No abstract

Results of publised pesticide mixure toxicity experiments conducted with aquatic organisms were compiled and evaluated to assess the accuracy of predictive mixture models. Three types of models were evaluated: Concentration addition (CA), independent action (IA), and simple interaction (SI). The CA model was the most often tested (207 experiments), followed by SI (59) and IA (37). The reviewed experiments are listed in the Supplemental material to provide a resource for future investigators. The predictive accuracy of each model was quantified for each experiment by the model deviation ratio (MDR), which was calculated by dividing the predicted toxicity by the observed toxicity. Eighty-eight percent of all experiments that evaluated the CA model had observed effective concentrations within a factor of 2 of predicted values (MDR values from 0.5-2.0). The median MDR was 1, about 5% of MDRs were less than 0.5, and about 5% were greater than 2, indicating unbiased estimates overall. The predictive accuracy of CA and IA models was influenced, however, by the different modes of action (MOA) of the pesticides. For experiments with pesticides with the same MOA, CA more accurately predicted effective concentrations for more experiments compared to IA, which tended to underpredict toxicity. The IA model was somewhat more accurate than the CA model for most mixtures with different MOAs, but in most cases there were relatively small differences between the models. Additionally, 80% of SI experiments had an MDR value below 2.0 despite a bias towards experiments that are likely to have an interaction. Thus, results indicate that the CA model may be used as a slightly conservative, but broadly applicable model with a relatively small likelihood of underestimating effects due to interactions.

No abstract

The first phase of intensive data collection for the National Water-Quality Assessment (NAWQA) was completed during 1993-1995 in 20 major hydrologic basins of the United States. Groundwater land-use studies, designed to sample recently recharged groundwater (generally within 10 years) beneath specific land-use and hydrogeologic settings, are a major component of the groundwater quality assessment for NAWQA. Pesticide results from the 41 landuse studies conducted during 1993-1995 indicate that pesticides were commonly detected in shallow groundwater, having been found at 54.4% of the 1034 sites sampled in agricultural and urban settings across the United States. Pesticide concentrations were generally low, with over 95% of the detections at concentrations less than 1 µg/L. Of the 46 pesticide compounds examined, 39 were detected. The compounds detected most frequently were atrazine (38.2%), deethylatrazine (34.2%), simazine (18.0%), metolachlor (14.6%), and prometon (13.9%). Statistically significant relations were observed between frequencies of detection and the use, mobility, and persistence of these compounds. Pesticides were commonly detected in both agricultural (56.4%; 813 sites) and urban (46.6%; 221 sites) settings. Frequent detections of pesticides in urban areas indicate that, as is the case with agricultural pesticide use in agricultural areas, urban and suburban pesticide use significantly contribute to pesticide occurrence in shallow groundwater. Although pesticides were detected in groundwater sampled in urban areas and all nine of the agricultural land-use categories examined, significant variations in occurrence were observed among these categories. Maximum contaminant levels (MCLs) established by the U.S. Environmental Protection Agency for drinking water were exceeded for only one pesticide (atrazine, 3 µg/L) at a single location. However, MCLs have been established for only 25 of the 46 pesticide compounds examined, do not cover pesticide degradates, and, at present, do not take into account additive or synergistic effects of combinations of pesticide compounds or potential effects on nearby aquatic ecosystems.

A recurring difficulty encountered in investigations of many metals and organic contaminants in ambient waters is that a substantial portion of water sample concentrations are below limits of detection established by analytical laboratories. Several methods were evaluated for estimating distributional parameters for such censored data sets using only uncensored observations. Their reliabilities were evaluated by a Monte Carlo experiment in which small samples were generated from a wide range of parent distributions and censored at varying levels. Eight methods were used to estimate the mean, standard deviation, median, and interquartile range. Criteria were developed, based on the distribution of uncensored observations, for determining the best performing parameter estimation method for any particular data set. The most robust method for minimizing error in censored-sample estimates of the four distributional parameters over all simulation conditions was the log-probability regression method. With this method, censored observations are assumed to follow the zero-to-censoring level portion of a lognormal distribution obtained by a least squares regression between logarithms of uncensored concentration observations and their z scores. When method performance was separately evaluated for each distributional parameter over all simulation conditions, the log-probability regression method still had the smallest errors for the mean and standard deviation, but the lognorm'al maximum likelihood method had the smallest errors for the median and interquartile range. When data sets were classified prior to parameter estimation into groups reflecting their probable parent distributions, the ranking of estimation methods was similar, but the accuracy of error estimates was markedly improved over those without classification.

Results of published pesticide mixture toxicity experiments conducted with aquatic organisms were compiled and evaluated to assess the accuracy of predictive mixture models. Three types of models were evaluated: Concentration addition (CA), independent action (IA), and simple interaction (SI). The CA model was the most often tested (207 experiments), followed by SI (59) and IA (37). The reviewed experiments are listed in the Supplemental material to provide a resource for future investigators. The predictive accuracy of each model was quantified for each experiment by the model deviation ratio (MDR), which was calculated by dividing the predicted toxicity by the observed toxicity. Eighty-eight percent of all experiments that evaluated the CA model had observed effective concentrations within a factor of 2 of predicted values (MDR values from 0.5-2.0). The median MDR was 1, about 5% of MDRs were less than 0.5, and about 5% were greater than 2, indicating unbiased estimates overall. The predictive accuracy of CA and IA models was influenced, however, by the different modes of action (MOA) of the pesticides. For experiments with pesticides with the same MOA, CA more accurately predicted effective concentrations for more experiments compared to IA, which tended to underpredict toxicity. The IA model was somewhat more accurate than the CA model for most mixtures with different MOAs, but in most cases there were relatively small differences between the models. Additionally, 80% of SI experiments had an MDR value below 2.0 despite a bias towards experiments that are likely to have an interaction. Thus, results indicate that the CA model may be used as a slightly conservative, but broadly applicable model with a relatively small likelihood of underestimating effects due to interactions.

During the 20 years from 1992 to 2011, pesticides were found at concentrations that exceeded aquatic-life benchmarks in many rivers and streams that drain agricultural, urban, and mixed-land use watersheds. Overall, the proportions of assessed streams with one or more pesticides that exceeded an aquatic-life benchmark were very similar between the two decades for agricultural (69% during 1992-2001 compared to 61% during 2002-2011) and mixed-land-use streams (45% compared to 46%). Urban streams, in contrast, increased from 53% during 1992-2011 to 90% during 2002-2011, largely because of fipronil and dichlorvos. The potential for adverse effects on aquatic life is likely greater than these results indicate because potentially important pesticide compounds were not included in the assessment. Human-health benchmarks were much less frequently exceeded, and during 2002-2011, only one agricultural stream and no urban or mixed-land-use streams exceeded human-health benchmarks for any of the measured pesticides. Widespread trends in pesticide concentrations, some downward and some upward, occurred in response to shifts in use patterns primarily driven by regulatory changes and introductions of new pesticides.

No abstract