Session: Health 2
Author: Nancy Thorpe
This hypothesis-generating study explored spatial patterns of childhood cancers in Maryland and investigated their potential associations with herbicides and nitrates in groundwater. Cancer data were obtained from the Maryland Cancer Registry (MCR) for bone, brain, leukemia, and lymphoma, for ages 0 – 17, during the years 1992 – 1998. Cancer clusters and relative risks were generated. Higher relative risk areas and potential clusters appeared in several counties. Contingency table analysis indicated potential association with several herbicides and nitrates. Cancer rates for the four types had an Odds Ratio OR = 1.93 with exposure to atrazine, and an OR = 1.54 for metolachlor. Mixtures of three compounds gave an OR = 7.56. A potential association was indicated between leukemia and nitrates, OR = 1.81, and bone cancer with metolachlor, OR = 2.26. These results gave insight to generate a hypothesis of the potential association between exposure to these herbicides and nitrates and specific types of childhood cancer.
In the United States, cancer is the second leading cause of death in those 17 years old or younger (NCI 1997). Of these childhood cancer cases, approximately 33% have leukemia, 20% have brain and spinal cord cancers, 4% have non-Hodgkin’s lymphoma, and about 2.4% have bone cancer (ACS 1997). These four types of cancer have been linked to environmental factors, particularly the increased use of agricultural and household pesticides (Daniels et al. 1997, Zahm and Ward 1998). Research has indicated the four most commonly used agricultural herbicides, atrazine, simazine, alachlor, and metolachlor, and nitrates have been associated with chromosomal aberrations, tumor growth, non-Hodgkin’s lymphoma, leukemia, and breast, bone and brain cancers (Alexson 1987, Zahm and Vineia 1988, Schwartz et al. 1988, Donna et al. 1989, Pinter et al. 1989, Tsutsui et al. 1989, Hajek et al. 1989 and 1993, Weisenburger 1990, L’Haridon et al. 1992, Meisner et al. 1992, Zahm et al. 1993, Ward et al. 1996, Kettles et al. 1997, Sathiakumar and Delzell 1997, Taets el al. 1998, Hardell et al. 1998, Hardell and Eriksson 1999, Sanderson et al. 2001).
The purpose of this hypothesis-generating study is to explore the spatial patterns of various types of childhood cancers and investigate potential correlations between them and the distribution of nitrates and the four most commonly detected herbicides in groundwater supplies throughout Maryland. Geographic Information Systems (GIS) methods, specifically ArcView 3.2 and Spatial Analyst 1.1 (Esri 1999), are the main tools of analysis for this project. Maryland Cancer Registry (MCR) data on childhood cancers along with the Census Bureau regional density maps are used to calculate and create relative risk and cluster analysis maps.
Childhood cancers were chosen for this study due to the unique differences of children from adults. Children detoxify chemicals differently and have higher metabolism rates than adults. This may allow chemicals to stay in their bodies longer while their cells reproduce faster, resulting in more chromosomal damage. The diseases in children have shorter latency periods than adults, which makes it easier to identify exposure to environmental toxins.
Computerized cancer case reports were obtained from the Maryland Cancer Registry, Maryland Department of Health and Mental Hygiene (DHMH). A request was made for the number of new childhood cancers (ages 0 – 17) reported in Maryland during the years 1992 – 1998 for four specific types of cancer, leukemia, non-Hodgkin’s lymphoma, brain and bone cancer. Each case report included the ICD-O-2 coded cancer diagnosis, age at diagnosis, gender, race and street address with zip-code.
Using the street addresses, each case was geocoded and a point theme of cases was generated. For patient confidentiality, all files and maps that contain the actual coordinates or that showed actual locations of the cases were destroyed. Once the cases were placed on the map, these points were aggregated to polygons of geographic units of both counties and census tracts. Using Census Bureau population data for these units, the incidence rate of the various cancers were computed for both county and census tract levels.
To overcome the problem of statistical significance with small numbers and to maintain patient confidentiality, the SaTScan software (SaTScan 1998), developed by the National Cancer Institute, was employed to test for statistically significant clusters across the state. This software generated information that was easily input into ArcView to create maps of relative risk and disease clusters. The program investigated potential clusters at both the county and census tract levels for all the four types of cancers together and separately for all ages 0 – 17 and also analyzed for the covariates of age, gender, and race. Various maps were generated to show the spatial distribution of the various childhood cancers, but only those at the county level are reported.
Contingency tables with chi-square analysis were used to investigate any potential associations between childhood cancer incidence and the spatial distribution of the four herbicides and nitrates in groundwater. Analysis of the tables was performed using GraphPad Prism version 3.02 for Windows, 1999, GraphPad Software, San Diego, CA, U.S.A.. In a previous study, the spatial distributions of herbicides and nitrates were determined and maps were generated of two-mile buffer zones around wells with any detectable concentrations, thus delineating the areas of exposure to the herbicides such as atrazine (Figure 1). By aggregating the points of case locations to the two-mile buffer zone polygons, the number of cases living within the exposed areas was calculated. Cases living in areas outside the buffer zones were the unexposed cases. Using Census data at the block group level, the total exposed and non-exposed disease-free population was calculated by overlaying the census block group layer map with the two-mile buffer zone layer. These calculated values of exposed and non-exposed populations were input into the contingency tables and the odds ratios were generated.
A total of 689 new cases of childhood cancers, ages 0 – 17 years old, were reported between the years 1992 – 1998 for the four types of childhood cancers specified. The overall incidence rate at the state level for these cancer cases is 8.5/100,000 per year, which is less than the nationwide rate of 14.2 per 100,000 (SEER 1997).
Results from the SaTScan analysis at the county level indicate the most likely cluster for the four types of cancer studied is Worcester County, although the cluster is not statistically significant. Eleven other counties, including Anne Arundel and Baltimore Counties in the Piedmont Province (central Maryland) and Caroline, Harford and Queen Anne’s Counties in the Coastal Plain Province (eastern Maryland) are listed as secondary clusters (Figure 2). After adjusting for the covariate of age, gender, and race, Worcester County was still the most significant cluster. Maps of the relative risk for the four cancer types were also generated, indicating Worcester County with the highest relative risk for these cancers (Figure 3). Spatial distribution maps of the individual cancer types were also generated but are not reported here. Leukemia showed potential clusters in the eastern region of the state, bone and brain cancers in the middle region, and non-Hodgkin’s lymphoma in the western region.
Results from the contingency table and chi-square analyses are listed in Tables 1 and 2, which list the odds ratio, confidence interval, and p-values for exposure to the herbicides and nitrates. The results show only a few associations are statistically significant.
Table 1. Risk of selected childhood cancers in Maryland (ages 0 –17 years old) by exposure to all detectable concentrations of selected herbicides and nitrates.
All ages 1.18 (0.98 –1.43) 0.0889
0 – 4 years 1.26 (0.94 –1.69) 0.1392
5 – 17 years 1.14 (0.89 –1.46) 0.3331
Atrazine 1.10 (0.78 – 1.56) 0.6577
Simazine 1.14 (0.75 – 1.75) 0.6193
Metolachlor 1.54 (1.14 – 2.07) 0.0061
Alachlor 1.49 (1.02 – 2.17) 0.0506
Nitrates 1.49 (1.22 – 2.83) 0.0001
Nitrate/Atrazine 1.18 (0.63 – 2.21) 0.7224
Nitrate/Atrazine/ 7.56 (4.16 – 13.73) <0.0001
Nitrate/Metolachlor/ 5.31 (2.84 – 9.93) <0.0001
Table 2. Risk of selected childhood cancers in Maryland (ages 0 – 17 years old) by exposure to all detectable concentrations of selected herbicides and nitrates; categorized by cancer type.
Bone 1.29 (0.70 – 2.37) 0.5196
Non-Hodgkins 1.03 (0.75 – 1.42) 0.9167
Brain 1.01 (0.54 – 1.87) 0.8906
Leukemia 1.35 (1.02 – 1.78) 0.0412
Bone/Metolachlor 2.26 (0.97 – 5.24) 0.0995
Bone/Atrazine 1.89 (0.76 – 4.70) 0.2821
Bone/Nitrates 1.28 (0.63 – 2.59) 0.6233
Brain/Nitrates 1.23 (0.86 – 1.75) 0.2913
Lymphoma/Nitrates 1.41 (0.74 – 2.68) 0.3877
Leukemia/Nitrates 1.81 (1.35 – 2.42) <0.0001
Leukemia/Atrazine 1.43 (0.89 – 2.30) 0.1850
Leukemia/Metolachlor 1.48 (0.93 – 2.36) 0.1256
Although the cluster analysis at the county level did not show any statistically significant clusters, the higher relative risk values indicate potential problem areas. One possible reason for the lack of significant clusters is that the number of cases is too small, thus making the analysis unstable and non-significant. Also, analysis of large geographic areas introduces errors of wide area non-uniformity. Cluster analysis at the census tract level resulted in a few statistically significant clusters that were located in the counties indicated as the most significant clusters.
The contingency table results indicate statistically significant odds ratio values for five different exposures. Children exposed to a mixture of nitrates, atrazine, and metolachlor have a 7.6 times greater chance of developing one of the four types of cancer studied than unexposed children; a 5.3 greater chance with exposure to a mixture of nitrates, simazine, alachlor, and metolachlor; and a 1.5 greater chance with exposure to metolachlor alone, nitrates alone, or alachlor alone. The high odds ratio values for the mixtures might indicate a synergistic effect of these compounds. Leukemia indicates an association with any exposure to the five compounds with 1.4 times the risk than non-exposed, and a 1.8 time the risk with exposure to nitrates alone. Bone cancer is 2.3 times the risk with exposure to metolachlor. No other specific cancer type was statistically significant, which might be due to the number of cases being too small to five conclusive results and the need of a larger population size for statistical significance.
Key findings for this study include the following: 1) the overall incidence rate for these cancer cases is 8.5 per 100,000; 2) Maryland counties that have higher relative risks for these various types of childhood cancer are located mainly in the Coastal Plain and Piedmont Provinces; and 3) the county with the greatest relative risk of cancer cluster for the childhood cancer types investigated in this study is Worcester County. It is strongly recommended that a more in-depth study be performed on childhood cancers for this county and others in the eastern part of the state.
The integration of the GIS software, ArcView, and the program SaTScan aided in the task of analyzing the cancer data. The SaTScan program was easy to use and understand. The use of these programs allowed for accurate analysis of pin-pointing potential problem areas in the state while maintaining patient confidentiality.
Contingency table analysis indicates possible associations of exposure to low-levels of atrazine, simazine, alachlor, metolachlor and nitrates, alone and in combination, with various childhood cancers, particularly leukemia and bone cancer. Combinations of the herbicides and nitrates indicate a possible synergistic affect of these compounds.
The use of contingency tables did not allow for analysis of various covariates, such as age, gender and race. The study itself was not designed to incorporate variables such as parental exposure, timing of exposure, not actual source of drinking water. GIS allowed for calculations of exposure amounts that are as accurate as possible, but some assumptions were made in the process that could have introduced errors. Based on this type of ecologic study, the given limitations, and possible sources of error, the method of analysis used for associating childhood cancers and exposures was useful and as accurate as possible.
ACS - American Cancer Society. 1997. Cancer Facts and Figures - 1997. ACS website. http://www.cancer.org/statistics/97cff/97chilr.html.
Axelson, O. 1987. Pesticides and cancer risks in agriculture. Medical Oncology Tumor Pharmacotherapy. Vol 3. No 3-4. pp. 207-217.
Daniels, Julie, Andrew Olshan, and David Savitz. 1997. Pesticides and Childhood Cancers. Environmental Health Perspectives. Vol.105. No.10. pp. 1068-1077.
Donna, A., P. Crosignani, F. Robutti, P. Betta, R. Bocca, N. Mariani, F. Ferrario, R. Fissi, and F. Berrino. 1989. Triazine herbicides and ovarian epithelial neoplasms. Scandinavian Journal Work Environmental Health. Vol. 15. pp. 47-53.
Esri. 1999. Environmental Systems Research Institute, Inc. Redlands, CA.
Hajek, R.A., S. Pathak, A. Boddie, and L. Jones. 1989. Aneuploidy of mouse cervicovaginal epithelium induced by perinatal estrogen treatment. Proceedings of the American Association of Cancer Research. Vol. 30. p. 299.
Hajek, R.A., N.T. Van, D.A. Johnston, and L.A. Jones. 1993. In vivo induction of increased DNA ploidy of mouse cerviovaginal epithelium by neonatal estrogen treatment. Biological Reproduction. Vol. 49. pp. 908-917.
Hardell L. and M. Eriksson. 1999. A case-control study of non-Hodgkin’s lymphoma and exposure to pesticides. Cancer. Vol. 85. No 6. pp. 1353-60.
Hardell L., G. Lindstrom, B. van Bavel, M. Fredrikson, and G. Liljegren. 1998. Some Aspects of the Etiology of Non-Hodgkin’s Lymphoma. Environmental Health Perspectives. 106 (Supplement 2) pp. 679-681.
Kettles, M.K., S.R. Browning, T.S. Prince, and S.W. Horstman. 1997. Triazine herbicide exposure and breast cancer incidence: an ecologic study of Kentucky counties. Environmental Health Perspectives. Vol. 105. No. 11. pp. 1222-1227.
L’Haridon J., M. Fernandez, V. Ferrier, and J. Bellan. 1992. Evaluation of the Genotoxicity of N-nitrosoatrazine, N-nitrosodiethanolamine and Their Precursors In Vivo Using the Newt Micronucleus Test. Water Resources. Vol. 2.
Meisner, L.F., D.A. Belluck, and B.D. Roloff. 1992. Cytogenetic effects of alachlor and/or atrazine in vivo and in vitro. Environmental Molecular Mutagenicity. Vol.19. No.1. pp. 77-82.
National Cancer Institute (NCI). 1997. SEER 1973-1994 Public-Use CD-ROM.
Pinter, A., G. Torok, M. Borzsonyi, A. Surjan, M. Csik, Z. Kelecsenyi, Z. Kocsis. 1990. Long-term carcinogenicity bioassay of the herbicide atrazine in F344 rats. Neoplasma. Vol. 37. No 5. pp. 533-539.
Sanderson, J. Thomas, Robert Letcher, Marjoke Heneweer, J.P. Giesy, and Martin van den Berg. 2001. Effects of Chloro-S-Triazine Herbicides and Metabolites on Aromatase Activity in Various Human Cell Lines and on Vitellogenin Production in Male Carp Hepatocytes. Environmental Health Perspectives. Vol. 109. Number 10. pp. 1027-1032.
Sathiakumar, N., and E. Delzell. 1997. A review of epidemiologic studies of triazine herbicides and cancer. Critical Reviews in Toxicology. Vol. 27. No. 6. pp. 599-612.
SaTScan Software. 1998. National Institutes of Health. Division of Cancer Prevention. http://srab.cancer.gov/satscan
Schwartz, D., and J. Loderfo. 1988. Congenital Limb Reduction Defects in the Agricultural Setting. American Journal of Public Health. June. Vol. 78. No. 6. pp. 654-57.
SEER (Surveillance, Epidemiology, and End Results). 1997. SEER 1973-1994 Public-Use CD-ROM. Cancer Incidence Public-Use Database and Cancer Statistics Review. National Cancer Institute.
Taets, Carrie, Susanne Aref, and A. Lane Rayburn. The Clastogenic Potential of Triazine Herbicide Combinations Found in Potable Water Supplies. Environmental Health Perspectives. Vol. 6, No. 4. pp. 197-201.
Tsutsui, T., N. Suzki, J. McLachlan, and J. Barrett. 1989. Induction by estrogens and their metabolites of morphological transformation and chromosome aberrations in cultured Syrian hamster embryo (SHE) cells. Proceedings of the American Association of Cancer Research. Vol. 30. pp. 154.
Ward, M.H., S.D. Mark, K.P. Cantor, D.D. Weisenburger, A. Correa-Villasenor, and S.H. Zahm. 1996. Drinking water nitrate and risk of non-Hodgkin’s lymphoma. Epidemiology. September. Vol. 7. No 5. pp. 465-71.
Weisenburger, D.D. 1990. Environmental epidemiology of non-Hodgkin’s lymphoma in eastern Nebraska. American Journal of Industrial Medicine. Vol. 18. No. 3. pp. 303-305.
Zahm, S., and P. Vineia. 1988. Immunosuppressive Effects of Dioxin in the Development of Kaposi’s Sarcoma and Non-Hodgkin’s Lymphoma. Lancet. January.
Zahm, S., and M. Ward. 1998. Pesticides and Childhood Cancer. Environmental Health Perspectives. Vol. 106. Supplement 3. pp. 893-908.
Zahm, S., D.D. Weisenburger, R.C. Saal, J.B. Vaught, P.A. Babbitt, and A. Blair. 1993. The Role of Agricultural Pesticide Use in the Development of Non-Hodgkin’s Lymphoma in Women. Archives of Environmental Health. September/October. Vol. 48. No. 5. pp. 353-358.
Hagerstown Community College
11400 Robinwood Drive
Hagerstown, MD 21742