The United States is the world's third largest producer of edible peanuts (Arachis hypogaea L.), behind China and India. In 2023, the U.S. harvested 636,315 hectares with an average yield of 4,192 kg/ha for a total in-shell production of 2.67 million metric tons (National Agricultural Statistics Service, 2024). U.S. peanuts are produced in three major regions consisting of the Southeast (Alabama, Florida, Georgia, Mississippi, South Carolina), Southwest (New Mexico, Oklahoma, Texas, Missouri, Louisiana), and Virginia-Carolina (North Carolina, Virginia). There are four different commercial peanut market types (runner, Spanish, Valencia, and Virginia), each with distinguishing characteristics related to final product use. Runner peanuts comprise approximately 80% of U.S. production, are grown primarily in the Southeast region, and primarily used for producing peanut butter. The Spanish type, which accounts for about 4% of total U.S. production, is primarily used in the production of peanut snacks, with the majority produced in Texas and Oklahoma. Valencia peanuts represent about 1% of U.S. production and are produced in New Mexico and Texas, and mostly used as roasted or boiled in-shell. Virginia peanuts comprise 15% of U.S. production and are produced in Virginia, North Carolina, South Carolina, and Texas. The primary use of Virginia peanuts is roasted in-shell (Agricultural Marketing Resource Center, 2024).

The U.S. peanut supply chain is unique compared to most commodities in that peanuts are delivered from the field in-shell (farmer stock) and must undergo further processing to remove the shell and then size peanut kernels (shelled stock) specific to final use. At harvest, farmer stock peanuts are delivered to peanut buying points. Peanut buying points are strategically located in the peanut producing regions for efficiency related to farmer stock peanut transport and represent the first point of consolidation within these smaller geographic regions. Buying point functions include cleaning, drying, grading, and storing peanuts. All farmer stock peanuts must be inspected following strict guidelines mandated by the USDA Agricultural Marketing Service and conducted by trained Federal State Inspection Service (FSIS) inspectors located at each buying point (Archer, 2016) to determine quality factors of farmer stock peanuts. Farmer stock peanuts are then stored in specially designed warehouses until they are delivered to shelling facilities for further processing. The basic function of peanut shelling is to remove peanut hulls from kernels and size peanuts into specified commercial market grades.

While inspection at the buying points focuses on the physical characteristics of the peanuts (Cowart et al. 2016), our focus in this analysis is to provide an evaluation of other factors, specifically the presence, if any, of pesticide residues in the U.S. peanut crop.

U.S. peanut growers use a variety of pesticides at various stages of production. These include herbicides such as glyphosate and 2,4-D used during preplant burndown, 2,4-DB used post planting, Fusilade (fluazifop-P-butyl) and Select (clethodim) as postemergence annual grass control; fungicides including Propulse (fluopyram and prothioconazole)) and Lucento (fluatrifol and bixafen); nematicides such as Velum (fluopyram) and Propulse; and a wide variety of insecticides including Brigade (bifenthrin) and Asana (esfenvalerate) (University of Georgia Extension, 2024).

The mere presence of pesticide residues does not necessarily imply that the food is unsafe. The toxicity of a particular pesticide residue depends on a risk assessment which accounts for the amount of the residue present, the degree of exposure, consumption patterns, and the pesticide’s toxicology (US Environmental Protection Agency, 2000). Most developed economies have their own determination of what level of residues are deemed unsafe.

In the United States, the EPA sets a maximum legal residue limit (MRL) or tolerance for each residue for each treated food. The pesticide residue tolerances apply to food grown in the U.S. and imported food. In setting the MRL for pesticide use, the EPA uses a "reasonable certainty of no harm" safety threshold which considers the toxicity of the pesticide directly, the toxicity of its breakdown products, how much of the pesticide is applied and how often, and how much of the pesticide residue remains in or on food by the time it is marketed (EPA.gov, 2024).

Internationally, since 2008, the European Union (EU) has established MRL standards that facilitate trade within the EU and imports into the EU (European Communities, 2008). In addition to MRLs, the European Chemicals Agency (ECHA) has derived a set of safety standards that take into account the consumption pattern of foods that may have pesticide residues. The acceptable daily intake (ADI) provides an estimate of the amount of a substance in food or drinking water that can be consumed per day “over a lifetime without presenting an appreciable risk to health.” (WHO, 1987 as cited by ECHA, 2016). The acute reference dose (ARfD) provides an estimate of the amount of a substance in food or drinking water that can be ingested over a short period of time, e.g., one meal or one day, without appreciable health risk to the consumer (JMPR, 2002 as cited by ECHA, 2016). Both standards are expressed in terms of body weight, e.g., milligrams per kilogram of body weight per day.

Empirical studies that examine the presence of pesticide residues on foodstuffs and provide an associated risk assessment are wide-ranging, encompassing many different types of crops, various chemical residues, and several different geographic regions. Most of these studies share a common methodology in the testing for residues as well as their risk assessments. These risk assessments typically involve using the ARfD to measure acute exposure health risk and the ADI to measure chronic exposure health risk.

The research presented here has two main objectives. First, it would be instructive to provide descriptive statistics for the level of pesticide residues detected in the test results. However, due to the high degree of censoring, estimates of the descriptive statistics can be misleading. Instead, we present 95% confidence intervals for the probability that the residue would be detected in another random sample of peanuts. Second, based on internationally accepted safety standards and accepted methodologies, we estimate short-term and long-term health exposure risk assessments for consumers of peanuts.

The study was conducted during the 2020-2022 crop years at farmer stock delivery to peanut buying points. The laboratory samples were chosen during the farmer stock grading process at each buying point when peanuts are split for internal damage detection. At the end of the day, the FSIS samples used in the grading process were comingled and a randomly drawn 2-kg sample was removed for laboratory testing. These samples were sent to the Eurofins laboratory located in New Orleans for pesticide screening. To get a representative sample of U.S. peanut production, samples were selected from the three major growing regions, the Southeast (SE), Southwest (SW), and the Virginia, North Carolina, South Carolina (VC) region. Samples included all three major market types: runner (all regions), Virginia (SW and VC), and Spanish (SW only), drawn from buying point locations in nine states. The proportion of samples drawn from each region was chosen to represent, as closely as possible, that region’s percentage of total peanut acres and market types. Table 1 describes the sample selection in more detail. The following analysis assumes that the samples tested represent random samples of peanut production for each region and overall U.S. peanut production.

Once at the Eurofins laboratory, each 2-kg sample was composited and homogenized. A representative sample was then taken for testing. Although testing packages varied from year to year, a common set of 537 pesticide residues was tested in each of the three years. Table 2 describes the Eurofins testing codes used and the associated analysis type.

Even though extensive testing for pesticide residue was conducted, only the 14 residues described in Table 3 were detected. Each of these residue concentrations was below the European Union’s maximum residue limit for food safety. All other residues were undetectable given the detection limits of the testing equipment.

Due to the high degree of censoring, estimates of descriptive statistics for the detected residues can be misleading. As a result, for each of the 14 residues that were detected, we instead focus on estimation of 95% confidence intervals for the probability that the residue would be detected in another random sample of peanuts. To do so we follow the methodology described in Helsel (2012), by first calculating the sample estimate for the proportion of the data below the detection limit for each residue, p=c/n, where c is the number of censored observations and n is the total number of observations. If not all test results censored, c<n, a 95% confidence interval around p is constructed as LL, UL where LL and UL are the lower limit and upper limit of the confidence interval given by

and

When all test results are censored, c=n, where LL and UL become

and UL=1. In each of these cases, the confidence interval for the proportion of test results above the detection limit can be constructed as [L, U] where L=1-UL and U=1-LL.

In addition to the probability confidence intervals, we also provide an assessment of the risks faced by consumers of peanuts due to the possibility of pesticide residues. To do so, we follow a methodology that has been used to assess pesticide residues on a wide variety of other agricultural products. Recent examples include the assessment of pesticide residues in samples of chestnuts, walnuts, and pine nuts produced in China (Liu et al., 2016), in dried nuts and fruits collected from bazaars and markets in Turkey (Aydin and Ulvi, 2019), and on vegetables commonly consumed in Mexico (Calderon, et al., 2022). The risk assessment consists of an acute or short-term health risk index, aHI, and a chronic or long term measure, also called the hazard quotient, HQ. Each of these measures are calculated for each pesticide residue detected. The measure of acute health risk is calculated as

where ESTI is an estimated short-term intake and is calculated for each residue as

and ARfD is the acute reference dose. The greater the value of aHI, the greater the short-term health risk, and an aHI>100 indicates an unacceptable short-term health risk.

The measure of chronic health risk commonly used in the literature is based on the average level of the residue in question. However, since we lack enough data to accurately measure the average, we instead calculate an upper limit of the mean, ULM, by replacing the censored values with the limit of detection. This results in an overestimation of the health risk posed. The measure of chronic risk presented here is calculated as

where EDI is an estimated daily intake and is calculated for each residue as

and where ADI is the acceptable daily intake. Basing our measure on the upper limit mean rather than the average level of the residue results in an upper limit of the estimated health risk posed. The greater the value of HQ, the greater the long-term health risk, and an HQ>100 indicates an unacceptable long-term health risk.

The ARfD and ADI values used in the above calculations are established jointly by the World Health Organization (WHO) and the Food and Agriculture Organization of the United Nations (FAO) by their Joint FAO/WHO Meeting on Pesticide Residues (JMPR) and compiled in the JMPR database (WHO, 2021).

Intuitively, each of the health risk indices measures the daily intake of a pesticide residue measured in milligrams per kilogram of bodyweight relative to safety consumption standards also measured in milligrams per kilogram bodyweight, either the amount that can safely be consumed in one day, the ARfD, or the amount that can be safely consumed over a lifetime, the ADI. If the amount consumed exceeds the safety standard, continued health may be at risk.

For each of the pesticide residues that were detected (see Table 3), we describe the sample data and the detection results for each of the three years included in the study and provide a confidence interval for the probability of selecting a sample in which the residue is undetectable. We then discuss the health risk assessment for each residue.