrogen dioxide observed in Canadian time-series studies. J Expo Sci Environ Epidemiol 2007;17(Suppl 2):S36–44. 10.1038/sj.jes.7500626 [DOI] [PubMed] [Google Scholar]
- 12.Atkinson RW, Kang S, Anderson HR et al. Epidemiological time series studies of PM2.5 and daily mortality and hospital admissions: a systematic review and meta-analysis. Thorax 2014;69:660–5. 10.1136/thoraxjnl-2013-204492 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177–88. 10.1016/0197-2456(86)90046-2 [DOI] [PubMed] [Google Scholar]
- 14.Huedo-Medina TB, Sanchez-Meca J, Marın-Martınez F et al. Assessing heterogeneity in meta-analysis: Q statistic or I2 index? Psychol Methods 2006;11:193–206. 10.1037/1082-989X.11.2.193 [DOI] [PubMed] [Google Scholar]
- 15.Begg CB, Berlin JA. Publication bias and dissemination of clinical research. J Natl Cancer Inst 1989;81:107–15. 10.1093/jnci/81.2.107 [DOI] [PubMed] [Google Scholar]
- 16.Egger M, Davey Smith G, Schneider M et al. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315:629–34. 10.1136/bmj.315.7109.629 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Duval S, Tweedie R. A nonparametric “trim and fill” method of accounting for publication bias in meta-analysis. J Am Stat Assoc 2000;95:89–98. [Google Scholar]
- 18.Shang Y, Sun Z, Cao J et al. Systematic review of Chinese studies of short-term exposure to air pollution and daily mortality. Environ Int 2013;54:100–11. 10.1016/j.envint.2013.01.010 [DOI] [PubMed] [Google Scholar]
- 19.Chiusolo M, Cadum E, Stafoggia M et al. ; EpiAir Collaborative Group . Short-term effects of nitrogen dioxide on mortality and susceptibility factors in 10 Italian cities: the EpiAir study. Environ Health Perspect 2011;119:1233–8. 10.1289/ehp.1002904 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Alessandrini ER, Faustini A, Chiusolo M et al. Air pollution and mortality in twenty-five Italian cities: results of the EpiAir2 project. Epidemiol Prev 2013;37:220–9. [PubMed] [Google Scholar]
- 21.Chen R, Samoli E, Wong CM et al. ; CAPES Collaborative Group . Associations between short-term exposure to nitrogen dioxide and mortality in 17 Chinese cities: the China Air Pollution and Health Effects Study (CAPES). Environ Int 2012;45:32–8. 10.1016/j.envint.2012.04.008 [DOI] [PubMed] [Google Scholar]
- 22.Moolgavkar SH, McClellan RO, Dewanji A et al. Time-series analyses of air pollution and mortality in the United States: a subsampling approach. Environ Health Perspect 2013;121:73–8. 10.1289/ehp.121-a73 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Stieb DM, Judek S, Burnett RT. Meta-analysis of time-series studies of air pollution and mortality: effects of gases and particles and the influence of cause of death, age, and season. J Air Waste Manag Assoc 2002;52:470–84. 10.1080/10473289.2002.10470794 [DOI] [PubMed] [Google Scholar]
- 24.Stieb DM, Judek S, Burnett RT. Meta-analysis of time-series studies of air pollution and mortality: update in relation to the use of generalized additive models. J Air Waste Manag Assoc 2003;53:258–61. 10.1080/10473289.2003.10466149 [DOI] [PubMed] [Google Scholar]
- 25.Committee on the Medical Effects of Air Pollutants (COMEAP). Cardiovascular disease and air pollution 2006. http://www.comeap.org.uk/documents (accessed 18 Mar 2013).