Development of a CDC-certified total testosterone assay for adult and pediatric samples using LC–MS/MS

Highlights

• •LC-MS assay to measure total testosterone in pediatric & adult serum/plasma samples.
• •Sensitivity of 1 ng/dL from 100 µL sample volume achieved.
• •Sample preparation includes simultaneous protein precipitation and derivatization.
• •Methods certified by the CDC Hormone Standardization program for total testosterone.

Abstract

Background

Highly accurate and sensitive method to measure testosterone in hypogonadal male, female and children is vital for proper diagnosis of hormone-related conditions and their treatment.

Objective

To develop an accurate and robust total testosterone ESI-LC-MS/MS quantification method with a simple sample preparation workflow and sufficient sensitivity for serum or plasma samples of all gender and age groups, via ketone functional group derivatization (using Amplifex™ Keto Reagent).

Method

A simple sample preparation method to accommodate both low and high numbers of samples was developed using simultaneous protein precipitation and derivatization with Amplifex™ Keto reagent, followed by centrifugation and direct injection of supernatant into an LC-MS/MS system (SCIEX Topaz™ IVD LC-MS/MS, in which MS is equivalent to a SCIEX 4500MD Mass Spectrometer). Total testosterone in human serum or plasma samples was quantified using an external calibration curve generated by calibrators spanning a broad concentration range of ∼1–2000 ng/dL (10–20,000 pg/mL), traceable to NIST 971 SRM. ¹³C₃-enriched testosterone was used as an internal standard to correct for both analyte loss during sample preparation and matrix effect during analysis (Supplementary Information: SI Fig. 4C). Two methods, one using a 96-well filter plate and another using Eppendorf tubes, were developed. Both methods were certified by the Centers for Disease Control (CDC) hormone standardization (HoSt) program for total serum testosterone. The feasibility of implementing the method for plasma and serum samples was tested via a small-scale method comparison study between matched pediatric serum and plasma samples derived from the same donor. In addition, plasma samples originating from the same donor collected in two different anticoagulant tube types (Li-heparin and K₂EDTA) were compared.

Results

Using in-house formulated NIST 971-traceable calibrators, the method was linear (r² > 0.999) between 1 and 2000 ng/dL (10 and 20,000 pg/mL) with a limit of detection of approximately 1 ng/dL (10 pg/mL). The testosterone concentration bias against 40 reference samples from the HoSt certification program was absolute <3% with an average %CV of ∼3–4%. More than 78% of samples passed the CDC bias criterion of ±6.4%. Comparison between pediatric matched serum and plasma samples resulted in high correlation (r² = 0.997) and bias of <5%. The calculated % difference between matched adult serum and plasma samples was ∼1%.

Conclusions

Feasibility for an accurate and streamlined method suitable for measuring total testosterone in all human samples was demonstrated with a choice of sample preparation workflow to suit low or high number of samples. The method can potentially be used for plasma matrix from different blood collection tubes (Li-Heparin and K₂EDTA).

1. Introduction

Clinical diagnostics lab assays should be simple, cost-effective, quick, and robust. In addition, alignment of the assay results (bias and imprecision) to a gold standard, such as the CDC reference method [1], has important implications for patient health. The increasing trend is to draw as little blood as possible from a patient, especially from pediatric and neonatal populations [2], thus driving assays to meet the sensitivity and accuracy requirements for both small sample size and the lowest reference level of the desired analyte in a population. In young adult females, testosterone levels are more than 15-fold lower than those in young adult males (2–45 ng/dL and 250–1100 ng/dL, respectively). In pediatric samples, total testosterone levels are the lowest in females and males aged 2–10 years and can be <20 ng/dL. As children reach puberty, the levels increase significantly until adulthood, but during tanner stage I the levels can be ≤5 ng/dL [3], [4]. In children, measurement of testosterone is important for identifying cases of inborn errors of sex steroid metabolism [4], [5] and delayed or precocious puberty [6], [7]. In adult females and males, low testosterone levels lead to infertility, sexual dysfunction, fatigue, loss of muscle mass, and mood swings [8], [9], [10]. The diagnosis of androgen deficiency in males and females, as well as the assessment of testosterone levels in pediatric samples, requires a highly specific, accurate, and sensitive measurement procedure [11].

Small- to medium-sized hospital labs analyzing relatively low sample volumes (<100/day) could benefit from a highly sensitive method utilizing manual sample preparation, which is easy for beginner-level operators. Simple and economical automated sample preparation with rapid analysis would be desirable for large, or core, clinical laboratories analyzing high sample volumes. The present method provides a simple, robust, and sensitive solution for all laboratories performing low-level testosterone analysis.

Though our previous testosterone analysis publication [12] was aimed to describe an ultra-high sensitivity method designed to enhance sensitivity for the quantification of free testosterone, dried blood spots, female, and/or pediatric serum samples, the method presented here is aimed to enhance simplicity and robustness. Analyte extraction and derivatization are performed simultaneously, and the incubation time during derivatization could also aid in improving the protein precipitation and testosterone extraction efficiency, as it is common practice to incubate the samples upon extraction [13]. In addition, the LC run time is shorter, and a more robust analytical column is used with larger diameter and particle size to better suit the analytical lab requirements. Last but not least, in the previous publication, testosterone analysis was performed on a higher sensitivity instrument (SCIEX QTRAP® 5500 LC-MS/MS), whereas in the current study we used SCIEX TOPAZ™ IVD LC-MS/MS System in which the mass spectrometer is a mid-range sensitivity, more affordable, equivalent to SCIEX 4500MD.

The method presented here enables quantification of total testosterone from human serum (or plasma) using a small sample volume of 100 µL with a lower limit of quantitation (LLOQ) of approximately 1 ng/dL (10 pg/mL). We describe two variations of the sample preparation workflow to accommodate small or large clinical lab requirements. The first workflow involves manual operation by which testosterone extraction is performed in an individual tube. It is designed for a smaller sample load (<100/day). The second workflow uses 96-well filter plates on an automated sample preparation platform that can also be used with manual pipetting; should an automated pipetting platform not be available. Both the individual tube and 96-well filter plate methods have been certified by the CDC Hormone Standardization program.

2. Experimental section

2.1. Chemicals and reagents

Certified standard solutions of testosterone (100 µg/mL in acetonitrile, 99.8% pure), 2,3,4-¹³C₃-testosterone (10 µg/mL in acetonitrile, 99.8%), and epi-testosterone (1 mg/mL in acetonitrile, 99.7% pure) were purchased from Cerilliant (Round Rock, TX). System suitability solution (SST) was prepared in-house from the above testosterone, epi-testosterone and 2,3,4 ¹³C₃ testosterone in MeOH/H₂O 1/1 (V/V). ZnSO₄·7H₂O was purchased from Sigma Aldrich and prepared as a 0.4 M solution with ultra-pure water (18 MΩ Elga LabWater system, USA). No further pH adjustment was necessary. Methanol, acetonitrile and formic acid, mass spectrometry grade solvents were purchased from VWR International (Radnor, PA). Bovine serum albumin (BSA, part number A2153) was purchased from Sigma Aldrich (St. Louis, MO) as lyophilized powder. Testosterone Standard Reference Material (SRM) NIST 971 was purchased from National Institute of Standards and Technology (Gaithersburg, MD).

CDC serum samples were analyzed as part of the Hormone Standardization (HoSt) certification program for serum total testosterone (CDC Atlanta, GA). Pediatric plasma (lithium heparin (LH), K₂EDTA)/serum matched samples were purchased from iSpecimen (Lexington, MA). The Amplifex™ Keto Reagent kit was obtained from SCIEX (Framingham, MA). The contents of the reagent vial and the diluent vial were mixed 1:1 (V/V) prior to use to obtain the final Amplifex™ Keto Reagent solution. Details of the derivatization procedures, chemistry, and geometric isomer formation upon derivatization are described in Star-Weinstock et al. [12].

For sample extraction, microcentrifuge tubes, 96-well filter plates, 96-well collection plates, as well as the filter plate covers were obtained from SCIEX (P/N 5045588). The LC autosampler tubes and snap caps were purchased from VWR International.

2.2. Blanks, calibrators, and controls

Calibrators and controls were prepared in-house using 5% BSA as matrix, which was determined to be free of endogenous testosterone by our assay. Testosterone working stock solution was prepared from the certified standard solution of testosterone (100 µg/mL in acetonitrile, Cerilliant). This working solution was used to spike the 5% BSA solution to make the highest concentration calibrator of ∼2000 ng/dL (∼20,000 pg/mL). The highest concentration calibrator was further diluted with 5% BSA to prepare the other calibrator concentrations (1.8, 8, 16, 59.9, 144.7, 844.2, and 1842.6 ng/dL) and quality control levels (4, 32.7, and 476.8 ng/dL). The concentrations of the calibrators and controls are traceable to NIST 971, Standard Reference Material (SRM) of testosterone in human serum matrix.

2.3. System suitability (SST) solution

The SST to monitor the performance of the LC-MS/MS system before analyzing calibrators and unknown samples was prepared as follows: A mixture of testosterone, 2,3,4-¹³C₃-testosterone, and epi-testosterone (10, 5, and 3 µg, respectively) was derivatized with 1000 µL of final Amplifex™ keto reagent solution for 120 min at ambient temperature in a microcentrifuge vial. The derivatized solution was transferred to a 100 mL volumetric flask and the volume was filled up to 100 mL using 1:1 (V/V) methanol–water, resulting in a stock solution of 0.1, 0.05, and 0.03 µg/mL in testosterone, 2,3,4-¹³C₃-testosterone, and epi-testosterone, respectively. This stock solution was further diluted 500- fold using 1:1 (V/V) methanol–water to obtain the final SST: Testosterone, 200 pg/mL (20 ng/dL); 2,3,4-¹³C₃-testosterone, 100 pg/mL (10 ng/dL); and epi-testosterone, 60 pg/mL (6 ng/dL). The SST was stored at −20 °C.

2.4. Internal standard (IS) solution

To correct for any variability in sample preparation and analysis, a methanol solution of 2,3,4-¹³C₃-testosterone (20 ng/mL, prepared by diluting the 10 µg/mL certified standard) was used as an IS. This solution was spiked into each sample, calibrator, or control before sample preparation.

3. Methods

3.1. Sample preparation

As illustrated in Fig. 1, the sample preparation may be performed using either individual tubes or a 96-well filter plate. The plate option is to allow for easy automation of high volume testing.

Fig. 1.

Sample preparation workflow that accommodates both single tube and 96-well plate methods.

Microcentrifuge vial sample preparation: The following reagents were added to a 1.5 mL microcentrifuge vial: 85 µL final Amplifex keto reagent solution, 25 µL precipitation solution (ZnSO₄ 0.4 M), and 25 µL IS solution. 100 µL of serum or plasma sample, calibrator, or control was then transferred to each individually labeled microcentrifuge vial. After vortex mixing for 15–30 s, the protein precipitated sample was incubated for 30 min at room temperature and centrifuged for 2 min at 15,000 rpm. A total of 140 µL of the supernatant was then transferred to an autosampler vial (polypropylene with 300 µL insert) for LC-MS/MS analysis.

96-well plate sample preparation: The following sample preparation steps can be performed with either manual pipetting or an automated robotic platform. The automated sample preparation in our lab was performed partially on a Topaz™ Prep Station (SCIEX, Framingham, MA) platform using a 96-well filter plate and a collection plate. The following order of reagents was dispensed manually onto the filter plate: 85 µL of Amplifex keto reagent solution, 25 µL precipitation solution, 25 µL of IS solution, and 100 µL of calibrators, controls, or patient samples. Next, the non-covered filter plate was loaded onto a specific rack of the Prep Station and the collection plate on a separate rack. An assay-specific programmed script was initiated for the robot to perform plate shaking at 600–800 rpm for 30 min at ambient temperature (20–30 °C) and centrifugation of 4000 rpm for 6 min (The robotic arm attaches the filter plate and the collection plate inside the integrated centrifuge). The collection plate (with the collected filtrate) was covered with an adhesive seal and placed in the autosampler of the LC-MS/MS system for analysis.

If an automated platform is not available, shaking and centrifugation of the filter plate can be performed on any standard 96-well plate shaker, such as the Eppendorf Thermomixer R, and a centrifuge with plate rotors, such as the Eppendorf 5430 with A-2-MTP rotor.

3.2. LC-MS/MS method

Prepared samples (prepared using either individual vials or 96-well filter plate method) were analyzed on the Topaz IVD LC-MS/MS System (4500MD Mass Spectrometer). The SST was injected in triplicate prior to analysis to qualify the LC-MS/MS system performance (Fig. 2). The injection volume was 50 µL each for the unknown sample, calibrator, control and SST. The LC gradient allowed for baseline separation of the geometric isomers of Amplifex™ ketotestosterone formed after derivatization. The HPLC mobile phases were 0.1% formic acid in water (Mobile Phase A) and 0.1% formic acid in acetonitrile (Mobile Phase B). A Phenomenex Kinetex biphenyl column (50 × 3.0 mm, 5 µm) was used at 40 °C with a gradient profile, as shown in Table 1, for analytical separation. The diverter valve on the mass spectrometer was used to bypass the excess reagent elution and LC column wash.

Fig. 2.

System Suitability Solution (SST) consisting of Amplifex™ Keto derivatized testosterone, epi-testosterone, and internal standard ¹³C₃-testosterone. Geometrical (E-Z) isomers of Amplifex Keto-testosterone oximes are separated as double peaks.

Table 1.

LC gradient profile.

Time (min)	Mobile Phase B (Percentage)	Flow rate (mL/min)	Diverter Valve switching program
0	10	0.8	At 1.0 min diversion from waste to LC column
0.5	30	0.8	To MS (Analysis)
2.5	38	0.8
2.6	95	0.8
2.7	95	1.2
3.1	95	1.2	At 3.0 min from LC column to waste
3.2	10	0.8
3.7	10	0.8

The MS/MS fragments used as quantifier and qualifier MRM transitions were 403.3 → 164.2 and 403.3 → 152.2, respectively. The transitions for the derivatized ¹³C₃-testosterone IS quantifier and qualifier were 406.3 → 167.2 and 406.3 → 155.2, respectively. The declustering potential (DP) was 80 V, collision energy (CE) was 55 and 60 eV (respectively), ion spray energy 3000 V, and the cell exit potential (CXP) 5 V. The source temperature was set to 650 °C and curtain gas (CUR) to 25 psi.

4. Data analysis and quantification

4.1. Determination of testosterone concentrations in unknown samples

Unknown sample concentrations were measured based on the calibration curve generated with the calibrators described in section 2.2. The concentration versus the area ratios between Amplifex Keto-testosterone (analyte) and Amplifex Keto-¹³C₃-testosterone (IS) was used for the linear regression analysis. Geometric isomer peaks were integrated and added for both the analyte and the IS. SCIEX Analyst® MD 1.6.2 and MultiQuant MD 3.0.2 programs were used for LC-MS/MS method optimization, data acquisition, data processing and quantification. The acceptable accuracy from the assigned value for the calibrators and controls was ±15%, except from the lowest calibrator, for which acceptable accuracy was ±20%.

The LLOQ was estimated after serial dilutions of NIST 971 female reference sample of 27.7 ng/dL (in triplicate) to the following concentrations: 6.93, 2.77, 1.39, 0.69 and 0.34 ng/dL with 5% BSA as diluent. The LLOQ criteria were %CV ≤ 15 (n = 3) and a signal-to-noise (S/N) ratio ≥ 10. In addition, the S/N ratio was evaluated for the lowest calibrator level on four different days and the concentration at which the S/N ratio would equal 10 was estimated.

Method interference studies are described in detail in a previous study from our group [12].

4.2. Method certification with CDC hormone standardization program

The CDC certification process is described in detail on the CDC website under “Laboratory Quality Assurance and Standardization Programs” [14]. Assay certification by the CDC was obtained for both sample preparation methods: the individual vials and the 96-well filter plate. First, the method was calibrated with 40 Phase 1 samples from the program and two NIST 971 SRMs. The reference concentrations of these 42 samples were compared to those determined by our method using Pearson correlation and percent bias calculation. Once the method passed the CDC acceptable bias criterion of ±6.4%, we enrolled in Phase 2 of the certification program in which 10 blinded samples were shipped to our lab for analysis every 3 months (Q1 through Q4). Our measured concentrations were reported back to the CDC for cumulative bias and imprecision estimation (bias estimation by CDC was based on CLSI guidelines EP9-A2 “Method comparison and bias estimation using patient samples”). The acceptable imprecision is ≤5.3%, and was calculated as the average % CV of all 40 samples (n = 4 for each sample). The Phase 2 program is renewed every four quarters. The data provided here are from the first certification cycle.

Method correlation (Pearson) and percent difference plots were generated using the Analyse-it software package and Excel 2016.

4.3. Comparison between matched serum and plasma samples

Using the individual vial sample preparation method, we compared the measured concentrations for serum and plasma obtained from the same individual’s blood draw (matched samples). Two separate studies were conducted. In one study, we compared matched samples from adult donors (9 female and 8 male) aged 21–65 years. Each donor’s blood collection was divided into three different tubes: (i) serum, (ii) plasma lithium heparin (LH), and (iii) plasma K₂EDTA. In this study, we compared not only the measured concentrations between serum and plasma, but also between different common plasma anticoagulation tubes, as described in the Supplementary Information (SI) Fig. 1. In the other study, we compared matched serum and plasma LH samples from 20 presumably healthy pediatric donors (10 female and 10 male) aged 7–18 years from mixed ethnicity and race.

5. Results and discussion

5.1. Method characteristics

A representative chromatogram of our lowest calibrator of 1.8 ng/dL (18 pg/mL) is shown in Fig. 3A. Based on repeated serial dilutions of NIST 971 female sample and from the S/N ratio of the lowest calibrator level over different days, the LLOQ was estimated to be 1 ng/dL (10 pg/mL). The results are shown in SI Fig. 2 and SI Table 1. The LLOQ is adequate for measuring testosterone in serum and plasma samples from female and pediatric sources for the vast majority of clinical purposes [3], [4], [15], [16]. The averaged %CV of each calibrator level over four different days ranged from 5.1 (lowest calibrator) to 2.8 (highest calibrator) as shown in SI Table 2. This simple sample preparation procedure involving simultaneous protein precipitation, analyte derivatization and extraction (without the need to dry and re-constitute the samples) is linear between 1 and 2000 ng/dL (10–20,000 pg/mL) with r² > 0.99 after 1/x weighting (Fig. 3B).

Fig. 3.

A. Representative chromatogram of lowest calibrator, 1.8 ng/dL (18 pg/mL) B. Representative calibration curve comprised of 6 concentration levels, 1.8–1843 ng/dL. The equation of the straight line was y = 1.04 + 0.00146 (r = 0.999).

As mentioned above, the calibrators and quality controls were traceable to NIST SRM 971. On each day of analysis, we added the two concentration levels of NIST 971 as an extra quality control from a different source. Fig. 4 shows a representative overlaid chromatogram of male and female NIST 971, in which testosterone concentrations are 643.5 ng/dL and 27.72 ng/dL, respectively (values obtained from the Certificate of Analysis of NIST 971).

Fig. 4.

NIST 971 Standard Reference Material (SRM). Female sample, 27.72 ng/dL; male sample, 643.5 ng/dL.

5.2. CDC HoSt certification

The first step in the process of method certification with the CDC HoSt program was to pass the acceptable criterion for the 40 value-assigned serum samples from HoSt Phase 1, which is a bias of ±6.4%. Comparing our method to the CDC reference method (Mean Difference plot, Fig. 5), the mean bias over the measurement range (1.8–1843 ng/dL) was −0.73% and the correlation coefficient (Pearson evaluation) r was 0.999 (Fig. 6), indicating high correlation and agreement with CDC reference concentrations, which enabled us to proceed to Phase 2 of the HoSt certification program for serum total testosterone.

Fig. 5.

Method comparison (mean difference plot) between CDC reference and SCIEX concentration values using 40 serum samples from CDC HoSt program for testosterone (Phase 1). The blue line of the Mean Difference plot represents the mean bias.

Fig. 6.

Pearson correlation between CDC reference and SCIEX concentration values using 40 serum samples from CDC testosterone HoSt program (phase 1). Correlation coefficient r = 0.999. The red lines define the 95% Confidence Interval.

A summary of method performance results, including a comparison with the CDC method is given in Table 2. The average absolute bias (% difference) of <3% between SCIEX measured concentrations and the CDC reference concentrations is below the accepted ±6.4%. In addition, the proportion of samples meeting CDC bias criterion [17] being >78% is another indication of high method accuracy. The reproducibility, expressed as the averaged %CV of all samples, was 3.02% and 3.94% for the individual vial and plate methods, respectively (data is shown in SI Table 3, as well as the summary of all samples %CV and %bias sorted by the gender). While the %CV of female samples was slightly higher than that of male samples due to a much lower testosterone concentration range, it remained within the acceptable value of ≤ 5.3% for both sample preparation methods.

Table 2.

Method performance after the first cycle year of HoSt Phase 2 program for total testosterone.

CDC Certification Report
	Method
Characteristic	Tube Method (Manual)	Plate Method (Automated)
LLOQ (ng/dL)	1	1
Dynamic Range (ng/dL)	1–2000	1–2000
% Bias against CDC Reference Method	Average: −1.2 Min: −11.6 Max: 4.4	Average: −2.9 Min: −11.2 Max: 7.1
95% Confidence Interval of Mean % Bias	−2.6 to 0.2	−4.2 to 1.6
Pearson Correlation Coefficient (r)	0.999	0.999
Individual Sample Pass Rate (%)	88	78
%CV Female (n = 20)	3.7 (5.2–45.8 ng/dL)	5.2 (7.3–45.8 ng/dL)
%CV Male (n-20)	2.4 (160–746 ng/dL)	2.7 (160–821 ng/dL)

Fig. 7A and B illustrate the individual sample bias throughout the cycle year for each of the sample preparation methods. Overall, individual sample concentrations were very close to the reference value, however, for each of the methods there was one quarter in which the bias with CDC reference concentrations was higher than for the rest of the year (more significantly observed for the plate method), due to a problematic calibrator set that caused the plate method to return a lower individual sample pass rate than the tube method.

Fig. 7.

Percent difference between the CDC reference and SCIEX mean concentration values (ng/dL) throughout the cycle year of the phase 2 HoSt certification program for serum total testosterone. A. Individual tube sample preparation. B. 96-well filter plate sample preparation method. The filter plate method started three months after the individual tube method, therefore, its first quarter (Q1) is concurrent with the second quarter (Q2) of the individual tube method.

5.3. Matched serum/plasma sample comparison

The CDC does not offer a parallel certification program for plasma samples. We also attempted to show the ability of our method to reliably measure testosterone in plasma samples and to test whether there is a difference in the results between different anticoagulants in plasma collection tubes. As plasma samples include fibrinogen and, therefore, have higher protein content, the concentration of an analyte in serum and plasma may differ depending on the analyte [18], [19]. As a preliminary evaluation, we measured testosterone in 37 matched serum/plasma samples from both genders and a broad age range over two experiments. Analysis of 20 pediatric samples, aged 7–18 years, indicated a high correlation between endogenous serum and plasma concentrations (Fig. 8). The measured testosterone concentrations were 1.7–934 ng/dL, and the average bias was < 5%, indicating no clinically significant difference in testosterone concentration between serum and plasma samples. Similarly, the analysis of 17 matched adult samples of serum versus plasma LH or plasma K₂EDTA resulted in comparable testosterone concentrations. For each of the 17 samples the %CV among serum, plasma LH, and plasma K₂EDTA, was ≤ 8 (n = 3), with an overall average of 4% (n = 51, 3 matched samples for each of the 17 donors, each was measured once). The results are shown in Fig. 9.

Fig. 8.

Measured testosterone concentrations of 20 matched serum/plasma pediatric samples and the corresponding correlation plot. The measured concentrations of the plasma samples are shown on the top of the bar chart.

Fig. 9.

Testosterone concentrations measured for 17 matched adult samples: Serum, plasma Li-heparin, and plasma K₂EDTA. In the X axis the sample name includes information on gender and age (e.g., F-27 originated from female, age 27 years). The displayed measured concentrations in the bar chart are from the serum samples.

In addition, looking at the percent difference among the three samples from each individual donor, the average between serum and plasma, or between plasma LH and plasma K₂EDTA was ∼ 1%, ranging from 0.04 to 11.6%. Individual sample information, as well as the measured testosterone concentrations in each sample, is shown in SI Tables 4 and 5.

Twenty pediatric serum samples, aged 2–15 years, with mixed gender and race (not plasma matched) were analyzed to produce the results shown in the SI Table 6. The measured testosterone concentrations ranged from 0.9 to 662 ng/dL. The lowest concentration was measured for a 2-year-old male and the highest for a 13-year-old male.

6. Summary and conclusions

A specific, sensitive, and robust sample preparation and LC-MS/MS method to measure testosterone was demonstrated over a broad range of sample concentrations and sample types. Our first published LC-MS/MS method for analyzing testosterone from serum [12] was intended for ultra-high sensitivity (≤0.1 ng/dL), it involved a separate derivatization step and a larger sample volume. The sample preparation for the method described here attempted to reduce time, extraction solvent, sample volume, and cost. We also eliminated the need for special sample drying equipment. The enhanced sensitivity is enabled by derivatization to attach a charged moiety to carbon 3 of testosterone using Amplifex™ Keto Reagent. However, unlike other methods that involve a separate derivatization step, in this method the derivatization reagent solution is used simultaneously for protein precipitation in one simple analyte extraction step. This method can be performed easily on an automated robotic platform with a 96-well filter plate attached to a collection plate. Both workflows, the individual vials and the 96-well filter plate (Prep Station robot), were successfully certified by the CDC HoSt program for serum total testosterone with a high accuracy score. Both methods are suitable for quantifying testosterone samples on a SCIEX Topaz IVD LC-MS/MS System (4500MD Mass Spectrometer) within a broad dynamic range of approximately 1–2000 ng/dL (10–20,000 pg/mL). Thus, the method met a level of sensitivity and accuracy that is required to reduce repeat samples due to ambiguous results and assure measurement consistency across different laboratories. The method can be used to analyze plasma samples with either LH or K₂EDTA as anticoagulant, but for plasma samples there would be a need to evaluate a larger number of matched serum/plasma samples from various anticoagulant tubes (n > 40, or as recommended in the Clinical and Laboratory Standard Institute guidelines) [20]. In addition, it would be necessary to confirm serum/plasma concentration equivalency with the 96-well filter plate workflow.

Appendix A. Supplementary data

The following are the Supplementary data to this article: