Method Comparison and Investigation of Interference Variables of a Hand‐Held Hemoglobinometer(HemoCue Hb 201⁺) in Cats

ABSTRACT

Background

Regular blood sampling to monitor RBC mass in anemic cats can exacerbate anemia. Laboratory‐based reporting can delay clinical decisions. A hand‐held hemoglobinometer, HemoCue Hb 201⁺ (HC‐201), requires only one drop of blood (10 μL) and provides results within 1 min.

Objectives

This preliminary study aimed to evaluate the utility of HC‐201 in cats and investigate the impact of potential interferents on its performance.

Methods

One hundred and fifty‐four venous blood samples in EDTA from 93 cats were analyzed. Hemoglobin concentration was measured once using an ADVIA 2120 analyzer and compared to the mean of 2–3 replicate measurements from an HC‐201. Agreement and systematic bias between HC‐201 and ADVIA results, along with precision between HC‐201 replicates, were assessed using Lin's concordance correlation coefficient, non‐parametric Bland–Altman, Passing‐Bablok regression, and intraclass correlation coefficient. The performance of HC‐201 in the presence of anemia, leukocytosis, azotemia, lipemia, icterus, hemolysis, and peripheral versus jugular venipunctures was assessed using Wilcoxon rank‐sum tests.

Results

Passing‐Bablok analysis revealed a significant constant bias (intercept = −2.242, 95% CI: −4.042 to −0.667) but no significant proportional bias (slope = 1.015, 95% CI: 1.000–1.032). HC‐201 demonstrated excellent agreement (ρ _c = 0.989) and precision (ICC = 0.997) with a median bias of −0.67 g/L (p = 0.001). The total observed error was 3.02%, within the allowable limits defined by international standards. Neither anemia, leukocytosis, azotemia, lipemia, nor venipuncture site influenced HC‐201 measurements. Samples with icterus and hemolysis were insufficient in number for statistical comparison.

Conclusions

This preliminary study indicates that HC‐201 offers reliable point‐of‐care monitoring for hemoglobin concentration in cats.

1. Introduction

Hospitalized feline patients often require daily diagnostic testing to monitor changes in their hematologic biomarkers, particularly anemic patients experiencing active blood loss, hemolysis, or inadequate bone marrow production. However, frequent blood collection can lead to iatrogenic anemia, sometimes accumulating in sufficiently large volumes to necessitate blood transfusions or prolong hospital stays [1]. Thus, clinicians must carefully balance acquiring crucial clinical data while minimizing potential harm to the patient.

Common methods to assess oxygen‐carrying capacity include laboratory‐based hematologic analysis of RBC mass and point‐of‐care evaluations of packed‐cell volume (PCV). Hematology analyzers provide accurate and precise assessments of multiple hematologic measures, such as direct cell counts and hemoglobin concentration, which can be used to characterize anemia. However, analysis on these benchtop analyzers often necessitates sending samples to an external laboratory, leading to increased turnaround times and delayed clinical decision‐making. Additionally, laboratories stipulate submission of blood volumes (0.5–1.0 mL) that can exacerbate anemia if repeated testing is required. Alternatively, PCV offers bedside measurement of RBC volume. Although this test requires less sample volume than a CBC, it still requires time to centrifuge the blood sample and at least 70 μL to fill the hematocrit tube [2].

Hemoglobin concentration serves as an indirect measure of oxygen‐carrying capacity. The cyanmethemoglobin method is the gold standard for measuring hemoglobin concentration; however, it requires a benchtop analyzer and generates toxic waste [3]. The ADVIA 2120 hematologic analyzer provides a cyanide‐free alternative. In this method, RBCs are lysed, and the iron in the free hemoglobin is oxidized to its ferric state, which then reacts with hydroxide and water to form a green monoaquomonohydroxyferri‐porphyrin. The concentration of this compound is measured colorimetrically at 565 or 546 nm [4, 5].

The HemoCue Hb 201⁺ System (HC‐201) is one of five hemoglobinometers made by HemoCue that provide a rapid alternative for assessing anemia in humans. This device requires only 10 μL of blood to measure hemoglobin concentration and provides results within 60 s [6]. The HC‐201 relies on a chemical reaction to lyse RBC and form azidemethemoglobin from hemoglobin. The light absorbance by the molecule is then measured at two wavelengths, 570 and 880 nm, to determine hemoglobin concentration [6]. As interferents in the blood, including lipemia, hemolysis, and icterus, can affect hemoglobin readings in methods that rely on light absorption [7, 8], the dual‐wavelength approach aims to improve measurement accuracy in the presence of lipemia and icterus.

The HC‐201 has shown promising reliability in human medicine for use in children [9], liver transplant patients [10], and blood donor screening [11]. However, other studies caution against its use due to evidence of systematic bias, ie, the tendency to consistently measure lower or higher than a reference method, and inadequate agreement in repeated measurements [12, 13]. This variation in precision and accuracy suggests that the HC‐201's performance may depend on pre‐analytical factors including sample handling methods and venipuncture site [14, 15, 16]. Thus, verifying the sample handling protocol and the performance of HemoCue devices for the specific population of interest is critical before using them in clinical practice.

Studies using various HemoCue models in animals showed promising results, including in horses [17], dogs [18], pigs [19], cows [20], and cats [18, 21]. Most of these investigations identified good correlation with reference laboratory analyzers; however, most studies did not investigate agreement, accuracy, or precision [18, 20, 21]. Furthermore, the studies used different HemoCue models, some without stating the model used [17, 21], making direct comparisons challenging. To date, two studies examined using the HemoCue device to measure hemoglobin concentration in cats [18, 21], and revealed a strong correlation between hemoglobin measurements by HemoCue (unspecified models) and laboratory reference methods (ADVIA 120 and Coulter Electronics). While correlation indicates that the measurements trend in the same direction, it does not test whether they provide the same values when measuring the same analyte. An agreement, which was not examined in these two studies, is needed to evaluate this. Additionally, one of these studies assessed precision but only at two concentrations created by diluting blood with plasma in vitro [21].

Comparison of any new laboratory method or equipment should include an assessment of agreement, accuracy, and precision [22]. The total allowable error (TE_a) measurement combines the tolerable limit of imprecision with accuracy to ensure high clinical utility for a given test [22]. The TE_a for hemoglobin concentration recommended by the American Society for Veterinary Clinical Pathology (ASVCP) is 10%, while it is 4% for humans based on the Clinical Laboratory Improvement Amendments (CLIA) [22, 23]. The calculated allowable error for a machine (TE_obs) should be less than the recommended TE_a. Notably, none of the currently published method comparison studies in veterinary species provides a calculated TE_obs for any HemoCue model tested.

This preliminary study aimed to evaluate the agreement, accuracy, and precision of the HC‐201 compared to the ADVIA 2120 in cats with varying hemoglobin levels in a clinical setting, while also assessing compliance with ASVCP TE_a requirements. Secondary objectives included assessing the impact of clinically observed levels of leukocytosis, azotemia, lipemia, hyperbilirubinemia, hemolysis, and venipuncture site on HC‐201 performance and comparing its accuracy in anemic versus non‐anemic samples.

2. Materials and Methods

An Animal Use Protocol exemption was granted by the Animal Care Committee of the University of Guelph, as this research used remnant blood collected from each cat during routine diagnostic testing. Consent to use these samples was obtained from each cat's owner.

2.1. Sample Collection

A total of 154 blood samples for this study were obtained from 93 cats presented to a veterinary teaching hospital in Ontario, Canada, between January 2023 and May 2023, with samples submitted for a CBC. Despite repeated evaluations, each cat's sample was considered an independent sample because the patient's clinical condition changed daily in response to therapy. All samples were collected into either 500 μL (MiniCollect EDTA Tubes, Greiner Bio‐One, North Carolina, USA) or 2 mL EDTA tube (BD Vacutainer K2 EDTA, Becton, Dickinson and Company, New Jersey, USA).

When possible, the site of venipuncture was recorded. Central sampling sites included samples collected from the jugular vein, while peripheral veins included all other vessels (i.e., cephalic, lateral or medial saphenous, or femoral).

2.2. Method Comparison

2.2.1. Timing and Sample Measurements

The time, rounded to the nearest half hour, between blood collection and HC‐201 analysis, as well as between analysis by the ADVIA 2120 and the HC‐201, was recorded. If the time of blood collection was not documented, it was assumed to be the earliest possible time (i.e., the start of the patient's appointment, the time of arrival for triage during an emergency presentation, or within the first hour of treatment for hospitalized patients).

2.2.2. ADVIA 2120

The hemoglobin concentration of each blood sample was measured at the Animal Health Laboratory (AHL, Guelph, Ontario, Canada) using an ADVIA 2120 (Siemens Healthcare Limited, Ontario, Canada) hematology analyzer, following the laboratory's standard operating protocol. This measurement served as the reference standard for this study.

2.2.3. HC‐201

After analysis by the ADVIA 2120, each sample was mixed by gently inverting the EDTA tube 8–10 times before hemoglobin was measured using the HemoCue Hb 201+ (HC‐201; California, USA). Briefly, three 15 μL aliquots were collected from each EDTA tube using a 20 μL pipette (PuroPET S Series, Luna Nanotech, Ontario, Canada). If the blood sample had an inadequate volume, only one or two aliquots were used. Each aliquot was ejected onto a hydrophobic surface (Parafilm M, MilliporeSigma Canada Ltd., Ontario, Canada). Sequentially, each drop was aspirated into the HC‐201 microcuvette (HemoCue, California, USA) for replicate analyses. The microcuvette was inspected for air bubbles and confirmed to be completely filled before insertion into the analyzer. Hemoglobin concentration was measured within 10 s of blood aspiration.

2.3. Investigation of Potential Confounding Factors

2.3.1. Potential Interferents

According to the manufacturer, the performance of the HC‐201 may be impacted by leukocytosis (> 600 × 10⁹/L), azotemia (creatinine > 2652.6 μmol/L or 30 mg/dL, urea > 83.3 mmol/L or > 5500 mg/dL), lipemia (Intralipid > 4000 mg/dL), hyperbilirubinemia (> 684 μmol/L or > 40 mg/dL), and high doses of certain medications, including acetaminophen (> 1322 μmol/L or > 20 mg/dL), ibuprofen (> 1939 μmol/L or > 40 mg/dL), salicylic acid (> 3620 μmol/L or > 50 mg/dL), and tetracycline (> 450 μmol/L or > 20 mg/dL) [6]. However, the reported interference values for leukocytosis, azotemia, Intralipid as a surrogate for hypertriglyceridemia, and hyperbilirubinemia are exceedingly high and unlikely to be encountered in veterinary clinical settings. The presence of potential interferent was defined as the following: leukocytosis as WBC above reference range of > 13.0 × 10⁹/L on ADVIA 2120, serum creatinine concentrations above the feline reference intervals of > 190 μmol/L as measured by Cobas 6000 c501 (Hoffmann‐La Roche Limited, Mississauga, Canada) on a concurrent blood sample as the CBC, and the presence of lipemia, icterus, and hemolysis were based on visual inspection by the certified medical laboratory technologist (present or absent). Due to the limited number of samples with potential interferents, samples were grouped binarily based on whether interferent levels were within or above reference intervals to allow for a preliminary assessment.

All medications listed as interfering with the HC‐201's performance, except tetracycline, are either toxic to cats or rarely used in veterinary medicine. Serum concentrations of tetracycline were not measured in this study as the levels at standard doses are unlikely to surpass the manufacturer's guidelines for interference (> 450 μmol/L).

2.3.2. Assessment of Anemia

HC‐201 performance in measuring hemoglobin in anemic samples was assessed. Anemia was defined as a hemoglobin concentration below the lower reference limit of the ADVIA 2120 (< 93 g/L).

2.3.3. Quality Control

The reference hemoglobin method to which the ADVIA 2120 is calibrated is approved by the National Committee for Clinical Laboratory Standards and uses certified standards traceable to the International Council for Standardization in Hematology (ICSH) [4]. The AHL participates in the Veterinary Laboratory Association Quality Assurance Program. Quality control testing is performed on the ADVIA 2120 daily, every 8 h, whenever a reagent is changed, and at three reference concentrations (within, below, and above the reference interval) using hematology reference materials compatible with the machine (ADVIA 3•in•1 TESTpoint Hematology Controls, Siemens Healthineers, Oakville, Canada) to ensure that the machine's performance meets acceptable precision and accuracy standards.

The HC‐201 is factory‐calibrated against the international reference method for hemoglobin determination, as established by the ICSH, and holds a CLIA waiver, which eliminates the need for additional calibration prior to clinical use [24]. Additionally, this analyzer features an internal self‐test quality control system that activates each time the device is turned on [24]. Despite these safeguards, we performed quality control assessments before the start of the study period, after 1 month of use, and at the end of the study period using HemoTrol Level I (76–84 g/L) and Level II (113–125 g/L) hemoglobin control (EuroTrol, Kentucky, USA). The analyzer operated within the testing reference interval at all three time points. HemoTrol Level III was not used in quality control analyses due to the infrequent occurrence of hemoconcentration in the presented cat population during this study.

2.3.4. Statistical Analyses

Descriptive statistics for the age and weight of the cats, the time between repeated samples of the same cat, the averages of time intervals from sample collection to analysis, hemoglobin concentrations obtained from both the ADVIA 2120 and replicate HC‐201 hemoglobin measurements, WBC counts, and creatinine concentrations were calculated using Microsoft Excel 2020 (Microsoft Corporation, Washington, USA) and RStudio 2023.12.0.369 (Posit Software, MA, USA). Normality of the descriptive data was assessed using the Shapiro–Wilk test and visual inspection of a histogram. Descriptive statistics are presented as mean ± SD for normally distributed data, and as median with interquartile range (IQR) for non‐normally distributed data to most accurately represent the population and data. All other statistical analyses were performed using SAS/STAT 9.4 (SAS Institute Inc., North Carolina, USA).

Passing‐Bablok regression analysis was performed to assess constant and proportional bias within the data. Lin's concordance correlation coefficient was used to assess the agreement between the ADVIA 2120 and the mean of the HC‐201 replicate measurements. The intraclass correlation coefficient (ICC) [25] for the HC‐201 replicate readings was used to assess assay precision. Grading schemes for Lin's concordance and ICC are presented in Table 1 [25, 26]. Bland–Altman analysis was performed to assess accuracy, or bias, between instruments. Kolmogorov–Smirnov, Anderson Darling, and Shapiro–Wilk tests were used to assess the residuals on the paired t‐test for bias. As the data were not normally distributed, the non‐parametric Wilcoxon rank‐sum test was performed to evaluate for statistically significant bias between methods. The 97.5% and 2.5% quantiles were used as the upper and lower limits of agreement (LOA), respectively. In the bias analysis, two outliers exceeding a three SD difference from the mean were identified after scrutiny of the Bland–Altman plot, and it was determined that they likely arose from insufficient mixing of minimal sample volumes in MiniCollect tubes. These outliers were excluded from calculations for concordance, bias, and interference variable analyses, but retained in ICC calculations as ICC relied solely on HC‐201 replicate readings.

TABLE 1.

Grading schemes for Lin's concordance correlation coefficient and intraclass correlation coefficient [25, 26].

Lin's concordance correlation coefficient
Poor	< 0.900
Moderate	0.900–0.950
Substantial	0.950–0.990
Almost perfect	> 0.990
Intraclass correlation coefficient
Poor reliability	< 0.50
Moderate reliability	0.50–0.75
Good reliability	0.75–0.90
Excellent reliability	> 0.90

The formula used to calculate total allowable error is: [22]

$${\mathrm{TE}}_{\mathrm{obs}}=\text{bias}\%+2\mathrm{CV}\%=\left[\frac{{\text{Mean}}_{\text{target}}-{\text{Mean}}_{\text{measured}}}{{\text{Mean}}_{\text{target}}}\times 100\%\right]+\left[2\left(\frac{\mathrm{SD}}{\text{Mean}}\times 100\%\right)\right]$$

where CV is the coefficient of variation.

Bias was determined from the mean difference between ADVIA 2120 and HC‐201 measurements. The SD and CV% were calculated for each set of HC‐201 replicate readings, and the average SD and CV% were used in the TEobs calculation.

When analyzing the impact of different subgroups of potential interferents, normality was assessed using the Kolmogorov–Smirnov test for sample size > 50 (anemic, jugular vein sampling) and the Shapiro–Wilk test for sample size ≤ 50 (azotemic, lipemic, leukocytosis, peripheral vein sampling). A general linear mixed model was attempted to compare biases in the subgroup comparisons. However, due to the lack of normality in the data, a Wilcoxon rank‐sum test was used to compare samples with results within the reference interval and those with anemia, leukocytosis, azotemia, lipemia, as well as between sampling sites.

2.3.5. Sample Size

After 34 samples were assayed, a preliminary analysis of Lin's concordance coefficient and ICC was performed and showed excellent agreement and precision, suggesting a sample size of 34 was sufficient. However, to bolster the interferent subgroup sample sizes, additional samples were collected until resource capacities were fully utilized.

3. Results

A total of 154 blood samples were acquired from 93 cats, of which 48 (52%) were neutered males, 44 (47%) were spayed females, and 1 (1%) was an intact male. The average age of the cats was 8.0 ± 4.5 years, and the median weight was 4.50 (IQR: 3.73–5.87) kg. Among these 93 cats, 62 (67%) had their blood analyzed once, 16 (17%) twice, 5 (5%) three times, 7 (8%) four times, 1 (1%) five times, and 2 (2%) six times spanning the data collection period. The median time between samples from the same cat is 18 (IQR: 7–30) days. The primary presenting diseases of the cats at the time of blood collection are summarized in Table 2, with neoplasia and blood dyscrasia being the most common.

TABLE 2.

Major presenting diseases of the cats (N = 93) at the time of blood collection. Seven cats had concurrent diseases, including: two with blood dyscrasia and inflammatory bowel disease (IBD); one with IBD and chronic kidney disease; one with congestive heart failure and non‐neoplastic nasal disease (epistaxis); one with hyperthyroidism and pancreatitis; one with hyperthyroidism and pleural effusion; and one with blood dyscrasia and diabetes mellitus.

Disease	Number (N)	Percent
Neoplasia	23	24.5
Blood dyscrasia	16	17.0
Acute kidney injury	9	9.6
Gastrointestinal disease	8	8.5
Chronic kidney disease	5	5.3
Non‐neoplastic nasal disease (e.g., epistaxis, congestion)	4	4.3
Metabolic disease	4	4.3
Non‐specific signs of illness (e.g., lethargy, hyporexia)	4	4.3
Cardiac disease	4	4.3
Neurological disease	4	4.3
Respiratory disease	3	3.2
Hepatic disease	2	2.1
Airway masses	2	2.1
Congenital disease	2	2.1
Lower urinary tract disease	2	2.1
Other diseases
Pancreatitis	1 each	1.0 each
Sepsis
Polydipsia
Hemoabdomen
Diaphragmatic hernia
Vaginal discharge
Thomboembolism
Abdominal pain
Unilateral transected ureter

The times from blood collection to analysis by the HC‐201 and the time from analysis by the ADVIA 2120 to analysis by HC‐201 were known for 146 (95%) and 140 (91%) samples, respectively. The median time between blood collection and HC‐201 analysis, and between ADVIA 2120 and HC‐201 analysis, was 7.0 (IQR: 5.0–10.0) hours and 4.8 (IQR: 2.5–6.5) hours, respectively. Among the 154 blood samples, 139 (90%) had three replicates, 13 (8%) had two replicates, and 2 (1%) had one replicate measured.

3.1. Comparison Between ADVIA 2120 and HC‐201 Performance

The Passing‐Bablok analysis indicated good comparability between the two methods (Figure 1). As the 95% CI for the slope included 1.0, there was no significant proportional bias. However, the 95% CI for the intercept did not include zero, indicating a significant constant bias (Figure 1). The Lin's concordance correlation coefficient (ρ _c) between the ADVIA 2120 and HC‐201 was 0.989 (95% CI: 0.987–0.993). Replicate HC‐201 measurements revealed an ICC of 0.997 (95% CI: 0.988–0.999). The mean hemoglobin concentrations measured by the ADVIA 2120 and HC‐201 were 104.28 ± 28.0 g/L and 103.44 ± 28.53 g/L, respectively. The HC‐201 demonstrated a statistically significant median bias of −0.67 (95% CI: −12.33–7.33) g/L (p = 0.008) (Figure 2).

FIGURE 1.

Passing‐Bablok regression comparing hemoglobin measurements by ADVIA 2120 and HemoCue Hb 201⁺. Slope = 1.015 (95% CI: 1.000–1.032). Intercept = −2.242 (95% CI: −4.041 to −0.667).

FIGURE 2.

Bland–Altman plot comparing hemoglobin concentration measurements by HemoCue Hb 201⁺ (HC‐201) and ADVIA 2120. HC‐201 has a slight negative bias compared to the reference method. Hb, hemoglobin.

The average SD of replicate HC‐201 measurements was 1.09 g/L, and the average CV was 1.08% (range 0%–16.1%). The % bias was 0.89%, and the TE_obs for the HC‐201 was 3.02%, which was within the limits of acceptability established by the ASVCP and CLIA [22].

3.2. Comparison of Interferent Factors

Of the 154 samples, 52 (33.8%), 32 (20.1%), 23 (14.9%), 17 (11.0%), 7 (4.5%), and 6 (3.9%) samples demonstrated anemia, leukocytosis, azotemia, lipemia, icterus, and hemolysis, respectively. Of the 154 samples, 147 (95.5%), 146 (94.8%), 137 (88.9%), 121 (78.6%), 100 (64.9%), and 73 (47.4%) were non‐hemolytic, non‐icteric, non‐lipemic, with normal leukocyte count, with normal hemoglobin concentration, and non‐azotemic, respectively. One sample had unknown icterus status, and another had unknown hemolysis status, and both were excluded from the evaluation. Fifty‐six samples (36.6%) did not have concurrent creatinine measurements; thus, the azotemic status was unknown, and these were excluded from the analysis. The descriptive statistics for WBC count in cats with leukocytosis and creatinine concentrations in azotemic cats are shown in Table 3.

TABLE 3.

Descriptive statistics of white blood cell (WBC) counts in cats with leukocytosis and serum creatinine concentrations in azotemic cats.

Analyte	Median	Interquartile range
WBC (×109/L)	23.8	17.7–29.8
Creatinine (μmol/L)	272.5	219.0–416.5

When assessing the impact of leukocytosis, azotemia, and lipemia on HC‐201 performance, none of the median differences between subgroups with and without these interferents were statistically significant, and agreement was moderate to substantial across these subgroups (Table 4). Due to the limited number of samples with icterus and hemolysis, statistical analysis of their impact on HC‐201 performance could not be performed.

TABLE 4.

Subgroup analysis of potential interference variables: Median bias, precision, and agreement of HemoCue Hb 201⁺ readings compared to ADVIA 2120.

Factor	Bias, median (IQR) (g/L)	LOA (g/L)	Wilcoxon rank‐sum test, p	Lin's concordance correlation coefficient, ${\rho }_c$ (95% CI)
Anemic (Hb < 93 g/L)	−0.67 (−2.33–0.33)	−12.33–10.50	0.257	0.946 (0.916–0.976)
Non‐anemic (Hb 93–153 g/L)	−0.67 (−2.17–1.33)	−22.33–9.00		0.969 (0.956–0.981)
Normal creatinine (50–190 μmol/L)	−1.00 (−2.33–0.67)	−14.00–9.00	0.070	0.990 (0.986–0.995)
High creatinine (> 190 μmol/L)	0.00 (−1.00–1.67)	−4.00–4.67*		0.979 (0.961–0.997)
Lipemic	−1.33 (−2.67–0.67)	−22.33–10.50*	0.809	0.964 (0.927–1.000)
Non‐lipemic	−0.67 (−2.33–0.67)	−13.25–10.00		0.992 (0.989–0.995)
Normal WBC (4.2–13.0 × 109/L)	−0.67 (−2.00–1.00)	−12.33–11.00	0.546	0.993 (0.991–0.996)
High WBC (> 13.0 × 109/L)	−0.67 (−2.67–0.06)	−22.33–10.50*		0.966 (0.941–0.992)
Jugular	−0.67 (−2.33–0.33)	−12.33–2.67	0.700	0.986 (0.978–0.994)
Peripheral	−0.67 (−3.00–0.67)	−5.00–4.00*		0.992 (0.986–0.998)

Abbreviations: ${\rho }_c$, Lin's concordance coefficient; CI, confidence interval; Hb, hemoglobin; IQR, interquartile range, LOA, limit of agreement.

^*5% and 95% quantiles reported as small sample size prohibited 97.5% and 2.5% quantile estimates.

3.3. Investigation of Effect of Anemia on HC‐201 Performance

The agreement between HC‐201 and ADVA 2120 measurements was not significantly different between subgroups of samples with anemia and those with hemoglobin concentrations within the reference interval (Table 4).

3.4. Comparison of Sampling Sites

The venipuncture site was recorded for 83 of the 154 samples (54%). Fifty‐one (61%) were collected from the jugular vein and 32 (39%) from a peripheral vein. There was no significant difference in the bias of HC‐201 hemoglobin measurements between the sampling site (Table 4).

4. Discussion

The HC‐201 demonstrated excellent agreement (ρ _c = 0.989) with the reference method and high precision (ICC = 0.997) in technical replicate measurements for assessing hemoglobin concentration in venous blood from cats. Passing‐Bablok analysis revealed a slight constant negative bias, or a tendency to consistently measure below the reference value. This is supported by a median bias of −0.67 g/L from non‐parametric Bland–Altman analysis. Although the HC‐201 tended to measure slightly lower than the ADVIA 2120, this difference is unlikely to be clinically significant, as discussed below. None of the potential pre‐analytical factors or interferents, including anemia, leukocytosis, azotemia, lipemia, or sampling site, affected HC‐201 performance.

The current study demonstrated excellent agreement (ρ _c = 0.989, slope = 1.015) between the HC‐201 and the reference method. A lack of proportional bias identified in the Passing‐Bablok analysis indicated that the bias is expected to be similar throughout the range of hemoglobin concentration measurements. Two previous studies in cats and two in humans determined correlation coefficients between various types of HemoCue analyzers and reference methods that were all 0.98 or above [18, 21, 27, 28]. In contrast, another study in humans identified less robust agreement between the HC‐201 and the Sysmex XT‐1800i [12]. However, in this study, the HC‐201 measurements were performed on capillary samples [12], which may be associated with reduced accuracy and precision, as discussed below. Further, while useful to determine whether measurements are trending together and cover a similar range of values, correlation coefficients neither indicate whether the measurements are accurate nor if they have good agreement.

In the present study, the HC‐201 demonstrated excellent precision (ICC = 0.997). A literature search did not identify any animal studies reporting the ICC of the HC‐201. However, a study in cats suggested good precision for an unspecified HemoCue model, as indicated by a low CV% (0.24%–1.00%) in replicate measurements [21], which is similar to the CV% observed in the current study. The average imprecision (% CV) of the current study and the CV% in the previous study both fell within the optimal range (< 1.5%) according to ASVCP guidelines [22]. Similarly, a human study demonstrated a high correlation coefficient between replicate HC‐201 measurements (r = 0.92), further supporting the device's excellent precision. Despite variation in methods to assess precision, our findings highlight the strong performance of the HC‐201 in providing reliable hemoglobin measurements.

The current study found a slight negative median bias of −0.67 g/L in HC‐201 measurements compared to the reference method. Given that PCV is considered approximately three times the hemoglobin concentration in g/dL [21], a 0.67 g/L difference in hemoglobin translates to about a 0.2% difference in PCV, which is unlikely to be clinically relevant. However, the LOA for the bias is wide (−12.33–7.33), and this 19.66 g/L range could translate to possible deviations of 5%–6% in PCV, which would impact clinical decision‐making. The wide interval is likely due to the non‐parametric analysis. In comparison, another study comparing the HC‐201 with the ADVIA 2120 in 60 humans reported no statistically significant differences in mean venous hemoglobin measurements [29].

Previous human studies comparing venous hemoglobin measurements by HC‐201 to various automated analyzers have found variable degrees of bias, including a negative but not statistically significant bias in two studies [10, 30], a positive bias in three studies [11, 13, 28], and no significant bias in one study [27]. Interestingly, one study reported a consistent negative bias ranging from −1.1 to −0.3 g/L (LOA −10.4 to 8.6 g/L) when comparing multiple HC‐201 units to a reference automated analyzer (Sysmex XP‐100) [15], which suggests that some degree of bias may arise from the analytical method used by the HC‐201 rather than from individual device variation, as the negative bias was consistent across different HC‐201 units. However, the variability in bias observed across the different studies may still be partially attributed to pre‐analytical factors such as differences in the reference laboratory analyzer used, the species of test subjects (humans versus cats), and sample handling methods.

A secondary objective of the current study was to make a preliminary assessment of interferent effects on HC‐201 performance in clinically encountered serum concentrations. Blood samples from clinical patients were used, as the serum matrix from an ill patient is expected to be different from that of a sample spiked with interferent, as classically performed in interferent studies [31]. Further, the leukocyte count, creatinine, urea, and bilirubin concentrations reported to interfere with the HC‐201 are extremely high and clinically rare [24]. No patients in this study met the cut‐off values specified in the HC‐201 manual. However, patients with values for these interferents above the reference interval were included in the analysis. In the current study, leukocytosis, azotemia, and lipemia did not impact the performance of HC‐201 when above reference intervals but below the manufacturer's reported range of concern. A previous study found no significant difference in the variation of HemoCue measurements compared to laboratory references in dogs with leukocytosis and lipemia [18], also suggesting that these two interferents do not significantly impact the performance of HemoCue. However, that study found higher variability when triglycerides were > 5.65 mmol/L and WBC counts were > 50.0 × 10⁹/L [18], although this was not statistically analyzed, likely due to the small sample size. Similarly, the current study did not compare variability within these subgroups due to low sample size. It is important to note that significant icterus, lipemia, and hemolysis can cause discrepancies in hemoglobin readings between colorimetric and light‐scatter analysis in the ADVIA 2120 [7]. Therefore, since HC‐201's performance was compared to the ADVIA 2120, it cannot be ruled out that the lack of significant differences in HC‐201's performance may be due to similar effects on ADVIA 2120 measurements. Overall, these results suggest that the dual‐wavelength system used in the HC‐201 may adequately adjust for clinically relevant levels of pigmentation or light scatter in serum, providing readings with accuracy comparable to a reference analyzer.

The current study found no significant difference in bias between HC‐201 and ADVIA 2120 for anemic and non‐anemic samples. This is further supported by the absence of proportional bias in the Passing‐Bablok regression analysis. Similarly, a study in dogs and cats found a strong correlation (r > 0.99) and low bias (0.5–0.9 g/L) between an unspecified HemoCue model and a reference laboratory hematology analyzer (Coulter Electronics) in both anemic and polycythemic subjects [18]. Together, these results suggest that the HC‐201 is likely to perform well across a range of hemoglobin concentrations.

The current study found no significant difference in HC‐201's performance between jugular and peripheral sampling sites. Similarly, a study in humans reported no significant difference in hemoglobin concentration between peripheral venous, central venous, and peripheral arterial blood samples [32]. However, several human studies have shown discrepancies in HemoCue models when comparing capillary versus venous blood, with capillary samples showing greater bias [13, 15], lower precision [33, 34], and weaker correlation with reference laboratory analyzers [11]. The differences in hemoglobin concentration between central, peripheral, and capillary samples may be explained by the Fahraeus–Lindqvist effect, which describes the phenomenon of decreasing concentration of RBC in narrower blood vessels [35]. Other possible factors include dilution of blood by tissue fluid [15], particularly if pressure is applied to the site to encourage blood flow [11], variations in colloidal suspensions of blood in capillaries of different diameter [29], difficulties in cuvette filling from smaller volumes in capillary drops [30], and the effect of temperature on blood flow [11]. Although the HC‐201 has demonstrated promising accuracy and precision with venous samples in the current study, it would be interesting to investigate whether there are differences in agreement with capillary and arterial samples in cats. If capillary samples prove to have comparable accuracy to venous samples, the HC‐201 could be invaluable for monitoring anemia in feline patients without the need for repeated venipuncture, as such measurements could be obtained via an ear or toe prick.

4.1. Limitations

Some cats had repeated samples taken, which could introduce bias if inherent characteristics of individual cats influenced analyzer performance. Although the samples were obtained from the same cat, they were analyzed as independent samples due to expected changes in patient treatment, disease stages, physiological conditions, and hematological markers of the cats between the days of blood collection, particularly after a median duration of 18 days between the samples.

The duration for which samples were stored before analysis by the HC‐201 and ADVIA 2120 varied, which may affect the perceived accuracy and precision of the measurements. This factor was not statistically analyzed in the current study due to the sample size. When validating a new analyzer, limiting the interval between measurements by the reference and study analyzer to no more than 2–4 h is recommended to minimize discrepancies due to sample degradation [36]. Although adhering to this timeline was not feasible in the current study, strong evidence supports that hemoglobin is generally stable. A 2017 study showed that storing blood for up to 4 days at 10°C and 30°C resulted in < 1% change in hemoglobin concentration compared to baseline when measured by the HC‐201 [37]. Similarly, two other studies found that hemoglobin measurements did not significantly differ from baseline for up to 72 h, whether stored at room temperature (23°C–25°C) or refrigerated (4°C) [38, 39]. Thus, even the longest delay in HC‐201 measurement in this study of 26 h was not expected to cause significant changes in hemoglobin concentration.

The performance of the HC‐201 was not compared to that of the ADVIA 2120 in subgroups with icterus and hemolysis due to small sample sizes. Since the HC‐201 measures at two wavelengths to account for pigmentation, comparing the two analyzers' performance with a larger sample size would be critical. Further, only three blood samples had hemoglobin concentrations above the reference range, preventing statistical analysis of the impact of hemoconcentration on analyzer performance. This represents a potential area of interest for future studies.

The analysis of agreement, bias, and precision was performed on a collective sample, including those with interferents, to ensure a larger sample size and greater generalizability. Although, interferents have been shown to have an insignificant impact on HC‐201 performance at high levels [6], the current study did not find leukocytosis, azotemia, and lipemia to impact HC‐201 performance at clinically relevant levels. However, additional subgroup analyses assessing the agreement, bias, and precision of samples without interferents are provided in Appendix S1. Overall, HC‐201's performance remains excellent, with values comparable to those from the full dataset.

This study used remnant blood from EDTA tubes applied to a hydrophobic surface, which differs from the direct capillary sampling via lancet protocol in the HC‐201 manual. Capillary sampling was not performed to optimize animal welfare by preventing a second puncture, as this study aimed to provide an initial assessment of HC‐201's performance in client‐owned cats with naturally varying hemoglobin concentrations. Given HC‐201's demonstrated accuracy and precision with venous samples in this study, future research could explore its performance with capillary samples, enabling anemia monitoring with minimal blood volume and no repeat venipuncture.

Finally, this study did not examine drugs that may impact HC‐201 performance. However, many medications cited as concerns by the manufacturer are not used in veterinary medicine, making their study irrelevant. The only medication commonly used for cats is tetracycline, namely doxycycline. Although serum tetracycline concentrations were not measured in the current study, one study reported a peak plasma concentration of 51.7 μmol/L in cats following a 5 mg/kg IV dose [40, 41]—well below the 450 μmol/L interference threshold [24]. Given standard dosing (5 mg/kg every 12 h or 10 mg/kg once daily), tetracyclines are unlikely to significantly affect HC‐201 measurements in clinical settings.

5. Conclusions

The HC‐201 measures hemoglobin concentration in feline patients with excellent precision and accuracy, falling within the acceptable TE_a limits set by the ASVCP and CLIA. It demonstrates a slight negative bias compared to the ADVIA 2120, which is unlikely to be clinically significant. Anemia, leukocytosis, azotemia, lipemia, and sampling site did not significantly impact HC‐201's performance. This study focused on venous hemoglobin measurements with the HC‐201, and further investigation into its accuracy and precision with capillary or arterial blood, as well as with hemoconcentrated, hemolyzed, and icteric blood samples, would be beneficial.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1.

Figure S1. Passing‐Bablok regression of the subset of data with no interferents (n = 49). Slope = 1.024 (95% CI: 0.994–1.055). Intercept = −3.619 (95% CI: −7.622 to −0.090).

Figure S2. Bland–Altman plot comparing hemoglobin concentration measurements by ADVIA 2120 and HemoCue Hb 201⁺ in the subset of samples with no interferents (n = 49). Hb, hemoglobin.