Evaluation of testosterone, estradiol and progesterone immunoassay calibrators by liquid chromatography mass spectrometry

Materials

Steroid standards for measurements used Certified Reference Materials for T (M914b, National Measurement Institute, Australia (NMI), Sydney, uncertainty ±1.9 %, coverage factor of 2), E₂ (E060, 1 mg/mL E₂ solution in acetonitrile, Cerilliant, Round Rock, Texas, USA, ±3.0 %; BCR-576, ±4.4 %, BCR-577 ± 5.8 %, BCR-578 ± 5.2, European Commission’s Institute for lyophilized human serum as reference materials) and P₄ (P-069, Cerilliant, USA, ±0.6 %). Stable isotope internal standards were d3-T (D507a, NMI), d4-E₂ (DLM2487-0, Cambridge Isotope Laboratories, USA) and d9-P₄ (D5385, CDN Isotopes, Canada) for calibration and quantification. All chemical of HPLC grade were purchased from Lab-Scan Analytical Science (Dublin, Ireland) unless otherwise specified.

Sample preparation

The steroid UPLC-MS and its validation were described previously [3]. Briefly, aliquots of serum, standards and quality control samples were mixed with deuterated steroid internal standards and methyl-tert butyl ether (MTBE). The organic layer was decanted, the solvent evaporated with dried extracts re-suspended in a methanol:water mix and transferred into a microtitre plate for injection into the UPLC-MS.

UPLC-MS steroid analysis

The UPLC conditions comprised elution of steroids using a methanol/water gradient from a Kinetex Phenyl Hexyl column with a Phenomenex guard cartridge at a column temperature of 45 °C and flow rate of 0.35 mL/min and a total run time of 13 min. The chromatography provided baseline separation for each steroid with estradiol at 5.74 min (in negative ionization mode) and testosterone at 7.07 min and P4 at 10.88 min (in positive ionization mode).

The original mass spectrometry conditions comprised an API-5000 triple-quadrupole mass spectrometer (Applied Biosystems/MDS SCIEX, Foster City, CA/Concord, Ontario, Canada) equipped with an atmospheric pressure photoionization (APPI) source. When the API-5000 mass spectrometer became unavailable, the same profiling method was established on a Sciex 6500+ triple-quadrupole mass spectrometer (SCIEX, Foster City, CA/Concord, Ontario, Canada) equipped with an electrospray ionization (ESI) source operating in both positive and negative modes. The curtain gas – 40psi, GS1-70psi, GS2-60psi, temperature 600deg, IS-4500V, CAD-9. Nitrogen was employed as curtain, nebulizer, collision gases. Unit mass resolution was set in both mass-resolving quadrupole Q1 and Q3. Multiple reaction monitoring (MRM) was used to quantify the steroids in both systems. The MRM’s quantified, their dwell time (ms) and collision energies (V) were E₂ (271-> 145, 200, −57 V) and d4-E₂ (275-> 147, 100, 57) in negative ionization mode and T (289-> 109, 200, 35), d3-T (292-> 109, 100, 35) and P₄ (315-> 97, 200, 34) and d9-P₄ (324->100, 100, 34) in positive ionization mode.

The limits of detection (LOD) and limit of quantitation (LOQ) were defined according to FDA [4] and European Medicines Agency [5] guidelines (Supplementary Material, Table 1). Reproducibility was based on the coefficient of variation (%) at four quality control (QC) levels for T and P₄ and three levels for E₂ with 10 within-day measurements as well as replicate analysis of manufacturer’s calibrators as shown below. Between-day variation was measured with 15 replicates over 3 days. Extraction recovery was based on the proportion of a pre-extraction spiked sample that was recovered in the analysis and matrix effect is based on the recovery of post extraction spiked samples [6].

Long-term stability of steroid measurements in this method was established from in-house internal quality control (QC) samples for testosterone (414 runs) at high (target 8.0 ng/mL [27.8 nmol/L], 97 ± 0.3 [SEM] %), mid (target 1.6 ng/mL [5.6 nmol/L], 100 ± 0.4 %) and low (target 0.35 ng/mL [1.2 nmol/L], 99 ± 0.5 %) range, for estradiol (345 runs) at high (target 400 pg/mL [1,468 pmol/L], 103 ± 2 %), mid (target 80 pg/mL [294 pmol/L], 99 ± 1 %) and low (target 20 pg/mL [73 pmol/L], 97 ± 1 %) range and for progesterone (65 runs) at high (target 16.0 ng/mL [50.9 nmol/L], 99 α ± 0.7 %), mid (target 3.2 ng/mL [10.2 nmol/L], 98 ± 1.2 %) and low (0.8 ng/mL [2.5 nmol/L], 98 ± 1.5 %) range.

The aim of the present study is to identify whether the true steroid concentrations in the manufacturers’ calibrators differ from the assigned values by more than pre-specified amounts. The true value is best described by the CRMs used to calibrate the assays, i.e. the line of identity in the assays. To the pre-specified limits must be added the following uncertainties: uncertainty of purity of reference material; uncertainty of preparation of MS assay calibrators; within run imprecision of the assays (noting that all manufacturer’s standards for each steroid were tested in the same run), bias due to manufacturer’s calibrator matrix and the stated uncertainty of value assignment of the manufacturer’s calibrators. Obtained results will be classified as exceeding the set criteria at the 95 % confidence level after these uncertainties have been included.

Steroid immunoassay calibrators

As direct steroid immunoassays employ manufacturer supplied calibrators, we investigated the accuracy of the calibrators provided with the assays for three direct sex steroid immunoassays and compared the assigned steroid concentrations with Certified Reference Materials (CRM) for T, E₂ and P₄ measured in a validated ultra-pressure liquid chromatography-mass spectrometry method (UPLC-MS) [3] (Table 1). Calibrators were obtained for methods from the following manufacturers: Abbott Architect, Beckman-Coulter Access, Roche Cobas, Siemens Atellica (Supplementary Table 2). The steroid immunoassay calibrators were prepared according to manufacturer’s instruction and measured in as many replicates as possible from the stock vials (from 4 to 6 per calibrator standard).

Assessment of calibrator bias was performed using criteria of ±6.4 % for testosterone, ±12.5 % for estradiol with ±2.5 pg/mL [9.2 pmol/L] for values below 20 pg/mL [73 pmol/L] (both taken from the CDC HoST steroid accuracy program bias criteria https:// www.cdc.gov/labstandards/hs_standardization.html accessed 3/12/20). For progesterone a pragmatic criterion of ±10 % was used. A wider criterion of values outside ±20 % was also used for all analytes. Additionally, assessment was made relative to the stated uncertainty of the materials.

To verify the validity of steroid analysis in the matrix of the calibrators, spike-recovery experiments (6 replicates) using the certified reference materials as spike and performed in the calibrator matrix (with or without spike) using the zero calibrator, if available, otherwise another calibrator, to estimate the recovery (%) after LCMS measurement including pre-analytical solvent extraction.

Manufacturers’ descriptions of the calibrators

In addition to the lot number details of the calibrators tested, Supplementary Table 2 shows the provided traceability information from the manufacturers including the reference material used at the top of the traceability chain. Supplementary Table 3 shows the calibrator compositions with generally a limited description of the source and nature of the steroid as well as the calibrator matrix. The claimed measurement uncertainty of each level of each calibrator along with the assigned values and results is shown for the 15 T calibrators (Supplementary Table 2), 15 E₂ calibrators (Supplementary Table 3) and 12 P₄ calibrators (Supplementary Table 4) from four manufacturers (Abbott, Beckman, Roche, Siemens). Of note, the Roche calibrators have different values assigned for different lot numbers of reagents making the assigned values approximate only.

Measurement of calibrators

For T (Figure 1, Table 1), the mean of the measurement results for six of the 14 (43 %) non-zero calibrators were outside the 6.4 % bias criteria, with five passing this level at the 95 % probability level with all uncertainties included. Four were more than 20 % from the stated value and for four calibrators, their range of values excluded the assigned value. The two lowest of five Abbott non-zero calibrators were below 80 % of the expected value, the Beckman values ranged between 98 and 113 %; the two Roche calibrators were 82 and 97 % and Siemens were over 129 % of stated values.

Figure 1:

Bland-Altman deviance plot (left) and Passing-Bablok regression plot (right) for testosterone calibrators for five different commercial calibrators. The line of identity is shown as a dashed black line.

For E₂ (Figure 2, Table 2), the mean of measurement results for eight of the 14 (57 %) calibrators were outside the selected limits with six passing this level at the 95 % probability level with all uncertainties included. Seven deviated greater than 20 % from the assigned value and for nine calibrators, their range excluded the assigned value. Additionally, only four were within the claimed measurement uncertainty from the manufacturer. Abbott showed a gradual rise from 92 % to over 130 % with increasing concentration while Beckman showed a fall from over 400–106 % over the range. The higher Roche calibrator was 158 % of the assigned value.

Figure 2:

Bland-Altman deviance plot (left) and Passing-Bablok regression plot (right) for estradiol calibrators for four different commercial calibrators. The line of identity is shown as a dashed black line.

For P₄, (Figure 3, Table 3) all 11 calibrators were below the assigned value with eight of the 11 (73 %) calibrators outside 10 % of the assigned value, five with 95 % probability, and four were outside 20 %.

Figure 3:

Bland-Altman deviance plot (left) and Passing-Bablok regression plot (right) for progesterone calibrators for four different commercial calibrators. The line of identity is shown as a dashed black line.

For testosterone calibrators, the spike-recovery was 93.4 ± 1.1 % (Architect), 91.8 ± 0.7 % (Beckman), 87.4 ± 0.4 % (Roche) and 94.9 ± 0.9 % (Siemens). For estradiol calibrators, the spike-recovery was 107.5 ± 1.6 % (Architect), 95.0 ± 1.3 % (Roche) and 88.8 ± 1.3 % (Siemens). For progesterone calibrators, the spike-recovery was 95.5 ± 9.0 % (Architect), 106.2 ± 3.5 % (Beckman), 103.3 ± 7.8% (Roche) and 107.8 ± 3.5 % (Siemens). The quantitative recovery of spiked standards and uncertainty of recovery was included in assessing the statistical likelihood of exceeding the prescribed bias between the different calibrator measurement results.