Year : 2015  |  Volume : 19  |  Issue : 1  |  Page : 8--13

Application environmental epidemiology to vehicular air pollution and health effects research

Rajan R Patil1, Satish Kumar Chetlapally2, M Bagvandas2,  
1 Division of Epidemiology, School of Public Health, SRM University, Chennai, India
2 School of Public Health, SRM University, Chennai, Tamil Nadu, India

Correspondence Address:
Rajan R Patil
Division of Epidemiology, School of Public Health, SRM University, Kattankulathur - 603 203, Potheri, Chennai, Tamil Nadu


Vehicular pollution is one of the major contributors to the air pollution in urban areas and perhaps and accounts for the major share of anthropogenic green-house gases such as carbon dioxide, carbon monoxide, nitrogen oxides. Knowledge of human health risks related to environmental exposure to vehicular pollution is a current concern. Analyze the range health effects are attributed varied constituents of vehicular air pollution examine evidence for a causal association to specific health effect. In many instances scenario involves exposure to very low doses of putative agents for extended periods, sometimes the period could mean over a lifetime of an individual and yet may result in small increase in health risk that may be imperceptible. Secondary data analysis and literature review. In environmental exposures, traditional epidemiological approaches evaluating mortality and morbidity indicators display many limiting factors such as nonspecificity of biological effects latency time between exposure and magnitude of the effect. Long latency period between exposure and resultant disease, principally for carcinogenic effects and limitation of epidemiological studies for detecting small risk increments. The present paper discusses the methodological challenges in studying vehicular epidemiology and highlights issues that affect the validity of epidemiological studies in vehicular pollution.

How to cite this article:
Patil RR, Chetlapally SK, Bagvandas M. Application environmental epidemiology to vehicular air pollution and health effects research.Indian J Occup Environ Med 2015;19:8-13

How to cite this URL:
Patil RR, Chetlapally SK, Bagvandas M. Application environmental epidemiology to vehicular air pollution and health effects research. Indian J Occup Environ Med [serial online] 2015 [cited 2022 Jul 1 ];19:8-13
Available from:

Full Text


Vehicular pollution is one of the major contributors to the air pollution in urban areas and perhaps and accounts for the major share of anthropogenic green-house gases such as carbon dioxide, carbon monoxide, nitrogen oxides. Apart from this direct pollutants emitted by vehicular pollution there are also secondary byproducts contributed by the motor vehicles for example ozone [1],[2] also aerosols, inorganic, and organic acids. Motor vehicles are also an important source for noncombustion particulate matter pollutants. Traffic pollution has been linked to wide range disease asthma, respiratory disease, cardiovascular diseases, adverse reproductive outcome, cancers, etc. However, these findings are surrounded with enormous scientific debates because of weak strength of associations and lack of toxicological evidence. [3]

The majority of situations involve exposures to low doses of pollutants for very long periods of time, which in many cases involve an individual's entire lifetime and result in small increments in health risks. In environmental exposures, traditional epidemiological approaches of evaluating mortality and morbidity indicators display many limiting factors such as nonspecificity of biological effects, long latency period between exposure and resultant disease, principally for carcinogenic effects and limitation of epidemiological studies for detecting small risk increments. [4],[5] The present paper discusses the methodological challenges in studying vehicular epidemiology and highlights issues that affect the validity of epidemiological studies in vehicular pollution.


Epidemiological studies require comparability between study and control groups prior to evaluating causal inference. Similarity between the groups can be considered at two levels that are environmental equivalence and population equivalences.

Environmental equivalence

It involves ensuring that both study and control groups are exposed similarly to all environmental conditions and other nonstudy exposures except for the traffic exposure. Environmental conditions can be additional exposures such as air pollutants, indoor environment or social environment, etc., within which vehicular pollution is occurring. These additional exposure are important in epidemiological evaluation as they could potentially be confounder or interactions with vehicular pollution resulting in additive or multiplicative effects giving rise to differential adverse health risk outcome. [6],[7]

Population equivalence

This concept is derived from an important epidemiological concept called study base principle. This principle requires both study and control groups to have similar baseline risks for developing hypothesized adverse health outcome in the absence of exposure or will result in similar magnitude of adverse health when both groups are exposed.


Exposure assessment is an important step in environmental health risk assessment to determine the association between exposure and the subsequent risk associated with adverse health outcome. To understand the health risk of vehicular air pollution, different methods of exposure ascertainment are adopted by different researchers as illustrated below.

Indirect exposure measurement

Majority of the studies related to vehicular air pollution rely on indirect assessment methods that are based on proxies of exposure measures to vehicular pollution e.g. distance from the vehicular traffic roadways or source intensity or geographic information systems generated computer modeling of exposure, etc. Other methods may involve giving weightage to different exposures in accordance with the time spent in each micro-environment. Researchers have also devised varied innovative methods to carry out indirect exposure assessment for vehicular pollution based on types of vehicles and engines, traffic density in terms of congestion, frequency and velocity of vehicular movement. Some researchers have used local geographical peculiarities like traffic within city limits or on highways and side road ways. While others have used habitats of exposed population near the traffic, distance from the roadways, amount of time spent in traffic jams, traffic counts. Outdoor sampling of PM2.5, soot, and NO 2 on roofs or other premises of building structures have also been collected for indirect exposure measurement. [8]

Direct personal measurements

Personal monitoring of exposures are the most accurate measurements but can only be carried out on a small but defined population. Direct assessment of exposure through personal measurement yields precise and scientifically valid information however the economics, equipment and the logistics required for personal measurements prove to be a major deterrent. Direct methods enhance the scientific validity of exposure ascertainment. Biomarkers have proven to be one of the most important tools in measurement of exposures in epidemiological studies. With advances in molecular biology, many biomarkers have been identified. Some of the common biomarkers in epidemiological studies have been makers of DNA damage as an endpoint. Assessment DNA adducts as resultant of exposure to vehicular air pollutants have been attempted by number of studies. Identification of specific exposure biomarkers not only bring in precision in the exposure assessment but also greatly help in shedding light on the possible mechanism of action of exposure. [9] Increasingly researchers are making extra effort looking for biomarkers which actually help in quantification of effects rather than merely classifying the outcome in a dichotomous category of presence or absence of end points. This provides greater flexibility in choice of epidemiological study design including cross-sectional design [10] Biological markers have unique advantage as outcome measures especially if they reflect prognosis of diseases as they may also give clue about susceptibility of disease or indicate presence of cofactors. [11]


The issue in taking distance as proxy for dose measurement in vehicular pollution epidemiology is based on an assumption that all individuals are equally exposed to the pollution within the given radius when in fact it is not, as only small fraction of population in a given radius proximity to traffic density will be highly exposed and rest of the individual within that radius will either be moderately or even minimally exposed. Hence, the assumption of equivalence of exposure to all the individuals within a given radius will invariably result in over estimation of exposure which in turn result in under estimation of relative risk. [12]

Surrogate and the actual estimate

Exposure measurement in vehicular pollution epidemiology heavily relies on proxy or surrogate measurement of exposure. However, it has been well-demonstrated by the researchers that correlation between surrogate and actual measure of exposure is not so good and the difference between the two measures gets even wider when logistic regression is applied to ascertain odds ratios for measuring the measure effect for the two measure. [13] While good correlation have been observed between blood lead level and it is surrogate X-ray fluorescence of tibia [14] at the same time for correlation between questionnaire and actual biological samples have be shown to be very poor around. [15] Furthermore, poor correlation have been found between surrogate measure based on location in herbicide spray area and actual measurement of adipose tissue dioxin level. [16]


The adverse health effects documented by many epidemiological studies need to be scrutinized for their diagnostic validity. Very often, these studies are questionnaire-based rather than objective evaluation of clinical diseases. For example majority of the studies documenting disease asthma are based on history of wheeze which may not be representative asthma. While wheeze is an important sign of asthma but presence of absence of wheeze alone cannot be diagnostic criteria of asthma. [17],[18]

Clinical significance of early biomarkers

While it is true that biomarkers help in reducing bias but very often biological end points are measured in terms of variations within the physiological range, hence they may not reflect any disease presence. They may be changes in hematological values or immune markers or changes in the sperm profile. Very often, clinical significance of such physiological changes is not well understood and hence has very little clinical significance. In general lower doses of environmental exposure may at most be associated physiological or sub-clinical conditions with nonspecific signs and symptoms rather than full-fledged classical disease. [19] Traditionally, epidemiological studies will have morbidities and mortalities as endpoints in contrast toxicological studies look for pathophysiological outcomes as endpoint which are merely biological response to entry of the foreign element in the body and may not necessarily be disease. [20]


One of the bigger challenges in interpreting presently available epidemiological association between the health effects and environmental air pollution is a lack of biological plausibility. It is further compounded by the lack of confirmation from the laboratory-based toxicological studies. Experimental animal studies have failed to show the pathological effects of exposure at the levels air pollutants present in the ambient air. [21] The problem with toxicological studies on animal models is that they often use higher concentration of pollutants to study biological effects. The ambient air level concentration pollutants very rarely produce any toxicological effect. It is well-established that toxicological effects are dose dependent outcomes; higher concentration exposure has very different modus operandi of pathophysiological pathways as compared to those at lower concentration. Second, the biological effects of single pollutants studied in the lab is very different from the effect of same pollutant when combined with other mixtures in the ambient air or in differential proportion in the same mixture at different traffic sites. [22]

Lab experiments with animals have demonstrated that the high concentrations of vehicular exhaust components especially from diesel and gasoline do induce cancer. There are many hypothesis of the mechanism, e.g. oxidative stress [23] but one cannot say with conviction if this can be attributed to the vehicular exhausts. However, when the epidemiological observations and findings from toxicological studies are evaluated together, it appears that there is likely to be an association between vehicular exhaust pollutants and cancer mortality.


Confounders need to be ruled out while considering the health effects due to environmental exposure to vehicular exhaust. The occupational exposure to lung carcinogens and other social-class related exposures are some of the confounders that need to be factored in. Retrospective studies have propensity for misclassification which invariably lead to underestimation of the magnitude of the effect. In theory high ambient air pollutants due to vehicular exhaust may increase the risk of lung cancer among never smokers, [24] however present day knowledge and evidence do not permit precise risk estimation. For example, many studies have looked at the cardiovascular effects of the traffic exposure based on physiological variations keeping variations in cardiac functioning as outcome measure. However, such outcome analysis may be questionable unless the potential confounders like stress and noise are adjusted in the traffic studies they could very likely be the actual risk factor. [25]

Many studies have shown a significant association between adverse birth outcomes and traffic density. It has been shown that such associations can be explained by confounding effect of socioeconomic status (SES) and also smoking status, owing to their association with the SES apart from being an effect modifier. Further analysis has shown association was stronger for the women who hailed from lower SES regions. This is also explained by the fact that normally mothers of lower SES resided in more polluted regions as compared to those belonging to higher SES. [26] Heat wave effect on mortality can be confounded by air pollution due to the presence of ozone and PM10. It is hypothesized that heat wave mortality is higher when ozone levels are higher hence it become imperative to adjust for pollution variables. [27]

Respiratory health effects are the most common end points studied in vehicular pollution therefore it is important to find out the role played by smoking status in risk of disease development. It is important to analyze whether the joint effect of individual constituent of vehicular pollution has played confounding or interactive role in risk development. A good analogy can be found in radon exposure and lung cancer, an additive effect along with smoking will result in similar excess risk in of lung cancer both in smokers and nonsmokers. However, since lung cancer among nonsmokers is not very common hence proportion increase and hence the etiological fraction of lung cancer among nonsmoker may seem much larger than smoker. If joint effect of smoking and radon is multiplicative in nature, then the proportional increase in lung cancer attributable to radon would be similar in both smoker and nonsmoker. However, the absolute risk increase among smoker would be much higher than in nonsmoker. This aspect has a great connotation in health risk assessment and influence on public health policy. Errors in measurement of covariate alter substantially the effect modification and may seem to have interaction when none exist. [28]


It is well-known that traffic police is highly stressful job due to long hours of outdoor duty in the hot sun and working amid high decibel traffic noise. It is also well established by many studies that traffic police personnel are highly stressed as compared to workers in other occupation. Stress is also an independent risk factor for most of the diseases or conditions that have so far been associated with vehicular pollution. There is a very high probability that stress might have confounded, mediated or modified the impact measures of the health risk that have been traditionally associated with air pollution. Stress might also play it is role indirectly by influencing or altering the risk behavior. Stress may be the primary reason for traffic police to seek treatment or have higher probability of reporting positively for the presence of adverse health effect. It is also reported that concern of environment exposure may induce alternate adverse outcome among exposed other than those that are hypothesized due to the specific environmental exposure. Studies on Chernobyl disaster have shown that perception of danger associated with living near the nuclear reactor in itself increase anxiety and distress levels. [29],[30]


Epidemiological studies of health effects due to vehicular air pollution have been controversial since the strength of associations have been very low, as a result the studies with lower relative risks are riddled with wide range of methodological issues. The basic argument being that the background "noise" (bias, confounding) in these studies was larger than the signal. The concept of clinical vulnerability subsumes past injuries/insults and related pathophysiological changes in the body that put an individual at greater risk of disease. The clinical vulnerability concept helps to explain reasons behind seemingly out of proportion effects to a small change in exposures. [31],[32]

Not only is correctness of exposure measurement important but more important is identifying the critical window period of exposure when it is biologically active is equally crucial. This is well-demonstrated in studies on exposure related to Down's syndrome. Recently it has been discovered that exposure occurring before fertilization determines the risk of development of Down's syndrome and hence all the exposure collected postfertilization is biologically irrelevant. It has been found that for Down's syndrome to occur, exposure at the time of first meiotic division of germ cell is the most crucial period. The timing aspect of exposure measurement need to be built into the research study right at the design phase. [33]

Since environmental epidemiology deals with identifying weak associations, it becomes imperative to improve exposure measurement. Exposure measures further need to be refined to increase sensitivity and avoid exposure misclassification. Outcome measures in environmental epidemiology should be clearly defined to identify adverse health condition. Specificity of the disease is crucial so as to avoid unnecessarily broaden the disease group there by the risk of diluting the measures of association. [34]

Major concern in studying vehicular air pollution in environmental epidemiology is the lack of clarity in exposure to pollutants and it is health consequences. In such situations, everybody gets exposed to low doses over entire lifetime, and that results in elevation of risk in very small units. Complexities get accentuated due to unusually long latency time between exposure and consequent outcomes which very often turn out to be small magnitude of the effect. Lack of specificity between exposure and outcome in humans complicates any ascertainment of causality (Franco et al., 2008). It is obvious that once we discount for all the exposure that do not have biological implications from measurement of associations then it will yield higher relative risk for a given association. [35]

Long lead time between the exposure and disease development pose a major challenge in vehicular pollution epidemiology. To establish the causal association for such slow developing disease, an epidemiologist must either rely on cohort studies which are uneconomical or logistically infeasible. Therefore, researchers invariably rely on retrospective exposure assessment with all it is known short comings. [36]

False positive-skepticism

Epidemiological study results are the only source of evidence possible, when experimental studies cannot be carried out due to ethical reasons. In the recent decades, epidemiology has advanced public health through regulatory, legal and medical measures. It is important to understand epidemiological studies are conducted in real world settings and not in controlled settings. It is natural that these studies are riddled with numerous biases. It would be against scientific spirit to selectively list evidence and keep harping on the possible methodological flaws which would then amount to abuse of epidemiology in public health. The concept of "false positives" results was used for many years by the tobacco industry to downplay the growing epidemiological evidence associating smoking to cancer. It is said that rare finding is rarely correct hence is more likely to be false positive. In spite of this understanding, many researchers tend to push their "novel and new" findings which can be actually detrimental to the society. [37],[38]

Amidst all the methodological issues associated with vehicular pollution and adverse effects, it is important to remember that the ultimate aim of public health is not establishing a causal association between individual constituents of vehicular exhaust and resultant disease. Public health is all about initiating action for disease prevention; environmental epidemiology findings should be scrutinized based on their ability to achieve public health impact at the community level. Any public health intervention should be based on strong epidemiological logic, and their potential impact should be demonstrated through a reduction in the incidence of disease associated with vehicular pollutions. [39]


One of the criteria for a causal association as propounded by Sir Bradford Hill is the need for replication of studies. An isolated study will not be good enough for drawing causal inference. In Epidemiological scheme of things, a positive or a negative finding gains wider acceptance if it is consistently validated by multiple studies in multiple settings. Another necessity for replication studies arise from the point of extrapolation of the findings. An evidence consistently coming from multiple studies from different geographical areas lends itself for greater extrapolation of results beyond the population studied, provided the representativeness is maintained as mere replication will not address the generalizability. [40],[41] Finally, There is a need to give greater attention to the effect of vehicular air pollutions on specific subgroups of the population who are at greater risk. Exposure is observed to increase in proportion with the volume of traffic. While the population level exposure estimates are highly valuable, however, they mask the extremes of exposure experienced by some at risk groups. At the risk, groups are people in extreme age group (geriatric and pediatric group) residing near traffic path. School children near roadways, people whose occupation require them most of their time traveling or working in heavy traffic regions e.g. traffic police, street vendors, shop keepers, etc., can be categorized as high risk group. [42],[43],[44]


Vehicular pollution epidemiology entails measurement of exposure whose resultant risk is small in magnitude and may not even be clinically significant but yet may have great public health importance because of widespread exposure in the community. From the perspectives of epidemiological causal concept as propounded by Rothman and Greenland, [45] the ambient air with vehicular exhaust pollutants can potentially cause adverse outcomes like cancer or cardiovascular diseases after prolonged exposure over long periods of time. One of the argument and analogy used is the ability of environmental tobacco smoke (ETS) to cause cancer. Contrary to the present day knowledge, these adverse health effects happen at relatively lower dose of tobacco as compared to well-known carcinogenic exposure in the environment. [46] Toxicologically speaking the strength of ETS is 1/100 as compared to active smoking. The absence of threshold and ability of very low dose exposure could be potentially explained by acquired vulnerability and cumulative effects of multiple exposures. This is explained well be Rothman by stating that the immediate cause is merely an antecedent event and that conglomeration of necessary conditions had already preexisted for the occurrence of the disease at that point in time.

Put in other way, the cause could be seen as that final component which finally completes an incomplete chain of events that initiate the cascade of events resulting into a specific outcome. This concept can be further comprehended well if it is understood from the perspectives of clinical vulnerability. The vulnerability can be due to genetic predisposition, or it can be as a consequence of antecedent insults or preexisting pathophysiological changes all of which increases the risk of disease. Hence, ETS may not cause of the disease but it can be viewed as the missing link that completed the chain of events due to preexisting vulnerabilities. This concept very well explains the effects of small changes in environmental exposure that results in seemingly out of proportion effects, especially among vulnerable subjects. It is substantiated by demonstrating that ETS has far larger effects among ex-smokers as compared to nonsmoker. This could be explained by the reasoning that ex-smokers are more vulnerable because of preexisting mutations and epigenetic changes and exposure to ETS helped to initiate a cascade of event that resulted in cancer. [47]

In summary, Rio Earth Summit 1992 principle 15 [48] states "where there are threats of serious or irreversible damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation" the tenets of precautionary principle behoves us to take steps to reduce vehicular pollution even if health effects evidence are weak.


1Liu CH, Leung DY. Numerical study on the ozone formation inside street canyons using a chemistry box model. J Environ Sci (China) 2008;20:832-7.
2Modig L, Dahgam S, Olsson D, Nyberg F, Wass K, Forsberg B, et al. Short-term exposure to ozone and levels of exhaled nitric oxide. Epidemiology 2014;25:79-87.
3US EPA. Air Quality Criteria for Particulate Matter. EPA/600/P-95/001aF; 1996.
4Franco SS, Nardocci AC, Günther WM. PAH biomarkers for human health risk assessment: A review of the state-of-the-art. Cad Saude Publica 2008;24 Suppl 4:s569-80.
5HEI. Panel on the Health Effects of Traffic-Related Air Pollution. Traffic-Related Air Pollution: A Critical Review of the Literature on Emissions, Exposure, and Health Effects. HEI Special Report 17. Boston, MA: Health Effects Institute; 2010.
6Samet JM. Air pollution risk estimates: Determinants of heterogeneity. J Toxicol Environ Health A 2008;71:578-82.
7O'Neill MS, Kinney PL, Cohen AJ Environmental equity in air quality management: Local and international implications for human health and climate change. J Toxicol Environ Health A 2008;71:570-7.
8Krzyzanowski M, Kuna-Dibbert B, Schneider J, editors. Health Effects of Transport-related Air Pollution. Copenhagen: World Health Organization, Regional Office for Europe; 2005.
9Huang HB, Lai CH, Chen GW, Lin YY, Jaakkola JJ, Liou SH, et al. Traffic-related air pollution and DNA damage: A longitudinal study in Taiwanese traffic conductors. PLoS One 2012;7:e37412.
10Demetriou CA, Raaschou-Nielsen O, Loft S, Møller P, Vermeulen R, Palli D, et al. Biomarkers of ambient air pollution and lung cancer: A systematic review. Occup Environ Med 2012;69:619-27.
11Hulka B, Wilcosky T, Griffith J. Biological markers in epidemiolgy. New York: Oxford University Press; 1990.
12Rothman KJ, Poole C. A strengthening programme for weak associations. Int J Epidemiol 1988;17:955-9.
13Rosner B, Spiegelman D, Willett WC. Correction of logistic regression relative risk estimates and confidence intervals for measurement error: The case of multiple covariates measured with error. Am J Epidemiol 1990;132:734-45.
14Somervaille LJ, Chettle DR, Scott MC, Tennant DR, McKiernan MJ, Skilbeck A, et al. In vivo tibia lead measurements as an index of cumulative exposure in occupationally exposed subjects. Br J Ind Med 1988;45:174-81.
15Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol 1985;122:51-65.
16Kang HK, Watanabe KK, Breen J, Remmers J, Conomos MG, Stanley J, et al. Dioxins and dibenzofurans in adipose tissue of US Vietnam veterans and controls. Am J Public Health 1991;81:344-9.
17Marshall JR, Graham S. Use of dual responses to increase validity of case-control studies. J Chronic Dis 1984;37:125-36.
18Hui SL, Walter SD. Estimating the error rates of diagnostic tests. Biometrics 1980;36:167-71.
19Hemminki K. Methodological issues in the use of biomarkers. Toxicol Lett 1995;77:119.
20Botti C, Combab P, Forastierec F, Settimib L. Causal inference in environmental epidemiology: The role of implicit values. Sci Total Environ 1996;184:97-101.
21Phillips AM, Silbergeld EK. Health effects studies of exposure from hazardous waste sites - where are we today? Am J Ind Med 1985;8:1-7.
22Weed DL. Environmental epidemiology: Basics and proof of cause-effect. Toxicology 2002;181-182:399-403.
23Lodovici M, Bigagli E. Oxidative stress and air pollution exposure. J Toxicol 2011;2011:487074.
24Samet JM, Avila-Tang E, Boffetta P, Hannan LM, Olivo-Marston S, Thun MJ, et al. Lung cancer in never smokers: Clinical epidemiology and environmental risk factors. Clin Cancer Res 2009;15:5626-45.
25Hampel R, Rückerl R, Yli-Tuomi T, Breitner S, Lanki T, Kraus U, et al. Impact of personally measured pollutants on cardiac function. Int J Hyg Environ Health 2014;217:460-4.
26Shah PS, Balkhair T; Knowledge Synthesis Group on Determinants of Preterm/LBW births. Air pollution and birth outcomes: A systematic review. Environ Int 2011;37:498-516.
27Analitis A, Michelozzi P, D'Ippoliti D, De'Donato F, Menne B, Matthies F, et al. Effects of heat waves on mortality: Effect modification and confounding by air pollutants. Epidemiology 2014;25:15-22.
28Greenland S. The effect of misclassification in the presence of covariates. Am J Epidemiol 1980;112:564-9.
29Dohrenwend BP, Dohrenwend BS, Kasl SV, Warheit GJ, Bardett GS, Chisholm RF, et al. Report of the public health and safety task group on behavioral effects. In: Staff Reports to the President's Commission on the Accident at Three Mile Island Public Health and Safety Task Force. No 052-003-00732-1. Washington, DC: U.S. Government Printing Office; 1979. p. 257-308.
30Dickman S. Chernobyl effects not as bad as feared. Nature 1991;351:335.
31Lioy PJ. Exposure analysis and assessment for low-risk cancer agents. Int J Epidemiol 1990;19 Suppl 1:S53-61.
32Armstrong BK, White E, Saracci R. Principles of Exposure easurement in Epidemiology. Oxford: Oxford Medical Publication; 1992. p. 236-65.
33Antonarakis SE. Parental origin of the extra chromosome in trisomy 21 as indicated by analysis of DNA polymorphisms. Down Syndrome Collaborative Group. N Engl J Med 1991;324:872-6.
34Florey CD. Weak associations in epidemiological research: Some examples and their interpretation. Int J Epidemiol 1988;17:950-4.
35Hatch M, Thomas D. Measurement issues in environmental epidemiology. Environ Health Perspect 1993;101 Suppl 4:49-57.
36Rothman KJ. Methodologic frontiers in environmental epidemiology. Environ Health Perspect 1993;101 Suppl 4:19-21.
37Boffetta P, McLaughlin JK, La Vecchia C, Tarone RE, Lipworth L, Blot WJ. False-positive results in cancer epidemiology: A plea for epistemological modesty. J Natl Cancer Inst 2008;100:988-95.
38Blair A, Saracci R, Vineis P, Cocco P, Forastiere F, Grandjean P, et al. Epidemiology, public health, and the rhetoric of false positives. Environ Health Perspect 2009;117:1809-13.
39Patil RR. Application of PHEL-'Public Health Epidemiological Logic' of Public Health Intervention and Public Health Impact. Int J Prev Med 2013;4:1331-6.
40Inferring Causation, Oxford University Press. Inferring causation in epidemiology: Mechanisms, black boxes, and contrasts. This is a post-peer-review but pre-copy-edited version of a paper due to appear. In: Illari PM, Russo F, Williamson J, editors. Causality in the Sciences.  Oxford, United Kingdom: A Volume Under Contract With Oxford University Press; 2011.
41Brunekreef B, Noy D, Clausing P. Variability of exposure measurements in environmental epidemiology. Am J Epidemiol 1987;125:892-8.
42James W, Jia C, Kedia S. Uneven magnitude of disparities in cancer risks from air toxics. Int J Environ Res Public Health 2012;9:4365-85.
43Morello-Frosch R, Jesdale BM. Separate and unequal: Residential segregation and estimated cancer risks associated with ambient air toxics in U.S. metropolitan areas. Environ Health Perspect 2006;114:386-93.
44White K, Borrell LN. Racial/ethnic residential segregation: Framing the context of health risk and health disparities. Health Place 2011;17:438-48.
45Rothman KJ, Greenland S, editors. Modern Epidemiology. 2 nd ed. Philadelphia: Lippincott Williams and Wilkins; 1998.
46Zhong L, Goldberg MS, Parent ME, Hanley JA. Exposure to environmental tobacco smoke and the risk of lung cancer: A meta-analysis. Lung Cancer 2000;27:3-18.
47Vineis P, Khan AE, Vlaanderen J, Vermeulen R. The impact of new research technologies on our understanding of environmental causes of disease: The concept of clinical vulnerability. Environ Health 2009;8:54.
48United Nations Conference on Environment and Development (UNCED), Rio de Janeiro; 3-14 June, 1992.