Spurious hyperglycaemia impairs automated leucocyte counting. A pilot study with two different haematological analysers

Introduction

Pre-analytical errors are a frequent problem that can compromise the quality of the total testing process and, along with analytical mistakes, can have a negative impact on the reliability of test results and patients’ safety¹. There is now consolidated evidence that the vast majority of pre-analytical problems emerge from mishandling or scarcely standardised procedures during collection, management and preparation of biological specimens². In particular, sample contamination by exogenous fluids (e.g., saline) or therapeutics (e.g., antibiotics, thrombolytic agents, glucose or potassium solutions) is a frequent source of problems in diagnostic testing, wherein test values may be variably biased by dilution or direct interference from the specific contaminating substance³. A paradigmatic example is contamination of biological samples with glucose-containing solutions. The administration of intravenous standard glucose solutions, typically 5% (i.e., 278 mmol/L) dextrose in water, is commonplace in clinical practice not only for the treatment of hypoglycaemia, but also for maintaining tissue hydration after acute disease or surgery, and as a means of parenteral nutrition⁴. Although it has been recommended that blood samples should be drawn from the opposite arm from which glucose solutions and other intravenous fluids are administered⁵, blood samples are frequently collected from infusion routes, in order to avoid a second venipuncture and save precious time in short stay or critical wards such as emergency departments and intensive care units. Even the best practice of the Clinical and Laboratory Standards Institute for collecting blood from intravenous lines, that is clearance of fluid before the sample is collected by discarding an adequate amount of fluid (i.e., typically 5 mL or 6 times the dead space volume)⁶, is frequently overlooked. Therefore, blood sample contamination is a relatively common occurrence and may lead to diagnostic errors and adverse consequences for patients’ health, including inappropriate therapeutic correction of spurious abnormalities (e.g., hypoglycaemia or hyperkalaemia), as well as unjustified transfusion of blood components in the case of spurious dilution of blood^7,8.

It has been previously shown that blood sample contamination by glucose solutions may generate a significant bias in clinical chemistry⁹ and coagulation testing¹⁰. Nevertheless, to the best of our knowledge, no information exists on the effect of spurious hyperglycaemia on leucocyte counts and differential. This is of paramount importance, since biased results may confound diagnostic reasoning and lead to inappropriate therapeutic decisions. The aim of this study was, therefore, to evaluate whether contamination of whole blood samples with standard glucose solutions (as may occur during normal venipuncture contaminated by exogenous glucose) may affect leucocyte counts measured by two widely used haematological analysers (ADVIA 2120 [Siemens Diagnostic Solutions, Milan, Italy] and XE-2100 [Dasit SpA, Cornaredo, Italy]), characterised by different analytical technologies.

Materials and methods

The study population consisted of 12 ostensibly healthy subjects recruited from the laboratory staff, for whom routine laboratory testing was requested. A total of 9 mL of blood was collected into three evacuated blood tubes containing K₂EDTA (13×75 mm, 3.0 mL BD Vacutainer^® Plus plastic whole blood tube containing 5.4 mg spray dried K₂EDTA [Becton Dickinson Italia SpA, Milan, Italy]). The autologous blood was pooled and then divided in four aliquots of 2.0 mL each. The first specimen was left untreated (i.e. without contamination), whereas scalar concentrations of a standard 5% glucose-containing solution (25 g of glucose monohydrate in 500 mL of water, 278 mOsm/L; Baxter SpA, Rome, Italy) were added to the remaining three aliquots, to achieve final glucose contaminations of 5%, 10% and 20%. The aliquots were left capped in upright position for exactly 1 hour at room temperature, and then analysed on an ADVIA 2120 and XE-2100^11,12. Both instruments were used properly, according to manufactures’ instructions. After the haematological analyses had been completed, all aliquots were centrifuged and plasma glucose was assayed by the reference hexokinase method, using the clinical chemistry analyser Beckman DxC (Beckman Coulter Inc., Brea, CA, USA).

The normality of the distribution of values was preliminarily verified by the Kolmogorov-Smirnov test and values were then reported as mean and standard error of the mean (SEM). Difference between the uncontaminated samples and the contaminated aliquots were assessed with Student’s paired t-test. Linearity was evaluated with linear regression analysis. Statistical analyses were performed using Analyse-it (Analyse-it Software Ltd., Leeds, UK). The subjects provided written consent to participation in the study, which was carried out in accordance with the Declaration of Helsinki and under the terms of all relevant local legislation.

Results

The main results of this study are shown in Table I. As intended, the concentration of plasma glucose increased progressively from the uncontaminated samples to the glucose-contaminated aliquots. As regards white blood cells (WBC), the total WBC, neutrophil and lymphocyte counts appeared to be significantly decreased starting from 5% glucose contamination with both analysers, the monocyte count from 10% glucose contamination with both analysers, and the basophil count from 20% contamination with both analysers. The eosinophil count was significantly decreased starting from 10% glucose contamination with ADVIA 2120 and from 20% glucose contamination with XE-2100. The large and unstained cell count was significantly decreased after 20% contamination.

Table I.

Effect of spurious contamination of diagnostic blood samples on total white blood cell (WBC), neutrophil (NEU), lymphocyte (LYM), monocyte (MONO), eosinophil (EOS), basophil (BASO) and large and unstained cell (LUC) counts (mean ± standard error of the mean).

	No contamination	Glucose solution contamination
		5%		10%		20%
	Value	Value	p	Value	p	Value	p
WBC×109/L
- ADVIA 2120	6.58±0.37	6.24±0.39	0.022	5.98±0.36	<0.001	5.21±0.32	<0.001
- XE-2100	6.56±0.40	6.23±0.40	<0.001	5.97±0.38	<0.001	5.54±0.35	<0.001
NEU×109/L
- ADVIA 2120	3.81±0.30	3.61±0.29	0.007	3.47±0.27	<0.001	3.01±0.25	<0.001
- XE-2100	3.66±0.30	3.47±0.29	<0.001	3.36±0.28	<0.001	3.09±0.25	<0.001
LYM ×109/L
- ADVIA 2120	2.22±0.13	2.10±0.13	0.049	2.02±0.12	<0.001	1.76±0.11	<0.001
- XE-2100	2.33±0.15	2.21±0.13	0.005	2.10±0.13	<0.001	1.98±0.13	<0.001
MONO×109/L
- ADVIA 2120	0.38±0.03	0.37±0.03	0.192	0.35±0.03	0.007	0.32±0.03	<0.001
- XE-2100	0.41±0.03	0.39±0.03	0.164	0.37±0.03	0.009	0.34±0.03	<0.001
EOS×109/L
- ADVIA 2120	0.14±0.02	0.13±0.02	0.408	0.12±0.01	0.045	0.11±0.01	0.001
- XE-2100	0.13±0.02	0.12±0.02	0.103	0.12±0.02	0.227	0.10±0.02	<0.001
BASO×109/L
- ADVIA 2120	0.03±0.01	0.03±0.01	0.319	0.03±0.01	0.107	0.02±0.01	0.005
- XE-2100	0.03±0.01	0.03±0.01	0.087	0.03±0.01	0.080	0.02±0.01	0.013
LUC×109/L
- ADVIA 2120	0.13±0.04	0.13±0.04	0.256	0.12±0.04	0.208	0.11±0.04	0.003
Glucose (mmol/L)	4.4±0.4	19.2±0.6	<0.001	33.2±1.0	<0.001	62.1±1.9	<0.0001

The significance of values is compared to the reference, uncontaminated aliquot.

When these changes were compared to the expected variations attributable to effect of dilution (Figures 1A and 1B), a different pattern was observed with the two haematological analysers. More specifically, the percentage decreases of WBC (r=1.00 and p=0.005), neutrophil (r=1.00 and p=0.002), lymphocyte (r=1.00 and p=0.002) and monocyte (r=0.99 and p=0.004) counts on ADVIA 2120 followed the expected trend precisely, whereas the decreases of eosinophil (r=0.84; p=0.070), basophil (r=0.80; p=0.103) and large and unstained cell (r=0.79; p=0.109) counts were non-linear and virtually unpredictable. On the XE-2100 analyser the percentage decreases of all WBC populations appeared to be linear up to 10% glucose contamination, whereas a substantial deviation from linearity was observed starting from aliquots with 20% glucose contamination (WBC: r=0.97 and p=0.057; neutrophils: r=0.97 and p=0.066; lymphocytes: r=0.95 and p=0.236; monocytes: r=0.97 and p=0.108; eosinophils: r=0.96 and p=0.156; basophils: r=0.98 and p=0.086). More specifically, all the leucocyte counts appeared to be overestimated by 3% to 5% compared to their theoretical values corrected for the dilution coefficient.

Figure 1A.

Percentage variation of white blood cell (WBC), neutrophil (NEU) and lymphocyte (LYM) counts in samples contaminated by a glucose standard solution.

The dotted line indicates the theoretical variation according to the dilution factor.

Figure 1B.

Percentage variation of monocyte (MONO), eosinophil (EOS) and basophil (BASO) counts in samples contaminated by a glucose standard solution.

The dotted line indicates the theoretical variation according to the dilution factor.

Discussion

Improper or inaccurate sample collection is a well-recognised source of problems in diagnostic testing¹³, including automated leucocyte counting¹⁴. Although current guidelines recommend that blood samples should not be collected through intravenous routes^5,6, daily practice is rather different. The placement of intravenous lines is commonplace in several healthcare settings, especially emergency departments, intensive care units and other clinical wards in which continuous infusion of fluids, nutrients or therapeutic agents is needed⁴. Accordingly, rather frequently, blood samples may be collected from infusion routes, with inadequate procedures for preventing contamination by intravenous fluids³. Although it was previously shown that spurious hyperglycaemia dramatically affects clinical chemistry and coagulation testing^9,10, no evidence has been provided that spurious contamination of blood samples with glucose may also generate a significant bias in leucocyte counting.

The results of this study clearly demonstrate that a 5% to 20% contamination of whole blood samples with a standard glucose solution (i.e., 25 g of glucose monohydrate in 500 mL of water) generates a bias in leucocyte enumeration, not only decreasing total WBC count, but also affecting the enumeration of most leucocyte subpopulations. Although we also observed that the bias was evident with two widely used haematological analysers employing different technology for automated leucocyte analysis, the trends were found to be globally different. More specifically, the changes recorded for WBC, neutrophil, lymphocyte and monocyte counts on the ADVIA 2120 were linear and consistent with the dilution factor, whereas those observed on XE-2100 showed a paradoxical trend, with significant deviation from linearity (i.e., 4% to 5% overestimation) at the highest glucose contamination (Figure 1). The trend for eosinophil and basophil counts was substantially similar to that of WBC, neutrophil and lymphocyte counts on XE-2100, with an overestimation between 3% and 4% in samples with 20% glucose contamination. The changes of eosinophil, basophil and large and unstained cell counts were less predictable on ADVIA 2120. Specifically, the pattern of variation did not follow a linear trend, with basophil count exhibiting the most paradoxical behaviour (i.e., almost unaffected up to 10% glucose contamination, with a sharp decrease afterward) (Figure 1); this may be attributable to the very low counts and the high coefficient of variation observed in this study which make it difficult to draw definitive conclusions about these cellular elements.

Regardless of the underlying causes of bias, which are probably a combination of glucose-induced effects on leucocyte biology and the different analytical techniques used for leucocyte enumeration and differentiation in ADVIA 2120 and XE-2100^11,12, the results of this study demonstrate that spurious hyperglycaemia (i.e., the presence of high glucose concentration in blood) is a source of bias in automated leucocyte counting. This may be of greatest significance in patients with hyperglycaemia and bacterial infections, in whom the total WBC count is probably the most important parameter, so that a spurious decrease may affect diagnostic reasoning.

Conclusions

The first and rather obvious conclusion that can be made from these findings is that contamination of diagnostic samples by glucose solutions may impair the reliability of leucocyte enumeration, thus jeopardising clinical decisions and therapeutic management. This has at least two important implications. One is the need to reinforce current recommendations to avoid drawing blood from intravenous lines^5,6 and, especially, to suppress test results from specimens with for which there is a high suspicion of spurious hyperglycaemia¹⁵. A second potential implication regards the impact of hyperglycaemia in vivo. The fact that a high glucose concentration in blood may interfere with accurate leucocyte enumeration with various haematological analysers, producing an often unpredictable bias, is a serious concern in longitudinal analysis of data, especially in patients at risk of acute hyperglycaemia. Specifically, an increase of glucose concentration up to 33 mmol/L (i.e., approximately 590 mg/dL) may be observed frequently in diabetics as well as in patients with acute illnesses, infections or surgery. According to our data, it cannot be excluded that a sudden variation of glucose homeostasis in the circulation may decrease the accuracy of leucocyte enumeration. It is also worth noting that blood is frequently collected through vascular devices in many healthcare settings, especially in emergency departments and intensive care units, and blood may, therefore, be contaminated by various amount of standard glucose solutions when an insufficient volume is discarded to clear the intravenous line. This possibility should be clearly acknowledged, to prevent spurious data from automated leucocyte counting leading to improper clinical and therapeutic decisions, especially in patients with haematological disorders.