Vol. 11 •Issue 2 • Page 43
Calibration, Control of Hematology Analyzers
Ensuring correct calibration of hematology instruments can prevent physicians from making the wrong clinical decisions.
Generally, action limits used in the clinical interpretation of test results depend on the reference ranges established for the test. If an instrument response changes, some of the resulting test values will not correspond to the clinician’s reference points and incorrect conclusions may be drawn. Specific action limits used for each clinical decision depend on the clinical problem, the circumstances for the specific patient and may also be related to the normal reference interval of the pertinent analyte. Specific action limits do not take into account the assumption that reference intervals may be biased.
It is often stated that the distribution of analyte values in a healthy population follows a Gaussian curve (i.e., a normal distribution). Similarly, it is stated that the reference interval, previously often called “normal range,” is represented by the center 95 percent of the distribution. Assay results that fall above the upper or below the lower reference limit are considered abnormal. However, these statements are not entirely correct. Analyte values in healthy populations distribute according to a lognormal distribution and the reference interval is represented by the 95 percentiles, not by the center 95 percent of the analyte distribution. For example, in an analyte distribution starting with a zero value, it would be incorrect to cut those values off. Many analyte value distributions in hematology, such as the hemoglobin concentration, appear to be Gaussian, but are not; many, such as the platelet count, do not appear Gaussian at all and are lognormal.
The Figure illustrates a distribution of hemoglobin values in a healthy population (solid line); the dotted and dashed lines model the effect of a 3 percent (3.75 g/L) and a 5 percent (6.23 g/L) negative calibration bias. A 3 percent bias at the lower reference limit of 105 g/L translates into an error of -3.2 g hemoglobin, a -5 percent bias into an error of -5.3 g. The number of subjects falling below the lower reference limit increases from 2.5 percent of the population to 3.8 percent, respectively 4.7 percent, resulting in an incremental increase in false-positive values (i.e., values incorrectly placed outside the reference limit) of 52 percent and 88 percent, respectively. This example clearly demonstrates that biased reference limits can easily lead to diagnostic errors. Thus, calibration and calibration verification are significant components of good laboratory practice.
Basic automated hematology analyzers provide a complete blood count (CBC). The CBC consists of a red blood cell count (RBC), a white blood cell count, the hemoglobin concentration (Hb), the mean (red) cell volume (MCV) and, calculated from these measured quantities, the (electronic) “hematocrit” (hct, calculated as MCV(fL) X RBC(#/L)), the mean (red) cell hemoglobin (MCH (pg) = Hb (g/L) / RBC (#/L)), and the mean (red) cell hemoglobin concentration (MCHC (g/L) = Hb (g/L) / hct (L/L)). Certain analyzers also include the platelet count and the mean platelet volume. A further class of automated analyzers includes measurement of a white cell differential count, either as 3-part differential (granulocytes, mononuclear cells, lymphocytes) or as 5-part differential (basophilic, eosinophilic and neutrophilic granulocytes, lymphocytes and monocytes), the red cell distribution width and the platelet distribution width.
Automated instruments to measure the various components of the routine CBC are not direct measurement devices, but comparators that require careful calibration and internal quality control. The number of iterative measurements that will result in a valid calibration value is dependent on the degree of precision that can be obtained for a specific calibrant or calibrator material, at a specific value for that material, on a specific analyzer. It is, therefore, not possible to describe a standard procedure that will lead to acceptable results for each circumstance. To obtain measurements that are comparable with those obtained by other systems and methods, adjustment of the instruments by means of calibrators with assigned values of defined accuracy is required. Calibrators may be produced from fresh human or animal blood, preserved blood, artificial materials or mixtures of some or all of these items. Several hematological techniques are available for assigning values directly to fresh blood calibrators.
Fresh blood calibrator values assignment 1
Three fresh blood specimens are obtained from apparently healthy subjects known to have an MCV of 86 fL – 96 fL and an MCHC of 330 g/L – 345 g/L (33.0 g/dL – 34.5 g/dL). The specimens are anticoagulated with K2EDTA or K3EDTA, 1.4 mg/mL – 1.6 mg/mL. (Note: K3EDTA causes a shrinkage of cells resulting in an approximate decrease of 2 percent of the packed (red) cell volume (PCV; » hct)). The specimens are stored at room temperature and must be tested within four hours. Pipettes used must be Class A, calibrated; volumetric flasks must be Class A, calibrated, made of borosilicate glass; special disposable capillary tubes for the determination of PCV must meet specifications of the American Society for Testing and Materials;2 the cell counting vials, plastic or glass, should have a minimum volume of 10 mL.
The hemoglobin concentration of the specimens (two dilutions) is determined using the International Committee for Standardization in Haematology (ICSH) haemiglobincyanide reference method.3,4 The PCV (hematocrit) is determined (four capillary tubes) using the ICSH selected method for microhematocrit determination.5,6
Note: the oxygenation state of the blood samples and the time elapsed between taking the sample and its analysis influence red cell volume—a fully deoxygenated cell is significantly larger than when fully oxygenated. Storage of samples also makes red cells increase in volume by taking up water.7
The red cell count is obtained using a semi-automated single channel counter that permits counting, by electronic means, of all the cells in a known displaced volume of the diluted blood sample. At present only aperture impedance particle counters meet this specification.8 A sterile, non-toxic, buffered saline solution is used as diluent. The diluent must contain < 5 x 104 particles/L in a size range of 20 fL Ð 120 fL, should neither crenate or lyse red blood cells, nor alter the MCV by > 2 fL over a 30-minute period. A primary dilution of 0.05 mL blood + 25 mL diluent is made, followed by a secondary dilution of 0.2 mL primary diluted sample + 20 mL diluent; this results in an overall 50 601-fold dilution. The diluted sample is transferred to the counting vial and counts performed within five minutes of completing the final dilution step. Two dilutions are made; each is counted twice.
The white cell count is obtained using an instrument as for red cell counting. The lytic agent used must be capable of completely lysing red cells, leaving no residual material capable of contributing to the count. The count signal from the leukocytes should fall into the size range equivalent to 45 fL – 450 fL and the count should be stable for 15 minutes after preparation. Two dilutions are made; each is counted at least twice.
The platelet count is obtained by the ICSH_ISLH RBC /platelet ratio method.9 Two dilutions are made.
Direct whole blood calibration of automated hematology analyzers is rarely, if ever, used in medical institutions. However, it is the cornerstone of the recommended method for assigning values to stabilized blood calibrators and, as such, is used by the manufacturers of these calibrators.
Stabilized blood calibrator values assignment 10
Calibrators prepared from stabilized human and/or animal blood enjoy wide use. These calibrators should have the following characteristics. Assigned values should be conferred by transfer of whole blood reference values. The assigned values should not change during the labeled shelf life, and the labeling should identify the types and models of analyzers for which the calibrator is suitable.
Values should be assigned to the stabilized blood calibrators as follows. CBC values are first assigned to fresh human blood specimens, the “reference specimens,” following the protocol described above (fresh blood calibrator value assignment). The assayed “reference specimens” are presented to one or more automated hematology analyzers (assignment analyzers) of the types/models for which the calibrator is intended. The values given by the assignment analyzer(s) to the reference specimens is compared to the reference values of these specimens and provides the calibration factor for the assignment analyzer(s). The candidate calibrator is assayed repeatedly by the calibrated assignment analyzer(s) and the mean of the replicated assays is assigned to the calibrator as assigned value(s).
Stabilized blood calibration method
Calibrators with assigned values are used to calibrate automated hematology analyzers. For calibration, the manufacturer’s instructions must be followed. Usually calibration is recommended after a major repair or a significant change of baseline accuracy (drift). The Clinical Laboratory Improvement Amendments of 198811 (CLIA) require (re)calibration at least every six months and when a complete change of reagents is introduced, when there is major preventive maintenance or replacement of critical parts, or when control materials reflect an unusual trend or shift of analyzer performance (par. 493.1217).
The number of calibration assays will influence the precision of the calibration factor. The error range of the calibration factor may be estimated using the standard error of the mean (s.e.m.) equation; greater assay iteration will result in a lower s.e.m.:
where s.d. = the analyzer imprecision and n = the number of replicated assays. The analyzer imprecision (s.d.) is calculated by making a number of consecutive (preference of 31) assays of the same well-mixed or stabilized blood specimen and applying the equation
where xi is the result of an individual assay, xa is the mean of n assay results, and n = the number of repeat assays. Note that any error in the calibration factor will be propagated to patient assays.
Stabilized blood controls with assigned values are sometimes used as less costly substitutes for calibrators. However, this practice is not acceptable. Although calibrators and controls from the same manufacturer may have a superficial resemblance, there may be sufficient differences in the manufacturing process, labeling, assignment of values, performance control and post-market surveillance to make calibrators and controls non-interchangeable.
Once calibrated, there should be a provision for independent verification of the calibration status. CLIA also requires such verification (par. 493.1217). Verification may be performed by assaying clinical specimens of which the results are known through measurement with prior-lot calibrated instruments and are compared to the values obtained with the current-lot calibrated instrument. Properly validated reference or comparison values may also be used.
Internal Quality Control
Once an analyzer’s performance baseline has been determined, its control is based on methods that will detect loss of accuracy, loss of precision or both. The methods must be sensitive enough to reveal loss of performance that could compromise patient assay values, yet cannot be oversensitive and signal errors where no errors exist. Control procedures must be carried out as an integral part of patient assay runs. A run should be related to analyzer workload and stability over time (drift) and not be simply based on time lapsed since the last control sample measurement. For control purposes, stabilized blood controls are generally used. These controls may have manufacturer-assigned values or laboratory-assigned values. Controls should have at least two levels of analyte concentration. These levels should reflect performance at “normal” analyte levels and at, or close to, the low limit of the reportable range. CLIA also requires at least two levels of analyte concentration (par. 493.1218). If the results of daily control assays are within laboratory-defined limits, e.g., ± 2 standard deviations (s.d.) of the assigned or conferred value, it may be assumed that the results of patient assays performed within a control-bracketed “run” are valid and reportable. The mean of replicate assays of control material is more informative than a single measurement and, thus, paired assays are often recommended. The mean of a pair of measurements provides a more exact value and, in addition, the difference between members of a pair provides an estimate of imprecision.
Out-of-limits control results should be able to distinguish between the following.
Random events vs. imprecision
With control limits defined as ± 2 s.d., one value in 20 will exceed these limits as a random event. Random events are excluded by two further assays of the control material that gave the out-of-limit result. If the difference between these assay results is less than the s.d. limits, the out-of-control result was likely a random event. If the difference is greater than ± s.d., loss of precision should be assumed and may be confirmed by two further assays. If at least two of the total of five results exceed the ± s.d. limits, precision has deteriorated.
Loss of accuracy
An out-of-limits control result due to loss of accuracy may be verified by the repeated assay method outlined above. If inaccuracy has developed, the mean of the replicate results will differ by at least 1 s.d. from the mean of previous control results.
If the accuracy change (bias) affects only one of the two (or more) control levels, deterioration of the control material should be suspected.
Control with Weighted Moving Average Method
This method monitors bias changes of the red cell indices, MCH, MCHC and MCV.
The effectiveness depends on the biological stability of the indices, i.e., an apparent change of any of these indices in an individual is more likely to be due to analytical error than to physiological factors.11 The method is practical because most modern hematology analyzers provide automated data acquisition and computation.
Patient assay results are automatically entered into an equation that gives a “running mean” and progressively compensates for variations among incoming results. The data batch sizes of n = 20 are suitable to provide stable means for comparison with target values. It is advisable, however, to minimize the use of clinical specimens with unusually high rates of red cell index dyscrasias.
Batch means are calculated and displayed and/or printed for review. Statistical comparison of mean values with target values for within-limits performance is determined and also displayed for review. Provision for editing outlier data points is provided. A reasonable goal for recognition of change is a 3 percent departure from a previously established and verified target value.
White Cell Differential Counts
Calibration and control of the white cell differential output of automated hematology analyzers is effected by following manufacturer’s directions. Comparison with manual differential counts can, however, be part of the overall quality control.
Dr. van Assendelft is director of Diagnostic Laboratories, National Center for Infectious Diseases, Centers for Disease Control and Prevention, U.S. Public Health Service, Department of Health and Human Services, Atlanta. Dr. Houwen is with the Department of Pathology and Human Anatomy, Loma Linda University (CA) Medical School.
1. International Committee for Standardization in Haematology; Expert Panel on Cytometry. The assignment of values to fresh blood used for calibrating automated blood cell counters. Clin Lab Haemat 1988;10:203-212.
2. American Society for Testing and Materials. Standard specification for disposable glass blood sample capillary tube (microhematocrit) designation. 1980, E734-780.
3. International Council for Standardization in Haematology. Recommendations for reference method for haemoglobinometry in human blood (ICSH Standard 1995) and specifications for international haemiglobincyanide reference preparation (4th edn). J Clin Pathol 1996; 49:271-274.
4. NCCLS. Reference and Selected Procedures for the Quantitative Determination of Hemoglobin in Blood; Approved Standard. NCCLS document H15-A3, 2000. NCCLS, 940 West Valley Road, Suite 1400, Wayne, PA 19087-1898.
5. International Committee for Standardization in Haematology; Expert Panel on Cytometry. Recommended method for the determination of packed cell volume by centrifugation. World Health Organization 1989, WHO LAB/89.1.
6. NCCLS. Procedure for Determining Packed Cell Volume by the Microhematocrit Method; Approved Standard. NCCLS document H7-A3, 2000. NCCLS, 940 West Valley Road, Suite 1400, Wayne, PA 19087-1898.
7. Bryner MA, Houwen B, Westengard J, Klein O. The spun micro-hematocrit and mean red cell volume are affected by changes in the oxygenation state of red blood cells. Clin Lab Haem 1997;19:99-103.
8. International Council for standardization in Haematology; Expert Panel on Cytometry. Reference method for the enumeration of erythrocytes and leukocytes. Clin Lab Haemat 1994;16:131-138.
9. International Council for Standardization in Haematology; Expert Panel on Cytometry and International Society of Laboratory Hematology; Task Force on Platelet Counting. Platelet counting by the red blood cell / platelet ratio method: a reference method. Am J Clin Path 2001;115:460-464.
10. NCCLS. Calibration and Quality Control of Automated Hematology Analyzers; Proposed Standard. NCCLS document H38-P, 1999. NCCLS, 940 West Valley Road, Suite 1400, Wayne, PA 19087-1898.
11. Health Care Financing Administration. Clinical Laboratory Improvement Amendments of 1988; Final Rule. Federal Register 1992;57:7107-86.
12. Korpman RA, Bull BS. The implementation of a robust estimator of the mean for quality control on a programmable calculator or a laboratory computer. Am J Clin Path 1976;65:252-253.