Feature Articles

Published: April 4, 2012
Find more content on:
Find Qualified IVD Industry Suppliers at Qmed:

Developing a Separation Matrix for Measuring Percent HbA1c

A recently developed technology performs the matrix binding and direct measurement of total hemoglobin, and subsequently of glycated hemoglobin, in a five-minute timeframe.

By: Ralph P. McCroskey and Lowry J. Messenger

Developing a separation matrix for measuring percent HbA1cHemoglobin A1c (HbA1c) is uniquely suited for diagnosing diabetes and monitoring its treatment with a specific number of tests per year (typically on a quarterly basis). Recent initiatives by the United Nations, the World Health Organization, and the International Diabetes Federation (IDF) have addressed the worldwide diabetes epidemic and the need for better diagnosis and treatment. At the 2010 National Glycohemoglobin Standardization Program (NGSP) Manufacturers’ Forum, the president of IDF asked manufacturers of commercial HbA1c assay systems to consider how they could provide accurate, reliable HbA1c assays to underdeveloped countries that do not have the resources to make currently marketed tests available.
Some current percent hemoglobin A1c (%HbA1c) assay methods include ion exchange or affinity column chromatography-based tests, which are expensive and not

Figure 1. Reflectance change during a lateral-flow assay.

mobile. Other methods incorporate immunoassays (requiring refrigerated storage) and/or relatively expensive cartridges and meters. A technology has now been developed that allows for the design of low-cost systems that meet the needs for diabetes testing in developing countries.
At the heart of this technology (described in U.S. patents 7195923 and 7695973) is a separation matrix that has been chemically modified to contain both negatively charged groups and boronate groups. When the pH of the buffer flowing through the matrix is acidic, positively charged glycated and non-glycated hemoglobin bind to the negatively charged groups. When the pH is basic, hemoglobin loses its charge and is released from the ionic binding. At basic pH, boronate binds not only cis-diols but also the glucose of glycated hemoglobin holding the glycated hemoglobin in the matrix. The hemoglobin bound in the matrix under each condition is quantified by taking reflectance measurements of the membrane using a glucometer-type meter. The membrane can be incorporated into simple strips that can be stored without refrigeration. Systems utilizing this technology allow for more widespread diagnosis and monitoring of diabetes in developing and developed countries.
Separation Matrix Preparation
The base membrane used for the separation matrix is a hydrophilic, permeable membrane with a pore size large enough to allow red blood cells to enter. Both the Sartobind A membrane by Sartorious and the Lateral-Flo membrane by Porex Corp. (Fairburn, GA) can be used as base membranes. Movement of the sample and buffer through the membrane can be either lateral or vertical.

Figure 2. Parallel variation in two measurements.

The work described below used the membrane in a lateral-flow configuration. The membrane thickness affects the measurement sensitivity, with a thicker membrane capable of binding more hemoglobin. A 10-mil-thick membrane did not bind enough hemoglobin to provide a suitable signal in the prototype assay system. A thickness of at least 15 mil was needed to provide adequate binding and signal. Since reflectance measurements are taken on the membrane, the membrane surface should be relatively uniform to minimize strip-to-strip variations in results. Both membranes that were tested contained covalently linked carboxylic acids groups, which function as the negatively charged groups in the technology.
Boronate groups were added to the membranes by covalently binding m aminophenylboronate (APB) to carboxyl groups already in the membrane using the carbodiimide 1-ethyl-3 (3-dimethylaminopropyl) carbodiimide (EDC). As the reaction time progresses, more boronate groups are added to the carboxyl groups on the membrane. The number of carboxyl groups is reduced, and total hemoglobin binding decreases. As the number of boronate groups increases, glycated hemoglobin bindingreaches a plateau. The reaction timing is controlled to achieve the optimal glycated binding without significant total binding loss.
Assay Details
The assay is started by adding the first buffer (pH 6.5) and taking a reference reflectance measurement. The 100% reflectance for each strip is set to the measured reflectance of a reference material. The reflectance of the membrane immediately after the strip is inserted into the meter is the logical reference to use. However, the assay precision is improved by using the reflectance of the membrane with buffer flowing through it as the reference. In one experiment, the precision of assays of a normal sample was 5.5% if the reference was the membrane alone. The precision was 2.9% when the reference was the membrane with buffer flowing through it.

Figure 4. Schematic of the test strip used in the prototype assay system.

The blood sample is added and rinsed through the membrane as the buffer continues to flow. This buffer also contains a detergent that hemolyzes the red blood cells to release the hemoglobin. At pH 6.5, non-glycated and glycated hemoglobin are positively charged and bind ionically to the negatively charged groups on the matrix. This binding is rapid and occurs as the hemoglobin moves through the matrix. At pH 6.5, the boronate groups are in a configuration that does not favor the binding of glycated hemoglobin, therefore the bound hemoglobin will contain the same proportion of glycated hemoglobin as the sample. A sufficient volume of buffer is added to remove non-bound (excess) hemoglobin from the matrix. The amount of hemoglobin bound to the matrix is measured as total hemoglobin using reflectance measurements. The light source is an LED at a nominal 430-nm wavelength, using the 415-nm absorbance peak of hemoglobin for quantitation.
After the first measurement, a buffer at pH 9.5 is added and rinsed through the membrane. The charge on the hemoglobin changes, and the ionic binding is reversed. The boronate groups change to a configuration that favors the binding of glycated hemoglobin as it is released from the negatively charged groups. The binding to the boronate groups is rapid and occurs as the hemoglobin is moving through the matrix. The resulting amount of hemoglobin that is bound to the matrix is measured as

Figure 5. Plots of ratios vs. %HbA1c for five patient samples.

glycated hemoglobin using reflectance measurements. Both binding steps are rapid and allow the matrix to be adapted to either flow-through or lateral-flow methods.
Figure 1 shows the change in reflectance (%R relative to the dry strip reflectance) as an assay progresses. When the first buffer is added (1), the reflectance drops because the matrix becomes more translucent. After the blood is added (2), the reflectance drops as the front of the high concentration of hemoglobin moves through the matrix, absorbing the light. As the non-bound hemoglobin rinses through the matrix (3), the reflectance increases to a plateau, due to the ionically bound hemoglobin remaining in the matrix, and a reflectance measurement (A) is made. When the second buffer is added (4), the reflectance initially drops and then increases as the non-glycated hemoglobin front rinses through the matrix (5). The reflectance increases to a plateau due to the glycated hemoglobin remaining in the matrix, and the second reflectance measurement (B) is made. The two reflectance measurements (A and B) are used to calculate the total and glycated hemoglobin concentrations. The ratio of glycated hemoglobin to total hemoglobin is used to calculate the %HbA1c in the sample as described below.

Figure 6. Bias plot (Scripps reference method).

The assay as described can be performed using 2-5 μL of sample. The membrane thickness and the number of hemoglobin binding sites affect the minimum sample volume. The sample tested can be capillary blood, venous blood drawn in an anticoagulant (normally EDTA), frozen venous blood drawn in anticoagulant, and controls. Using the prototype system, 85 μL of the first buffer and 55 μL of the second buffer are adequate to perform the rinses. The assay time is approximately five minutes. After adding the first buffer and reading the wet strip blank, too long of a delay in adding the blood could allow too much buffer to pass, and rinsing will not be complete. Therefore, the blood sample should be added within 15 seconds of reading the wet strip reflectance. The timing of adding the second buffer is not critical.
The assay methodology offers several advantages, including the following:

•    In the ionic binding step of the assay, not only excess hemoglobin is rinsed through the strip but also other non-binding elements in the blood sample. This removes possibly interfering substances from the measurement area before the reflectance measurements are made.

•     The hemoglobin bound with the first buffer will have non-glycated and glycated hemoglobin amounts in proportion to their concentrations in the sample. Therefore, a change in the amount of hemoglobin bound with the first buffer will result in a proportional change in glycated hemoglobin bound with the second buffer. Variations in the total hemoglobin measurement are therefore parallel to variations in the glycated hemoglobin measurement. Using the ratio of glycated to total hemoglobin to calculate %HbA1c eliminates the effect of the variations. Strip-to-strip variations due to component and manufacturing variations are therefore minimized. This can be seen in Figure 2, which shows the results from multiple assays of the same sample. The variation in the concentration of the total hemoglobin bound in the different assays is parallel to the variation in the concentration of the glycated hemoglobin. The precision results provided in Figure 2 illustrate that the imprecision of the ratios is a lot smaller than would be predicted from the total and glycated hemoglobin measurement imprecision, if these two were completely independent.

•     Both total and glycated hemoglobin are measured at the same location on the matrix using the same light source and detector. Variations between the two measurements are minimized, resulting in improved assay precision.

•     The membrane is composed of only covalently bound, negatively charged carboxyl and boronate groups, and does not contain any labile proteins or labeling compounds. The membrane is inherently stable without refrigeration, making it ideal for use in developing countries.

•     The membrane is the assay’s only active component and can be incorporated into a number of designs that can be simple and inexpensive, or more complex.
Prototype System
To demonstrate the feasibility of the technology to measure %HbA1c, it was incorporated into a simple assay system. The components of the assay system include buffer A (pH 6.5), buffer B (pH 9.5), the test strips, which are stored in a desiccated

Figure 7. Stability of %HbA1c assay results.

(silica gel) vial, an applicator for easily adding samples in a line across the membrane, and a handheld glucometer-type reflectance meter (nominal 430-nm LED; see Figure 3). The test strip holder on the meter contains a well that is positioned at the end of the membrane when the strip is inserted. Buffers are added to this well, which flow into the membrane. The test strip holder was also designed with slots to hold the two legs of the applicator and correctly position the line containing the sample onto the membrane.
A Porex Lateral-Flo membrane with added boronate groups was prepared as described above and was used in a test strip design to take advantage of the lateral-flow properties of this membrane (see Figure 4). The membrane sits on a plastic support, a reflective layer lies over the read area, and an absorbent sink collects the buffers and rinsed blood. Buffer is added at one end of the membrane through the well on the meter, and the blood sample is added just downstream from the buffer. The buffer moves the sample through the membrane to the sink at the other end. Binding occurs throughout the membrane, and reflectance measurements are taken between the blood application site and the sink through a hole in the plastic support.
Test strips were assembled as cards using a manual lamination process. Ribbons of the different components were positioned onto the plastic base, which was pre-coated with adhesive and pre-punched with holes. The components were held together using layers of double-stick adhesive tape. Each card produced 50 test strips, which were cut from the card using a guillotine cutter. Batches of up to 5000 test strips were made using this process.
Calculations
In the prototype system assay, the meter measures the reflectance of total hemoglobin and glycated hemoglobin as described above. These reflectance measurements are converted to hemoglobin concentrations, which calculate the glycated to total hemoglobin ratios. The ratios are converted to reference %HbA1c values by standardizing the ratios to the %HbA1c results of a reference method. The reference method used in the studies described below is a boronate affinity HPLC method which has had its %HbA1c test results standardized to the Diabetes Control and Complications Trial (DCCT).

Table 1. Linearity and precision.

The relationship between reflectance and concentration is non-linear and is determined experimentally for the type of membrane being used. Samples with different hemoglobin concentrations are passed through the membrane under conditions in which no hemoglobin binds. Reflectance measurements are made for each concentration as the sample flows through the measurement area. The concentration versus reflectance data are fitted with a four parameter logistic equation. This equation converts the reflectance measurements made during the assay (e.g., at points A and B in Figure 1) to their corresponding hemoglobin concentration values.
The glycated hemoglobin concentration (point B measurement) is divided by the total hemoglobin concentration (point A measurement) to obtain the glycated-hemoglobin-to-total-hemoglobin ratio. This ratio is linearly related to the HbA1c concentration measured by the reference method (see Figure 5). Figure 5 shows the data obtained from multiple assays of five patient’s blood samples. This linear relationship (slope and intercept) converts ratios to referenced %HbA1c values.
Performance of the Technology
The performance of the technology was studied using the prototype system. The study included comparing assay results to a reference method, measuring assay precision, determining assay linearity, and determining test-strip stability using temperature stress.
Results from assays of blood samples using the prototype system were compared to results of assays of the same samples using an NGSP secondary reference method. NGSP provides a means to standardize HbA1c test results to two studies that established the relationship between HbA1c concentrations and long-term problems in patients with diabetes: DCCT and the United Kingdom Prospective Diabetes Study (UKPDS). NGSP sets the criterion for comparing test results from new technologies to reference method test results, based on the bias between the two sets of results. The test method is certified when the criterion is met.
Frozen whole blood samples from forty patients were obtained from the University of Missouri, a primary reference lab for NGSP. These samples were assayed by the boronate binding HPLC method which NGSP uses as a secondary reference method. They were then assayed in duplicate using the prototype system. The samples ranged from 5% HbA1c to 11% HbA1c, with nine between
5% and 6%, seven between 6% and 7%, and twenty-four between 7% and 11%.
The least squares regression analysis of the results from the assays using the prototype system versus the results of the reference method gave a regression formula of Y = 1.21X – 0.12, in which Y is the prototype A1c and X is the reference A1c. Statistical analysis shows that the slope was 1 (P=0.394) and the intercept was 0 (P=0.397). The correlation coefficient R2 was 0.971, showing that the %HbA1c results obtained using the prototype assay demonstrate a good correlation to the results obtained using the reference method. Figure 6 shows the bias plot between the prototype assay results and the reference method assay results. The 95% confidence interval range of the bias results falls within the 2011 NGSP criterion for certification of traceability to the DCCT and UKPDS study results.
The precision of the prototype system was measured by the results of multiple assays of venous blood samples (stored at 4° C), which were conducted by two operators using two lots of strips for two days. Each operator ran eight assays each day for each lot (32 total assays for each lot). Overall results for lot one were a mean %HbA1c of 5.2 with a CV of 3.1%. The results for lot two were a mean %HbA1c of 5.4 with a CV of 2.8%. A two-way analysis of variance of the data showed there was no statistically significant day-to-day variation (P=0.238) or operator-to-operator variation (P=0.238).
The linearity of the assay was demonstrated by the results of assays of admixtures of a normal and elevated control. The results illustrated that the assay has a linear response up to at least 16% HbA1c and that the precision is consistent throughout the linear range (see Table I).
The stability of the membrane was determined by putting the test strips though an accelerated aging process and storing them at an elevated temperature of 45° C. (An earlier study showed that the ionic binding of hemoglobin to the membrane was affected by exposure of the membrane to 75% relative humidity for an extended period of time. The test strips were therefore stored in closed vials containing silica gel desiccant.) Assays of a venous blood sample that was stored frozen in aliquots were carried out at different storage intervals and used a new vial each time. Test strips that were stored at room temperature were also used to assay the blood at each storage interval as a control for possible changes in the blood samples over time. Figure 7 provides %HbA1c assay results for the different storage intervals for both sets of strips.
Compared to the day-zero results, the assay results demonstrated no evidence of test-strip decay at 45° C in measured %HbA1c for at least 365 days. The test strips stored at 45° C provided the same results as the room-temperature-stored strips at each storage interval. Results from these assays also showed that there was no decay in the total hemoglobin binding or the glycated hemoglobin binding at either temperature for 365 days. This demonstrated that the negatively charged groups and the boronate groups maintained their functionality at both temperatures during the storage time.
Preliminary results testing possible interferents indicated that hemoglobin variants HbAE, AD, AC, AS, bilirubin, and lipemia did not alter assay results. It is expected that there will be little interference by blood components since most will wash through the read area before measurements are taken. Compounds must not only bind but also absorb light at the measurement wavelength in order to interfere directly. In addition, any reduced binding of total hemoglobin would be compensated by a parallel reduction in glycated hemoglobin due to the use of the ratio to calculate %HbA1c.
Conclusion
With the use of the described binding technology, it is possible to develop a low-cost HbA1c assay system. The system can utilize a handheld reflectance meter for measurements. The chemically modified membrane is stable, which allows the system to have adequate stability without refrigeration. It is therefore possible to develop a relatively low-cost system that can be used in developing countries. The technology can also be incorporated into more sophisticated systems for point-of-care use.


Ralph McCroskey, PhD is manager of R&D at Scripps Laboratories Inc. He can be reached at mccroskey@scrippslabs.com

Lowry Messenger, PhD is director of business development at Scripps Laboratories Inc. Reach him via Philip Baddour at pbaddour@scrippslabs.com


No votes yet

Login to post comments