Quality Assurance and Quality Control in a Soil Test Laboratory David H. Hardy, Rao Mylavarapu, and Tony Provin Introduction Soil testing laboratories provide soil fertility recommendations for farmers and consultants to make management decisions in crop production. Labs are also used relevant to environmental issues such as phosphorus loss and heavy metal loading as governed by state and federal regulations. Although end uses of soil test data may be extremely different, the primary goal of the lab is to provide reliable, consistent, and valid data. The need for accurate results imposes great demand on labs in terms of quality assurance and quality control. Quality assurance (QA) in a laboratory is governed at the management level to ensure that laboratory data provided to the customer meet a defined level of confidence. This implies that data are generated under statistical control by trained staff in the laboratory where activities are driven by standard operating procedures (SOPs), good laboratory practices (GLPs), frequent inspection, and thorough documentation to ensure quality control. Quality control (QC) occurs in the laboratory. The quality assessment of data is essentially a monitoring process of internal or external reference standards meeting defined statistical limits to ensure accuracy and precision. Here, we provide helpful, applied information based on practices found to be important in the operation of public soil testing labs analyzing routine grower samples for agronomic recommendations. We begin this chapter discussing detection limits of chemical measurements to provide a general understanding of this topic as pertaining to instruments and methods. Afterward, quality assessment and quality control are discussed in detail. Limits Associated with Chemical Measurements Laboratories generate thousands of data daily with quantitative instruments but are obligated to report values only when certainty exists that a given value is real or measurable with confidence. This is especially true for environmental labs involved in ultra-trace measurements. For soil testing labs performing agronomic tests, there is more leniency in reporting in general but an understanding of chemical measurements as related to “detection limits” is necessary and beneficial. Three limits associated with chemical measurement are— limit of detection (LOD), limit of quantitation (LOQ), and limit of linearity (LOL). These may be defined as follows: LOD: the concentration of a given analyte where there is certainty of a non-zero measurement. The term method detection limit is also used to describe LOD. LOQ: a concentration above the LOD where quantitative measurement is meaningful for a given analyte and the method. The term method reporting limit (MRL) is sometimes used to describe LOQ. LOL: the highest concentration or upper limit, of the linear calibration range, both for the analyte and the method. We can further think about detection limits as related to instruments as well as methods. Instrument manufacturers provide specified detection limits for a given analyte. This instrumental LOD is typically lower than what can be attained in a soil testing lab where a given analyte such as phosphorus is present in the sample matrix with other constituents after extraction. For extraction methods of elements from soils, we refer to the LOD as being the method detection limit (MDL). The MDL is approached in soil testing by analyzing reagent or extractant blanks. A measureable amount of an analyte may exist simply by performing the method itself due to small concentrations of elements being present in reagents, extraction vessels and so forth. Blanks are probably most important in measurements of low concentrations and in micronutrient analysis. In routine soil fertility testing, methods should be evaluated with blanks to determine if significant contributions occur. Occasionally, preparation blanks must be used due to the inability of the laboratory to exactly duplicate the analyzed solution matrix. This may be due to the laboratory accepting solutions for analyses and the client failed to provide adequate matrix solution for the development of standards. Although it is customary to not report zero values, some soil testing labs when testing soils for agronomic recommendations may do so since reporting a non-value such as below MDL may be confusing to growers. Given the variability in soil fertility across an agricultural field and methods used in sample collection relative value and subsequent recommendation are most important. For environmental testing, values below the MDL should be qualified and reported as such. More common in agronomic laboratories to establish a MRL that is sufficiently low enough to insure all agronomic fertilizer or remediation requirements/recommendations can be utilized. In this instance, a less than symbol is placed in front of the MRL. With this understanding, we now discuss assessing data for accuracy and precision. Quality Assessment of Accuracy and Precision through Statistical Control of Data 2 Accuracy and precision are two words used when qualifying data. Accuracy is defined as the closeness of the analytical value to the “true quantitative value” for a given analyte. If one thinks of a target, great accuracy is hitting the bull’s eye. A lab should define its confidence in reporting a value. Precision is a measure of the consistency of reproducing a value; it is a measure of scattering of data around the mean. Tightly clustered data indicate great precision but precision does not suggest accuracy. When data are tightly clustered but inaccurate, random error is small but there is a strong indication of bias or occurrence of systematic error. “Control” or “check” samples, also referred to as internal reference samples are commonly used in labs to assess accuracy and precision. Check samples are often used by labs that analyze a large number of samples to circumvent the very high cost of the purchase of prepared reference samples. The following guidelines should enable someone to prepare a “check” sample for quality assessment. Select soils similar in soil chemistry, texture and organic matter to those analyzed for clients o Gather several different check samples to encompass a range of results Dry, grind and screen to 2mm size just as preparing clients’ samples Thoroughly mix the sample to acquire homogeneity. o An affordable, effective method of mixing is through placement of the ground, screened soil on a clean tarp; the sample is split and mixed by pulling alternate corners of the tarp toward the center. This effectively divides the sample in half upon each pull. o Another method of mixing is through use of a cement mixer. Subsample the thoroughly mixed sample 30 times to acquire subsamples for soil analysis. Samples should be run in slugs of 10 times on three different days. After collection of 30 subsamples from the prepared check soil as above, each are tested for analytes using SOPs of the lab. A time of at least 4 hours should separate each analysis; some labs analyze samples on different days. o Labs preparing check soils sometimes exchange samples with other labs using same methods in an effort to attain the best estimate of a true value. Variability in measurements will come from inherent variability within the samples, each step in sample preparation related to equipment (scoops, pumps) or technician technique, and instrumentation. If a lab has multiple instruments, check samples should be analyzed on each instrument to compare results. Bias can be investigated by repeatedly analyzing the samples using different chemists on different instruments and on different days to further establish baseline data— the “true” values for the check samples. Statistical analysis of the data from the analysis, the mean (X) and standard deviation (s), is used to establish the required accuracy to accept data. For example, data for phosphorus found in Table 1 are used to define the following limits. 3 Upper Warning Limit (UWL): X + 2s Lower Warning Limits (LWL): X - 2s Upper Control Limit (UCL): X + 3s Lower Control Limits (LCL): X - 3s Assuming a normal distribution of data about the mean, these limits include 95% and 99% of the population for warning and control, respectively. Table 1. Phosphorus data from a bulk check soil sample used to establish accuracy.a Sample P Sample P Sample P ppm ppm ppm 1 8.63 2 8.13 3 7.48 6 7.92 7 7.55 8 7.42 11 7.86 12 7.56 13 8.15 15 8.18 17 7.64 18 7.93 Mean = 7.83; s = 0.31, n = 20 LWL = 7.21, UWL = 8.44, LCL = 6.91, UCL = 8.75 a Twenty subsamples extracted and analyzed by ICP. Sample 4 9 14 19 P ppm 7.92 7.55 7.96 7.77 Sample 5 10 15 20 P ppm 7.75 7.41 7.71 8.03 The analytical limits are typically plotted to create X-Quality control charts as seen in Figure 1. Figure1. Phosphorus data from Table 1 plotted as an X-Quality control chart. 10 9.5 UCL P mg dm -3 9 8.5 UWL 8 7.5 LWL 7 LCL 6.5 6 0 5 10 15 Observation Number 4 20 25 During daily lab analyses if data for a “check” sample falls outside of the established warning or control limits for an analyte, it warrants investigation. Lab supervisors and chemists should determine the cause and the necessary corrective action. Consider if All analytes are involved or if only one analyte is involved New standards are recently involved The problem is isolated to one instrument or technician A recent trend in data is occurring The “check sample” is becoming segregated relative to particle size. One element, especially a micronutrient, may sometimes fall just outside of limits without concern. If the check sample is segregated, new control limits may need to be established. Whatever the case may be, it is imperative that a thorough discussion and investigation among supervisors and chemists occurs to ascertain if a problem exists. The frequency of check samples is a laboratory management decision. There is a monetary cost associated with check samples, both in time and analytical overhead. In most soil labs, a check sample is analyzed for every 20 to 30 grower samples. The checks are inserted and analyzed as part of the daily work with grower samples. This allows for measurement of systematic error, especially if the checks are “blind” so the analyst cannot differentiate the check from grower samples. Blind check samples are possibly preferred but check samples that are not blind can also serve a valuable role too. They can identify a problem during the actual analysis, before numerous samples are potentially analyzed in error. Even if the check samples are not true blind samples, the analyses data, means and standard deviation criteria should not be shared with laboratory technicians. Limiting these data to the laboratory manager, director and QA officer will better insure check sample integrity. Although check samples are most common and helpful to monitor accuracy on a daily basis, it is recommended that labs participate in proficiency testing programs that are external to the lab. These programs involve analysis of various soil samples on a systematic basis (usually quarterly) throughout the year. Typically, multiple labs using various methods are involved and data from labs are statistically analyzed to estimate a “true” value for analytes. This approach is excellent as long as the population of participating labs for a given method is not too small. Proficiency programs are affordable and voluntarily utilized by many private and public labs across North America as well as internationally. Two examples of proficiency soil testing programs available in the United States are the North American Proficiency Testing (NAPT) program administered by the Soil Science Society of America and the Agricultural Laboratory Proficiency Testing (ALP) program. These proficiency testing programs are not certification programs which do exist but are much more costly. The samples used in these programs have been finely ground to reduce heterogeneity and allow the laboratory to pinpoint mechanistic and analytical difficulties within the laboratory. 5 Accuracy can also be evaluated through use of standard reference materials (SRMs) that offer certified values of analytes. However, SRMs are expensive and with few exceptions fail to provide mean data from commonly used soil testing methods. Check samples can also be used to evaluate precision or repeatability in acquiring the same data over time. Precision can also be estimated by analyzing duplicates of grower samples or other unknowns. As an example in monitoring precision, we use K data presented in Table 2 from 11 grower samples analyzed twice (duplicates— Xanalysis1, Xanalysis2) by the lab. The absolute relative difference percent (RD%) is calculated where X = K data as follows. RD% = 100 | [(Xanalysis1 - Xanalysis2) ÷ (Xanalysis1 + Xanalysis2)÷ 2) ] | The mean and the standard deviation of the RD% values are calculated. The upper control limit is calculated as the mean %RD + 3s which is equal to 22.61 for our example. Data for K are under statistical control for precision if the calculated %RD fall between 0 and 21.98 Table 2.Potassium data acquired from duplicate analysis of grower soil samples. Sample 1 2 Analysis 1 Analysis 2 -3 ------------------K mg/dm ---------------8.6 8.1 9.8 9.7 %RD 6.0 1.0 3 5.5 4.9 11.5 4 21.7 19.5 10.7 5 17.4 16.8 3.5 6 59.8 63.2 5.5 7 5.9 5.7 3.4 8 7.9 7.5 5.2 9 4.1 3.4 18.7 10 3.5 3.4 2.9 11 16.1 15.0 7.1 Mean %RD = 6.87%, s = 5.04, Mean %RD + 3s = 21.98 The %RD values of the 11 samples can be plotted as a range chart (R chart) as presented in Figure 2 with the upper control limit. Although these data are from duplicate analyses of grower samples, the data for check samples presented in Table 1 could be used for similar purposes. 6 Data can also be evaluated for precision by calculating the absolute difference between duplicate measurements: | Xanalysis1 - Xanalysis2 |. The mean of the absolute average difference (R) is calculated and control limits are set for as follows: UCL = 3.267 R for a 99% confidence interval UWL = 2.512 R for a 95% confidence interval LWL & LCL = 0 A range chart (R) chart (not shown) can be used to plot the absolute differences and the associated limits. All data should fall under the UCL for it to be considered under statistical control. Also, data points should be randomly distributed within the warning limits when under a state of statistical control. Figure 2. Potassium data presented in Table 2 plotted as a range 25 UCL = 21.98% Relative Difference % 20 15 10 5 0 0 2 4 6 8 10 12 Observation Number Quality Control in the Laboratory In any laboratory, consistency in daily work regardless of a task is extremely important to attain precise data. Quality control is a method to minimize errors and foster reproducibility in all aspects of lab work. The following practices and guidelines are important in attaining good quality control. Standard Operation Procedures 7 The foundation of QC in daily work is by use of standard operating procedures (SOPs) for both technical and nontechnical work. The word “standard” implies an operation is performed the same way on each occasion. Any part of a procedure where personal preference or judgment of chemists could be imposed is to be included in the SOP. At a minimum, SOPs in soil testing are needed for sample receiving and drying, sample grinding, sample scooping, solution preparation, soil acidity measurements- pH and exchangeable acidity, soil extraction, element analysis by ICP, soil disposal, and dishwashing. An SOP needs to describe information in sufficient detail such that a competent person unfamiliar with the method or task could obtain acceptable results and / or conduct a reliable review of results using the procedure. The SOP for a given task needs to include: title page with reference method, lab name/location, authorization signature, and approval dates; table of contents; scope of application; summary of method; safety/waste handling; interferences; apparatus/equipment; reagents/chemicals; sample collection, preservation, shipment, and storage; calibration; sample preparation; procedure; calculations; quality control; data validation; and preventative maintenance. Guidance for preparing SOPs is found at http://www.epa.gov/quality/qs-docs/g6-final.pdf Instruments and Equipment— Operation, Maintenance, & Records Highly optimized, operational instrumentation or equipment is essential to producing consistent, accurate data. Users should refer to the operation manual for the manufacturer’s guidelines in correct operation, calibration, maintenance and trouble shooting. Many manufacturers offer on-site training at the purchase of an instrument as well as regional seminars; laboratories should take advantage of such training opportunities. Table 3 lists guidelines for daily operation of laboratory equipment in a highvolume lab setting for non-research samples. A more stringent calibration schedule may be desirable for labs analyzing small volumes of samples or research samples. Records on equipment are extremely important. A record of purchase, repair, maintenance and calibration should be archived for all equipment and instruments, including items such as automated pipettes. The operator’s manual can be used if space allows for non-daily information. An instrument log can be used for daily calibration. Each entry should be dated and initialed. Details are needed as to who performed a given activity. As equipment ages, repairs often become more frequent. Management needs to be informed of increasing downtime and potential need for replacement. Sensitivity of analysis too becomes important for some instruments such as ICPs. Chemicals, Standards, and Lab Solutions 8 Analytical methods should specify the quality or grade of reagents needed to achieve desired results. Instruments that rely on gases for their operation will specify the purity of gas needed in the operator’s manual; if not found, consult the manufacturer. Just as any instrument, a record of chemical purchase is needed. Chemical inventory records should be kept so lab supervisors are aware of supply on hand. Ordering should be adequate to prevent uninterrupted lab operation. An inventory of chemicals for the entire lab should occur yearly, preferably after the lab’s busiest season. Care should be taken in disposal of chemicals no longer needed due to changes in methods or expiration dates of the chemical. Upon receipt, each chemical container should be inspected as to its contents and physical soundness, dated and initialed only by designated chemists, preferably supervisors. An expiration date should be circled or written on the container. The container should be properly stored according to the material safety data sheet notebook (MSDS). A copy of the MSDS should be placed in the MSDS laboratory notebook with replacement of the out-dated copy if needed. At the time of opening a container for initial use, the date opened should be recorded and initialed on the container. Table 3.Guidelines for equipment calibration and maintenance. Equipment ICP Calibration After initial warm-up Check calibration every 30 samples Special Concerns Temperature Humidity A recalibration is necessary if periodic QC samples fail to be within control limits Profile 2-hr intervals Maintenance Schedule routine professional maintenance yearly or as suggested by manufacturer All unknown since run since last acceptable QC sample must reanalyzed Significant changes in torch, tubing, gas flow, or method Spectrophotometers After initial warm-up pH Meter & probes Check calibration every 30 samples Hourly or every 6090 samples, read back samples frequently 9 Temperature Check probes frequently for wear; make sure electrodes are filled. Conductivity Meter Balances pH 4.0 and 7.0, consider 10.0 for saline soils with high pH Check known samples for calibration frequently Temperature and fouling of probe surfaces Daily prior to use Check frequently for wear and tear of the probes and as per the manufacturers specifications Clean after each use Annual certification & cleaning by qualified company Macrodispensers Automated Pipettes Soil scoops Daily using balance and DI water or solution of known density Each work session and should be calibrated quarterly. Micropipettes should be calibrated quarterly. Annually Wear of scoop top Logbooks are needed for preparation of standards or solutions such as extractants. Entries in the logbook should include: Time / date of preparation Receptacle container for the solution being prepared Name, amount and lot of chemicals used order mixed in preparation Required measurements (for example, initial pH for buffered solutions) Associated soil samples analyzed Chemist’s initials Again, careful attention should be given to chemical expiration dates relative to the date on the container and also to the longevity of the use for the end solution in preparation. Lastly any chemical or reagent removed from its original container if unused should never be returned back to it. Chemist Notebooks and Laboratory Records 10 The purpose of laboratory records whether in a chemist notebook or a logbook is to provide a historical reproduction of an occurrence. Without documentation, a chemist cannot prove or possibly remember what was done. Records shall: Allow for historical reproduction Always be entered with blue or black ink, never pencil Be accurate, legible, objective, and recorded in a permanent manner Be recorded immediately Follow an established retention policy Protect and maintain integrity and security in data Record any modification in procedure Contain information on all calculations to the final result if needed Document levels of review Occasionally in documentation, errors will occur. Guidelines to follow when this occurs are: Draw a single line through the error such that original entry remains legible Write the correction adjacent to the error Never use white-out Initial and date change Offer a reason for the change if confusion is likely Use revisions to electronic files if involved to maintain history Labware and Glassware In many high volume labs, plastic has replaced glassware vessels in many day-today operations. Plastic is affordable, usually unbreakable, and can often be used without contamination if properly cleaned; in some cases disposable glassware offers more of an economical advantage than washing, especially for certain tasks. Some metal contaminates, primarily zinc, can be present in high volume disposable plastic manufactured products. Laboratories should purchase these supplies in sufficient quantities to reduce the number of manufacture production runs or lots. Prior to use of the new lot, multiple blank extractions should be conducted and compared to existing inventory. Most plastic vials used in daily extractions can be configured in a grouping of ten to twelve bottles that will allow for ease of handling as well as washing. Automated washing stations allow for grouped bottles to be thoroughly washed efficiently. Manifolds with nozzles that allow for bottles to rest upon with water controlled by a footpedal in such an apparatus has proved effective. Three levels of rinse are made with the first two being with tap water and the last being made with deionized, distilled water. Filtering devices can be washed similarly at a sink. Labware should be washed before it becomes dry. Many high volume labs use “solo” cups that are dispensable for pH. 11 When glassware is involved, scrub to remove any residue, soak if needed, wash with phosphate-free detergent and triple rinse. If standards are prepared in volumetric flasks, the flasks should be soaked in a 0.1 N HCl for a few hours, then rinsed thoroughly with deionized water. Volumetric pipettes can be washed using standard pipette washers and should be washed after each use. These should also be soaked in an dilute (not to exceed 1N HCl) acid bath prior to washing. All pipettes should be thoroughly dried prior to use. Store any labware not in use in a clean environment as dust-free as possible. When possible, closed cabinets or drawers are best. Any stored or new labware should be thoroughly washed prior to use. Daily Traceability of Samples and Laboratory Processes The importance of records in logbooks and lab notebooks has already been mentioned. The documentation of chain of events for the progression of various steps of receiving, handling and analysis that a sample undergoes is extremely beneficial and important to attain, although routine samples may not require a formal chain-of-custody. In the end, it is important to know what happened to each sample, who performed each step, where solutions were attained, what instruments were used, when each task was performed and so forth. This is of tremendous assistance if error is found in the analysis and also to attain accountability of technicians or chemists. Table 4 lists guidelines for sample traceability for various lab processes. Table 4.Guidelines in tracking lab processes. Lab Process Sample receiving Technician / Chemist Activity Record client name, shipment information Method of Record Logbook / bar code scanner / hardcopy submittal forms, LIMS system Date / time of arrival Method of shipment Condition of samples Location of storage if needed Lab ID assignment Special considerations if quarantine Organize working groups of samples Soil receiving logbook Designate QA/QC position Lab ID assigned to sample box/bag 12 Lab ID assigned / recorded on information sheet Grinding Scoop Extraction ICP analysis pH / BpH Date / time of process Grind soil samples Label vials with lab IDs for tests Scoop samples and controls Addition of extractant Filtering Assign samples to instrument Additions of water / buffer Read pH / BpH Grinding logbook Sample Prep logbook Extractant / solutions logbook ICP logbook or LIMS Meter / electrode logbook or LIMS Meter / electrode logbook or LIMS Training & Error Prevention Written, formal SOPs are excellent tools to train new employees and refresh experienced chemists. Again, training offered by instrument and chemical manufacturers is valuable too. The QAQC and/or laboratory manager should maintain a logbook documenting all trainings to insure existing and new staff are fully aware of all laboratory procedures and changes in methodology, instrumentation and similar differences which could impact overall laboratory performance. An important part of quality in laboratory work is a prepared worker, not only from a technical perspective but also with mental attitude. Prior knowledge of work volume expectations is extremely important. In labs as any work environment, emergencies and illness will occur and employees may be absent unexpectedly. Having chemists with a wide range of abilities in the event of employee illness can be very beneficial. Cross-training to allow for such flexibility and also to avoid monotony sometimes associated with lab work is recommended. No matter how diligent workers are, errors will occur. Anticipating errors through a thorough understanding of specific work is extremely important. Data should be reviewed by the technicians generating the data and further reviewed by supervisors/QA officers to prevent erroneous data from being reported to users. A conscientious chemist can prevent generation of poor quality data before data are submitted to further scrutiny, thus resulting in efficiency. Some known errors that may occur in soil testing labs are listed below. Details are given as to the lab location or work related, actual error, result if uncorrected; and action to minimize the potential for recurring error. Soil receiving; error— misalignment of samples / incorrect lab numbers as related to information on the client information sheet; result— client receives incorrect sample results and recommendations; action— worker carefully reviews work after assigning lab numbers to sample box/bag and information sheets. 13 Soil receiving; error— not placing a control sample in a grouping of samples as specified; result if uncorrected— loss of QA/QC for the group of samples involved and uncertainty as to data quality; action— worker carefully reviews work. Grinding; error— placing samples out of sequence as samples are ground; result— lab numbers and samples are now out of sequence so client will receive incorrect data and recommendations; action— working in verified order without deviation and review of lab number order before and after scooping. Grinding; error action— samples not sufficiently dry, so soil adheres to the grinder and contaminates samples; result— potential inaccurate results of samples; action— thoroughly checking samples for dryness, further dry if necessary, and clean grinder if needed. Sample preparation or soil scooping; error— using the wrong size scoop for an analysis; result— inaccurate data; action— paint handles of similar size scoops the same color and store scoops of different sizes apart from one another. Sample preparation or soil scooping; error— samples scooped out of order; result— client receives incorrect sample results and recommendations; check sample data fails; action— train workers to systematically scoop samples and to ascertain control samples are scooped in correct positions; control samples should be scooped as grower samples are scooped, not at one time. Soil Extraction; error— filtrate is poured out of sequence; result— client receives incorrect samples results and recommendations; action- label each extraction rack with a color code and beginning lab number with the same done for the extraction filtering apparatus. ICP analysis; error— data are 0 or near 0 values for many elements due to blocked tubing; result— client receives incorrect results and recommendations; action— review data prior to importing into LIMS and free blockage. References Taylor, J. K. 1987. Quality Assurance of Chemical Measurements. Lewis Publishers, Inc. Chelsea, MI. Watson, M.E., J. Kotuby-Amacher, P. Chu, C. Focht, W. Greig, M. Nathan, T. Provin, K. Reid, and D.A. Horneck. 2005. NAPT Quality Assurance / Quality Control Model Plan for Soil Testing Laboratories. Soil Sci. Soc. of Am. 14 Hoskins, B. Chapter 1- Laboratory quality assurance programs. IN Recommended Soil Testing Procedures for the Northeastern United States. 3rd. Ed. Northeastern Regional Publication No. 493. Reeuwijk, L. P. and V.J.G. Houba. 2001. Guidelines for quality management in soil and plant laboratories. Food and Agriculture Organization of the United Nations FAO Soils Bulletin 74. Daya Publishing House, Shastrinagar Delhi, India. 15