The U.S. EPA is commissioned by the Safe Drinking Water Act and the Clean Water Act to develop routine methods for the analysis of drinking water and wastewater, respectively. Results of these methods are used to establish whether water is safe to drink or if an industrial or municipal effluent is in compliance with a permit. There is a concentration for each pollutant, called a Maximum Contaminant Level (MCL); the sample is analyzed by the approved method, and if the contaminant concentration is lower than the MCL, the water is good.
There are difficulties with this approach, including sampling, sample preservation, sample extraction or digestion, analysis and even quality control. This article is not to imply that our approach is wrong; it is merely to share some observations.
Rapid methods are designed to analyze many samples for several constituents in the shortest time possible. The methods are sometimes developed with almost total disregard for variability in matrices or simply assuming that all matrices are created equal. For instance, a method approved for wastewater may have been validated using the required nine matrices, including all the required detection limit and recovery studies, but it still only applies to the matrices tested. There is no way the method can be applied to all matrices and, even if one matrix was a “sewage treatment plant effluent,” the method may not be applicable to all treatment plant effluents.
For this reason, matrix spikes and matrix spike duplicates are analyzed. If a matrix spike fails the method acceptance criteria, the failure is credited to matrix interference and the method performance is referenced to an interference free laboratory control sample (LCS). Then it’s on to the next samples. The actual analyte concentration from that matrix is not reported; instead of reporting results as they really are, they are reported as they are not.
It is not possible, and it is even discouraged, for a laboratory to evaluate each matrix for potential interferences. Laboratory technicians are pressed to run as many samples as they can per day, and regulators do not allow method modifications anyway. (Limited method modification is permitted for wastewater at 40 CFR Part 136.6 and in some drinking water methods; however, once a modification is made, the laboratory applies the newly modified method to all samples.)
Method development organizations, such as U.S. EPA, ASTM and Standard Methods, devise rapid methods with known accuracy and precision. The validation processes may vary, but all evaluate various matrices, sample preservation and holding times, interferences, and conduct interlaboratory trials to estimate single laboratory and multiple laboratory precision. Efforts are usually made to estimate accuracy, either by analysis of known concentration quality control samples or spiked sample recovery. Even with all these precautions, the method developed still only applies to the matrices tested and each method will usually say so.
An example of a validated method that fails to perform in all matrices is the determination of total cyanide by distillation with sulfuric acid solution. The method performs well with adequate accuracy and precision in waters that have no matrix interferences. However, a wastewater effluent that has been de-chlorinated using sodium sulfite may exhibit a negative bias. Spike results on this sample, particularly if the sample was spiked with KCN, will be low. This spike recovery properly evaluated that the sample results are too low. If, however, the sample contained thiocyanate and nitrate, the results are biased high. A recovery of 100% seems good, but fails to detect that the sample result is high because the matrix reacted during the distillation process to create cyanide. Only analysis by another interference-free method will detect the bias.
This leads to problems with quality control (QC) procedures, particularly the use of spiked samples to evaluate accuracy. It used to be that accuracy of a method was estimated by the analysis of the analyte using several accepted techniques in the hands of skilled analysts. If the results agreed, within known and accepted precision, a “true” value was established. This was a common practice in the certification of National Institute of Standard (NIST) reference materials, such as NIST SRM 120c Phosphate Rock (Florida). Environmental samples, however, are not so easy to analyze by multiple analysts and methods as ore minerals due to holding time requirements, among other things. So now, matrix spike recovery is used. Suppose, however, that a water sample is being analyzed for free cyanide by ASTM D7237. The spike is added and the sample analyzed only to find very low recovery. Is the low recovery due to a method flaw, or a laboratory blunder or QC failure? Or is the low recovery because trace metals in the sample complexed the free cyanide added and will continue to do so until enough cyanide is added to fully react with the sample? In the latter case, the low recovery is correct, and it is the use of spike recovery that fails to demonstrate that the method is accurately measuring free cyanide.
In another example, suppose diesel fuel is spiked onto a high-clay mineral sample and then extracted and analyzed for total petroleum hydrocarbons (TPHs). The clay mineral captures the diesel and refuses to let go of it, causing spike recovery to be low. Does this mean the analysis failed? Or are things okay because the TPH recovery from the QC sample prepared by spiking diesel fuel onto a sand matrix was reported? Are methods developed to extract and analyze samples as they are, or are methods developed using clean, interference-free matrices ensuring good recovery resulting in analysis of things the way they are not?
In today’s world, laboratories are challenged to run thousands of samples as fast as they can and the results are used to determine whether the sample is okay to drink or complies with a permit. Rapid methods are needed that are capable of maintaining these very high sample loads. Methods need to be standardized so that no matter who does the analysis or where the testing takes place, the results are similar. That said, care must be taken to ensure that chemical test methods are capable of analyzing samples the way they are, and are not developed using samples as they are not.
William Lipps is environmental/chemical business unit manager, Shimadzu Scientific Instruments, Inc., 7102 Riverwood Dr., Columbia, Md. 21046, U.S.A.; tel.: 410-381-1227, ext. 1802; e-mail: [email protected]; www.shimadzu.com