Independent Consultants in Environmental and Forensic Chemistry
Volume 4, Issue 2, Spring 2001
Are Your Corrected Data Correct?
Methods used for the analyses of environmental samples often require the use of internal and surrogate standards. In USEPA SW-846 method 8082A for determining polychlorinated biphenyls (PCBs) in solid and aqueous matrices, internal and surrogate standards are defined. An internal standard is A... added to each sample extract prior to analysis ...@ A surrogate standard is A... added to each sample prior to extraction.@
A surrogate standard is a compound that has properties similar to the target analyte(s) that a particular analytical method is designed to identify and measure. The surrogate compound is not expected to be in an environmental field sample and should not interfere with the identification or quantification of the target analytes. By demonstrating that the surrogate compound can be recovered from the sample matrix with reasonable efficiency, the surrogate standard performs a quality control function on the suitability of the analytical method for the intended analyses and on the ability of the laboratory to execute that method with reasonable proficiency. If a surrogate compound is not recovered, an analyte of concern also may not be recovered.
The purpose of an internal standard is to provide a reference concentration against which the responses of the target analytes are compared. By adding the internal standard to the sample just prior to instrumental analysis, the quantity present is not affected by extraction efficiencies or other sample handling procedures. The internal standard can then compensate for relatively minor fluctuations in instrument sensitivity to provide more accurate quantification of the target analytes.
Recent trends in the analyses of environmental samples are contrary
to the basic concepts and purposes of the internal and surrogate standards. One such trend is to correct target analyte concentrations based on the recovery of the surrogate standard. If a surrogate standard has a 50% recovery in a sample, multiplying the surrogate recovery by two would yield a 100% recovery for the surrogate standard. Similarly, the target analyte results are corrected by multiplying the result by two in order to represent a 100% recovery of the target analyte. However, unless data are available to demonstrate that the surrogate and target compounds experience equal recoveries in the analytical method, the practice of correcting for the surrogate recovery is without foundation and, consequently, invalid.
In EPA method 525.2, the internal standard is allowed to be added to the sample prior to extraction. The internal standard is no longer an internal standard; it is a surrogate standard. When sample results are calculated against this Ainternal standard,@ the analyst is, in effect, correcting for surrogate recovery.
If analytical chemical data are to be valid and usable, results cannot include corrections for surrogate recovery without substantiating evidence that the recoveries for the surrogate and target analytes are equivalent.
Can You Get Bad Results from a Good Method?
Various agencies publish methods to analyze for various compounds or species in samples. Depending on the agency, these methods undergo considerable testing prior to publication to ensure the ability of the procedures to accurately identify and measure the compound or species present in the samples. However, unless the method is performed correctly, poor results may be obtained.
Performing the method correctly involves not only conducting each step of the procedure exactly as published in the method but also understanding the chemistry involved in the process. When analytical methods are developed, the samples tested often contain only the analytes of concern. Potential interfering compounds are sometimes added to the samples to ensure that the compound of concern can be distinguished from interferents. However, environmental samples often contain more interferents or interferents in greater quantity than when tested during the method development process. If methods are applied as written to environmental samples, the suppression of interferents may not be adequate to produce accurate results.
In the analyses of metals, one of the sample preparation steps involves oxidation to convert the metals to a soluble state in which they can be analyzed. For many metals, nitric acid is used; for mercury, potassium permanganate is the oxidizing agent. A problem arises if other oxidizable materials are present in the sample. The other materials will use up the oxidizing agent, and the metals may not be totally converted to the soluble form. This creates two problems.
If metals are not totally converted to the soluble form, more metal may be present in the sample than will be measured. The resulting reading will be biased low. On the other hand, if interfering materials are still present in the sample after exhausting the oxidizing agent, the interferents may provide a response to the instrument that may be misconstrued as due to a metal and will provide a false positive result for the metal.
Analyses for mercury in sediments must be monitored carefully in the preparation stage. Many sediment samples contain a host of organic compounds. When samples are prepared for digestion, the potassium permanganate addition provides a characteristic purple or pink color to the solution. If, after digestion is complete, the purple or pink color persists, this is an indication that sufficient permanganate was present to oxidize any mercury and/or organic material present. If, however, the sample contains no color, this is an indication that all of the permanganate was exhausted, and the oxidation process may have been incomplete. All of the mercury may not be in a soluble state, and the measured mercury result may be biased low. If organic material is still present in the sample, the material may provide a response that may be interpreted as a false positive for mercury. The appropriate action on observing no color in a digested mercury sample is to add more permanganate and digest the sample again. However, unless the laboratory technician is aware of this situation, the sample may be forwarded to the next stage of processing to eventually produce an inaccurate result.
The color due to permanganate is easily detected in the mercury preparation step. The use of colorless acids for the oxidation of other metals does not provide for a similar easily discernible indication that the sample has not been digested completely. For this reason, extreme care must be exercised when analyzing samples containing, or potentially containing, high concentrations of organic materials. Otherwise, inaccurate results may be obtained even if the analytical method was performed as published.
What Happened to those Acid Compounds?
Many environmental investigations include the analyses for a class of chemicals referred to as semivolatile organic compounds (SVOCs) which include materials that are acidic, basic, and neutral (neither acidic nor basic) in character. In order to monitor the ability to detect these compounds in a sample, surrogates are spiked into samples, and laboratory blanks spiked with the SVOCs are analyzed. Sometimes, review of a semivolatiles data package reveals that few or no acid compounds were detected, and the recoveries for the spiked acid compounds were also either very low or were not detected. What happened to those compounds? Many times the laboratory will dismiss the poor recoveries with a simple statement that the poor recoveries were due to a Amatrix effect.@ If the spiked acids were not detected, should you believe the results from your samples?
Many factors may play a role in reduced recoveries of acid surrogates and compounds. In some cases, the perceived loss of acid compounds is simply that the acids were never spiked into the sample. Sometimes, improper adjustment of the pH to less than two will prevent complete extraction of acids from the sample, particularly if the sample is initially highly alkaline. If temperatures greater than those specified in the associated preparation methods are applied to speed extraction and/or concentration, compounds may escape the cooling condensers, and acid compound recoveries may decrease. Taking the extract to less than 1 mL may also result in loss of semivolatile analytes.
However, a more insidious cause of poor acid recoveries exists. Groundwaters taken from anoxic conditions may contain soluble manganese. When a sample containing manganese is made alkaline to extract the basic compounds, the manganese will be oxidized to manganese dioxide. Manganese dioxide has been shown to oxidize phenolic compounds, particularly the acid surrogates 2-fluorophenol, phenol-d5, and 2,4,6-tribromophenol. This problem can be circumvented by extracting the acid compounds before the basic compounds. However, EPA methods 625 and 1625 require the extraction of the basic compounds before the acid compounds. EPA method 604 for phenols allows treating the sample with base to remove interferences but cautions that problems could occur with recoveries of phenol and 2,4-dimethylphenol. If these methods are followed as written, loss of acid compounds may occur.
The loss of surrogate and spiked compounds can be caused by several reasons. Further examination of the samples and data must be performed in order to determine if the loss was caused by matrix effect or laboratory error. Regardless of the cause, the inability to recover compounds known to be present in a sample indicates that that method or the performance of that method was not adequate to accomplish the desired analyses.
How do you tell if your laboratory cheated?
Sometimes you cannot, especially if you asked for just the answers! Presuming you requested and received a full data package from the laboratory, it depends on how well you review the data and on your expertise in reviewing raw data.
Why should a laboratory cheat in the first place? Many times a laboratory is in the position of having too many samples to analyze in too short a time. This situation is brought on by the inability to control the arrival of samples at the laboratory coupled with short holding times or quick turnaround requirements. When problems occur with instruments, production schedules may not be met.
One of the more frequent problems associated with instruments is the inability to establish or maintain proper calibration. These types of problems provide opportunities for creative manipulations of data. If appropriate responses for calibration standards cannot be obtained, the easiest manipulation to be made is outright fabrication of desired responses. This practice can be detected by simply matching the raw instrument output to the values used for establishing the calibration.
A more surreptitious practice concerns manipulation of data in which instrument responses are based on the area under a chromatographic peak. If that area yields a value that is higher than desired for the calibration curve, the operator can simply perform a manual integration of the peak to reduce the area included. In some cases, manual integration is warranted when a peak must be distinguished from other peaks or baseline noise. However, when the area of a peak is deliberately and unreasonably reduced to produce a desired value (Apeak shaving@) to meet a calibration, this type of manipulation is inappropriate. Conversely, when the area of a peak is increased (Apeak enhancement@) by inappropriately including portions of the baseline or another peak, this is also cheating.
Practices of peak shaving and/or peak enhancement cannot be detected unless original chromatograms are reviewed and manual integrations are indicated. Comparison of calibration data with the quantitation reports may not reveal that manual integrations were performed since, while instrument software will annotate the manual integrations with an AM@ in the quantitation report, this notation can be removed by the operator. The reviewer must then refer to the chromatogram to determine if the reported area is consistent with the integrated peak size.
The visual inspection of raw data that is sometimes necessary to determine if reported data are reliable must be conducted by a trained data reviewer.