The White Blood Cell Differential. Evaluation of Rapid Impression Scanning Versus the Routine Manual Count

Transcription

1 628 BRODY ET AL. A.J.C.P. May 987 culating lymphocytes may reflect their release from reticuloendothelial organs, entry of new lymphocyte cohorts from lymphoid tissue, or spontaneous reexpression of epitopes on the same cells that lost these markers during surgery. The unfailing decrease of the T4 subset may reflect a special vulnerability of this phenotype to mechanical or immunogenic injury. The clinical relevance of these observations may be viewed in at least two ways. The first is that transient reductions in the T4/T8 ratio may be part of a normal immune response to blood transfusions. Temporary inactivation of immunoreactive lymphocyte clones may be beneficial and prevent, or at least limit, immunization to blood products or other antigens. 0 The second perception is that these circumstances could make patients susceptible to a number of infectious diseases occurring after blood transfusions, such as hepatitis, the postperfusion syndrome caused by the cytomegalovirus and Epstein-Barr virus, 8 and transmission of human lymphotropic virus (HTLV) III. 9 The fact that most immunocompetent patients undergoing CPB do not develop postoperative viral infections attests to their ability to sustain themselves against stimuli potentially injurious to their immunity. On the other hand, occult, defective immunoregulation may be accentuated during open heart surgery, and patients so impaired could be liable to infectious sequelae. The reversal of the The White Blood Cell Differential T4/T8 ratio two hours after open heart surgery and reports of acquired immune deficiency syndrome developing after this procedure 4 ' 6 make the foregoing speculation a potential reality. References. Black PH: Shedding from normal and cancer-cell surfaces. N Engl J Med 980;0: Boyum A: Isolation of mononuclear cells and granulocytes from human blood. Scand J Clin Lab Invest 968; 2(suppl 97): Collett B, Alhaf A, Abdullah NM, et al: Pathways to complement activation during cardiopulmonary bypass. Br Med J 984; 289: Delpre G, Ilfeld D, Pitlik S, et al: AIDS after coronary bypass surgery (letter). Lancet 984; :0 5. Gascom P, Zoumbos NC, Young NS: Immunologic abnormalities in patients receiving multiple blood transfusions. Ann Intern Med 984; 00: Jett JR, Kurnitsky JA, Hornburger HA: Acquired immunodeficiency syndrome associated with blood product transfusion. Ann Intern Med 98;99: Kessler CM, Schulof RS, Goldstein A, et al: Abnormal T-lymphocyte subpopulations associated with transfusions of blood derived products. Lancet 98; : McMonigal K, Horwitz CA, Henle W, et al: Post-perfusion syndrome due to Epstein-Barr virus. Transfusion 98; 2:5 9. Perkins HA: Transfusion-associated AIDS. Am J Hematol 985; 9: Woodruff MFA, varood JJ: Possible implications of the effect of blood transfusion on allograft survival. Lancet 98; :20-20 Evaluation of Rapid Impression Scanning Versus the Routine Manual Count CHARLES F. ARKIN, M.D., L. JEFFREY MEDEIROS, M.D., LEONID Z. PEVZNER, M.D., PH.D., BETHANNE P. GUERTIN, MT(ASCP), PAULA J. KOBOS, MT(ASCP)SH, JANE W. PHELPS, MT(ASCP), AND SANDRA J. SMITH, MT(ASCP) A quick, one and a half-minute qualitative microscopic scan was investigated as an alternative approach to the more labor-intensive 00-cell differential white blood cell count. The scanning results of 400 randomly selected hospital cases were compared with the on-line results of the 00-cell counts. Additionally, SO cases selected to have a high percentage of abnormal results were each scanned and manually counted by four different readers. The results indicate that the scanning differential is equivalent to the 00-cell manual count in the detection of the presence of abnormal cell types such as immature granulocytes and blasts. Its ability to properly estimate the relative proportions of normal Received April 7, 986; received revised manuscript and accepted for publication October 29, 986. Address reprint requests to Dr. Arkin: Department of Pathology, New England Deaconess Hospital, 85 Pilgrim Road, Boston, Massachusetts Laboratory of Pathology, New England Deconess Hospital, Boston, Massachusetts, and Laboratory of Surgical Pathology, Stanford University Medical Center, Stanford, California cells, especially lymphocytes, however, does not appear as reliable as the manual count. Most importantly, the analysis demonstrates that the scanning differential count exhibits a set of advantages and disadvantages that is complementary to those of the "three-part differential" technic provided by the newer generation automated hematologic analyzers. The authors therefore propose that these two procedures used in combination offer a suitable alternative to the manual 00-cell differential count. (Key words: Differential leukocyte count; Leukocytes; Three-part differential; White blood cells) Am J Clin Pathol 987; 87: Downloaded from by guest on 0 September 208

2 Vol. 87 No. 5 BRIEF SCIENTIFIC REPORTS 629 THE WHITE BLOOD CELL differential count (DIFF), one of the most widely performed of clinical laboratory procedures, is nonetheless controversial with regard to its medical utility. 205 It is highly labor intensive and relatively expensive to perform. In addition, its inherent sampling variation, along with unavoidable interoperator differences in classification criteria, combine to limit the reliability and usefulness of the test. 6 " 8 Automated technology has become available over the past 5 years, both in the form of sophisticated computerized pattern recognition systems and sensitive flow cytometry devices. These technics have not, as yet, replaced the manual DIFF on a widespread basis because each has substantial drawbacks of its own such as high costs, mechanical problems, and difficulties in interpreting abnormal specimens. ' 5 ' 9 ' 7 ' 9 As an alternative to the manual 00-cell DIFF, we investigated the utility of performing DIFFs by low-power microscopic scanning of large areas of a peripheral smear and qualitatively estimating the results much as one does when scanning a vaginal cytologic smear. We have compared these estimates with their respective manual counts to determine the presence and extent of any important differences in test performance. s The scanning technic for DIFFs was performed on peripheral blood smears stained and spread by our routine technics i.e., a Wright's-Giemsa stain and a centrifugal peripheral smear device. Four medical technologists participated in the scanning. All had several years of experience in reading differential counts (4-7 years), and their duties are almost entirely limited to the hematology laboratory. They were charged to examine the peripheral smear for no longer than.5 minutes, using a stopwatch. Instructions required this time were divided roughly into a 5-second general impression at 0X magnification, a second scan at 20X magnification, and a 50- second scan at 50X magnification. The time included red blood cell and platelet morphology and a platelet estimate, but not report time. After the scan, the technologist completed a worksheet that indicated any perceived abnormality in white blood cells, red blood cells, or platelets. The study was conducted in two parts. In part, 400 randomly selected patient slides on which a differential count had already been performed were distributed among the four scanners. Each manually counted DIFF was compared with its respective scanned DIFF, and the numbers of agreements and disagreements with regard to the levels of neutrophils, bands, eosinophils, monocytes, and lymphocytes were tallied along with the presence of the following abnormal cell types: atypical lymphs (greater then per 00 cells counted), myelocytes, metamyelo- Table. The Range of Cell Counts that Determine the Designations "Low," "," "Borderline," and "High" for Manual Differential Counting for Each Major Neutrophil Band Eosinophil Monocyte Lymphocyte Low <40 < Borderline 7-4 High >70 >4 >5 >2 >50 cytes, promyelocytes, blasts, and nucleated red blood cells. Part was intended to evaluate the impact of the scanning DIFF on the hospital population in general. Part 2 of the study was designed to allow a closer look at the performance of the scanning DIFF in abnormal specimens and to allow comparison of the interoperator reproducibility of the scanning technic with the manual technic. In this part 50 patient specimens selected for their prominent abnormalities were each scanned by the four scanning technologists and read manually by four other medical technologists. The results of the four technologists for each case for both scanning and manual counting methods were then compiled and compared. reproducibility (interoperator agreement) was judged by comparing the number of cases showing four operator agreements, three operator agreements, and two operator agreements. Contingency tables were prepared from data from parts and 2 for each white blood cell type to compare scanning and manual counts. For the normally present cell types, the categories were designated as "normal," "increased," or "decreased" (Table ). values were based on previous in-laboratory studies on 0 subjects. Band counts were considered elevated only when above the borderline range. Abnormal cell types were categorized as being present or not. Chi-square analysis was used to test for the presence of differences between methods. A case-by-case review of the data was carried out by a pathologist when it appeared necessary for a more meaningful interpretation. In part 2 truth was defined as the mean value of the four manual counts. Results The total numbers of designations (normal, increased, etc.) for each cell type by both scanning and counting are shown in contingency table form for part and part 2 data in Tables 2 and, respectively. These tables show the relative "liklihood" of each technic in making a certain designation. Table 4 shows interoperator reproducibility in terms of the number of readers in agreement for both Downloaded from by guest on 0 September 208

3 60 ARKIN ET AL. AJ.C.P. May 987 Table 2. Contingency Tables for Each Major White Blood Giving the Numbers of Designations Obtained by Both Scanning and Counting Technics Table. Contingency Tables for Each Major White Blood Giving the Numbers of Designations Obtained by Both Scanning and Counting Technic from 400 Randomly Selected Hospital Patients Lymphocytes* Polymorphonuclear neutrophilsf Band neutrophilsf Monocytes Eosinophils^ Immature granulocytes /><0.00. tp<0.05. tp<0.0i. Designation Decreased Decreased and borderline Counting Scanning the scanning technic and the manual count for those cell types where there are adequate numbers of cases. Lymphocyte Counting The data from both the random (part ) and selected (part 2) cases reveal a clear tendency for the scanners to overlook low lymphocyte counts. There appears to be a mild tendency to overcall elevated numbers of lymphocytes in the randomly chosen cases, but data from the selected cases show good agreement in cases of lymphocytosis. Analysis of the data from the 50 selected cases shows an overall sensitivity to lymphopenia of 62.5% and to lymphocytosis of 92.9%. Case-by-case review shows that the tendency to miss lymphopenia extends to lymphocyte counts below 0% as 9 of 68 such cases were misclassified as normal. Interestingly, however, one of the four scanning technologists showed virtually perfect lymphocyte assessment. As would be expected, the level of operator agreement for lymphocyte counting is statistically significantly better by manual counting than by scanning (Table 4). Polymorphonuclear Leukocyte Counting Table 2 suggests a tendency for the scanners to overlook increased proportions of polymorphonuclear leukocytes. This may be an extension of their failing to notice decreased lymphocytes. Because the relative numbers of Lymphocytes* Band neutrophils Monocytes Eosinophils Normoblasts Immature granulocytes Blasts Designation and borderline Counting Scanning Agreement levels are based on the number of technologists in agreement. The data base is SO patient peripheral smears selected to have a high number of abnormalities. 7><0.00. polymorphonuclear leukocytes are generally inversely related to lymphocyte counts, these data were not extensively analyzed. Table 4. Contingency Tables Showing the Number of Cases Falling into Various Levels of Agreement Among Four Technologists for Both Scanning and Counting s by Lymphocyte counts (50 cases both normal and abnormal)* No. of operators in agreement of 4 Manual method Scanning method Band counts (50 cases both normal and abnormal) No. of operators in agreement Manual method Scanning method 5 of 4 5 Immature granulocytes (0 cases detected; No. of operators detecting the finding of 4 Manual method 0 Scanning method 0 Blasts (9 cases detected) No. of operators detecting the finding Manual method Scanning method lof of 4 of Agreement levels are based on the number of technologists in agreement. The data base is 50 patient peripheral smears selected to have a high number of abnormalities. P<0.0. Downloaded from by guest on 0 September 208

4 Vol. 87 No. 5 BRIEF SCIENTIFIC REPORTS 6 Band Neutrophils Counting Looking at the ability of the scanning technic to detect frankly elevated band counts (greater than 4%), data from the 400 randomly chosen cases show the scanners detecting 54 of the 6 cases detected by manual counting (88.5%). Data from the 50 selected cases suggest that both technics report similar numbers of normals and abnormals (Table ), but a further breakdown of the data reveals that different cases were designated as abnormal by the two methods so that only 27 of cases (6.4%) having elevated bands by manual counting were detected by scanning. By looking at these cases individually, it appears that the scanning technic may be most effective at band counts of more than 8% where 9 out of 20 scans were correctly classified as abnormal (95%). Monocyte Counting Case-by-case analysis of the randomly selected cases shows 27 cases of monocytosis called normal by scanning. Seven occurred in cases of monocytosis above 6% and four above 8%. Data analysis of the selected cases shows that scanning detected monocytosis in of 20 opportunities, while the manual method detected 4 of 20. Eosinophil Counts The random cases but not the selected cases suggest that the scanning technic undercalls elevated eosinophil counts. Further examination of this data reveals that of 2 cases designated as elevated by counting but normal by scanning, only eosinophil count was as high as 0% and the large majority were borderline counts of 6% and 7%. At this level there is also substantial disagreement among the manual counts. The data from the 50 selected cases shows 8 cases of eosinophilia were detected by scanning as compared with nine by counting (Table ). Abnormal s (immature myeloid cells, blasts, nucleated red blood cells) Tables 2-4 indicate that the above-abnormal cell types were detected with similar sensitivities and with similar interoperator agreement. Further analysis of the data from Table 2 reveals that of the cases with immature myeloid cells, 9 were detected by scanning only and 9 were detected by manual count only, while 27 cases were detected by both. Of the selected cases, there were nine with blasts, five of which had a blast cell count of less than 0%. The scanning technic detected blasts times of 6 opportunities, while the manual method detected blasts only 24 times. Basophil Counts and Atypical Lymphocytes The case populations under study in both parts of the study had insufficient numbers of basophils and atypical lymphoctyes to provide a meaningful evaluation. Discussion There has long been a legitimate question as to whether the information conveyed by the DIFF can, in general, justify its relatively high cost and inconvenience. 82 The findings are usually nonspecific, especially when alterations in the count are less than striking. This results in part from the variety of physiologic alterations that can affect bone marrow function and ultimately the DIFF. 4 This nonspecificity is also a result of the large variations in the DIFF both between persons and in the same person from day to day. 6 Furthermore, there is sizable imprecision of the procedure itself because of sampling error. 27 This effect is obligatory and becomes proportionately greater as the count of the particular cell type deviates from 50%. Differences in cell classification criteria, nonhomogenous distribution of cells on the slide, and technologist fatigue are additional factors influencing DIFF count imprecision. 047 Automation of the DIFF has not been completely successful in circumventing the problems presented by the manual DIFF. The highly sophisticated computerized pattern recognition devices have been extensively evaluated and are credited with acceptable performance.,5,97 ' 9 They have not, however, met with widespread acceptance in the marketplace, probably because of their marginal cost savings and frequent mechanical problems. Their use is, therefore, limited to a relatively few large clinical laboratories. Flow cytometric technics have become increasingly popular in recent years, primarily because they are incorporated into the newer generation multiparameter hematologic cell analyzers. 7 " They use conventional cellsizing technics and markedly improve the precision of cell classification by counting large numbers of cells. On the other hand, most systems in use place leukocytes into only three categoriesneutrophils, lymphocytes, and monocytes 4,6 and there is no attempt to subclassify eosinophils, basophils, atypical lymphocytes, immature myeloid cells, blasts, or nucleated red blood cells. Also, although they provide an index of anisocytosis and an analysis of red blood cell size distribution, these instruments do not classify specific red blood cell morphologic findings, nor do they perform platelet estimates. The study presented herein examines an alternative approach to performing DIFFs, i.e., '/2-minute qualitative scan, to assess the relative cell proportions and the presence of abnormal cell types. This approach is similar to Downloaded from by guest on 0 September 208

5 62 ARKIN ET AL. A.J.C.P.-May 987 that of other commonly performed cytologic procedures like vaginal cytologic examination, where one has similar goals. The concept of a scanning DIFF is not a great departure if one puts the evaluation of a peripheral blood smear in the context of the standard procedures for evaluating cytologic smears of other body fluids. The findings of this study indicate that the scanning technic is as reliable as the manual count in the detection of abnormal cell types such as blasts, immature granulocytes, and nucleated red blood cells. There were insufficient numbers of cases with basophilia or atypical lymphocytes to judge their detectability. Similarly, the ability of this procedure to detect monocytosis and eosinophilia was not clearly demonstrated because so many cases fell into the borderline areas. Although scanning DIFFs overlook many elevated band counts, they did seem to efficiently detect those at high levels. The presence of lymphocytopenia and, to a lesser extent, lymphocytosis was clearly more reliably detected by manual counting. In short, the scanning DIFF shows the greatest proficiency at finding abnormal cell types and a lesser ability to estimate the relative proportions of normal cells. Presumably the performance of the scanning technologists would improve with more experience and feedback as to their errors. This study did not include such feedback in order to eliminate the possibility of giving the operators access to the results of the manual counts and thereby influence their objectivity. The superior performance of one technologist in the estimation of lymphocyte counts does suggest that finer tuning of the other operators may be possible. Although this study does not convincingly prove that the scanning DIFF can equivalently replace the 00-cell manual count, there is an alternative approach to its use, that of combining the scanning DIFF with the "threepart" DIFF provided by flow cytometry. These two procedures complement each other and compensate the others' deficiencies. Specifically, the scanning DIFF is proficient in the detection of abnormal cell types, red blood cell evaluation, and platelet estimates, while being relatively deficient in estimating cell proportions. The threecellflowcytometric technic, on the other hand, is an excellent estimator of the proportions of the common cell types but does not identify abnormal cells or evaluate red blood cell morphologic characteristics. The combination of the three-part DIFF and the scanning DIFF should, therefore, provide a suitable alternative to the manual DIFF in many clinical laboratories. In this scheme, the scanning DIFF would replace the manual DIFF for those cases that require morphologic review in addition to a three-part differential. The criteria for which cases require morphologic review are not addressed by this study, but they certainly represent a high proportion of cases in the acute care environment given the high prevalence of abnormal white blood cell counts, thrombocytopenia, and anemia. References. Arkin CF, Sherry MA, Gough AG, Copeland BE: An automated leukocyte analyzer. Am J Clin Pathol 977; 67: Bull B, Korpman RA: Characterization of the WBC differential count. Blood Cells 980; 6:4-49. Cavill I, Jacobs A, Fisher J, desousa P: Sequential blood counts and their variation in normal subjects. Clin Lab Hematol 98; : Charache S, Nelson L, Keyser E, Metzger P: A clinical trial of threepart electronic differential white blood cell counts. Arch Intern Med 985; 45: Cornbleet J: Spurious results from automated hematology cell counters. Lab Med 98; 4: Cornbleet J, Kessinger S: Evaluation of Coulter S-Plus three-part differential in population with a high prevalence of abnormalities. Am J Clin Pathol 985; 84: Cox CJ, Habermann TM, Payne BA, Klee GG, Pierre RV: Evaluation of the Coulter Counter Model S-Plus IV. Am J Clin Pathol 985; 84: Diggs LW: Differential leukocyte counts in medical practice, Differential leukocyte counting. Edited by JA Koepke. Skokie, College of American Pathologists, 978, pp Dutcher TF, Benzel JE, Egan JJ, Hart DJ, Christoper EA: Evaluation of an automated differential leukocyte counting system, Parts I, II, III. Am J Clin Pathol 974; 62: England JM, Bain BJ: Total and differential leukocyte count. Br J Haematol 976;:-7. Greendyke RM, Kanter DR, DeBoover L, Savage L, VanGelder S: A comparison of differential white blood cell counts using manual technique and the Coulter S-Plus IV. Am J Clin Pathol 985; 84: Klee GG, O'Sullivan MB: Screening versus diagnostic differential leukocyte counts, Differential leukocyte counting. Edited by JA Koepke. Skokie, College of American Pathologists, 978, pp Nelson L, Charache S, Keyser E, Metzger P: Laboratory evaluation of the Coulter "three-part electronic differential." Am J Clin Pathol 985; 8: Nicheles TF, Fristensky R: The state of the art for leukocyte countinga clinician's view, Differential leukocyte counting. Edited by J A Koepke. Skokie, College of American Pathologists, 978, pp Shafer JA: Quality assurance for differential white cell counts performed manually, Differential leukocyte counting. Edited by JA Koepke. Skokie, College of American Pathologists, 978, pp Statland BE, Winkel P: Physiological variability of leukocytes in healthy subjects, Differential leukocyte counting. Edited by JA Koepke. Skokie, College of American Pathologists, 978, pp Talstad I: Problems in microscopic and automated cell differentiation of blood and cell suspension. Scand J Haematol 98; 26: VonAssendelft OW, McGrath C, Murphy RC, Schmidt RM: The differential distribution of leukocytes, Differential leukocyte counting. Edited by JA Koepke. Skokie, College of American Pathologists, 978, pp Wardlaw S, Rathbone R, McPhedran P: An instrument for white blood cell subclassincation. American Clinical Products Review 985; 2-26 Downloaded from by guest on 0 September 208