ORIGINAL ARTICLE Particulate Air Pollution and Fetal Health A Systematic Review of the Epidemiologic Evidence Svetlana V. Glinianaia, Judith Rankin, Ruth Bell, Tanja Pless-Mulloli, and Denise Howel Background: Research on the potential impact of air pollution on the health of adults and children has grown rapidly over the last decade. Recent studies have suggested that air pollution could also be associated with adverse effects on the developing fetus. This systematic review evaluates the current level of epidemiologic evidence on the association between ambient particulate air pollution and fetal health outcomes. We also suggest further research questions. Methods: Using database searches and other approaches, we identified relevant publications published between 1966 and 2001 in English. Articles were included if they reported original data on birthweight, gestational age at delivery, or stillbirth related to directly measured nonaccidental exposure to particulate matter. Results: Twelve studies met the inclusion criteria. There was little consistency in the evidence linking particulate air pollution and fetal outcomes. Many studies had methodologic weaknesses in their design and adjustment for confounding factors. Even in well-designed studies, the reported magnitude of the effects was small and inconsistently associated with exposure at specific stages of pregnancy. Conclusions: The currently available evidence is compatible with either a small adverse effect of particulate air pollution on fetal growth and duration of pregnancy or with no effect. Further research should be directed toward clarifying and quantifying these possible effects and generating testable hypotheses on plausible biologic mechanisms. (Epidemiology 2004;15: 36 – 45) Submitted 10 December 2002; final version accepted 12 September 2002. From the School of Population and Health Sciences, University of Newcastle, Newcastle-upon-Tyne, United Kingdom. This study was funded by the Department of Epidemiology and Public Health, University of Newcastle. Correspondence: Svetlana V. Glinianaia, School of Population and Health Sciences, William Leech Building, Faculty of Medical Sciences, University of Newcastle, Framlington Place, Newcastle upon Tyne, NE2 4HH, UK. E-mail: svetlana.glinianaia@ncl.ac.uk Supplemental material for this article is available with the online version of the Journal at www.epidem.com Copyright © 2003 by Lippincott Williams & Wilkins ISSN: 1044-3983/04/1501-0036 DOI: 10.1097/01.ede.0000101023.41844.ac 36 T here is widespread evidence that short-term increases in ambient air pollution result in increased mortality and morbidity in adults and children, even at levels below current air quality standards.1–10 Considerable consistency across studies has been observed for many health effects, particularly for those linked with particulate air pollution, although the biologic mechanisms of the health effects are unclear.2,3 The effects are reported to be more pronounced in susceptible population groups such as the elderly or people with preexisting cardiovascular and respiratory conditions.11–14 In addition to short-term exposures, long-term exposure to particulate air pollution has been linked to an increase in total mortality, cardiopulmonary mortality, and respiratory morbidity.3,11 Fetuses are thought to be a further subgroup of the population who could be vulnerable to the effects of air pollutants.11 Studies of the association between particulate matter and adult health have been broadly consistent across different geographic areas of the world. At this early stage of evidence, it is of value to summarize the strength of the evidence for fetal outcomes, examine if the results are similarly consistent, and identify further research needs. This article focuses on the associations between exposure to particulate air pollution and particular fetal health outcomes (fetal growth retardation, preterm birth, and stillbirth), with the aim of identifying, appraising, and summarizing the current epidemiologic evidence. METHODS Identification of Publications and Review Process This work is part of a broader formal systematic review of the association between ambient air pollution and fetal and infant health outcomes. The process of the review was based on the guidelines published by the U.K. National Health Service Centre for Reviews and Dissemination.15 We identified papers through searches of medical and biologic databases (Medline 1966 –2001; Embase 1980 –2001; CAB Health 1973–2001; Science Citation Index, Web of Science 1981–2001; Conference Papers Index via Cambridge Scientific Abstracts databases Epidemiology • Volume 15, Number 1, January 2004 Epidemiology • Volume 15, Number 1, January 2004 1982–2001), environmental databases (Environmental Science and Pollution Management 1981–2001; Pollution Abstracts 1981–2001; Agricultural and Environmental Biotechnology Abstracts 1993–2001; Toxicology Abstracts 1981–2001; TOXLINE 1993–2001; Health and Safety Sciences Abstracts 1981– 2001), and WorldCat 1966–2001 using a comprehensive list of fetal/perinatal/infant and environmental keywords and phrases (available with the electronic version of this article at www. epidem.com). We also searched web-based resources, scanned reference lists (of review articles, books, and published studies), searched the “gray” literature (U.K. Environment Agency reports, Department of Environment, Food and Rural Affairs publications, and WHO publications), and consulted with key experts in the field. Identified publications were scanned by one of the authors (SG). The inclusion criteria for articles in this review were 1) nonaccidental exposure to directly measured particulate matter (PM); 2) a fetal health outcome; 3) publication between January 1, 1966, and December 31, 2001, in the English language; and 4) availability through the British Library or the Internet. We excluded papers describing outcomes related to occupational or accidental exposure, as well as papers on spontaneous abortions and congenital anomalies. The former was excluded because we were interested in the effects of ambient particulate air pollution on fetal health. Publications covering spontaneous abortions and congenital anomalies were excluded because there were too few for consideration. Results presented only as abstracts were excluded after the reviewing process because they did not contain sufficient information for meaningful interpretation. Papers meeting the inclusion criteria were formally reviewed by pairs of reviewers using a data extraction form based on previous reviews.16,17 This form was piloted before starting the data extraction process. The information extracted included study design, measurement methods for pollutants and outcomes, statistical techniques, confounding factors, and results. Table 1 presents a summary of the study design and results for each fetal outcome. There are multiple entries for publications that reported results for more than 1 outcome. Results were referenced back to the specific publication (rather than to the study) from which they were produced. When possible, effect estimates (odds and risk ratios, mean change) are recalculated as the expected change in outcome for an increase in air pollution levels by 10 ␮g/m3 (total suspended particulates or TSP, PM10, PM2.5). This scale is not comparable across studies measuring different particle size fractions, but its use facilitates comparison among studies using the same particle size measurements. It was not appropriate to combine the results in a formal meta-analysis because the study methods vary so widely. © 2003 Lippincott Williams & Wilkins Particulate Air Pollution and Fetal Health Exposure Measurements TSP and dust were used as measures of particle concentration in early publications. More recently, PM10 (particulate matter smaller than 10 ␮m aerodynamic diameter) or PM2.5 (particulate matter smaller than 2.5 ␮m aerodynamic diameter) have been used as measures of exposure. The exposure estimates in 1 recent study were based on the sum of annual TSP and sulfur dioxide (SO2) emissions data (TSPSO2).18 Fetal Outcomes The following definitions were used: low birthweight (LBW), birthweight less than 2500 g; very low birthweight (VLBW), birthweight less than 1500 g; intrauterine growth retardation (IUGR), birthweight at a given gestational age and sex less than 10th percentile based on national standards for live births; preterm birth, birth at gestational age less than 37 completed weeks; and stillbirth, fetal loss at gestational age 28 weeks and more, birthweight 1000 g or more, or fetal length 35 cm or more. LBW can be a consequence of either preterm birth (appropriate-for-gestational-age babies born prematurely) or retarded fetal growth (small-for-gestational-age babies born at term). Reduced fetal growth and duration of gestation have different risk factors,19 and different biologic mechanisms are likely to lead to these different outcomes. We considered LBW to be a measure of fetal growth if the analysis was adjusted for gestational age, or if LBW in full-term babies (born at 37 or more completed weeks of gestation) was used as an outcome. IUGR is a measure of restricted fetal growth at any gestation. We categorized VLBW (used as an outcome measure in 1 study18) as a measure of preterm birth, because over 97% of VLBW babies are likely to be born preterm.20 –22 Study Design A study was described as ecologic if both fetal and exposure data were measured or estimated at a geographic area-based level rather than at an individual level. One ecologic study was described as time-series because it included data over time for a geographically defined population. For the remaining study designs, the measurement level was predominantly semi-individual (in which fetal data were collected at an individual level and air pollution data were measured at an area-based level) or occasionally individual (both fetal and air pollution data were collected or estimated at an individual level). In the cohort studies, the sample comprised all births in a defined area and time period, and the air pollution exposure before birth and subsequent fetal events were identified. In the only relevant case-control study (fetal-event cases and control births), the exposure to air pollution was estimated by an annual mean rather than exposure before birth. 37 Glinianaia et al 38 TABLE 1. Studies Investigating the Association Between Particulate Matter and Birthweight Parameters, Preterm Birth and Stillbirth Author Country and Data Collection Period Study Design China 1988–1991 Cohort Maisonet28 USA 1994–1996 Cohort Bobak24 Czech Republic 1991 Cohort Ha30 S. Korea 1996–1997 Cohort Findings by Exposure Periods Estimate (95% CI)* Low birthweight (LBW) adjusted for gestational age Means of exposure in each Per 10-␮g/m3 increase in TSP in third trimester, whole pregnancy trimester (26 wks–delivery): ⫹ lagged moving average, AOR of LBW ⫽ 1.01 ie., 1, 2, 3, . . .n weeks (1.005–1.014); reduction in mean before birth BW ⫽ 0.7 g (SE ⫽ 0.14); Data not reported for associations in other exposure periods Means of exposure in each Per 10-␮g/m3 increase in PM10: trimester of pregnancy First trimester: AOR ⫽ (measurements taken at 0.93 (0.85–1.00) every sixth day averaged) Second trimester: AOR ⫽ 0.93 (0.85–1.02) Third trimester: AOR ⫽ 0.96 (0.88–1.06) Reduction in mean BW not reported Average daily means in each Per 10-␮g/m3 increase in TSP: First trimester of pregnancy trimester: AOR ⫽ 1.02 (0.99–1.07) Second trimester: AOR ⫽ 1.03 (0.98–1.08) Third trimester: AOR ⫽ 1.03 (0.99–1.08) Per 10-␮g/m3 increase in TSP in first trimester: reduction in mean BW ⫽ 2.2 (0.6–3.7) Average daily means in the Per 10-␮g/m3 increase in TSP: first and third trimesters of First trimester: AOR of LBW ⫽ pregnancy 1.03 (1.00–1.06) Third trimester: AOR of LBW ⫽ 0.96 (0.93–1.00) First trimester: reduction in mean BW ⫽ 4.5 g (3.0–6.0) Comments Adjusted for key confounders† at individual level. Not adjusted for other pollutants examined (SO2) Adjusted for other pollutants examined (CO and SO2) and key confounders† at individual level, including maternal smoking. Adjusted for key confounders† at individual level. Results adjusted for other pollutants examined (SO2, NOx) not given. Adjusted for key confounders† at individual level. Not adjusted for other pollutants examined (SO2, CO, NO2, O3). Continued on next page Epidemiology • Volume 15, Number 1, January 2004 © 2003 Lippincott Williams & Wilkins Wang29 Exposure Period(s) Considered Author Country and Data Collection Period Study Design Czech Republic 1994–1996 Cohort Dejmek27 Czech Republic 1994–1998 Cohort Bobak24 Czech Republic 1991 Cohort Sakai31 Japan 1973–1977 Ecological Bobak25 Czech Republic 1986–1988 Ecologic Rogers18 USA 1986–1988 Casecontrol Findings by Exposure Periods Estimate (95% CI)* Intrauterine growth retardation (IUGR) Average daily means in each Per 10-␮g/m3 increase in PM10 in gestational month first month: AOR ⫽ 1.22 (1.07–1.40) Per 10-␮g/m3 increase in PM2.5 in first month: AOR ⫽ 1.16 (0.99–1.35) Continuous data results not reported for other periods Average daily means in each Per 10-␮g/m3 increase in PM10 in gestational month first month: Teplice region: AOR ⫽ 1.19 (1.06–1.33); Prachatice region: AOR ⫽ 1.04 (0.86–1.27) Estimates not reported for 2–9 months Means of exposure in each Per 10-␮g/m3 increase in TSP in first trimester of pregnancy by trimester: averaging daily means AOR ⫽ 0.98 (0.94–1.01) No associations in other trimesters as well, but the estimates not reported LBW without adjustment for gestational age Annual mean Correlation coefficient ⫽ ⫺0.455 (P ⬎ 0.1) for annual mean dust particle levels and LBW rate Annual mean Per 10-␮g/m3 increase in TSP: AOR ⫽ 1.01 (0.99–1.03) Very LBW not adjusted for gestational age Annual exposure mean of Compared with TSPSO2 ⬍9.9␮g/m3: TSPSO2 at birth address TSPSO2 ⬎56.8 ␮g/m3: AOR ⫽ 2.88 (1.16–7.13) TSPSO2 25.2–56.8 ␮g/m3: AOR ⫽ 1.27 (0.68–2.37) TSPSO2 9.9–25.2 ␮g/m3: AOR ⫽ 0.99 (0.51–1.72) Comments Adjusted for key confounders† at individual level, including maternal smoking. PM10 results not adjusted for PM2.5 and vice versa. Adjusted for key confounders† at individual level including maternal smoking. Not adjusted for other pollutants examined (c-PAHs, PM2.5) Adjusted for key confounders† at individual level. Not adjusted for other pollutants examined (SO2, NOx). No adjustment for confounders, including other pollutants (SO2, NO2, NO). Poor exposure information. Adjusted for socioeconomic factors at area-based level. Also adjusted for other pollutants examined (SO2, NOx), but the results not given here. Imprecise exposure information. Individual exposure estimates from environmental transport model. Adjusted for a range of important confounders at individual level. continued on next page 39 Particulate Air Pollution and Fetal Health Dejmek26 Exposure Period(s) Considered Epidemiology • Volume 15, Number 1, January 2004 © 2003 Lippincott Williams & Wilkins TABLE 1. Continued Glinianaia et al 40 TABLE 1. Continued Author Country and Data Collection Period Study Design Exposure Period(s) Considered © 2003 Lippincott Williams & Wilkins Czech Republic 1991 Xu33 China 1988 Cohort Ritz32 USA 1989–1993 Cohort Bobak25 Czech Republic 1986–1988 Ecologic Sakai31 Japan 1973–1977 Ecologic Annual mean Pereira34 Brazil 1991–1992 Timeseries 0–14-day lagged moving averages of exposure For annual mean dust particle levels and spontaneous fetal death rate: correlation coefficient ⫽ ⫺0.351 (P ⬎0.1) Per 10-␮g/m3 increase in PM10 on the concurrent day: ARR of daily intrauterine deaths ⫽ 1.01 (1.00–1.02) Comments Adjusted for maternal age, parity and socioeconomic factors at individual level. Not adjusted for other pollutants examined (SO2, NOx). Adjusted for some confounders. Other analyses adjusted for other pollutants examined (SO2) (results not given here). Adjusted for a range of maternal and socioeconomic factors at individual level. Other analyses adjusted for other pollutants examined (CO, NO2, O3) (results not given here) Adjusted for socioeconomic factors at area-based level. Other analyses adjusted for other pollutants examined (SO2, NOx) (results not given here). Imprecise exposure information. No adjustment for confounders, including other pollutants (SO2, NO2, NO). Poor exposure information. Adjusted for limited confounders at area-based level. Other analyses adjusted for other pollutants examined (CO, NO2, SO2, O3) (results not given here). *When possible, study results were reexpressed as the estimated effect of increasing air pollution levels by 10 ␮g/m3 (PM10, TSP). † Key confounders for LBW typically included gestational age, maternal age, sex of the baby and season; in addition, some studies also adjusted for parity, socioeconomic factors, maternal smoking, and alcohol consumption. LBW (low birthweight), birthweight ⬍2500 g; very LBW, birthweight ⬍1500 g; IUGR (intrauterine growth retardation), birthweight at given gestational age and gender ⬍10th percentile vs. country-based standards for live births; preterm birth, birth at ⬍37 wks of gestation; AOR, adjusted odds ratio; ARR, adjusted rate ratio; range in brackets after AOR or ARR is 95% confidence interval; PM10, particulate matter smaller than 10 ␮m aerodynamic diameter; PM2.5, particulate matter smaller than 2.5 ␮m aerodynamic diameter; TSP, total suspended particulates; c-PAHs, carcinogenic fraction of polycyclic aromatic hydrocarbons; O3, ozone; SO2, sulphur dioxide; NOx, nitrogen oxides; CO, carbon monoxide. Epidemiology • Volume 15, Number 1, January 2004 Preterm birth ⴞ mean gestational age Average daily means in each Per 10-␮g/m3 increase in TSP: First trimester: AOR ⫽ 1.03 trimester of pregnancy (1.01–1.06) Second trimester: AOR ⫽ 1.02 (0.99–1.05) Third trimester: AOR ⫽ 1.02 (0.99–1.06) 0–11-day lagged moving Per 10-␮g/m3 increase in TSP and 7-day lag: AOR preterm bith ⫽ averages of exposure 1.01 (1.00–1.02) reduction in mean gestation ⫽ 0.0042 wks (SE ⫽ 0.012) Means of exposure during Per 10-␮g/m3 increase in PM10: First month: AOR ⫽ 1.02 (1.00–1.04) whole pregnancy, first 2 6 wks before birth: AOR ⫽ 1.03 months and 1, 2, 4, 6, 8, (1.01–1.06) 12, and 26 wks before Estimates not reported for other birth. (measurements taken periods at every sixth day averaged) Stillbirth Annual mean Per 10-␮g/m3 increase in TSP: AOR of stillbirth rate ⫽ 0.99 (0.95– 1.04) Bobak24 Cohort Findings by Exposure Periods Estimate (95% CI)* Epidemiology • Volume 15, Number 1, January 2004 Quality of the Studies Study quality was described by the extent to which the design, conduct, and analysis minimized selection, measurement, and confounding biases. The following methodologic domains have been suggested to describe the quality of observational studies: study design, comparability of subjects, exposure measurement, outcome measurement, statistical analysis, and funding or sponsorship.23 The main variations among the reviewed studies were in 1) study design (semi-individual vs. ecologic); 2) exposure measurement (exposure periods used; validity and reliability of the measurement method; comparability of measurement across the study groups; distance of residence from the monitoring stations, when applicable; and spatial or compositional heterogeneity among the examined districts); 3) adjustment for key confounders (in particular, gestational age, maternal smoking, socioeconomic factors); and 4) statistical analysis (appropriateness and whether multiple comparisons were taken into consideration in the interpretation of the results). Key points in relation to these criteria are given in Table 1 and taken into account in the narrative summary. RESULTS Studies Identified We identified 312 papers, reports, and abstracts, of which 46 met the original inclusion criteria for the full systematic review of all air pollutants and fetal/infant outcomes. (We excluded 225 based on the abstract and a further 41 based on reviewing the full publication.) Twelve publications relating to fetal outcomes and directly measured particulates met the inclusion criteria for this article and are reported here. Study Methods Table 1 summarizes the findings of the 12 publications exploring the association of maternal exposure to particulate air pollution with various birthweight parameters,18,24 –31 preterm birth,24,32,33 and stillbirth.25,31,34 Two Czech studies26,27covered the same study area, with the later study27 extending the study period by 2 years.26 The 2 Chinese studies29,33 also covered the same study area, but examined slightly different populations (all births vs. full-term births) and fetal outcomes (preterm birth vs. birthweight). Particulate air pollution was measured as TSP,24,25,29,30,33 TSPSO2,18 dust particles,31 PM10,26 –28,32,34 or PM2.5.26,27 Six publications reported outcomes relating to fetal growth: term LBW,28 –30 LBW adjusted for gestational age,24 and IUGR.24,26,27 Four publications reported outcomes relating to preterm birth,18,24,32,33 including 1 with VLBW as an outcome measure.18 In 2 studies, the outcome was LBW unadjusted for gestational age,25,31 which can be considered as either a consequence of preterm birth or growth restriction. The studies varied by study design, geographic region, PM source and composition, co-pollutant exposures, exposure period investigated, and summary statistics used (Table © 2003 Lippincott Williams & Wilkins Particulate Air Pollution and Fetal Health 1). Eleven of the 12 studies used direct measurements of particulates from routine monitoring of the ambient air pollution level by automated network stations in the study areas,24 –34 and the remaining 1 used air pollution data based on industrial emission and meteorologic data.18 Eight studies using direct measurements were semi-individual cohort studies using area-based estimates of maternal exposure to pollutants.24,26 –30,32,33 One case-control study used industrial emissions and modeled the pollutant level at each home address.18 The remaining 3 studies used ecologic or timeseries design.25,31,34 In the cohort studies, the hourly or daily measurements of PM concentrations from monitors allowed estimation of individual maternal exposure during the whole or part of the pregnancy, whereas 1 case-control and 2 ecologic studies used the annual mean concentrations of particles. The exposure periods differed among the cohort studies. Many estimated maternal exposure to PM by trimester of pregnancy, but some examined gestational months26,27 or other periods before birth.32 All 6 cohort studies examining fetal growth adjusted for some important confounding factors, including gestational age, maternal age, and fetal sex. Some studies additionally adjusted for season,24,26 –29 parity and socioeconomic factors,24,26 –28,30 maternal smoking and alcohol consumption,26 –28 and other air pollutants.28 One ecologic study did not adjust for any confounding factors.31 Because there is no consistency as to whether associations with PM are reported after adjustment for other pollutants, in Table 1 we have quoted results unadjusted for other pollutants when possible. Study Findings Fetal Growth Four of the 6 cohort studies on fetal growth reported associations between decreased fetal growth and particulates,26,27,29,30 whereas 2 had inconclusive findings24,28 (Table 1). The association with a change of 10 ␮g/m3 in particulate matter was strongest (odds ratio [OR], approximately 1.2) for 2 of the studies in which IUGR was the clinical outcome,26,27 but this was estimated fairly imprecisely. The exposure period associated with reduced fetal growth was not consistent across studies. Ha et al.30 found a weak association (OR ⫽ 1.03) between reduced fetal growth in term babies and increased PM during the first trimester of pregnancy only, whereas Wang et al.29 found a weak association (OR ⫽ 1.01) with exposure in the third trimester but not in the first. Two Czech publications reported a slightly stronger association in the first gestational month, but not in the other examined periods.26,27 These 2 latter publications are not independent datasets, with one26 being a subset of the other.27 Two ecologic studies found little evidence of an association between PM level and LBW.25,31 However, because 41 Epidemiology • Volume 15, Number 1, January 2004 Glinianaia et al there was no adjustment for gestational age, this outcome included both growth-restricted and preterm fetuses. Preterm Birth Weak associations with preterm birth (OR, ⬍1.03) were found with a 10-␮g/m3 increase in PM during pregnancy in 3 cohort studies.24,32,33 However (and like with fetal growth), the relevant period of exposure varied among studies (Table 1). The only case-control study found a higher risk of VLBW with increased estimated levels of TSPSO2 after adjustment for important socioeconomic and maternal confounding factors, including maternal and passive smoking; the odds ratio comparing highest and lowest exposure levels was 2.88, but the precision of this estimate was very low18 (Table 1). Because there was no adjustment for gestational age, VLBW babies in this study are very likely to be born preterm,20 –22 thus contributing to the findings of “preterm birth” studies. Stillbirth The 3 studies reported little evidence of an association between exposure to PM and stillbirth rate. However, the “stillbirth” studies were ecologic or time-series in design with their known potential for bias. In our judgment, the current evidence is insufficient to assess a possible association between PM and stillbirth. In summary, some results from studies of reasonably high quality suggested a weak association between ambient PM levels and fetal growth restriction and preterm birth, but others did not. All of the reported positive associations were small with odds ratios close to 1.0. Furthermore, the critical exposure period, if any, remains unclear. DISCUSSION The available evidence suggests that if there is any effect of particulate air pollution on fetal outcomes, it is unlikely to be large. It remains important to establish whether a causal association exists, even one of small magnitude, because at a population level, even a small reduction in mean birthweight or gestational duration could have a substantial health impact. The current evidence is compatible with either a small true association or with no association. Although some good-quality studies did report an association between exposure to higher levels of particulate air pollution and adverse fetal outcomes, a number of potential biases and errors need to be considered. Publication Bias Some studies lacking statistically significant associations might not have been published. For published studies, it is not known whether the authors chose to report all computed associations, regardless of statistical significance. Publication bias, and the exclusion of papers not published in English, could have decreased the number of results available for review. 42 Multiple Comparisons Most papers reported the results relating to various combinations of pollutant, exposure period, and outcome. The findings should be interpreted with caution in these circumstances because of the increased likelihood of a positive finding occurring by chance. All relevant comparisons should be reported, whatever the findings. Exposure Assessment Semi-individual studies have the limitation that air pollution exposure is estimated by a few monitors that might not represent individual exposures. This could result in misclassification of exposure, which biases effect estimates toward the null. The potential for bias is also affected by the monitoring possibilities or decisions (eg, annual or daily means, where and how measured, distance of residence from the monitoring stations, residential mobility, and so on). A further issue could be the absence of information about the level of indoor air pollution. These factors could lead to an underestimate of the size of any association. Studies exploring the health effects of PM are complex to summarize because the definitions and measurement techniques have varied over time. Even the toxicity of equal-sized PM depends on its dispersion and chemical composition, which, in turn, depend on a particular source.35 The reviewed studies varied by geographic region, with different sources and composition of PM, average levels and ranges, and co-pollutant exposures. Particulate air pollution was measured as PM10 in 4 geographic settings (Northeastern United States, Southern California, Brazil, Czech Republic), as PM2.5 in 1 (Teplice, Czech Republic), TSP in 3 (China, South Korea, Czech Republic), TSPSO2 in one (Georgia, United States), and as dust particle levels in 1 (Japan). The composition of TSP in different geographic regions can differ considerably. According to the U.S. Environmental Protection Agency, the PM10-to-TSP ratio is between 50% and 60% for U.S. sampling sites.2 In the Czech Republic, in contrast, PM10 has been estimated to constitute approximately 80% of TSP.25 In China, coal combustion is the dominant source of particulate air pollution,29 whereas in the Czech Republic, industrial air pollution plays a major role in the overall ambient pollution.24 In South Korea, automobile exhaust emission is the major air pollution source.30 Differences in PM level, size, and composition could have affected the strength of association between PM and fetal growth in the different geographic settings. For example, in 4 comparable cohort studies,24,28 –30 2 (in China29 and South Korea30) found a weak association between PM level and reduced fetal growth in different trimesters of pregnancy, but the other 2 (from the United States28 and the Czech Republic24) found little evidence of an association in any trimester of pregnancy. The ambient levels of PM in the Chinese study were much higher © 2003 Lippincott Williams & Wilkins Epidemiology • Volume 15, Number 1, January 2004 than those from South Korea or the Czech Republic, which were themselves higher than in the U.S. study, and this could contribute to differences in studies. Confounding Any observational study has the potential problem that an association can result from confounding. This can arise from uncontrolled confounding factors or residual confounding when the measurement of confounding factors is insufficiently precise. Birthweight is sensitive to numerous factors such as gestational age, maternal age, parity, maternal prepregnancy weight, maternal height, weight gain during pregnancy, socioeconomic status, smoking, and infant sex; of these, gestational age is the most important factor. To act as confounders, these factors must also be associated with the exposure. This could occur indirectly through the association of air pollution with socioeconomic status. Problems with control of confounding are partially associated with study design. It is more difficult to control sufficiently for confounders in ecologic than in individual or semiindividual studies because of the additional potential sources of bias in ecologic studies.36 For instance, in an ecologic Czech study,25 the adjustment for socioeconomic characteristics at a district level substantially reduced estimates for LBW. Given the potential impact of socioeconomic factors at both area and individual levels, it would be prudent to investigate whether both types of socioeconomic status measures should be included as confounding factors. Incomplete adjustment for nonseasonal confounding factors such as socioeconomic characteristics is not likely to be a major issue when the comparison is among pregnancies at different times rather than in different areas. Most semi-individual studies in this review chose to control for key confounding factors (ie, gestational age, maternal age, infant sex) at an individual level. However, adjustments were made less often for other important individual risk factors such as smoking, socioeconomic status, and environmental exposures, including other air pollutants (eg, SO2, NO2). In the 3 studies reporting estimates with and without adjustment for other pollutants, the effect of this adjustment on the association between PM and fetal outcomes was not consistent across the studies.25,32,33 Without adequate control for confounders, the magnitude of any association between low-level particulate air pollution exposure and fetal health would be difficult to quantify, especially if such an association is small in relation to associations with other etiologic factors. Biologic Plausibility Of all the air pollutants thought to affect health, only the mechanism of the toxic effect of CO on the fetus is well understood.37 This lack of understanding of other possible mechanisms makes it difficult to design studies to investigate potential associations, given the wide choice of key design © 2003 Lippincott Williams & Wilkins Particulate Air Pollution and Fetal Health factors. Any effect of air pollution can be mediated by multiple mechanisms (eg, an air pollutant can exert its effects directly on fetal growth by passing across the placenta or indirectly by impairing maternal health).38 Three potential mechanisms have been put forward to explain the effects of particulate air pollution on adult mortality and morbidity: an inflammatory response that alters blood coagulation, an allergic immune response, and an alteration to cardiac autonomic function resulting in the reduction of heart rate variability. All these potential mechanisms can be relevant to the fetus. In addition, maternal exposure to particulate air pollutants during pregnancy can result in decreased efficiency of the transplacental function with consequent deterioration in fetal growth and development. It is not known yet whether toxic components of PM or other measured (SO2, CO, NOX, ozone) or unmeasured compounds that are correlated with PM (such as polycyclic aromatic hydrocarbons) interfere with mechanisms regulating fetal growth and development. PM can serve as a surrogate for a complex mixture of air pollutants. When considering the overall strength of evidence, we drew on the following 3 domains: quality (summarizing the quality evaluation of the studies), quantity (magnitude of effect, numbers of studies), and consistency (the extent to which similar findings are reported using similar and different study designs).23 The evidence reviewed here gives a mixed picture, with a number of findings showing a weak association over some exposure periods, and other studies reporting little or no evidence of an association between particulate air pollution and fetal health outcomes. This picture is consistent with either a causal effect of small magnitude or with no such effect. It currently remains unclear whether the reported associations reflect a true effect or have arisen from the various potential sources of bias. The critical period of exposure is also unclear, and it is possible that the critical exposure period is different for LBW if it were caused by IUGR or by preterm birth. Despite the growing number of studies investigating the relation between air pollution and fetal outcomes, the evidence for a causal association remains weak, albeit plausible. Future research is needed to clarify whether there is a small adverse effect of particulate air pollution on fetal health. Further ecologic studies are unlikely to add to the evidence. A time-series approach could be justified if the study examines the potential effect of short-term changes in air pollutant levels on acute events (eg, preterm birth, stillbirth), but it would not be useful when examining birthweight as an outcome variable. More refined methodologic designs are needed such as large population-based cohort or casecontrol studies using individual fetal outcome and covariate data and high-quality exposure data. Studies are more likely to find evidence for a small effect if they involve settings with wide variation of air pollution levels. 43 Epidemiology • Volume 15, Number 1, January 2004 Glinianaia et al There is currently no strong evidence or hypothesis favoring one fetal outcome measure over another. In our judgment, all reviewed outcomes (ie, fetal growth, preterm birth, or stillbirth) merit further investigation. Moreover, because it is not known whether there are critical time windows of exposure during fetal development, future studies should examine a number of exposure periods during pregnancy (gestational months or trimesters) rather than using annual means of pollution levels. They should also address whether effects are cumulative and, if so, what duration of exposure to ambient air pollution is likely to cause cumulative effects. Although the role of other air pollutants is unclear, it would be advisable to measure these in addition to PM, and to report results with and without adjustment for other air pollutants. To enable estimation of exposure at different periods of pregnancy, daily measurements of air pollutants, information on gestational age at delivery, and date of birth should be available. Studies using individual measurements of exposure to ambient air pollution are now both feasible and desirable. To avoid spurious positive findings, care should be taken to form individually testable hypotheses, given the problems associated with multiple comparisons. For future studies involving birthweight, it is essential that adjustment for gestational age and other confounders (eg, smoking, socioeconomic status, and maternal factors) is made at an individual level. Reliable methods for the accurate estimation of gestational age, based on fetal ultrasound evaluation in addition to traditional obstetric methods, are advisable. PM source and chemical composition, as well as size, affect PM toxicity. High-quality exposure assessment is essential for future studies. If further research delivers consistent evidence that particulate air pollution adversely affects fetal health, there will have to be more attention to particles arising from different sources such as vehicular traffic or industrial processes. Finally, the use of biomarkers for monitoring individual maternal and fetal exposure to environmental air pollution (biomarkers of exposure)39,40 or the biologic response to exposure (biomarkers of effect) is at a very early stage. Future studies would benefit from interdisciplinary collaboration between environmental and perinatal epidemiologists, molecular biologists, and environmental toxicologists. Further development is necessary to advance our understanding of a possible mechanism for any impact of air pollution on fetal health. REFERENCES 1. Holgate ST, Samet JM, Koren HS, et al., eds. Air Pollution and Health. London: Academic Press; 1999. 2. Dockery DW, Pope CA III. Acute respiratory effects of particulate air pollution. Annu Rev Public Health. 1994;15:107–132. 3. Pope CA III, Bates DV, Raizenne ME. Health effects of particulate air pollution: time for reassessment? Environ Health Perspect. 1995;103: 472– 480. 44 4. Committee on the Medical Effects of Air Pollution (COMEAP). The Quantification of the Effects of Air Pollution on Health in the United Kingdom. London: HMSO; 1998. ISBN 0-11-322102-9. 5. Katsouyanni K, Touloumi G, Spix C, et al. Short-term effects of ambient sulphur dioxide and particulate matter on mortality in 12 European cities: results from time series data from the APHEA project. BMJ. 1997;314:1658 –1663. 6. Lebowitz MD. Epidemiological studies of the respiratory effects of air pollution. Eur Respir J. 1996;9:1029 –1054. 7. Sydbom A, Blomberg A, Parnia S, et al. Health effects of diesel exhaust emissions. Eur Respir J. 2001;17:733–746. 8. Künzli N, Kaiser R, Medina S, et al. Public-health impact of outdoor and traffic-related air pollution: a European assessment. Lancet. 2000;356: 795– 801. 9. Schwartz J. Air pollution and daily mortality: a review and meta analysis. Environ Res. 1994;64:36 –52. 10. Brunekreef B, Dockery DW, Krzyzanowski M. Epidemiologic studies on short-term effects of low levels of major ambient air pollution components. Environ Health Perspect. 1995;103:3–13. 11. Pope CA III. Epidemiology of fine particulate air pollution and human health: biologic mechanisms and who’s at risk? Environ Health Perspect. 2000;108:713–723. 12. Gouveia N, Fletcher T. Time series analysis of air pollution and mortality: effects by cause, age and socioeconomic status. J Epidemiol Commun Health. 2000;54:750 –755. 13. Saldiva PH, Pope CA III, Schwartz J, et al. Air pollution and mortality in elderly people: a time-series study in Sao Paulo, Brazil. Arch Environ Health. 1995;50:159 –163. 14. Schwartz J. What are people dying of on high air pollution days? Environ Res. 1994;64:26 –35. 15. NHS Centre for Reviews and Dissemination. Undertaking Systematic Reviews of Research on Effectiveness: CRD’s Guidance for Those Carrying Out or Commissioning Reviews. Report No. 4, 2nd ed. York: CRD; 2001. 16. Bell R, Luengo S, Petticrew M, et al. Screening for ovarian cancer: a systematic review. Health Tech Assess. 1998;2:1– 84. 17. Rankin J, Carlin L, White M. A Review of the Literature on Interventions to Reduce Mortality From Breast and Cervical Cancer in Minority Ethnic Groups. London: Cancer Research Campaign; 1998. 18. Rogers JF, Thompson SJ, Addy CL, et al. Association of very low birth weight with exposures to environmental sulfur dioxide and total suspended particulates. Am J Epidemiol. 2000;151:602– 613. 19. Kramer MS. Intrauterine growth and gestational duration determinants. Pediatrics. 1987;80:502–511. 20. Mongelli M, Gardosi J. Birth weight, prematurity and accuracy of gestational age. Int J Gynecol Obstet. 1997;56:251–256. 21. Alexander GR, Kogan M, Bader D, et al. US birth weight/gestational age-specific neonatal mortality: 1995–1997 rates for whites, Hispanics, and blacks. Pediatrics. 2003;111:e61– e66. 22. Savitz DA, Ananth CV, Berkowitz GS, et al. Concordance among measures of pregnancy outcome based on fetal size and duration of gestation. Am J Epidemiol. 2000;151:627– 633. 23. West S, King V, Carey TS, et al. Systems to Rate the Strength of Scientific Evidence. US Department of Health and Human Services. Agency for Healthcare Research and Quality. AHRQ Publication No. 02-E016; 2002. 24. Bobak M. Outdoor air pollution, low birth weight, and prematurity. Environ Health Perspect. 2000;108:173–176. 25. Bobak M, Leon DA. Pregnancy outcomes and outdoor air pollution: an ecological study in districts of the Czech Republic 1986 – 8. Occup Environ Med. 1999;56:539 –543. 26. Dejmek J, Selevan SG, Benes I, et al. Fetal growth and maternal exposure to particulate matter during pregnancy. Environ Health Perspect. 1999;107:475– 480. 27. Dejmek J, Solansky I, Benes I, et al. The impact of polycyclic aromatic hydrocarbons and fine particles on pregnancy outcome. Environ Health Perspect. 2000;108:1159 –1164. 28. Maisonet M, Bush TJ, Correa A, et al. Relation between ambient air pollution and low birth weight in the Northeastern United States. Environ Health Perspect. 2001;109:351–356. © 2003 Lippincott Williams & Wilkins Epidemiology • Volume 15, Number 1, January 2004 29. Wang X, Ding H, Ryan L, et al. Association between air pollution and low birth weight: a community-based study. Environ Health Perspect. 1997;105:514 –520. 30. Ha EH, Hong YC, Lee BE, et al. Is air pollution a risk factor for low birth weight in Seoul? Epidemiology. 2001;12:643– 648. 31. Sakai R. Fetal abnormality in a Japanese industrial zone. Int J Environ Stud. 1984;23:113–120. 32. Ritz B, Yu F, Chapa G, et al. Effect of air pollution on preterm birth among children born in Southern California between 1989 and 1993. Epidemiology. 2000;11:502–511. 33. Xu X, Ding H, Wang X. Acute effects of total suspended particles and sulfur dioxides on preterm delivery: a community-based cohort study. Arch Environ Health. 1995;50:407– 415. 34. Pereira LA, Loomis D, Conceicao GM, et al. Association between air pollution and intrauterine mortality in Sao Paulo, Brazil. Environ Health Perspect. 1998;106:325–329. © 2003 Lippincott Williams & Wilkins Particulate Air Pollution and Fetal Health 35. Mage DT. A particle is not a particle is not a PARTICLE. J Expo Anal Environ Epidemiol. 2002;12:93–95. 36. Morgenstern H, Thomas D. Principles of study design in environmental epidemiology. Environ Health Perspect. 1993;101: 23–38. 37. World Health Organization Occupational and Environmental Health Team. Guidelines for Air Quality. Geneva: WHO; 2000. 38. Vorherr H. Factors influencing fetal growth. Am J Obstet Gynecol. 1982;142:577–588. 39. Perera FP, Whyatt RM, Jedrychowski W, et al. Recent developments in molecular epidemiology: a study of the effects of environmental polycyclic aromatic hydrocarbons on birth outcomes in Poland. Am J Epidemiol. 1998;147:309 –314. 40. Tabacova S, Baird DD, Balabaeva L. Exposure to oxidized nitrogen: lipid peroxidation and neonatal health risk. Arch Environ Health. 1998; 53:214 –221. 45