HbA1c levels measured by enzymatic assay during off-site health checkups are lower than those measured by on-site HPLC assay

Abstract

HbA1c is widely used as a therapeutic target marker and as a diagnostic marker for diabetes mellitus. This has led to an increasing frequency of HbA1c measurements in current health checkups throughout Japan. In the present study, we compared the HbA1c levels measured by an enzymatic assay (EA-HbA1c) off-site during health checkups with the HbA1c levels measured by on-site ion-exchange high-performance liquid chromatography (HPLC; HPLC-HbA1c) in a hospital. A total of 96 individuals (53 males and 43 females; age, 68.9 ± 8.4 years old; 70 diabetic and 26 non-diabetic individuals) whose HbA1c levels were measured by both the methods listed above were included in the study. Since no HPLC-HbA1c levels were measured on the day of the health checkup, HPLC-HbA1c levels were estimated using HPLC-HbA1c levels measured before and after the health checkup. A significant correlation of HbA1c levels was observed between the two groups (R = 0.973; p < 0.001). However, EA-HbA1c levels measured off-site during health checkups are lower than estimated HPLC-HbA1c levels measured on-site (6.37 ± 0.75% vs. 6.69 ± 0.75%; p < 0.001). Since lower EA-HbA1c levels measured during health checkups, which diverged from on-site measurements, may lead to underestimating diabetes mellitus, accurate measurement of HbA1c is required irrespective of the measuring method. Further investigation of the cause of falsely low EA-HbA1c levels and the strategy for reconciling HbA1c to reflect plasma glucose accurately are warranted.

Introduction

HbA1c is widely used as a therapeutic target marker and as a diagnostic marker for diabetes [1]. This has led to an increasing frequency of HbA1c measurements in current health checkups throughout Japan. Consequently, HbA1c has increasingly come to be used not only in diabetes practice but also in the field of health checkups recently.

Although HbA1c levels had previously been measured by the ion-exchange high-performance liquid chromatography (HPLC) assay alone, other techniques, such as affinity chromatography, immunoassays, and enzymatic assays, have been developed and HbA1c levels have currently been measured by various methods. Among these methods, an enzymatic assay is the most widely used in health screening settings, because it is less expensive and can process a larger number of samples more quickly than the other methods. Therefore, facilities that employ the enzymatic assay are increasing for health checkups [2].

Most devices and kits for measuring HbA1c levels have been approved by the certification body of NGSP [3]. Moreover, more or less normal HbA1c levels have been made available using techniques in Japan based on an external accuracy management investigation by the Japan Medical Association [4]. However, Otabe et al. reported that HbA1c levels measured by an enzymatic assay (EA-HbA1c) were approximately 0.3% lower than the measurements by the HPLC assay (HPLC-HbA1c) in a study of diabetic patients [5]. In the present study, we attempted to compare EA-HbA1c levels measured during off-site health checkups with on-site HPLC-HbA1c levels measured in a hospital.

Subjects and methods

Subjects

A total of 96 individuals (53 males and 43 females; age, 68.9 ± 8.4 years; 26 non-diabetic and 70 diabetic individuals) visiting the Kawanishi City Hospital whose EA-HbA1c levels were measured during health checkups and HPLC-HbA1c levels were obtained in the hospital before and after measurement of EA-HbA1c levels in the hospital were included (Table 1). Nine laboratory facilities measured HbA1c levels during health checkups. Patients whose HbA1c levels were measured in health screening facilities and those with HbA1c levels obtained by kits other than the Metabo Lead for health checkups were excluded from the study. For patients with multiple EA-HbA1c measurements recorded during health checkups, the mean levels of EA-HbA1c and HPLC-HbA1c were calculated and used for this study. The reported investigations were carried out in accordance with the principles of the Declaration of Helsinki as revised in 2000. The study was approved by the institutional research board of Kawanishi City Hospital (date of approval; December 25, 2018, approval no.: 30022).

Table 1.

Clinical characteristics in study individuals and the comparison of various values between the non-diabetic individuals (non-DM) and diabetic individuals (DM)

	Total	Non-DM	DM	p
n	96	26	70	–
Age (years)	68.9 ± 8.4	68.7 ± 10.0	68.9 ± 7.8	0.9131
Male (%)	53 (55.2)	13 (50.0)	40 (57.1)	0.5317
EA-HbA1c (E2) (%)	6.37 ± 0.75	5.55 ± 0.24	6.68 ± 0.64	< 0.001
pre-HPLC-HbA1c (H1) (%)	6.68 ± 0.88*	5.85 ± 0.27*	6.99 ± 0.83*	< 0.001
eHPLC-HbA1c (H2) (%)	6.69 ± 0.75*	5.85 ± 0.26*	7.01 ± 0.70*	< 0.001
post-HPLC-HbA1c (H3) (%)	6.71 ± 0.76*	5.93 ± 0.57*	7.00 ± 0.60*	< 0.001
ΔHbA1c (EA-HbA1c − eHPLC-HbA1c) (%)	− 0.32 ± 0.18	− 0.30 ± 0.14	− 0.33 ± 0.19	0.460
%HbA1c (EA-HbA1c/HPLC-HbA1c) (%)	95.2 ± 2.5	94.9 ± 2.3	95.3 ± 2.6	0.538
Period from the day of HPLC-HbA1c measurement prior to the health checkup until the day of the health checkup (a) (days)	35 (16, 54)	82.5 (0, 190)	33.5 (19.5, 47.5)	< 0.001
Period from the day of the health checkup until the day of HPLC-HbA1c measurement after the health checkup (b) (days)	43 (14.8, 71.3)	100 (34, 166)	34.5 (13.5, 55.5)	< 0.001

Abbreviations of E2, H1, H2, H3, a and b were used in Fig. 1

*p < 0.001 vs. EA-HbA1c

DM diabetes mellitus, pre-HPLC-HbA1c HPLC-HbA1c measurement prior to the health checkup, eHPLC-HbA1c estimated HPLC-HbA1c, post-HPLC-HbA1c HPLC-HbA1c measurement after the health checkup

Laboratory methods

At Kawanishi City Hospital, the HPLC assay in standard mode (G8; Tosoh Co., Tokyo, Japan) was used for HbA1c measurements. Blood samples were collected with sodium fluoride-containing blood collection tubes for plasma glucose measurements and HbA1c were measured in whole blood. As for the health checkup, blood samples were collected with sodium fluoride-containing blood collection tubes for plasma glucose measurements and transported to the laboratory facilities followed by HbA1c measurements with the enzymatic assay (Metabo Lead; Hitachi Chemical Diagnostics Systems Co., Ltd., Tokyo, Japan) within 1 day after blood sampling [6]. The samples were centrifuged at 800g for 5 min for measuring HbA1c levels in the blood cells. All HbA1c levels were indicated according to the National Glycohemoglobin Standardization Program (NGSP) [3].

Calculation of estimated HPLC-HbA1c levels at EA-HbA1c measurement

Since no HPLC-HbA1c levels were measured on the day of the health checkup, HPLC-HbA1c levels were estimated. Based on HPLC-HbA1c levels measured before and after the health checkup, HPLC-HbA1c levels after adjustment for the number of days from the checkup were obtained by the following procedure: HPLC-HbA1c levels measured before and after the health checkup were referred to as H1 and H3, respectively, and EA-HbA1c levels measured for the health checkup as E2. In addition, estimated HPLC-HbA1c on the day of the health checkup was referred to as H2. The period from the day of the HPLC-HbA1c measurement prior to the health checkup until the day of the health checkup was set as “a”, and another period from the day of the health checkup until the day of the HPLC-HbA1c measurement after the health checkup was set as “b”. H2 was calculated by the following formula (Fig. 1):

$$\text{H2}=\text{H1}+\left(\text{H3}-\text{H1}\right)\times \text{a/(a}+\text{b}).$$

Fig. 1.

Estimation of HPLC-HbA1c levels at EA-HbA1c measurement. HPLC-HbA1c levels measured before and after health checkups were referred to as H1 and H3, respectively, and EA-HbA1c levels measured during health checkups as E2. In addition, estimated HPLC-HbA1c on the day of the health checkup was referred to as H2. The period from the day of HPLC-HbA1c measurement prior to the health checkup to the day of the health checkup was set as “a” and another period from the day of the health checkup until the day of HPLC-HbA1c measurement after the health checkup was set as “b”. H2 was calculated by the following formula: $\text{H2}=\text{H1}+\left(\text{H3}-\text{H1}\right)\times \text{a/(a}+\text{b})$

ΔHbA1c was obtained by subtracting H2 (estimated HPLC-HbA1c) from E2 (EA-HbA1c). The ratio of EA-HbA1c to estimated HPLC-HbA1c was obtained by dividing E2 by H2 and then multiplying by 100 (%).

Statistical analyses

Continuous variables are shown as mean ± SD, when distribution was normal and as medians with interquartile range when distribution was skewed. For statistical analyses, the unpaired Student’s t test, Mann–Whitney U test, or Chi-square test was used to compare the two groups, as appropriate. To analyze the correlations between two parameters, Pearson’s correlation coefficient was performed with the StatView computer program (Version 5.0 for Windows, Abacus Concepts, Berkeley, CA). p values < 0.05 were considered statistically significant.

Results

Among the 96 individuals, 70 individuals (72.9%) had type 2 diabetes mellitus (Table 1). EA-HbA1c levels (6.37 ± 0.75%) were significantly lower than both pre-HPLC-HbA1c levels (6.68 ± 0.88%; p < 0.001) and post-HPLC-HbA1c levels (6.71 ± 0.76%; p < 0.001). Furthermore, EA-HbA1c levels were also significantly lower than the estimated HPLC-HbA1c (6.69 ± 0.75%; p < 0.001). The differences between EA-HbA1c levels and estimated HPLC-HbA1c levels [Δ(EA-HbA1c − estimated HPLC-HbA1c)] were − 0.32 ± 0.18% and the EA-HbA1c/estimated HPLC-HbA1c ratio was 95.2 ± 2.5%. Both values were not significantly different between non-diabetic subjects and diabetic patients (Table 1).

A strong correlation between EA-HbA1c and estimated HPLC-HbA1c was significant (R = 0.973, p < 0.001); the regression equation for both levels was as follows: EA-HbA1c = 0.976 × estimated HPLC-HbA1c − 0.158 (Fig. 2a). Both regression lines shifted downward as compared to the line of y = x. Furthermore, ΔHbA1c was not significantly correlated with estimated HPLC-HbA1c (Fig. 2b).

Fig. 2.

Comparison of EA-HbA1c levels with estimated HPLC-HbA1c levels. a Correlation of EA-HbA1c with estimated HPLC-HbA1c. The line y = x is shown with a dotted line. b Bland–Altman plot. Difference between EA-HbA1c and estimated HPLC-HbA1c [Δ(EA-HbA1c − HPLC-HbA1c)] and correlation with estimated HPLC-HbA1c are indicated

Changes in HbA1c levels obtained for the health checkups from two patients in the same laboratory facility are indicated in Fig. 3. Both the patients had been receiving oral anti-diabetic drugs. The laboratory facilities had measured HbA1c levels by the HPLC method until January 2017 and changed the testing technique to the enzymatic assay from February 2017. In both the patients, off-site HPLC-HbA1c levels measured for the health checkups were comparable to the on-site HPLC-HbA1c levels, while off-site EA-HbA1c levels measured for the health checkups were lower than the on-site HPLC-HbA1c levels.

Fig. 3.

Changes in the levels of on-site HPLC-HbA1c and HbA1c measured during health checkups in two subjects. Changes in the levels of on-site HPLC-HbA1c and HbA1c measured during health checkups are shown in two subjects who visited the same health screening facility (a 48-year-old male; b 52-year-old male). The facilities used for the health checkup switched the HbA1c measurement technique from the HPLC method to the enzymatic assay from February 2007, the time point marked with a down arrow. ● HbA1c measured by the HPLC method (G8; Tosoh Corporation) on-site (Hp-HPLC); ■ HbA1c measured by the HPLC method (G8; Tosoh Corporation) at the health checkup (HC-HPLC); □ HbA1c measured by the enzymatic assay at the health checkup (HC-EA)

Discussion

In the present study, we compared the EA-HbA1c levels during health checkups measured off-site after sample transfer with the estimated HPLC-HbA1c levels obtained immediately after blood sampling in a hospital. As a result, EA-HbA1c levels were approximately 0.3% lower than estimated HPLC-HbA1c with significance. These results of falsely low HbA1c levels underestimate diabetes mellitus. The possibility that diabetes mellitus cannot be diagnosed accurately based on HbA1c was suggested by the difference in HbA1c levels, depending on the measurement technique used in the present study.

HbA1c had originally been measured by the HPLC assay and HPLC assay has currently been the golden standard for HbA1c measurement. Since 1990, some techniques, such as immunoassays, affinity chromatography assays, and enzymatic assays, have been developed. The external accuracy management investigation by the Japan Medical Association showed that HbA1c levels were roughly comparable between both the techniques [4]. In papers regarding the Metabo Lead, EA-HbA1c levels were almost similar to HPLC-HbA1c levels [6]. However, EA-HbA1c levels measured during health checkups were lower than the estimated HPLC-HbA1c levels in the present study and we considered the mechanism as follows.

There are two types of health checkup, including on-site health checkups for which hematologic tests are performed by the same medical institution, such as a health screening facility, and off-site health checkups for which hematologic tests are conducted by a laboratory facility outside the institution. For an on-site health checkup, HbA1c has been measured immediately after blood sampling without transport of the samples. On the other hand, for an off-site health checkup, the collected blood samples have been transported to a laboratory facility to measure HbA1c within 1 day. Since HbA1c levels were measured for the off-site health checkup in most patients included in the present study, those with HbA1c measured on-site were excluded from this study. Are lower HbA1c levels obtained by the off-site health checkup associated with HbA1c measurement at certain time points after blood sampling? With regard to HbA1c measured by the Metabo Lead, the samples may not affect the measurement for at least 7 days under refrigeration after blood collection [7]. Therefore, lower HbA1c levels cannot be explained by the time point of the HbA1c measurement after blood sampling.

Recently, most facilities have used sodium fluoride-containing collection tubes as blood collection tubes for HbA1c to measure plasma glucose levels from the same samples. It has been known that hemolysis tends to occur in the samples due to the increase in osmolality with high sodium concentrations in the collection tubes for plasma glucose [8]. In general, when HbA1c levels are rapidly measured on-site, the correct levels can be obtained without hemolysis of the samples, whereas hemolysis may be highly likely to occur, because of the effect of vibrations while transporting samples for an off-site health checkup. Senescent erythrocytes have been known to be susceptible to the effect of hemolysis compared to young erythrocytes [8], which led to higher HbA1c levels [9]. In the HbA1c measurement by enzymatic assay, HbA1c in the blood cells has been measured by collecting samples from the erythrocyte compartment after centrifugation generally at low speed [6]. With hemolysis in the samples for the HbA1c measurement, the erythrocyte compartment after centrifugation may show higher counts of young erythrocytes with less HbA1c, resulting in low HbA1c levels. In the study by Otabe et al., HPLC-HbA1c levels were measured on-site and EA-HbA1c levels were obtained by an external laboratory facility [5], which may result in lower EA-HbA1c levels than HPLC-HbA1c levels. Similar to EA-HbA1c, there is a report that showed lower HbA1c levels measured by immunoassay (IA-HbA1c) [10]. In IA-HbA1c measurements, HbA1c in the erythrocyte compartment has also been measured after centrifugation of samples collected with a collection tube for plasma glucose as performed for EA-HbA1c. Consequently, IA-HbA1c measurements may show lower HbA1c levels with a similar mechanism to EA-HbA1c measurements.

Next, we would like to consider the strategies for accurate measurement of EA-HbA1c based on the causes discussed above. We should consider using transport containers to avoid vibrations during transportation, while using ethylenediaminetetraacetic acid (EDTA) tubes with a lower incidence of hemolysis rather than collection tubes for blood glucose with a higher tendency for hemolysis, and measuring HbA1c in whole blood without the effect of hemolysis and not HbA1c in the blood cells. Furthermore, GA, which is measurable in serum, can be used as a glycemic control marker [11]. A case–cohort study of DCCT/EDIC revealed that HbA1c and GA had similar associations to retinopathy and nephropathy, which were strengthened when both measures were considered together [12]. Furthermore, the Hisayama study in Japan has shown that GA is associated with diabetic retinopathy and carotid intima-media thickness (IMT) of carotid atherosclerosis to the same extent as HbA1c [13, 14].

There were some limitations in the present study. The first was the small sample size. Although nearly 100 individuals are sufficient for statistical analyses, further verification of findings obtained by this study should be performed in many subjects. The second was to include multiple health screening and laboratory facilities in this study. Differences in distance between the health screening and laboratory facilities may affect ΔHbA1c. Measurement errors may also be present in HbA1c between the facilities. An investigation in a single health screening facility and a single laboratory facility should be performed in the future. The third was that HPLC-HbA1c and EA-HbA1c levels from samples collected on the same day could not be measured simultaneously. HPLC-HbA1c levels measured before and after the EA-HbA1c measurement were used to estimate HPLC-HbA1c levels for the EA-HbA1c measurement; if improvement or deterioration of glycemic control was observed in a short period of time in some diabetic patients, the possibility that HPLC-HbA1c levels could not be estimated accurately cannot be ruled out. EA-HbA1c levels in non-diabetic subjects, whose plasma glucose levels are stable, were significantly lower than estimated HPLC-HbA1c levels. EA-HbA1c levels in diabetic patients were also significantly lower than estimated HPLC-HbA1c levels and Δ(EA-HbA1c − estimated HPLC-HbA1c) levels in both groups were not significantly different. Although measurement error of ΔHbA1c maybe slightly higher compared to simultaneous measurement, the number of subjects with improved glycemic control and those with deteriorated glycemic control were comparable in a short period of time. Therefore, such data may not affect the conclusion. However, further prospective studies for simultaneous measurements are warranted.

It was shown that EA-HbA1c levels measured during off-site health checkups were approximately 0.3% lower than on-site HPLC-HbA1c levels. This suggests that hemolysis is highly likely to occur from the transportation of samples collected with sodium fluoride-containing blood collection tubes, and HbA1c levels may become lower with the effect of hemolysis when HbA1c levels in the blood cells are measured after centrifugation of the samples. The causes of lower EA-HbA1c levels should be elucidated immediately to establish a strategy for the future.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflicts of interest.

Ethical approval

All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1964 and later versions.

Informed consent

Informed consent or a substitute for consent was obtained from all patients prior to participation in this study.