An isotope dilution-liquid chromatography-tandem mass spectrometry based candidate reference measurement procedure for the simultaneous quantification of 25-hydroxyvitamin D3 and 25-hydroxyvitamin D2 in human serum and plasma

Chemicals and reagents

25-Hydroxyvitamin D3 (25OHD3, CAS: 19356-17-3) and 25-hydroxyvitamin D2 (25OHD2, CAS: 21343-40-8) from MedChemExpress (MCE®, Monmouth Junction, NJ, USA) purchased via Hoelzel Biotech (Cologne, Germany) were used. The internal standard, 25-hydroxyvitamin D3-[23,24,25,26,27-13C5] in ethanol (25OHD3-13C5, CAS: 19356-17-3), was obtained from Entegris Inc – IsoSciences (Ambler, PA, USA). The internal standard, 25-hydroxyvitamin D2-[252627-13C3] (25OHD2-13C3, CAS: 1325307-59-2), was synthesized by Endotherm – Life Science Molecules (Saarbruecken, Germany). 3-Epi-25-hydroxyvitamin D3 in ethanol (CAS: 73809-05-9) and 3-Epi-25-hydroxyvitamin D2 (CAS: 908126-48-7) in ethanol were both obtained from Entegris Inc – IsoSciences. 25-Hydroxyvitamin D calibration solution (SRM 2972a [11]) and certified secondary reference material (SRM 972a [12], SRM 2970 [13]) were obtained from the National Institute of Standards and Technology (NIST, Gaithersburg, MD, USA).

Sodium carbonate (CAS: 497-19-8), sodium hydrogen carbonate (CAS: 144-55-8), ethanol absolute for analysis (CAS: 64-17-5), ethyl acetate (CAS: 141-78-6) and 2-propanol (CAS: 67-63-0) were purchased from Merck (Darmstadt, Germany). Methanol (CAS: 67-56-1) and formic acid (CAS: 64-18-6), both LC-MS grade, were obtained from Biosolve (Valkenswaard, The Netherlands).

Vitamin D depleted human serum was obtained from Roche Diagnostics GmbH (Penzberg, Germany) and from Golden West Diagnostics (Temecul, USA). Mixed gender native plasma pools (Li-heparin [CUST-BB-17022020-4C], dipotassium (K2) ethylene diamine tetraacetic acid (EDTA) [CUST-BB-17022020-4D] and tripotassium (K3) EDTA [CUST-BB-17022020-4E] were purchased from Biotrend (Cologne, Germany). The native human serum pool consisted of five samples. According to the Declaration of Helsinki, all native serum and plasma patient samples were anonymized samples.

qNMR for determining the purity of standard materials

qNMR experiments were performed on a Jeol 600 MHz NMR equipped with a Helium-cryoprobe. For both the analytes, single-pulse-¹H{¹³C}NMR was utilized for the quantitation (alkene proton; δ=6.24 ppm; 1H; tecnazene as qNMR internal standard; CDCl₃ as solvent) with an inter-scan delay of 70 s (Supplemental Material 3, Figures 1–4). All the qNMR details about impurities, quantitative resonance, and chemistry in solution have already been published [5]. Further details relevant to this manuscript are available in the Supplemental Material 2.

Additionally, we also performed quantum chemical calculations to determine the most reactive chemical bond towards H-atom transfer (HAT analysis) was also performed owing to the alleged photo-instability of the analytes. Finally, Density Functional Theory (DFT) single-point calculations were performed at the wB97X-V/def2TZVP level of theory and all the information is presented in the Supplemental Material 2. These theoretical analyses contributed in the determination of ideal quantitative resonance for both the analytes along with experimentally available data in the literature.

Preparation of matrix based calibrators and quality control samples

Matrix-based calibrators and quality controls (QCs) containing both 25OHD2 and 25OHD3 were used. Three independently weighed calibrator stock solutions were prepared for 25OHD2 and 25OHD3 using the qNMR characterized material. For each stock solution, 5 mg were weighed using an ultra-microbalance (XP6U/M, Mettler Toledo) and dissolved in 10 mL ethanol in a volumetric flask (0.500 mg/mL; 1.21 mmol/L for 25OHD2 and 1.25 mmol/L for 25OHD3). The purity of each material (25OHD2: 83.7 % ± 0.5 %, 25OHD3: 93.6 % ± 0.4 %, determined by qNMR) was taken into account for the calculation of the final stock solution concentrations. Three working solutions containing both analytes were prepared from these individual stock solutions in 50 % ethanol, all with a concentration of 10 μg/mL for both analytes (24.1 μmol/L for 25OHD2 and 24.7 μmol/L for 25OHD3). These working solutions were used to prepare spike solutions for the eight matrix matched calibrators (ranging from 1.50 to 180 ng/mL; 3.64–436 nmol/L for 25OHD2 and 3.74–449 nmol/L for 25OHD3) and three QC samples (4.50 ng/mL, 80.0 ng/mL and 150 ng/mL; 10.9, 194, 364 nmol/L for 25OHD2 and 11.2, 200, 374 nmol/L for 25OHD3). Both, calibrators and QCs, had the same concentration of 25OHD2 and 25OHD3. Vitamin D depleted human serum was used as surrogate matrix for calibrators and controls. In addition, secondary reference material (NIST SRM 2970, 25OHD2: 23.5 ng/mL, [56.9 nmol/L]; 25OHD3: 9.63 ng/mL, [24.0 nmol/L]) and one native patient sample (25OHD2: <LLMI, 25OHD3: ∼35.0 ng/mL [87.4 nmol/L]) were used to monitor the method performance over time.

Internal standard (ISTD) solution

An internal standard solution containing both 25OHD2-13C3 and 25OHD3-13C5 was prepared by dissolving 25OHD2-13C3 powder material in ethanol and further dilution to a 100 μg/mL (241 μmol/L) 25OHD2-13C3 working solution. Then, the 25OHD2-13C3 working solution and the commercially available 100 μg/mL (247 μmol/L) 25OHD3-13C5 solution were combined and further diluted with 50 % ethanol, resulting in a 165 ng/mL (397 nmol/L for 25OHD2-13C3 and 407 nmol/L for 25OHD3-13C5) ISTD spike solution.

Sample preparation

Serum and plasma matrices (lithium-heparin plasma, K₂-EDTA and K₃-EDTA) and vitamin D depleted serum can be used as sample matrix.

200 µL of sample (unknown native sample, calibrator or QC) was pipetted into a 0.5 mL brown screw cap vial. 20 µL internal standard spike solution (25OHD2-13C3 and 25OHD3-13C5 in 50 % ethanol) was added and incubated on an overhead shaker (Sarmix M2000; Sarstedt, Nümbrecht, Germany) for 15 min at room temperature. Samples were then treated with 100 µL of carbonate buffer on a thermomixer (Eppendorf 5382 thermomixer C, 10 min, 1,400 rpm), to release analytes from Vitamin D binding protein. Further purification was carried out via supported liquid extraction (Novum SLE 3 cc; Phenomenex, Aschaffenburg, Germany). The obtained extract was evaporated to dryness with an N₂ evaporator (Biotage, Uppsala, Sweden; 30 °C, 1.3 mL/min, 20 min; Biotage, Uppsala, Sweden) and reconstituted in 100 µL 80 % methanol. In a final step, the samples were filtered using centrifuge filter tubes and transferred into 1.5 mL amber sample vials with an insert. Vitamin D metabolites are sensitive to strong UV light, such as direct sunlight, as well as heat, so that precautions must be taken to avoid potential degradation.

Liquid chromatography-tandem mass spectrometry (LC-MS/MS)

The measurement of 25OHD2 and 25OHD3 were performed via a heart-cut two-dimensional LC-MS/MS method on a triple quadrupole Thermo Fisher TSQ Altis coupled to a High Pressure Liquid Chromatography Thermo Fisher Vanquish Horizon with two pumps and a two position 6-port valve. All method details (gradient, source and Multiple Reaction Monitoring [MRM] parameters) are shown in Supplementary Material 1.

10 µL of each sample were injected to a Thermo Fisher Hypersil Gold C4 column (50 × 2.1 mm, 1.9 µm) and transferred onto a Phenomenex Kinetex F5 column (150 × 2.1 mm, 2.6 µm) in the second dimension. Both analytes were transferred simultaneously via a heart-cut of 0.25 min width. Column temperature was kept at 30 °C. The HPLC flow rate was set to 0.3 mL/min in the first dimension and 0.5 mL/min in the second dimension. Mobile phases consisted of ultrapure water (A1), ultrapure water containing 0.1 % formic acid (A2) and methanol (B). Electrospray ionization was performed in positive mode. The MRM transitions for the analytes and their internal standards were: 25OHD2, m/z 413.3 → 395.3 (quantitation ion (QI)), m/z 413.3 → 271.2 (confirmation ion (CI)); 25OHD2-13C3: m/z 416.3 → 398.3 (ISTD QI), m/z 416.3 → 271.2 (ISTD CI); 25OHD3, m/z 401.3 → 383.3 (QI), m/z 401.3 → 365.3 (CI); 25OHD3-13C5, m/z 406.3 → 388.3 (ISTD QI), m/z 406.3 → 370.3 (ISTD CI). The total runtime per sample was 11 min.

Structure of analytical series, calibration and data analysis

A system suitability test (SST) was performed prior to every analytical series to examine sensitivity, retention time, chromatographic separation of the analytes and their corresponding epimer and to evaluate potential analyte carry-over. Two levels (SST1 and SST2) were prepared in methanol/water (8 + 2, v + v) both containing 25OHD2, 25OHD2-13C3, 3-Epi-25OHD2, 25OHD3, 25OHD3-13C5 and 3-Epi-25OHD3 corresponding to the concentrations of calibrator level 1 and level 8. To pass the SST, each signal-to-noise ratio (S/N) had to be ≥10 for SST1 of the respective quantifier transition of both analytes. Both epimers of the analytes are transferred to the second dimension and had to be baseline separated from 25OHD2 or 25OHD3 with a chromatographic resolution of R >1.5. To examine a potential carry-over, SST2 was injected followed by the injection of two solvent blanks. The analyte peak area observed in the first blank needed to be ≤20 % of the analyte peak area of SST1.

For each measurement sequence, a set of calibrator and quality controls were measured at the beginning and the end of the analytical series. Calibrators were measured in increasing concentrations. The calibration functions (using both injections of the 8 calibrators) were obtained by linear regression of the area ratios of the analyte and internal standard (AR=area ratio) against the analyte concentration (c_A) resulting in the function AR=a*c_A + b with 1/c_A weighting.

The sequence setup varied due to specific measurement requirements. For reference value assignment, the number of sample preparations (n=x) was dependent on the desired measurement uncertainty and samples were measured on at least two different days. Method comparison or complaint samples were prepared with n=1.

Data processing of the raw data files was done with the Thermo Fisher software Chromeleon™ Version 7.2.10. Please refer to Supplementary Material 1 for more details.

Selectivity/specificity

Selectivity was evaluated by spiking the analytes 25OHD2, 25OHD3, the corresponding internal standards 25OHD2-13C3, 25OHD3-13C5 and the interferences 3-Epi-25OHD2 and 3-Epi-25OHD3 into a native human serum pool. The chromatographic separation of 25OHD2 and 25OHD3, 3-Epi-25OHD2 and 25OHD2, as well as the separation of 3-Epi-25OHD3 and 25OHD3 was evaluated and analytes and interferences had to show a resolution R ≥1.5 in the second dimension.

Vitamin D depleted serum was spiked with the internal standards 25OHD2-13C3 and 25OHD3-13C5 and measured to determine the amount of residual unlabeled analyte in the internal standard solution. Both, the vitamin D depleted serum and the native human serum pool were checked at the expected retention time to identify possible interfering matrix signals.

Qualitative and quantitative matrix effects

Possible matrix effects were determined by a post-column infusion experiment. Therefore, a neat solution of 150 ng/mL each of 25OHD2 (364 nmol/L), 25OHD2-13C3 (361 nmol/L), 25OHD3 (374 nmol/L) and 25OHD3-13C5 (370 nmol/L) (prepared in methanol/water [1 + 1, v + v]) was infused at a constant flow rate of 10 μL/min. Processed samples (native human serum pool, methanol/water [8 + 2, v + v], vitamin D depleted serum and plasma matrices [K2-EDTA, K3-EDTA, Li-heparin]) were injected to evaluate ion enhancement or suppression effects.

In addition, a quantitative experiment based on the comparison of standard line slopes was performed. For both analytes, an ethanolic calibration solution from NIST (SRM 2972a) is available and was used for the preparation of a calibrator set. Additionally, calibrator sets were prepared in methanol/water (8 + 2, v + v), native human serum pool, vitamin D depleted serum, and lithium-heparin plasma. Although the absolute peak areas of the analytes may vary across different matrices at the same concentration, it is crucial to ensure that the ratios of analyte to internal standard and the resulting slopes do not differ. Each sample was measured once, and a comparison was made between the slopes, correlation coefficients, and confidence intervals of the slopes, which needed to overlap.

Furthermore, the neat calibration was set as standard and the calibrator levels prepared in matrix were examined as controls. Recoveries were reported as the percentage of recovery of the measured concentration relative to the nominal concentration.

Precision, trueness and accuracy

Precision of the method was assessed by performing a 5-day validation experiment, as previously described by Taibon et al. [14]. An analysis of variance (ANOVA)-based variance component analysis (VCA) was used to determine the total variability (type A uncertainty) of the method. An internal statistic program based on the VCA Roche Open Source software package in R was used [17].

Two levels of the certified secondary reference material from NIST (SRM 972a level 3 25OHD2: 13.3 ng/mL, [32.2 nmol/L]; SRM 972a level 3 25OHD3: 19.8 ng/mL, [49.4 nmol/L] and SRM 972a level 4 25OHD3: 29.4 ng/mL, [73.4 nmol/L]), three levels of spiked vitamin D depleted serum covering the measuring range of both analytes (4.50, 80.0 and 150 ng/mL; 10.9, 194, 364 nmol/L for 25OHD2 and 11.2, 200, 374 nmol/L for 25OHD3), and one native serum sample (25OHD2: approx. 48.0 ng/mL, [116 nmol/L]; 25OHD3: approx. 12.3 ng/mL, [30.7 nmol/L]) were used. All samples were prepared in triplicates for two separate measurement parts (part A and part B) and injected twice for each part, resulting in n=12 measurements per day and n=60 measurements over 5 days. For each part, individual calibrator levels were prepared.

The same samples with the exception of the native serum sample were used to determine accuracy and trueness of the method. In addition, to evaluate accuracy and trueness in plasma, lithium heparin plasma was fortified with the same concentration levels as serum samples. For the determination of the nominal concentration the concentration of residual endogenous 25OHD2 and 25OHD3 was taken into account. All samples were prepared in triplicate for each part on a single day resulting in n=6 sample preparations.

For 25-hydroxyvitamin D, the acceptance criteria for the RMP are defined as CV_ref ≤1.7 %, B_ref ≤2.6 % and TE_ref ≤5.4 %.

Lower limit of the measuring interval (LLMI)

The lower limit of the measuring interval (LLMI) was determined using a spiked sample in vitamin D depleted serum at the concentration of the lowest calibrator level (1.50 ng/mL for both analytes; 3.64 nmol/L for 25OHD2 and 3.74 nmol/L for 25OHD3). This sample was prepared fivefold on one day. Recovery, bias and precision were required to be within the range of acceptance criteria defined within the accuracy and precision experiment.

Linearity

To evaluate linearity, calibrator levels with a calibration range extended by 20 % at the lower and upper end of the range (1.20–216 ng/mL, 2.91–523 nmol/L for 25OHD2 and 3.00–539 nmol/L for 25OHD3) were prepared in triplicate in vitamin D depleted serum. Each calibrator set was prepared using individual weighings based on the described test procedure. Calibration curves were evaluated using a linear and quadratic fit, both with 1/x weighting. Suitable mathematical regression must have a correlation of determination R² ≥0.99, and the residuals had to be randomly distributed.

In addition, to confirm the linearity of the method for both analytes high concentrated native samples were serially diluted with a low concentrated native sample for 25OHD2 and vitamin D depleted serum for 25OHD3. To achieve the desired concentration for the upper limit of the measuring ranges, the high concentrated native samples were spiked with 25OHD2 and 25OHD3. Final samples were prepared by dilution following specific ratios (10 + 0, 9 + 1, 8 + 2, 7 + 3, etc.). Recoveries were reported as the measured concentration recovered relative to the nominal concentration of the sample pools.

Sample stability

Stability of processed samples stored at 6 °C in the autosampler was investigated for a period of 17 days. Therefore, four different concentration levels from the precision experiment were re-measured (n=1 for each time point). Recoveries were calculated by comparing the measured values with the nominal concentration (t=0 d).

Additionally, the stability of spike solutions was assessed by utilizing spike solutions from three calibrator levels (low, mid, high) that were stored at −80 °C for a duration of 21 weeks. These spike solutions were then employed to prepare matrix-based levels, which were quantified using a calibrator set that was spiked with freshly prepared spike solutions.

Furthermore, matrix based calibrator and QC material stored at −80 °C was evaluated for a period of 33 weeks and recoveries were determined utilizing freshly prepared calibrator levels. All samples were prepared and measured once for each time point.

Stability can be guaranteed for a measurement interval of 2–28 days (x) for x−1 day, and for a measurement interval of >4 weeks (y) for y−1 week.

Uncertainty of measurements

Measurement uncertainty was determined according to the GUM [3] and was previously described in Taibon et al. [14]. The calculation considered the following steps: purity of the qNMR characterized reference material, weighing, preparation of stock, working, spike and calibrator solutions, preparation of ISTD solution, sample preparation of the calibrators, measurement of calibrators and generation of the calibration curve, preparation of unknown samples as well as the measurement and evaluation of sample results. The uncertainty estimation for the preparation of calibrators (unc_cal) was carried out using type B evaluation. For all other aspects, such as precision (unc_prec), type A evaluation was performed.

Reference or target values were determined by averaging multiple sample preparations conducted on different days. The measurement uncertainty was obtained by combining the uncertainty in calibrator preparation (unc_cal) with the uncertainty (standard deviation) of the mean of the measurement results (unc_mean). The resulting uncertainty was then multiplied by a coverage factor of k=2 for a confidence level of 95 %, assuming a normal distribution. For more detailed information, please refer to Supplemental Material 3.

Equivalence to JCTLM listed reference measurement procedures

Equivalence to JCTLM listed reference measurement procedures was demonstrated through the quarterly participation in the CDC Vitamin D Standardization-Certification Program (VDSCP). The reference measurements for the target value assignment of samples were performed in the CDC Vitamin D Reference Laboratory using the CDC ID-LC-MS/MS Reference Method for 25-hydroxyvitamin D2 and D3 [18] and/or by Ghent University using ID-LC-MS/MS Reference Method for 25-hydroxyvitamin D2 and D3 [19]. According to the CDC scheme, sample preparation must be conducted on two different days with duplicates on each day. Individual measurement results are then reported.

Furthermore, to show comparability to JCTLM listed reference laboratories, annual participation was carried out in the RELA (IFCC External Quality Assessment Scheme for Reference Laboratories in Laboratory Medicine) scheme. For this scheme, samples were prepared in triplicate on two separate days, and the mean of the measurement results (target value) along with the corresponding measurement uncertainty were reported.

Traceability to SI units

The most important aspect of the reference measurement procedure is its traceability to the SI-unit kilogram, which has been established by utilizing Tecnazene as the qNMR ISTD which is directly traceable to the NIST PS1 (primary qNMR standard). Utilizing the above-mentioned parameters, n=6 measurements for 25-hydroxyvitamin D3 yielded an absolute content of 93.6 ± 0.8 % (k=2). For 25-hydroxyvitamin D2 an absolute content value of 83.7 ± 1.0 % (k=2) was determined with n=6 measurements.

Selectivity/specificity

Chromatographic separation of 25OHD2, 25OHD3 and their interferences 3-Epi-25OHD2 and 3-Epi-25OHD3 was achieved using a two-dimensional heart cut LC approach. The combination of Thermo Fisher Hypersil Gold C4 column and Phenomenex Kinetex F5 column achieved a satisfactory chromatographic resolution (R) of ≥1.7 for 25OHD2 and 3-Epi-25OHD2, and ≥2.0 for 25OHD3 and 3-Epi-25OHD3 in the second dimension.

The evaluation of quantifier and qualifier transition for both analytes showed no interfering signals at the respective retention time in all tested matrices. The peak area of the residual unlabeled analyte from the injection of the isotopically labeled internal standard was found to be 0.3 % for 25OHD2 and 3.6 % for 25OHD3, which are well below the ≤20 % criteria.

Qualitative and quantitative matrix effects

The qualitative post-column infusion experiment showed no significant ion enhancement or suppression at the respective retention time for 25OHD2 or 25OHD3 and their ISTDs in all tested matrices.

In the quantitative experiment that involved comparing the slopes of standard lines, slopes for 25OHD2 were found to be 0.090 (95 % confidence interval [CI] 0.088 to 0.092) for the neat solution, 0.085 (95 % CI 0.083–0.086) for NIST SRM 2972a calibration solution, 0.089 (95 % CI 0.086–0.091) for the vitamin D depleted serum, 0.094 (95 % CI 0.089–0.098) for native human serum pool and 0.087 (95 % CI 0.086–0.088) for Li-heparin plasma. The coefficients of determination were 1.00 independent of the calibration matrix. For 25OHD3 slopes were found to be 0.070 (95 % CI 0.068–0.072) for neat solution, 0.067 (95 % CI 0.66–0.068) for NIST SRM 2972a calibration solution, 0.070 (95 % CI 0.067–0.072) for vitamin D depleted serum, 0.072 (95 % CI 0.069–0.076) for native serum pool and 0.069 (95 % CI 0.067–0.071) for Li-heparin plasma. Coefficients of determination were ≥0.999 independent of the calibration matrix. The overlapping confidence intervals of the slopes indicate no significant difference between them, suggesting the absence of matrix effects.

Recoveries for the vitamin D depleted serum, the native serum matrix and the Li-heparin plasma matrix were found to range between 101 and 105 % (25OHD2), and 97 and 104 % (25OHD3), respectively.

Precision, trueness and accuracy

Accuracy and precision of the method were determined by a 5-day validation experiment. Intermediate precision was found to be less than 4.0 % for 25OHD2 and less than 3.6 % for 25OHD3. Repeatability CV ranged from 1.6 to 3.3 % (25OHD2), and 1.4–2.9 % (25OHD3), respectively, over all concentration levels (Table 1).

Accuracy and trueness were demonstrated using certified secondary reference material from NIST (SRM 972a Level 3 and 4 for 25OHD3, SRM 972a Level 3 for 25OHD2) (Figure 1). In addition, trueness was evaluated in spiked vitamin D depleted serum and Li-heparin plasma. NIST SRM 972a Level 3 was found with a mean bias (n=6 preparations) of −1.0 % for 25OHD3. NIST SRM 972a Level 4 was found with a mean bias of 0.6 % for 25OHD2 and – 1.1 % for 25OHD3. In vitamin D depleted serum the mean bias was found to range between −2.6 % and 1.0 % for 25OHD2 and between −2.5 % and 0.9 % for 25OHD3, whereas for the plasma levels the mean bias was between −4.2 % and −4.1 % for 25OHD2 and −3.9 to 0.1 % for 25OHD3. High concentration levels were diluted with vitamin D depleted serum before sample preparation. The bias ranged from −0.4 to −1.0 % (25OHD2) and −1.8 to 0.9 % (25OHD3), respectively (Table 2).

Figure 1:

LC-MS/MS analytical readouts. (A) Chromatogram of the NIST SRM 972a with a 25OHD3 concentration of 19.8 ng/mL (49.4 nmol/L). Analyte (25OHD3) left and internal standard (25OHD3-13C5) right. (B) Chromatogram of a native patient sample with a 25OHD3 concentration of 12.5 ng/mL (31.2 nmol/L). Analyte (25OHD3) left and internal standard (25OHD3-13C5) right. (C) Chromatogram of the NIST SRM 972a with a 25OHD2 concentration of 13.2 ng/mL (32.0 nmol/L). Analyte (25OHD2) left and internal standard (25OHD2-13C3) right. (B) Chromatogram of a native patient sample with a 25OHD2 concentration of 44.5 ng/mL (108 nmol/L). Analyte (25OHD2) left and internal standard (25OHD2-13C3) right.

Lower limit of the measuring interval (LLMI)

The LLMI was determined using spiked vitamin D depleted serum and corresponded to the lowest calibrator level (1.50 ng/mL for both analytes, 3.64 nmol/L for 25OHD2 and 3.74 nmol/L for 25OHD3). The mean bias (n=5 preparations) was 0.4 % and the CV was 3.5 % for 25OHD2. The evaluation for 25OHD3 showed a mean bias of 4.9 % and a CV of 1.7 %.

Linearity

Linearity was evaluated for an extended range of 20 % at the lower and upper end of the range. A linear regression model was chosen for 25OHD2 and 25OHD3 due to the random and equal distribution of residuals. The coefficients of determination were equal or better than 0.999 and 0.998, for 25OHD2 and 25OHD3, respectively. The measurement results of serially diluted samples showed a linear dependence with a correlation coefficient of 0.999 for 25OHD2, and 0.998 for 25OHD3. The recovery of the measured concentration relative to the nominal concentration ranged from 92 to 100 % for 25OHD2, and 98–106 % for 25OHD3.

Sample stability

The stability of processed samples (autosampler stability) was determined at 6 °C by re-analyzing four different levels (three spiked stripped serum levels and NIST SRM 972a Level 4) with freshly prepared calibrators. Processed samples were found to be stable for 16 days with recoveries ranging from 93 to 105 % for 25OHD2 and 95–103 % for 25OHD3.

The stability of neat spike solutions was evaluated based on three calibrator levels, which were stored at −80 °C. Vitamin D depleted serum was spiked with these solutions and the recovery was determined using freshly prepared calibrators. It was found that the neat spike solutions remained stable when stored at −80 °C for 21 weeks. Recoveries ranged from 98 to 107 % for 25OHD2 and 94–106 % for 25OHD3.

Stability of matrix-spiked calibrator and QC material were evaluated using four different levels (three spiked stripped serum levels and NIST SRM 972a Level 4). Samples were found to be stable for 33 weeks, when stored at −80 °C. Recoveries ranged from 97 to 104 % for both analytes.

Uncertainty of results

The overall uncertainties for a single measurement for 25OHD2 ranged from 2.3 to 3.9 % and from 2.4 to 3.6 % for 25OHD3 (Table 3). By multiplying the overall uncertainty by a coverage factor of k=2 (corresponding to a confidence level of 95 % assuming a normal distribution), the corresponding expanded uncertainty varied between 5.0 % and 8.1 % (25OHD2) and 5.0–7.4 % (25OHD3), respectively. These uncertainties can be reduced by performing multiple measurements.

For the reference or target value assignment, multiple sample preparations were carried out for each sample on at least 2 days. The average of these results, calculated using n=6, was used. The overall uncertainties were found to range from 0.5 to 1.7 %, with expanded uncertainties ranging from 2.1 to 3.9 % for 25OHD2. For 25OHD3, the overall uncertainties ranged from 0.7 to 1.4 %, with expanded uncertainties ranging from 2.1 to 3.2 %, as shown in Table 4.

Equivalence to JCTLM listed reference measurement procedures

Equivalence to JCTLM listed RMPs was demonstrated through the participation in the CDC Vitamin D Standardization-Certification Program as well as the RELA scheme, as illustrated in Table 5.

Over a period of a year (Q2 2023 to Q1 2024) 40 samples were measured with n=2 preparations over 2 days, resulting in n=4 preparations per sample. Sample means for total 25OHD (sum of 25OHD2 and 25OHD3) are then compared to target values (nmol/L) provided by the CDC. The Passing-Bablok regression showed a very good agreement and yielded a regression equation with a slope of 1.02 (95 % CI 1.00 to 1.05), an intercept of −0.9 (95 % CI −2.5 to −0.7) and the Bland-Altman plot with a bias of 0.6 % (95% CI -0.44 to 1.7) (Figure 2).

Figure 2:

25OHD total results (nmol/L) for the quarterly participation in the Vitamin D Standardization-Certification Program (VDSCP) provided by CDC. (A) Passing-Bablok regression analysis: Slope 1.02 (95 % CI 1.0–1.05), intercept −0.9 (95 % CI −2.5 to −0.7). (B) Bland-Altman plot mean bias 0.6 %.

In addition, the secondary reference material SRM 2970 from NIST was measured in each sequence (n=8 results). We achieved a mean 25OHD2 concentration of 23.9 ng/mL (57.9 nmol/L) with an average bias of 1.6 % and a mean 25OHD3 concentration of 9.77 ng/mL (24.4 nmol/L) with an average bias of 1.4 % (Table 6).

Furthermore, participation in the RELA scheme showed a good agreement with results from JCTLM listed laboratories. Results are shown only for 25OHD3 in ng/mL (Table 7).