We purchased unlabeled standards for estradiol (E2), progesterone (P4), and ethinyl estradiol (EE2) from Cerilliant (Round Rock, TX) and unlabeled standards for etonogestrel (ENG), levonorgestrel (LNG), medroxprogesterone acetate (MPA), and norethindrone (NET) from Sigma (St. Louis, MO). We purchased internal standards E2-d5 and P4-d9 from Cerilliant and internal standards EE2-d7, ENG-d7, MPA-d6, LNG-d6, and NET-d6 from Toronto Research Chemicals (North York, ON, Canada). Ammonium fluoride (NH₄F) and dichloromethane (DCM) were purchased from Sigma. We purchased charcoal stripped human serum from BioChemed Services (Winchester, VA) and normal human serum from Golden West Biologicals (Temecula, CA). LC-MS grade water and methanol were purchased from Honeywell Burdick & Jackson.

150 µl of standard, QC, and serum samples were pipetted into 350 µl 96-well microtiter plates. 100 µl of ultrapure water containing 2.5 ng/ml E2-d5, 3.0 n g/ml P4-d9, 7.0 ng/ml EE2-d7, 6.0 ng/ml ENG-d7, 1.8 ng/ml MPA-d6, 1.8 ng/ml LNG-d6, and 4.0 ng/ml NET-d6 internal standards were added. A double blank of 0 ng/ml standard (150 µl) and 0 ng/ml deuterated hormones (100 µl) in ultrapure water was also prepared in duplicate. Plates were shaken to mix for 5 min and the contents of each well transferred to a 400 µl 96-well Isolute supported liquid extraction (SLE+) plate (Biotage, Uppsala, Sweden). We loaded the contents of each well into the SLE+ plate using a Pressure+ positive pressure manifold (Biotage) and incubated for 5 min at room temperature. Steroid hormones were eluted into a 2 ml glass-coated 96-well deep well plate (ThermoFisher, Waltham, MA) with 2 × 900 µl DCM and dried under forced air at 40°C in a TurboVap 96 automated evaporation system (Biotage). Plates were rinsed with 400 µl DCM and dried under forced air. After careful reconstitution in 50 µl of 25:75 methanol:water (v:v) we transferred the contents of each well to a 350 µl V-bottom 96-well microtiter plate (Shimadzu) and analyzed by LC-MS/MS. Two plates with a single curve were typically prepared each day in this manner.

For calibration curves, we prepared a mixture of unlabeled standards for E2, P4, EE2, ENG, MPA, LNG, and NET in methanol, each at a final concentration of 10 ng/ml in charcoal-stripped human serum. Two-fold serial dilutions of this mixture were made in charcoal-stripped human serum to yield the final twelve-point calibration curve, which ranged from 0.010 ng/ml to 10 ng/ml and included calibrators at 0, 0.010, 0.020, 0.039, 0.078, 0.156, 0.312, 0.625, 1.25, 2.5, 5, and 10 ng/ml. Serum standards were prepared daily immediately prior to sample preparation and methanolic stocks of each unlabeled standard were stored at −80°C. We prepared quality control (QC) samples by spiking a mixture of unlabeled hormone standards into normal human serum (Table 1). This QC was prepared and assayed in triplicate at the front and back of each assay. In addition, we generated QC pools of human serum samples for each of the synthetic progestins tested, by combining samples known to be positive for an individual hormone. Samples for both individual positive and negative controls, and pooled sera of defined HC characteristics were from the Partners PrEP Study, a clinical trial of HIV-1 pre-exposure prophylaxis that enrolled Kenyan and Ugandan participants from 2007–2012 [10]. We analyzed these individual pools for ENG, LNG, MPA, and NET in duplicate in each assay. A pool for EE2 was not created from subject samples in the Partners PrEP Study as there was insufficient sample volume with high enough levels of EE2 for pooling.

Microtiter plates were prepared as described above and loaded onto a SIL-30ACMP autosampler set at 10°C (Shimadzu). 25 µl of each standard, QC or sample were injected onto a Raptor 2.7µm Biphenyl 50mm × 2.1mm column (Restek, Bellefonte, PA) with an inline guard column (2.7 µm Biphenyl 5 mm × 2.1 mm, Restek) at 40°C using reversed phase chromatography. Mobile phase A was 0.15 mM NH₄F in water; mobile phase B was 100% methanol. The LC time gradient was created using two Nexera LC-30AD pumps (Shimadzu) as follows: 0.00–0.80 min, 70%–71% B; 0.80–3.10 min, 71%-71% B; 3.10–3.50 min, 73%–84% B; 3.50–3.60 min, 84%–94% B; 3.60–4.01 min, 94%–100% B; 4.01–5.55 min, hold at 100% B; 5.55–8.30 min, return to 70% B and hold for re-equilibration. The entire gradient was run at a flow rate of 0.25 ml/min. Heated electrospray injection in both positive (P4, ENG, LNG, MPA, Net) and negative (E2, EE2) modes with ultra-fast polarity switching and scheduled multiple reaction monitoring (MRM) on a Shimadzu LCMS-8050 was used for detection of steroids. The interface temperature was 300°C, desolvation line temperature was 150°C, and heat block temperature was 500°C. Gas was supplied by a Peak Genius 1051 nitrogen and air generator (Peak Scientific, Inchinnan, UK). Nitrogen gas was used for nebulizing and drying gases, while air was used for heating gas. Nebulizing gas flow was 3 L/min, heating gas flow was 10 L/min, and drying gas flow was 10 L/min. We used a capillary B needle in the interface and needle protrusion was set to 1.0 mm. Argon (Airgas, Radnor, PA) was used for the collision induced dissociation at 270 kPa. The MS/MS conditions for each target were optimized using the automated MRM optimization procedure in LabSolutions (Shimadzu). The MRM transitions for all steroids analyzed in this method can be found in Table 2.

We determined analyte recovery (extraction efficiency) by spiking the target hormone into separate aliquots of charcoal-stripped serum both before and after extraction, but before the dry-down step. Samples were analyzed in triplicate. We evaluated matrix effects by comparing blank serum spiked with target hormone after extraction with reference standards spiked at the same concentration into 25:75 methanol:water. These samples were analyzed in quadruplicate.

We analyzed the specificity of each assay in the method by comparing extracted blank serum (analyte-free matrix) against unlabeled standards spiked into extracted blank serum samples. For each compound at least two MRMs were monitored to ensure that no interference was present at the retention time for each analyte.

We tested assay precision by analyzing multiple replicates of QCs within each assay (intra-assay precision) and over multiple assays (inter-assay precision) (n=14). Precisions were reported as coefficients of variation (CV). We evaluated accuracy by comparing the analyzed concentration against a nominally spiked concentration of each analyte in normal human serum.

Stability after preparation was tested during the validation procedure. This was done to cover the anticipated run time for a typical analytical batch (approximately 25 h), during which prepared samples would be resident in the autosampler at 10°C. Prepared samples were kept at autosampler temperature for up to 72 h and analyzed with fresh samples for reference.

Data were analyzed using LabSolutions version 5.72 (Shimadzu). Target reference ion ratios were set according to the reference ion ratio of the highest standard. Default ion allowance for peak identification was 30% relative to this target ratio. Linear regression with 1/C weighting was used for analysis of calibration curves.

We performed the simultaneous quantitation of contraceptive steroids EE2, ENG, LNG, MPA, and NET, and endogenous steroids E2 and P4 in a single analysis using LC-MS/MS (Figure 1). Excellent linearity was observed for each target hormone within the calibration range (R>0.99). Limits of quantitation for each target can be found in Table 3 and were determined by the lowest concentration calibrator that was successfully integrated in LabSolutions with accuracy between 80–120% [11]. This resulted in lower limits of quantitation (LLOQ) for E2, EE2, and P4 at 0.010 ng/ml, for ENG, LNG, and MPA at 0.020 ng/ml, and for NET at 0.040 ng/ml. LLOQs for ENG, LNG, MPA, and NET were set more conservatively because of matrix-associated peaks that were recognized by MRMs which potentially interfered with integration at lower concentration calibrators. These interfering peaks were at different retention times than found for the hormones, but adjacent to the hormone peaks (Figure 1). The affected MRMs were 313.20→109.25, 313.20→245.25, and 313.20→79.00 for LNG; 387.20→123.10 for MPA; 325.20→147.00 for ENG; and 299.20→109.25 for NET (Figure 1). We included these MRMs in the method as they resulted in robust quantitation of the target compounds at acceptable levels, and optimized chromatographic conditions allowed for sufficient resolution and separation of these compounds from the adjacent peaks (Figure 1). In total, 36 transitions were required for all seven analytes and their internal standards over approximately 5.2 min of analysis time. An additional 3 min of equilibration for instrument stabilization after each injection was required before the next injection was initiated.

We determined the method detection limit (MDL) for each compound using a previously published method [12]. The limit of detection (LOD) and LOQ are typically indicated as the concentration at which the signal to noise ratio for a particular analyte is 1:3 (LOD) or 1:10 (LOQ). However, many sources of noise in modern MS/MS instrumentation have been eliminated, making signal to noise comparisons difficult to determine if not meaningless to LOD and LOQ determinations [13]. The MDL is defined as the minimum concentration of an analyte that can be measured and reported with 99% confidence that the analyte concentration is greater than zero. The MDL is determined from analysis of replicate standard injections in matrix at concentrations near the LOQ to evaluate the uncertainty in the system [13]. For determination of MDLs in this method we performed five replicate injections on two separate days of the lowest concentration calibrator for each analyte and calculated each MDL using the formula described by the Environmental Protection Agency [12]. MDLs for the analytes in this method can be found in Table 3. An amount of analyte equal to or greater than the MDL is both detectable and distinguishable from the background with 99% confidence [13]. We calculated specific MDLs in lieu of determining the LOD for each compound in the method.

We selected chromatographic conditions to provide good peak shapes and optimal separation for all compounds, with a focus on the need for adequate sensitivity of the estrogen compounds E2 and EE2. These requirements were met by using a biphenyl column with mobile phase consisting of 0.15 mM NH₄F in water (A) and methanol (B) paired with the gradient program described in 2.4 above. In this method, separation of the progestins from each other and potential interfering matrix peaks were critical to obtain acceptable levels of sensitivity for these compounds. We selected ammonium fluoride as an additive, as it is known to give up to 11 times more sensitivity in negative mode and 2.4 times more sensitivity in positive mode than other commonly used additives such as ammonium hydroxide [14,15], thus eliminating the need for derivatization during sample preparation to achieve appropriate sensitivity.

We performed validation of the analytical method largely according to FDA guidelines [11], by assessing the following parameters: specificity; precision, accuracy, and recovery; calibration curve; sensitivity; and reproducibility.

We used our method to analyze serum samples obtained from 29 women with self-reported use of implant (ENG or LNG; n=5), injectable (MPA; n=5) or oral HC (EE2, LNG; n=4), or women self-reporting no HC use (n=15, Table 4). We also generated sample pools from women known to be administered ENG (n=6 women), LNG (n=14 women), MPA (n=7 women) or NET (n=5 women). We applied the method to these pools (4 pools, 1 pool for each of the hormones, Table 5). Our results were similar to levels previously reported in the literature for normally cycling women or women taking the appropriate HC [7,16–19].