Quality Control Practical Applications for the Clinical Laboratory

Transcription

1 C OAGULATION 8 Dianne DeLucia Quality Control Practical Applications for the Clinical Laboratory

2 Dianne DeLucia, B.S. MT (ASCP) Product Manager Coagulation Systems Instrumentation Laboratory Quality Control Practical Applications for the Clinical Laboratory

3 Introduction The purpose of a well-defined quality control program is to adequately monitor performance characteristics, such as precision and accuracy, so as to alert the operator to sources of unsatisfactory analytical performance. Through careful analysis of quality control information, the decision is made to either accept or reject patient data. A comprehensive quality control program has both an internal and an external component. An internal program involves the routine analysis of stable control materials that are as similar as possible to the patient samples being run. By setting limits and rules which must be followed, imprecision and bias can be measured and monitored on a daily, monthly and long-term basis. In an external program, a laboratory analyzes unknown samples sent from a source outside of the laboratory. The outside source is either a voluntary proficiency testing organization or an agency appointed by the government to be responsible for laboratory accreditation and licensure. Participation in these types of programs allows a laboratory to measure its ability to obtain the correct result on an unknown specimen as well as to compare its performance with other laboratories using the same methodology. Another type of external quality control program is one in which control data from a laboratory s internal QC program is compared statistically to other labs who are using the same lot numbers of material. Quality control data is sent to a central processing location on a monthly basis. Cumulative statistics, such as mean, standard deviation and coefficient of variation, are reported as well as peer group comparison data. This allows the laboratory to assess their own in-house performance as well as note any biases they may have in comparison to other labs using the same methods and controls. Both internal and external programs are essential to a well-rounded quality control program. 3

4 Quality Control in practice Control Materials Materials used in QC testing should be as similar as possible to the samples being tested. Only matrix materials may be used when evaluating analytical performance as it relates to clinical interpretation. A laboratory should obtain enough of the same material to last a minimum of one year. The material should be stable over the period of use and exhibit minimal vial-to-vial variability. Non-matrix specimens may be included to supplement the matrix material. In an external QC program, for practical purposes, only matrix materials may be used. 1 If an assay is calibrated, the control material must be of a different batch, or lot, than the calibrator. 1 A minimum of two concentrations of each analyte being measured should be used and should be focused, when possible, at clinical value levels encountered in patient specimens. More than two levels of control may be used to monitor additional decision levels, if desired. 1 Frequency Placement of quality control samples Setting target values A minimum of two control samples of different concentrations must be analyzed at least once during each analytical run. 1 This should be no less than once per 8 hour shift or once every 40 samples, whichever comes first. 2 An analytical run, for the purposes of quality control, is a period of time, or series of measurements, within which the accuracy and precision of the measuring system is expected to be stable. 1 The manufacturer must define the length of a run during which accuracy and precision of both the instrument and reagents have been determined to be stable. The user may define a run length that is different from that recommended by the manufacturer as long as it does not exceed the manufacturer s recommendations. Important things to consider are workflow patterns, sample stability and reporting intervals of patient results. It is important to detect any variations that occur between runs. Therefore, control samples must be run during each analytical run. 1 Controls may be placed in fixed positions and analyzed at fixed intervals throughout a run or may be randomly placed along with patient samples. Placement at fixed intervals (ie. beginning and end of a run) allows for an assessment of drift throughout a run. Random placement gives a more valid assessment of the imprecision of patient data and is preferred. Target values must be determined by each individual laboratory. Control ranges provided by the manufacturer on the package insert are to be used as guidelines only. 1 Mean values should be set using a minimum of 20 values obtained from a minimum of 20 separate runs. In the low volume laboratory, less than 20 runs may be used, as long as no more than 3 data points are taken from each run. If a new lot of material is being analyzed, it must be run in parallel with the current lot. 1 4

5 Setting Control Limits Monitoring control limits Once a control s target mean has been set, limits of acceptability around that mean must be determined. If a control result is within established limits, the instrument is assumed to be functioning properly. If a control result is outside of these limits, the instrument must be assumed to be malfunctioning until the source of the problem is identified. Results of control samples must meet the laboratory s criteria for acceptability prior to reporting patient test results. 3 The traditional approach to setting control limits is the statistical method. It involves using the mean plus or minus a multiple of the standard deviation, for example +/- 2 SD. Another approach is to use medical usefulness limits.this means that the laboratory director sets control limits at a range that is consistent with the ability of clinicians to adequately treat patients. Ranges can be widened to 2.5 times the SD, for example. This is a pure judgment call. The Levey-Jennings chart is an extremely useful tool used to display quality control data points. It displays the mean and 2 SD lines above and below the mean. According to Gaussian statistics, the 2 SD lines indicate the area in which 95.5% of the points should fall. The values obtained from a single lot number of control should show a relatively even distribution of points above and below the mean. Once in approximately 20 tests, a value will fall outside of the 2 SD range (usually between 2 and 3 SD). This usually occurs by chance and doesn t necessarily indicate an out of control situation. Repeating the determination on that sample, or reconstituting a new vial of control usually results in the value falling within normal limits. +2 SD -2 SD 5

6 Shifts and trends can easily be monitored by L-J charts. Shifts are said to occur when six or more consecutive points fall either above or below the mean. They are usually caused by events such as a change in calibration or reference value. A trend is seen when six or more values are seen to steadily increase or decrease. The trend can start on one side of the mean and continue across to the other. This is usually caused by a gradual change in a reagent, control or instrument. 20 Shifts 10 Mean days 20 Trends 10 Mean days 6

7 Quality control rules Quality control rules can be designed to detect any bias or imprecision that has an impact on the quality of patient results. Although there is no requirement to use specific rules, many have been designed. Each rule detects a slightly different type of error. A combination of several rules, designed to detect both random and systematic error will greatly improve a laboratory s quality control program. Ideally, a rule will not indicate an error when there is none. This would lead to a high false rejection rate of runs due to random error. It should, however, indicate a high probability for error detection due to analytical errors. The simplest rule involves a target mean, obtained over time from multiple runs of control material, and limits around that mean. This has been discussed earlier in this monograph. Some of the other more commonly used rules follow. They are represented in the form A L, where A is the number of data points and L is a limit derived from Gaussian statistics. 1 2s A data point is outside a control s +/- 2 SD limits. Additional investigation should be performed. +2 SD -2 SD 1 3s A data point is outside a control s +/- 3SD limits. This is cause for rejection of a run. +3 SD -3 SD 7

8 2 2s Two consecutive data points exceed the same limit, the mean plus 2 SD s or the mean minus 2 SD s. This rule is applied either to two different control levels within the same run or to two consecutive results of the same material across two runs. In either case, it is cause for rejection of a run. +2 SD -2 SD R 4s The difference between two control values within the same run exceeds 4 SD s. This is cause for rejection of a run. +2 SD -2 SD 8

9 4 1s Four consecutive results are greater than the mean plus 1 SD or less than the mean minus 1SD. The results can be within one control material or across control materials. This is cause for rejection of a run. +2 SD -2 SD 10 x Ten consecutive results fall on one side of the mean. This can occur within one control level or across control levels. It is cause for rejection of a run. +2 SD -2 SD The 1 3s and R 4s rules generally detect random error while the 2 2s, 4 1s, 10 x rules and the 1 3s rule, when very large, detect systematic error. 4 Dr. James Westgard and his associates have developed an approach for using a combination of the previous rules as follows: If no control exceeds the 2 SD limit, the analytical run is considered to be in control and patient data may be reported. The 1 2s rule, in this case, is used as a warning. If this is exceeded, the control data should be tested using the 1 3s, 2 2s, R 4s, 4 1s and 10 x rules. If none of these rules are violated, the run is in control. If any one of them is violated, the run is not in control and patient data cannot be reported. Generally, if an error occurs on only one control material, the problem may be the control. If it occurs on more than one control, it is probably due to a system error, such as calibration, reagents, pipetting or optics. 9

10 Use of Quality Control data Analysis of data All patient test results obtained in an unacceptable test run or since the last acceptable test run must be re-evaluated to determine if patient test results have been adversely affected. Once a problem has been identified, the laboratory must take the action necessary to insure the reporting of accurate and reliable test results. 5 If a single control fails, it should be re-run, using a new vial. If the second result is in, patient results can be reported. If it is still out, the operator must consult the troubleshooting guide to determine the cause of the imprecision. In either case, both data points must be documented. Once the problem has been identified and corrected, the patient samples and controls should be re-run. If the second run is in control, the patient results can then be reported. It is good laboratory practice to evaluate overall quality control performance on a monthly basis. The mean and standard deviation should be recalculated each month using the previous month s data. (Data from unstable periods of operation should not be included, as this will cause the SD s to be too large and the control limits too wide.) If the data has not changed significantly, no action is required. If, however, there is a significant difference from the original data, the situation must be investigated and corrective action documented. Regardless of the system used to monitor control violation, detailed quality control guidelines are essential in every laboratory. In all cases, the rules must be applied before patient data is reported. Documentation The laboratory must document and maintain records of all quality control data for a minimum of two years. 6 10

11 References 1. Internal Quality Control Testing: Principles and Definitions, NCCLS Document C24-A, Vol.11 No Collection, Transport, and Processing of Blood Specimens for Coagulation Testing and Performance of Coagulation Assays - Second Edition, NCCLS Document H21-A2, Vol. 11 No Clinical Laboratory Improvement Amendments of 1988, [ (e)]. 4. Westgard, J.O.; Barry, P.L.; Hunt, M.R.; Torgny, G. A Multi-Rule Shewhart Chart for Quality Control in Clinical Chemistry. In: Clin. Chem. 27/3, (1981). 5. Clinical Laboratory Improvement Amendments of 1988, [ (b)]. 6. Clinical Laboratory Improvement Amendments of 1988, [ ]. 7. Portnoy, A.L., Ph.D. A Review of the Principles of Quality Control. 8. Westgard, J.O., Ph.D. Planning and Validating QC Procedures, College of American Pathologists Surveys Manual,

12 Part. No