Residential Exposure to Traffic in California and Childhood Cancer

Reynolds, Peggy*; Von Behren, Julie*; Gunier, Robert B.*; Goldberg, Debbie E.†; Hertz, Andrew‡

doi: 10.1097/
Original Article

Background: Motor vehicle emissions are a major source of air pollution in California. Past studies have suggested that traffic-related exposures can increase the risk of childhood cancer, particularly leukemia.

Methods: From California’s statewide, population-based cancer registry, we identified cancers diagnosed in children younger than 5 years of age between 1988 and 1997. We matched these cases to California birth certificates. For each case, we randomly selected 2 control birth certificates, matched by birth date and sex. For each mother’s residential address at the time of her child’s birth, we calculated road density by summing the length of all roads within a 500-foot radius of the residence. Traffic density was based on road lengths and vehicle traffic counts for highways and major roads.

Results: The distributions of road and traffic density values were very similar for the 4369 cases and 8730 matched control subjects. For all cancer sites combined, the odds ratio (OR) for the highest road density exposure category, compared with the lowest, was 0.87 (95% confidence interval [CI] = 0.75–1.00). For all sites combined and for leukemia, the ORs were also below 1.0 for the highest traffic density exposure category (0.92 for both). For central nervous system tumors, the OR was 1.22 (CI = 0.87–1.70).

Conclusions: In a large study with good power, we found no increased cancer risk among offspring of mothers living in high traffic density areas for all cancer sites or leukemia.

From the *California Department of Health Services, Environmental Health Investigations Branch, Oakland, California; the †Public Health Institute, Berkeley, California; and ‡Impact Assessment Inc., Oakland, California.

Submitted 28 June 2003; final version accepted 23 September 2003.

Funded by National Cancer Institute Grant number R01 CA71745.

Correspondence: Peggy Reynolds, California Department of Health Services, Environmental Health Investigations Branch, 1515 Clay Street, Suite 1700, Oakland, CA 94612. E-mail:

Article Outline

Multiple studies have suggested an increased risk for childhood cancer, particularly leukemia, associated with traffic-related exposures.1–7 Parental occupational exposures to petroleum hydrocarbons and solvents8,9 have also been implicated in the development of childhood cancer. Parental work in motor vehicle-related occupations (eg, drivers and mechanics) has been implicated in the development of childhood brain tumors.10 However, recent California analyses have found little or no association between traffic volume and childhood leukemia11,12 or other cancers in children.13 A carefully designed Danish case-control study also found no relationships of traffic density, estimated nitrogen dioxide concentrations, or estimated benzene concentrations with childhood leukemia or central nervous system (CNS) tumors.14 Although evidence from human health studies to date is mixed, the potential relationship between air quality and childhood cancer risk is of continued concern.

Motor vehicle emissions are a major source of air pollution in California. Cars and trucks accounted for most of the estimated emissions of benzene (66%), 1,3-butadiene (57%), and diesel particulate matter (49%) in the state during 1996.15 Benzene is a known cause of adult leukemia,16 and butadiene and diesel exhaust are probable human carcinogens based on limited evidence from occupational and laboratory animal studies.17,18 Traffic counts and proximity to roads have commonly served as surrogates for exposure to these potential carcinogens; concentrations of these compounds are higher within 500 to 1000 feet of busy roads and freeways based on measured traffic-related air pollutant levels.19–21

We examined traffic measures in relation to cancer in very young children. This large case-control study linked information from California’s population-based cancer registry with data from California birth certificates.

Back to Top | Article Outline


Selection of Cases and Controls

We obtained information on childhood cancer cases from California’s statewide, population-based cancer registry (the California Cancer Registry) identifying cases diagnosed in children younger than 5 years of age between 1988 and 1997. A total of 5177 cases, born between 1983 and 1997, met these criteria. Using probabilistic record linkage,22 we successfully matched 4369 (84%) of these cases to California birth certificates. For each case, we randomly selected 2 control birth certificates, matched by birth date and sex. To maintain consistency with case selection criteria, we selected control births from mothers who were California residents at the time of the birth. We then matched the controls’ birth certificates to the California Birth Cohort files, which contain linked birth and death records for infants in the first year of life. We excluded 73 controls who died in infancy and replaced them with the next eligible live birth certificate from the birth cohort file (again matching for birth date and sex). Controls were replaced only if their death occurred at an age younger than the corresponding case’s age at diagnosis. The California Health and Human Services Agency, Committee for the Protection of Human Subjects approved our use of these files for this project.

Back to Top | Article Outline

Address Information

From each case and control birth certificate, we abstracted the mother’s residential address at the time of her child’s birth and used a geographic information system (GIS) to assign the address a latitude and longitude. We then used this geocoded point to locate the address on a georeferenced road network. Using 1990 U.S. Census information,23 we also assigned each address to a U.S. census block group and used the block group’s median family income as a measure of the family’s socioeconomic status.

Back to Top | Article Outline

Traffic Measures

We considered 2 traffic measures: road density and traffic density. To calculate road density, in miles per square mile (mi/mi2), we summed the total length of all roads within a 500-foot radius of each subject’s residence at birth and then divided by the area of the 500-foot radius (0.0282 square miles). Geographic Data Technology (Lebanon, NH) provided our road network database. We removed low traffic volume roads (private, forest service, and 4-wheel drive roads) from the database before performing our analysis.

To calculate traffic density, we used a combination of road lengths and vehicle traffic counts available for highways and major roads in California. Vehicle traffic counts, which are collected and reported by the California Department of Transportation, represent the number of vehicles per weekday traveling both ways on a road segment; this is also referred to as the average daily traffic. We calculated the vehicle miles traveled both ways measured road segments by multiplying the average daily traffic by the length of the road segments within a 500-foot radius of each subject’s residence at birth. We estimated traffic density in vehicle miles traveled per square mile for each subject by summing the vehicle miles traveled within a 500-foot radius of each subject’s residence and dividing by the area of that radius (0.0282 square miles).

To account for regional levels of traffic-related air pollutants that vary by urbanization, we obtained county level benzene emissions data,15 and then ranked each California county by its estimated benzene emissions divided by its land area (pounds/square mile). We assigned each county a categorical value of 1 to 4 based on its benzene emission quartile.

Back to Top | Article Outline

Statistical Analysis

To account for the matched nature of cases and controls, we calculated odds ratios (OR) and 95% confidence intervals (CI) using conditional logistic regression. All models included terms for race/ethnicity. We performed analyses for cases of all cancer sites combined, as well as for leukemias (with matched controls) and CNS cancers (with matched controls).

We divided road density and traffic density values into 5 categories based on the controls’ distributions. We used subjects in the lowest exposure quartile (<25th percentile) as the reference group and subjects in the second and third exposure quartiles as the medium-exposure group. We divided the last exposure quartile into 2 groups and assigned the highest-exposure category to subjects with traffic measures at or above the 90th percentile.

Back to Top | Article Outline


We successfully matched 4369 cases of childhood cancer to California birth certificates and included them in our analysis. These included 1728 leukemia cases, of which 1407 (81%) were acute lymphoid leukemia as well as 746 CNS tumor cases. We abstracted birth certificate information for 8738 matched controls, of whom 8 were dropped because their mothers lived outside California at the time of their birth. We successfully geocoded 98% of both the case (n = 4299) and control (n = 8596) residential birth addresses.

Table 1 shows the distribution of cases and controls by age, sex, and race/ethnicity, as well as distributions for maternal demographic factors and the child’s birth weight. Table 2 shows the road and traffic density distribution values for cases and controls. Only 204 (1.6%) subjects (70 cases and 134 control subjects) had missing road and traffic density distribution values; this was because their addresses could not be geocoded. The traffic density value was zero for 1262 cases (29%) and 2382 controls (28%) because there were no highways or major roads with traffic count data within the 500-foot radius of these subjects’ residences; they were categorized in the lowest quartile of potential exposure (the referent category).

As expected, road and traffic density values were highest for children living in metropolitan areas of California and lowest for children living in the state’s rural areas. Median road density was 21.7 mi/mi2 in metropolitan areas, 21.2 mi/mi2 in smaller cities, and 10.7 mi/mi2 in rural areas. The difference between urbanized and rural areas was more pronounced for traffic density. The median traffic density value was approximately 53,000 vehicle miles traveled per square mile in metropolitan areas, approximately 22,000 vehicle miles traveled per square mile in smaller cities, and zero in rural areas (most rural children had no measured road segments within a 500-foot radius of their residence). Road density values were distributed similarly for cases and control subjects (Table 3); 9% of cases and 10% of controls fell into the highest road density category. Results were similar for traffic density. The distributions of road and traffic density were also similar between cases of leukemia and CNS tumors and their respective control subjects.

Table 4 shows ORs for our traffic measures and childhood cancers. For all cancer sites combined, the OR for the highest road density exposure category (compared with the lowest) was 0.87 (CI = 0.75–1.00). The OR for leukemia was also below 1.0 (0.80; 0.64–1.01). For CNS tumors, the OR for the highest road density exposure was 1.03 (0.75–1.43). The ORs were also below 1.0 for the highest traffic density exposure for all sites combined (0.92; 0.80–1.06) and for leukemias (0.92; 0.73–1.15). For CNS tumors, the OR was 1.22 (0.87–1.70) based on a 1% difference in number of cases and controls in the highest exposure category (10% vs. 9%). Additional adjustment for maternal age, birth weight, and neighborhood income had little effect on point estimates for road and traffic density (data not shown). We also examined ORs for the 2 major leukemia types, acute lymphocytic leukemia and acute nonlymphocytic leukemia. The point estimates for road density and traffic density exposure categories were similar to those for all leukemias combined and all the confidence intervals included 1.0 (data not shown).

We repeated our analyses for traffic density within a 1000-foot radius of subjects’ birth residences to evaluate the sensitivity of our findings on the definition of road proximity. The ORs in these analyses were similar to those for our 500-foot radius analyses. The ORs for all traffic density categories, for all cancer sites combined, were slightly below 1.0, with an OR of 0.95 for the highest category (CI = 0.82–1.09). For leukemias, the OR for the highest traffic density category was 1.01 (CI = 0.82–1.29). The OR for the highest traffic density category for CNS cancer in this larger area was also below 1.0 (OR = 0.88; CI = 0.63–1.24).

Based on county-level benzene emissions data and land area, the most urban areas of California ranked highest in regional, traffic-related pollution levels. When we added our 4-level categorical variable for benzene emissions to the models, adjusting for differences in background levels of air pollution based on the subject’s county of residence, the resulting traffic-related point estimates for cancer risk did not change (data not shown).

Back to Top | Article Outline


In this large population-based study, we found little or no evidence for increased cancer risk among children born in high-traffic areas. For all sites combined and for the leukemias, the ORs were at or below unity for all exposure categories. For the CNS tumors, the OR for the highest traffic density exposure was elevated (1.2), but there was no dose-response pattern with increasing exposure categories. This finding was based on a very small difference (1%) between number of cases and controls in the highest exposure category. Two previous studies have reported positive associations between CNS cancers and traffic.4,13 Estimates were only slightly elevated13 or based on small numbers of cases in the highest exposure group.4

Although several studies of childhood cancer and traffic have reported increased risk with higher proxy exposures,1–7 the 4 most recently published studies on this topic have not.11–14 These 4 recent studies (3 from California) comprise much larger samples than earlier studies; the Danish study14 incorporated more rigorously modeled traffic attributes. The only positive association from the United States comes from 2 analyses of data from Denver, Colorado.1,5 A major methodologic improvement in our current study over this previous study was in the selection procedure for controls; the other researchers used random-digit dial (RDD), which tends to result in control subjects with higher socioeconomic status (SES), especially in children.24,25 In California, lower SES children are much more likely than upper SES children to live in high traffic areas.26 In addition, the Denver analyses examined traffic patterns near the diagnosis residence, whereas our study assessed traffic near the birth address.

Our findings were similar to those reported by Langholz et al.12 in their case-control study of childhood leukemia in Los Angeles County, which included cases from a broader age range and earlier time period. Although the Los Angeles County study found somewhat elevated risk ratios, ours were mostly below 1.0. The null results from these 2 case-control studies are consistent with those from a statewide ecologic study of childhood cancer rates by neighborhood measures of traffic volume.13

Because we know little about the causes of childhood cancer, it is not clear which time windows of exposure could be most important. Most studies of traffic and childhood cancer have examined exposure potential based on a child’s residence at diagnosis or current residence for control children.1,2,4–6,12 Studies of parental hydrocarbon-related occupational exposures have based their analyses largely on birth certificate data8,9; to the extent that occupations do not change, these studies can also reflect exposures closer to the time of diagnosis. It is worth noting that the most recent studies of traffic-related associations have produced generally null results, whether measuring exposure for recent residence,12,13 birth residence,11 or a child’s complete residential history encompassing an exposure window from 9 months before birth through the child’s date of diagnosis.14

Measures of traffic intensity have been of interest in childhood cancer studies because they provide proxy measures of exposures to chemical mixtures from mobile sources. Because ambient air quality data are scarce, it is often difficult to assess the degree to which proxy measures actually inform such exposures in residential areas. In Denmark, an evaluation of traffic exposure estimates using measured levels of benzene found correlation coefficients (r) of 0.55 for traffic density, which improved to 0.68 when using a model that included meteorologic data and background pollution levels.27 Using the methods described for the Danish analysis, we found a similar correlation (r = 0.69) between measured benzene and traffic density levels within a 500-foot radius of monitoring sites.13 Traffic volume is also associated with other neighborhood characteristics. Statewide data suggest that traffic density is higher in areas of California that are more urban, more industrial, more socioeconomically disadvantaged, and more likely to be populated by people of color.26 Further adjustment for these factors, or for background measures of air quality, did not change the results of our analyses.

It is possible that populations living near roadways in California, compared with other areas of the world, have very different exposures to mobile source pollutants. Traffic levels in California’s urban areas tend to be much higher than those reported in the earlier studies on this topic. Although the traffic exposure measure used in our current study is not identical to those used in previous childhood cancer studies, precluding direct comparisons, estimates of relative traffic levels could be informative. In the original Denver study,1 traffic volume on the subject’s road for the highest exposure category was greater than 10,000 cars per day. The percentage of subjects in this highest exposure category was 4% in Denver1 and 2% in Denmark.14 A reanalysis of the Denver study found that 8% of subjects, using the original control subjects, and 19% of subjects, using citywide control subjects, had distance weighted-traffic density greater than 10,000 cars per day within a 500-foot radius circle of their residence,28 whereas in Los Angeles, 30% of the subjects were at or above this exposure level.12 A traffic density of 10,000 cars per day in our study would be 67,000 vehicle miles traveled per square mile. Thus, more than one third (35%) of our subjects had a traffic density value above this level, similar to the percentage in Los Angeles. Based on this indirect comparison, it is unlikely that the traffic levels in our study were too low to detect the same effect as in the Denver study. However, emissions per vehicle have varied over time and could also be considerably different in California, Denver, and Denmark. California has some of the most stringent fuel and emission control standards in the world. Vehicle miles increased 26% in California from 1988 to 1997, but during the same period, ambient air concentrations decreased 30% on average for traffic-related pollutants such as carbon monoxide, ozone, and particulate matter.29

In a separate analysis, we recently evaluated childhood cancer incidence rates in California by census tract-based hazardous air pollutant levels as modeled by the U.S. Environmental Protection Agency (EPA).30 Unlike our proxy measure of traffic density for air contaminant exposure, the hazardous air pollutant models provided overall ambient air quality estimates that incorporated emissions from multiple sources, including traffic, large industrial manufacturing facilities, small area sources (such as dry cleaners and gas stations), and residential solvent use. Although the hazardous air pollutant study found no significant childhood cancer excess (for all sites combined) in census tracts with the highest exposure levels, the rate ratios for leukemia did appear elevated in these areas. These findings suggest that traffic-related pollutants in combination with pollutants from other sources could require further study in relation to childhood cancer.

One limitation of this study was our assignment of exposure based solely on residence at birth. Although we did not have residential history information for our study subjects, it was possible to compare address information for 2 points in time for case subjects. Many of our case families moved during the child’s gestation and in the first years after birth. When we compared cases’ birth addresses to their addresses at diagnosis, we found the 2 were the same for only 50% of cases. Because our study relied only on existing records, we could not obtain potentially useful information on parental occupational exposures, the amount of time children spent commuting, and other sources of air toxicants such as environmental tobacco smoke, which is known to contain benzene and other carcinogenic chemicals. Additionally, we did not have access to individualized air monitoring or other direct measures of subjects’ individual personal exposure. Although such direct measuring methods can provide significant predictive data, the literature suggests that even measured concentrations of air pollutants in children’s homes can only partially explain their personal exposures to combustion-related air pollutants.31,32

The records-based nature of our study affords several advantages over traditional interview studies. Both our cases and control subjects were population-based, eliminating common problems of selection or participation bias. Exposures were not self-reported, removing potential recall bias. Our study was population-based for the entire state of California and included a large number of subjects with a wide variety of traffic exposures ranging from rural areas to very urban inner cities, thus offering considerable heterogeneity in exposure potential.

We did not find evidence of increased cancer risk among young children born in high traffic areas in California. We also found no increased risk for traffic exposure and leukemias, the most common type of childhood cancer. We observed a slight increase for CNS tumors among offspring of mothers living in high traffic density areas, with no evidence of a dose-response. These data, along with recent results from other research in this area suggest that traffic-related exposures, in particular those in California during recent decades, are not associated with childhood cancer occurrence. However, other important influences on air quality can present risks for the development of cancer in young children, either alone or in combination with mobile source pollutants. Studies that can better characterize patterns of residential mobility and measured air quality will be important in further elucidating these risk relationships.

Back to Top | Article Outline


We thank Theresa Saunders for manuscript preparation and Sarah Hines-Stephans for providing the photograph used in this article.

Back to Top | Article Outline


1.Savitz DA, Feingold L. Association of childhood cancer with residential traffic density. Scand J Work Environ Health. 1989;15:360–363.
2.Nordlinder R, Jarvholm B. Environmental exposure to gasoline and leukemia in children and young adults—an ecology study. Int Arch Occup Environ Health. 1997;70:57–60.
3.Knox EG, Gilman EA. Hazard proximities of childhood cancers in Great Britain from 1953–80. J Epidemiol Community Health. 1997;51:151–159.
4.Feychting M, Svensson D, Ahlbom A. Exposure to motor vehicle exhaust and childhood cancer. Scand J Work Environ Health. 1998;24:8–11.
5.Pearson R, Wachtel H, Ebi K. Distance-weighted traffic density in proximity to a home is a risk factor for leukemia and other childhood cancers. J Air Waste Manag Assoc. 2000;50:175–180.
6.Harrison RM, Leung P, Somervaille L, et al. Analysis of incidence of childhood cancer in the West Midlands of the United Kingdom in relation to proximity to main roads and petrol stations. Occup Environ Med. 1999;56:774–780.
7.Michelozzi P, Capon A, Barca A, et al. Traffic exposure and childhood leukaemia in Rome, Italy [Abstract]. Epidemiology. 2001;12:501.
8.Colt JS, Blair A. Parental occupational exposures and risk of childhood cancer. Environ Health Perspect. 1998;106:909–925.
9.Savitz DA, Chen J. Parental occupation and childhood cancer: review of epidemiologic studies. Environ Health Perspect. 1990;88:325–337.
10.Cordier S, Mandereau L, Preston-Martin S, et al. Parental occupations and childhood brain tumors: results of an international case-control study. Cancer Causes Control. 2001;12:865–874.
11.Reynolds P, Elkin E, Scalf R, et al. Case-control pilot study of traffic exposures and early childhood leukemia using a geographic information system. Bioelectromagnetics. 2001;Suppl 5:S58–S68.
12.Langholz B, Ebi KL, Thomas DC, et al. Traffic density and the risk of childhood leukemia in a Los Angeles case-control study. Ann Epidemiol. 2002;12:482–487.
13.Reynolds P, Von Behren J, Gunier RB, et al. Traffic patterns and childhood cancer incidence rates in California, United States. Cancer Causes Control. 2002;13:665–673.
14.Raaschou-Nielsen O, Hertel O, Thomsen BL, et al. Air pollution from traffic at the residence of children with cancer. Am J Epidemiol. 2001;153:433–443.
15.California Air Resources Board. 1996 California Toxics Inventory. November 20, 2002. Available at: Accessed December 6, 2002.
16.Evaluation of the Carcinogenic Risk of Chemicals to Humans, 29th ed. Lyon, France: International Agency for Research on Cancer; 1982.
17.United States Environmental Protection Agency. Integrated Risk Information System (IRIS). November 5, 2002. Available at: Accessed December 6, 2002.
18.International Agency for Research on Cancer. IARC Monographs Programme on the Evaluation of Carcinogenic Risks to Humans. Available at: Accessed December 6, 2002.
19.Measuring Concentrations of Selected Air Pollutants Inside California Vehicles. Sacramento: California Air Resources Board, Research Division; 1998. Publication no. 95–339.
20.Kinney PL, Aggarwal M, Northridge ME, et al. Airborne concentrations of PM2.5 and diesel exhaust particles on Harlem sidewalks: a community-based pilot study. Environ Health Perspect. 2000;108:213–218.
21.Dubowsky S, Wallace L, Buckley T. The contribution of traffic to indoor concentrations of polycyclic aromatic hydrocarbons. J Expo Anal Environ Epidemiol. 1999;9:312–321.
22.Jaro M. Probabilistic linkage of large public health data files. Stat Med. 1995;14:491–498.
23.Census of Population and Housing, 1990: STF3. Washington, DC: US Bureau of the Census; 1992.
24.Poole C, Trichopoulos D. Extremely low-frequency electric and magnetic fields and cancer. Cancer Causes Control. 1991;2:267–276.
25.Olson SH, Kelsey JL, Pearson TA, et al. Evaluation of random digit dialing as a method of control selection in case-control studies. Am J Epidemiol. 1992;135:210–222.
26.Gunier RB, Hertz A, Von Behren J, et al. Traffic density in California: socioeconomic and ethnic differences among potentially exposed children. J Expo Anal Environ Epidemiol. 2003;13:240–246.
27.Raaschou-Nielsen O, Hertel O, Vignati E, et al. An air pollution model for use in epidemiological studies: evaluation with measured levels of nitrogen dioxide and benzene. J Expo Anal Environ Epidemiol. 2000;10:4–14.
28.Ebi KL, Zaffanella LE, Greenland S. Application of the case-specular method to two studies of wire codes and childhood cancers. Epidemiology. 1999;10:398–404.
29.The 1999 California Almanac of Emissions and Air Quality. Sacramento: California Air Resources Board, Planning and Technical Support Division; 1999.
30.Reynolds P, Von Behren J, Gunier R, et al. Childhood cancer incidence rates and hazardous air pollutants in California: an exploratory analysis. Environ Health Perspect. 2002;111:663-668.
31.Linaker CH, Chauhan AJ, Inskip H, et al. Distribution and determinants of personal exposure to nitrogen dioxide in school children. Occup Environ Med. 1996;53:200–203.
32.Raaschou-Nielsen O, Lohse C, Thomsen BL, et al. Ambient air levels and the exposure of children to benzene, toluene and xylenes in Denmark. Environ Res. 1997;75:149–159.

Cited By:

This article has been cited 1 time(s).

Childhood Leukemia and Socioeconomic Status in Canada
Borugian, MJ; Spinelli, JJ; Mezei, G; Wilkins, R; Abanto, Z; McBride, ML
Epidemiology, 16(4): 526-531.
PDF (308) | CrossRef
Back to Top | Article Outline
© 2004 Lippincott Williams & Wilkins, Inc.