Evaluation of the validity of a rapid method for measuring high and low haemoglobin levels in whole blood donors

Abstract

Background

Haemoglobin screening methods need to be highly sensitive to detect both low and high haemoglobin levels and avoid unnecessary rejection of potential blood donors. The aim of this study was to evaluate the accuracy of measurements by HemoCue in blood donors.

Materials and methods

Three hundred and fourteen randomly selected, prospective blood donors were studied. Single fingerstick blood samples were obtained to determine the donors' haemoglobin levels by HemoCue, while venous blood samples were drawn for measurement of the haemoglobin level by both HemoCue and an automated haematology analyser as the reference method. The sensitivity, specificity, predictive values and correlation between the reference method and HemoCue were assessed. Cases with a haemoglobin concentration in the range of 12.5–17.9 g/dL were accepted for blood donation.

Results

Analysis of paired results showed that haemoglobin levels measured by HemoCue were higher than those measured by the reference method. There was a significant correlation between the reference method and HemoCue for haemoglobin levels less than 12.5 g/dL. The correlation was less strong for increasing haemoglobin levels. Linear correlation was poor for haemoglobin levels over 18 g/dL. Thirteen percent of donors, who had haemoglobin levels close to the upper limit, were unnecessarily rejected.

Discussion

HemoCue is suitable for screening for anaemia in blood donors. Most donors at Yazd are males and a significant percentage of them have haemoglobin values close to the upper limit for acceptance as a blood donor; since these subjects could be unnecessarily rejected on the basis of HemoCue results and testing with this method is expensive, it is recommended that qualitative methods are used for primary screening and accurate quantitative methods used in clinically suspicious cases or when qualitative methods fail.

Introduction

Haemoglobin screening is used both to safeguard the health of potential donors and to ensure an adequate quality of blood products for recipients. According to world blood transfusion standards, potential donors should have haemoglobin levels over 12.5 g/dL in order to be accepted as donors^1,2. Blood donation with haemoglobin levels less than 12.5 g/dL can cause anaemia or deterioration of conditions that could lead to health problems in blood donors. In addition, the blood products from these donors are not of sufficient quality because of their low haemoglobin content³. On the other hand, blood from donors with high haemoglobin levels can also be of poor quality and have a negative effect on the recipient's safety. Based on the Iranian Standard Operation Procedure, the range of blood donor haemoglobin levels considered acceptable is from 12.5 to 17.9 g/dL and potential donors with higher or lower haemoglobin levels are deferred from blood donation (00.TM. 062. SOP/01). The initial methods that were used for haemoglobin screening in blood donors, such as copper sulphate and the haemoglobin colour scale, are qualitative or semiquantitative². They can only detect low haemoglobin levels and cannot be used for detecting high haemoglobin levels.

Nowadays, the quantitative methods used in many blood banks can determine a vast range of haemoglobin levels^4,5 and can detect blood donors with haemoglobin levels higher or lower than normal, who should not donate blood. However, the validity of these methods should be evaluated before use and methods with sufficient sensitivity and specificity should be used in order not to expose blood donors and recipients to risk or to lose potential donors.

Numerous studies have been done to evaluate the diagnostic value of these rapid methods for determining low haemoglobin levels and anaemia^6–11. The goal of these studies was to select highly sensitive and accurate methods with very low false-deferral and false-pass rates. However, few studies have evaluated their diagnostic value for determining high haemoglobin levels. We, therefore, performed this study to assess the diagnostic accuracy of a rapid, quantitative method (HemoCue 201+, HemoCue AB, Angelholm. Sweden) for detecting both high and low haemoglobin levels in blood donors.

Material and methods

In this cross-sectional study, 320 subjects were randomly selected from 9,000 blood donors attending Yazd blood transfusion centre over a 3-month period. All participants were fully informed of the aim of the study and their consent was documented by signing a consent form before blood sampling. Six donors who refused to participate were removed from the sample.

Capillary haemoglobin was measured using a HemoCue 201+ instrument (HemoCue AB, Angelholm, Sweden) calibrated in accordance with the manufacturer's instructions. The calibration of the HemoCue was verified daily before the first measurement by a control cuvette which was available for each set. Results were compared with venous haemoglobin levels determined by a Cobas-Micros automated haematology analyser (Roche ABX Cobas Micros, Montpellier, France) as the standard method. The Cobas analyser was calibrated using the standardised reference method and haemoglobin values were measured using the haemiglobincyanide (cyanmethaemoglobin) method by a spectrophotometer that was calibrated periodically and subject to regular quality controls. The parameters of the cell counter were also verified regularly by using control blood samples. In addition, external quality control was performed periodically by the Iranian reference laboratory for Cobas.

Capillary blood samples were obtained by lancing the middle fingertip previously disinfected with ethanol. The first two drops of blood were wiped away and the third drop was evaluated by HemoCue 201+. Two millilitres of venous blood were obtained by venipuncture in EDTA Vacutainer tubes immediately after taking the capillary sample and the haemoglobin concentration in this venous sample was measured by both HemoCue 201+ and the Cobas-Micros automated haematology analyser that was calibrated using a Pharmacia Biotech spectrophotometer, model 80-2088-64. The haemoglobin values determined by the Cobas-Micros automated haematology analyser were taken as the standard reference against which to compare the values obtained with HemoCue 201+. Two trained physicians performed the capillary and venous tests by HemoCue201+, separately, and a haematology laboratory technician performed the Cobas-Micros measurements. All of these people were blind to the results of the other tests. The criterion for acceptance for blood donation was a haemoglobin concentration in the range of 12.5 to 17.9 g/dL; potential donors with haemoglobin levels less than or equal to 12.4 g/dL or greater than or equal to 18 g/dL were deferred.

The sensitivity, specificity, positive and negative predictive values (PPV and NPV) and accuracy of high and low haemoglobin levels were determined for HemoCue 201+ using the following formulae:

$$\begin{array}{c}\text{Sensitivity}=\text{TP}/(\text{TP}+\text{FN})\\ \text{Specificity}=\text{TN}/(\text{TN}+\text{FP})\\ \text{PPV}=\text{TP}/(\text{TP}+\text{FP})\\ \text{NPV}=\text{TN}/(\text{TN}+\text{FN})\\ \text{Accuracy}=(\text{TP}+\text{TN})/\text{n}\end{array}$$

where TP is the number of blood donors correctly rejected on the basis of the HemoCue measurement; FN is the number of blood donors falsely passed by using the HemoCue measurement; FP is the number of blood donors falsely (inappropriately) rejected by using the HemoCue measurement; TN is the number of blood donors correctly passed for donation on the basis of the HemoCue measurement; and n is the number of samples.

The 95% confidence interval (95% CI) for sensitivity and specificity was calculated using the following formula:

$$95\%\hspace{0.17em}\text{CI}=\text{p}\pm 1.96\hspace{0.17em}{[\text{p\hspace{0.17em}}(1-\text{p})/\text{n}]}^{1/2}$$

where p is the sensitivity or specificity and n is the number of samples.

Correlations between the results of the HemoCue measurements and those of the automated analyser were calculated. Pearson's correlation coefficient, the receiver operating characteristic (ROC) curve, linear regression and paired t-test were used to analyse the data with SPSS 17.5 software (IBM Corporation, New York, USA).

Results

This study included 314 potential blood donors aged between 18 and 63 years old (mean: 34.2±10.1 years). The results obtained with HemoCue 201+ and the standard, reference method are summarised in Table I. Using HemoCue 201+, the mean capillary haemoglobin level was higher than the venous haemoglobin level, but the difference was not statistically significant (0.05±0.46 g/dL). Comparing the results of the venous blood samples analysed by the standard method and the capillary samples analysed by HemoCue, the mean capillary haemoglobin was higher (0.79±0.70 g/dL, P <0.001).

Table I.

Haemoglobin (Hb) values of blood donors.

Gender	Donation status	N	HemoCue 201+		Standard method
			Capillary Hb (g/dL) Mean±SD (range)	Venous Hb (g/dL) Mean±SD (range)	Venous Hb (g/dL) Mean±SD (range)
	First time	54	17.5±1.53 (13.6–22.1)	17.3±1.47 (13.9–22.3)	16.7±1.46 (13.1–20.8)
Male	Repeat	153	17.3±1.77 (11.3–21.6)	17.2±1.68 (12.0–21.6)	16.3±1.50 (11.3–20.8)
	Total	207	17.3±1.71 (11.3–22.1)	17.2±1.62 (12.0–22.3)	16.4±1.49 (11.3–20.8)
	First time	70	13.6±1.64 (8.1–16.1)	13.7±1.51 (8.7–16.6)	13.1±1.56 (7.9–15.7)
Female	Repeat	37	13.2±1.93 (7.8–16.1)	13.3±1.82 (7.9–16.0)	12.4±1.69 (7.0–15.2)
	Total	107	13.5±1.74 (7.8–16.1)	13.5±1.63 (7.9–16.6)	12.8±1.63 (7.0–15.7)
Total		314	16.0±2.52 (7.8–22.1)	15.9±2.40 (7.9–22.3)	15.2±2.28 (7.0–20.8)

Legend

The results of HemoCue 201+ and the standard reference method test performed on donors are summarised in this table. Using HemoCue 201+, the mean capillary haemoglobin level was higher than the venous haemoglobin level, but the difference was not statistically significant (0.05±0.46 g/dL). On comparing venous samples analysed by the standard method and capillary samples, the mean capillary haemoglobin was higher (0.79±0.70 g/dL, P <0.001).

Of the 314 blood donors evaluated, 107 (34%) were female and 207 (66%) were male. Twenty-six percent of the males and 65% of the females, with mean haemoglobin levels determined by the standard, reference method of 16.7±1.46 g/dL and 13.1±1.56 g/dL, respectively, were first-time donors. Thirty-four (31.7%) females and five (2.4%) males had haemoglobin levels less than 12.5 g/dL and were considered truly anaemic. Three men and five women within the truly anaemic group were falsely passed for donation. The lowest haemoglobin levels that were inappropriately passed was 11.2 g/dL in a woman and 12.1 g/dL in a man. Twenty-seven (8.6%) subjects had haemoglobin levels of 18 g/dL or above and 113 (36%) had haemoglobin levels between 16.1 and 17.9 g/dL, as determined by the standard method. Using the capillary haemoglobin measurements, 18.4% of the potential donors were falsely deferred because of high haemoglobin levels. All of these unnecessarily deferred donors were male. The lowest haemoglobin level that led to a false deferral was 15.8 g/dL for capillary samples and 16.2 g/dL for venous samples. Tables II and III show the distribution of high and low haemoglobin levels, using HemoCue 201+ compared to the standard method.

Table II.

Frequency of donors with low haemoglobin (Hb) levels detected by HemoCue 201+ using capillary and venous measurements.

HemoCue 201+		Standard method		Total
		Donors with Hb <12.5 g/dL, N (%)	Donors with Hb <12.5 g/dL, N (%)
Donors with Hb <12.5 g/dL	Capillary	31 (79.5)	0	31 (9.9)
	Venous	25 (64.1)	0	25 (8.0)
Donors with Hb ≥12.5 g/dL	Capillary	8 (20.5)	275 (100)	283 (90.1)
	Venous	14 (35.9)	275 (100)	289 (92.0)
Total		39 (100)	275 (100)	314 (100)

Legend

McNemar test: P <0.001.

Thirty-four (31.7%) females and five (2.4%) males had haemoglobin levels below 12.5 g/dL and were considered truly anaemic. Three men and five women in the true anaemic group were falsely passed for donation. This table shows the distribution of low haemoglobin levels as determined by HemoCue 201+ and the standard method.

Table III.

Frequency of donors with high haemoglobin (Hb) levels detected by HemoCue using capillary and venous Hb.

HemoCue 201+		Standard method		Total
		Donors with Hb ≥18 g/dL N (%)	Donors with Hb <18 g/dL N (%)
Donors with Hb ≥18 g/dL	capillary	27 (100)	58 (20.2)	85 (27.0)
	venous	26 (96.3)	41 (14.3)	67 (21.3)
Donors with Hb <18 g/dL	capillary	0	229 (79.8)	229 (73.0)
	venous	1 (3.7)	246 (85.7)	247 (78.7)
Total		27 (100)	287 (100)	314 (100)

Legend

McNemar test: P <0.001

Because of high capillary haemoglobin levels, 18.4% of donors were incorrectly deferred. All of the deferred subjects were male. This table shows the distribution of high haemoglobin levels using HemoCue 201+ and the standard method.

Table IV shows the specificity, sensitivity, PPV, NPV, false deferrals and false passes using HemoCue 201+ for haemoglobin screening of lower or higher than normal haemoglobin levels in blood donors. Pearson's correlation coefficient was determined for the results obtained with this method and the haemoglobin concentrations determined using the standard, reference method. The correlation between HemoCue 201+ and standard method results was high for haemoglobin levels lower than 12.5 g/dL (for capillary haemoglobin by HemoCue 201+: r =0.950, P <0.001 and for venous haemoglobin by HemoCue 201+: r =0.958, P <0.001). However, the correlation coefficient decreased with increasing haemoglobin levels and was poor at levels of more than 18 g/dL (for HemoCue 201+ capillary and venous haemoglobin: r =0.628 and r =0.683, respectively). The linear correlation between capillary and venous haemoglobin concentrations measured by HemoCue 201+ was high at all haemoglobin levels. Figures 1 and 2 show the correlations between haemoglobin levels determined by HemoCue 201+ and the standard method using linear regression analysis.

Table IV.

Diagnostic value of HemoCue 201+ for evaluating high and low haemoglobin (Hb) levels.

The table shows the specificity, sensitivity, positive and negative predictive values, false deferral and false-pass rates using HemoCue 201+ for haemoglobin screening of lower or higher than normal haemoglobin levels in blood donors.

Validity	Hb ≥18 g/dL		Hb <12.5 g/dL
	Venous	Capillary	Venous	Capillary
Sensitivity	96.3	100	64	79.5
95% CI	94.3–98.3		58.7–69.3	75.1–83.9
Specificity	85.7	79.8	100	100
95% CI	81.9–89.5	75.4–84.2
Positive predictive value	38.8	31.7	100	100
Negative predictive value	99.6	100	95	97.2
Accuracy	86.6	81.5	95.5	97.5
False pass rate of total blood donors	0.3	0	4.4	2.5
False deferral rate of total blood donors	13.1	18.4	0	0

Figure 1.

Linear regression analysis of venous haemoglobin levels determined by HemoCue 201+ vs venous haemoglobin levels determined by the standard method.

Figure 2.

Linear regression analysis of capillary haemoglobin levels determined by HemoCue 201+ vs venous haemoglobin levels determined by the standard method.

Discussion

According to the Iranian Standard Operation Procedure, the upper limit of haemoglobin level for acceptance for blood donation is 17.9 g/dL and volunteers with haemoglobin levels of 18 g/dL or more are deferred (00.TM. 062. SOP/01). As quantitative haemoglobinometers are being used regularly nowadays for determining a vast range of haemoglobin levels, these meters need to be validated before use.

The manufacturers claim that these instruments give linear results over a broad range of haemoglobin levels. However, as the validation of these methods has been evaluated only for lower levels of haemoglobin, in this study we assessed their validation also at higher levels of haemoglobin.

In this study we found that haemoglobin levels measured with HemoCue 201+ were higher than venous haemoglobin levels measured by a standard method using a Cobas-Micros automated analyser, confirming results of other studies^6,12–14. In certain studies, the difference was 2–3% and even 8–10% as compared to the standard method¹⁵. Portable haemoglobinometers (such as HemoCue) are also subject to technical errors (incomplete filling, bubbles in cuvettes) and abnormally high haemoglobin values are generated in some donors². Tong et al. showed that at the lower end of the normal range of haemoglobin concentrations, venous haemoglobin levels were higher than capillary levels. However, they showed that as the haemoglobin levels increased, the difference between the venous and the capillary measurements reversed¹⁶. The results from HemoCue can be influenced by measuring blood samples of different volumes or from different fingertips¹⁷. We found that as haemoglobin levels increased, the discrepancy between the haemoglobin levels measured with HemoCue 201+ and the reference method also increased. In other studies as the true haemoglobin levels increased, so did the haemoglobin levels measured with HemoCue 201+¹⁸. The results of our study showed that at low haemoglobin levels, there was a strong correlation between HemoCue and reference method measurements, with this correlation being similar to that found in other studies. In Radtke's study, the correlation between HemoCue 201+ and standard method measurements was 0.9¹¹. In the present study, the specificity of the HemoCue 201+ method for screening for haemoglobin levels lower than 12.5 g/dL was 100% for both capillary and venous samples. This was similar to the results of other studies^6,12,19. The sensitivity of this method for screening for anaemia (defined as a haemoglobin <12.5 g/dL) in this study was 79.5% for capillary blood and 64% for venous samples. Therefore, considering both the sensitivity and specificity mentioned above, the discriminating power for detecting anaemia was greater for capillary samples than for venous samples. In studies mentioned above, the sensitivity ranged from 37.9 to 99%. There were eight false-passes (4.4% for venous samples and 2.5% for capillary samples in donors with haemoglobin concentrations less than 12.5 g/dL). There was no false-deferrals. All of the donors with a haemoglobin level less than 11.2 g/dL were correctly deferred, similarly to the results of studies by Mendrone et al⁶. and Ziemann et al^6,20. Consequently, based on this study, the rapid HemoCue method is suitable for anaemia screening in blood donors. This is a very important finding, because a previous study showed that there was more iron-deficiency anaemia among female donors in the city of Yazd than in their counterparts in western countries²¹.

In our study, as the haemoglobin level increased, the correlation between results obtained with HemoCue and the standard method decreased, and at levels above 18 g/dL, the correlation was not significant. An evaluation of the diagnostic value of the HemoCue method for screening for higher than normal haemoglobin levels (≥18 g/dL) showed that, although the method had sufficient sensitivity, its specificity and PPV were, respectively, 79.8% and 31.7% for capillary samples and 85.7% and 35.8% for venous samples. Thus, based on its use on capillary and venous samples, the false deferral rates were 18.4% and 13.1%, respectively. In a study by Akbari et al. with similar results, the false-failure rate for HemoCue 201+ was 13% (unpublished data). The present results show that a large number of potential blood donors could be unnecessarily rejected if screened by this method. Losing many potential donors despite the continuously increasing need for blood and blood products could threaten the blood supply.

Unlike most countries that have an almost equal female to male ratio of donors, in Iran most donors are male. Therefore, a higher proportion of donors have haemoglobin concentrations close to the upper limit for acceptance (36% have a haemoglobin concentration of 16 to 17.9 g/dL and 8.6% have a concentration over 18 g/dL). The mean haemoglobin concentration in first-time male blood donors was 16.6±1.46 (13.1–20.8) g/dL. It seems that the haemoglobin concentration is higher in Iranian men than in European males. In a study in Denmark 3.9% of healthy men had a haematocrit of more than 54%²² but in this study 13% of male donors had a haemoglobin concentration of more than 18 g/dL. Some possible reasons for this high haemoglobin level are dehydration due to the hot, dry climate in Yazd, Iran, the high altitude over sea level and tobacco smoking in Yazd blood donors. Furthermore, an important cause of high haemoglobin levels in blood donors is that polycythemic people are frequently referred for blood donation because of their health, in accordance with their doctor's or family's recommendations.

Increasing the cut-off haemoglobin level to 19 g/dL with a specificity of 93% could decrease the deferral rate considerably, but the sensitivity would decrease to 67% resulting in an increased false-pass rate. Alternatively, the cut-off could be eliminated because it seems that taking a full donation from clinically healthy men with high haemoglobin levels is most unlikely to do harm but recommendations about life-style modifications such as smoking cessation and adequate water and fluid intake, specifically before blood donation, should be given.

Another major limitation to the use of rapid haemoglobinometry methods for primary screening is the high cost per test, which is not economical for the organisation. This is the reason for not using these methods as the first-line screening test in many centres, such as UK blood services, where the copper sulphate method is the first-line screening test for evaluating haemoglobin. The studies performed recommended using HemoCue in cases in which the copper sulphate method fails¹⁷.

The limitation of this study is that there were few readings of values at the lower end of the range of haemoglobin concentrations, an aspect that should be considered in future studies.

In summary, considering the high percentage of male donors, which results in a large proportion of blood donors with high haemoglobin levels and accordingly a high deferral rate, it is proposed that qualitative methods be used for primary haemoglobin screening of blood donors. Accurate quantitative methods should be used when there is a clinical suspicion of haematological disorders and/or qualitative methods fail. However, two concerns will remain: the first is the possibility of accepting people with haematological disorders although this is rare, while the second is a high ratio of red blood cell content to preservative solution, which could compromise the quality of units of red blood cells. Adjusting the ratio of the collected volume of blood to the volume of preservative solution may solve this problem. It would be advisable that these issues are considered in the standard operation procedures of the Iranian blood transfusion organisation.