Evaluation of Automated Reticulocyte Counts and Their Reliability in the Presence of Howell-Jolly Bodies

Transcription

1 HEMATOPATHOLOGY Evaluation of Automated Reticulocyte Counts and Their Reliability in the Presence of Howell-Jolly Bodies KAREN G. LOFSNESS, MS, MONICA L. KOHNKE, BS, AND NANCY A. GEIER, BS An automated reticulocyte procedure using a flow cytometer and the fluorescent dye thiazole orange was evaluated for clinical use. The mean reticulocyte count on 118 hematologically healthy adults was 1.56% (standard deviation SD] 0.54), with virtually no difference in percentage between women and men. The mean absolute values were 68.4 X 1O'/L (SD 24.6) and 75.7 X 10»/L (SD 27.2), respectively. When compared with the standard microscopic technique, the automated method showed excellent correlation (r = 0.98) and greatly improved precision (coefficient of variation CV] 4.1%) over the manual method (CV 22.8%). Preanalytic storage of blood samples at 4 C for up to 48 hours did not significantly affect results, nor did varying the incubation time of diluted samples from */i to 2 hours. In a group of patients with appreciable numbers of Howell-Jolly bodies, automated reticulocyte counts were spuriously elevated. The difference between the manual and automated counts on these patients approximated the percentage of Howell-Jolly bodies observed on their Wright-Giemsa stained blood smears. (Key words: Automated counting; Flow cytometry; Howell-Jolly bodies; Reticulocytes) Am J Clin Pathol 1994;101: Reticulocytes are immature red blood cells (RBC) that have recently left the bone marrow and still contain intracellular RNA. The effectiveness of erythropoiesis can be measured by staining blood with a supravital dye and determining the number of RBCs that contain this residual RNA. Reticulocyte counts provide clinically useful information, not only in the diagnosis and classification of anemias, but also in monitoring therapeutic response. The most widely used method for counting reticulocytes is a manual microscopic procedure. 1 Although it is inexpensive and relatively simple to perform, this method is labor intensive and imprecise. Individual differences in the preparation and staining of the sample, distributional variation on the smear, the limited number of cells actually counted, and interobserver subjectivity in identifying reticulocytes all contribute to the poor reproducibility of the manual procedure. Recently, automated reticulocyte methods, using flow cytometry and fluorescent dyes that bind RNA, have shown improved precision and cost effectiveness. 2 " 12 However, most of the currently used fluorescent dyes bind polynucleotides nonspecifically, and thus stain DNA as well as RNA.' 3 Nucleated cells, because of their high DNA content, have fluorescent activity that is many times greater than that of reticulocytes and can usually be excluded or "gated out" from the RBC population. Small lymphocytes, such as those seen in chronic lymphocytic leukemia, occasionally cause interference. Although the rela- From the Department of Laboratory Medicine and Pathology, University of Minnesota, Minneapolis. Minnesota. Manuscript received October 8, 1992; revision accepted February 1, Address correspondence to Ms. Lofsness: Box 198 UMHC, University of Minnesota, Minneapolis, MN tive concentration of nucleated RBCs and erythrocytes containing malarial parasites or Howell-Jolly bodies is usually very low, these inclusions also fluoresce and have the potential to spuriously elevate an automated reticulocyte count. The objectives of this study were to evaluate the Coulter Epics Profile II flow cytometer (Coulter Electronics, Hialeah, FL) for reticulocyte counting in a routine hematology laboratory, to use this instrument to establish reference values, and to study the effect of significant numbers of Howell-Jolly bodies on automated reticulocyte counts. MATERIALS AND METHODS Automated Reticulocyte Counting Automated reticulocyte counts were performed on a Coulter Epics Profile II flow cytometer, using a modification of the method of Lee and coworkers. 13 A 1:10,000 working solution of thiazole orange was prepared fresh daily by diluting 5 jul stock solution (1 mg/ml thiazole orange in methanol) in 50 ml Isoton II (Coulter Electronics, Hialeah, FL); the solution was then stored in a brown bottle. For each blood specimen, two dilutions (labeled "blank" and "thiazole orange") were prepared in 10 X 75 mm capped polypropylene tubes. Each blank tube contained 5 nl blood diluted in 2.0 ml Isoton II, and each thiazole orange tube contained 5 nl blood diluted in 2.0 ml working thiazole orange solution. Both tubes were mixed well and incubated in the dark at room temperature for more than 30, but less than 120, minutes before analysis. The tubes were remixed two to three times during this incubation period, and again immediately before counting. The following instrument settings were used for reticulocyte counting. The laser was set at 15 m W. The green fluorescence 85

2 86 HEMATOPATHOLOGY of 50,000 RBC (measured with 525 nm band pass and 550 nm short pass dichroic filters) was collected on a 256 channel 4 generation log analyzer. Forward scatter and log side scatter gates were used to encompass the RBC in a bitmap. The high voltage and gain were set so that 99.9% of thefluorescenceof an unstained sample fell between channels 0 and 50. The sample flow rate was set at 50 ^L/min ( cells/sec). The instrument start-up procedure recommended by the manufacturer was performed daily, and any necessary adjustments were made before the analysis of the blood samples. For each determination, the blank sample was analyzed first, and results were checked to ensure that the background count was approximately 50 events, the reported percentage was 0.1%, and the histogram did not appear abnormal. The range bar cursor was adjusted if necessary, and the thiazole orange sample was analyzed using the same settings as the blank (Figure 1). To correct for the background count, 0.1% was subtracted from the reticulocyte percentage obtained on the thiazole orange sample. Manual Reticulocyte Counting Equal amounts of blood and new methylene blue were combined and incubated at room temperature for 5-15 minutes. The dilutions were remixed, and three wedge smears were prepared from each sample. The number of reticulocytes per 500 erythrocytes was determined microscopically by each of two technologists on separate slides (1000 cells total). If the number of reticulocytes did not agree withinfivecells, a third technologist counted 500 cells on the third slide, and results were averaged. A reticulocyte was defined as an RBC containing at least two granules of reticulum. Specimen Collection and Method Evaluation Venous blood samples were collected in Na 2 EDTA (1.5 mg/ ml), and reticulocyte counts were determined as described. Complete blood counts (CBC) were performed by either a Coulter S-Plus IV or STKR (Coulter Electronics, Hialeah, FL), and blood smears were prepared and stained by a Wright- Giemsa method. Results were analyzed statistically using Statview II software (Abacus Concepts, Berkeley, CA). Healthy Subjects Reference values were determined by performing automated reticulocyte counts on blood samples from 118 healthy nonhospitalized subjects, aged 16 to 80 years. All 67 women and 51 men were judged to be hematologically healthy on the basis of their automated CBC. Absolute reticulocyte counts were calculated by multiplying the relative count (in percent) by the RBC count determined on the same blood sample. Patient Samples For correlation studies, 215 random blood samples from hospitalized patients, on whom reticulocyte counts had been ordered, were analyzed by both the manual and automated methods. Reproducibility was evaluated by performing three replicate analyses on 10 patient samples by both the manual and automated methods. To determine the effect of preanalytic specimen storage time on automated reticulocyte results, 15 patient blood samples were first analyzed within 4 hours of B MR MAX CODNT LFL1 PBRCBHT HEAR SD I gpcv MIR MAI 1 O.ÍOTTÍÍX LFL1 COURT PEBCEHT 1527 rr HEAR TUT SO ÜTÍ7T \ BPCJ FIG. 1. Histograms of blank (A) and thiazole orange (B) dilutions. Cell number is shown on the vertical axis (count), andfluorescentintensity, plotted on a four-generation log scale, is shown on the horizontal axis (LFL1). The range bar cursor [1] is set so that the event count on the blank is around 50, and the percentage of cells showingfluorescenceis 0.1%. The thiazole orange sample is analyzed using the same settings, and 0.1 % is subtracted from the result. This patient's reticulocyte count was 4.9%. collection, then stored at 4 C and reanalyzed at 24 and 48 hours. The stability of the fluorescence in the blood-working thiazole orange dilution was measured by comparing the results on 10 patient samples after 'A, 1, and 2 hours incubation time. Forty-one blood samples from 19 patients, on which Howell- Jolly bodies had been observed microscopically, were used to study the potential interference of these inclusions on automated reticulocyte counts. The automated and manual reticulocyte percentages on these patients were obtained by the methods described above. The percentage of Howell-Jolly bodies was determined independently on a Wright-Giemsa stained AJ.C.P.-January 1994

3 TABLE 1. AUTOMATED RETICULOCYTE REFERENCE VALUES LOFSNESS, KOHNKE, AND GEIER 87 Automated Reticulocyte Counts Mean ±2SD Mean (XIO'/L) ±2SD 118 adults (ages yr) women men s 5 blood smear of the same sample by having two technologists count the number of RBCs containing Howell-Jolly bodies per 500 cells (1000 cells total) in the manner of a manual reticulocyte count. RESULTS Reference Values and Correlation Studies Reticulocyte results for 118 hematologically healthy adults are shown in Table 1. The mean reticulocyte percentage was 1.56%, with a standard deviation (SD) of 0.54%. The average absolute reticulocyte count was 71.5 X 10 9 /L (SD 25.9). For the 67 women, the means were 1.56% (SD 0.55) and 68.4 X 10 9 /L (SD 24.6), with observed ranges of 0.4% to 3.4% and 17 to 136 X 10 9 /L, respectively. The 51 men had means of 1.55% (SD 0.54) and 75.7 X 10 9 /L (SD 27.2), and observed ranges of 0.6% to 2.9% and 30 to 144 X 10 9 /L. Figure 2 shows the frequency distribution of reticulocyte percentages. Because this distribution was not Gaussian, the 95% reference interval was also determined by percentile rank. When the lowest 2.5% and highest 2.5% of the values were excluded, the range was 0.6% to 2.7%. Figure 3 illustrates the frequency distribution of absolute reticulocyte values for these 118 subjects. The correlation between manual and automated reticulocyte results on 215 patient samples with manual percentages ranging from 0.1% to 28.2% is shown as Figure 4. By linear regression analysis, the coefficient of correlation (r) was The coefficient of determination (i 2 ) was 0.95, and y = 0.92x Relic absolute « FIG. 3. Frequency distribution of absolute reticulocyte values (X 10 9 /L) by flow cytometer on 118 healthy adults (67 women, 51 men). Reproducibility and Stability The 10 patient samples used to evaluate reproducibility had automated counts ranging from 1.2% to 9.3%. For the automated method, the average coefficient of variation (CV) was 4.1%, with a range of 0.9% to 7.7%. By the manual method, the same samples showed a mean CV of 22.8% (range 11.1% to 36.1%). The 15 random bloods used to study the effect of preanalytic specimen storage time had initial automated reticulocyte values ranging from 1.0% to 10.3%, with a mean of 3.3%. After 24 hours of storage at 4 C, the results ranged from 0.8% to 10.3%, with a mean of 3.4%. Only one 24-hour sample varied as much as 0.5% from the original value. After 48 hours at 4 C, the mean was still 3.4% and the range was 0.9% to 10.3%. One 48-hour sample was 0.7% higher than the initial result; the rest were all within ± 0.5%. These differences were not statistically significant by paired /-tests. The 10 samples that were analyzed after l h, 1, and 2 hours incubation in the working thiazole orange solution had initial automated reticulocyte percentages of 1.1% to 6.8%, with a mean of 2.4%. After 1 hour of incubation, they ranged from 1.0% to 6.5% (mean 2.4%), and after 2 hours the range was 1.0% to 6.4% (mean 2.4%). One 2-hour dilution was 0.4% lower than the initial value; the rest were all within ± 0.2%. These differences were not statistically significant by paired i-tests. y Six *.17, r > 48, r-aqu«i*d m M (n 21S) ,w,..,VK.i...v..., Relic % FIG. 2. Frequency distribution of reticulocyte percentages byflowcytometer on 118 healthy adults. FIG. 4. Flow cytometer versus manual method, reticulocyte percentages. Scattergram and regression coefficients. Vol. 101 No. I

4 88 HEMATOPATHOLOO T TABLE 2. EFFECT OF HOWELL-JOLLY BODIES Patient No n HJ (%) Manual reticulocyte (%) HJ + manual reticulocyte (%) Automated reticulocyte (%) l OS "> ? ? I HJ = Howell-Jolly bodies. Effect of Howell-Jolly Bodies Forty-one samples in which Howell-Jolly bodies were easily found on the Wright-Giemsa smear were selected for reticulocyte analysis by both the manual and automated methods. These samples had manual reticulocyte counts ranging from 0.2% to 19.0%, and automated counts from 0.7% to 24.3%. In every case, the automated percentage was higher than the manual. The percentage of Howell-Jolly bodies was determined microscopically and ranged from 0.2 to 8.8%. When this percentage was added to the manual reticulocyte count, the result approximated the automated count in most cases. Table 2 shows data from the initial samples of the 19 patients with increased Howell-Jolly bodies. The flow cytometer histogram patterns on these samples with Howell-Jolly bodies did not appear to show any unusual characteristics (Fig. 5). When the thiazole orange dilutions from one of the Howell-Jolly samples was examined by fluorescent microscopy, the Howell-Jolly bodies appeared as bright green discrete intracellular dots, whereas the reticulocytes were diffusely fluorescent. DISCUSSION The clinical utility of the reticulocyte count is compromised by the fact that, in most laboratories, it is still determined microscopically. Although other routine hematologic counts are NIK COOHT LFL1 PERCENT HEAR SD \ HPCV FIG. 5. Histogram of thiazole orange dilution on a patient with an automated reticulocyte count of 5.4%, a manual count of 2.2%, and a Howell-Jolly body count of 2.5%. now performed by complex instruments that give highly accurate and reproducible results, the standard method for counting reticulocytes has not changed for many years and remains imprecise and unreliable. 14 ' 15 The recent development of practical techniques usingflowcytometry and fluorescent dyes will allow many clinical laboratories to automate reticulocyte counting and bring it to an acceptable level of reliability. On the basis of an internal study performed in 1966, the reference range for reticulocytes in our laboratory was considered to be 0.5% to 3.3% for women and 0.5% to 2.1% for men. As a result of the present study, the range has been adjusted to 0.5% to 2.7% for both sexes. Although we found virtually no difference between the reticulocyte percentages for women and men, reports in the literature vary on this point. Earlier studies by both manual 16 and automated 4 methods describe significantly higher reticulocyte percentages in women than men. However, other authors report no appreciable difference by flow cytometry, 2 ' 3 ' 7 and one study of 265 adults actually found a slightly lower mean percentage in women (0.92%) than in men (1.07%)." Because method comparison for reticulocyte counting necessitates using the admittedly imprecise manual technique as the reference by which to judge a new automated system, it is somewhat surprising to find a high degree of correlation between the two methods, especially in the higher clinical ranges. Our correlation and reproducibility data were similar to those reported by others, 4-6 ' 7 ' 10 ' 12 and showed that the automated method agreed closely with the manual, and was far more precise. The greatly improved precision obtained withflowcytometry reflects not only the much larger number of cells counted (~ 50,000 versus 1000 with the microscope), but also the elimination of subjectivity on the part of the observer. In a preliminary report from the CAP reticulocyte project, Savage and coworkers 15 recommend that the r value for an automated reticulocyte system should be greater than 0.95, with a y intercept of 0.25 or less when correlated with manual reticulocyte results. They also state that the system CV should be 15% or less. The automated system evaluated in this study, with r equal to 0.98, a y intercept of 0.17, and a CV of 4.1%, easily meets these criteria. The stability of this system was not affected by specimen storage time, either before analysis or following dilution in thiazole orange. Although some studies of sample storage reported a progressive fall in reticulocytes after blood was stored more than 24 hours at 4 C, 4-9 we found no significant differences up to 48 hours. Because refrigerated blood may be held as long as 2 days without appreciable changes in reticulocyte values, routine samples can be batched and run every other day, with specimens requiring immediate attention being performed by the manual method. Selected samples can also be stored at 4 C and reanalyzed with the next batch as a quality control check of system reproducibility. Because the thiazole A.J.C.P.-January 1994

5 LOFSNESS, KOHNKE, AND GEIER 89 Automated Reticulocyte Counts LFL1 HIN HAX COUNT PERCENT HEAD SD % HPCV T 27T TTJOT X5IT FIG. 6. Histogram of thiazole orange dilution on patient with chronic lymphocytic leukemia. The second peak caused by the WBC (200 X 10 9 /L) is evident. The initial range bar cursor setting [1] included some of the WBC, and gave a reticulocyte value of 2.6%. Range bar cursor setting [2] was extended to include all of the WBC, whereas setting [3] excluded the WBC peak and gave afinalresult of 2.2%. orange incubation timing is somewhatflexible,individual samples do not have to be analyzed in any specific order. Samples are easily added or repeated, and batches may be even interrupted for a reasonable period of time. Thiazole orange and the otherfluorescentdyes used in automated reticulocyte counting bind to DNA as well as RNA, thus nuclear material can potentially cause interference. Most leukocytes are so large and highly fluorescent that they are easily excluded from the RBC population being analyzed. Occasionally, it is difficult to "gate out" the small lymphocytes seen in chronic lymphocytic leukemia. However, the second peak they produce is evident on the histogram, and adjustment of the bitmap and the upper range bar cursor will usually provide acceptable results (Fig. 6). Large numbers of nucleated RBC would probably require similar adjustments. Erythrocytes containing smaller inclusions of DNA, such as malarial parasites or Howell-Jolly bodies, are more similar to reticulocytes in both size and fluorescent activity. When present in appreciable numbers, these RBC inclusions have the potential to elevate automated reticulocyte counts incorrectly. A recently published method for quantitating malarial parasites in in vitro culture systems actually uses flow cytometry with thiazole orange as the binding dye. 17 Some studies evaluating automated reticulocyte counting mention the possibility of spurious results in samples containing Howell-Jolly bodies. 78 However, others state that either Howell-Jolly bodies do not interfere, 2 ' 10 '" or their concentration is generally so low that they have no significant effect. 61 ' 8 Our investigation of the interference caused by Howell-Jolly bodies was prompted by a patient sample that had a manual reticulocyte count of 1.8% and an automated count of 6.0%. Howell-Jolly bodies were noted on the new methylene blue preparation, and examination of the patient's Wright-Giemsa stained blood smear showed several Howell-Jolly bodies in everyfield.by counting the number of Howell-Jolly bodies per 1000 RBC in the manner of a manual reticulocyte count, 4.1% Howell-Jolly bodies were found on this blood smear. Because the total number of microscopically determined reticulocytes and Howell-Jolly bodies so closely approximated the automated reticulocyte count, it was decided to select samples with noticeable numbers of Howell-Jolly bodies deliberately to see if this pattern persisted. Although this method of determining the percentage of Howell-Jolly bodies is admittedly as imprecise as the manual reticulocyte method itself, in every case the automated count was demonstrated to be higher than the manual, and the numeric difference between the two was somewhat close to the Howell-Jolly percentage. Howell-Jolly bodies are nuclear remnants that may be seen in erythrocytes following splenectomy. They may also be seen in hemolytic anemia, megaloblastic anemia, and hyposplenic states. 19 The question might well be raised as to the prevalence of samples containing appreciable numbers of Howell-Jolly bodies in a patient population. The samples in this study were from 19 patients who were seen in our 550-bed tertiary care hospital over a span of 4 weeks. Because such samples are not rare we recommend screening Wright-Giemsa blood smears on all samples submitted for reticulocyte counts, and performing manual counts on those that have more than one Howell-Jolly body per oil immersion (X 1000) field. Automated reticulocyte counting, which has consistently proven superior to the manual technique, should eventually replace the manual technique as the method of choice. In our experience, agreement between the two methods has been good, except for the occasional sample containing Howell-Jolly bodies or otherfluorescinginterferences. Errors associated with these factors can be minimized by carefully monitoring histograms and screening Wright-Giemsa smears for RBC inclusions. Although automated reticulocyte counting may not be feasible for smaller laboratories with low test volume, it should be a practical alternative for many clinical laboratories, particularly those that have already acquired aflowcytometer to perform immunologic marker studies. REFERENCES 1. National Committee for Clinical Laboratory Standards. Method for reticulocyte counting: Proposed standard. Villanova, Pa: NCCLS, 1985, Publication H16-P. 2. Batjer JD, Lin LI, Fritsma GA. Stability and reproducibility of reticulocyte analysis by flow cytometry. Hematology monograph. McGaw Park, 111: Baxter Diagnostics, 1991, pp Bowen D, Bentley N, Hoy T, Cavill I. Comparison of a modified thiazole orange technique with a fully automated analyser for reticulocyte counting. J din Pathol 1991;44: Carter JM, McSweeney PA, Wakem PJ, Nemet AM: Counting reticulocytes by flow cytometry: use of thiazole orange. Clin Lab Haemal 1989; 11: Corash L, Rheinschmidt M, Lieu S, Meers P, Brew E. Enumeration of reticulocytes usingfluorescence-activatedflowcytometry. Pathol Immunopathol Res 1988;7: Davis BH, Bigelow NC. Flow cytometric reticulocyte quantification provides clinically useful reticulocyte maturity index. Arch Pathol Lab Med 1989; 113: Ferguson DJ, Lee S, Gordon PA. Evaluation of reticulocyte counts by flow cytometry in a routine laboratory. Am J Hematol 1990;33: Vol. 101-No. 1

6 90 HEMATOPATHOLOGY 8. Hackney JR, Cembrowski GS, Prystowsky MB, Kant JA. Automated reticulocyte counting by image analysis andflowcytometry. LabMed 1989;20: Kojima K, Niri M, Setoguchi K, Tsuda I, Tatsumi N. An automated optoelectric reticulocyte counter. Am J Clin Palhol 1989:92: Nobes PR, Carter AB. Reticulocyte counting usingflowcytometry. J Clin Pathol 1990;43: Tatsumi N, Tsuda I, Kojima K, Niri M, Setoguchi K. An automated reticulocyte counting method: Preliminary observations. MedLabSci 1989;46: Tichelli A, Gratwohl A, Driessen A, et al. Evaluation of the Sysmex R-1000: An automated reticulocyte analyzer. Am J Clin Pathol 1990;93: Lee LG, Chen C, Chu LA. Thiazole orange: A new dye for reticulocyte analysis. Cylometry 1986; 7: Peebles DA, Hochberg A, Clarke TD. Analysis of manual reticulocyte counting. Am J Clin Pathol 1981; 76: Savage RA, Skoog DP, Rabinovitch A. Analytic inaccuracy and imprecision in reticulocyte counting: A preliminary report from the College of American Pathologists reticulocyte project. Blood Cells 1985:11: Deiss A, Kurth D. Circulating reticulocytes in normal adults as determined by the new methylene blue method. Am J Clin Pathol 1970:53: Makler MT, Lee LG, Recktenwald D. Thiazole orange: A new dye for Plasmodium species analysis. Cytometry 1987; 8: Davis BH, Bigelow NC. Clinicalflowcytometric reticulocyte analysis. Pathobiology 1990; 58: Bull BS, Breton-Gorius J, Beutler E. Morphology of the erythron. In: Williams WJ, Beutler E, Erslev A, Lichtman MA, eds. Hematology. 4th ed. New York, NY: McGraw-Hill, 1990, p A.J.C.P.-January 1994