Quality Control into Quantitative

of new exercises. The goodness of analytical data is judged on the basis ..... spreadsheets for data analysis, statistics, and calibration curves. As ...
5 downloads 6 Views 91KB Size
In the Classroom

Integration of Quality Assurance/Quality Control into Quantitative Analysis Suzanne C. Bell* and Jeff Moore Department of Chemistry, Eastern Washington University, 526 Fifth Street, Cheney, WA 99004

Quality assurance/quality control (QA/QC) has become paramount in industrial, government, and research laboratories. However, this vital aspect of chemistry is often peripherally discussed in chemistry curricula. The importance of teaching these ideas is now being addressed (1). Quantitative analysis (“quant”) offers the ideal venue to introduce QA/QC. In this course, students can learn how QA/QC and statistics are interpreted; how errors can be isolated; and how QA/QC is used to troubleshoot and repair faulty procedures from inception through reporting. Furthermore, existing laboratories can be modified to encompass QA/QC without implementation of new exercises. The goodness of analytical data is judged on the basis of accuracy and precision. Quant unknowns are in a sense unrealistic, since there is (usually) a true value associated with that unknown. However, real samples do not have a known value; the analyst must determine it, gauging reliability by QA/QC data. Although traditional unknowns are essential for evaluating student performance, it is equally essential that students learn to interpret results when there is no known standard. Numerous techniques can be employed. For example, students can estimate the accuracy of their analyses using known samples provided to them. Precision can be gauged through replicates. Coupling these to the analysis of blanks, spikes (fortified samples), and traditional unknowns helps students critique their own work while developing troubleshooting skills. Thus, QA/QC helps students develop invaluable analytical judgment. Skills that emerge from the study and application of QA/QC include problem identification and problem solving. In traditional courses, students turn in an unknown result with little, if any, sense of how they have done or whether problems were encountered. QA/QC allows students to identify problems, study and optimize their techniques and procedures, and apply realistic uncertainties. Coupled to an understanding of statistics and underlying chemical principles, QA/QC helps students to evaluate their work rather than having a professor or teaching assistant (TA) tell them what went wrong after it is too late to do anything about it. Such “postmortem” feedback leads to unnecessary discouragement and does not reflect analytical chemistry as practiced in most laboratories outside academia.

limit, it is considered unacceptable (“out of control”) and corrective measures must be taken. Although the inclusion of a control chart exercise in quant is laudable, it fosters the idea that QA/QC is a separate topic rather than a continuing theme. Thus, the idea of a control chart can and should be expanded. In quant for example, students often perform iodometric titrations using thiosulfate, a reagent subject to chemical and microbial degradation (4, 5). Control charting can be coupled with iodometric titrations as follows. Students at Eastern Washington University (EWU) prepare thiosulfate solutions without boiling the reagent water and without pH control. Additionally, the solutions are stored in sunlight. The solutions are initially standardized using 3–5 replicates. These data are used to generate the warning and control limits, here defined as ±1 and 2 standard deviation units, respectively. Although more initial replicates would be desirable, given time and reagent limitations, this represents a reasonable compromise. Warning and control limits are specified as ± 2 and 3 standard deviation units in Standard Methods for the Examination of Water and Wastewater (6 ). However, tighter limits are advantageous for several reasons. Since standardization occurs early in the quarter, student technique (particularly reproducibility) is still improving; large standard deviations (and thus large warning and control regions) often result. Under the quarter system, the time that

New Laboratory Exercises A survey of the literature indicates that a preferred method for integration of QA/QC into courses is through control charts (2, 3). In such exercises, a simple analysis is repeated throughout the term and the results are plotted. A range of acceptable values is set, along with warning limits and control limits that are also shown on the chart. Typically, determination of the warning and control limits is based on standard deviation. When a measured value exceeds a control *Corresponding author.

874

Figure 1: Representative control charts. “No Precautions” indicates that thiosulfate was made and stored without preservative steps. The four lines are the upper and lower warning and control limits; the x axis is molarity of the thiosulfate.

Journal of Chemical Education • Vol. 75 No. 7 July 1998 • JChemEd.chem.wisc.edu

In the Classroom

elapses from solution preparation to the final titration can be as little as eight weeks. Thus, by narrowing the control and warning limits, there is a greater chance these limits will be crossed and students can discuss the implications. After initial standardization, students titrate the thiosulfate weekly using KI/KIO3 and plot the data. Representative control charts are shown in Figure 1. Note that the control and warning limits for the upper chart, “All Precautions Taken”, are larger than those for the lower chart. This is due to the larger standard deviation of the replicates used to generate the former and reflects differences typically seen among students. Interestingly, after running this laboratory for several quarters, many trends have been observed in addition to the steady decline in concentration. These varying responses could be due to improvement in titration technique, evaporative loss from the volumetric flask, some combination of these factors, or others as yet undetermined. Over a semester (several weeks longer than a quarter), clearer trends would be expected. Existing Laboratories Laboratories constituting the backbone of traditional quant courses (gravimetric, volumetric and electrochemical) need not be discarded to include QA/QC. Consider the determination of the percent copper in a brass sample using a thiosulfate/iodine titration: 1. A method blank is included in which all preparatory steps are performed in the absence of brass. 2. A known sample is provided; students perform a single titration and calculate their results. The percent error between the experimental and true value is calculated and students decide whether further practice is needed before unknowns are attempted. Within the constraints of time and reagent supply, students are allowed to repeat the known titration as many times as they wish.

Addition of these two QA/QC samples helps students detect systematic errors such as impure reagents, errors in calculations, and contaminated glassware. Thus, they can proceed to the unknown with confidence in their technique and reagents. 3. Replicates of the unknown (n = 3) are added. This affords students a measure of precision and is used to calculate a percent relative standard deviation (%RSD, the standard deviation divided by the mean and multiplied by 100) and a confidence interval (95%) for their result.

Target values for percent errors on knowns and for %RSD are based on the technique and performance of past classes. Students are graded on accuracy (gauged by their known and unknown) and precision (gauged by %RSD). Using all QA/QC data (blank, known, and replicates), students identify systematic errors, postulate sources, offer solutions, and predict the impact on their unknown results. For example, assume a student performed the titrations described above. The blank showed residual copper or other reactive components, the known showed a positive error, and the replicates generated a %RSD of 1%. The student would be expected to predict good precision but a high bias on the unknown results due to a systematic error traceable to contamination. This is a classic example of precise but inaccurate results, easily associated with target analogies used in many textbooks (4, 5).

Use of Spikes and Fortified Samples Blanks, replicates, and knowns suffer from the limitation that they are separate from the actual unknown sample. Thus, analytical problems arising from the matrix (such as masking, adsorption) escape detection. This problem has been addressed using internal standard calibration, standard addition, and fortification (spiking) of samples. Two types of spikes are commonly used, matrix spikes and surrogate spikes. Consider the analysis of mouthwash for fluorine using an ion-selective electrode. Students calibrate the electrode using an aqueous solution of NaF, and both knowns and unknowns using this matrix typically work well, assuming ionic strengths are kept constant. However, it is not unusual for the results of a mouthwash analysis to differ significantly from the label value. Students investigate potential sources of the error using a matrix spike sample. Students take a sample of mouthwash that does not contain fluoride and add a known amount of F ᎑ to the matrix such that the concentration falls in the middle of the calibrated range. The percent recovery of spiked fluorine provides the students with the data required to distinguish a matrix problem from problems related to uncertainty in the label value and calculations. Coupled with other QA/QC results (blanks, knowns, and replicates), the students can discuss the reliability of their analysis with confidence. Surrogate spikes are another type of QA/QC geared toward identification of matrix effects or other problems associated with a specific sample. Surrogates are elements or compounds that are chemically related to the target analyte but are not likely to be found in the sample. By determining the percent recovery of the surrogate, the analyst can estimate the reliability of the result obtained for the target analyte. Use of surrogates requires multielement or multicomponent capability such as afforded by gas chromatography (GC). In addition, organic analyses are well suited for surrogate spikes since compounds similar to the analyte are readily available and by definition, amenable to the same chromatographic conditions as the target analyte. Fluorinated or deuterated compounds are a good choice for surrogates when using GC and GC coupled to mass spectrometry (GC–MS) because these compounds are rarely found in natural matrices such as soil and water. At EWU, students analyze swimming pool water for chloroform that forms as a by-product of chlorination. Purgeand-trap (PT) preconcentration is used, coupled to GC–MS, and an Environmental Protection Agency (EPA) method format is followed. The instruments and experimental conditions are summarized in the box. Students are responsible for sample collection based on EPA protocols. They collect two sets of samples; the first with no headspace in the vial, the second allowing large amounts of headspace. Prior to the laboratory, the instructor performs the MS tuning and initial calibration. When students arrive, they are responsible for the tune check, calibration check, and loading of all samples up to the autosampler capacity of 16. With group sizes typically 3–4, each student has the chance to load multiple samples. This loading procedure includes addition of surrogate and internal standards to each sample using microsyringes. Because pool water rarely imparts a matrix effect, students include a synthetic sea water sample in their batch to simulate

JChemEd.chem.wisc.edu • Vol. 75 No. 7 July 1998 • Journal of Chemical Education

875

In the Classroom Instruments and Experimental Conditions Instrument

Hewlett-Packard 6890 GC–MSD

Column and temperature program

HP-624, 25.0 m × 200 µm × 1.12 µm; 35 °C for 2 min, then 6°/min to 180 °C

Purge-and-Trap

Hewlett-Packard Purge and Trap concentrator with 16-position heated autosampler unit

Trap and conditions Software

Vocarb 3000; 11-min purge time, 40 mL/min, desorb for 2 min at 225 °C HP EnviroQuant, Rev. C

Sample size

5.0 mL

Standards

internal: α,α,α-trifluorotoluene, 50.0 ppb surrogate: 4-bromofluorobenzene, 50.0 ppb

potential matrix interferences, which should be reflected in poor recovery of the surrogate standard. After samples are loaded, students complete the sequence table in the GC–MS software that controls operations including analysis, quantitation, and reporting. The sequence is started and samples are run overnight, yielding several pages of quantitative data, QA/QC data, and calibration information. Since the pool samples are not prepared by the instructor or TA, there is no way to know a “true” value and no way to calculate a percent error based on student results. Therefore, students rely on QA/QC to evaluate their data. Typical student results (five student groups) are presented in Table 1. Note that pool water samples were collected over several days; the concentration of chloroform was not constant. Variation among student groups is clearly evident, and in some cases, the concentration of chloroform in the vial with headspace was found to be higher than in vials without headspace. Surrogate recoveries were also low. Variations were attributed to the students’ inexperience with microsyringes. The poor reproducibility likely resulted from different students loading the individual samples. The goal here was not so much perfection of the technique as exposure to EPA-type analyses and the accompanying blizzard of information (analytical and other) that results. While the volume of data can be daunting, it is an accurate reflection of commercial analytical data reduction. Thus, this

876

laboratory provides an ideal capstone laboratory which integrates laboratory, instrument, QA/QC, and data analysis skills. Other laboratories such as those involving spectrophotometry and high pressure liquid chromatography can be modified into similar capstone exercises. Since the instrument has to be calibrated for one sample the same as for a dozen, such laboratories are time and cost-effective. Other Procedures QA/QC is not limited to the bench. Rather, it encompasses the analytical process from sample collection through data reporting. Appreciation for this “cradle-to-grave” philosophy of QA/QC can also be addressed in quant, again, with modest alteration of existing procedures. At EWU, students use spreadsheets for data analysis, statistics, and calibration curves. As part of QA/QC, they verify formulas used in the spreadsheets and document this in their laboratory notebook. An appreciation for calibration is covered in the discussion of calibration of analytical balances using certified weights. Calibration check standards obtained from commercial sources are used to verify calibrations and students are required to document all relevant information, including lot numbers and expiration dates. Thus, the concept of traceability is stressed in lectures and reinforced in the laboratory. Much QA/QC is either directly related to or included in the laboratory notebook. Traditional instruction for keeping this vital document has been relegated to a few pages in most texts. However, given the current litigious climate and recent scandals surrounding notebook documentation, it is crucial that students learn to keep a legally defensible notebook at the earliest opportunity. For many, this opportunity is quant. In the EWU course, provisions for notebooks are stringent. A partial list of requirements includes: • • •

All pages, front and back, must be signed and dated with a full legal signature Black or blue ink is required. Typed procedures (reduced copies) are pasted or taped in the notebook, with initials through at least two points separating the inserted paper and the notebook page. This way, if material is ever removed, the initials will be partially removed as well and the removal will be immediately obvious.

Journal of Chemical Education • Vol. 75 No. 7 July 1998 • JChemEd.chem.wisc.edu

In the Classroom • • • •

• • •

Sources of all data, information, and reagents must be completely documented. Copies of all spreadsheets, graphs, and final reports must be included. Blank pages must be either marked through or explained; e.g, “The next ten pages are left blank for spacing.” Any corrections made must be indicated by a single line through the discarded sections. Furthermore, the student must initial any mark-throughs and briefly explain why the data was discarded. Serial numbers plus complete descriptions (make, model, etc.) of all instrumentation are required. Any computer documents included must be clearly identified by name, date, and file name. The “ten-year rule” applies: The information in the notebook must be complete enough that the entire experiment could be duplicated without any ambiguity ten years hence.

Initially, students struggle with the format; after that the rigorous procedures become habit. The best student notebooks are used as examples for later classes. More importantly, these notebooks can demonstrate to potential employers that the student is capable of generating defensible documentation. Like an understanding of QA/QC, this is a valuable and transportable skill that will serve students long after they have left the Chemistry Department. Time Management and Other Considerations The above procedures, with some variations, have been in use at EWU since late 1994. Several students have received jobs or internships in which they reported that the course format was especially helpful. These include positions with local environmental laboratories, forensic laboratories, and pharmaceutical companies. In comments on the course students typically report that although it was more work than expected, they appreciated the attention to technical writing, details of laboratory technique, and especially the skills obtained with spreadsheets and word processors. However, redesign of the course required careful planning and a continually evolving, flexible approach. One of the limitations with integration of QA/QC is the added time and effort involved. This is a tradeoff made in commercial laboratories as well, seeking a balance between reasonable and appropriate QA/QC versus the time and effort required to generate it. However, a balance is found and QA/ QC has become ubiquitous, the purpose being to insure the goodness of analytical results. Thus, inclusion of QA/QC in quant, a course geared toward teaching measurement science, should not be a question of “if ” but rather “how”. In turn, “how” requires a thoughtful assessment of resources, class size, and curricular concerns. Typically, students require more hands-on guidance with this approach in the earlier weeks of the term. In turn, this requires additional laboratory management and supervision by the instructors and experienced TAs to insure success. More reagents are needed and larger quantities of waste are generated. All these factors need to be weighed, but the main issue is the additional time needed to complete labs when QA/QC samples are added. The amount of additional time required varies depending on the procedure. In general, laboratories utilizing

a calibration curve are least impacted. Consider again the analysis of mouthwash for fluoride using an ion selective electrode. The bulk of the laboratory time is devoted to preparation of calibration standards. Adding a blank, known, matrix spike, and replicate unknowns still allows for completion of the exercise, including preparation of a calibration curve and analysis of a calibration check solution, in a single 3-hour laboratory period. On the other hand, the determination of copper in brass by a redox titration as mentioned above typically takes 5–6 laboratory hours from initial preparations through reagent standardization, analysis of blanks, knowns, and 3 replicates of the unknowns. This compares with an estimated 3–4 hours when QA/QC samples are not included. To offset additional time requirements, several approaches are taken, not the least of which is teaching students to work efficiently. The framework of the course is such that students are essentially free to organize their time. Reagents are made well in advance, and instrument access is scheduled early in the quarter. Moreover, students are not allowed to work outside the scheduled laboratory hours, even to wash glassware. Given the large number of laboratories required (11–14) and the inflexible and regular due dates (1–2 reports per week), students quickly become adept at allocating their time. Interestingly, this includes arriving prepared for the work ahead rather than reading a laboratory for the first time the same day they plan to do it. With time rationed, they learn how wasteful this practice is. In addition, students quickly realize that good results on the known predict good accuracy on the unknown, which in turn represents a large proportion of their grade. It is not unusual to find students doing additional replicates of the knowns, testing the purity of reagents, or doing additional replicates of the unknown rather than leaving 15 minutes early. Thus, the design of the course is such that students are motivated to learn time management and to exploit the value of QA/QC within the context of inflexible deadlines and limited samples and reagents. In the workplace, no less will be expected of them. Acknowledgments Suzanne Bell is grateful to the National Science Foundation (ILI grant 95-51724) and to several sections of quant students who patiently and with enthusiasm performed and critiqued the cited experiments. Jeff Moore, who did his portion of this work while an undergraduate at EWU, is now employed with Bayer Corporation, Pharmaceutical Division, Analytical Quality Assurance Laboratory, Spokane, WA. Literature Cited 1. Perone, S. P.; Englert, P.; Pesek, J.; Stone, C. J. Chem. Educ. 1993, 70, 847. 2. Marcos, J.; Rios, A.; Valcarcel, M.; J. Chem. Educ. 1995, 72, 947. 3. Laquer, F. C.; J. Chem. Educ. 1990, 67, 900. 4. Skoog, D. A.; West, D. M.; Holler, F. J.; Analytical Chemistry: An Introduction, 6th ed.; Saunders: Philadelphia, 1994. 5. Harris, D. Quantitative Chemical Analysis, 4th ed.; Freeman: New York, 1995. 6. Standard Methods for the Examination of Water and Wastewater, 19th ed.; Eaton, A. D.; Clesceri, L. S.; Greenberg, A. E., Eds.; EPA Group: Hanover, MD, 1995; Section 1-6.

JChemEd.chem.wisc.edu • Vol. 75 No. 7 July 1998 • Journal of Chemical Education

877