Dr. Carol Weisskopf, Analytical Chemist, WSU
The Food Quality Protection Act (FQPA) focuses pesticide risk evaluation on human exposure, the bulk of which is considered to be from food. Dietary exposure is not a particularly easy number to determine. The calculation of exposure levels seems straightforward: one can take the average pesticide concentration found on a particular food item multiplied by the amount of that food in the average diet. Contributions of a pesticide from all food items in the average diet would make up the total exposure for that chemical. This exposure level could be compared to some level known or thought to be safe. Safety factors taking into account variability in pesticide concentrations, consumption patterns and individual susceptibility would be applied, and one could decide if the (overused metaphoric) risk cup is overflowing. Each step of this extrapolation provides ample grounds for debate.
One component of this equation falls directly into the laps of the analytical chemists: determining the concentration of pesticide on a food item. The chemist uses or modifies an existing analytical procedure, or develops a new one, for analysis of a particular chemical on a specific food item. Often this process is simple and occasionally it's not, but residue chemists do these types of analyses regularly. The quality of the data (are the numbers right?) is easy to check. To ensure that the analyses accurately represent the distribution of concentrations present in a food (are the numbers typical?), large numbers of samples can be analyzed. Most of the argument can then center on what the numbers mean the province of toxicology and risk analysis, not chemistry.
All of this works quite well when there are actual concentration numbers to deal with. One might think that not finding pesticides in a sample would make data evaluation easier than when pesticides were found. Actually, dealing with 'non-detects' is more difficult. Detections and detection limits, their impact on data analysis, and their bearing on FQPA implementation are the subjects of this and next month's contributions to the AENews.
What is a detection limit? It all depends on how you look at it. In this discussion, I'll start with the equipment and work backwards. First, there is the detector sensitivity. That is the minimum amount of a chemical that we need to deliver to a particular detector to 'see' it. Our senses (particularly taste and smell) are good analogues for instrumentation. I am capable of detecting 0.1mg of sucrose in a direct delivery (I can taste 1 grain of sugar when I put it right on my tongue). This is an amount, not a concentration. But, just as we rarely encounter sugar in pure form in our diet, this is not the most typical sample description for our instrumentation.
We are rarely able to put small amounts of pure material (like sugar) right in one of our detectors. We usually need to have the chemical in a solvent, a small amount of which is introduced into the chromatographic system preceding the detector. In a 2 milliliter (mL) delivery, I can detect sucrose in dihydrogen oxide at a concentration of 3 milligrams (mg)/mL (in a sip, I can taste a concentration equal to a quarter of a teaspoon of sugar in one cup of water). In this example, I have had to deliver 60 times more sugar to be able to taste it, which is the case with our instruments as well. Sample delivery is important. If a sprayer rather than a spoon was used for tasting, there would have to be a lot more sugar to detect it. This concentration is the instrumental detection limit. It is a function of the detector, the chromatographic system and the injection method.
This detection limit is not the same as the limit for the concentration in our sample. We can take our original cup full and boil it down to a final volume of 2 mL (a sip). All the sugar should stay, although I haven't actually tried it. At the instrumental detection limit (taste threshold) of 3 mg/mL (3,000 ppm) previously established, boiling down the sample would allow the concentration of sugar in the original (pre-boiled) water sample to be 0.024 mg/mL, which is 24 ppm! This is the method detection limit, and depends on the sample size. If we boil down a one gallon sample, the method detection limit would be 1.5 ppm.
Detector sensitivity varies somewhat by manufacturer. A single chemical may also be analyzed by more than one detector, at different sensitivities (there is a discernable smell to sugar water, but it has to be extremely concentrated). The sample size can vary, depending on availability. Finally, the detection limit can vary according to the matrix, or type of sample. Water is easy. If we took a gallon of coffee and boiled it down to one teaspoon, we would need a lot of sugar to be able to taste it. In fact, with a gallon of boiled-down coffee, the amount of sugar necessary to taste probably wouldn't fit in a teaspoon.
Coffee, in comparison to water, is what chemists refer to as a 'dirty' sample. Chemists might like our coffee strong (our horseradish hot and our garlic robust) but we prefer our samples wimpy. A dirty sample needs cleanup steps in the sample work-up, and still can't reach the same method detection limits as a clean sample. Dirty samples are harder to analyze, take longer, the results are usually more variable, and the detection limit higher than is the case for clean samples. Things that our personal detectors work well on (highly flavored, odored or colored) also produce high backgrounds in chemical analyses. Clean commodities are water and iceberg lettuce. Hard ones include onions, the cabbage family, horseradish, and hops.
Variabilities in sample types (water, coffee, lemon juice) inevitably lead to differences in method detection limits among commodities. Variabilities in instrumental detection limits because of the detector used in an analysis (taste vs. smell), as well as efficiency of the chromatographic systems, lead to differences between labs. We also need to throw in differences among chemicals analyzed. By both odor and taste, we could detect sugar more easily than corn starch in coffee or in water.
What happens when we can't taste a quarter of a teaspoon of an unidentified white powder in our cup of coffee? If it were sugar, it wouldn't be an issue even if we were overweight. If it were a plutonium salt the folks next door at Hanford would certainly get excited. It would be convenient if the most toxic compounds had correspondingly low detection limits. However, among the most easily detected compounds in residue chemistry are phthalates. Several of our detectors respond well to these industrial chemicals. They are present in plastics and a host of other products (including the plastic wrap in which vegetables are occasionally packaged), are ubiquitous in the laboratory and in the environment. The banned herbicide Dinoseb is more difficult to detect, although considerably more toxic.
Easy and hard crops, sensitive and unresponsive pesticides, new and old instruments all conspire to make detection limits moveable objects. It's not rocket science. We don't usually need to hit the moon. A method detection limit comfortably below an established tolerance or environmental criteria has been good enough. In the past, there has rarely been a compelling reason to push to the absolute limits of method or instrument sensitivity in routine analyses.
A method that is more sensitive than needed is not cost-effective. Lower sensitivities are achieved with increase in personnel time, supplies and equipment costs. The detection limit required for an analysis depends on the intended use of the data. If one were looking at the spray pattern from a piece of application equipment, the concentrations are expected to be high, and we would probably be making dilutions rather than pushing our detection limits. If we were making sure that a pesticide concentration was below the tolerance levels established for a commodity, a method detection limit one-tenth of the tolerance would be comfortable. Answering some environmental fate questions, such as the transport of pesticides in fog, snow or dust, might require maximum sensitivity from our methods and instruments.
When we complete a study, we generally have some samples in which a pesticide
was found, and some in which it was not detected. When we find a pesticide,
the use of the data is usually straightforward: concentrations exceeded
some regulatory level or not; they increased or decreased; they were higher
or lower here than there. Next month, I'll talk about what we think is happening
in the imaginary land below the detection limits, how we deal with non-detects,
and how FQPA has raised the ante on detection limits in tolerance setting
and pesticide use.
Return to the Table of Contents for theAugust 1998 issue
Return to the Agrichemical & Environmental News Index