Critical and systematic evaluation of data for estimating human exposures to 2,4-dichlorophenoxyacetic acid (2,4-D) – quality and generalizability

ABSTRACT

The herbicide 2,4-dichlorophenoxyacetic acid (2,4-D) has been commercially available since the 1940’s. Despite decades of data on 2,4-D in food, air, soil, and water, as well as in humans, the quality the quality of these data has not been comprehensively evaluated. Using selected elements of the Biomonitoring, Environmental Epidemiology, and Short-lived Chemicals (BEES-C) instrument (temporal variability, avoidance of sample contamination, analyte stability, and urinary methods of matrix adjustment), the quality of 156 publications of environmental- and biomonitoring-based 2,4-D data was examined. Few publications documented steps were taken to avoid sample contamination. Similarly, most studies did not demonstrate the stability of the analyte from sample collection to analysis. Less than half of the biomonitoring publications reported both creatinine-adjusted and unadjusted urine concentrations. The scope and detail of data needed to assess temporal variability and sources of 2,4-D varied widely across the reviewed studies. Exposures to short-lived chemicals such as 2,4-D are impacted by numerous and changing external factors including application practices and formulations. At a minimum, greater transparency in reporting of quality control measures is needed. Perhaps the greatest challenge for the exposure community is the ability to reach consensus on how to address problems specific to short-lived chemical exposures in observational epidemiology investigations. More extensive conversations are needed to advance our understanding of human exposures and enable interpretation of these data to catch up to analytical capabilities. The problems defined in this review remain exquisitely difficult to address for chemicals like 2,4-D, with short and variable environmental and physiological half-lives and with exposures impacted by numerous and changing external factors.

Introduction

The herbicide 2,4-dichlorophenoxyacetic acid (2,4-D) is widely registered globally for use on crops, lawns, rights-of-way, and for aquatic and forestry applications (USEPA 2005). This herbicide has been commercially available since the 1940’s. A variety of formulations are available and may contain different forms of 2,4-D, including acid, amine, and ester forms (Peterson et al. 2016). Human exposures may occur through consumption of foods with 2,4-D residues, or from exposure to (1) drinking water and recreational water use, (2) residential and/or occupational uses, and (3) post-application exposures associated with use on sites such golf courses, parks and lawns (USEPA 2005).

Table 1. Elements of the Biomonitoring, Environmental Epidemiology, and Short-Lived Chemicals (BEES-C) Instrument used in this assessment (from, 2015; LaKind et al. 2014, 2015)

Study assessment components	Tier 1	Tier 2	Tier 3
Biomarker Selection and Measurement
Biomarker stability	Samples with a known history and documented stability data or those using real-time measurements.	Samples have known losses during storage but the difference between low and high exposures can be qualitatively assessed.	Samples with either unknown storage history and/or no stability data for target analytes.
Sample contamination	Samples are contamination-free from time of collection to time of measurement (e.g. by use of certified analyte-free collection supplies and reference materials, and appropriate use of blanks both in the field and lab). Research includes documentation of the steps taken to provide the necessary assurance that the study data are reliable.	Incomplete documentation of QC steps taken to avoid contamination.	There are known contamination issues and no documentation that the issues were addressed.
Matrix adjustment	Study provides results either in the main publication or as a supplement for creatinine-adjusted or SG-adjusted and non-adjusted urine concentrations and reasons are given for adjustment approach.	Study only provides results using one method (matrix-adjusted or not).	No established method for adjustment (e.g. adjustment for hair).

Table 2. Summary of publications relevant for human exposure assessment by matrix

Matrix	Reviewed publications*	Publications assessed with BEES-C*
Food	16	7
Soil/Dust	24	16
Air	40	32
Water	78	44
Urine	77	52
Blood/Semen	6	5
Total	241	156

*Some publications evaluated more than one matrix and are included more than once

Understanding the extent, duration, variability, and timing of human exposures is essential for adequately characterizing and classifying exposure in epidemiological research and human health-risk assessment. At the same time, assessing exposures to 2,4-D for specific and general populations is difficult due to a combination of factors. First, as noted above, various forms and formulations 2,4-D are available for use with different physicochemical properties affecting fate and transport, although different forms break down to the acid form (Walters 1999). Second, exposures associated with applications of 2,4-D are impacted by meteorological conditions, type of application equipment used, extent of application, personal protective equipment, and other factors (Peterson et al. 2016). Third, 2,4-D has a short environmental and physiological half-life, raising complexities related to capturing temporal variability in exposures and determining both peak and long-term exposure levels (USEPA 2005).

The research on 2,4-D concentrations in various environmental media and human matrices using biomonitoring approaches is extensive. In addition, there are multiple epidemiology studies examining associations between 2,4-D exposures and human health effects, including those relying on self-reported use and measures of 2,4-D in environmental media and urinary biomarker data (Burns and Swaen 2012). With the current shift away from lab animal studies, there will likely be an increased emphasis on the use of human data as the basis for public health decision-making. An assessment of the quality of data for use in human exposure assessments is an integral part of the decision-making process (Barroso et al. 2016). Surprisingly, to date, no review has critically examined the quality of the published exposure data to determine whether the 2,4-D data are robust, reliable, and/or generalizable.

The aim of this review was to first examine the quality of the published, peer-reviewed 2,4-D monitoring data from studies of food, air, soil, water, and human matrices by focusing on four quality elements of the Biomonitoring, Environmental Epidemiology, and Short-lived Chemicals (BEES-C) instrument: documentation of avoidance of sample contamination, documentation of analyte stability from time of sample collection to sample analysis, and matrix adjustment methods (specific to interpretation of urinary measurements). Reports were also examined for information on number of samples collected and time between 2,4-D application and sample collection with an eye toward understanding whether the data offered information on peak or long-term exposures as well as the likelihood for exposure misclassification.

Our objective in this review was to examine the strengths and limitations of the international 2,4-D exposure database to inform future research. Our aim was to answer the following questions: (1) Are the published data of sufficient quality for use in assessing general population exposures and/or exposures to specific communities? (2) Are the published data generalizable to broader populations (i.e. can the data be used to generate information on typical peak and/or background exposures in populations other than those specific to an individual study)? (3) What additional information can be gleaned from large-scale population- and media-based studies available from governmental reports (e.g. NHANES, US Department of Agriculture Annual Pesticide Data Program [PDP])? To be clear, this review process results in an overall assessment of various aspects of individual study quality. It is the responsibility of the end user to decide if the data are fit-for-purpose.

Methods

Identification and selection of studies

Recommendations for systematic reviews (Moher and Tricco 2008; Shea et al. 2007; Sutton et al. 1998) and elements of the “assessment tool of multiple systematic reviews” (AMSTAR) guidelines that pertain to exposure data were followed (Shea et al. 2007). The search and selection of relevant studies were conducted independently by two study authors (C.J.B and J.S.L.). Electronic data sources including PubMed and Web of Science™ Core Collection were used to conduct the initial literature search. Using keywords “2,4-D,” ‘2,4-dichlorophenoxyacetic acid,” “air,” “blood,” “dust,” “food,” “soil,” “urine,” and “water” as well as various combinations of these keywords, we selected articles that included environmental- and/or biomonitoring-based measurements of 2,4-D. References of retrieved publications were reviewed to identify publications not captured by the electronic search.

The inclusion criteria for this review were as follows:

1. study includes measurements of 2,4-D in environmental media and/or humans (as opposed to obtaining exposure information through other approaches such as questionnaires or job title),

2. investigation was published in the peer-reviewed published literature,

3. publication appeared in English prior to 17 May 2016 (end of literature search),

4. study measures environmental levels of 2,4-D (as opposed to spiked samples, etc.), and

5. study provides data that can be linked to human exposure.

The following categories of studies were excluded:

1. mineralization, sorption, and recovery studies as well as batch and column studies,

2. methodological studies using pooled urine samples,

3. groundwater studies with no evidence of potential human exposure to the groundwater,

4. toxicology studies, including PBPK, lab animal studies, in vitro studies,

5. studies on exposures to Agent Orange/dioxins,

6. studies of poisonings via self-administration,

7. reviews,

8. regeneration studies and other effects on plants, including plant toxicity,

9. environmental fate studies that rely upon experimental application, not real world uses (including bioaccumulation),

10. studies on decontamination/remediation, and

11. human health studies with no exposure data.

The final body of literature was examined independently by 4 co-authors (C.J.B., J.S.L., J.S.N., A.J.B.). Data from each study were tabulated, and resulting summary tables were again cross-checked (by C.J.B., J.S.L., J.S.N., A.J.B.).

Assessment of individual studies

Principles for characterizing the quality of human studies that rely on biomarkers of chemicals with short physiologic half-lives are described in an evaluative instrument referred to as the BEES-C (LaKind et al. 2015, 2014). While originally developed for biomonitoring data, the evaluative elements on temporality, selection of analyte, sampling and analysis, and study design are also applicable to evaluating the quality of environmental data. These elements capture core components of the overall quality of measurement data.

To evaluate the scientific quality of each study and strengths and limitations for formulating conclusions (steps 7 and 8 of the AMSTAR guidelines), two co-authors (C.J.B. and J.S.L.) independently reviewed each retrieved study using three exposure-related elements of the BEES-C instrument. The BEES-C instrument uses a three-tier system with Tier 1 indicating the highest quality (LaKind et al. 2014). While the instrument includes numerous aspects of study quality, the focus was on elements that are relevant to practices for conducting and reporting exposure: sample contamination, analyte stability, and matrix adjustment (Table 1). Temporal variability was explored but was not included in the tiered assessment as many of the investigations were not epidemiology studies and not applicable to the BEES-C tiering approach. The four elements are described here as follows.

Sample contamination

A high-quality study (Tier 1) includes documentation of the procedures used to avoid contamination by the analyte of interest and validates that samples are contamination-free from the time of sample collection to the time of analysis. Approaches include the appropriate use of blanks both in the field and lab. A Tier 2 study has incomplete documentation of the steps taken to avoid contamination, while a Tier 3 study includes no documentation.

Analyte stability

A study is assigned a Tier 1 ranking if it includes details on the history of the sample and has documented analyte stability. If losses of the analyte over time are known, a Tier 2 study has accounted for these losses in the exposure estimates. Finally, a Tier 3 study provides no information on this element.

Matrix adjustment

Measurements of analytes in urine are commonly measured and reported in units of concentration (mass per volume urine). However, urinary output is highly variable and affected by factors including diet, exercise, hydration, and age. A common approach to correcting for variable urine dilution is by adjusting with urinary creatinine concentration or urine specific gravity. However, urinary creatinine levels are also variable and may be affected by age, gender, race/ethnicity, body mass index, fat-free mass, time of day of collection of urine samples, exercise, diet, and health (LaKind and Naiman 2015). Currently, there is no consensus on the best method(s) for “correcting” urinary biomarkers measurements for variable urine dilution (LaKind et al. 2014). Minimally, both the volume-based and corrected concentrations need to be included to enable appropriate comparison across studies (LaKind et al. 2014). A Tier 1 study includes both adjusted and unadjusted urinary concentration data.

Temporal variability

Temporal variability was not formally assessed with a tiering approach. However, study-specific available information on time between 2,4-D application and sample collection, as well as number of samples collected (one-time sampling, multiple samples, and/or longitudinal samples) was tabulated to assess whether the data capture information on short-term or long-term variability/changes in exposure. As noted by Meeker et al. (2013): “Characterizing temporal variability in exposure metrics, especially for biomarkers of nonpersistent compounds… is a critical step in designing and interpreting an epidemiology study related to the potential for exposure measurement error.” If not properly accounted for, temporal variability will impact the ability to interpret study results because of a lack of information on whether peak, chronic, or background concentrations are captured.

Results

Literature search and review

The total number of citations identified from the literature search using PubMed and Web of Science™ was 2628. After removal of publications based on the above-noted exclusion criteria, the total number of publications evaluated was 241. After further review, many of the publications were found not to contain have relevant data (see exclusion criteria in the Methods section), leaving 156 for quality assessment (Table 2). The following sections describe the results of the quality review. Full details of the quality assessment of the literature using the BEES-C elements described above are provided in Tables S1-S6.

Exposure data quality assessment

2,4-D and food

Sixteen publications were identified in the peer-reviewed literature that included measurements of 2,4-D in food items. However, nine of these publications did not contain data relevant for human exposure assessment, either because they were primarily methodological papers that used a limited number of samples with no information on source or field handling of food items (Anastassiades and Schwack 1998; Lancas, Rissato, and Galhiane 1999; Lancas, Rissato, and Mozeto 1996; Vdovenko et al. 2013), because the items measured were in a form not relevant to direct consumption (e.g. wheat straw) (Cessna and Hunter 1993; Gowen, Wiersma, and Tai 1976; Lancas, Rissato, and Galhiane 1999; Lancas, Rissato, and Mozeto 1996), because no primary concentration data were provided (Gartrell et al. 1985; Shin et al. 2011) or because 2,4-D results were not included (Whitmore et al. 1994).

Seven studies (Table S1) contained sufficient information to be evaluated for exposure data quality (Frank, Carpentier, and Mackenzie 1987; Melnyk, Berry, and Sheldon 1997; Morgan et al. 2008; Morgan, Wilson, and Chuang 2014; Wilson, Chuang, and Lyu 2001; Wilson et al. 2003, 2010). One study evaluated residue on game fish following treatment for Eurasian water milfoil in two Canada lakes (Frank, Carpentier, and Mackenzie 1987). The other studies either collected duplicate diet samples of foods and beverages in household and daycare settings (Morgan et al. 2008; Morgan, Wilson, and Chuang 2014; Wilson, Chuang, and Lyu 2001; Wilson et al. 2003, 2010) or determined food in households from pesticide applicators (Melnyk et el. 1997).

The 5 studies published after 2000 included both field and lab blanks and were therefore considered of highest quality in terms of addressing avoidance of sample contamination (Table S1). Melnyk, Berry, and Sheldon (1997) documented the use of lab blanks (Tier 2). For assessing analyte stability from time of sampling through time of analysis, none of the studies demonstrated that storage time and conditions did not adversely impact analyte stability.

In addition to the quality elements for internal validity described above, strengths and limitations of the data were also considered for use for more generalized exposure assessments. A strength of the 2,4-D/food studies is that they provide measurements of foods and beverages consumed, as opposed to crops or grains that have not yet been processed or are parts of foods not typically consumed. However, several issues are noted: (1) because the foods were combined into an estimate of 2,4-D in solids and liquids, data on individual foods cannot be extracted and used for people whose diets differ from these cohorts; (2) there is limited information on sources of food and their proximity to use of 2,4-D and so changes in diets throughout a given season or year may place an individual in a different 2,4-D exposure category; (3) studies did not clarify whether these data represent cohort exposures at other times of the year and there is no information on whether the food collected is representative of longer-term food and drink consumption habits; and (4) the data are only from Canada and two states in the USA and the cohorts are relatively small.

Minimal information was provided on timing of the most recent use of 2,4-D related to food and drink-related exposures. The source of 2,4-D was clearly apparent in the study of fish (Frank, Carpentier, and Mackenzie 1987) since the lakes were treated with 2,4-D and repeated samples were collected post treatment. Only Melnyk, Berry, and Sheldon (1997) attempted to determine the source of the residue on the household food by evaluating the food from a duplicate meal collected in the field during a 2,4-D application. However, this was limited to a single sample. Only one home owner out of the 135 sampled by Morgan et al. (2008) used a 2,4-D product within 3 days of sample collection. Thus, one cannot determine whether these foods and drinks represent peak or background exposure. These combined issues limit the generalizability of the available food data.

2,4-D and soil/dust

Twenty-four studies were identified that measured 2,4-D in soil/dust. Methodological studies with the objective of evaluating an analytical or data collection approach were excluded (Colt 1998; Colt et al. 2008; De Amarante et al. 2003; Roberts and Camann 1989). In addition, omitted from further consideration were studies designed to determine the dissipation of 2,4-D following applications and related evaluations of the microbial characteristics of soil (Grover et al. 1985; Taskar et al. 1982; Voos and Groffman 1997; Wilson, Geronimo, and Armbruster 1997).

The remaining 16 publications included sufficient exposure-related information for quality assessment (Table S2). The studies were all conducted in the USA or Canada. Two studies collected soil from agricultural settings or after an application beneath power lines (Gowen, Wiersma, and Tai 1976; Meru et al. 1990) and one was an occupational setting (Frank, Campbell, and Sirons 1985). The other investigations examined indoor soil/dust (i.e. dust on bare surfaces and carpet dust). Several investigations included measurements of soil outside the home and/or daycare settings (Morgan et al. 2008; Morgan, Wilson, and Chuang 2014; Wilson, Chuang, and Lyu 2001; Wilson et al. 2010). Regarding documentation of avoidance of sample contamination, four studies included both field and lab blanks and are considered Tier 1 (Morgan et al. 2008; Morgan, Wilson, and Chuang 2014; Wilson, Chuang, and Lyu 2001; Wilson et al. 2010). Three studies documented the use of either field blanks or lab blanks and are categorized as Tier 2 (Curwin et al. 2005b; Deziel et al. 2015; Metayer et al. 2013).

Regarding assessment of analyte stability from time of sample collection through time of analysis, only two of the studies provided information on sample holding time or recovery of 2,4-D in field samples (Tier 2) (Curwin et al. 2005b; Nishioka et al. 1999); the remaining studies are considered Tier 3.

A strength of the studies based in homes and day care centers is that they report exposure data on residues specific to the living environment of adults and children (Curwin et al. 2005b; Wilson et al. 2010). The generalizability to other geographic regions is limited since variables related to deposition include the structure of the building, cleaning frequency, climate, and number of persons living within the household and might differ from region to region and country to country. The method of collecting samples from household vacuum bags is another limitation. While the approach is simple and noninvasive, the investigator cannot determine the timeframe of soil/dust deposition (i.e. the vacuum bag may have been in use for days, weeks, or longer). Notably, when investigators required that subjects owned their rugs for 5 years or more and used their vacuum within the past year, the resulting sample size was reduced by half (Colt et al. 2004). Another limitation of vacuum bags is that the vacuum can be used in another place, such as in a vehicle. Deziel et al. (2015) concluded “While measurements of carpet dust provide information on active ingredients, only interview data can provide information on household behaviors such as rooms occupied by children, timing of pesticide use, and other covariates important to evaluating pesticide exposure.”

An additional issue is raised by the studies by Gowen, Wiersma, and Tai (1976), Frank et al. (1985), Meru et al. (1990) and Whitmore et al. (1994); these may have provided relevant exposure data at the time they were conducted, but various aspects of 2,4-D application practices have changed since the 1970’s–1980’s and before using these data to estimate current exposures, applicability of past practices to current ones must be considered (Peterson et al. 2016).

2,4-D and air

Forty publications were identified that included 2,4-D concentration data in air (indoor and outdoor ambient air and samples from personal monitors) from a variety of settings (e.g. urban and rural/farm areas, homes, occupational spraying activities). The studies were conducted in the USA, Canada, Europe, Malaysia and South Africa, and were published over a time span from the 1960’s to the 2000’s.

Eight of these publications did not contain data relevant for human exposure assessment and are not considered further. Reasons for exclusion include: no data or insufficient data (e.g. only minimum concentrations, units in mass per sampling device) were provided (Cessna et al. 2000; Grover et al. 1985; Messing et al. 2014; Raeppel et al. 2015; Trevisan et al. 1993), 2,4-D was measured in precipitation and data on meteorological conditions or rainfall totals were not included (Asman et al. 2005), or measurements focused on 2,4-D butyl ester which was replaced by other 2,4-D formulations (Peterson et al. 2016) in the 1980’s (Farwell et al. 1976; Hee, Sutherland, and Vetter 1975).

Thirty-two studies (Table S3) contained sufficient information to proceed with a quality assessment. These studies included a wide range of study locations, settings and approaches. In terms of documentation of procedures used to prevent and/or quantify the extent of sample contamination, 10 studies included both field and lab blanks and are considered Tier 1 (Aulagnier et al. 2008; Hines et al. 2001; Messing et al. 2011, 2013; Morgan et al. 2008; Morgan, Wilson, and Chuang 2014; Thomas et al. 2010; Waite et al. 2002a; Wilson, Chuang, and Lyu 2001; Wilson et al. 2010). Five studies included only field blanks (Curwin et al. 2005b; Whitmore et al. 1994; Yao et al. 2006) or lab blanks (Baraud et al. 2003; Hawthorne et al. 1996) and ranked as Tier 2. The remainder did not document avoidance of sample contamination and are thus Tier 3 for this quality element.

For assessing analyte stability, four studies (Curwin et al. 2005b; Messing et al. 2013; Nishioka et al. 2001; Sandmann, Debeer, and Vandyk 1991) provided some information on sample storage time (Tier 2, Table S3) but only two (Abbott et al. 1987; Lavy et al. 1982) specifically assessed whether any “2,4-D was broken down prior to analysis” (Lavy et al. 1982). These two are considered to be Tier 1 studies for this quality element. Most of the air samples in the study by Lavy et al. (1982) had concentrations of 2,4-D below the limit of detection (LOD), thus information on analyte stability could not be extracted. Abbott et al. (1987) reported that “Results of field and laboratory recovery tests showed that all 2,4-D applied to the sampling media was recovered in the analytical process.” The remaining studies were considered Tier 3 for analyte stability.

Several of the studies collected data on the relationship between time of sampling and pesticide application (Baharuddin et al. 2011; Curwin et al. 2005b, 1986a; Grover et al. 1976; Harris, Solomon, and Stephenson 1992; Hawthorne et al. 1996; Hines et al. 2001; Kolmodin-Hedman and Erne 1980; Kolmodin-Hedman, Hoglund, and Akerblom 1983; Lavy et al. 1982; Libich et al. 1984; Thomas et al. 2010). These studies may likely provide the most useful data for assessing short-term exposure potential associated with either occupational or non-occupational use of 2,4-D. These reports may also be informative regarding whether exposure estimates represent peak or background exposure. In contrast, the timing and/or sources of exposure were not specified for the other investigations.

The generalizability of these data for other populations in other geographic regions and/or other conditions is unclear. Ambient air concentrations are impacted by agricultural practices and meteorological conditions and none of the reviewed publications provided information that would allow for translation of study results to other settings. Even for indoor air sampling (Curwin et al. 2005b; Morgan et al. 2008; Morgan, Wilson, and Chuang 2014; Whitmore et al. 1994; Wilson, Chuang, and Lyu 2001), monitoring results may be heavily dependent on the type of formulation, extent of application, and time(s) since last application as well as structural factors such as room size and ventilation. Further, longitudinal sampling would be needed to understand whether peak exposures were captured as well as the rate of decrease in air concentrations over time.

2,4-D and water

A total of 78 publications were identified that included 2,4-D water concentration data from a variety of settings such as wetlands, streams, lakes, and surface water. The studies were conducted in the USA, Canada, Mexico, Chili, Korea, Malaysia, Australia, and Europe, and were published over a time span from the 1960’s to the 2000’s.

Thirty-four of these publications did not contain data relevant for human exposure assessment and are not considered further. The excluded studies were experimental, analytical, dissipation, or rainfall studies (Amani et al. 2011; Anirudhan and Alexander 2014; Degenhardt et al. 2011; Farenhorst, Andronak, and McQueen 2015; Hill et al. 2002, 2003; Matthiessen et al. 1992; Meru et al. 1990; Nam et al. 2014; Omidi et al. 2014; Osborne et al. 2015; Qin, Wei, and Li 2002; Quaghebeur et al. 2004; Ryals, Genter, and Leidy 1998; Sandmann, Debeer, and Vandyk 1991; Siemering, Hayworth, and Greenfield 2008; Trevisan et al. 1993; Waite et al. 1995, 2002b; Watson 1977; Yamini and Saleh 2013), provided insufficient data (data only in graphical format) (Botta et al. 2012; Cerejeira et al. 2003; Davis and Lydy 2002; Gilliom 2007; Mitchell, Brodie, and White 2005; Muir and Grift 1987; Nagy and Painter 1985; Schultz and Harman 1974; Schultz and Whitney 1974; Spalding, Junk, and Richard 1980; Whitmore et al. 1994), were a review (Hallberg 1989) or monitored a unique site not applicable to general population exposures (Buczynska and Szadkowska-Stanczyk 2005).

Forty-four studies (Table S4) contained sufficient information to be evaluated for exposure data quality. These studies included a wide range of study locations, settings and approaches. In terms of steps taken to avoid and/or quantify the extent of sample contamination, 18 studies are considered Tier 2 as the investigators documented use of either field blanks (Bartlett et al. 2016; Loos et al. 2009, 2010a; Loos, Locoro, and Contini 2010b; Phillips and Bode 2004) or lab blanks (Birch et al. 2015; Buser and Muller 1998; DeLorenzo et al. 2012, 2007; Donald et al. 2001; Frank et al. 1990; Frank, Logan, and Clegg 1991; King and Balogh 2010; McManus et al. 2014; Park et al. 2011; Rawn et al. 1999b; Woudneh et al. 2006, 2007). Nine studies included both and are considered as Tier 1 (Cohen et al. 1990; Ensminger et al. 2013; Felix-Canedo, Duran-Alvarez, and Jimenez-Cisneros 2013; Glozier et al. 2012; Messing et al. 2011; Palma et al. 2004; Struger and Fletcher 2007; Waite et al. 2002a; Wijnja, Doherty, and Safie 2014). All the other studies are Tier 3.

Eight of the studies provided some information on sample storage time (Table S4) (Cohen et al. 1990; DeLorenzo et al. 2012; Herrero-Hernandez et al. 2013; King and Balogh 2010; Loos et al. 2009, 2010a; Loos, Locoro, and Contini 2010b; Wiergowski et al. 2000) but did not demonstrate that 2,4-D was stable over time from collection through analysis (Tier 2). The remainder of studies were Tier 3 for demonstration of analyte stability.

In terms of capturing data on temporal changes in 2,4-D concentrations in water, many of the water studies collected multiple samples, often weekly or monthly, for example, to track decreases in 2,4-D concentration following an application (Wojtalik, Hall, and Hill 1971), to understand differences in land use and agricultural practices (Cessna and Elliott 2004; Donald et al. 2001) or to follow a presumed high exposure “spill” (Frank et al. 1990). Other studies sought to capture the impact of seasonal effects of surface water stratification on 2,4-D water concentrations. The findings of investigations were mixed. Woudneh et al. (2006) observed the highest levels of 2,4-D in surface water during spring sampling (as compared to fall), which was consistent with time of herbicide application. An evaluation of both ground water and surface water in Bulgaria concluded that concentrations were associated with intensity of herbicide use and properties of the pesticide (Balinova and Mondesky 1999). Similarly, Hall et al. (1993) noted that 2,4-D in urban creeks mainly had detectable levels of 2,4-D shortly after herbicide applications. In contrast, Waite et al. (1992) found no marked correlation with 2,4-D in surface water and pesticide use in a small agricultural watershed, and similarly found no association between groundwater 2,4-D levels with season or rainfall events. Donald et al. (2001) noted no marked differences in 2,4-D concentrations in wetlands near fields with and without 2,4-D application; in some cases, 2,4-D had not been applied to these lands for many years.

Buser and Muller (1998) detected appreciable variations in 2,4-D levels in Swiss surface water samples collected about a month apart but not timed according to pesticide application. Similarly, seasonal variations in 2,4-D concentrations were observed in some but not all waters collected from urban, agricultural, and mixed sites by Woudneh et al. (2007). However, Glozier et al. (2012) did not report seasonal differences in 2,4-D levels in urban and agriculturally-impacted rivers and streams in a national survey in Canada.

Precipitation might also impact 2,4-D levels in water, either through dilution or run off. In studies of urban surface waters with no agricultural inputs, concentrations of 2,4-D were significantly higher in surface waters during rainstorms (Ensminger et al. 2013) and after precipitation events (Glozier et al. 2012). In agricultural areas, Donald et al. (1999) found that wetlands had higher levels of pesticides when these chemicals were applied prior to significant rainfall events. Palma et al. (2004) and Phillips and Bode (2004) also observed higher 2,4-D levels in surface waters after rain events.

Difficulties in capturing temporal changes in 2,4-D were highlighted by Rawn et al. (1999a) who noted “that detailed temporal patterns of herbicides and nutrients could only be established if high sampling frequency was maintained during such critical periods such as application periods and during major runoff events. An unexpected finding was that herbicide concentrations were not related to runoff losses, but instead they did correspond to increased levels in precipitation and air measured within the watershed.” These investigators recognized the complicated determinants of movement of herbicides through a watershed, which are influenced by factors such as growing season, soil type, terrain, and meteorological conditions.

Overall, data on 2,4-D levels in water included several types of water systems from various countries and climates, ranging from snowmelt in Canada to rice fields in Malaysia. Most investigations collected multiple samples in a watershed. The most relevant for human exposure were those studies of potable water sources. Higher quality investigations collected frequent samples to determine the range of concentrations during a growing season (Glozier et al. 2012; Palma et al. 2004; Woudneh et al. 2007). However, the factors described above impact the generalizability of 2,4-D water data.

2,4-D and urine

A total of 77 publications were identified that included data on 2,4-D levels in urine from adults and/or children. The studies were conducted in the USA, Canada, Europe, Thailand, Caribbean, Nicaragua, and Costa Rica, and six studies with unspecified locations. The study publication years ranged from 1980 to 2015.

Twenty-five of these publications were not considered further. Many contained data described in other publications reviewed here or were analytical methods papers (Adgate et al. 2001; Aprea, Sciarra, and Bozzi 1997; Arbuckle et al. 2002, 2006; Arcury et al. 2010; Baker et al. 2000; Barr et al. 1999; Cessna et al. 2000; Coble et al. 2005, 2011; Davis et al. 2013; De Felip et al. 1988; Figgs et al. 2000; Grover et al. 1986b; Harris et al. 2001, 2002; Hill et al. 1995b; Holler et al. 1989; Knopp, Schmidt, and Niessner 1993; Morgan and Jones 2013; Morgan et al. 2006; Murphy, Kutz, and Strassman 1983; Scher et al. 2007; Semchuk et al. 2003; Shafik, Sullivan, and Enos 1971; Taskar et al. 1982).

Fifty-two studies (Table S5) included relevant information on urinary 2,4-D measurements in urine and were evaluated for exposure data quality. Three of these studies included documentation of steps taken to avoid and/or quantify the extent of sample contamination in both the field and lab (Tier 1) (Bradman et al. 2015; Morgan 2015; Morgan et al. 2008). Tier 2 studies included either field blanks (Thomas et al. 2010) or lab blanks (or blanks of an unspecified nature) (Arbuckle et al. 2005; Arcury et al. 2007, 2009; Curwin et al. 2005a; Hill et al. 1989; Hines et al. 2003; Kutz et al. 1992 ; Nigg and Stamper 1983 ; Panuwet et al. 2008; Raymer et al. 2014 ; Shealy et al. 1996 ; Wilson et al. 2010). The remaining studies provided no documentation and are considered Tier 3 studies.

Three studies included information on amount of time between sample collection and analysis (Bhatti et al. 2010; Knopp 1994; Lavy et al. 1982). Lavy et al. (1982) specifically demonstrated an effort to assess whether 2,4-D was stable in the samples until analysis.

Because of the uncertainties with interpreting urinary biomonitoring data introduced by urinary dilution, LaKind et al. (2014) recommended that studies include results for volume-based measures and those adjusted for urinary dilution. Thus, for urinary studies, the additional BEES-C element of matrix adjustment was considered. Reporting both creatinine-adjusted and unadjusted urine concentrations has not been widely adopted in the publications reviewed here, with only 13 of the publications including both (Tier 1) (Alexander et al. 2007; Arcury et al. 2007; Bradman et al. 2015; Curwin et al. 2005a; Hill et al. 1995a, 1989; Jurewicz et al. 2012; Knopp and Glass 1991; Morgan 2015; Morgan et al. 2008; Panuwet et al. 2009, 2008; Rodriguez et al. 2012).

A key strength of many of these studies is the collection and reporting of information on the relationship between time of urine sampling and time of pesticide application. In addition, several investigators collected urine samples prior to and after 2,4-D application (Alexander et al. 2007; Arbuckle and Ritter 2005; Hines et al. 2003; Kolmodin-Hedman and Erne 1980; Lavy et al. 1982; Thomas et al. 2010; Zhang et al. 2011). This information might be used to distinguish peak exposures from “background” exposures.

A limitation of many of these studies is that only one spot sample per participant was collected; because of the short physiological half-life of 2,4-D (Sauerhoff et al. 1977), this one measure does not provide information on temporal variability. Other studies collected 12- or 24-hour samples or conducted repeated sampling. These longitudinal studies offer the best opportunities to draw robust conclusions regarding exposures in the study population and to apply their results to other populations.

2,4-D and other biomonitoring matrices

Five studies evaluated 2,4-D in blood (Knopp 1994; Knopp, Schmidt, and Niessner 1993; Kolmodin-Hedman and Erne 1980; Semchuk et al. 2004, 2003), but one (Knopp, Schmidt, and Niessner 1993) was excluded from further consideration as it was a case study of an unusually highly exposed individual (Table S6). Of the remaining studies, Semchuk et al. (2003, 2004) reported the use of different types of lab blanks (Tier 2); none of the studies documented the use of field blanks. Only Knopp (1994) provided information on sample storage time (Tier 2); none of the studies included assessments of analyte stability from time of sample collection through sample analysis.

One study on 2,4-D in semen was identified (Arbuckle et al. 1999) (Table S6). Lab blanks were included in the protocol, but use of field blanks was not documented (Tier 2). This is interesting as the likelihood of contamination from handling might be difficult to avoid. No data were provided on analyte stability in semen (Tier 3).

Discussion

In this review, the quality of more than 150 published, peer-reviewed studies that reported concentration data on 2,4-D in various media and matrices was examined using elements of the BEES-C instrument. To our knowledge, this is the first comprehensive quality assessment of 2,4-D exposure data. The focus was on the avoidance of sample contamination and the potential for loss of analyte during collection, shipping and storage of the samples. For urinary measures of 2,4-D, it was noted if results were reported both on a volume and creatinine-adjusted basis. Finally, the extent to which studies provided information from multiples measures and time between 2,4-D application and sample collection was examined. These data may determine the degree to which the investigations captured variability in exposures and whether the data represent peak or background exposures.

In the following sections, three questions are discussed: (1) Are the published data of sufficient quality for use in assessing exposures? (2) Are the published data generalizable to broader populations? And (3) What additional information can be gleaned from large-scale population- and media-based studies available from governmental reports?

Are the published data of sufficient quality for use in assessing exposures?

In terms of the elements of the studies reviewed here, no single study was assigned to the highest quality tier (Tier 1, Tables S1–S6) by documenting the steps taken to avoid sample contamination in the field and lab and by demonstrating analyte stability during sample collection, shipping and storage; both sample contamination and analyte stability may impact the internal validity of individual studies (LaKind et al. 2014). Sample contamination was found for many other short-lived, commonly used chemicals, despite great efforts taken to avoid contamination (Barr et al. 1999; Calafat and Needham 2008, 2009; LaKind et al. 2014; Needham, Calafat, and Barr 2007) and inclusion of field and lab blanks is an integral component of quality control (Bartram and Balance 1996; USEPA 2008, 2009). Because 2,4-D is widely used, contamination at low levels is possible; residues on surfaces and equipment might be the source of exposure and/or transferred to a sample (Zhang et al. 2011). The transfer of trace amounts of 2,4-D to condoms may have contributed to detection of 2,4-D in semen samples (Arbuckle et al. 1999). Exposure of collection materials or matrix to environmental media might result in falsely elevated concentrations of the target analyte.

As shown in Figure 1, analytical LOD for 2,4-D in urine have declined, and therefore the accuracy of measurements of trace levels is increasingly important compared to periods when LOD were comparatively high. For urine biomonitoring, many of the studies reviewed did not include information describing whether participants were instructed on keeping their hands and other body parts free of 2,4-D while collecting samples or how to keep sampling containers free of 2,4-D. Similarly, few studies described field equipment cleaning procedures. This, in combination with lack of blanks, raises concerns regarding the internal validity for these studies, (Figure 1).

An important factor for internal validity of each study is documentation of 2,4-D stability from the time of sample collection to the time of analysis. Considering the short environmental half-life of 2,4-D and potentially long sample holding times under variable conditions, substantial degradation may occur. Walters (1999) noted that environmental half-lives-ranging 1 to 393 days- depended upon medium, pH, presence of sunlight, oxygen, and microbial activity. 2,4-D in water samples stored at low temperatures (4° C) declined substantially over a period of less than a month (McManus et al. 2014). Data on sample stability for 2,4-D in air were available for only 1 week (25°C) (NIOSH 1994). While short-term concentrations of 2,4-D-spiked urine in gauze pads kept at 37° C in a water bath did not change markedly over 24 hr (Hu et al. 2000), Davis et al. (2013) indicated that additional work is needed to determine the long-term storage stability of 2,4-D conjugates in urine. Harris et al. (2002) who stored urine samples in the freezer for up to 9 months prior to analyses reported no significant differences in mean recoveries among five different storage methods. Most publications reviewed here did not report sample storage times and conditions (Tables S2–S6). Researchers may store samples before and/or after shipping for months or years. A lack of information on storage times or conditions and information on stability of analyte raises questions regarding robustness of the generated data.

Figure 1. Temporal decline of LODs in 2,4-D literature on urine biomonitoring.

Uncontrolled sample contamination may lead to falsely elevated concentrations while analyte degradation during storage may result in falsely lower concentrations. Both erode the accuracy of the exposure information, the validity of using the data in exposure assessments, and ultimately the ability to employ the data to make informed public health decisions. Documentation of quality control is essential. The increasing availability of online publications without page limits provides an opportunity for researchers to make more quality information available, either from primary investigators or by contract labs. Tier 2 and Tier 3 study data may be informative depending upon the purpose of the data. Not all uses require that all elements be Tier 1.

Figure 2. 2,4-D publications by medium/matrix over time.

Are the published data generalizable to other populations?

The published studies reviewed here provide an opportunity to develop an understanding of relative sources and ranges of exposures to 2,4-D in a variety of populations. The data provide a starting point for comparing concentrations in the context of use patterns from one location to another. However, the external validity of the body of literature is also important.

Geography and setting

Most of the reports reviewed focused on relatively small, specific geographic regions. To some extent, this is a strength for exposure assessment as this approach enables coordination with local applications and collection of data on conditions relevant to exposures in food, soil, air, and water including meteorological conditions, food consumed, and contact with environmental media since time of application. Equally, exposure models are often population- and geography-specific. At the same time, assumptions regarding generalizability of these data needs to be made with care. With respect to geography, diet studies reviewed were all from North America, which are not necessarily representative of foods consumed elsewhere. Similarly, 2,4-D data for soil and dust were limited to the USA and Canada. Transfer of contaminated soil into the home depends upon the structural characteristics of the dwelling, which vary by geographic region, culture, climate, and economic status. Further, actual exposure to the residents may also be impacted by personal hygiene practices such as showering after work and frequency of cleaning the home.

In contrast, the studies of water and air exhibited a more global scope, spanning North America, Europe, and other continents. However, the generalizability of a given air sample to another population/geographic region is similarly complex as factors that impact air and water concentrations such as meteorological conditions also vary across regions and countries.

Many of the water studies were conducted to assess changes in concentration in a water body over various time periods post-application. While this provides useful surveillance information, other information needs to be known before applying these data to more general models of human exposures. This includes details of the 2,4-D application and meteorological conditions, as well as other factors that affect surface water concentrations such as thermal gradients in the water body. Perhaps the most useful water data are those derived from investigations of drinking water, including dugouts used for potable water which represent a worst-case exposure scenario due to their proximity to tilled farmland (Cessna and Elliott 2004; Grover et al. 1997; Waite et al. 2002a). However, even these data are difficult to interpret; as observed by Donald et al. (2001), levels of 2,4-D in wetlands were similar regardless of whether 2,4-D was applied a few days prior or not for years. This suggests that 2,4-D may be transported from the site of application and be regionally and evenly distributed (Donald et al. 2001).

In terms of the biomonitoring studies, there are numerous issues that limit the generalizability of urinary 2,4-D data. These include variations in (1) direct contact with the 2,4-D application process, (2) number of acres treated, (3) gloves and other personal protective equipment use, (4) time between exposure and sample collection, (5) formulation, and (6) environmental factors such as temperature, humidity, wind speed, and direction, and application method (Alexander et al. 2007; Arbuckle et al. 2005; Garry et al. 2001). Collection of these types of data enhance the ability to use biomonitoring results to inform exposure modeling of other global settings, where, as noted by Panuwet et al. (2008), farmers may handle pesticides inappropriately.

Temporal changes

The studies reviewed here span 5 decades (Figure 2) and data were generated using a wide variety of sampling and analytical approaches. The year/decade the experiment was conducted might be highly relevant in weighing the value of data for conducting current exposure assessments as 2,4-D formulations, applications techniques, farming practices, and analytical methods have advanced since the time of the earliest studies. An assessment of temporal changes related to these issues is necessary to understand whether older study results are relevant to today’s exposures in North America and elsewhere.

Various changes occurred since the 1960’s that likely resulted in decreased human exposure to 2,4-D. Arbuckle et al. (2005) found lower urinary 2,4-D concentrations in farmer applicators after 1995 and attributed improvements in education, equipment, and labeling to declining occupational exposure. Early investigators acknowledged that differences in volatility by formulation type impact amounts of drift and volatilization (Farwell et al. 1976). The form of the 2,4-D in the formulation also determines its behavior in the environment (Waite et al. 2002a). More volatile 2,4-D esters such as propylene glycol butyl ester, methyl ester, isopropyl ester and butyl ester were replaced in the 1980’s by less volatile 2,4-D esters, thereby reducing volatilization after application (Peterson et al. 2016). These changes in moiety and agricultural practices are positive steps toward reducing both occupational and non-occupational exposures. In addition, agricultural practices changed to reduce soil erosion and water runoff. In turn, these temporal changes led to reductions in herbicide residue loss. Frank, Logan, and Clegg (1991) attributed decrease in herbicides in water levels from the mid-1970’s to the time of their study in the late 1980’s to improved educational programs and conservation practices. As a result, care needs to be taken in reconstructing exposure in the distant past if using data from the present (and vice versa), particularly for use in epidemiology investigations (Alexander et al. 2007). While exposure opportunities today may be higher due to increased uses for agricultural practices and urban encroachment, it may also be lower due to other temporal variations.

Analytical methods have also changed over time resulting in lowered LOD for 2,4-D in urine (Figure 1). The LOD for 2,4-D in water samples ranged from 50 ng/L for samples collected in 1989 (Waite et al. 2002a) to 0.2 ng/L for samples collected in 2013 (Birch et al. 2015). This presents a challenge when comparing results across studies. Between 1988–1994, initial general population surveys in the US demonstrated minimal to no detectable 2,4-D exposures in urine sampled when LOD was 1 µg/L (Centers for Disease Control and Prevention, 2015; Hill et al. 1995a). By 2009–2010, the LOD was almost an order of magnitude lower (0.15 µg/L), and levels of urinary 2,4-D were detectable for half of the sampled population (Centers for Disease Control and Prevention, 2015).

Peak versus background exposures

A challenge to estimating human exposures is understanding whether exposure data represent peak or background. For 2,4-D concentration data in environmental media, the half-life varies by environmental conditions such as amount of sunlight, presence of oxygen and microbes, as well as weather conditions. In assessing environmental data for external validity, these factors need to be measured and considered.

For interpreting biomonitoring data, the short physiological half-life in humans (17.7 hr) (Sauerhoff et al. 1977; Walters 1999) also complicates the ability to infer exposure to the past or future months or years. In general, one cannot state with confidence that a single measurement represents a reliable longer-term exposure estimate. Studies that collected repeated samples during and after work/application demonstrated temporal variability in urinary 2,4-D levels (Alexander et al. 2007; Arbuckle et al. 2005; Kolmodin-Hedman and Ene 1980). To illustrate, among corn farmers, Bakke et al. (2009) detected higher urinary 2,4-D levels compared to controls, even during periods when pesticides were not recently applied, suggesting that farmers may be exposed to 2,4-D through sources that are not related to the actual application of 2,4-D. More detailed information is needed when attempting to use study-specific data to generalize to other populations.

More to this point, many of the investigations reviewed here collected only one sample. The validity of a single sample improves with repeatability of the activity or exposure source (Aylward et al. 2014). Arcury et al. (2009) recommended that research should “include multiple measures of dose or exposure collected at frequent intervals (daily or weekly) over an extended period with close evaluation of risk factors for exposure. Repeated exposure measures will allow us to discern the trajectory of exposure and the factors that are associated with exposure.” In addition, timing of urine collection is important when assessing exposure. In conclusion, improper use of the available biomonitoring data for estimating exposures within or beyond the examined population may result in exposure misclassification (LaKind et al. 2014).

What additional information can be gleaned from large-scale population- and media-based studies available from governmental reports?

Large-scale governmental surveys are undertaken to shed light on sources, routes, and levels of exposure to the general population. Surveys such as the US Pesticide Data Program and US Geological Survey National Water Quality Assessment Program evaluate up to thousands of samples annually in food and drinking water, respectively (Gilliom 2007; USDA 2004). The European Food Safety Authority also reviews residues on plants and foods as part of its risk assessment procedures (EFSA 2016). These large studies report the presence of 2,4-D in various fractions of environmental samples and humans. In the USA, 2,4-D was detected in 16% of 24 drinking water samples in 2002 but in no food samples in 2014 (USDA 2004, 2015). In Europe, 2,4-D was found in fewer than 5% of the food samples (PDP 2014 ; EFSA 2014). A compilation of ambient air data from the USA (14 locations excluding remote locations and areas with direct source impacts; N > 1163) reported a range of ambient air concentrations for 2,4-D (from < LOD to 4 ng/m³⁾ (Kelly et al. 1994), although these data may not be relevant to today’s air concentrations due to changes in practices and extent of applications since 1994.

As previously noted, nationally representative biomonitoring studies of urinary 2,4-D from the US and Canada from 1999 to 2010 reported detectable concentrations for up to half of the sampled population (Canada 2010; Centers for Disease Control and Prevention 2015) (LOD ≤ 0.2 µg/l). This population percentage with detectable levels suggests routes of exposure other than actual applications of 2,4-D and drift from proximity to application. However, these surveys did not include specific information on sources of exposure. Associated data reported by the Add, Centers for Disease Control and Prevention (CDC) as part of the US National Health and Nutrition Examination Survey (NHANES) may be used to test certain hypotheses related to 2,4-D exposure. The publicly available NHANES data were utilized on fasting time as a proxy for contribution of dietary exposure to urinary 2,4-D levels. In adults 21 years and older (N = 1,925), the associations between fasting hours and unadjusted and creatinine-adjusted log urinary 2,4-D levels were not statistically significant. This agrees with results of the study of pregnant women (Lewis et al. 2014), in which no significant associations were noted between urinary 2,4-D and consumption of nearly all fruits, vegetables, and legumes (except collards or spinach). In contrast, for children participating in NHANES (<21 years of age, N = 822), significant inverse associations between fasting hours and log urinary 2,4-D were observed (unadjusted: slope = −0.001, p-value = 0.00083; creatinine-adjusted: slope = −0.015, p-value = 0.00033). In other words, longer fasting times were correlated with lower concentrations of 2,4-D in urine, suggesting food as a possible source of exposure in US children. This is in agreement with results of an exposure study of children (Wilson et al. 2010) which found that dietary exposures accounted for almost 90% of the aggregate dose of 2,4-D.

The NHANES questionnaire information on application of lawn herbicides was utilized as a proxy for contact with 2,4-D recognizing that not all herbicides contain 2,4-D. The relationship was examined between urinary 2,4-D concentrations and response to the NHANES question “In the past 7 days, were any chemical products used in the lawn or garden to kill weeds?”. The association was positive and significant for adults (N = 1,715) for both unadjusted and creatinine-adjusted urinary levels (p = 0.00021 and 0.0011, respectively); the correlation was not significant for children (N = 807). In adults, a positive response to this question had the effect of increasing urinary 2,4-D concentration by an estimated 75%. Thus, recent use of these products appeared to exert an observable effect on the general US population.

In summary, for governmental reports described here, large studies up to thousands of samples offer information regarding the distribution of exposure within a region or population. However, in the absence of data on timing and uses of 2,4-D, generalizability of this information regarding specific target groups or peak versus background exposure is poor. Nonetheless, data from large-scale governmental investigations may generate hypotheses on which to base future research.

Conclusions

The quality of human health studies is directly reliant upon quality of exposure data. With growing interest in using human exposure and epidemiology data for public health decision-making, the importance of high-quality exposure data has increased. A published quality assessment instrument was utilized to systematically review the large body of literature on 2,4-D exposure data and it was found that few publications (1) documented steps taken to avoid sample contamination in the field and/or lab or (2) demonstrated the stability of the analyte over the time from sample collection through analysis. The possibility that this may reflect incomplete documentation in the final publication(s) is recognized. Only a small number of studies included duration of time between 2,4-D application and sample collection, limiting our ability to evaluate whether results reflected peak or background exposures. Other investigations focused on general surveillance and did not seek to examine the relationship between reported concentrations of 2,4-D and temporal or geographic proximity to 2,4-D applications.

To use the 2,4-D concentration data effectively to support epidemiological investigations, risk assessments and regulatory decisions, improved demonstration of quality control measures in existing and future published studies is needed. Studies with high quality elements were identified in this review, demonstrating the feasibility of using and reporting on quality control elements, either in the body of a publication or as supplemental material. Difficulties in conducting exposure studies of short-lived chemicals with variable formulations, uses, and application processes are recognized. Additional complexities are introduced by factors such as meteorological conditions and human behavior. It is necessary to emphasize that categorization of a quality element as Tier 2 or 3 does not necessarily preclude its use; this decision would depend on professional judgment.

Perhaps the greatest challenge for the exposure community is not the inclusion and documentation of quality control measures, but rather the ability to reach consensus addressing problems specific to short-lived chemical exposures in observational studies. A more comprehensive conversation is needed to address critical issues such as (1) feasibility of timed data collection to minimize exposure misclassification, (2) minimum data requirements for distinguishing peak from background concentrations, and (3) study design improvements to enhance generalizability of study data for advancing our understanding of human exposures to short-lived chemicals, as well as to allow interpretation of these data to catch up to analytical capabilities. Exposure science problems described in this review remain exquisitely difficult to address for chemicals like 2,4-D, with short half-lives and with exposures impacted by numerous and changing external factors including those related to application practices and formulations.

Acknowledgments

The Task Force was not involved in the design, collection, management, analysis, or interpretation of the data; or in the preparation or approval of the manuscript.

Funding

This work was supported by the Industry Task Force II on 2,4-D Research Data. The 2,4-D Research Task Force is made up of those companies owning the technical registrations on the active ingredient in 2,4-D herbicides. They are Dow AgroSciences (U.S.), Nufarm, Ltd. (Australia) and Agro-Gor Corp., a U.S. corporation jointly owned by Alb augh, LLC. (U.S.) and PBI-Gordon Corp. (U.S.).

Supplemental data

Supplemental data for this article can be accessed at the publisher’s website.

Additional information

Funding

This work was supported by the Industry Task Force II on 2,4-D Research Data. The 2,4-D Research Task Force is made up of those companies owning the technical registrations on the active ingredient in 2,4-D herbicides. They are Dow AgroSciences (U.S.), Nufarm, Ltd. (Australia) and Agro-Gor Corp., a U.S. corporation jointly owned by Alb augh, LLC. (U.S.) and PBI-Gordon Corp. (U.S.).