Abstract
Purpose
In epidemiologic studies, neighborhood characteristics are often assigned to individuals based on a single residence despite the fact that people frequently move and, for most cancer outcomes, the relevant time-window of exposure is not known. The authors evaluated residential mobility patterns for a population-based series of childhood leukemia cases enrolled in the Northern California Childhood Leukemia Study.
Methods
Complete residential history from one year prior to birth to date of diagnosis was obtained for 380 cases diagnosed between 1995 and 2002. All residences were assigned U.S. Census block group designations using a geographic information system.
Results
Overall, two-thirds (65.8%) of children had moved between birth and diagnosis, and a third (34.5%) moved during the first year of life. Approximately 25% of the mothers had moved during the year before the child's birth. Multivariable analysis indicated greater residential mobility to be associated with older age of the child at diagnosis, younger age of the mother at child's birth, and lower household income. Among those who had moved, residential urban/rural status for birth and diagnosis residences changed for about 20% of subjects, and neighborhood SES for 35%.
Conclusions
These results suggest that neighborhood attribute estimates in health studies should account for patterns of residential mobility. Estimates based on a single residential location at a single point in time may lead to different inferences.
Keywords: Childhood leukemia, Epidemiology, Exposure classification, Residential mobility, Socioeconomic status
INTRODUCTION
It is becoming increasingly more common in epidemiological studies to use residential location to examine the effects of neighborhood characteristics and hazardous environmental exposures in relation to certain health outcomes, particularly with the expanding use of geographic information system tools (1). Studies examining the association between socioeconomic status (SES) and risk of childhood leukemia have used these methods to assign area-based measures of SES to children based on the geographic location of a single residence in time, usually birth, diagnosis, or death residence (2). Previous studies of childhood leukemia have also evaluated risks associated with proximity of a child's residence to sources of air pollution, traffic emissions (3), and agricultural pesticides (4). However, such data fail to account for the degree of misclassification which may result from residential mobility of the subjects. Such concerns are compounded by the uncertainty regarding the potentially different effects of exposures during critical time windows of development, the importance of which is becoming more evident based on recent studies (5, 6).
The current study takes advantage of an unusual opportunity to utilize complete residential histories from one year prior to birth to the date of diagnosis for a population-based series of childhood leukemia cases enrolled in the Northern California Childhood Leukemia Study (NCCLS). We evaluated both the degree of residential mobility and characteristics associated with mobility in this population. In addition, among children who moved, we evaluated the differences between the child's birth and diagnosis residence for two residential attributes of ongoing interest in the literature on childhood cancer: neighborhood levels of urbanization and SES.
MATERIALS AND METHODS
Study population
The NCCLS is an ongoing case-control study designed to investigate the etiology of pediatric leukemias. Beginning in December of 1995, newly diagnosed childhood leukemia cases were rapidly ascertained from every major pediatric hospital located in a 17-county San Francisco Bay Area study region, which was expanded in 1999 to 35 counties in Northern and Central California. Initial comparisons with population-based surveillance data from the California Cancer Registry (1997-1999) demonstrated the NCCLS rapid case ascertainment protocol to have effectively identified nearly 90 percent of all age-eligible newly diagnosed childhood leukemia cases among residents of the study counties. Cases interviewed between 1995 and 2002 are included in this current analysis (N = 380). For each eligible case, the statewide birth registry maintained by the Center for Health Statistics of the California Department of Public Health was utilized to generate a list of randomly selected controls that matched the case on date of birth, gender, maternal race, and Hispanic status (has a biological parent who is Hispanic).
Cases and controls were considered eligible if they were under 15 years of age, resided in the study region at the date of diagnosis (or corresponding reference date for controls), had a parent or guardian who spoke either English or Spanish, and had no prior history of malignancy. Approximately 85 percent of eligible cases have consented to participate. Residential mobility patterns were presented for the case series only. The control group may be, by design, less mobile compared to the source population as a result of our control selection strategy which required tracing people from California birth addresses. A detailed description of control selection in the NCCLS is reported elsewhere (7).
The study protocol was approved by the Institutional Review Boards of University of California, Berkeley and all collaborating institutions, and a written informed consent was obtained for all participating subjects.
Data collection
Interview data
A detailed in-home personal interview with the primary caretaker of each subject was conducted shortly after consent was obtained. Residential street addresses for all residences one month or longer were collected for children from birth to the date of diagnosis, and for mothers one year prior to child's birth. Information pertaining to the child used to evaluate characteristics associated with residential mobility included age, sex, race/ethnicity, birth order, breastfeeding, daycare attendance and complete residential history from birth to the time of diagnosis. Relevant parental characteristics included mother's age at child's birth, parental education, annual household income at the time of interview, mother's place of birth, mother's residences during the year before child's birth and whether she smoked some time during the three years after the child was born.
Residential attribute data
We geocoded all residences of the child from birth to the date of diagnosis. We used a geographic information system (GIS) to automatically match addresses to a road network and assigned a latitude/longitude coordinate, which was then used to determine the corresponding U.S. Census block group. When possible, we manually located all addresses that could not automatically be matched. Geocoding was performed using ArcView GIS software (8) (Environmental Systems Research Institute, Redlands, CA) and street databases (9-11). We did not attempt to geocode addresses outside California.
We used data from the U.S. Census Bureau to derive neighborhood measures of urbanization and SES. We assigned measures of urbanization to all residences (large metro, city and rural/town) based on U.S. Census 1990 and 2000 block group data for residences with move-in dates during or prior to 1995 and move-in dates after 1995, respectively. The “large metro” category was defined as areas within a Metropolitan Statistical Area (MSA) or Consolidated Metropolitan Statistical Area (CMSA) of ≥ 1 million people. The middle category, “city,” included areas in MSA/CMSAs of < 1 million people. The final category, “rural/town,” included small towns (< 50,000 people) and rural areas outside of census-designated urbanized areas.
We created a summary SES metric based on the 1990 and 2000 census data, incorporating block group measures of occupation, education and income (9, 12). To do this, we first ranked all California block groups separately by education level (percentage of adults over age 25 having completed a college degree or higher), income (median family income) and occupation (percentage of adults employed in managerial or professional occupations) according to quartiles based on the statewide adult population. This resulted in a score of one through four for each of these SES attributes (13). We then summed the scores for each attribute and categorized them into three groups based on quartiles: low SES (the lowest quartile of this score for the statewide population), medium SES (second and third quartiles) and high SES (the fourth and highest quartile).
Statistical analysis
Subjects who reported having two or more residences between birth and diagnosis were considered movers. The Pearson chi-square test was used to compare the characteristics of movers and non-movers among childhood leukemia cases. To evaluate factors associated with residential mobility, rate ratios (RR) and 95 percent confidence intervals (CI) were calculated using multiple linear regression assuming a Poisson distribution for the number of child's residences from birth to diagnosis. A scaling factor was specified to adjust for the observed under-dispersion of the count data, a situation where the variance for the Poisson distribution of the data is smaller than the mean. Among childhood leukemia cases who moved, birth residence and diagnosis residence were compared in terms of two residential attributes: (1) urban/rural status and (2) SES. Kappa statistics were calculated to evaluate the categorical agreement of residential attribute levels between birth and diagnosis residence. The characteristics of children experiencing “no change” in residential attributes between birth and diagnosis residence were compared to children experiencing a change in residential attributes using the Pearson chi-square test.
RESULTS
Table 1 shows the patterns of residential moves among all childhood leukemia cases, and separately for younger and older age groups. Overall, 65.8 percent (250/380) of the study children had two or more residences between birth and the date of diagnosis. As expected, child's residential mobility varied considerably by age with about 55 percent of children aged 0 to 4 years having moved at least once compared to 80 percent of children aged 5 to 14 years. A large proportion of children that moved had at least one residence outside of the county of the birth residence (38.5 percent among children ages 0 to 4 and 46.6 percent among children ages 5 to 14), including areas outside of California and the U.S. Among children who had moved, slightly greater than half (131/250) moved for the first time during the first year of life. Finally, about a quarter (101/380) of children's mothers moved during the year before the child's birth.
TABLE 1.
All Ages |
Ages 0-4 |
Age 5-14 |
||||
---|---|---|---|---|---|---|
Characteristic | N | % | N | % | N | % |
Mother moved during year before child's birth | ||||||
No | 279 | 73.4 | 157 | 74.4 | 122 | 72.2 |
Yes | 101 | 26.6 | 54 | 25.6 | 47 | 27.8 |
Child's number of residence(s) between birth and date of diagnosis | ||||||
1 (never moved) | 130 | 34.2 | 96 | 45.1 | 34 | 20.4 |
2-3 | 156 | 41.1 | 93 | 43.7 | 63 | 37.7 |
4-6 | 67 | 17.6 | 20 | 9.4 | 47 | 28.1 |
7 or more | 27 | 7.1 | 4 | 1.9 | 23 | 13.8 |
Greater than 1 (has moved) | 250 | 65.8 | 117 | 54.9 | 133 | 79.6 |
Median number of residences (range) | 2 (1-14) | 2 (1-10) | 3 (1-14) | |||
Among children that have moved: † | ||||||
Moved during the first year of life | 131 | 52.4 | 67 | 57.3 | 64 | 48.1 |
Moved out of US at least once | 26 | 10.4 | 9 | 7.7 | 17 | 12.8 |
Moved out of California at least once‡ | 49 | 19.6 | 16 | 13.7 | 33 | 24.8 |
Moved out of county of birth residence§ | 107 | 42.8 | 45 | 38.5 | 62 | 46.6 |
Duration at birth residence (months (SD)) * | 14.4 (13.1) | 22.7 (24.3) | ||||
Duration at diagnosis residence (months (SD)) * | 14.7 (12.4) | 40.7 (34.6) |
Mean number of months; SD, standard deviation
Percent among children that have moved
Includes subjects that have moved out of the US
Includes subjects that have moved out of the US and/or California
Table 2 presents the characteristics of movers (2 or more residences between birth and diagnosis) and non-movers (1 residence). Compared to non-movers, movers appeared to be older at diagnosis, have younger mothers, have parents with less formal education, have mothers who smoked, and have a diagnosis residence located in an area of lower SES. Also, there appeared to be a larger proportion of Hispanic and Black children than non-Hispanic White and “other” race/ethnicities among the movers. The results of the multivariable analyses presented in table 3 indicate that, similarly, a greater number of residences is associated with older age of the child at diagnosis, younger age of the mother at child's birth, lower household income, and living in an area classified as “city” (versus rural or town) at diagnosis. Analyses stratified by child's age at diagnosis produced similar results as the combined analysis (data not shown).
TABLE 2.
Non-movers (N = 130) |
Movers (N = 250) |
||||
---|---|---|---|---|---|
Characteristics | N | % | N | % | χ2P-value† |
Child Characteristics | |||||
Child's sex | |||||
Male | 72 | 55.4 | 133 | 46.8 | 0.685 |
Female | 58 | 44.6 | 117 | 53.2 | |
Child's race/ethnicity | |||||
White, non-Hispanic | 64 | 49.2 | 111 | 44.4 | 0.034 |
Hispanic | 45 | 34.6 | 108 | 43.2 | |
Black, non-Hispanic | 2 | 1.5 | 12 | 4.8 | |
Other | 19 | 14.6 | 19 | 7.6 | |
Child's age at diagnosis (years) | |||||
0-1 | 33 | 25.4 | 21 | 8.4 | < 0.001 |
2-4 | 63 | 48.5 | 96 | 38.4 | |
5-9 | 23 | 17.7 | 88 | 35.2 | |
10-14 | 11 | 8.5 | 45 | 18.0 | |
Mean (SD) * | 4.3 (3.5) | 6.2 (3.6) | |||
Birth order | |||||
First | 50 | 39.0 | 109 | 44.7 | 0.568 |
Second | 45 | 35.2 | 80 | 32.8 | |
Third or higher | 33 | 25.8 | 55 | 22.5 | |
Breastfeeding | |||||
No | 22 | 17.0 | 47 | 19.3 | 0.675 |
Yes | 107 | 83.0 | 197 | 80.7 | |
Daycare attendance | |||||
No | 54 | 41.5 | 98 | 39.2 | 0.659 |
Yes | 76 | 58.5 | 152 | 60.8 | |
Parental Characteristics | |||||
Maternal age at child's birth (years) | |||||
< 20 | 5 | 3.9 | 33 | 13.2 | < 0.001 |
20-24 | 12 | 9.3 | 80 | 32.0 | |
25-29 | 36 | 27.9 | 58 | 23.2 | |
30-34 | 48 | 37.2 | 59 | 23.6 | |
≥ 35 | 28 | 21.7 | 20 | 8.0 | |
Mean (SD) * | 30.8 (5.3) | 26.6 (5.9) | |||
Maternal education | |||||
≤ High school | 47 | 36.2 | 118 | 47.2 | 0.002 |
Some college or other school | 31 | 23.9 | 76 | 30.4 | |
Bachelor's degree | 52 | 40.0 | 56 | 22.4 | |
Paternal education | |||||
≤ High school | 49 | 38.6 | 137 | 56.2 | 0.005 |
Some college or other school | 29 | 22.8 | 43 | 17.6 | |
Bachelor's degree | 49 | 38.6 | 64 | 26.2 | |
Household income ($1,000/year) | |||||
< 15 | 13 | 10.0 | 44 | 17.6 | 0.107 |
15-29 | 18 | 13.9 | 51 | 20.4 | |
30-44 | 21 | 16.2 | 41 | 16.4 | |
45-59 | 27 | 20.8 | 38 | 15.2 | |
60-74 | 15 | 11.5 | 25 | 10.0 | |
≥ 75 | 36 | 27.7 | 51 | 20.4 | |
Maternal place of birth | |||||
California | 62 | 47.7 | 114 | 45.8 | 0.885 |
US, not California | 23 | 17.7 | 49 | 19.7 | |
Outside US | 45 | 34.6 | 86 | 34.5 | |
Mother moved during the year before child's birth | |||||
No | 97 | 74.6 | 182 | 73.1 | 0.749 |
Yes | 33 | 25.4 | 67 | 26.9 | |
Mother smoke post-natally (3 years) | |||||
No | 114 | 87.7 | 196 | 78.7 | 0.035 |
Yes | 16 | 12.3 | 53 | 21.3 | |
Residential Area Characteristics | |||||
Urban/rural at diagnosis residence | |||||
Rural or town | 31 | 24.0 | 61 | 24.9 | 0.953 |
City | 30 | 23.3 | 59 | 24.1 | |
Large metro | 68 | 52.7 | 125 | 51.0 | |
SES at diagnosis residence | |||||
Low | 18 | 14.0 | 58 | 23.7 | 0.067 |
Medium | 65 | 50.4 | 117 | 47.8 | |
High | 46 | 35.6 | 70 | 28.6 |
SD, standard deviation
P-values determined using two-sided Pearson's chi-square test.
TABLE 3.
Characteristics | RR | 95% CI * |
---|---|---|
Child's Characteristics | ||
Age at diagnosis (years) | ||
0-1 | 1.00 | Ref |
2-4 | 1.47 | 1.19, 1.84 |
5-9 | 2.07 | 1.66, 2.60 |
10-14 | 2.61 | 2.06, 3.31 |
Race/ethnicity | ||
White, non-Hispanic | 1.00 | Ref |
Hispanic | 0.93 | 0.80, 1.07 |
Black, non-Hispanic | 0.92 | 0.67, 1.26 |
Other | 1.03 | 0.82, 1.29 |
Parental Characteristics | ||
Maternal age at child's birth (years) | ||
< 20 | 1.00 | Ref |
20-24 | 0.77 | 0.64, 0.92 |
25-29 | 0.66 | 0.54, 0.80 |
30-34 | 0.55 | 0.44, 0.68 |
≥ 35 | 0.45 | 0.34, 0.60 |
Paternal education | ||
≤ High school | 1.00 | Ref |
Some college or other school | 0.84 | 0.71, 1.01 |
Bachelor's degree | 1.00 | 0.83, 1.20 |
Household income ($k/year) | ||
< 15 | 1.00 | Ref |
15-29 | 0.90 | 0.75, 1.08 |
30-44 | 0.85 | 0.70, 1.04 |
45-59 | 0.66 | 0.54, 0.82 |
60-74 | 0.65 | 0.51, 0.84 |
≥ 75 | 0.72 | 0.57, 0.92 |
Mother moved during the year before child's birth | ||
No | 1.00 | Ref |
Yes | 1.07 | 0.94, 1.22 |
Residential Area Characteristics | ||
Urban/rural at diagnosis residence | ||
Rural or town | 1.00 | Ref |
City | 1.26 | 1.07, 1.48 |
Large metro | 1.09 | 0.94, 1.28 |
Rate ratios (RR) and 95% confidence intervals (CI) were determined using multiple linear regression adjusting for all variables in the table and assuming a Poisson distribution for number of child's residences. A scaling factor was specified to adjust for the under-dispersion of the data. Variables included in the initial model included child's age at diagnosis, child's race/ethnicity, maternal age at child's birth, maternal education, paternal education, household income at diagnosis, mother moved during the year before child's birth, maternal smoking after child's birth, residential urban/rural status at diagnosis residence and residential SES at diagnosis residence.
Among children who moved and had both birth and diagnosis residences geocoded (N = 201), we evaluated the differences between the child's birth and diagnosis residence for two attributes of interest, urban/rural status and SES. Tables 4 and 5 present the concordance and discordance between the two residences in terms of these residential attributes. On average, approximately 80 percent of children who moved experienced no change in urban/rural status between the two residences (kappa = 0.672, P < 0.001). Of those who did experience change, a larger number of children appeared to move from large metro areas to rural areas (14 subjects) compared to moves in the opposite direction (0 subjects). For SES, approximately 64 percent of children maintained the same residential SES between birth and diagnosis residence (kappa = 0.428, P < 0.001), a slightly lower percentage compared to what was observed for urban/rural status. Of those who did change in residential SES, the direction appeared to be from lower to higher SES.
TABLE 4.
Birth Residence | ||||||||
---|---|---|---|---|---|---|---|---|
Rural / Town |
City |
Large Metro |
||||||
N | % | N | % | N | % | Total | ||
Diagnosis Residence | ||||||||
Rural / Town | 26 | 86.7 | 8 | 16.0 | 14 | 11.6 | 48 | 23.9 |
City | 4 | 13.3 | 37 | 74.0 | 8 | 6.6 | 49 | 24.4 |
Large Metro | 0 | 0.0 | 5 | 10.0 | 99 | 81.8 | 104 | 51.7 |
Total | 30 | 14.9 | 50 | 24.9 | 121 | 60.2 | 201 | 100.0 |
TABLE 5.
Birth Residence | ||||||||
---|---|---|---|---|---|---|---|---|
Low |
Medium |
High |
||||||
N | % | N | % | N | % | Total | ||
Diagnosis Residence | ||||||||
Low SES | 31 | 62.0 | 18 | 16.7 | 1 | 2.3 | 50 | 24.9 |
Medium SES | 16 | 32.0 | 67 | 62.0 | 11 | 25.6 | 94 | 46.8 |
High SES | 3 | 6.0 | 23 | 21.3 | 31 | 72.1 | 57 | 28.3 |
Total | 50 | 24.9 | 108 | 53.7 | 43 | 21.4 | 201 | 100.0 |
When examining the characteristics associated with change in residential attributes between birth and diagnosis residence, daycare attendance appeared to be less common among children who experienced change in urban/rural status compared to children with no change (data not shown). Also, children who experienced change were more likely to have mothers that moved during the year before the child's birth. No marked associations were observed for changes in SES between birth and diagnosis residence (data not shown).
DISCUSSION
This descriptive analysis of residential mobility for a population-based series of California childhood leukemia cases showed subjects to be highly mobile, with 65.8 percent of subjects having changed residences at least once (55 percent among children aged 0 to 4 years and 80 percent among children aged 5 to 14 years) during the period between birth and the date of diagnosis. In a study conducted in Northern England, Alexander et al. observed that 30.8 percent of children with leukemia and non-Hodgkin's lymphoma had changed residence by the age of 2 (14). The California children in our study were more mobile, with 47.1 percent of children having changed residence by the age of 2 years. In addition, we observed a greater proportion of mothers moving during the year before child's birth (26.6 percent) than in the Northern England study (16.5 percent). Our observation for mother's residential mobility during the year before child's birth is more consistent with U.S. studies of birth defects which have reported findings ranging from 20 percent to 33 percent (15-17). The authors of the Northern England study acknowledge, however, that their study population is likely to be less mobile compared to the general population since selection into the study required children to have been living in the same area at birth and at diagnosis. As expected, our results are similar to those from a study of leukemia in children less than five years of age conducted in San Diego, California which found that more than half of those young children had moved at least once between birth and diagnosis (18).
In this current study, several factors were found to be associated with residential mobility including older age of the child, younger age of the mother at child's birth, lower household income and diagnosis residence located in “city” compared to “rural or town.” Somewhat unexpectedly, there did not appear to be a significant difference in the number of residences between the race/ethnicity groups after accounting for the influence of SES in the multivariable analysis. We are not aware of other studies in the epidemiologic literature that describe characteristics associated with child's residential mobility. However, studies evaluating the mobility of mothers during pregnancy have reported similar findings associated with age, race and SES (15, 19). Urban/rural status was not found to be associated with mobility in these studies of pregnant women.
Many studies of SES and childhood leukemia have used neighborhood measures of SES based on a single residence (2). A recent case-control study from Denmark, which was able to compare the risk of childhood leukemia associated with SES based on birth residence and diagnosis address, observed an increased leukemia risk for children born in low-income areas, but no difference based on diagnosis address or based on individual SES information (job class of parents) (20). This study highlights the fact that different inferences may be drawn by relying on a single residence at one point in time, and this may account for some of the inconsistencies in the literature on SES and childhood leukemia (2, 21).
Several studies, including one recent California state-wide study, have shown that low-income neighborhoods and communities of color are more likely to be located near sources of pollution (22-24). This suggests that the changes in SES between birth and diagnosis residence observed in our study population may also be accompanied by changes in exposures to environmental pollutants, including motor vehicle emissions and industrial pollutants. Most previous studies have estimated such exposures based on either child's birth residence, residence at diagnosis, or current residence primarily due to the lack of residential history information. Our limited knowledge of the causes of childhood leukemia and the role of timing of exposure makes it difficult to assess which of these estimates may be most appropriate, and this resulting “misclassification” may contribute to the inconsistent results for many of these factors. In our comparison of child's birth and diagnosis residence with regards to two residential attributes, we found that a large proportion of subjects tended to move to areas of similar urban/rural status (80 percent) and SES (64 percent). However, this may be an overestimate since this analysis was limited to subjects with both birth and diagnosis residences geocoded, and as a result, excluded a small number of subjects born outside of California (2.9 percent) and those with residences that could not be geocoded for some reason (~10 percent of subjects' birth or diagnosis residences). Furthermore, this analysis does not evaluate changes in residential attributes for residences other than those at birth and diagnosis.
Our focus on a population-based series of childhood leukemia cases enrolled in the NCCLS as opposed to the controls was guided by our concern that the control group may be artificially less mobile compared to the case source population as a result of our control selection strategy. In preliminary analyses, cases do appear to be more mobile than controls, however, the degree to which this difference in mobility can be attributed to our selection strategy has not been determined. Although the two groups differed with respect to total number of residences, the factors associated with residential mobility among controls were largely similar to those presented here for cases. In addition, the concordance and discordance between birth and diagnosis residence with regard to urban/rural status and neighborhood SES did not appear to significantly differ from that of cases.
In summary, results of this descriptive analysis demonstrate this population-based series of cases to be highly residentially mobile, with patterns that differ by demographic subgroups. These results suggest that for epidemiologic studies of children, even very young ones, there are serious implications in using a single residential location to determine area-based exposure assignments. Additionally, in view of the uncertainties about which time periods may be of primary importance for leukemia risk and in order to examine exposures during different time periods, attempts should be made to account for patterns of residential mobility. With full residential histories of the children, area-based measures can be estimated for all residences, and changes in estimates due to residential mobility can be accounted for in the analysis. In the absence of pertinent residential history data in studies where children are suspected to be residentially mobile, one strategy may be to limit the study population to children with certain demographic characteristics that have been associated with less mobility, such as those identified in this analysis.
ACKNOWLEDGEMENTS
This study was supported by grants from the National Institute of Environmental Health Sciences (P42 ES04705 and R01 ES09137) and the National Cancer Institute (R01 CA92674). We thank the participating hospitals and clinical collaborators. We would also like to acknowledge the participants of this study and the Northern California Childhood Leukemia Study staff for their effort and dedication.
List of Abbreviations
- SES
socioeconomic status
- NCCLS
Northern California Childhood Leukemia Study
- GIS
geographic information system
- US
United States
- CA
California
- MSA
metropolitan statistical area
- RR
rate ratios
- CI
confidence interval
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
REFERENCES
- 1.Boscoe FP, Ward MH, Reynolds P. Current practices in spatial analysis of cancer data: data characteristics and data sources for geographic studies of cancer. Int J Health Geogr. 2004;3:28. doi: 10.1186/1476-072X-3-28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Poole C, Greenland S, Luetters C, Kelsey JL, Mezei G. Socioeconomic status and childhood leukaemia: a review. Int J Epidemiol. 2006;35:370–384. doi: 10.1093/ije/dyi248. [DOI] [PubMed] [Google Scholar]
- 3.Raaschou-Nielsen O, Reynolds P. Air pollution and childhood cancer: a review of the epidemiological literature. Int J Cancer. 2006;118:2920–2929. doi: 10.1002/ijc.21787. [DOI] [PubMed] [Google Scholar]
- 4.Alavanja MC, Ward MH, Reynolds P. Carcinogenicity of agricultural pesticides in adults and children. J Agromedicine. 2007;12:39–56. doi: 10.1300/J096v12n01_05. [DOI] [PubMed] [Google Scholar]
- 5.Buffler PA, Kwan ML, Reynolds P, Urayama KY. Environmental and genetic risk factors for childhood leukemia: appraising the evidence. Cancer Invest. 2005;23:60–75. [PubMed] [Google Scholar]
- 6.McNally RJ, Parker L. Environmental factors and childhood acute leukemias and lymphomas. Leuk Lymphoma. 2006;47:583–598. doi: 10.1080/10428190500420973. [DOI] [PubMed] [Google Scholar]
- 7.Ma X, Buffler PA, Layefsky M, Does MB, Reynolds P. Control selection strategies in case-control studies of childhood diseases. Am J Epidemiol. 2004;159:915–921. doi: 10.1093/aje/kwh136. [DOI] [PubMed] [Google Scholar]
- 8.Environmental Systems Research Institute I . ArcView. version 3.3 ESRI; Redlands, CA: 2000. [Google Scholar]
- 9.United States Bureau of the Census . TIGER2000 [data file] US Census Bureau; Washington D.C.: 2002. [Google Scholar]
- 10.Geographic Data Technologies GD GDT Home Page. 2002 Available at http://www.geographic.com/home/index.cfm.
- 11.Navigational Technologies NavTech Home Page. 2002 Available at http://www.navtech.com/
- 12.United States Bureau of the Census . Census of population and housing, 1990: Modified age/race, sex and hispanic origin (MARS) state and county file [data file] US Bureau of the Census; Washington, D.C.: 1992. [Google Scholar]
- 13.Reynolds P, Hurley SE, Quach AT, Rosen H, Von Behren J, Hertz A, et al. Regional variations in breast cancer incidence among California women, 1988-1997. Cancer Causes Control. 2005;16:139–150. doi: 10.1007/s10552-004-2616-5. [DOI] [PubMed] [Google Scholar]
- 14.Alexander FE, McKinney PM, Cartwright RA. Migration patterns of children with leukaemia and non-Hodgkin's lymphoma in three areas of northern England. J Public Health Med. 1993;15:9–15. doi: 10.1093/oxfordjournals.pubmed.a042826. [DOI] [PubMed] [Google Scholar]
- 15.Canfield MA, Ramadhani TA, Langlois PH, Waller DK. Residential mobility patterns and exposure misclassification in epidemiologic studies of birth defects. J Expo Sci Environ Epidemiol. 2006;16:538–543. doi: 10.1038/sj.jes.7500501. [DOI] [PubMed] [Google Scholar]
- 16.Khoury MJ, Stewart W, Weinstein A, Panny S, Lindsay P, Eisenberg M. Residential mobility during pregnancy: implications for environmental teratogenesis. J Clin Epidemiol. 1988;41:15–20. doi: 10.1016/0895-4356(88)90004-2. [DOI] [PubMed] [Google Scholar]
- 17.Shaw GM, Malcoe LH. Residential mobility during pregnancy for mothers of infants with or without congenital cardiac anomalies: a reprint. Arch Environ Health. 1992;47:236–238. doi: 10.1080/00039896.1992.9938355. [DOI] [PubMed] [Google Scholar]
- 18.Reynolds P, Elkin E, Scalf R, Von Behren J, Neutra RR. A case-control pilot study of traffic exposures and early childhood leukemia using a geographic information system. Bioelectromagnetics. 2001;Suppl 5:S58–68. doi: 10.1002/1521-186x(2001)22:5+<::aid-bem1024>3.3.co;2-0. [DOI] [PubMed] [Google Scholar]
- 19.Fell DB, Dodds L, King WD. Residential mobility during pregnancy. Paediatr Perinat Epidemiol. 2004;18:408–414. doi: 10.1111/j.1365-3016.2004.00580.x. [DOI] [PubMed] [Google Scholar]
- 20.Raaschou-Nielsen O, Obel J, Dalton S, TjOnneland A, Hansen J. Socioeconomic status and risk of childhood leukaemia in Denmark. Scand J Public Health. 2004;32:279–286. doi: 10.1080/14034940310022214. [DOI] [PubMed] [Google Scholar]
- 21.Smith A, Roman E, Simpson J, Ansell P, Fear NT, Eden T. Childhood leukaemia and socioeconomic status: fact or artefact? A report from the United Kingdom childhood cancer study (UKCCS) Int J Epidemiol. 2006;35:1504–1513. doi: 10.1093/ije/dyl193. [DOI] [PubMed] [Google Scholar]
- 22.Gunier RB, Hertz A, Von Behren J, Reynolds P. Traffic density in California: socioeconomic and ethnic differences among potentially exposed children. J Expo Anal Environ Epidemiol. 2003;13:240–246. doi: 10.1038/sj.jea.7500276. [DOI] [PubMed] [Google Scholar]
- 23.Neumann CM, Forman DL, Rothlein JE. Hazard screening of chemical releases and environmental equity analysis of populations proximate to toxic release inventory facilities in Oregon. Environ Health Perspect. 1998;106:217–226. doi: 10.1289/ehp.98106217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Perlin SA, Sexton K, Wong DW. An examination of race and poverty for populations living near industrial sources of air pollution. J Expo Anal Environ Epidemiol. 1999;9:29–48. doi: 10.1038/sj.jea.7500024. [DOI] [PubMed] [Google Scholar]