Evaluating the reliability of environmental concentration data to characterize exposure in environmental risk assessments.

Environmental risk assessments often rely on measured concentrations in environmental matrices to characterize exposure of the population of interest-typically, humans, aquatic biota, or other wildlife. Yet, there is limited guidance available on how to report and evaluate exposure datasets for reliability and relevance, despite their importance to regulatory decision-making. This paper is the second of a four-paper series detailing the outcomes of a Society of Environmental Toxicology and Chemistry Technical Workshop that has developed Criteria for Reporting and Evaluating Exposure Datasets (CREED). It presents specific criteria to systematically evaluate the reliability of environmental exposure datasets. These criteria can help risk assessors understand and characterize uncertainties when existing data are used in various types of assessments and can serve as guidance on best practice for the reporting of data for data generators (to maximize utility of their datasets). Although most reliability criteria are universal, some practices may need to be evaluated considering the purpose of the assessment. Reliability refers to the inherent quality of the dataset and evaluation criteria address the identification of analytes, study sites, environmental matrices, sampling dates, sample collection methods, analytical method performance, data handling or aggregation, treatment of censored data, and generation of summary statistics. Each criterion is evaluated as "fully met," "partly met," "not met or inappropriate," "not reported," or "not applicable" for the dataset being reviewed. The evaluation concludes with a scheme for scoring the dataset as reliable with or without restrictions, not reliable, or not assignable, and is demonstrated with three case studies representing both organic and inorganic constituents, and different study designs and assessment purposes. Reliability evaluation can be used in conjunction with relevance evaluation (assessed separately) to determine the extent to which environmental monitoring datasets are "fit for purpose," that is, suitable for use in various types of assessments. Integr Environ Assess Manag 2024;20:981-1003. © 2024 Society of Environmental Toxicology & Chemistry (SETAC). This article has been contributed to by U.S. Government employees and their work is in the public domain in the USA.

Using wastewater-based epidemiology to estimate drug consumption-Statistical analyses and data presentation.

AIM: Analysis of wastewater samples can be used to assess population drug use, but reporting and statistical issues have limited the utility of the approach for epidemiology due to analytical results that are below the limit of quantification or detection. Unobserved or non-quantifiable-censored-data are common and likely to persist as the methodology is applied to more municipalities and a broader array of substances. We demonstrate the use of censored data techniques and account for measurement errors to explore distributions and annual estimates of the daily mean level of drugs excreted per capita. MEASUREMENTS: Daily 24-hour composite wastewater samples for 56days in 2009 were obtained using a random sample stratified by day of week and season for 19 municipalities in the Northwest region of the U.S. METHODS: Methamphetamine, benzoylecgonine (cocaine metabolite), 3,4 methylenedioxymethamphetamine (MDMA), methadone, oxycodone and hydrocodone were identified and quantified in wastewater samples. Four statistical approaches (reporting censoring, Maximum Likelihood Estimation, Kaplan-Meier estimates, or complete data calculations) were used to estimate an annual average, including confidence bounds where appropriate, dependent upon the amount of censoring in the data. FINDINGS: The proportion of days within a year with censored data varied greatly by drug across the 19 municipalities, with MDMA varying the most (4% to 94% of observations censored). The different statistical approaches each needed to be used given the levels of censoring of measured drug concentrations. Figures incorporating confidence bounds allow visualization of the data that facilitates appropriate comparisons across municipalities. CONCLUSIONS: Results from wastewater sampling that are below detection or quantification limits contain important information and can be incorporated to create a more complete and valid estimate of drug excretion.

Additional analysis of dendrochemical data of Fallon, Nevada.

Previously reported dendrochemical data showed temporal variability in concentration of tungsten (W) and cobalt (Co) in tree rings of Fallon, Nevada, US. Criticism of this work questioned the use of the Mann-Whitney test for determining change in element concentrations. Here, we demonstrate that Mann-Whitney is appropriate for comparing background element concentrations to possibly elevated concentrations in environmental media. Given that Mann-Whitney tests for differences in shapes of distributions, inter-tree variability (e.g., "coefficient of median variation") was calculated for each measured element across trees within subsites and time periods. For W and Co, the metals of highest interest in Fallon, inter-tree variability was always higher within versus outside of Fallon. For calibration purposes, this entire analysis was repeated at a different town, Sweet Home, Oregon, which has a known tungsten-powder facility, and inter-tree variability of W in tree rings confirmed the establishment date of that facility. Mann-Whitney testing of simulated data also confirmed its appropriateness for analysis of data affected by point-source contamination. This research adds important new dimensions to dendrochemistry of point-source contamination by adding analysis of inter-tree variability to analysis of central tendency. Fallon remains distinctive by a temporal increase in W beginning by the mid 1990s and by elevated Co since at least the early 1990s, as well as by high inter-tree variability for W and Co relative to comparison towns.

Summing nondetects: incorporating low-level contaminants in risk assessment.

Low-level contaminants often are present below the detection or reporting limits of a laboratory, resulting in values reported as a nondetect or less-than. How can these values be summed along with detected concentrations to obtain a total, particularly when weighting factors such as toxic equivalence factors (TEFs) are used? The most common method employed by environmental scientists for summing nondetects along with detected values is to substitute one-half the detection limit for each nondetect. This substitution allows the least precise measurements, data with high detection limits, to have a strong influence on the resulting total amount. Substitution methods have repeatedly been shown to provide substandard results in studies over the last 2 decades. Here an alternative, the Kaplan-Meier (KM) method used throughout the fields of medical and industrial statistics, is used to obtain the total. KM estimates are far less affected by the least precise data than are estimates computed using substitution. No assumptions about the distribution of data (whether they follow a normal or other distribution) need be made. Direct application of KM to computation of toxicity equivalence concentrations (TECs) is shown.

Much ado about next to nothing: incorporating nondetects in science.

A great many papers and one textbook have been published on the topic of how to incorporate 'nondetects', low-level values reported only as below a detection limit, into statistical analyses. This is of interest not only in occupational hygiene but also in environmental sciences and astronomy, among other fields. Here, the literature is reviewed from the earliest known publication on the topic >40 years ago and recommendations contrasted. I have tried to pull some unifying conclusions out of the mix, ending with four suggestions I believe all can agree on. See if you agree with me.

Methyl tert-butyl ether (MTBE) in public and private wells in New Hampshire: occurrence, factors, and possible implications.

Methyl tert-butyl ether (MTBE) concentrations > or = 0.2 /microg/L were found in samples of untreated water in 18% of public-supply wells (n = 284) and 9.1% of private domestic wells (n = 264) sampled in 2005 and 2006 in New Hampshire. In counties that used reformulated gasoline (RFG), MTBE occurred at or above 0.2 microg/L in 30% of public- and 17% of private-supply wells. Additionally, 52% of public-supply wells collocated with fuel storage and 71% of mobile home park wells had MTBE. MTBE occurrence in public-supply wells was predicted by factors such as proximity to sources of fuel, land use, and population density, as well as low pH and distance from mapped lineaments. RFG use, land-use variables, and pH were important predictors of private-well MTBE occurrence. Variables representing sources of MTBE, such as the distance to known fuel sources, were not significant predictors of MTBE occurrence in private-supply wells. It is hypothesized that private wells may become contaminated from the collective effects of sources in high population areas and from undocumented incidental releases from onsite or proximal gasoline use. From 2003 to 2005, MTBE occurrence decreased in 63 public-supply wells and increased in 60 private-supply wells, but neither trend was statistically significant.

Regional Kendall test for trend.

Trends in environmental variables are often investigated within a study region at more than one site. At each site, a trend analysis determines whether a trend has occurred. Yet often also of interest is whether a consistent trend is evident throughout the entire region. This paper adapts the Seasonal Kendall trend test to determine whether a consistent regional trend occurs in environmental variables.

Fabricating data: how substituting values for nondetects can ruin results, and what can be done about it.

The most commonly used method in environmental chemistry to deal with values below detection limits is to substitute a fraction of the detection limit for each nondetect. Two decades of research has shown that this fabrication of values produces poor estimates of statistics, and commonly obscures patterns and trends in the data. Papers using substitution may conclude that significant differences, correlations, and regression relationships do not exist, when in fact they do. The reverse may also be true. Fortunately, good alternative methods for dealing with nondetects already exist, and are summarized here with references to original sources. Substituting values for nondetects should be used rarely, and should generally be considered unacceptable in scientific research. There are better ways.

Stability of low levels of perchlorate in drinking water and natural water samples.

Perchlorate ion (ClO4-) is an environmental contaminant of growing concern due to its potential human health effects, impact on aquatic and land animals, and widespread occurrence throughout the United States. The determination of perchlorate cannot normally be carried out in the field. As such, water samples for perchlorate analysis are often shipped to a central laboratory, where they may be stored for a significant period before analysis. The stability of perchlorate ion in various types of commonly encountered water samples has not been generally examined-the effect of such storage is thus not known. In the present study, the long-term stability of perchlorate ion in deionized water, tap water, ground water, and surface water was examined. Sample sets containing approximately 1000, 100, 1.0, and 0.5 microg l(-1) perchlorate ion in deionized water and also in local tap water were formulated. These samples were analyzed by ion chromatography for perchlorate ion concentration against freshly prepared standards every 24h for the first 7 days, biweekly for the next 4 weeks, and periodically after that for a total of 400 or 610 days for the two lowest concentrations and a total of 428 or 638 days for the high concentrations. Ground and surface water samples containing perchlorate were collected, held and analyzed for perchlorate concentration periodically over at least 360 days. All samples except for the surface water samples were found to be stable for the duration of the study, allowing for holding times of at least 300 days for ground water samples and at least 90 days for surface water samples.

Complications with using ratios for environmental data: comparing enantiomeric ratios (ERs) and enantiomer fractions (EFs).

Complications arise when ratios are used to present environmental data because ratios are an unbounded, multiplicative scale that can lead to asymmetrical (skewed) data distributions. Enantiomeric ratios (ERs), historically used in discussions of chiral signatures, often are published as mean ER+/-single-value standard deviation. Application of statistical summaries, such as the widely used sample mean and standard deviation, to skewed ratio data is misleading and often inappropriate. Comparison of statistically summarized ER and enantiomer fraction (EF) data (which are based on a bounded, additive scale) for a range of hypothetical values reveals substantial discrepancies when conversion between ER and EF formats is used. These discrepancies are largest when the ratio data are greater than one and have large variability, because the data are more skewed. In many cases, the use of fractions instead of ratios can help to minimize misrepresentation of environmental data, including chiral data. The use of nonparametric statistical summaries, e.g., median and percentiles, provides a more robust indicator of the typical value and spread for both ER and EF data.