Biological Variations of Hematologic Parameters Determined by UniCel DxH 800 Hematology Analyzer

The participants were 40 healthy volunteers of Chinese ethnicity (21 women and 19 men) with ages ranging from 20 to 40 years old. None of the women were menstruating. All participants maintained their normal lifestyles, including no excess of alcohol, tea, and tobacco consumption, and did not participate in strenuous exercise during the study period.

The blood samples were drawn in duplicate at 8:00 am, noon, and 4:00 pm each day for 3 consecutive days. The participants were in sitting position for at least 15 minutes before drawing. All samples were collected in EDTA anticoagulation tubes (BD Inc, Franklin Lakes, New Jersey) by a single experienced phlebotomist and analyzed within 2 hours after specimen collection.

All samples were analyzed using a single DxH 800 hematology analyzer (Beckman Coulter Inc, Brea, California). Before each batch sample analysis, instrument quality controls were performed using the same lots of Coulter S-CAL Calibrator (lot No. 112753780; Beckman Coulter), and Coulter 6C Cell Control with 3 levels at different concentrations (lot No. for level 1, 122755170; for level 2, 132757540; for level 3, 142755190; Beckman Coulter) to allow consistent determination during the course of the study. This study protocol was approved by the hospital ethics committee.

Data collected from the DxH 800 included the conventional parameters of red blood cells, reticulocytes, and platelets, as well as many newly described hematologic parameters, such as LHD, microcytic anemia factor, mean sphered cell volume, red cell size factor, immature reticulocyte fraction, mean reticulocyte volume, high–light-scatter reticulocytes, reticulocyte distribution width, and mean platelet volume. Immature reticulocyte fraction was determined by using a supravital stain (new methylene blue) to highlight cytoplasmic RNA and a new flow cell design to support multiple angles of light scatter measurements, enabling enhanced data acquisition to allow not only for measuring the reticulocytes from the entire red blood cell population, but also for identification of a subpopulation of the immature reticulocytes with high light scatter. The ratio of the immature reticulocytes to the total reticulocyte population is defined as the immature reticulocyte fraction. Mean platelet volume was determined using volume, conductivity, and light scatter technology by measuring direct current impedance. Other parameters, such as LHD, microcytic anemia factor, mean sphered cell volume, and red cell size factor, were mathematically calculated by the instrument.

Nested analysis of variance and coefficients of variation (within-subject [CV_I] and between-subject [CV_G]) were performed using SPSS software, version 10.0 (SPSS, Chicago, Illinois) and Microsoft (Redmond, Washington) Excel 2003. Analytic coefficient of variation (CV_A) was calculated from 10 independent tests using Coulter 6C Cell Control. Reference change values (RCVs) were calculated using the formula RCV = 2^1/2 * Z * (CV_A² + CV_I²)^1/2, where Z scores are 1.65, 1.96, and 2.58 for probabilities of 90%, 95% and 99%, respectively. Comparison between 2 means was performed by Student t test. A P value less than .05 was considered significant.

We initially studied within-subject (CV_I) and between-subject (CV_G) biological variations on hematologic parameters. As shown in Table 1, we achieved similar CV_I and CV_G for the conventional erythrocyte or platelet parameters compared with published data.^1,2  The analytic variations (CV_A) of DxH 800 for these parameters appeared smaller than or comparable to those found in previous studies.^1,2,5  The CV_I and CV_G for newly described erythrocyte parameters, such as microcytic anemia factor, mean sphered cell volume, or red cell size factor, were less than 5%, except for LHD, which showed greater intraindividual and interindividual variations (Table 1). On the other hand, CV_I and CV_G for most reticulocyte parameters were higher than those of erythrocytes or platelets, likely because of higher analytic imprecision for these parameters (Table 1). The index of individuality, calculated as the simple ratio of CV_I:CV_G,¹  for most parameters examined, was less than 0.5 (Table 1). A low index of individuality indicates that conventional reference values for these parameters may be of little utility, particularly when deciding whether changes observed in an individual are clinically significant.¹ 

The intraday and interday biological variations on hematologic parameters were next investigated. As shown in Table 2, intraday biological variations for most hematologic parameters examined were constant except for reticulocyte distribution width coefficient of variation and reticulocyte distribution width SD, which showed a statistically significant difference between 8:00 am and noon measurements. We do not have good explanations for these fluctuations. No significant interday biological variations of the hematologic parameters examined were observed (data not shown).

Monitoring hematologic parameters allows the detection of various physiologic or pathologic states when their values are increased or decreased longitudinally in relation to themselves. It is important to know the RCV or critical difference, which defines the percentage change that should be exceeded given the analytic and biological variations inherent to a particular test, in that there is a significant difference between the 2 consecutive measurements. Using the formula described in “Materials and Methods,” we calculated RCV for all hematologic parameters investigated, as shown in Table 3. The percentage changes that were significant (95%) or highly significant (99%) for most conventional red cell parameters were smaller than or compatible with previous data.^{1, 2}  The smaller RCV was likely due to the smaller CV_A of the current analyzer, as illustrated in Table 1. The RCV for new red cell parameters except for LHD was also smaller. On the other hand, the RCV for most reticulocyte parameters was larger (Table 3), likely because of greater CV_A and intraindividual variations (Table 1).

The implications of a decreased analytic variation in gaining greater confidence in the inherent biological variations in healthy individuals should be obvious, especially in using an individual's baseline values to compute that individual's personal reference interval. Using the newest Coulter hematology analyzer, the DxH800, we demonstrated that CV_I or intraindividual and CV_G or interindividual biological variations of most conventional hematologic parameters were fairly constant among the subjects and during the time points investigated. The CV_I and CV_G for most newly described erythrocyte parameters except for reticulocyte parameters were small, suggesting these parameters are less variable around the homeostatic setting point intraindividually and interindividually. The higher CV_I and CV_G for most reticulocyte parameters were likely due to higher analytic imprecision caused by random error (Table 1). This may be due to low quantity of the reticulocytes, as we usually see poor correlation for basophil measurements between 2 instruments. In addition, the index of individuality for the newly described hematologic parameters was low. A low index of individuality suggests that these parameters may have marked individuality and the conventional reference values for these parameters may be of little utility.¹ 

Our results are slightly different from the previously published observations, which showed hour-to-hour and subject-specific diurnal fluctuation.^5,6  For example, the study by Sennels et al⁵  demonstrated diurnal variation of hematology parameters. In this study, samples were collected every third hour through 24 hours, 9 time points in total, in contrast to 3 time points during an 8-hour interval in our study. The greatest changes for most parameters were observed around 9:00 pm to midnight. One of the explanations for these differences is that the current study uses a more advanced analyzer with likely improved precision, which may minimize the fluctuations due to analytic variations. Another possibility would be that because the study investigated the changes at only 3 daytime points each day for a 3-day period, some changes, such as seasonal variations, might not have been observed.^3,5  Interestingly, the biological variations during the similar time points (from 9:00 am–3:00 pm) were also small in that study⁵  compared with our study (from 8:00 am–4:00 pm). The numbers from two 9:00 am points were also similar, suggesting smaller interday variation.⁵ 

Using the data on CV_I and analytic precision, we have also established the RCV or critical difference for both conventional and newer hematologic parameters (Table 3). The RCV may be used as reference to determine the changes that occur in an individual's serial results before the change is significant, and to generate objective delta-check values for use in quality management.¹  For example, a difference between 2 consecutive analyses for LHD that is 42.4% or greater, or 55.8% or greater, will be considered as a significant or highly significant change, respectively (Table 3). These results may be flagged, which suggest that they may have failed the delta check. The delta-check failure may be due to real change in the patient's condition or errors associated with samples (either the first or the second). Therefore, there are considerable advantages to using RCV to alert laboratory staffs that serial results in an individual have changed significantly (95%) or highly significantly (99%).¹  Another potential clinical application is to use the RCV in everyday practice just as we frequently use conventional population-based reference values. For example, a recent study¹⁴  demonstrated that the median value of LHD for healthy subjects (n = 90) was 2.3, with a range of 0.9 to 4.1, which is similar to our mean values of LHD (2.89 ± 0.82, 2.84 ± 0.79, and 2.68 ± 0.63 at 8:00 am, noon, and 4:00 pm time points, respectively) (Table 2). However, the median value of LHD for the patients with iron-deficiency anemia (n = 110) was 22.3, with a range of 5.5 to 54, which was much greater than the RCV (>99%) from this study, indicating a highly significant difference in LHD between healthy individuals and the patients with iron-deficiency anemia.¹⁴  Lastly, RCV may vary with changes in analytic precision and bias and can be lowered by improving precision. Therefore, if possible the analytic variations should be eliminated or certainly minimized by careful attention to quality management practices.

These observations may be clinically important. Monitoring an individual's health often requires assessment of serial laboratory test results. Repeat results are seldom identical. Changes in laboratory values may be due to preanalytic variation, analytic imprecision, biological variation, or a change in the individual's health condition. Comparing CV_I and CV_G may allow us to decide the utility of traditional population-based reference ranges.¹  Data on biological variation may also be used for determining the number of samples needed to get an estimate of the homeostatic setting point within a certain percentage with a stated probability, and deciding the best way to report test results, the best sample to collect, and the test procedure of greatest potential use.¹  Lastly, documentation of biological variations for newly described hematologic parameters is an essential prerequisite in the development of any new application clinically.¹