Certified primary reference standard for 17-β estradiol NMIJ CRM 6004-a lot 009 from National Metrology Institute of Japan (Ibaraki, Japan) with a purity of 98.4% was used as calibrators. [²H₅]-estradiol lot B0369 (Steraloids, Newport, RI), with an isotopic purity of 99.0%, was used as an internal standard (IS). HPLC-grade methanol and hexane, as well as reagent-grade ammonium carbonate and sodium acetate were obtained from Fisher Scientific (Suwanee, GA). Anhydrous ethanol was purchased from Sigma-Aldrich (St. Louis, MO). Ammonium fluoride was obtained from Acros (Pittsburgh, PA). Steroids analyzed for interference testing (Table 1) were obtained from Steraloids (Newport, RI), LGC Standards (Manchester, NH), Cerilliant (Round Rock, TX) and Sigma-Aldrich (St. Louis, MO). Three levels of serum-based certified reference materials (BCR 576, BCR 577, and BCR 578) were purchased from the Institute for Reference Materials and Measurements (IRMM, Geel, Belgium). Fresh-frozen, single-donor serum materials were obtained from Solomon Park Research Laboratories (Kirkland, WA). Other serum materials for QC preparation and matrix evaluations were obtained from BioreclamationIVT (Westbury, NY).

The trueness and imprecision of an RMP needs to be smaller than that of routine measurement procedures to ensure target values assigned have a sufficiently low uncertainty so that any uncertainty added by a routine method can then still lead to clinically meaningful results. RMPs should have a limit that is half the limit for a routine method for imprecision and a limit that is one-third the maximum bias.^32,33 The CDC HoSt Program for E2 certifies analytical performance criteria for a routine E2 assay with a difference of ±12.5% at concentrations <73.4 pmol/L (20.0 pg/mL) and ±2.5 pg at concentrations > 73.4 pmol/L (20.0 pg/mL) from the true value and includes a suggested imprecision of 11.3%.¹⁵ The described cRMP meets the suggested performance criteria for E2 with a bias within ±2.8% and imprecision of ≤5.7%. These characteristics are similar to those of established RMPs.^22–24

A comparison of our cRMP against an established RMP using 40 samples over a concentration range of 2.61–1 050 pM (0.71–285 pg/mL) using Deming regression analysis with a variance ratio of 1 and t test, α = 0.05 found no statistically significant difference (Table 2). The agreement between the candidate and reference measurement procedure did not change when the samples were evaluated separately for low (typically found in postmenopausal women), medium (typically found in postmenopausal women and men), and high (typically found in premenopausal women) E2 concentrations. The mean percent bias of the cRMP to the established RMP for the 40 samples was 0.5% (95% CI, 0.1–0.9%; SE 0.21%), which is well within the suggested bias criterion of ±2.8%. All measurement results obtained with the IRMM serum-based certified reference materials were within ±0.9% of the assigned certified values for BCR 576, 577, and 578 and showed no significant bias when evaluated following the protocol described in NIST Special Publication 829²⁸ (Table 3). Not only does the cRMP have an overall mean bias within suggested criteria, but each evaluated individual sample met this criteria, indicating no relevant sample specific bias.

The intra-assay and interassay imprecision was less than 2.7% for all three levels of QC material tested (Table 4). The combined or total imprecision was less than 2.4%. Uncertainty was assessed at three levels as shown in Table 4 for both type A and type B factors. The relative expanded uncertainty was ≤ 4.8% at each level. Type B uncertainty was minimized through the use of primary standards with purity >98% and the use of gravimetric measurements in place of volumetric measurements in preparation of standards and samples. The imprecision and uncertainty were consistent and independent of concentrations of E2 in the cRMP, meeting performance goals across concentrations in premenopausal women, men, and postmenopausal women serum samples.

The cRMP provided high levels of accuracy and precision. This was achieved by gravimetric measurements, bracketing of calibrators, enhanced chromatographic resolution, and mass transitions specific to E2. Volumetric steps were replaced with gravimetric measurements to minimize the contributions of imprecision and inaccuracy from pipets and volume changes due to changes in temperature.³⁴ Bracketing calibration around the target concentration, a common practice for reference methods, further increased precision and accuracy.^35,36 Variability at the extremes of the large dynamic concentration range commonly observed in regular calibration curves (2.61-1 100 to pM) was minimized with bracketing into three different levels for quantitation. These adjustments were used so that the analyte and IS were close in concentration, which reduced the variability of signal intensities and measurement imprecision by reducing potential concentration-dependent ion suppression effects that would result from significantly different concentrations of IS and analyte entering the mass spectrometer.³⁷

The structural analogues and potential interfering compounds of E2 listed in Table 1 either did not contain the same mass transitions or were chromatographically separated from E2 (11.3 min). When an IS-only sample was monitored, no additional peaks were observed in the analyte signal, confirming the absence of isotopic interferences from the use of the selected internal standard.

The mass transitions used for quantitation by the cRMP were highly specific to E2 and included the precursor ion for E2 (271 m/z) as well as a fragment ion (145 m/z). Several LC-MS/MS methods for E2, as well as the previously described LC-MS/ MS RMP, use dansyl chloride as a derivatizing agent to enhance the ionization of E2 in the source and to increase the method’s sensitivity. However, with these methods the dansyl fragment ion is used to quantitate E2, thereby limiting the specificity of the method. This limitation is overcome in our method by measuring E2 without derivatization.

In addition the specificity of the mass transitions to E2 without derivatization for both the QI and CI allowed for the use of QI/CI ratios to rule out any potential interferences. For this method, the QI/CI ratios in 40 single-donor serum samples were compared to the neat calibrators. The mean QI/ CI ratio of 9 replicates of each calibrator was 1.12 (95% CI of 1.11–1.13) and mean ratio of 40 samples from 2.61 to 1 050 pM was 1.11 (95% CI of 1.11–1.12). The QI/CI ratios of all 40 serum samples were well within ±20% of the mean QI/CI ratio of calibrators, indicating that the cRMP is not affected by interfering compounds.

Chromatographic resolution was important for the cRMP to ensure any potential interfering compounds that could be identified by the QI/CI were properly resolved. An example is shown in Figure 1, where several additional peaks can be found resolved from the E2 peak at 11.3 min in the serum sample (Figure 1B) as compared to the neat sample (Figure 1A). In example, if the peaks in the 2 min window (9.3–11.3 min) in Figure 1B were not properly resolved, the concentration of the serum sample in Figure 1B, with a true E2 concentration of 26.5 pM (7.21 pg/mL), would increase by approximately 173% to 72.7 pM (19.8 pg/mL).

A phenyl/hexyl column was used to improve the separation of aromatic compounds with a 9 min gradient from 48% to 62% of buffer B to further resolve E2. The change in percent buffer B (1.6% per min) is similar to the previously published LC-MS/MS RMP²⁴ (1.4% per min) but was accomplished in less than half the time, thus increasing the throughput while still achieving the required resolution. This chromatographic approach allowed us to resolve several additional peaks that could potentially contribute to inaccurate E2 results.

Maximizing the recovery of E2 during sample preparation is important in ensuring adequate signal strength during LC-MS/MS analysis. Low recovery can affect the method’s LOD as well as precision if ample analyte and IS are not recovered, especially at the low concentrations of E2. The overall extraction efficiency of the procedure was assessed by spiking IS prior to sample preparation and injection onto the LC-MS/ MS in duplicate at a concentration of 1.02 nM (279 pg/mL) as described previously.³⁸ An extraction recovery of 90.2% (7.7% CV) was achieved. To enhance ionization and sensitivity of the method without a derivatizing agent, 0.2 mM ammonium fluoride was used in the LC buffer to promote ionization of E2 under negative ionization mode.³⁹ Sensitivity sufficient for assigning target values to sera from men and pre- and postmenopausal women was achieved and is the intended use of this method. Through this approach, the limit of detection (LOD) (at a signal-to-noise ratio (S/N) of 10) was 2.61 pM (0.71 pg/mL), which is a slight improvement over the previously published LC-MS/MS RMP and has additional specificity achieved by avoiding dansyl chloride derivatization.

For evaluation of ME, the slope and r² values were determined for each matrix and the neat calibrators as well as the mean ME % for each matrix (Table 5). On the basis of the data presented, the sample matrix effect can be evaluated by two separate observations. First, as the slopes of all matrixes and matrix free curves are identical, the absence of any significant sample ME on quantification can be confirmed. Second, the sample ME % showed either no effect or a slight enhancement for the saline and charcoal stripped matrixes evaluated (>100%). The serum samples had a ME % of 96.2% and 92.2%, indicating slight ion suppression from the matrix, as expected with increased serum volume requirements for samples with lower concentrations of E2 such as male samples. However, our evaluations have shown no significant effect of this matrix effects on the accuracy and precision of our method in both male and female samples.

Sample preparation of the cRMP plays a key role in ensuring that no interferences and no significant matrix effects exist. This is achieved in part through the selected conditions used during the isolation of the analyte from the matrix. The first LLE isolated the lipids from polar compounds. A solution of ethyl acetate/hexane at 3:2, v/v, was selected to maximize isolation of the analyte from the matrix. The addition of a basic solution (ammonium carbonate buffer, pH 8.2) to the sample extract and the second LLE allowed for the removal of acidic impurities, such as fatty acids and phospholipids. These lipids are a source of ion suppression and matrix effects, particularly in LC-MS/MS analyses.³⁷ A higher pH for the ammonium carbonate buffer would increase the removal of impurities;²⁵ however, the pH of 8.2 was selected to be two pH units below the pKA of E2 (10.5) to avoid deprotonation and subsequent decrease in recovery of E2.⁴⁰

The cRMP has been automated using the Hamilton Starlet (Reno, NV) equipped with an analytical balance. The enhancement to the method doubled the sample throughput compared to the manual sample preparation which helps to meet the increasing demand for target value assignments.

The described ID LC-MS/MS cRMP for measuring total 17-β estradiol with sufficient sensitivity for assigning target values to sera from men, and pre- and postmenopausal women, has higher specificity at similar accuracy and precision performance compared to existing RMPs. High specificity is needed to ensure accurate target value assignment, especially in samples with low E2 concentrations, where small amounts of interfering compounds can have a high impact on the measurement result. The cRMP is suitable to address the current unmet needs of clinical laboratories, assay manufacturers, and PT providers by providing value assignments to serum for calibration and verification of method performance as part of the CDC HoSt Program.