CSA was calibrated using a 6-point calibration curve with 6Plus1 Multilevel ISD calibrators in whole blood (Chromsystems) with concentrations of 23.40, 128.40, 294.00, 471.00, 745.00, and 1,890.00 ng/ml, which were also used for linearity validation. Internal quality controls (QCs) were evaluated using Bio-Rad Lyphochek Whole Blood ISD Controls levels 1, 2, 4, and level 5 were added when there was a high-concentration sample (Bio-Rad Laboratories, Inc.). CSA Standard (CAS 59865-13-3, ≥95%) was provided by MilliporeSigma, and [²H₁₂]-Cyclosporin A (CSA-d12, AlsaChim) with a purity of 98% ²H was used as an internal standard (IS). Formic acid and zinc sulfate were obtained from MilliporeSigma (analytical grade). Other solvents and reagents were HPLC grade. Ultrapure water of 18.2 MΩ cm resistivity was obtained from a Milli-Q (MilliporeSigma) water purification system.

Samples were pretreated by mixing 20 µl EDTA anti-coagulated whole blood with 400 µl sample pretreatment reagent consisting of 0.05 M zinc sulfate and 30.0 ng/ml CSA-d12 in 50% methanol/water. Samples were vortexed vigorously for more than 20 sec and mixed for 5 min in a 55-well oscillator. After centrifuging for 5 min at 10,000 x g at 4˚C, the obtained supernatant was used for analysis.

LC-MS/MS analysis was performed on an LC-20AXR (Shimadzu Corporation) tandem AB SCIEX API4000 plus LC-MS/MS instrument (Applied Biosystems) at Guangzhou KingMed Center for Clinical Laboratory Co., Ltd. (Guangzhou, China). Two HPLC systems were paralleled via a six-port switching valve mounted on an MPX driver (Fig. 1). The system was controlled and data was processed using the Analyst software (AB SCIEX, version 1.6.2), followed by quantitative analysis using the MultiQuant software (AB SCIEX, version 3.0.1) with the default integration parameters except for a 0.05 min baseline sub. window and a weighting factor of 1/x for area linear regression.

Chromatographic separation was achieved on a C18 column (MilliporeSigma, 2.1x50 mm, 2.7 µm particles) with a C18 SecurityGuard column (Phenomenex, 2x2.1 mm) at 60˚C. A gradient elution program was set at a flow rate of 0.5 ml/min from 60% buffer B and changed to 100% buffer B (buffer A: 2 mM ammonia acetate and 0.1% formic acid in water; buffer B: methanol) at 1.5 min for 1.0 min. From 2.51 to 3.5 min the gradient was changed and kept at 60% buffer B to stabilize the column for the next injection (Table I). The injection volume was 10 µl.

Through the paralleled HPLC systems, during the 1.5 min interval from 0.8 to 2.3 min, the LC was set to the mass spectrometer, and another sample could be detected during the remaining time (Fig. 1). The autosampler needed an additional 0.4 min per sample for injection. In this way, within 4.3 min, two samples were detected, thus allowing for a shorter analysis time for each sample of 2.15 min, thus improving the instrument throughput.

An electrospray ionization source was used in positive ion mode (Applied Biosystems). Precursor/production pairs were used (1,219.9/1203.0 m/z for CSA and 1,232.0/1,215.2 m/z for CSA-d12) in multiple reaction monitoring (MRM) mode with a dwell time of 120 msec. Qualitative ion pairs were also monitored (1,221.0/1,204.0 m/z for CSA and 1,233.0/1,216.1 m/z for CSA-d12) with a dwell time of 60 msec. MRM pairs were processed at 60, 6, 30, and 20 V for declustering potential, entrance potential, collision energy, and collision cell exit potential, respectively. The curtain gas, collision gas, nebulizer gas, and auxiliary gas were set at 25, 5, 50, and 60 psi, respectively. The ion spray voltage was 5,500 V, and the source temperature was 400˚C.

The modified LC-MS/MS assay was evaluated according to the guidelines in the CLSI LC-MS C62-A documentation and the International Association of Therapeutic Drug Monitoring and Clinical Toxicology (IATDMCT) expert consensus group (12,13). HPLC-MS, affinity chrome-mediated immunoassay (ACMIA), chemiluminescent microparticle immunoassay (CMIA), electrochemiluminescence immunoassay (ECLIA), and cloned enzyme donor immunoassay (CEDIA) were performed as described previously (6,14-16).

Whole blood samples at three concentration levels were used to assess the intra-day precision and inter-day precision (n=20 for each concentration). Intra-day precision was processed on the same day and day-to-day variability was assessed by analysis of two sets of samples on 10 different days. The acceptance criterion for total imprecision was based on the recommendation of the IATDMCT expert consensus group of ≤10% (13).

A total of six samples of each concentration were measured in duplicate on 3 different days. Whilst 23.40, 128.40, 294.00, 471.00, 745.00, and 1,890.00 ng/ml were obtained from the 6-point calibration curve (Chromsystems), 5.85 and 7.80 ng/ml were obtained by dilution from 23.40 ng/ml by the blank point. The acceptance criterion was defined as a regression deviation <10% and the calibration curve had to have a correlation coefficient (r) of 0.99 or better. The lowest concentration that met the above criteria and had a signal-to-noise ratio >10 was accepted as the lower limit of quantitation (LLoQ).

The clinical reportable range was determined according to the linear range and the maximum dilution multiple. A high-concentration sample was diluted 2, 4, or 8 times with homogeneous drug-free whole blood (EDTA anticoagulant). Sample concentrations after dilution should be within the linear range and not threefold below the LLoQ. Taking the concentrations before dilution as references, the dilution multiple was determined by analyzing five replicates of patient samples before and after dilution separately. The recovery of the diluted samples should range from 85 to 115%, and the deviation of five duplicates should be <15%.

A total of six external quality assessment (EQA) samples [Laboratory of the Government Chemist (LGC) Institution] under the ISD program (CICTAC-Cyclosporin) were detected to assess the method's accuracy. A bias within ±25% and a Z score under ±3 was considered acceptable.

In recovery experiments, three concentration levels of whole blood samples were spiked with low, moderate, and high concentrations of standard solutions separately to calculate recoveries. All the spiked and unspiked samples were analyzed with three replicates. The spiked recovery, which refers to (spiked sample concentration - unspiked sample concentration)/standard spiked value, should be within 85-115%.

Matrix specificity was validated by evaluation of the presence of any peaks in the corresponding position for the CSA and CSA-d12 when the blank matrix (the point of STD-0 in the standard curve) and the blank patient samples were detected.

The matrix effect is independent of the presence of the analyte and influences the accurate quantification for CSA (10). In this matrix admixing experiment, the matrix effect between whole blood patient sample and QC or calibration was evaluated. Standard solutions (a and b: commercial QC at low and high concentrations respectively, Bio-Rad Laboratories, Inc; c-f: matrices at blank, low, moderate, and high concentrations of the calibration respectively, Chromsystems Instruments & Chemicals GmbH), patient samples, and mixed samples (standard solutions mixed with six samples A-F from different patients separately in different proportions) were prepared. The matrix effect was evaluated through the analysis of three replicates of the three different matrix samples above separately. The deviation between the response of the mixed samples and the response average of the patient samples and standard solutions should be <20%.

A post-column infusion experiment was processed to evaluate the matrix, in which the analytes dissolved in solvent (1.0 mg/l CSA or CSA-d12 in 50% methanol/water) was directly infused into the mass spectrometer using a syringe pump (30 µl/min) to gain a stable total ion count (TIC). Then three different extracted matrix samples were separately injected concurrently by the HPLC system. There should be no decrease or increase in TIC of the direct infusion observed at the time point when the matrix components are eluted from the column.

Carryover was evaluated through repeated (n=3) injections of samples with low concentration (C1, LLoQ), immediately followed by high concentration [C2, upper (U)LoQ] and low concentration (C3, LLoQ) injections. The acceptance criterion was ≤20% carryover, which was calculated using the following formula: Bias=(C3 mean-C1 mean)/C1 mean x100%.

This modified method was used for the detection of CSA concentrations in 41 nephrotic patients (33 with nephrotic syndrome and 8 with kidney transplants) who were treated with CSA between April 2016 and July 2021 in the Guangdong Provincial Hospital of Chinese Medicine (Guangdong, China). Among these patients, 10 were women and 31 were men, with an age range of 17-78 years, a median age of 44 years and an mean age of 43.46±16.54 years. EDTA anticoagulated whole blood was collected from patients before the morning CSA dose using the aforementioned LC-MS/MS method to obtain the CSA trough concentration. Patient information, such as clinical diagnosis, CSA trough concentration, laboratory indicators, and drug combinations were recorded. This study was approved by the Ethics Committee of Guangdong Provincial Hospital of Chinese Medicine (approval no. ZE2020-240-01). Written informed consent was obtained from all participants.

Chronic kidney disease (CKD) was graded on the basis of the estimated glomerular filtration rate (eGFR) according to the KDIGO guidelines (17): CKD1, eGFR ≥90 ml/min/1.73 m²; CKD2, eGFR 60-90 ml/min/1.73 m²; CKD3, eGFR 30-59 ml/min/1.73 m²; CKD4, eGFR 15-29 ml/min/1.73 m²; CKD5, eGFR <15 ml/min/1.73 m².

Statistical analyses were performed using MedCalc version 20.0.22 (MedCalc Software bvba), SPSS Statistics version 20.0 (IBM Corp.), and GraphPad Prism version 8.3 (GraphPad Software, Inc.). The results of proficiency testing were evaluated using Passing-Bablok regression and Bland-Altman plots. Continuous data are presented as the mean ± SD for the normally distributed data or otherwise as the median (range). For intergroup comparisons, normally distributed data were analyzed using a Welch one-way ANOVA test with a Games-Howell post hoc test. Non-normally distributed data were assessed using a Mann-Whitney U test or a Kruskal-Wallis test. P<0.05 was considered to indicate a statistically significant difference.

The retention times for CSA and CSA-d12 were 1.74 and 1.71 min (0.94 and 0.91 min for the MS monitoring period), respectively. For the blank matrix and blank patient sample, no peak was detectable in the corresponding position. A high-concentration sample (1,692.87 ng/ml) and a low-concentration sample (10.10 ng/ml) were used to assess carryover. And cross-contamination between samples was negligible.

A total of six whole blood samples were analyzed to assess precision, three of these were tested as 20 parallels in 1 day and the remaining three samples were analyzed in duplicate, one run per day over 10 days (Table II). For the two HPLC systems, the total imprecision was <3.64%, which met the goal of ≤10% for all concentrations.

The method showed good correlation in the linear range of 5.85-1,890.00 ng/ml (weighting 1/x, Fig. 2) with an r >0.99, and the LLoQ was set to 5.85 ng/ml (Fig. 3). The linear regression equation was y=0.0016x+0.00144 (r=0.99948) for HPLC system 1, and y=0.0016x+0.00284 (r=0.99954) for HPLC system 2. The dilution performance was assessed by diluting a high-concentration sample with homogeneous drug-free whole blood (EDTA anticoagulant) 2, 4, or 8 times. Dilution recoveries were within the pre-defined acceptance limits and the deviation of five duplicates was <4.15%. Hence, the sample could be quantified by dilution when the concentration of the analyte exceeded the ULoQ (1,890.00 ng/ml), which meant that this method could obtain quantitative results for samples with concentrations between 5.85 and 15,120 ng/ml.

A total of six EQA samples were in good agreement with the results from the LGC institution (44.05-2,129.50 ng/ml) and met the acceptable range for the LC-MS/MS method (bias <±4.71%, Z scores <±0.57%, Fig. 4).

A total of three concentration levels of whole blood samples (59.73±0.63, 124.87±1.79, and 365.21±0.78 ng/ml) were spiked with low, moderate, and high concentrations of standard solutions (54.04, 104.49, and 511.73 ng/ml) separately, and the spiked recoveries ranged from 89.4-104.7%.

Post-column infusion experiments showed no observable decrease or increase in the TIC of the direct infusion when the matrix components were eluted from the column (Fig. 5). For the matrix admixing experiment, standard solutions and patient samples were mixed in a 1:4 ratio. The deviations of the mean responses ranged from -1.9-5.5%, thus meeting the acceptance criterion of <20% and confirming no relative matrix effect.

A total of 12 standard samples with a concentration range of 121.74-1,439 ng/ml were applied to compare the modified LC-MS/MS with five other CSA detection methods, including HPLC-MS, ACMIA, CMIA, ECLIA, and CEDIA. The Z score represents the deviation between the measured value and the target value. In the modified LC-MS/MS, the Z scores of the 12 samples were between -0.5 and 1, which were statistically different from ACMIA and CMIA (Fig. 6). However, the Z score of the LC-MS/MS method (median=0.1,100; 25th-75th percentile=-0.0925-0.4375) was less discrete when compared with HPLC-MS (median=-0.425; 25th-75th percentile=-0.830-0.410), ECLIA (median=0.395; 25th-75th percentile=-0.1425-0.7025) and CEDIA (median=0.405; 25th-75th percentile = -0.245-0.700). Both ECLIA and CEDIA are a type of immunoassay based on the specific reaction of antigen and antibody. The obvious disadvantage was cross-reactivity with metabolites, particularly in liver transplant recipients where hepatic dysfunction can lead to alterations in CSA metabolism and elimination (18). CEDIA had cross-reactivity for CSA metabolites AM1, AM4n, and AM9, resulting in a positive bias of the results (16). Likewise, Shigematsu et al (19) evaluated the analytical performances of immunoassays in monitoring tacrolimus, using LC-MS/MS as a standard, and observed a bias of 7.46% in ECLIA. Thus, the accuracy and stability of the modified LC-MS/MS showed improvement to some extent.

Patients with kidney transplants received more frequent monitoring than those with nephrotic syndrome (NS). The CSA concentration was significantly higher for patients with kidney transplants than for those with NS (P<0.05). Of the 33 patients with NS, 13 underwent renal biopsies, and nine were diagnosed with membranous nephropathy, whereas four showed minimal changes in the disease. The results of the monitoring are shown in Table III.

Concomitant medications were more common in patients with NS. The CSA treatment was most frequently combined with antihypertensive agents. The results of the monitoring of patients with different drug combinations are shown in Table IV.

Patients with NS in the CKD3 stage had higher blood concentrations than those in the CKD1 stage (P<0.05). However, no statistically significant differences between other grades in patients with either NS or kidney transplants were observed (P>0.05, Table V).