An isotope dilution-liquid chromatography-tandem mass spectrometry-based candidate reference measurement procedure for the quantification of cortisone in human serum and plasma

Chemicals and reagents

Cortisone (CAS 53-06-5, TraceCERT^®, lot BCBZ5582 and BCCD4087) and the internal standard (ISTD) solution, cortisone-2,3,4-¹³C₃ (¹³C₃-cortisone, CAS 53-06-5 [unlabeled]), 100 μg/mL in methanol, were obtained from Sigma Aldrich (Taufkirchen, Germany).

LC-MS grade methanol (MeOH, CAS 67-56-1) was purchased from Biosolve (Valkenswaard, The Netherlands). Ammonium fluoride (≥99.99 % trace metals basis, CAS 12125-01-8) was bought from Sigma Aldrich, ethyl acetate CAS 141-78-6 from Merck (Darmstadt, Germany) and phosphate-buffered saline (PBS) from VWR (Ismaning, Germany). Water was purified in-house using a Millipore Milli-Q^® IQ 7000 system from Merck (Darmstadt, Germany).

Steroid-free human serum (Mass Spect Gold Human Serum, ultra-low hormones & steroids) was obtained from Golden West (Temecula, California) and albumin BPLA 1 (Assay Quality, Ref. ID 117265360001) from Roche Diagnostics GmbH (Mannheim, Germany), respectively. Mixed gender native plasma pools (Lithium-heparin [Li-heparin; CUST-BB-08022022-2A], dipotassium (K₂)-ethylene diamine tetraacetic acid [EDTA; CUST-BB-17022020-4C], and tripotassium (K₃) EDTA [CUST-BB-17022020-4D]) were obtained from Biotrend (Cologne, Germany). As surrogate matrix, 60 g/L albumin in phosphate-buffered saline was used. The native human serum pool was prepared by pooling eight native serum samples. All native sample pools were prepared in accordance with the Declaration of Helsinki.

General requirements for laboratory equipment

A comprehensive list of laboratory equipment, which has been calibrated and certified by the manufacturer, along with general requirements, is available in Supplementary Material 1.

qNMR characterization of reference material

In this RMP, we utilized a commercially available qNMR-characterized SI-traceable reference material. The mass fraction value was compared using our own developed in-house qNMR method (see Supplementary Material 2).

Additionally, we also performed quantum chemical calculations to determine the most reactive chemical bond towards deprotonation. 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 calibrators and quality control (QC) samples

Two independently weighed calibration stock solutions were prepared. For each stock solution, 5 mg of cortisone was weighed using an ultra-microbalance (XP6U/M, Mettler Toledo) and dissolved in 10 mL methanol to obtain a concentration of 500 μg/mL (1,387 μmol/L) The exact concentration of the stock solutions was calculated based on the certified value from the certificate (TraceCERT^® BCCD4087, 98.4 % ± 1.2 %, k=2). Primary stock solutions were further diluted with 40 % methanol to produce three working solutions: two with a concentration of 10.0 μg/mL (27.7 μmol/L) and one with a concentration of 1.00 μg/mL (2.77 μmol/L), respectively. These working solutions were used to prepare eight calibrator spike solutions, with concentrations of 4.00–6,000 ng/mL (11.1–16,646 nmol/L), which were further diluted with albumin surrogate mix (60 g/L) to obtain eight spiked matrix calibrator solutions with concentrations ranging from 0.0800 to 120 ng/mL (0.222–333 nmol/L). An additional zero sample was prepared by combining albumin surrogate mix (60 g/L) with 40 % methanol.

A quality control (QC) stock solution was prepared by weighing 5 mg of cortisone using an ultra-microbalance (XP6U/M, Mettler Toledo) and dissolving this in 10 mL methanol, to obtain a concentration of 500 μg/mL (1,387 μmol/L). This primary stock solution was further diluted in 40 % methanol to produce two QC working solutions with concentrations of 10.0 μg/mL (27.7 μmol/L) and 1.00 μg/mL (2.77 μmol/L), respectively. These working solutions were used to prepare three spike QC solutions with concentrations of 4.00–5,900 ng/mL (11.1–16,368 nmol/L), which were further diluted with albumin surrogate mix (60 g/L) to obtain three spiked QC samples with concentrations ranging from 0.0800 to 118 ng/mL (0.222–327 nmol/L).

In addition to the three spiked QC samples, two anonymized native patient samples were used as native QC samples.

Internal standard (ISTD) solution

The commercially available ISTD ¹³C₃-cortisone (CAS 53-06-5) in methanol 100 μg/mL (275 μmol/L) was used as an ISTD stock solution. The stock solution (75 µL) was diluted with 40 % methanol to produce an ISTD working solution with a concentration of 150 ng/mL (413 nmol/L).

Sample preparation

Serum, plasma (Li-heparin plasma, K₂EDTA plasma and K₃EDTA plasma) and the surrogate matrix, were used as a sample matrix.

Samples were prepared by pipetting 100 µL of either a native sample, calibrator, or QC sample into a 0.5 mL screw-cap micro tube, followed by the addition of 20.0 µL of the ISTD working solution. These were then incubated in a thermomixer (Eppendorf 5382 thermomixer C) for 5 min at 1,400 rpm and room temperature, combined with 330 µL of Milli-Q water, and returned to the thermomixer to be shaken for 5 min at 1,400 rpm and 23 °C.

Purification was performed by loading 400 µL of the sample/water mixture onto a 3 cc Novum Support Liquid Extraction (SLE) cartridge (Phenomenex) then applying a vacuum for 5 s. Following this, the sample was allowed to equilibrate for 20 min. Elution was performed twice using 1.0 mL ethyl acetate, with the extract evaporated to dryness in a nitrogen evaporation system (Biotage TurboVap LV; 50 °C, 1.3 L/min, 15 min) and the residue reconstituted in 200 µL of 40 % methanol. The solution was filtered using 0.22 µm centrifuge filter tubes (23 °C, 15,000 rcf, 5 min) and transferred into an HPLC vial with an 0.1 mL insert.

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

A QTRAP mass spectrometer (AB Sciex QTRAP 6500⁺) coupled to an HPLC with two pumps and a two position six-port switching valve (Agilent 1290 Infinity II LC) was used for cortisone analysis.

Chromatographic separation of cortisone was achieved using two-dimensional heart-cut LC with a combination of two orthogonal stationary phases to minimize matrix effects and the co-elution of isobaric interferences. In the first dimension a Waters Acquity BEH C18 (50 × 2.1 mm; 1.7 µm) was utilized, and in the second dimension a Restek Raptor Biphenyl (150 × 2.1 mm, 2.7 µm), both with an UltraShield UHPLC PreColumn Filter, (Restek, 0.5 µm frit). In both dimensions, mobile phases consisted of 0.2 mM ammonium fluoride in Milli-Q water (A) and methanol (B). Separation at 50 °C was achieved using individual gradient programs over 10 min. HPLC flow rates were set to 300 μL/min in the first dimension and 500 μL/min in the second dimension, respectively. The injection volume was 5 µL.

Cortisone was detected by an AB Sciex QTRAP 6500+ mass spectrometer operating in multiple reaction monitoring and positive electrospray ionization modes. The quantifier ion transition (m/z 361.1 → 163.2) and corresponding ISTD transition (m/z 364.1 → 165.9) served as the base for cortisone quantification, while the additional qualifier transition (m/z 361.1 → 121.1), and corresponding ISTD transition (m/z 364.1 → 123.9) were monitored to screen for unknown interferences. LC and MS methods are fully reported in Supplementary Material 1.

System suitability test (SST)

Prior to every sequence, a system suitability test (SST) was performed to check instrument performance and to generate data for the long-term evaluation of method performance. Two sample levels (SST1 and SST2) were prepared in 40 % methanol, with cortisone at the concentrations corresponding to calibrator level 1 and calibrator level 8, respectively. Both samples contained ISTD at a concentration of 9.68 ng/mL (26.6 nmol/L).

Instrument sensitivity was evaluated based on the signal-to-noise ratio (S/N) of the SST1 analyte quantifier trace. Noise was calculated as the SD (using mean of zero) of all chromatographic data points in a retention time window of 7.20–7.70 min. To pass the SST, the S/N was required to be ≥50.

To examine potential carryover, sample SST2 was injected, followed by a solvent blank (40 % methanol). The analyte peak area observed in the first blank was required to be ≤20 % of the peak acquired during SST1 analysis to pass the SST. Retention time of the analyte was required to be 8.6 ± 0.5 min in both SST samples.

Calibration data processing and structure of analytical series

Calibrator levels were measured at increasing concentrations, to yield a calibrator block, at the beginning and the end of each sequence. The calibration function was obtained by quadratic regression, using the area ratios of the analyte and ISTD (y) against the effective analyte concentrations (x), resulting in the function y=ax² + bx + c (where a, b, and c are factors obtained by the weighted regression).

The individual sequence setup varied due to specific measurement requirements. For assignment of target values, the number of sample preparations (n=x) was dependent on the desired MU, and samples were measured on at least two different days. In case of a method comparison study, or complaint sample measurement, samples were prepared with n=1.

Data evaluation was performed using the Analyst software, with an IntelliQuant integration algorithm. Settings for automatic peak integration are reported in Supplementary Material 1.

Method validation

Assay validation was conducted following the Clinical & Laboratory Standards Institute’s C62A Liquid Chromatography-Mass Spectrometry Methods [37] and the Guide to the expression of uncertainty in measurement (GUM) [38].

Selectivity/specificity

To assess the specificity of the candidate RMP, evaluations were conducted using different matrices. The surrogate matrix contained neither cortisone nor the ISTD while the native serum pool contained low levels of endogenous cortisone. To determine whether there would be any interference with the quantifier and qualifier transitions of the analytes, cortisone and/or the ISTD were spiked to the matrices and examined at the anticipated retention time. Furthermore, the surrogate matrix was spiked with the ISTD to investigate the presence of any residual unlabeled analyte in the stable isotope-labelled ISTD material in order to confirm its purity and suitability.

To evaluate separation of cortisone and the ISTD from potential interferents, the surrogate matrix spiked with cortisone, the ISTD, cortisol, prednisone, prednisolone, 6α-hydroxycortisol, aldosterone, and α-hydroxytriazolam was analysed.

Matrix effects

In a qualitative post-column infusion setup, a solution containing 50 ng/mL (139 nmol/L) of cortisone in 2 mM ammonium fluoride in Milli-Q water/methanol (1+1, v/v) was infused into the eluent stream from the HPLC column using a T-piece at a flow rate of 10 μL/min. A neat solution composed of methanol/Milli-Q water (2+3, v+v) as well as several prepared samples were analysed using this setting: the surrogate matrix, steroid-free human serum, the native serum matrix pool and three plasma matrices (Li-heparin, K₃EDTA, K₂EDTA). The obtained chromatograms were examined on any changes in the MS signal at the anticipated retention time of cortisone and ISTD, which might indicate a possible impact of matrix components on ionization.

In addition, a comparison of standard line slopes was conducted using both calibrations in neat solvent (methanol/Milli-Q water [2+3, v+v]) as well as in matrix, respectively: the surrogate matrix, the native serum pool and the Li-heparin plasma. For the native serum and plasma matrix, the highest calibrator level was eliminated due to its nominal concentration being outside of the working range of the method, which is caused by the endogenous cortisone content of the native samples. Neat solvent calibrators were diluted to match the final concentration of processed calibrator levels, while matrix calibrator levels were prepared according to the sample preparation protocol. Calibration slopes and coefficients of determination were compared.

Furthermore, matrix-based calibrator samples were assessed as controls, with the neat calibration serving as the standard. Recoveries were then reported as percentages of the measured concentration recovered relative to the nominal concentration.

Linearity

To assess linearity, three distinct calibrator sets were prepared, each involving two weighings of the reference material. To expand the calibration range of this method (0.0800–120 ng/mL, 0.222–333 nmol/L) by ±20 %, two additional calibrator levels (0.0640 ng/mL [0.178 nmol/L] and 144 ng/mL [400 nmol/L], respectively) were prepared. The peak area ratio of the analyte to each corresponding ISTD was plotted against the respective nominal analyte concentration to determine the correlation coefficient (≥0.99) and residuals for each calibration curve.

Furthermore, the method linearity was proven by examining the recovery of serially diluted samples using the preferred regression model for calculation (n=1). In this scenario, sample 1 was a spiked steroid-free human serum, containing approximately 0.0800 ng/mL (0.222 nmol/L) cortisone while sample 11 was the native serum pool fortified to a final cortisone concentration of approximately 120 ng/mL (333 nmol/L). Subsequently, samples 2–10 were prepared by serially diluting sample 11 with sample 1 at specific ratios (e.g. 9+1 for sample 2, 8+2 for sample 3, 7+3 for sample 4, etc.). The measurement results have to show a linear dependence with a correlation coefficient of ≥0.99. Recovery was expressed as a percentage of the measured concentration recovered relative to the nominal concentration of the sample pools.

Lower limit of the measuring interval (LLMI)

The method’s precision and accuracy at the lower limit of the measuring interval (LLMI) were determined by measuring a spiked surrogate matrix sample matching the lowest calibrator level (0.0800 ng/mL cortisone, 0.222 nmol/L). This experiment was conducted over five days, with two calibrator preparations performed each day, resulting in two measurement sequences: part A and part B, respectively. For each part, the sample was prepared once and injected twice, resulting in a total of 20 measurements. Recovery, bias and precision were determined.

Precision, trueness and accuracy

Precision was assessed through a multi-day validation experiment with the aim of determining the overall variability of the method. Various sources of variability, including between-injection, between-preparation, between-calibration and between-day variability, were evaluated using an ANOVA-based variance components approach.

Three spiked samples (0.240 ng/mL [0.666 nmol/L] cortisone in surrogate matrix, 80.0 ng/mL [222 nmol/L] cortisone in steroid-free serum and 118 ng/mL [327 nmol/L] cortisone in steroid-free serum, respectively) as well as two native patient samples containing approximately 10.0 ng/mL (27.7 nmol/L) and 32.0 ng/mL (88.8 nmol/L) cortisone, respectively, were included in the evaluation. The experiment was performed on five experimental days, with an individual calibrator set being prepared for each of part A and part B, respectively. Each sample was prepared in triplicate for each part A and part B, respectively, and subjected to two injections each, resulting in a total of 12 measurements per day and 60 measurements across the five-day timeframe. The sample measurement sequence followed a predetermined order, commencing with calibrators, followed by spiked samples and native samples, and concluding with calibrators.

Data obtained from the experiment were evaluated using Biowarp, an internal statistical program based on the VCA Roche Open Source software package in R [39]. The evaluation focused on assessing repeatability (variability between injections and between preparations) as well as intermediate precision (between-calibration and between-day variability). These measures were expressed as the SD and coefficient of variation (CV).

It is important to highlight that the independent production of calibrators was evaluated separately as type B uncertainty. This evaluation involved individual uncertainties resulting from preparation of stock and working solutions, spike solutions and matrix-based calibrator levels. For more comprehensive information, please refer to Supplementary Material 3.

To assess accuracy, spiked samples at cortisone concentrations of 0.240 ng/mL (0.666 nmol/L) in surrogate matrix, 80.0 ng/mL (222 nmol/L) in steroid-free serum, and 118 ng/mL (327 nmol/L) in steroid-free serum, were utilized, respectively. In addition, Li-heparin plasma samples with an endogenous cortisone concentration were fortified with concentrations of 0.240, 10.0, and 80.0 ng/mL (0.666, 27.7, and 222 nmol/L) cortisone, respectively. The validity of the dilution protocol was evaluated using a native serum matrix pool fortified with 118 ng/mL (327 nmol/L) and 200 ng/mL (555 nmol/L) cortisone, respectively.

The experiment was performed on one day with individual calibrator preparations for part A and part B, respectively. For each part, samples were prepared in triplicate. Accuracy was determined by calculating the percentage recovery of the measured concentration compared to the nominal concentration. Trueness was reported as percentage recovery of the mean measured concentration relative to the spiked concentration.

Sample stability

The stability of processed samples was assessed at a temperature of 8 °C. To accomplish this, three calibrator levels from across the measuring range were re-measured (n=3 for each time point, four time points) over a period of 23 days in storage. The determined average area ratio (analyte/ISTD) was compared to the average area ratio obtained on day zero.

Furthermore, the stability of three levels of spike solutions (n=1 for each time point, eight time points) and three matrix-based calibrators (n=1 for each time point, nine time points) covering the measuring range were assessed after being stored at −80 °C over a period of 32 weeks. Recoveries were determined using freshly prepared calibrator levels. Total error (TE) was utilized to determine the acceptance criteria for stability. TE was estimated to be ±6.7 %, considering the analytical performance specifications.

Equivalence of results between independent laboratories

To assess the agreement of results between two independent laboratories (Roche Diagnostics GmbH in Penzberg [Laboratory 1] and Clinic for Children and Adolescents, University Hospital Erlangen [Laboratory 2]), a method comparison study was conducted. This study involved a total of 83 samples, including anonymized residual patient samples (39 serum and 44 plasma samples). Each sample was prepared once and measured over a three-day period.

In addition, Laboratory 2 conducted a three-day precision experiment following the same experimental design. Spiked samples used in this experiment were provided by Roche Diagnostics Penzberg (Laboratory 1). Both laboratories independently prepared calibrator levels using the certified qNMR-characterized reference material. In addition, adaptations to the sample preparation protocol were made in the second laboratory to use the 96-well plate format. All steps for sample preparation before filtration were done in one single 96-well deep well plate from polypropylene: 100 µL sample and 20 µL internal standard were incubated before adding 330 µL water. The sample was mixed and 400 µL of the diluted sample was loaded on a 96-well Novum SLE plate (Phenomenex) and placed on one set of glass inserts supplied by Hirschmann (Eberstadt, Germany). Following equilibration, the samples were eluted into the glass inserts using ethyl acetate and the extract was evaporated to dryness. The reconstituted sample was loaded on a 96-well centrifuge filter plate (Restek, Resprep PPT3) and placed on a fresh set of glass inserts. After centrifugation the filtrate was sealed and measured.

To establish comparability between our RMP and an accredited routine LC-MS/MS method (in accordance with DIN EN ISO 15189 and ISO 17025 standards) for determination of cortisone in biological matrices, a second method comparison study was conducted between Laboratory 1 and the Institute of Laboratory Medicine at Leipzig University Hospital (Laboratory 3). Laboratory 3 utilized an online-solid phase extraction-LC-MS/MS method, which was calibrated using a commercial kit, as outlined by Gaudl et al. [40]. The objective of this study was to demonstrate the comparability of the two methods using commercially available calibration material.

Estimation of measurement uncertainty (MU)

MU was assessed according to the GUM [38]. Two key factors, the uncertainty in calibrator production (unc_cal) and the calculated uncertainty in the precision experiment (unc_prec), were considered representative for the overall measurement MU.

To avoid error underestimation, for each individual sample concentration level, unc_cal of the calibrator level with the higher uncertainty is chosen for this combination of uncertainties. Target values were established by averaging multiple sample preparations performed on different days. MU was obtained by combining the unc_cal with the uncertainty (standard deviation [SD]) of the mean for individual measurement results (unc_mean). The resulting uncertainty was multiplied by a coverage factor of k=2 for a confidence level of 95 %, assuming a normal distribution (see Supplemental Material 3).

qNMR characterization of reference material

The traceability to the SI unit kilogram, which is the most important parameter for a reference measurement procedure, has been achieved using a certified reference material where the certified value is established qNMR. The reference material is traceable to the primary qNMR standard NIST PS1. The certified material (TraceCERT^®, lot BCBZ5582) has an absolute content of 98.4 % ± 1.2 % (k=2). In order to address situations where commercially available reference materials are not accessible, we have developed an in-house qNMR method (see Supplementary Material 2). The value obtained from our method is 98.6 % and is within the uncertainty of the reference material.

Selectivity/specificity

During the analysis of both the surrogate matrix sample and the native matrix pool, no interfering signals were observed in any of the recorded MRM transitions at the expected retention time of cortisone. No significant signal in the cortisone quantifier trace caused by residual unlabeled cortisone from the ISTD material was detected under the sample preparation conditions, therefore confirming the ISTD material suitability for use in this specific RMP.

The method selectivity was further proven during the interference testing experiment. From the six added potential interferents, only aldosterone was transferred to the second dimension, where it was baseline-separated from cortisone with a resolution of 7.6.

Matrix effect

During the post-column infusion experiment, no significant ion suppression or enhancement effects were observed in chromatograms of the tested matrices at the expected retention time of cortisone and ISTD.

Furthermore, a comparison of standard line slopes was performed using calibrations in both solvent and various matrices. Calibrator level 2 was excluded from all sets due to the identification of statistical outliers, and level 3 was excluded from the native serum matrix set because of a pipetting error. Additionally, calibrator level 8 was excluded from both the native serum and Li-Heparin plasma matrix sets, as its value exceeded the upper limit of quantification. The slopes of the standard lines were determined to be 0.039 (95 % confidence interval [CI]: 0.037–0.040) for the neat solution, 0.038 (95 % CI 0.037–0.039) for the surrogate matrix, 0.036 (95 % CI 0.032–0.039) for the native serum matrix, and 0.037 (95 % CI 0.034–0.041) for the Li-heparin plasma matrix, respectively. The overlapping confidence intervals of these slopes indicate that there is no significant difference between them, confirming the absence of a matrix effect. Strong correlation coefficients of ≥0.999 were consistently observed regardless of the matrix used for calibration, further supporting the conclusion that there is no matrix effect.

Additionally, the recoveries of all calibrator levels in the surrogate matrix, native serum, and Li-heparin plasma ranged from 94 to 103 % when evaluated as controls using the neat calibration as the standard.

Linearity

A quadratic model with a 1/x² weighting was determined to be the preferred regression model as it exhibited a more balanced distribution of residuals around zero, compared to a linear regression model. Furthermore, all individual calibration curves achieved a correlation coefficient of r ≥0.999, indicating a high level of linearity.

In addition, the linearity of the method was confirmed by a linear relationship between measured and theoretical concentration of serially diluted samples with a correlation coefficient of ≥0.995. The recoveries ranged from 95 to 102 %, except for one sample found at 110 %.

Lower limit of the measuring interval (LLMI)

The LLMI was determined using a spiked matrix sample with a concentration of approximately 0.0800 ng/mL (0.222 nmol/L) (Figure 1). Two measurement values had to be excluded due to a laboratory error. The relative deviation (n=18) was 1.1 % and CV was 4.4 %.

Precision and accuracy

To assess the overall variability of the RMP, a multi-day validation experiment was conducted, with CVs provided for easier interpretation of the results. The intermediate precision associated with between-day, calibration preparation and injection variance ranged between 2.1 and 3.4 %. The repeatability CV, including variances from between-preparation and injection, ranged from 1.9 to 2.9 % across all concentration levels, as indicated in Table 1.

Regarding method accuracy, the results showed a relative mean bias ranging from −3.7 to 2.8 %, independent of the nature of the sample (see Table 2). These findings indicate that the method exhibits no significant bias and demonstrated independence of the matrix.

Stability

Processed samples were proven to be stable at 8 °C for the complete duration of the experiment of 23 days, with a relative difference between −2.9 and 5.6 %, compared to day zero. Consequently, the stability of processed samples at 8 °C can be guaranteed for 22 days.

Additionally, recoveries for spike solutions stored at −80 °C ranged from 95 to 105 %, with the exception of two measuring levels which were found to be 108 and 111 %. Recoveries for matrix-based calibrators and control materials ranged from 93 to 105 %. This demonstrates stability over the experimental duration of 32 weeks. Therefore, the stability can be ensured for up to 31 weeks.

Figure 1:

Cortisone LC-MS/MS derived analytical readouts. (A) Chromatogram of calibrator level 1 with a cortisone concentration of 0.080 ng/mL (0.222 nmol/L) spiked in albumin surrogate mix (60 g/L); (B) pooled patient serum sample with a concentration of 31.4 ng/mL (87.1 nmol/L); analyte quantifier trace (left) and ISTD quantifier trace (right); (C) pooled Li-heparin plasma patient sample with a concentration of 31.0 ng/mL (86.0 nmol/L); analyte quantifier trace (left) and ISTD quantifier trace (right).

Equivalence of results between independent laboratories

Out of the 83 anonymized patient samples analyzed at Laboratories 1 and 2, three samples were identified as outliers performing the LORELIA outlier test [41]. Passing-Bablok analysis showed strong agreement between the laboratories with a regression slope of 1.03 (95 % CI 1.01–1.05) and an intercept of −0.21 (95 % CI −0.40 to −0.01) (Figure 2A). Pearson’s correlation coefficient was ≥0.993. Bland-Altman analysis indicated a mean bias of 1.0 % (95 % CI −0.1 to 2.0), the relative differences ranged from −8.1 to 10.0 % (95 % CI interval 3.5 %) (Figure 2B). A three-day precision experiment at Laboratory 2 showed similar CVs to Laboratory 1, with repeatability ranging from 0.4 to 2.2 % and intermediate precision ranging from 1.3 to 3.2 %, respectively.

Figure 2:

Results from the patient sample-based cortisone method comparison study performed between two independent laboratories (Laboratory 1: Roche Diagnostics GmbH, Penzberg and Laboratory 2: Department of Pediatrics and Adolescent Medicine, University Hospital Erlangen). (A) Passing-Bablok regression analysis demonstrated strong agreement between the laboratories, with a slope of 1.03 (95 % CI 1.01–1.05) and an intercept of −0.21 (95 % CI −0.40 to −0.01). The Pearson’s correlation coefficient was ≥0.993. (B) Bland-Altman analysis indicated a mean bias of 1.0 % (95 % CI −0.1 to 2.0 %) and a range of relative differences between −8.1 and 10.0 % (95 % CI interval 3.5 %).

Naturally, both measurements come along with a certain measurement error. The error of the calculated difference of both measurements is respectively larger. The performance specifications of CV_ref ≤3.0 % and B_ref ≤1.7 % must be met by each laboratory. Following the total error concept, the maximum allowable difference is calculated as 2 × B_ref + 1.96 × sqrt(CV²_ref + CV²_ref), resulting in 11.7 %. Thus, the presented results align well with the pre-determined acceptance criteria based on the biological variability and the intermediate precision data from individual laboratories and indicate that the proposed RMP is transferable between laboratories.

During the comparison study of the candidate RMP (Laboratory 1) and a routine assay of Laboratory 3, 80 of the same anonymized patient samples, also used for the RMP transfer study, were used. One sample was detected as outlier. Passing-Bablok regression resulted in a slope of 1.12 (95 % CI 1.05–1.19) and an intercept of −0.88 (95 % CI −1.58 to −0.01) (Figure 3A). Pearson’s correlation coefficient was ≥0.962. Although a positive slope and a negative intercept was observed, the Bland-Altman analysis showed a good agreement between the routine assay and the candidate RMP with a mean bias of 5.9 % (95 % CI 3.4–8.5). The 2S interval of the relative differences ranged from −16.4 to 28.3 % (95 % CI interval 8.8 %) (Figure 3B), which is in agreement with the overall variability of the routine assay [40].

Figure 3:

Results from the patient sample-based cortisone method comparison study performed between the accredited routine LC-MS/MS method (calibrated with a commercially available kit) and the RMP. (A) Passing-Bablok regression analysis demonstrated a good agreement with a slope of 1.12 (95 % CI 1.05–1.19) and an intercept of −0.88 (95 % CI −1.58 to −0.01). Pearson’s correlation coefficient was ≥0.962. (B) Bland-Altman analysis showing the agreement between the routine assay and the reference method (RMP). The mean bias was 5.9 (95 % CI 3.4–8.5) with relative differences ranging from −16.4 to 28.3 % (95 % CI interval 8.8 %).

Estimation of measurement uncertainty (MU)

Estimation of MU was performed based on the description in Supplemental Material 3. The measurement process for serum samples exhibited an overall MU ranging from 2.3 to 3.7 %, regardless of sample concentration, as indicated in Table 3. To account for a confidence level of approximately 95 %, assuming a normal distribution, the resulting overall MU was multiplied by a coverage factor of k=2. As a result, the expanded measurement uncertainties fell within a range of 4.7–7.4 %.

For the establishment of reference or target values, multiple sample preparations were carried out for each sample on at least two different days. The average of these results, calculated using n=6, was used. The overall uncertainties were found to range from 1.1 to 2.8 %, with expanded uncertainties ranging from 2.1 to 5.5 % (k=2), as shown in Table 4.