Development and Validation of a Rapid ¹³C₆-Glucose Isotope Dilution UPLC-MRM Mass Spectrometry Method for Use in Determining System Accuracy and Performance of Blood Glucose Monitoring Devices

Abstract

Background:

There is currently considerable discussion about the accuracy of blood glucose concentrations determined by personal blood glucose monitoring systems (BGMS). To date, the FDA has allowed new BGMS to demonstrate accuracy in reference to other glucose measurement systems that use the same or similar enzymatic-based methods to determine glucose concentration. These types of reference measurement procedures are only comparative in nature and are subject to the same potential sources of error in measurement and system perturbations as the device under evaluation. It would be ideal to have a completely orthogonal primary method that could serve as a true standard reference measurement procedure for establishing the accuracy of new BGMS.

Methods:

An isotope-dilution liquid chromatography/mass spectrometry (ID-UPLC-MRM) assay was developed using ¹³C₆-glucose as a stable isotope analogue to specifically measure glucose concentration in human plasma, and validated for use against NIST standard reference materials, and against fresh isolates of whole blood and plasma into which exogenous glucose had been spiked. Assay performance was quantified to NIST-traceable dry weight measures for both glucose and ¹³C₆-glucose.

Results:

The newly developed assay method was shown to be rapid, highly specific, sensitive, accurate, and precise for measuring plasma glucose levels. The assay displayed sufficient dynamic range and linearity to measure across the range of both normal and diabetic blood glucose levels. Assay performance was measured to within the same uncertainty levels (<1%) as the NIST definitive method for glucose measurement in human serum.

Conclusions:

The newly developed ID UPLC-MRM assay can serve as a validated reference measurement procedure to which new BGMS can be assessed for glucose measurement performance.

The ability to accurately measure blood glucose levels in patients with diabetes is extremely important for the proper management of the disease. In recent years there has been a proliferation in the number of new handheld devices to measure blood glucose at home, at the physician’s office, and at outpatient diabetes centers. Inaccuracy of blood glucose meters can substantially impact measured glucose levels, and such variability can potentially lead to adverse and often severe consequences for people with diabetes. Establishing the accuracy of glucose meters across the range of their intended use is therefore extremely important under normal everyday use in the self-management of diabetes.

Currently there is considerable discussion about the accuracy of blood glucose meters^1-2 and the standards used to determine this accuracy. Although there are different ways to measure glucose concentrations in blood, to date the most widely accepted reference measurement procedures to determine accuracy of new blood glucose monitoring systems (BGMS) are based on the very same enzymatic glucose reactions, albeit on different stand-alone instrument platforms, that is being used in the BGMS tested. The use of such similar enzyme-based glucose assays to determine absolute accuracy in glucose concentration measurement across various BGMS is thus fraught with potential sources of error on many of the same levels. Some of these sources relate to the variables that can be encountered both in the blood sample matrix itself, as well as environmental factors that can adversely affect the often immobilized-enzyme reactions, producing variable results.^3-4 In addition to the analytical measurement itself, there are preanalytical and postanalytical sources of error at both the operator level and the instrument level.^5-7 However, given that the current glucose concentration reference measurement procedures used to determine accuracy across BGMS are more appropriately to be considered “comparative” methods rather than true “primary” reference measurement procedures, we sought a nonenzymatic glucose concentration assay technique that could truly serve as an orthogonal primary reference procedure to measure glucose levels in a highly accurate manner, and which could then serve to assess the true accuracy of enzyme-based handheld glucose meters.

One such method that satisfies these criteria for a primary method is isotope dilution liquid chromatography-mass spectrometry (ID-LC/MS). ID-LC/MS is an analytical method that combines the separation capabilities of liquid chromatography with the mass analysis capabilities of mass spectrometry. LC-MS has very high sensitivity and selectivity. The LC-MS method uses a liquid chromatography system that separates molecules in a mixture such as plasma or whole blood. Mass spectrometry (MS) is an analytical technique used for determining identification and quantification of specific molecules in a mixture separated by the chromatographic procedure, most often by using 2 stages of MS referred to as tandem mass spectrometry (MS/MS). MS/MS works by ionizing chemical compounds to generate charged molecules and fragments of the parent molecule and measuring their mass-to-charge (m/z) ratios. In a typical MS/MS procedure, a sample is loaded onto the MS instrument and the components of the sample are ionized in the instrument’s source, which results in the formation of charged particles or ions. The ions are then separated according to their m/z ratio in the mass analyzer by electromagnetic fields, and detected in the mass detector based on their m/z profile, creating a high resolution mass spectrum that can be used for identification and quantification purposes.

By looking at the unique fingerprint generated by the ionization of the glucose parent molecule (m/z 179), and by quantifying the fragment ions that result by MS/MS (m/z 89 and m/z 59), one can specifically identify and quantify the amount of glucose in a complex sample like blood or plasma. Furthermore, inclusion of a heavy (stable) isotope-labeled analogue form of glucose to the original sample prior to LC-MS analysis, such as ¹³C₆-glucose which has a parent m/z of 185 and MS/MS fragments of m/z 61 and m/z 92, but has identical physical and chemical properties as the endogenous glucose analyte except for an altered mass that can be differentiated in the mass spectrometer, acts as a spiked internal standard and affords even better quantitative measurements. Thus this type of glucose assay can compensate for any endogenous analyte loss or other analyte alteration that may occur during the sample preparation or LC-MS assay. In such studies, total recovery of the endogenous glucose analyte is not essential to the assay as the determined value of the final glucose concentration is derived from the ratio of the endogenous glucose analyte and the spiked glucose heavy isotope.

Indeed, MS-based techniques have been used before in several diverse methods to detect glucose in various types of samples.^8-13 The National Institute of Standards and Technology (NIST) use an isotope-dilution gas chromatography-mass spectrometry (GC-MS) technique as its definitive method for measuring glucose in human serum.^14-15 However, the GC-MS technique requires extensive sample preparation including glucose chemical derivitization, and takes several days to complete. Previous work by others has resulted in methods to measure serum glucose levels,^10,16 and has also helped establish its correlation with GC-MS techniques,¹² similar to that used by the NIST. We have established here a ID-LC/MS/MS procedure to work with ultra-performance liquid chromatography (UPLC) which results in a quick and highly accurate and reproducible method to measure glucose levels from small amounts of original samples, similar to amounts obtained from standard finger pricks with commercial lancets. We then used this ID-LC/MS/MS method as a standard reference measurement procedure to establish the accuracy and performance standards of a new cellular-enabled handheld glucose meter from Livongo Health, the In Touch® Blood Glucose Monitoring System, in side-by-side parallel glucose measurement studies.¹⁷

Materials and Methods

Materials and Preparation of Certified D-Glucose Reference Stock Solutions

Glucose used for preparing the glucose intervals spiked into whole blood/plasma samples and for exogenous standard curves was from a USP certified reference material D-glucose stock (PHR1000; Fluka). It was weighed out on a certified analytical balance, traceable to NIST calibrated weight standards, according to dry weight and an initial reference stock solution dissolved to 100 mg/ml in water/0.1% benzoic acid by gravimetric means using the density of water-benzoic acid solvent (ρ = 1 g/cm³). This stock was then further diluted as appropriate into PBS using calibrated pipettes at the desired concentrations to make the necessary glucose intervals for whole blood spikes, as shown in the text. ¹³C₆-D-glucose [99%] used as the heavy-isotope internal standard was obtained from Cambridge Isotope Laboratories, Inc (Andover, MA; CLM-1396), and was initially weighed out and dissolved as a 100 mg/ml stock solution in water/0.1%benzoic acid as described above, and then further diluted into water to make the internal standard spike solutions. Standard reference materials SRM 909c (normal frozen human serum) and SRM 965b (glucose in frozen human serum) were purchased from the National Institute of Standards and Technology (NIST; Gaithersburg, MD). SRM 965b consisted of 4 different certified levels of glucose concentrations in human serum, and SRM 909c had a single certified reference level of glucose. All other LC reagents and solvents were LC/MS grade and purchase from either Fluka or Fisher Scientific.

LC/MS/MS Parameters Used for ID LC-MRM Analysis

Glucose measurements obtained by ID UPLC-MRM analysis were accomplished on a UPLC/MS/MS system consisting of an Acquity UPLC (Waters Corporation, Milford, MA) connected in-line with a Xevo TQ mass spectrometer (Waters Corp). Buffer A consisted of 95% acetonitrile, 2.5% methanol, 2.4% water, and 0.1% ammonium hydroxide. Buffer B consisted of 50% acetonitrile, 25% methanol, 24.9% water, and 0.1% ammonium hydroxide. The weak needle wash solution consisted of 75% acetonitrile, 12.5% methanol, 12.4% water, and 0.1% ammonium hydroxide. Of each sample, 5 μl, or standard calibration mix, was injected onto the UPLC column using the partial loop injection mode in a 20 μl loop volume using a 0.5 um in-line filter. An Acquity UPLC BEH amide (1.7 μm particle size) column of 2.1 mm × 50 mm dimension was used. Separation was accomplished under isocratic conditions with 55% solvent A/45% solvent B, at a constant flow rate of 0.13 ml/min with the column temperature maintained at 85C. Under these conditions the glucose eluted at a retention time of 2.78 ± 0.03 min. Sample inject-to-inject cycle time was 4 min. The mass spectrometer was optimized for detection of glucose in the negative ion mode using electrospray ionization. MS conditions were as follows: source temp = 150C, desolvation temp = 400C, desolvation nitrogen gas flow = 800 L/hr, capillary voltage = 2.5 kV, and collision gas (argon) = 0.15 ml/min. The spectrometer was set to operate in the MRM mode, monitoring 4 molecular transitions (2 from glucose, and 2 from ¹³C₆-glucose, with collision energy [CE] shown for each) as follows: 179.1->59.0 (CE = 16 V), 179.1->89.0 (CE = 8 V), 185.1->61.0 (CE = 16 V), and 185.1->92.0 (CE = 8 V). Dwell times for monitoring each transition were set to 0.396 sec and cone voltage to 8 V. All LC/MS operating conditions and MS/MS data collection were accomplished using MassLynx software (v4.1, Waters Corp).

Isotope-Dilution UPLC-MRM Method Validation

The UPLC-MRM method was validated for specificity, accuracy, and precision in its ability to repeatedly determine the correct NIST certified concentration values (within the expanded uncertainty limits as delineated by the NIST) for glucose at 5 different levels in human serum using NIST SRM 909c and 965b. All external glucose standard curves to which the NIST SRM materials were compared were prepared from the independent USP certified reference D-glucose stocks prepared as described above, and analyzed as described below in the Data Analysis section. Determination of the glucose concentrations in the NIST SRMs by the UPLC-MRM assay was carried out over 3 completely independent measurements, using separate and independent sample preparations and external standard curves prepared fresh for each repeated measure, with duplicate injections of each sample analyzed. Repeatability of these measurements was ascertained between duplicate injections (intraassay precision) and between the separate repeated independent analyses (interassay precision) over 2 separate days (intermediate precision).

Isolation of Whole Blood

Fresh whole venous blood was drawn into 4 ml BD Vacutainer® lithium heparin anticoagulant tubes from venous puncture of the forearm medial cubital vein, immediately before measurements and kept at room temperature

Preparation of Whole Blood, Plasma, and Glucose-Spiked Blood

Whole blood was prepared and assayed for its basal level concentration of free glucose as described below. Alternatively, individual aliquots of fresh whole blood were spiked with an additional amount of exogenous glucose from reference standard glucose stock solutions and allowed to equilibrate for 15 min at room temperature to augment the endogenous levels of free blood glucose to the levels described in the text. These whole blood standards were then further processed as described below for UPLC-MRM analysis. Plasma for UPLC-MRM analysis was obtained from fresh whole blood, or from the glucose-spiked whole blood, by centrifugation of the whole blood at 500 × g for 5 min at room temperature in a microcentrifuge tube and transferring a 10 μl portion of the plasma supernatant into a fresh microcentrifuge tube for further processing for ID UPLC-MRM as described below.

Sample Preparation Specific for ID UPLC-MRM Assays

A total of 10 μl of whole blood, whole blood glucose-added interval standards, or plasma derived from these samples was mixed with 90 μl of the internal standard (¹³C₆-glucose) solution in water in microcentrifuge tubes. Samples were then immediately mixed with 300 μl of acetonitrile and mixed thoroughly for 30 seconds and allowed to sit undisturbed for 10 min at room temperature to precipitate blood/plasma proteins and other large molecules from the samples. Samples were then centrifuged to separate the soluble small molecules (including glucose) from the precipitate at 15,000 × g for 5 min. A portion of the resulting supernatant was removed to 96 well plates for analysis. All whole blood and plasma glucose assays performed by UPLC-MRM were measured in duplicate from 3 independently processed samples on multiple days. The total process time from withdrawal of blood to loading ¹³C₆-glucose spiked plasma samples on the UPLC column and obtaining MS results was less than 15 minutes.

Data Analysis

External standard curves were established for each of the light and heavy isotope molecular transitions monitored in the MRM assay using stock solutions. Percentage recovery of glucose from the protein crash-out sample prep methodology was calculated from the recovered amount of the heavy isotope (¹³C₆-glucose) internal standard for each sample, taking into account the 99.1% isotope enrichment of the original ¹³C₆-glucose. Measured glucose amounts (ie, light isotope glucose measurements representing total endogenous free glucose concentration + any spiked amount of exogenous glucose solutions added) were normalized to the percentage recovery of the ¹³C₆-glucose for determination of absolute amount of total glucose concentration in the blood/plasma samples. Calculations of glucose amounts from processed mass spectral data was facilitated by TargetLynx software (v 4.1; Waters Corp), and all other data normalization and calculations accomplished using Excel software (Microsoft Corp, Redmond, WA). Uncertainty budget parameters were dominated by dilution pipetting issues, and were minimal. Uncertainty values shown represent the expanded uncertainty (U), where U = ku_c and where u_c represents the standard uncertainty of the mean concentration, and where the coverage factor k = 2 (~95% confidence interval). Precision was determined by analysis of the coefficient of variation.

Results and Discussion

UPLC-MRM Method Development and Validation Specificity

The LC method was optimized for rapid, high resolution chromatography using UPLC conditions for fast sample-to-sample turn-around times and high sensitivity detection of glucose. Since glucose is a very polar molecule, hydrophilic interaction liquid chromatography (HILIC) conditions were used on an amide solid-phase column to effect the separation of glucose from other matrix substances. Isocratic gradient conditions were optimized to separate glucose from other hexose sugars that may be present in blood such as fructose, mannose, and galactose. As these hexose sugars all share the same chemical composition as glucose, differing only in their structural orientation of hydroxyl groups, their MS/MS fragmentation pattern in the mass spectrometer yields very similar fragment ions. Thus coelution of hexose sugars with glucose from the LC could potentially interfere with an MRM assay. Figure 1 shows this method was capable of effectively separating other hexose sugars from glucose under the optimized UPLC conditions, with the exception of galactose. However, galactose is normally present in blood at roughly a thousand fold lower concentration than glucose,¹⁸ so it is not expected to present a significant interference issue to the assay’s glucose specificity.

Figure 1.

Separation of hexose sugars by UPLC. Mass chromatograms (base peak relative intensity shown on y-axis) showing the relative retention times of glucose LC-elution in relation to other hexose sugars. Each sugar was injected on the UPLC independently, and retention time measured under standard conditions described in the methods section. The relative retention time for each sugar is shown at the peak apex, and illustrates the chromatographic resolving power of the method.

A tandem quadrupole mass spectrometer was optimized in the negative ion multiple-reaction-monitoring (MRM) mode to selectively isolate the 179 m/z and 185 m/z negative parent ions of glucose and ¹³C₆-glucose respectively in the first quadrupole of the mass spectrometer. Quantification of the 2 most energetic MS/MS fragment ions produced in the collision cell from each of the parent ions was accomplished by setting the last quadrupole of the mass spectrometer to scan for ions of m/z 59 and 89 from glucose, and ions of m/z 61 and 92 from ¹³C₆-glucose. Thus glucose and its heavy isotope analogue are each specifically detected and quantified from every sample using 2 distinct ion transitions, increasing both the specificity and the quantitative power of the assay by using this MRM technology. Combining the time resolution of the UPLC separation, where glucose elutes with a distinct retention time of 2.78 ± 0.03 min, with the specific glucose molecular transitions monitored in the MRM mass spec method results in a highly specific LC-MRM isotope dilution method for the detection and quantification of glucose from blood, serum, or plasma.

Detection Limits/Quantitation Limits/Linearity/Range

Figure 2 shows the range and linearity of the developed UPLC-MRM assay. In determining the limits of linearity, an expanded range of glucose concentrations (0.05 mg/dl to 12 800 mg/dl) was tested in the UPLC-MRM assay to establish the limits of detection (LOD) and limits of quantitation (LOQ). Figure 2 clearly shows that the assay becomes nonlinear at both very low levels of glucose (< 0.2 mg/dl) and at very high levels of glucose (> 800 mg/dl). However, there is a range of linearity that spans approx. 3.5 logs between 0.2-800 mg/dl where there is good linearity (r² >0.998). It was determined that the LOD was 0.05 mg/dl (signal-to-noise ratio of 3) and the LOQ was 0.2 mg/dl (signal-to-noise ratio of 10; CV 2.06% at n = 8).

Figure 2.

MS-MRM method development and performance characteristics of glucose-specific MRM assay. Representative MS-MRM method development curve of a single glucose MRM transition showing an overview of the method’s performance characteristics. The MS-MRM response curve is shown for a large dynamic range of glucose concentration showing the region of linear dynamic range. The coefficient of determination (r²_log) is for the linear region shown by the dotted line. LOD and LOQ were determined from similar curves at the extreme lower end of glucose sensitivity, and are as reported in the text.

Standard curves generated for the quantification of glucose from samples are established for every UPLC-MRM assay, and a separate individual standard curve is generated for every molecular transition monitored in the MRM assay. Figure 3 shows representative curves for each of the 4 MRM transitions monitored in the assay. At least 10 different mixtures of calibration standards are used to generate the external standard curves for every UPLC-MRM assay, typically spanning 10-800 mg/dl glucose. This spans the range that blood glucose concentrations may fall within for patients with diabetes, and where blood glucose monitors would have to typically perform. The endogenous amount of glucose in the sample is determined by using the average concentration measured across both m/z 179 transitions in the sample compared to the light glucose (m/z 179) transition standard curves, while the internal standard amounts are determined in an analogous fashion relative to the ¹³C₆-glucose (m/z 185) molecular transition curves.

Figure 3.

Representative standard curves for glucose and ¹³C₆-glucose MRM-transitions. Each molecular transition monitored in the mass spectrometer is represented by a unique standard curve from which the levels of both the endogenous glucose and internal standard (¹³C₆-glucose) concentrations are determined. Correlation coefficients as shown are for the log values.

The established UPLC-MRM assay thus demonstrated high specificity and sensitivity for glucose, with suitable dynamic range and linearity over the range necessary for establishing a reference measurement procedure to which personal point-of-care BGMS could readily be assessed. However, such an intended use of the assay also requires that the UPLC-MRM assay is both highly accurate and precise.

Accuracy and Precision

Accuracy of the developed method was determined by validating that the relative MS response units derived from the calibration standards were directly proportional to the dry weight of glucose loaded on column as determined gravimetrically, and by verifying that the glucose concentrations of NIST standard reference materials of glucose in human serum, as determined by the UPLC-MRM assay, were measured to within the stated uncertainty limits delineated by the NIST. Precision of the measurements was determined by coefficient of variation analysis.

The isotope-dilution UPLC-MRM assay can serve as a primary method for a reference measurement procedure for glucose quantification, as IDMS is an accepted technique for such purposes as expressed by the Consultative Committee on the Quantity of Material (CCQM) of the International Bureau of Weights and Measures (BIPM) and the NIST.¹⁹ Primary methods have the highest metrological qualities and need to be directly traceable back to SI units. Figure 4 and Table 1 illustrate that this criteria has been met by this UPLC-MRM assay, in that the detector response of the mass spectrometer as defined by the area under the curve (AUC) for both the light and heavy isotopes of the glucose calibration standards are directly traceable back to the gravimetrically determined concentrations of each as attained by direct weight and volume (by density) measures obtained on a NIST-calibrated balance and by NIST-calibrated volumetric pipettes. The data in Figure 4 are representative of how each glucose concentration is determined in the UPLC-MRM assay and is shown, for 1 molecular transition each, from 4 representative concentrations of glucose standards. Equal amounts of both glucose and heavy ¹³C₆-glucose isotopes are mixed by weight, as determined gravimetrically, and separated by the UPLC method and detected in the mass spectrometer. Figure 4 shows that the detector response elicited in the MS is essentially the same for each glucose moiety when present in equal abundance. Table 1 lists the relative AUC response ratios for glucose/¹³C₆-glucose over a very wide range of glucose concentrations. Table 1 shows that when the areas under the curves for each of the 2 molecular transition monitored in the mass spectrometer for each of the glucose and ¹³C₆-glucose isotopes are averaged together, the mean AUC values for each are in very good agreement with each other (~0.6% by CV), and directly proportional back to the gravimetrically determined glucose amounts assayed (r²_log = .9938). These results verify that the AUC as determined from the UPLC-MRM assay can be used to accurately determine glucose concentration from samples injected onto the UPLC column with less than 1% uncertainty in measurement, and thus can serve as the basis for an accurate primary method for glucose concentration determination.

Figure 4.

LC-MRM quantification of glucose by area under the curve (AUC). Representation of how LC-MRM assay quantification is carried out by analysis of the area-under-curve from the mass chromatograms. For validation studies, the AUC was correlated with the actual dry weight of glucose and ¹³C₆-glucose calibration standards injected on column, and for detector response correlation to each other at equal weight ratios. Direct traceability in the MS detector response to the dry weight of each standard is thus established validating this method as a primary method. Blood/plasma/serum samples are quantified by similar means by comparing AUC of samples to standard curves generated from calibration mixtures. Extended verification analysis is as shown in Table 1.

Table 1.

MS Relative Response Factors of Glucose Calibration Standards From Chromatographic Area Under the Curve (AUC) Measurements.

Calibration standard mix	Glucose concentration (mg/dl)	Mean AUC glucose	Mean AUC 13C6-glucose	AUC ratio glucose/13C6-glucose
1	0.29	5717	5649	1.012
2	0.70	11 546	11 625	0.993
3	1.7	22 572	22 744	0.992
4	10	48 953	49 378	0.991
5	20	87 814	88 516	0.992
6	25	155 987	156 452	0.997
7	50	276 508	276 016	1.002
8	75	394 924	392 807	1.005
9	100	486 200	484 045	1.004
10	150	660 710	657 658	1.005
11	200	773 920	772 717	1.002
12	250	882 226	882 024	1.000
13	300	942 811	941 261	1.002
14	350	1 000 343	992 564	1.008
15	400	1 053 827	1 047 491	1.006
16	450	1 084 243	1 078 460	1.005
17	550	1 169 493	1 165 478	1.003
18	625	1 191 777	1 186 677	1.004
19	700	1 264 505	1 271 648	0.994
	Average ratio glucose/13C6-glucose			1.001
			SD	0.006
			CV	0.599%

AUC values of both glucose and ¹³C₆-glucose are averaged from the 2. MRM transitions monitored from each compound in the mass spectrometer. Glucose concentration versus mean AUC r²_log = .994.

Table 1 shows that the UPLC-MRM assay can measure glucose calibrants made up in water and/or PBS accurately; however, glucose as found in blood, serum, or plasma is surrounded by different background ion compositions. The ability of the UPLC-MRM assay to accurately measure glucose concentrations in such background matrices is extremely important if the assay is to be used as a primary method for traceability of blood glucose monitors. Since it is difficult to completely deplete serum/plasma from background levels of glucose it is not practical to develop external standard curves in a normal serum/plasma background. We therefore verified the UPLC-MRM assay’s ability to accurately measure glucose in normal human serum by measuring the glucose levels in NIST SRMs 965b and SRM 909c, relative to the glucose external calibrants made in PBS. Table 2 compares the results obtained by the UPLC-MRM assay with those of the certified NIST values obtained by the NIST definitive method (isotope-dilution GC-MS). The results show that the relative deviations of the LC-MRM measured values are typically <1% from the certified values (avg % bias 0.67), within the stated expanded uncertainty levels for the samples. Table 2 also reports the measured precision of the LC-MRM assay for both intraassay precision (repeatability across individual duplicate measures of each of the 3 independent samples), as well as the intermediate precision of the assay (interassay precision, averaged over all repeated measures taken over multiple days), as described in the methods section. The data show that the UPLC-MRM method is both highly accurate and precise at measuring glucose from human serum, and thus would be expected to perform well as a primary method for establishing a standard reference procedure to measure the accuracy and performance of BGMS.

Table 2.

ID UPLC-MRM Method Validation: NIST Certified Values Versus ID UPLC-MRM Measured Values.

		ID UPLC-MRM measured values
NIST SRMs 965b/909c	NIST certified glucose concentrationa,b,d (mg/dl)	Glucose concentrationc,d (mg/dl)	Intraassay measurements		Interassay measurements
			SD	CV (%)	SD	CV (%)	% bias
Level 1	33.08 ± 0.48	33.14 ± 0.34	0.35	0.86	0.30	0.90%	0.18
Level 2	75.56 ± 1.06	74.85 ± 1.05	0.72	0.77	0.72	0.96%	−0.94
Level 3	118.50 ± 1.70	118.14 ± 1.27	1.79	1.22	1.17	0.99%	−0.30
Level 4	294.50 ± 3.60	295.36 ± 3.11	2.18	0.59	3.10	1.05%	0.29
Level 5	90.98 ± 1.07	89.51 ± 0.86	1.15	1.02	0.69	0.77%	−1.62
	Average valuese			0.89		0.93	0.67

^aLevel 1-4 values from NIST SRM 965b certificate of analysis.

^bLevel 5 from NIST SRM 909c certificate of analysis.

^cAveraged measured values of 3 independent replicates.

^dExpanded uncertainties are shown as ± values.

^eAverage values shown for intra/interassay CVs and for (the absolute values of) % bias.

UPLC-MRM Assay Performance With Whole Blood and Plasma Spiked With Exogenous Glucose

The UPLC-MRM assay was able to measure the NIST glucose-in-serum SRMs with high accuracy and precision when measured against the external glucose calibrants made in PBS. We thus wanted to test how the assay would perform in freshly isolated whole blood and plasma spiked with exogenous amounts of glucose in a similar fashion as the NIST SRMs. This would serve to mimic the type of samples BGMS typically measure, whole blood; and to compare those readings to the plasma isolated from such whole blood samples, as most BGMS convert the whole blood glucose readings to plasma equivalents for reporting purposes. Table 3 shows the results for whole blood spiked with 3 different levels of exogenous glucose, and the plasma fraction isolated from each, and assayed independently by the UPLC-MRM procedure. The basal level of endogenous glucose concentration was measured for each starting sample (no glucose spike) and subtracted from the total glucose concentrations measured for every spiked sample to give the reported measured values in Table 3. The measured glucose concentrations in the whole blood samples gave readings ~7% below the comparable plasma glucose concentration measurements. This is consistent with what would be expected for measured differences between whole blood and plasma, depending on actual hematocrit levels.^3,4,20

Table 3.

Glucose Concentrations of Whole Blood and Plasma Spiked With Glucose.

		ID UPLC-MRM measured values
		Replicate 1			Replicate 2			Replicate 3			Interassay values
Sample	Added Glucose (mg/dl)	Glucose conc.a (mg/dl)	SD	CV (%)	Glucose conc.a (mg/dl)	SD	CV (%)	Glucose conc.a (mg/dl)	SD	CV (%)	Average glucose conc. (mg/dl)	SD	CV (%)
Blood 1	100	93.6	1.28	0.53	92.8	1.83	0.54	90.0	2.61	3.61	92.1	1.92	2.09
Blood 2	200	188.4	0.92	0.88	188.4	1.88	0.67	185.0	2.38	2.20	187.3	1.98	1.06
Blood 3	400	372.6	0.94	1.74	372.9	1.65	1.46	368.6	5.62	2.08	371.4	2.40	0.65
Plasma 1	100	101.3	2.03	0.72	102.7	1.62	0.87	98.6	1.87	1.37	100.7	1.80	1.79
Plasma 2	200	200.1	1.40	0.23	204.3	4.27	0.89	203.0	2.25	1.96	202.7	1.82	0.90
Plasma 3	400	389.7	0.50	0.53	398.3	1.60	0.33	397.0	2.46	0.80	395.0	4.64	1.17

^aThe endogenous basal concentration of glucose in each blood and plasma sample was measured for each replicate sample and subtracted from the total glucose readings of the spiked samples to determine the measured values of glucose in the samples as shown in the table.

One can see the assay performed very well overall with measured glucose values in line with the expected values from the spike solutions, and CVs typically below 2% for both the whole blood and plasma samples. This indicates that the UPLC-MRM assay has equivalently high levels of accuracy and precision measuring glucose concentration levels in whole blood and plasma matrices, as in the serum matrix as measured in the NIST SRMs.

Conclusions

A highly specific and sensitive isotope-dilution UPLC-MRM method was developed and optimized to quickly assay glucose from samples consisting of whole blood, plasma, or serum. The method was shown to act as a primary method for glucose concentration determination, and validated in terms of its accuracy, precision, linearity, dynamic range, and limits of quantitation. The assay was shown to measure plasma/serum glucose levels with equally high accuracy as the NIST definitive method involving ID-GC/MS, but involves much less time for sample preparation and no sample derivatization prior to measurement. The method can accommodate small sample amounts, such as the amount of blood derived from finger sticks, with relative ease. The assay was shown to perform equally well for measuring glucose levels from freshly isolated whole blood and plasma, and would thus be suitable for use as a validated orthogonal standard reference measurement procedure to which enzymatic-based BGMS that use electrochemical or spectrophotometric detection to measure blood/plasma glucose concentration could be compared for accuracy. This new method has been used as a truly orthogonal primary standard reference measurement procedure in confirming the accuracy, and determining the performance characteristics of a new cellular-enabled personal BGMS, the In Touch® blood glucose monitor from Livongo Health, as reported in the accompanying study.¹⁷