Automated Platelet Counts

Transcription

1 Automated Platelet Counts Accuracy, Precision, and Range DENNIS W. ROSS, M.D., PH.D., LANIER AYSCUE, M.T.(ASCP), AND MARGARET GULLEY, B.S. Ross, Dennis W., Ayscue, Lanier, and Gulley, Margaret: Automated platelet counts. Accuracy, precision, and range. Am J Clin Pathol 74: , The automation of platelet counting is essential in laboratories performing a large number of procedures with high precision. Maintaining this precision and establishing accuracy require an understanding of the special problems of counting platelets, including: (1) the large dynamic range of the measurement, (2) the variable size and aggregability of platelets, and (3) the specific requirements of a quality-control method. In evaluating these problems, we designed experiments that measured the linearity and precision of two types of platelet counters, light scattering and electronic aperture impedance, and that evaluated the suitability of commercially available reference materials. The results show that the instrumentation is excellent and the reference materials are good; such that, given a well-planned quality-control method as presented here, the automated platelet count gives a rapid result with precision and accuracy. (Key words: Automation, Reference materials; Quality control; Accuracy; Precision.) THE AUTOMATION of platelet counting is essential in laboratories performing a large number of procedures; however, accuracy and precision of the automated platelet count are continual challenges and more difficult to achieve than for the erythrocyte or leukocyte counts. A review by Wertz and Koepke 5 of results of platelet counting conducted as part of a College of American Pathologists survey showed a much wider variation among laboratories for all methods than could be explained on an analytic basis. Several reasons for this problem in controlling accuracy are: (1) the large dynamic range of the platelet count for patients, (2) the lack of a primary standard to calibrate instruments, and (3) the small variable size and aggregability of platelets. 1 The dynamic range (distribution of values encountered by the laboratory's measurement system) spans three orders of magnitude from 10 9 to /1. This wide range, not encountered for other hematologic values, challenges the system to operate Received December 4, 1979; accepted for publication December 27, Based in part on work presented at the ASCP-CAP Meeting, October 28, 1979, Las Vegas, Nevada. Address reprint requests to Dr. Ross: Hematology Laboratories, School of Medicine, University of North Carolina, Chapel Hill, North Carolina Hematology Laboratories, School of Medicine, University of North Carolina, Chapel Hill, North Carolina linearly in the low, normal, and elevated platelet ranges. The automated instrument must be directly calibrated or indirectly controlled with a reference material, the accuracy of which is an important factor in determining the accuracy of the automated platelet count. Automated instruments depend on size thresholds that allow accurate distinctions among platelets, small debris, and erythrocytes. Thus, to test this function, the size distribution of the particles in the reference material should compare with that in the normal platelet population. In evaluating these problems, we designed various experiments: (1) to measure the linearity and precision of two types of platelet instrumentslight scattering and aperture impedance counters; (2) to count, compare, and size commercial reference materials; (3) to determine the dynamic range for our patient population and the normal individual platelet range; and (4) to evaluate quality-control measures by comparisons of accumulative daily platelet results. Instrumentation Materials and Methods The instruments utilized in platelet counting are the AutoCounter and Hemalog 8/90,* which operate on the light-scattering principle, 2 and the Ultra-Flo 100,t which uses the electronic aperture impedance method. The cell-sizing apparatus consists of a pulse-height analyzer and amplifier^ connected to the Ultra-Flo 100. The Ultra-Flo 100 generates electronic pulses proportional to cell size with very high resolution because of a hydrodynamic focusing aperture. These pulses are collected and sorted according to size by the pulseheight analyzer, which displays an X-Y plot of the cell * Technicon Instruments Co., Tarrytown, New York. t Clay-Adams, Parsippany, New Jersey. t Nuclear Data, ND-60 and PSA-1, Schaumsburg, Illinois /80/0800/0151 $00.80 American Society of Clinical Pathologists 151

2 152 ROSS, AYSCUE, AND GULLEY A.J.C.P. August (A) LATEX SPHERES (B) FIXEO GOAT RBC (E lij m Z UJ <J > io r < (C) FIXED ANIMAL PLATELETS (D) FIXED HUMAN PLATELETS FIG. 1. Particle-volume distribution curves: relative cell number versus cell volume, logarithmic scale in femtoliters, for four different types of platelet reference materials. 0 L I LOG CELL VOLUME (fl) size versus cell number. A hard copy of the display is obtained on an XY plotter (Fig. 1). Linearity and Precision of the Instruments Linearity was determined by assaying prepared bloods with platelet counts ranging from 6 to 990 x 10 9 /1 in intervals of 50 x 10 9 /1 made by mixing various amounts of platelet-rich plasma and washed erythrocytes suspended in platelet-free plasma. Each specimen was assayed following a saline blank on the AutoCounter, Hemalog 8/90, and the Ultra-Flo 100. Precision was analyzed by repeatedly assaying wellmixed donor bloods. Measurement Distributions The results of platelet counts in all patient samples submitted to the Hematology Laboratories, North Carolina Memorial Hospital for a period of five working days (1,331 measurements) were collated into a histogram to ascertain the dynamic range (Fig. 2). The data for the distribution of repeat counts on a specimen from a normal individual were obtained from daily whole-blood controls used to check calibration and thresholding. Forty milliliters of blood were drawn not more often than weekly from healthy donors for this purpose; these data constitute a series of repeat platelet counts for the same individuals. To compare all data, an average of seven platelet counts each from ten donors were normalized to 1.00 by dividing each donor's weekly platelet count by the same donor's mean platelet count. This distribution was then reprojected about a mean of 260 x 10 9 /1. For the normal distribution, a large series of 907 platelet counts were taken from the data of Stevens and Alexander, 4 who performed counts for blood donors aged 18 to 65 years, using a Coulter Thrombo-counter. They used visual-phase platelet counts as a reference for accuracy. Our own smaller series of platelet counts for blood donors did not differ significantly from this large, well-studied group. Quality Control by Data Analysis Using a method similar to that introduced by Bull 3 for monitoring erythrocytic parameters, we calculated and plotted the running average of every group of 20- patient platelet counts. The large dynamic range of platelet counts caused excessive variation in the mean values; therefore, this method was modified to include a running average based on only those platelet counts within one standard deviation of the mean of our dynamic range. Another quality-control method was used to compare the present platelet counts and the previous counts for each individual patient. The running average of a Coulter Electronics, Hialeah, Florida.

3 Vol. 74 No. 2 AUTOMATED PLATELET COUNTS 153 Fie. 2. Distribution of platelet counts for: (a) patientswhite background, the dynamic distribution; (b) blood donors crosshatched background, the normal distribution; (c) one person black background, the individual distribution. PLATELET COUNT platelet count for a patient today minus the count for that patient yesterday (if available) was calculated and monitored as a measure of the stability of automated platelet counts. Commercial Platelet Calibration Materials Platelet reference materials evaluated in this study were from five commercial sources; their specifications for stability, composition; and quantity are listed in Table 1. Our own estimation of each reference material was made by calculating results obtained on the Hemalog 8/90. Each reference sample was used as a calibrant against which the other 12 were assayed, yielding a 13 x 13 matrix of values (Vjj). The sixth row 4 corresponds to the various counts obtained for the sixth reference, when the instrument was calibrated to each of the other 12 references. The j lh column represents the counts obtained for the other 12 references when the instrument was calibrated to the j th reference sample. Thus, each reference material was analyzed Table I. Technical Characteristics of Platelet Reference Materials Haem-PC* PLT-4t TQCt Quanticel HP Quanticel HRP Teckni-cal' Platelet-Chex Plus' Stability, closed vial Stability open vial Count ranges available Source of calibrant 6 months 8 days Normal, low platelets 45 days Vial size 0.6 ml 3 ml * J. T. Baker Diagnostics, Bethlehem, Pennsylvania. t R & D Systems, Minneapolis, Minnesota. t Technicon Instruments. Tarrytown, New York. 2 weeks Normal, low, extra-low Animal platelets 6 months montl 12 months 90 days 1 year Normal Normal, extra-low Normal, low Normal, extra-low Goat erythrocytes platelets erythrocytes Latex particles and platelets 10 ml 10 ml 7 ml 10 ml BHP. Inc.. West Chester, Pennsylvania. ' Streck Laboratories. Omaha, Nebraska. 6 months 35 days Normal, low erythrocytes and platelets 3.5 ml

4 154 ROSS, AYSCUE, AND GULLEY A.J.C.P. August 1980 _. 700 "o 600 * 500 V) K400 uj300 & 200 a! 100 J I I L J L J I I I PER CENT FIG. 3. A linearity determination for the Hemalog 8/90 automated platelet counter; platelet count as a function of percentage of added platelet concentrate. as a calibrant and as a sample when compared with other calibrants. Our estimate of the platelet concentration in each reference material was made by normalizing this matrix. This is equivalent to averaging the performance of each sample as a calibrant (values in column j) and as a sample when the instrument was calibrated to other references (values in row i) 9. Linearity and Precision Results The linearity of the automated platelet counting instruments (Hemalog 8/90, AutoCounter, and Ultra- Flo 100) is within 2% in the dynamic range (Fig. 3). The intraassay precisions of the three instruments, measured as the coefficients of variation for repeat measurements, are 3%, 7%, and 4%, respectively. It was found that the Hemalog 8/90 has a 5% carry-over from the previous sample, which will distort the linearity below 60 x 10 9 /1 platelets if a low count is not preceded by a saline blank. Commercial Platelet Calibrant Materials The size distribution of each reference material depends on its composition. Those consisting of preserved animal platelets most closely match the volume distribution of human platelets. Two sources containing latex beads and fixed goat erythrocytes have narrower size distributions than human platelets, as shown in Figure 1. These materials, if properly vortexed before use, are satisfactory in terms of a control for the count, but will not serve as a check on upper and lower thresholds. Table 2. Platelet Reference Materials: Determination of Accuracy in Platelet Counts Manufacturers Reference Values Multiple Averaged Results Mean (xl0 9 /l) Range* Mean (xl0 9 /l) (±1 SD) BHP, Inc. Quanticel, HP, normal Quanticel, HP, extra-low Quanticel, HRP, normal Quanticel, HRP, low J. T. Baker Diagnostics Haem Pc, normal Haem Pc, low R&D Systems PLT-4, normal PLT-4, low PLT-4, abnormal Streck Laboratories Teckni-cal, normal Teckni-cal, abnormal Teckni-cal, low abnormal Technicon Instruments TQC lot #1 lot # * Range as given by manufacturer if available.

5 Vol. 74 No. 2 AUTOMATED PLATELET COUNTS Table 3. Distributions of Platelet Counts 155 Range Definition Mean (xl0 9 /I) Standard deviation (xl0 9 /l) Number of Measurements Normal Dynamic Individual The distribution of platelet counts in a population without known disease The distribution of platelet counts in a population of patients referred to this laboratory The distribution of platelet counts for an individual without known disease on repeated determinations over a 90-day period , Our estimation of the accuracy of each commercial reference material is given in Table 2. All 13 reference materials are, by our method, within 11% of the manufacturers' stated values, many within 5%. Measurement Distributions Each of the histograms representing the distribution of platelet counts obtained for patients, for healthy blood donors, and for repeated determinations for an individual are shown overlapping Figure 2. The mean values and standard deviations of these distributions are given in Table 3. The width at 2 SD of the dynamic distribution is 60% greater than that of the normal distribution. Furthermore, a large proportion of the clinically significant platelet values lies outside these limits. The standard deviation for the individual distribution is only one fourth that of the normal distribution. Quality Control by Data Analysis The moving average calculated using the modified method of Bull' 5 showed slowly varying random fluctuations during five days of continuous operation. At one point, a steady slow drift towards increasing platelet counts was seen. The flow cell in the platelet counter was washed out and thoroughly flushed. The moving average promptly returned to baseline without further drifting. This was interpreted as a high-noise condition due to accumulated debris in the flow cell. The second technic, using a moving average of the difference between a patient's current platelet count and the previous platelet count for the same patient, was inadequately evaluated owing to a continuing problem with our ability for online capture and analysis of laboratory data. A small study analyzed by manual methods demonstrated that this moving average is a stable parameter. However, an investigation using this parameter calculated in real-time over many days of operating will be necessary to determine whether it is suitable for monitoring drift and single aberrant values in the automated platelet count. Discussion From our study of the various aspects of accuracy and precision in automated platelet counting, we have drawn several conclusions. First, automated equipment is capable of producing accurate results for low, normal, and elevated platelet counts if the equipment is properly controlled and calibrated. The existing instrumentation and quality-control methods for platelet counting provide greater accuracy and precision than was reported in the review by Wertz and Koepke. 5 Both the light-scattering and aperture-impedance counters are, when properly operating, linear and precise. We have shown that the dynamic range of the platelet count for patients is wide, thus requiring a system that is accurate over three orders of magnitude in the platelet count. Therefore, linearity becomes a very important feature, and each laboratory needs to verify this feature with each instrument. Given a linear instrument, it becomes necessary to determine at which point the system should be calibrated or checked for maximum accuracy. The clinical significance of platelet counts is greatest in the low range, where decisions about therapy are often based on relatively small changes. The calibration control on a lightscattering instrument varies the length of time for platelet counting, which is equivalent to adding or subtracting a percentage from all counts. For example, if the reference material is assayed 20% too high, the calibrate control is adjusted to the correct platelet value, thus reducing the counting time by 20%. An error of six in the lower-level calibrant adjusts the normal range by 60, whereas an error of six in the normal level calibrant adjusts the lower range by only 0.6. Thus, the interests of accuracy and precision are best served by calibrating to a reference in the normal range and using a low or high abnormal reference material as a check on linearity. Second, from our study of individual platelet ranges and from our experience with quality-control measures involving the modification of the Bull delta-check, 3 we propose two quality-control procedures. The pro-

6 156 ROSS, AYSCUE, AND GULLEY A.J.C.P. August 1980 cedures compare the accumulative daily results for each patient in order to check interassay precision. Two ways of doing this, both involving the use of an online real-time data processor are: (1) calculate and monitor the running average of all platelet counts that are within one standard deviation of the mean for the dynamic range; or (2) calculate and monitor the running average of a platelet count for a patient today minus the count for the same patient yesterday. The wide dynamic range of platelet counts eliminates the easy recognition of an "impossible" value, such as would be the case for a hemoglobin of 50 or a mean of corpuscular volume of 200. To control each patient's platelet count, the relatively small individual range allows a comparison with the previous counts for the same patient. The running average of all patient counts controls the instrument function continuously. The diagnosis and correction of instrument dysfunction is more efficient and prompt when constantly controlled by this method. Third, the majority of the reference materials available are acceptable as standards to assure accuracy. However, each laboratory should evaluate each lot number and develop a protocol for its correct storage and use. Given this attention to detail, the reference materials are suitable as calibrant sources and as checks on normal operation. The accuracy, precision, and quality control of automated platelet counts can be acceptably managed despite the inherent problems. A program of instrument checks, internal control of the reference material, and continuous monitoring of results is necessary to achieve this end. References 1. Brecher GB, Schneiderman M, Cronkite EP: The reproducibility and constancy of the platelet counts. Am J Clin Pathol 23:15-26, Brittin GM, Dew SA, Fewell EK: Automated optical counting of blood platelets. Blood 38: , Bull BS: A statistical approach to quality control, Quality control in haematology. Edited by SM Lewis, JF Coster. New York, Academic Press, 1975, pp Stevens RF, Alexander MK: A sex difference in the platelet count. Br J Haematol 37: , Wertz RK, Koepke JA: A critical analysis of platelet counting methods. Am J Clin Pathol 68: , 1977.