Automated Leukocyte Differentials: A Review and Prospectus

Transcription

1 Symposium Automated Leukocyte Differentials: A Review and Prospectus Thomas F. Dutcher, MD A utomation, or more correctly, instrumentation, that provides a high degree of accuracy and precision in blood cell counting has become commonplace in clinical laboratories. When considering the sophistication of the cell counters available today, it should be remembered that almost 30 years have passed since the introduction of the Model A Coulter Counter". Some modern counting and sizing instruments are truly automatic; that is, they accept a sample of whole blood, perform the appropriate dilutions, analyze the numbers and sizes of the cells, carry out appropriate calculations, and publish the results. They also fulfill, at least in part, the definition of automation that requires the instrument to provide a quality control of its own processing. In contrast to the almost 30-year availability of cell counters, instrumentation for the morphologic examination of blood smears has been around for less than ten years. As might be expected, these instruments are not as close to true automation as are the cell counters; nor is their contribution to increased accuracy in the performance of leukocyte differentials and erythrocyte morphology examinations as clear as are the contributions of cell counters. Nevertheless, the instrumentation available for leukocyte differentials offers a From the Dept of Clinical Pathology, William Beaumont Hosp, Royal Oak, Ml higher level of precision and consistency of "cell calls" than can be attained among a group of technologists. History of Instrumentation Instrumentation for leukocyte classification has developed along two pathways: (1) cytochemical characteristics identified in a flow system (the generations of the Technicon instruments that are considered elsewhere in this issue of Laboratory Medicine) and (2) the attempts to mimic the human eyeball-brain combination by pattern recognition and image analysis using computers for cell recognition and a mechanically driven stage that rapidly moves the slide until a cell is "captured" and viewed for analysis by the computer. The seeds of the microscopic image analyzers were planted in the early 1950s when the National Cancer Institute of the National Institutes of Health (NIH) funded the development of a microscope-analyzer marriage that was to survey cervical cytology smears and diagnose cancer. As a resident at NIH in the late 1950s, I was aware that such an instrument was being developed and that it was called the Cytoanalyzer. For a number of reasons, this instrument never worked and never became commercially available. However, basic techniques of cell scanning by mechanized microscopic stages, cell capture, and embryonic decision logic for image analysis and pattern recognition were developed. My next awareness of developments stemming from the Cytoanalyzer came in By that time, Perkin-Elmer had been working for about seven years on an automated microscope linked to computers for the classification of peripheral blood leukocytes. Computers of that time were large and relatively slow, and the programs for image analysis were not well developed. Reportedly, the Perkin-Elmer instrument (dubbed Cellscan) could, in 1966, recognize lymphocytes, monocytes, and neutrophils with 90% accuracy and a 100cell differential could be accomplished overnight. With the advent of smaller and faster computers and the development of more sophisticated image analysis logic, the Perkin-Elmer effort was briefly marketed as the diff 3b in the mid-1970s. The production rights were purchased by Coulter Electronics in 1977 and the instrument, with modifications, has been marketed as the Coulter diff 3 a. Parallel developments using variations of cell scanning techniques and recognition logic were fostered by others, and eventually four microscopic-image analyzing leukocyte differential instruments became commercially available. The first was the Corning LARCb, introduced in 1975, and followed quickly by the Geometric Data Hematrak0, the Coulter diff 3, and somewhat later, the Abbott ADC 500d. Of these four in- LABORATORY MEDICINE VOL. 14, NO. 8, AUGUST

2 Comparison of Two Technologists' Precision and Hematrak 360 Precision on Duplicate Differentials of Normal Blood Smears* Number of Smears Technologists Hematrak Segmented Band Total Lym ytes Monocytes Eosinophils 11% 10% % 20% % 6% % 15% % 17% % 29% "Modified from Dutcher, Jakubowski, and Orser5 struments, only the Hematrak has survived into its third generation, alt h o u g h Coulter h a s announced a redesigned instrument of the image analysis type the Coulter diff 4. Comparisons with Technologist Performance As t h e s e i n s t r u m e n t s b e c a m e available, several reports appeared in the literature citing comparisons with technologist performance, particularly with regard to speed of analysis, precision, and accuracy. The best sole compendium of such evaluations is the book published by t h e College of American Pathologists (CAP) edited by John Koepke, MD, entitled Differential Leukocyte Counting.1 This book contains considerable food for thought concerning aspects of the differential count in addition to evaluations of individual instruments. 2-4 It contains one comparative evaluation of the performance of laboratory technologists, the LARC, Hematrak, and Technicon Hemalog D e. 5 The most recent comparative evaluation of three obsolete instruments (Geometric Data Hematrak 360, Coulter diff 3, and Technicon Hemalog D/90) appeared in the J a n u a r y 1983 issue of the Journal of Clinical Laboratory Automation.6 This journal has also recently carried evaluations of the Hematrak Model 480 d7 and Technicon H6000 f. 8 Findings by all evaluators of each instrument are similar and can be briefly stated as follows: (1) The instruments are equal to or better than technologists with regard to precision (repeatability on the same specimen or same blood smear), especially in the classification of bands and monocytes. (2) The instruments are equal to or better than technologists with regard to accuracy, at least relative to finding abnormal cells (accuracy, as thought of in chemistry or even cell counting, is nearly impossible to assess in differential counting because of the inherent variability in 100- or 200-cell differentials). (3) The instruments produce results faster than technologists as timed from presentation of the slide to results available for reporting. (4) The instruments segregate smears into normal or abnormal (abnormal because of cell distribution outside the normal range and the presence of abnormal cell types) as well as do technologists. Depending on the criteria established by the laboratory director for a manual review of smears when the instrument discovers some level of abnormality, the instrument can be expected to successfully process 70% to 80% of samples without a manual review and up to another 15% with no more than a rapid scan, often on the instrument, by a technologist. Clearly, t h e s e e s t i m a t e s r e l a t e directly to the proportion of patients in the hospital populace who do not have circulating abnormal cells. A representative comparative precision study of technologist performance a n d t h e p e r f o r m a n c e of t h e Hematrak 360 as determined by statistical analysis of the results of duplicate 200-cell differentials from the same smear for 40 normal persons is presented in the Table. This data is similar to the results on each of the image analysis instruments for which evaluations have appeared. The information indicates the accuracy of cell identification of normal cells as well as the precision of repeat analyses. Other applications of some first g e n e r a t i o n i n s t r u m e n t s were the interpretation of red cell morphology, estimation of platelet adequacy, and reticulocyte counting. In my experience, and as reported by Gulati and co-workers 7 for the Hematrak 480, the instrument performance is not as good in these areas as it is in leukocyte identification. Evaluation of the newer Hematrak Model 590 and the Coulter LABORATORY MEDICINE VOL. 14, NO. 8, AUGUST 1983 diff 4 will be of great interest in assessing better performance in these areas. Purchasing Decisions One of the most useful aspects of studying published evaluations is to direct attention toward purchase of any that are clearly outstanding. This was not a real factor in the first generation instruments because the performance level was similar in all cases and purchase decisions became an ancillary matter. It seems unlikely that there will be any overwhelming distinction among the newer instrum e n t s. T h e r e a r e other aspects to consider prior to purchase, and some of these relate to the size, type, and philosophy of the individual laboratory. Criteria concerning purchase for such aspects as utility and cost justification must be thought through by each l a b o r a t o r y ; h o w e v e r, some guidelines might be helpful. Utility We have considered utility within four categories. Capability By this is meant the capability of the instrument to reliably evaluate a high percentage of the smears presented to it without the necessity of a manual review. Unless the newer ins t r u m e n t s h a v e more reliable programs for the positive identification of monocytes (often r e g a r d e d as a "suspect" cell in the early models), technologist review will continue to be necessary for a number of normal differentials in which the instrument has flagged monocytes as suspect cells. A similar consideration will be necessary for the accuracy of programs for e r y t h r o c y t e m o r p h o l o g y a n d platelet estimation. Evaluation of the accuracy for identifying the presence of abnormal cells will also be necessary because the newer models are said

3 to have expanded their repertoire of cell identification. More need for manual review can be expected in hospitals with a busy hematology-oncology service than in laboratories that infrequently process such specimens. Capability also includes the type of specimen that can be processed. It is interesting, but not of proven causality, that the extinct instruments required spun smears and therefore only venipuncture specimens could be used. The n e w H e m a t r a k a n d t h e new Coulter accept either spun smears or appropriately spread wedge smears. This will be advantageous in laboratories dealing with a significant number of pediatric patients. Speed of analysis The second criterion of utility is speed of analysis. The newer instruments are said to process some 100 specimens per hour while unattended. How nearly this level is actually achieved will depend on the number of smears that need manual review. For instance, in the report by Gulati and co-workers, 7 the Hematrak flagged 39 out of 593 specimens (6.6%) because of cells of abnormal morphology. I assume that each of these would be subjected to manual review. Manual review, even if it can be done with the smear still on the instrument, decreases the effective throughput. If erythrocyte morphology recognition is not reliable, even more reviews will be necessary. However, noting a need for manual review of RBC morphology will be facilitated by a comparison of the routinely obtained erythrocyte indices with the morphology as reported by the instrument. I sometimes wonder about the emphasis on speed. After all, it takes ten minutes to make and stain the smear. Whether the analysis takes 45 seconds or 95 seconds is relatively inconsequential for the total turnaround time within the laboratory. Speaking of turnaround times, I question the value of stack loading up to 100 smears and then letting the instrument run in a "walkaway" mode. Can your laboratory stain 100 smears at a time? Also, the report of the first smears analyzed cannot be released pending possible m a n u a l review u n t i l t h e "walked away" technologist walks back after up to one hour. This can significantly times. prolong turnaround Availability A third element of utility assessm e n t is a v a i l a b i l i t y. This m e a n s availability for analysis of specimens 24 hours a day. Availability should not be a problem with the new microscopic analyzers because they can be in a constant standby mode and used even for a single specimen received in the wee hours of the morning. Availability also includes downtime and adequacy of service. This assessment will vary by individual instrument, by vendor, and by the care given to the instrument by the user. Technologist acceptance The fourth and last element is technologist acceptance. In general, this should be no problem. Yet there are t h o s e who j u s t don't t r u s t i n s t r u ments or have the misguided notion that if there are only a few differentials to do, as in the late evening or early morning, it is easier to do them manually. This practice defeats one of the major attributes of instrumentation; namely, consistency of cell identification, which is unlikely to be as good among a group of technologists. Cost justification The final item for consideration in purchasing is the cost justification. This also must be assessed individually by each laboratory, taking into account the workload, original price, continuing reagent costs, technologist time saved, technologist salary, charge for the procedure, benefits of patient care in consistency of cell identification and lesser turnaround times, and the capability for 200 or more cell differentials. It seems to be generally agreed that a daily workload of 250 to 300 or more differentials is necessary to justify purchase on purely financial grounds. Some Problems You may have noticed that much of the preceding portion of this article has been written in the past tense. Because of the current lull between the old and new generations of the instruments, there is not much to be said in the present tense. Yet it may be interesting to do a bit of speculation in the future tense. With the advent of "automated" leukocyte differentials, there was a brief period of excitement that the leukocyte differential might become an even more clinically valuable laboratory test. It is ironic that instead of this, the automation has fostered a renewed evaluation of the worth of differentials in patient care. Bands vs Polys One concern is the value of continu i n g to a t t e m p t t h e distinction of "bands" from "polys." Most evaluations of the differential instruments have noted that technologist variability in making this distinction is so great t h a t comparisons between technologists and instruments cannot be considered valid. The problem has been noted by the companies that make the instruments, and they have allowed for software manipulations that allow the user to maintain a "high band" or a "low band" environment. The problem has often been highlighted in the color transparencies sent out in the CAP Survey program. A few years ago, a cell that was clearly a band by the criteria established by the CAP Survey Hematology Resource Committee and so stated in the instruction manual for the Survey, was called a band by about half of the 2,000 participating laboratories and a poly by the other half. In the Table, the coefficient of variation () for bands is 20%, compared with only 8% for total neutrophils. In a similar study involving 12 technologists and a paired differential on 254 specimens, Gilmer 3 recorded a of 198% in the reporting of bands. Gulati and co-workers7 quote a for bands from a 1976 CAP Survey sample of 84%, and I assume that they accept this as not out-of-line with their laboratory. Variability of Cell Classification In addition to this remarkable variability among technologists, which makes any "band count" suspect, the inherent variability of cell classification in a perfectly made smear with random distribution of cells and an error-free technologist calling the cells has been studied by Rumke. 9 From tables he constructed, it can be appreciated that there is a statistical chance that a 100-cell differential from a blood smear with a "true" 12% band count might be reported as anything LABORATORY MEDICINE VOL. 14, NO. 8, AUGUST

4 between 5% and 20%. Although unlikely in practice, it is statistically possible to report an apparent fourfold increase in bands (5% to 20%) even though the true count is a constant 12%. Considered in a slightly different way, Rumke's data show that following a differential with a reported 12% bands (or any other cell type), a second differential would have to have greater than 23% bands or fewer than 3% bands before the difference could be considered statistically significant. The danger, of course, lies in making a significant clinical judgment on the basis of apparent changes in band counts that are not real but only reflect predictable statistical factors and wholly unpredictable technologist variability. Through discussion with the patient-care physicians at William Beaumont Hospital about such variability and the potential for making a clinical error on the basis of band counts, we have reached an agreement not to include bands in the differential count. This policy has been in effect for more than one year with no known detriment to patient care. Biologic Variability In addition to the variability of technologists reporting, particularly with regard to bands vs polys and the inherent variability of any cell type as highlighted by Riimke, there is a growing volume of information that indicates a personal biologic variability. Each person seems to have his own daily pattern 1 0 (W. Leeburg and T. F. Dutcher, unpublished data, 1982). In some persons the white cell count may increase by several thousand from early morning to early afternoon; in some, the white count may remain relatively stable; in others, it may decrease by several thousand; and in still others, t h e variability differs in direction from day to day. In general, this lability is a consequence of swings in the absolute neutrophil count and the absolute lymphocyte numbers are stable. This biologic variability makes the interpretation of a changing leukocyte count and differential difficult, especially so for sequential differentials or the "daily CBC and diff' that is so frequently ordered. There is also an increasing awareness that the combined technical and biologic variability is considerably greater than is the sensitivity of the differential for selecting out most diseases Exclusions from this generalization include such relatively rare diseases as leukemia and acute infections. Several articles 1314 in the literature attest to the overutilization and ineffectiveness of the leukocyte differential. A recently published study 15 states that in an analysis of 799 outpatient leukocyte differentials, 475 were considered to have been requested without any identifiable clinical indication. Only 63 (13.3%) of these yielded an abnormal differential cell count. Fifty of these abnormalities were mild to modest increases or decreases in one of the normal cell lines and associated with a normal total leukocyte count. None of these patients was subsequently determined to have a clinically significant disease related to the isolated abnormal differential. The other 13 outpatients had an abnormal differential in association with an abnormal leukocyte count and no follow-up information is presented. This study and others done on both outpatient and inpatient differentials suggest that as many as 95% of differentials t h a t are ordered without some specific clinical indication are wasted, in the sense of not discovering any disease associated with leukocyte abnormalities. This does not take into account any abnormalities of erythrocytes or platelets that may be discovered in a peripheral smear examination. The benefit of viewing smears for erythrocyte morphology in patients with anemia has been studied by J e n et al. 17 They reviewed the erythrocyte morphology of 183 patients with anemia and found that only in two patients did the review reveal information that was not predictable from the routine RBC indices, other aspects of t h e CBC, or by the obvious clinical picture. Future of Differential Counting Whither then goeth the diff? Crystal ball gazing is always subject to considerable chances for error, but I think the future of leukocyte differential counting is visible on the horizon. A few evaluations of the technology now available, selection of the most efficient technology for a particular application (large hospital laboratory, small hospital laboratory, inpatient or outpatient laboratory, reference LABORATORY MEDICINE VOL. 14, NO. 8, AUGUST 1983 laboratory, etc) and then, the most difficult task, education of the patient-care physicians, will yield remarkable changes within the next five years. The two technologies that require evaluation are the cytochemical differential exemplified by the Technicon H6000 f a n d t h e t h r e e - h u m p histogram of leukocyte size groupings now available from Coulter. By the time this article appears, Ortho r may offer a similar histogram. Each of the instruments from these corporations will also provide size distribution information on erythrocytes and platelets. Each also presents the leukocyte differential data as a percentage and as an absolute value. Information to the requesting patient-care physician will be routinely reported as the absolute count r a t h e r t h a n t h e now commonplace percentage differential. Without trying to predict the best instrument or combination of instruments, I believe that the general procedure for routine hematology studies will be roughly as follows: the blood specimen will be presented to the instrument for sampling and will be analyzed in a truly automated fashio with the instrument internally providing on-line process control and flagging abnormal features of the cytologic findings to identify those specimens that should be visually scanned. It may be t h a t the safest processing of the initial specimen from a hospitalized patient will include a scan of a smear as a precaution against the presence of a small population of abnormal cells such as blasts, nucleated RBCs, or abnormal lymphocytes, and also to be certain that peculiar cells such as Pelgeroid neutrophils, hypersegmented polys, vacuolated neutrop h i l s, etc, a r e not p r e s e n t. If no abnormal cells are found on a visual scan of at least 500 cells, subsequent "differentials" can be accepted from the instrument assuming that no new flagging situation has arisen. Notice that the predicted mode of processing does not include a full 100or 200-cell differential. It has been shown 18 that if no abnormal cells are seen in a population of 500 cells observed on a blood smear, there is a 99% confidence level that no abnormal cells are present. I am confident that technologists are competent to identify abnormal cells by scanning

5 with a x 25 oil immersion objective. In order for reliance on a scan to be valid, evaluations must show that the distribution of leukocytes determined by the instrumentation is valid. The flow system cytochemical differential has been around long enough that flagging criteria and interpretation of the oscilloscope patterns and cell classifications are well developed. Early indications are that the information presented in the volume distribution histogram format is reliable, but flagging criteria have not yet been firmly established. One early appreciation is that the cell size histograms, presented by Coulter and Ortho as determined during cell counting, are not necessarily lymphocytes, monocytes, and granulocytes, but cells of the appropriate size range. Methods of examination of the peripheral blood smear have been virtually unchanged since the early 1900s when giants like Arneth, Naegeli, and Schilling peered through their microscopes. Obviously, the peering at abnormal smears will continue, but I believe that the technology of today, including current developments of monoclonal antibodies that are spe- cific for individual types of blood cells, will soon replace the traditional leukocyte differential. The utilization of the available and coming technology will require a massive educational effort for those in the laboratory as well as those involved more directly in patient care References 1. Koepke JA (ed): Differential Leukocyte Counting. Proceedings of CAP Conference/Aspen, Skokie, IL, College of American Pathologists, Trobaugh FE, Bacus JW: Design and performance of LARC automated leukocyte classifier, in Koepke JA (ed): Differential Leukocyte Counting, Skokie, IL, CAP, 1978, chap Gilmer PR: Coulter Perkin-Elmer diff 3 automated counting system, in Koepke JA (ed): Differential Leukocyte Counting, Skokie, IL, CAP, 1978, chap Pierre RV: The Technicon Hemalog D automated leukocyte differential system, in Koepke JA (ed): Differential Leukocyte Counting, Skokie, IL, CAP, 1978, chap Dutcher TF, Jakubowski D, Orser B: Comparative evaluation of automated blood cell differential analyzers: Hematrak, LARC, and Hemalog D, in Koepke JA (ed): Differential Leukocyte Counting, Skokie, IL, CAP, 1978, chap Marchand A, VanLente F, Galen RS: Automated differential leukocyte counters: comparison of three systems. J Clin Lab Autom 1983;3: Gulati GL, Hyun BH, Ashton JD: Hematrak Model 480 automated differential system: clinical evaluation. J Clin Lab Autom 1982;2: Hosty TA, Harris LG, Stonacek SM, et al: Eval uation of automated continuous-flow instrument: Technicon H6000. J Clin Lab Autom 1982;2: Rumke CL: Statistically expected variability in differential leukocyte counting, in Koepke JA (ed): Differential Leukocyte Counting, Skokie, IL, CAP, 1978, pp Statland BA, Winkel P: Physiological variability of leukocytes in healthy subjects, in Koepke JA (ed): Differential Leukocyte Counting, Skokie, IL, CAP, 1978, pp Klee GG, O'Sullivan MB: Screening vs diagnostic differential leukocyte counts, in Koepke JA (ed): Differential Leukocyte Counting, Skokie, IL, CAP, 1978, pp Bull BS, Korpman RA: Characterization of WBC differential count. Blood Cells 1980;6: Brecher G, Anderson RE, McMullen PD: When to do diffs. How often should differential counts be repeated? Blood Cells 1980;6: Rock WA, Grogan JE: Demand vs need vs physician prerogatives in use of WBC differential. JAMA 1983;249: Rick EC, Crowson TW, Connelly DP: Effectiveness of differential leukocyte count in case findings in a m b u l a t o r y care setting. JAMA 1983;249: Jen P, Woo B, Rosenthal PE, et al: Utility of peripheral blood smear in diagnosis of anemia. Clin Res 1981;29:321A. Barnett CW, quoted by Necheles TF, Frestensky R: State of art for leukocyte counting a clinician's view, in Koepke JA (ed): Differential Leukocyte Counting, Skokie, IL, CAP, 1978, pp Suppliers a. b. c. d. e. f. Coulter Electronics, Inc, Hialeah, FL. Corning Medical, Medfield, MA. Geometric Data, Wayne, PA. Abbott Laboratories, North Chicago, IL. Technicon Instrument Corp, Tarrytown, NY. Ortho Diagnostic Systems, Raritan, NJ. LABORATORY MEDICINE VOL. 14, NO. 8, AUGUST