The median area sampled was 2.3 m2 and the inter-quartile range was 2.2 – 3.5 m2 . We obtained latitude and longitude coordinates at the front door of each home using a global positioning system device. We defined the season of dust sample collection as winter , spring , summer , or fall . We collected 16, 112, 91, 117, 69 and 29 samples each year from 2001 – 2006, respectively .Among the 50 pesticides measured in the dust samples,we selected eight insecticides for this analysis that were: used historically or currently for residential treatment of insects in California; frequently detected in carpet dust samples from the study populations ; and representative of different chemical classes of insecticides. Insecticides included in this analysis and their chemical classes are: chlordane and DDT , chlorpyrifos and diazinon , carbaryl and propoxur , and permethrin and cypermethrin . Dust samples were stored at −20 ºC prior to laboratory analysis in 2005–2008. Detailed laboratory methods have been published elsewhere.22 Briefly, carpet dust samples were sieved and approximately 0.5 g of dust extracted in a hexane:acetone solvent mixture, centrifuged, and concentrated. We quantified concentrations of insecticides using gas chromatography/mass spectrometry. In each batch,mobile grow rack quality control samples included duplicates spiked with 250 ng of each analyte and a solvent method blank. We spiked carbon-13 labeled surrogate recovery standards into all samples prior to extraction as a check on method performance. The limit of detection for the eight insecticides ranged from 1 – 5 ng/g of dust. Duplicate samples had average relative percent differences of 10 – 30%.
Mean recoveries of target analytes in spiked samples ranged from 85% – 118% and SRS recoveries averaged between 82 – 111% in quality control samples. SRS-corrected and uncorrected concentrations were highly correlated ; therefore, we report results for uncorrected concentrations.At the time of dust collection, we asked participants about treatments for pests in the home and garden during the previous 12 months. Specifically, we asked about insecticide treatments for ants and cockroaches; bees and wasps; fleas and ticks; flies and mosquitoes; carpenter ants and termites; any other indoor insects; and professional pesticide treatments inside and outside the home. Since several insecticides included in our analyses were also used in agriculture, we assessed the potential for pesticides to be transported into the home from the workplace by asking participants in the NCCLS whether anyone in the household worked during the previous 12 months in the following occupations: farm or ranch worker; gardener, landscaper, groundskeeper or nursery worker; agricultural packer; or pesticide applicator handling, formulating or mixing pesticides. We evaluated other potentially important factors including the year in which the residence was built and the age of the sampled carpet. We used the latitude and longitude coordinates of each home to assign the county and 2000 U.S. census tract. To account for potential regional differences in agricultural insecticide use, we grouped counties into three well-defined geographic regions of California including the urban San Francisco Bay Area, the agricultural Central Valley, and other . To account for potential neighborhood differences in the intensity of home insecticide use that might occur in more densely populated compared to less densely populated areas, we calculated the population density for each census tract by dividing the total population in each tract by the land area of the tract in square kilometers .
For participants that were not asked or did not know the year their residence was built , we used the median year of homes built in their census tract . Five of the insecticides had agricultural use during our study period , and we calculated the density of agricultural use near each residence using methods described elsewhere.Briefly, we used the California Department of Pesticide Use Reporting database from 2000 to 200624 to estimate use around the home in kilograms per square kilometer based on reported use within two distances for 365 days prior to collection of the dust sample, which coincided with the self-reported information on home and garden use.In our primary analyses, we included the 413 samples from NCCLS homes and the 21 samples collected during the first visit to APS homes. We summed the concentrations of all measured isomers of a compound for analyses. Individual insecticide concentrations in carpet dust samples were log-normally distributed. We evaluated bivariate relationships between concentrations of insecticides and potential determinants using non-parametric methods. We used the Kruskal-Wallis test to compare concentrations of insecticides in residences based on categorical variables and Spear man correlation coefficients for the continuous agricultural pesticide use variables. Potential determinants included self-reported home and garden use of insecticides by pest treated , possible occupational insecticide exposure , agricultural insecticide use near the home , year the residence was built , population density , age of sampled carpet , season sample collected and region . We used the non-parametric Mann-Kendall trend test25 to identify significant temporal trends of insecticide concentrations based on a single carpet dust sample from all 434 homes. We used Tobit regression to create separate models for each insecticide with the natural log-transformed concentration as the dependent variable and time in fractional years as the independent variable to estimate the average percentage change in concentrations of insecticides over time.
Tobit regression offers an unbiased approach for analyzing measurement data with observations below the LOD, where values below the LOD are treated as censored values that between zero and the LOD that fit the overall distribution of the measured data.We included covariates that were associated with insecticide concentrations in our bivariate analysis in the multi-variable model for each insecticide, and retained the covariates that remained significant in the final model. We estimated the average annual change in concentrations of each insecticide, adjusted for significant covariates, using Equation 1. For comparison, we also calculated the average percent annual change of insecticide concentrations in dust using linear regression models adjusted for the same covariates included in the Tobit models, but using the LOD/2 for concentrations below the LOD instead of a normally distributed value between zero and the LOD. We also ran the final models stratified by the two regions with the greatest number of samples to evaluate differences in temporal trends of insecticide concentrations between urban and agricultural areas. To evaluate temporal trends of insecticide concentrations in carpet dust within households, we used data from the 15 APS homes in Fresno where repeated dust samples were collected. We used mixed-effects models to estimate the change in insecticide concentrations over time while accounting for the correlation between measurements from the same household . We assumed mutually independent and normally distributed random intercepts for each residence. We imputed values below the LOD for APS homes with repeated measurements using a maximum likelihood procedure that assumed a lognormal distribution, which has been reported to yield unbiased estimates of regression parameters for variables that have at least 50% of measurements above the LOD.26 Individual insecticides were detected in 49% to 99% of carpet dust samples . The geometric mean concentrations were between 14 ng/g and 34 ng/g except for permethrin, which had a geometric mean concentration two orders of magnitude higher at 1,300 ng/g. Geometric mean concentrations of insecticides in carpet dust were significantly different by region for five of the eight insecticides . Concentrations of OC insecticides DDT and chlordane were highest in the more urban San Francisco Bay Area,mobile racking for growing whereas concentrations of OP insecticides chlorpyrifos and diazinon were highest in the agricultural Central Valley. Chlordane, DDT, chlorpyrifos and propoxur concentrations were higher in older homes than homes built more recently, while concentrations of cypermethrin were higher in newer homes. Permethrin concentrations were higher in samples collected from older carpets. Concentrations of carbaryl were higher in homes with self-reported use of flea and tick products. Concentrations of both permethrin and cypermethrin were higher in homes that had used products to control ants and roaches, fleas and ticks, and flies and mosquitoes. Concentrations of permethrin were also higher in homes that reported using products to control bees and wasps, while concentrations of cypermethrin were higher in homes that reported outdoor treatment by professional pest controllers.
Agricultural use accounted for a large proportion of chlorpyrifos but not propoxur sales during our study period .24 Concentrations of insecticides in carpet dust were not significantly higher in homes with pesticide-exposed workers and did not differ by the season in which the sample was collected . The Spearman rank correlation coefficients between carpet dust concentrations and agricultural use near the residence during the 12-months prior to sample collection were stronger for a 1,250 m radius buffer than for a 500 m radius, and the correlation was significant for three of the five study insecticides currently used in agriculture: chlorpyrifos , diazinon and carbaryl . After adjusting for significant covariates in Tobit regression models, we observed that concentrations of three of four insecticides no longer sold for indoor use decreased significantly across individual carpet dust samples from 434 homes during our study period of 2001–2006, while DDT concentrations did not change . Chlorpyrifos concentrations decreased an average of 31% per year : −38%, −22%, and diazinon 48% per year . These decreases were highly significant . In homes located in the Central Valley, the percentage changes for chlorpyrifos and diazinon, were not as large , albeit still highly significant . Chlordane levels decreased by 15% per year during the study period across our entire study area and this decrease was greater in the San Francisco Bay Area . Among insecticides sold for indoor use during our study period,only propoxur concentrations in carpet dust decreased significantly . Concentrations of carbaryl, cypermethrin and permethrin did not change significantly during our study period or differ by region. Propoxur concentrations in carpet dust decreased 34% per year and the estimated decrease was similar by region. For all insecticides evaluated, results from linear regression models were similar to those from Tobit models . Using the non-parametric Mann-Kendall trend test, we observed similar results with highly significant decreasing trends in carpet dust concentrations for chlorpyrifos, diazinon and propoxur, and also decreasing concentrations of chlordane during our study period 2001–2006 . Results were nearly identical when models were restricted to controls . Similar to results for a single dust sample from each home, we observed significant decreases in chlordane , diazinon and propoxur concentrations in repeated carpet dust samples collected from each of 15 residences in Fresno County from 2003 – 2005 . The calculated percentage decreases per year were very similar to the results based on models using dust concentrations from a single dust sample collected at each of the 434 study residences . The only differences in results produced by models using the repeated samples from 15 residences in Fresno County versus models using a single sample from all 434 residences were that concentrations of carbaryl decreased significantly in Fresno repeat homes instead of remaining constant, and that chlorpyrifos concentrations remained relatively constant in Fresno homes instead of decreasing. We observed significantly decreasing trends in carpet dust concentrations of the insecticides chlorpyrifos, diazinon, propoxur, and chlordane over the period 2001–2006 across 434 homes located in both urban and agricultural areas of California. The timing of this reduction coincided with the discontinuation of sales for residential use of chlorpyrifos in 2001 and diazinon in 2004. Restrictions on household use of propoxur were initiated by the USEPA in 1997, which could explain its similar annual rate of decline as for chlorpyrifos and diazinon .6 The persistence of chlordane and the time lapse between the chlordane ban and our study period likely explains the relatively lower annual rate of decline in concentration we observed. In contrast, we did not observe a decreasing trend for DDT, perhaps because it was banned from all use in the U.S. much earlier than chlordane, in 1972.1 However, we did record a relatively high detection frequency for DDT and a geometric mean concentration similar to chlordane . It is noteworthy that these two OC insecticides were still at detectable levels in carpet dust during the time period our samples were collected. The fact that these OC insecticides were detected in some homes built after these insecticides were removed from the market for indoor use suggests that an additional source of these compounds may be contaminated soil transported into the home.