Resumes & CVs

A study of adverse birth outcomes and agricultural land use practices in Missouri

Description
Missouri is an agriculturally intensive state, primarily growing corn and soybeans with additional rice and cotton farming in some southeastern counties. Communities located in close proximity to pesticide-treated fields are known to have increased
Categories
Published
of 7
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  A study of adverse birth outcomes and agricultural land use practicesin Missouri Kirsten S. Almberg a, n , Mary Turyk a , Rachael M. Jones b , Robert Anderson a , Judith Graber a,e ,Elizabeth Banda a , Lance A. Waller c , Roger Gibson d , Leslie T. Stayner a a University of Illinois at Chicago, School of Public Health, Epidemiology and Biostatistics Division, 1603W. Taylor Street, Chicago,IL 60607, United States of America b University of Illinois at Chicago, School of Public Health, Environmental and Occupational Health Sciences Division, 2121W. Taylor Street, Chicago,IL 60612, United States of America c Department of Biostatistics and Bioinformatics, Rollins School of Public Health, Emory University, 1518 Clifton Road NE, Atlanta,GA 30322, United States of America d Missouri Department of Health and Senior Services, Division of Community and Public Health, Bureau of Environmental Health, P.O.Box 570, Jefferson City, MO 65109, United States of America e Environmental and Occupational Health Sciences Institute, Rutgers University, 170 Frelinghuysen Road, Piscataway, NJ 08854, United States of America a r t i c l e i n f o  Article history: Received 25 July 2013Received in revised form17 March 2014Accepted 17 June 2014Available online 27 September 2014 Keywords: Low birth weightPreterm birthsAgricultural land usePoissonGEE a b s t r a c t Background:  Missouri is an agriculturally intensive state, primarily growing corn and soybeans withadditional rice and cotton farming in some southeastern counties. Communities located in closeproximity to pesticide-treated  fi elds are known to have increased exposure to pesticides and may beat increased risk of adverse birth outcomes. The study aims were to assess the relationship betweencounty-level measures of crop-speci fi c agricultural production and adverse birth outcomes in Missouriand to evaluate the most appropriate statistical methodologies for doing so. Methods:  Potential associations between county level data on the densities of particular crops and lowbirth weight and preterm births were examined in Missouri between 2004  2006. Covariates consideredas potential confounders and effect modi fi ers included gender, maternal race/ethnicity, maternal age atdelivery, maternal smoking, access to prenatal care, quarter of birth, county median household income,and population density. These data were analyzed using both standard Poisson regression models as wellas models allowing for temporal and spatial correlation of the data. Results:  There was no evidence of an association between corn, soybean, or wheat densities with lowbirth weight or preterm births. Signi fi cant positive associations between both rice and cotton densitywere observed with both low birth weight and preterm births. Model results were consistent usingPoisson and alternative models accounting for spatial and temporal variability. Conclusions:  The associations of rice and cotton with low birth weight and preterm births warrantfurther investigation. Study limitations include the ecological study design and limited availablecovariate information. &  2014 Elsevier Inc. All rights reserved. 1. Introduction Herbicides were applied to over 95% of the corn and soybeansgrown in the U.S. in 2001 (U.S. Environmental Protection Agency,2012). Corn and soybeans are the most common crops grown inMissouri, and the proportion of land dedicated to growing soy-beans and corn within a county can be as high as 58% and 37%,respectively (Fig. 1). Although agricultural chemicals have beenextremely useful in boosting crop production, they can migrateaway from application sites through the soil, water, air, and ontopersonal belongings (Chester and Ward, 1984; Rull et al., 2006).As a result, families and communities not occupationally involvedContents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/envres Environmental Research http://dx.doi.org/10.1016/j.envres.2014.06.0160013-9351/ &  2014 Elsevier Inc. All rights reserved.  Abbreviations:  (CDC), Centers for Disease Control and Prevention; (CI), con fi denceinterval; (EPHTN), Environmental Public Health Tracking Network; (GEE), gener-alized estimating equations; (MDHSS), Missouri Department of Health and SeniorServices; (NASS), National Agricultural Statistics Survey; (PNC), prenatal care; (RR),relative risk; (SES), socio-economic status; (SGA), small for gestational age; (USDA),United States Department of Agriculture; (WIC), Women, Infants, and Children n Corresponding author. E-mail addresses:  almberg@uic.edu (K.S. Almberg),mturyk1@uic.edu (M. Turyk), rjones25@uic.edu (R.M. Jones), tallbob@uic.edu (R. Anderson), jmg502@eohsi.rutgers.edu (J. Graber), ebanda2@uic.edu (E. Banda), lwaller@emory.edu (L.A. Waller), roger.gibson@health.mo.gov (R. Gibson).Environmental Research 134 (2014) 420 – 426  with agricultural pesticide use may also be exposed (Lu et al.,2000; Fenske et al., 2002).Atrazine and glyphosate are two commonly used herbicides inthe Midwest. Atrazine is a pre-emergence broadleaf herbicidemost commonly applied to corn  fi elds. Glyphosphate is commonlyused for weed control particularly on crops that are geneticallymodi fi ed to be resistant to this herbicide. Atrazine is one of themost commonly detected contaminants in surface water in the U.S.and particularly in the Midwest corn-belt region (U.S. Environ-mental Protection Agency, 2007).Nitrogen fertilizers are also commonly applied to agricultural fi elds in the U.S. Nitrate (NO 3- ) from fertilizers, livestock manure,and humanwaste is also a common contaminant of drinking watersources in agricultural regions (Nolan et al., 1997; Nolan and Hitt,2006).To date, epidemiological studies that have explored the asso-ciation of community-based exposure to atrazine with adversebirth outcomes have had inconsistent results. An increased risk of delivering small for gestational age (SGA) babies was found inIowa communities with drinking water contaminated by atrazine(Munger et al., 1997). An increased risk of SGA was also seen inIndiana when atrazine was present in drinking water during thethird trimester (Ochoa-Acuña et al., 2009). Rinsky et al. (2012), observed a signi fi cantly increased risk of preterm births inKentucky counties with the highest versus the lowest atrazinelevels in drinking water. Conversely, Ochoa-Acuña et al. (2009)reported no association between atrazine in drinking water andpreterm births. Maternal exposure to atrazine during pregnancyhas been associated with lower birth weight, length, and headcircumference (Chevrier et al., 2011), however, other studies haveshown no association between atrazine exposure during preg-nancy and birth weight (Villanueva et al., 2005; Sathyanarayanaet al., 2010).While there is evidence from animal studies that suggestsglyphosate may be genotoxic and disrupt endocrine function (deCastilhos and Cestari, 2012; Romano et al., 2012), the evidence forsuch an association from human studies is sparse. Savitz et al.,(1997) reported elevated odds of preterm birth among infantswhose fathers had been exposed to glyphosate-based pesticidesprior to and at the time of conception compared to those withoutexposure, although the results were not statistically signi fi cant.More recently, Gasnier et al. (2009) found glyphosate-based Fig. 1.  Crop coverage of corn, soybeans, wheat, rice, cotton, and total crop densities in Missouri, 2005.Data Source: United States Department of Agriculture Quick Stats 2.0, 2010 K.S. Almberg et al. / Environmental Research 134 (2014) 420 – 426   421  herbicides to be genotoxic and disrupt the endocrine system inhuman cells.Effects from exposure to nitrate from drinking water below themaximum contaminant level set by the Environmental ProtectionAgency (MCL  ¼ 10 mg/L) are not well understood. There is someevidence from epidemiologic studies that exposure to nitrate fromdrinking water sources may increase the incidence of neural tubedefects, oral cleft defects, limb de fi ciencies (Brender et al., 2004;Brender et al., 2013), cardiac defects (Cedergren et al., 2002), SGA and spontaneous abortions (Manassaram et al., 2006). Animalstudies also show moderate associations with fetal loss, neonatalmortality, and decreased litters and live births (Manassaram et al.,2006).Many of the previous studies were ecologic studies that failed toaccount for geographic and temporal patterns in regression models(Munger et al.,1997; Cedergren et al., 2002; Schreinemachers, 2003;Villanueva et al., 2005; Rinsky et al., 2012). Geographic correlationsin health outcomes may result from populations in different areasbeing exposed to greater risk or to protective factors. Further, it isimportant to consider the non-independence of observations whenstudying a geographic unit of observation, such as counties. Countiesthat are adjacent would be expected to be more similar in termsof socio-demographic characteristics than counties farther away(Burnett et al., 2001). For these reasons, incorporating spatial andtemporal data into ecologic analyses of environmental data andhealth outcome data may be important.Our primary aim in this study is to evaluate the relationshipbetween county-level measures of agricultural crop productionand low birth weight and pre-term birth in the state of Missouri in2004 – 2006. Agricultural crop production, measured by the densityof crop acreage in each county, was used as a surrogate forcommunity pesticide and fertilizer exposure. Owing to concernabout the in fl uence of geographical and temporal patterns onpossible associations between crop production and the two healthoutcomes, our secondary aim is to examine whether or notrepresentation of geographical and temporal patterns alters sta-tistical associations. Speci fi cally, we compare associations esti-mated by three statistical methods that account for geographicaland/or temporal patterns in different ways. 2. Methods  2.1. Study area The study area was the state of Missouri. The predominant crops produced inMissouri are corn, soybean, and wheat. Rice and cotton are grown in the southeastcorner of the state (Fig. 1).  2.2. Birth outcome data County-level counts of low birth weight and preterm births were provided bythe Missouri Department of Health and Senior Services for the years 2004, 2005,and 2006. Preterm births were de fi ned as infants born prior to 37 weeks gestation.Low birth weight births were de fi ned as infants weighing less than 2500 g at orabove 37 weeks gestation. All birth outcome data was strati fi ed by race/ethnicity,gender, month of birth, year of birth, and county.  2.3. Agricultural land use measurements Agricultural land use density, as measured by the percentage of county landdedicated to the production of speci fi c crops, was used as a surrogate measure forthe potential for community exposure to agricultural chemicals. The acreage of planted corn, soybean, wheat, rice, cotton, and total agricultural crops in eachcounty was obtained from the United States Department of Agriculture (USDA)National Agricultural Statistics Service's Quick Stats 2.0 database (United StatesDepartment of Agriculture, 2010). Annual crop density was calculated by dividingthe area of land planted with a speci fi c crop by the total land area of the county(United States Department of Commerce, 2011).  2.4. Data for other risk factors County-level covariates considered as potential confounders and effect modi- fi ers in this analysis included child's gender, mother's race/ethnicity, mother's ageat delivery, quarter of birth, county median household income, population density,maternal smoking, prenatal care status and percentage of mothers using theWomen, Infants, and Children (WIC) food and nutrition service.Values for race/ethnicity, gender, and month and year of birth were providedwith the health outcomes data from the Missouri Department of Health and SeniorServices. The race categories used in this analysis were non-Hispanic white, non-Hispanic black, non-Hispanic other race, and Hispanic. Quarter of birth was de fi nedas (1) January through March, (2) April through June, (3) July through September,and (4) October through December.County level maternal age groups were obtained from data provided by theCenters for Disease Control Environmental Public Health Tracking Network(EPHTN). The percentage of births born to mothers in each age group wascalculated from data on the number of full term singleton live births born to eachmaternalage group per county divided by the total number of full term singleton live birthsper county. Maternal age at delivery was divided into three categories:  o 20,20  39, and  4 40 years of age.The median household income in each county for years 2004, 2005, and 2006was used as an indicator of socio-economic status (SES) and was obtained from theAmerican FactFinder database (United States Census Bureau, 2010). Populationdensity of each county was used as an indicator of urbanicity. Annual county-levelrates of maternal smoking and access to prenatal care were obtained from theMissouri Department of Health and Senior Services (MDHSS) (Missouri Departmentof Health and Senior Services, 2013).Consistent with other studies (Schreinemachers, 2003; Rinsky et al., 2012), weexcluded counties with large metropolitan centers (population 4 300,000) toreduce potential confounding by unmeasured risk factors. This restriction excludedSt. Louis City and three counties. Although urban counties may have little to noexposure as classi fi ed in this study, urban mothers are likely to have very differentlifestyle and other pregnancy-related factors than rural mothers.  2.5. Statistical methods Poisson regression was used to evaluate the association between crop produc-tion patterns and counts of either low birth weight or preterm births (Method 1).The natural logarithm of the counts of full-term singleton births and the counts of singletons births were used as the offsets for the low birth weight and pretermbirth models, respectively. Poisson regression models were performed with PROCGENMOD using SAS s software, Version 9.2 (SAS Institute Inc., Cary, NC, USA 2002).Two additional model structures were used to account for the temporal and/orgeographic correlation among the county rates. To account for temporal correlation inthe data we used a generalized estimate equation (GEE) Poisson regression approach(Method 2). GEE is a quasi-likelihood estimation procedure that makes weakassumptions about the correlation structure among observations, relying on a  “ work-ing ”  correlation structure for each vector of repeated observations (Zeger and Liang,1986). All GEE Poisson regression models were performed with PROC GENMOD usingSAS software,Version 9.2.Adistancedecay random effectsPoissonmodelthat allowedfor spatially varying random effects was used to account for the highly variable andnon-random geographic distribution of crop production across the state (Method 3).This model accounts for spatial autocorrelation, which occurs when values of alocation depend on those at nearby locations, by using a county adjacency matrix toinform the correlation structure between observations. Distance decay models wereestimated in R using GAMEPHIT software (Burnett et al., 2001).Confounders and effect modi fi ers were examined in each analysis, includingmaternal race/ethnicity, child's gender, county rate of maternal age groups, countymedian income, county population density, county rates of maternal smoking andprenatal care status, quarter of birth, and year. An interaction term between theexposure variables and season of birth was examined to address potential seasonaldifferences in the exposure-outcome relationship. Season of birth was based onmonth of birth and de fi ned as January – March, April –  June, July – August, or Sep-tember – December. Effect modi fi cation was not detected and therefore no interac-tion terms were included in  fi nal models. 3. Results  3.1. Descriptive analysis Crop densities varied widely across counties in Missouri, withcorn and soybeans being the most commonly grown crops (Fig.1).More soybeans than corn were grown, with a median soybeandensity of approximately 14% (62% maximum), compared to amedian corn density of 6% (37% maximum). Considerably less K.S. Almberg et al. / Environmental Research 134 (2014) 420 – 426  422  wheat was grown, with median values of wheat density between0.25%  2%, although some counties had more than 20% wheatdensity during the study years. Rice and cottonwere grown only in fi ve to eight counties. In counties that did grow cotton, the mediandensity was 26.8%. Total agricultural land density was highlyvariable across the state with a median total crop density of 24%(range 0.12  100%). Among those counties producing rice andcotton, median total crop density was 82%).There were a total of 140,329 live births in Missouri between2004 and 2006 (Table 1). A large majority of mothers were non-Hispanic white (90%). There were slightly more males born thanfemales (51% versus 49%). The proportion of live births across thethree study years was roughly equal, as was the proportion of livebirths in each quarter of the year. Preterm births comprised nearly13% of all live singleton births. Roughly 2% of live full-termsingleton births in the study area were classi fi ed as low birthweight. The median percentage of mothers who smoked duringpregnancy was 25% (range: 10  65%), and the median percentageof women who received no prenatal care during pregnancy was13% (range: 0 – 97%).  3.2. Regression analyses Results from multivariable models of low birth weight andpreterm births are presented in Tables 2 and 3. All relative riskshave been computed to show the effect of a 10% increase in cropdensity.Cotton density was signi fi cantly associated with increased riskof low birth weight births across all three methodological appro-aches (Method 1 RR [95%CI]: 1.13 [1.06  1.19]; Method 2 RR [95%CI]: 1.13 [1.07  1.18]; Method 3 RR [95%CI]: 1.12 [1.05 – 1.19]). Ricedensity was also signi fi cantly associated with an increase in risk of low birth weight births (Method 1 RR [95%CI]: 1.22 [1.07  1.40];Method 2 RR [95%CI]: 1.22 [1.06  1.40]; Method 3 RR [95%CI]: 1.22[1.03  1.44]). Total crop density was signi fi cantly associated withan increase in risk of low birth weight in only Poisson regressionmodels (Method 1), although the effect size was small (Method1 RR [95%CI]: 1.02 [1.00  1.05]). No association was observedbetween corn, soybean, or wheat density and low birth weightbirths.Rice density was signi fi cantly associated with preterm birthsacross all three methodological approaches (Method 1 RR [95%CI]:1.21 [1.15  1.30]; Method 2 RR [95%CI]: 1.21 [1.16  1.28]; Method3 RR [95%CI]: 1.15 [1.03 – 1.28]) (Table 2). Cotton density was alsosigni fi cantly associated with preterm births, although the magni-tude of effect was smaller than that of rice density (Method 1 RR [95%CI]: 1.06 [1.03  1.09]; Method 2 RR [95%CI]: 1.06 [1.04  1.09];Method 3 RR [95%CI]: 1.05 [1.00 – 1.10]). There was a signi fi cantlyreduced risk of preterm births with increasing corn density inPoisson regression models (Method 1), but this effect was wea-kened in GEE and distance decay models (Methods 2 and 3). Therewas no observed association between soybean, wheat, or totalcrop densities and preterm births.There was strong geographic overlap among Missouri countiesthat produce rice and cotton (  ρ ¼ 0.56,  p o 0.0001). Regressionmodels containing both cotton and rice densities were  fi tted todetermine if the observed associations with cotton and rice wereindependent of one another (Table 3). In these models, only cottonremained signi fi cantly associated with low birth weight birthsacross all three methodological approaches (Method 1 RR [95%CI]:1.13 [1.04  1.18]; Method 2 RR [95%CI]: 1.11 [1.06  1.17]; Method3 RR [95%CI]: 1.11 [1.03 – 1.18), while rice density was only signi fi -cantly associated with low birth weight in GEE models (Method2 RR [95%CI]: 1.10 [1.01  1.19]). Only rice density remainedsigni fi cantly associated with preterm births in models containingboth rice and cotton across all three methodological approaches(Method 1 RR [95%CI]: 1.19 [1.11  1.27]; Method 2 RR [95%CI]: 1.19[1.11  1.26]; Method 3 RR [95%CI]: 1.13 [1.01 – 1.26). Cotton densitywas only marginally signi fi cantly associated with preterm births inany of the models containing both rice and cotton densities. 4. Discussion We observed signi fi cant positive associations between bothrice and cotton density separately and both low birth weight andpreterm births in this study. Only eight out of 115 counties inMissouri produced rice between 2004 and 2006. Five countiesproduced cotton in Missouri, all of which also produced rice. Inmodels that included both rice and cotton density, cotton densityremained a signi fi cant predictor of low birth weight births, butrice was no longer signi fi cantly associated with this outcome.Rice remained a signi fi cant predictor of preterm births in modelsthat included both crops, while the effect of cotton density wasweakened. The changes in strength of association suggest that itmay be hard to independently estimate the association of rice orcotton density with either outcome because of their strongcorrelation. It is not possible to ascribe the positive  fi nding of adverse birth outcomes being associated with rice and cottondensities to exposure from a particular pesticide. Pesticidesregistered for use on rice in Missouri include acephate, deltame-thrin, and imidacloprid. Glyphosate, azadirachtin, capsaicin, anddicamba are registered pesticides for use on cotton in Missouri(Missouri Department of Agriculture, 2013). In addition, the eightcounties that produce rice, cotton, or both are different from the  Table 1 Maternal and infant characteristics for births ( N  ¼ 140,329) occurring from 2004 to2006 in rural a Missouri counties.  Variable  N %BirthsLive births 140,329 100Live singletons 136,057 97Live term singletons 120,751 86Maternal raceNon-Hispanic white 125,952 90Non-Hispanic black 5398 4Hispanic 5900 4Other/Unknown 3079 2Maternal ageMedian % of mothers aged  o 19  –  13Median % of mothers aged 40 – 54  –  1GenderMale 72,116 51Female 68,213 49Season of birth Jan – Mar 33,610 24Apr –  Jun 34,730 25 Jul – Sep 36,942 26Oct – Dec 35,047 25Year2004 45,451 322005 46,629 332006 48,249 34Median percentage of maternal smoking  –  25Median percentage of mothers who received no prenatal care  –  13Percentage of mothers enrolled in WIC  –  51OutcomesPreterm births 17,630 13Low birth weight births 3007 2Data sources: Missouri Department of Health and Senior Services (MDHSS, 2013)and CDC Environmental Public Health Tracking Network. a Rural is de fi ned as a county with a population of   o 300,000. K.S. Almberg et al. / Environmental Research 134 (2014) 420 – 426   423  rest of the state on key demographic variables. These countieshave lower median income and higher rates of maternal smoking,prenatal WIC use, and teenage pregnancies.A signi fi cant negative association between corn density andpreterm births was observed. This association was signi fi cant onlyin normal Poisson regression (Method 1). There was no consistentevidence of an association observed between soybean, wheat, ortotal agricultural crop densities and either low birth weight orpreterm births in this study (Table 2). The wheat density resultsare consistent with those of  Schreinemachers (2003), who foundno effect of wheat coverage on either of these outcomes. The lackof associations between corn, soybean, and wheat densities andthe two health outcomes in this study may be a result of the crudenature of how the exposures were estimated. Crop density andassociated exposures to agricultural chemicals may vary substan-tially within counties and across the state, such that county-levelanalyses may obscure associations present at smaller geographicalscales. For example, Xiang et al., (2000) found that low birthweight was associated with total crop production, and speci fi callysugar beets and corn, within 300 m of the mother's residence.Future studies should pursue a more geographically re fi nedapproach of measuring residential crop proximity and density.Due to the geographic distribution of the exposure variablesand the between-year correlation of the observations in the study,three statistical approaches were used to explore associationsbetween crop density and the two birth outcomes. Each metho-dological approach - Poisson, GEE, and distance decay - yieldedsimilar results (Tables 2 and 3). The distance decay method allowsfor spatially varying random effects (in this case county is therandom effect), and produced similar point estimates, but widercon fi dence intervals than the other methodological approaches,in general. The rationale for using the distance decay model isthat adjacent counties are expected to be more alike with respectto unmeasured confounders than counties farther away fromeach other. The consistency of results across models types wasunexpected. The consistency of the results from the distancedecay model (Method 3) with those from Methods 1 and 2 suggestthat after controlling for the demographic and socio-economiccovariates, there is little additional variation by county. Conse-quently, the most parsimonious modeling approach, normalPoisson regression, may be most appropriate for modeling thesedata.Pesticide applications occur at speci fi c times throughout theyear, and as such, there is a conceptual basis to suggest that cropproduction might produce seasonally elevated rates of adversebirth outcomes. To test this hypothesis, we used a season byexposure interaction term to test the effect of seasonal pesticideapplication on low birth weight and preterm births; howeverthese coef  fi cients were not statistically signi fi cant. Strati fi ed ana-lyses of each crop density and birth outcome model by quarter of birth yielded consistent estimates of association across seasons forall but one model (See Supplemental Materials Table 4). Theassociation between rice density and low birth weight wasstrongest and signi fi cant for births occurring April to June. Theassociation between cotton and low birthweight was strongest infor births occurring between Jan and June. The timing of exposureto crops (and pesticide application) during gestationwas uncertainin this analysis because the exposure variables were annualmeasures, though the health outcomes were seasonal measures.Therefore, the results of this study do not provide evidence with  Table 2 Relative risks (RR) from adjusted single-exposure regression of low birth weight and preterm births on corn, soybean, wheat, rice, cotton, and total crop densities in Missouri,2004 – 2006, using three modeling approaches. Relative risks represent risk associated with 10% increase in crop density. Outcome Exposure variable Poisson  ( Method  1)  GEE  ( Method  2)  Distance Decay   ( Method  3) RR (95% CI) RR (95% CI) RR (95% CI) Low birth weight a Corn 1.01 (0.95 – 1.06) 1.01 (0.95 – 1.07) 1.00 (0.94 – 1.07)Soybean 1.02 (0.99 – 1.06) 1.02 (0.99 – 1.06) 1.02 (0.97 – 1.06)Wheat 1.05 (0.94 – 1.18) 1.05 (0.95 – 1.17) 1.05 (0.91 – 1.20)Rice 1.22 (1.07 – 1.40) 1.22 (1.06 – 1.40) 1.22 (1.03 – 1.44)Cotton 1.13 (1.06 – 1.19) 1.13 (1.07 – 1.18) 1.12 (1.05 – 1.19)All Crops 1.02 (1.00  1.05) 1.02 (1.00 c  1.04) 1.02 (1.00 c  1.04)Preterm births b Corn 0.97 (0.94 – 0.99) 0.97 (0.94 – 1.00) 0.96 (0.92 – 1.01)Soybean 1.00 (0.98 – 1.01) 1.00 (0.98 – 1.02) 1.00 (0.97 – 1.03)Wheat 1.01 (0.96 – 1.06) 1.01 (0.95 – 1.07) 1.02 (0.93 – 1.11)Rice 1.21 (1.15 – 1.30) 1.21 (1.16 – 1.28) 1.15 (1.03 – 1.28)Cotton 1.06 (1.03 – 1.09) 1.06 (1.04 – 1.09) 1.05 (1.00 – 1.10)All Crops 1.01 (1.00 c  1.02) 1.00 (0.99  1.01) 1.00 (0.99  1.02) a Models adjusted for maternal race, maternal ethnicity, gender, mother's age group, maternal smoking, county median income, county population density,season of birth, and year. b Models adjusted for maternal race, maternal ethnicity, gender, mother's age group, maternal smoking, no PNC, prenatal WIC use, county median income, countypopulation density, season of birth, and year. c Indicates actual value is less than 1.00, but has been rounded for presentation purposes.  Table 3 Relative risks from adjusted regression models of low birth weight and preterm births with rice and cotton in Missouri, 2004 – 2006 that included both rice and cotton, usingthree modeling approaches. Relative risks represent risk associated with 10% increase in crop density. Outcome Exposure variable Poisson  ( Method  1)  GEE  ( Method  2)  Distance decay   ( Method  3) RR (95% CI) RR (95% CI) RR (95% CI) Low birth weight a Rice 1.10 (0.93 – 1.29) 1.10 (1.01 – 1.19) 1.09 (0.92 – 1.31)Cotton 1.13 (1.04 – 1.18) 1.11 (1.06 – 1.17) 1.11 (1.03 – 1.18)Preterm births b Rice 1.19 (1.11 – 1.27) 1.19 (1.11 – 1.26) 1.13 (1.01 – 1.26)Cotton 1.03 (1.00 – 1.06) 1.03 (1.00 – 1.06) 1.04 (0.99 – 1.09) a Models adjusted for maternal race, maternal ethnicity, gender, mother's age group, maternal smoking, county median income, county population density,season of birth, and year. b Models adjusted for maternal race, maternal ethnicity, gender, mother's age group, maternal smoking, no PNC, prenatal WIC use, county median income, countypopulation density, season of birth, and year. K.S. Almberg et al. / Environmental Research 134 (2014) 420 – 426  424
Search
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks