The study population consisted of ten adult and ostensibly healthy volunteers recruited from the laboratory staff (five women, mean age 37.5±0.8 years and five males, mean age 35.0±7.4 years). All subjects were Caucasian, had no diabetes mellitus, hypertension and had not taken any medication for one month before the study. Six venous blood samples from each subject were collected in K3-ethylenediaminetetraacetic acid (K3-EDTA) tubes (Becton Dickinson, Franklin Lakes, NJ). All samples were analyzed immediately after venipuncture (i.e., within 30min). The analysis of the impact of different storage temperatures was then carried out by storing three blood tubes from each individual at room temperature (RT) and three blood tubes were divided in six aliquots and stored (refrigerated) at 4°C. Repeated measures were then performed on each sample after 2h, 4h, 6h, 8h, 24h, 36h and 48h of storage. An additional study was performed using 40 routine samples with abnormal values, that is, containing at least one abnormality of hemoglobin (Hb), platelet (PLT) or white blood cell (WBC) counts or morphological alterations (i.e., at least one morphological flag for WBC). Count abnormalities included Hb lower than 70g/dL, PLT lower than 100×109/L or higher than 400×109/L; WBC lower than 1.00×109/L or higher than 12.00×109/L. Hematological testing was performed immediately upon arrival in the laboratory (i.e., within 30min) and then each sample was divided into 8 aliquots, four were stored at 4°C and four were stored at RT. The tests were repeated after 4h, 8h, 24h 36h and 48h of storage. All measures (i.e., the baseline and the repeated analyses) were performed in duplicate and the final value was expressed as the mean of the two analyses at each time point.

The following parameters were assessed to check blood sample stability: extended complete blood count (CBC) profile parameters, including all basic CBC parameters [RBC, Hb, hematocrit (HT)], mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), RDW-SD, RDW-CV and NRBC. Moreover, the extended differential counts (DIFF) (including WBC, NE, LY, MO, EO, BA and HFC), PLT profile parameters (including PLT, PCT, MPV, PDW and P-LCR) and the RET profile (including RET, IRF, LFR, MFR and HFR) were evaluated.

The measurements at the different time points were concomitantly performed with both the XN-9000 and BC-6800 analyzers. The mean analytical characteristics of the two analyzers are summarized in Table 1. Briefly, the XN-9000 and the BC-6000 analyzers perform a 5-part DIFF, RET count, NRBC count, with flags appearing in the presence of abnormal results.7,14,15 Both analyzers use a combination of flow cytometry and fluorescence with lysing buffers for leukocyte DIFF and identification of abnormal cells. A separate channel for NRBC assessment is also available in the BC-6800.

The between-run imprecision of both the XN-9000 and BC-6800 was evaluated according to the CLSI document EP5-A3,16 by analysis in duplicate of three levels (i.e., level 1, 2 and 3) of control materials (XN-CHECK; Streck Laboratories Inc., Omaha, NE, USA and BC-6D, BC-BC-RET and NRBC; Shenzhen Mindray Bio-Medical Electronics, Shenzhen, China) for 40 consecutive working days. The study was carried out in accord with the Declaration of Helsinki, under the terms of all relevant local legislation and with prior approval of the Local Ethics Committee.

The significance of the difference of the parameters obtained in paired samples measured with the two analyzers was estimated according to the Steel–Dwass–Critchlow–Fligner test, with assessment by the Hodges–Lehmann location shift for multiple comparisons of means and medians between different groups, after verification of value distribution by the Shapiro–Wilk test. Statistical significance was set for p-values <0.05. The results were then reported as delta variations from baseline analysis immediately after collection, as ΔX (TX−T0), where “X” is the different timing and “0” is the baseline result. Percentage variations from the baseline result (T0) in samples with statistically significant differences were then analyzed using Bland–Altman plots (Bias%) and compared with the current quality specifications for optimal bias (OP-Bias%),17 that is calculated using intra-individual biological variability (CVi) and inter-individual biological variability (CVg) following the equation: OP-Bias%=0.250 (CVi+CVg)1/2. Bias% was also compared with the reference change values or critical difference (CD)18,19 when these indices were available. The CD percentage is the highest relative difference between two consecutive measurements, that, at a chosen level of probability, may still be due to the combined effect of the analytical (Va) and biological (Vi) variations. It is given by the following equation: CD%=K×(Va2+Vi2)1/2, where K depends on the chosen probability. The comparison between Bias% and CD% was performed only for those parameters exhibiting a statistically significant difference between Bias% and OP-Bias% throughout the study period. The statistical analysis was performed using Analyse-it software version 3.90.1 (Analyse-it Software Ltd.; Leeds, UK).

The median values obtained at baseline (i.e., T0) in the normal sample group did not significantly differ between the two analyzers for all the parameters tested, except for MCH and MCHC (Table 2). In this group of normal samples, the values of WBC, RBC, HB, MCH and NRBC were found to be stable up to 48h at RT and 4°C using both analyzers. Conversely, the HT values displayed a statistically significant increase 48h after collection in samples stored at RT (but not in those stored at 4°C) when measured with both analyzers. The comparison of the Bias% at different time points (TX) with the OP-Bias% showed that HT is stable at room temperature for up to 8h using both the XN-9000 (Bias%: 0.9) and BC-6800 (Bias%: 1.9), whereas the Bias% for HT in samples stored at RT remained lower than the CD% for up to 24h. The same analysis in samples stored at 4°C showed good stability for up to 24h after collection when compared with the OP-Bias%, whereas the Bias% always remained lower than the CD% throughout the study period (i.e., up to 48h) using both analyzers.

The values of MCV and RDW-SD displayed significant differences after 24h of storage at RT, whereas the results remained substantially unchanged for up to 48h after collection in samples stored a 4°C. Interestingly, the Bias% of MCV exceeded the OP-Bias% after 2h of storage at RT with the XN-9000 and after 8h of storage at RT with the BC-6800, respectively. The Bias% of MCV was lower than the relative CD% for up to 8h of storage using both analyzers at RT. At variance, the Bias% always remained lower than CD% throughout the study period in samples stored at 4°C. The RDW-CV exhibited significant variations after 24h from collection using both analyzers at RT, whereas significant differences were observed after 24h of storage in samples stored at 4°C using the XN-9000 but not with the BC-6800 (Table 2). When compared with the OP-Bias%, RDW-CV values were found to be stable for up to 8h at both RT and 4°C using the XN-9000, and for up to 2h at RT with the BC-6800. The Bias% of RDW-CV was always lower than the OP-Bias% for up to 48h of storage using the BC-6800.

At variance with previous parameters, the values of MCHC displayed a specific and instrument-dependent variation. More specifically, significant differences were observed after 4h of storage at RT and 2h of storage at 4°C with the XN-9000. Accordingly, the Bias% exceeded the OP-Bias% at 2h of storage at both temperatures, whereas the Bias% did not exceed the CD% for up to 8h of storage at RT and for up to 48h of storage at 4°C. As regards MCHC values obtained with the BC-6800, significant differences were found after 24h of storage at RT and at the 48h time point after storage at 4°C. The OP-Bias% was exceeded after 8h of storage at RT and 24h of storage at 4°C, whereas the Bias% remained lower than the CD% for up to 8h of storage at RT and for up to 48h of storage at 4°C (Table 2).

All these parameters appeared to be substantially stable for up to 8h at both RT and 4°C in the abnormal samples group using both analyzers. The most relevant exceptions are summarized in Table 3. Specifically, the NRBC measured with the XN-9000 showed significant variations 8h after collection in samples stored at RT and 4h after collection in those stored at 4°C, whereas significant variations of NRBC measured with the BC-6800 could only be observed after 8h of storage at 4°C. The comparison between the Bias% and OP-Bias% for the MCH and RDW-CV measured with the BC-6800 showed a significant variation after 4h of storage at 4°C. As regards the XN-9000, only the Bias% for MCHC increased over the OP-Bias% after 4h of storage at RT. Interestingly, the Bias% was found to be always lower than the CD% for all parameters with both analyzers for up to 8h of storage at both temperatures.

The baseline values of the different RET parameters were found to be always different on comparing the measurements of the two hematological analyzers (Table 2). In the group of normal blood samples, RET and percentage of HFR were found to be stable for up to 48h using both analyzers and at both temperatures. The Bias% of RET was found to be higher than the OP-Bias% after 36h of storage at both temperatures. The percentages of IRF, LFR and MFR were found to be higher than the T0 values after 24h of storage at RT with the XN-9000 and after 36h of storage at RT with the BC-6800 (Table 4). In the group of abnormal blood samples all the RET parameters were found to be stable for up to 8h using both analyzers and at both temperatures (Table 3).

The baseline values of Leukocytes and DIFF counts did not exhibit statistically significant variations throughout the study period in the subgroup of normal samples, using both analyzers and at both temperatures, with the only exception of the MO measured with the BC-6800 at the 48-h time point (Table 4). Accordingly, the Bias% of MO increased over the OP-Bias% after 36–48h of storage at RT and after 24h of storage at 4°C. Importantly, the Bias% was found to be always lower than the CD% throughout the study period, using both analyzers and at both temperatures.

In the abnormal samples group the various parameters were also found to be stable up to 8h of storage using both analyzers and at both temperatures (Table 5). The OP-Bias% was exceeded after 4h for NE, LY and MO at 4°C, and EO at RT using the BC-6800. The Bias% of EO also exceeded the OP-Bias% after 8h of storage at 4°C using the XN-9000 (Table 5).

With the exception of PLT and PCT, the baseline values of all the PLT parameters were found to be significantly different between the two analyzers in the group of normal blood samples (Table 4). In this group of specimens, PLT and PCT parameters were found to be stable for up to 48h at both RT and 4°C using both analyzers (Table 4). The remaining parameters (MPV, PDW, PCT and P-LCR) showed instrument-dependent variations.

The Bias% of the PLT count measured with the XN-9000 was found to be higher than the OP-Bias% after 2h of storage at 4°C, but remained always lower than the CD% for up to 48h of storage at this temperature. The values of MPV measured with the XN-9000 significantly increased after 24h of storage at RT and at 48h of storage at 4°C. However, the variation of the MPV was found to be higher than the OP-Bias% starting from 2h of storage at RT and 8h of storage at 4°C (Table 4). Importantly, the Bias% of the MPV never exceeded the CD% throughout the 48h of storage. As regards the BC-6800, the MPV values were significantly increased after 4h of storage at RT and 24h of storage at 4°C. A Bias% larger than the OP-Bias% was observed after 2h of storage at RT, although it remained lower than the CD% throughout the 24h of storage at RT (Table 4). The values of the PDW measured with both analyzers were found to be significantly different after 24h of storage at RT, whereas significant differences in samples stored at 4°C were observed after 24h of storage with the BC-6800 and after 48h of storage with the XN-9000 (Table 4).

In the group of abnormal samples, the PLT parameters were found to be stable for up to 8h using both analyzers at both temperatures (Table 5). Nevertheless, the Bias% of the MPV measured with the BC-6800 was found to be higher than the OP-Bias% after 2h of storage at both temperatures, whereas the Bias% of the MPV measured with the XN-9000 exceeded the OP-Bias% after 4h of storage at 4°C. In no case, however, the Bias% was found to be higher than the CD% in up to 8h of storage at both temperatures (Table 5).