Instruments: LCMS-8050 CL triple quadrupole liquid chromatography-tandem mass spectrometer (Shimadzu Corporation, Japan) equipped with an electrospray ionization (ESI) source and an LC-30AD liquid chromatography system; Agilent ZORBAX Eclipse Plus C18 chromatographic column (2.1 × 100 mm, 1.8 μm); TG16-WS high-speed refrigerated centrifuge (Xiangyi Laboratory Instrument Development Co., Ltd., Hunan, China); XH-E vortex mixer (Kangjian Medical Supplies Co., Ltd., Jiangsu, China); micropipettes (Eppendorf, Germany); Milli-Q ultrapure water system (Millipore, USA).

ReagentsandStandards: Haloperidol (purity >98%, batch no. H1025), amisulpride (>98%, batch no. A1028), olanzapine (>98%, batch no. O1015), clozapine (>98%, batch no. C1032), perphenazine (>98%, batch no. P1011) and their corresponding deuterated internal standards: haloperidol-d4, amisulpride-d5, olanzapine-d3, clozapine-d4, perphenazine-d5 were all purchased from Hunan Demeter Instrument Co., Ltd. Methanol and acetonitrile were MS grade (Thermo Fisher Scientific, USA); formic acid was HPLC grade (Merck, Germany); all water used was ultrapure water (resistivity ≥18.2 MΩ·cm).

BlankSerumandClinicalSamples: Blank serum was obtained from leftover serum samples of healthy volunteers (with informed consent and confirmed not taking relevant psychiatric drugs) undergoing physical examination at our hospital, and stored at −80˚C for later use. Clinical validation samples were retrospectively analyzed steady-state trough concentration serum samples (collected in the morning before medication) from 150 patients with psychiatric disorders (diagnosis conforming to ICD-10 criteria) who were hospitalized or treated as outpatients at Baise Second People’s Hospital and receiving relevant antipsychotic drug therapy from January 2025 to December 2025. This study was approved by the Medical Ethics Committee of Baise Second People’s Hospital (Approval No.: BSEY-LL-2025-018). Informed consent was obtained from all patients or their legal guardians.

Preparation of Standard Stock and Working Solutions: An appropriate amount of each drug standard was accurately weighed, dissolved, and diluted with methanol to prepare single standard stock solutions with a concentration of 1.0 mg/mL, stored at −20˚C protected from light. Mixed standard working solutions at appropriate concentrations were prepared by serial dilution with methanol before use. Internal standard stock and working solutions were prepared similarly.

PreparationofCalibrationCurveandQualityControl(QC)Samples: 190 μL of blank serum was pipetted into a 1.5 mL EP tube, followed by the addition of 10 μL of mixed standard working solutions at different concentrations and 10 μL of mixed internal standard working solution (final IS concentration of 50 ng/mL for each), vortex-mixed for 30 seconds to prepare calibration curve samples at concentrations of 10, 25, 50, 100, 150, and 250 ng/mL. QC samples at low (25 ng/mL), medium (100 ng/mL), and high (200 ng/mL) concentration levels were prepared in the same manner.

200 μL of serum sample (calibration curve, QC, or clinical sample) was aliquoted into a 1.5 mL EP tube. 20 μL of mixed internal standard working solution (final concentration 50 ng/mL) was added, followed by 500 μL of ice-cold acetonitrile as protein precipitation reagent. The mixture was vortexed for 3 minutes for thorough mixing and then centrifuged at 4˚C, 14,500 rpm for 10 minutes. Approximately 200 μL of the supernatant was carefully transferred into an insert vial. 2 μL was injected for LC-MS/MS analysis.To assess potential solvent effects, the impact of injecting the undiluted acetonitrile supernatant (approximately 70% organic phase) on chromatographic peak shape was evaluated. Results showed symmetrical peaks without significant distortion, indicating negligible solvent effects under the initial gradient conditions (15% acetonitrile); therefore, an additional dilution step was omitted.

ChromatographicConditions: Mobile phase A was 0.1% formic acid in water, and mobile phase B was acetonitrile. The gradient elution program was set as follows: 0 - 1.0 min, maintain 15% B; 1.0 - 4.0 min, B linearly increased from 15% to 70%; 4.0 - 5.0 min, maintain 70% B; 5.0 - 5.1 min, B rapidly increased from 70% to 95%, and maintained until 6.5 min; 6.5 - 6.6 min, B rapidly decreased from 95% to 15%, followed by re-equilibration at 15% B until 8.0 min. The flow rate was set at 0.35 mL/min. Column oven temperature: 40˚C. Autosampler temperature: 10˚C. Total analysis run time: 8.0 minutes.

Mass Spectrometric Conditions: Electrospray ionization (ESI) source was used with positive ion mode acquisition. Source parameters were set as: desolvation line (DL) temperature 250˚C; heat block temperature 400˚C; nebulizing gas (nitrogen) flow 3.0 L/min; drying gas (nitrogen) flow 10 L/min; heating gas (air) flow 10 L/min. The scan mode was multiple reaction monitoring (MRM). The optimized MS parameters for each analyte and corresponding IS are detailed in Table 1.

Table 1.Multiple Reaction Monitoring (MRM) parameters for the five antipsychotics and their deuterated internal standards.

*Note:The product ion marked with an asterisk is the quantifier ion transition.

Bbbnnn Method validation was performed systematically with reference to the “9012 Guidance for Validation of Quantitative Analytical Methods for Biological Samples” in the Chinese Pharmacopoeia (2020 Edition, Volume IV) and international standards [13].

1) Specificity: Six individual blank serum samples from different sources, blank serum spiked with analytes at LLOQ concentration and IS, and actual patient serum samples were analyzed to investigate the presence of interfering endogenous substances within the retention time windows of the target analytes and IS.

2) Linearity and Lower Limit of Quantification (LLOQ): Calibration curves were constructed by plotting the peak area ratio of each analyte to its corresponding IS versus the analyte concentration, using weighted (1/x²) least squares linear regression analysis. The LLOQ was defined as the lowest concentration with a signal-to-noise ratio (S/N) ≥ 10, accuracy within 80% - 120%, and precision (RSD) ≤ 20%.

3) Precision and Accuracy: Intra-day precision was assessed by analyzing five replicates of QC samples at low, medium, and high concentration levels within a single batch. Inter-day precision was assessed by analyzing QC samples at the same three levels in three separate batches on different days. Precision was expressed as relative standard deviation (RSD). Accuracy was expressed as the percentage of the mean measured concentration relative to the nominal concentration.

4)ExtractionRecoveryandMatrixEffect: Matrix effect was evaluated using the post-column infusion technique. Extraction recovery was calculated by comparing the analyte peak area from pre-extraction spiked QC samples (low and high concentrations) after sample pretreatment to the analyte peak area from post-extraction spiked samples where the same amount of standard was added to the supernatant of processed blank matrix. Previous studies indicate that effective control of matrix effect and stable extraction recovery are key to ensuring the accuracy of biological sample analysis [14].

5)Stability: The stability of QC samples was investigated under various conditions: at room temperature for 24 hours; after three complete freeze-thaw cycles; after storage at −20˚C for 14 days. Considering that clinical samples are typically analyzed shortly after collection, the 14-day stability data are sufficient for routine testing cycles. For samples requiring long-term storage, storage at −80˚C is recommended, and their long-term stability will be addressed in future studies. and processed samples were placed in the autosampler (10˚C) for 24 hours. Comprehensive stability verification has been confirmed as a necessary prerequisite for applying a method to routine clinical testing [15].

The validated method was integrated into the hospital’s routine TDM workflow. Serum concentrations were determined in 150 samples from patients receiving relevant antipsychotic therapy. Measured drug concentrations were compared with the therapeutic reference ranges recommended in consensus guidelines [4] to assess patients’ blood concentration levels. Combined with clinical efficacy records and adverse reaction reports, the results provided reference suggestions for clinicians regarding dosage adjustment. Studies have indicated that TDM based on LC-MS/MS technology also holds significant value in assessing medication adherence in psychiatric patients [16].

Chromatographic data were acquired and processed using LabSolutions software (version 5.97, Shimadzu, Japan). Statistical analyses, including calibration curve fitting, precision, and accuracy, were performed using Microsoft Excel 2019 and SPSS 30.0 software.

Through systematic optimization of mobile phase composition, gradient elution program, ion source parameters, and MRM transitions, an LC-MS/MS method capable of separating and detecting the five target drugs within 8 minutes was established. Figure 1 shows representative MRM chromatograms under optimized conditions for a blank serum sample, a blank serum sample spiked with analytes at LLOQ and IS, and a typical patient serum sample. The results show sharp and symmetrical chromatographic peaks for all analytes and corresponding IS, with stable retention times. No significant interfering peaks from endogenous substances were observed at the elution positions of the target analytes in blank serum.

Figure 1. Representative MRM chromatograms. (A) Blank serum; (B) Blank serum spiked with LLOQ concentrations (10 ng/mL) of the five drugs and corresponding internal standards; (C) A typical patient serum sample from a patient concurrently taking olanzapine and clozapine).

1)Specificity: Six individual blank serum samples from different sources showed no significant interfering chromatographic peaks (response intensity less than 20% of the LLOQ response) within the retention time windows of the target analytes and corresponding IS, indicating good specificity of the method.

2)LinearityandLLOQ: All five target drugs showed good linearity in the concentration range of 10 - 250 ng/mL. Typical regression equations and correlation coefficients are detailed in Table 2. The correlation coefficients (r²) for all analytes were greater than 0.998. The LLOQ was determined to be 10 ng/mL. At this concentration, the S/N for all drugs exceeded 15, accuracy ranged from 88.5% to 105.3%, and precision RSD was ≤12.8%, fully complying with the acceptance criteria for method validation. In similar studies, LLOQs for methods simultaneously detecting classical antipsychotics are mostly between 2 - 10 ng/mL [17]. The LLOQ set in this method aligns with clinical practical needs.

Table 2. Linear regression equations, correlation coefficients, and LLOQ for the five antipsychotics.

3)PrecisionandAccuracy: As shown in Table 3, the intra-day precision RSD for QC samples at low, medium, and high concentration levels ranged from 2.5% to 8.1%, and inter-day precision RSD ranged from 3.8% to 9.7%; accuracy ranged from 94.2% to 106.5%. All results met the general acceptance criteria for bioanalytical method validation. Studies have reported intra-day and inter-day precision less than 10% for similar detection methods [18]. The precision indices of this method are consistent, indicating good reproducibility of the results.

Table 3. Method precision and accuracy validation results.

4)MatrixEffectandExtractionRecovery: Evaluation results showed that for the five drugs at low and high concentration levels, the IS-normalized matrix factor values ranged from 85.6% to 113.2%, with RSDs all less than 8.5%, indicating that the matrix effect was within a controlled range and reproducible. Using the one-step acetonitrile precipitation method for pretreatment, the average extraction recoveries for each drug at low and high concentrations ranged from 86.3% to 102.4%, with RSDs less than 7.8%, meeting methodological requirements. Studies using similar protein precipitation methods for serum sample processing have reported extraction recoveries of 86.2% - 104.8% [19]. The extraction efficiency of this method is comparable and the operation is simpler and faster.

5) Stability: Stability experiment data confirmed that for QC samples placed at room temperature for 24 hours, subjected to three freeze-thaw cycles, stored frozen at −20˚C for 14 days, and processed samples placed in the autosampler (10˚C) for 24 hours, the variation in each drug concentration was within ±12%, indicating good stability of samples under these different storage and processing conditions.

From January 2025 to December 2025, the method developed in this study was applied to TDM of 150 clinical samples. The basic demographic characteristics and medication profiles of the patients are shown in Table 4. The distribution of detection results is shown in Figure 2: in 97 samples, the blood concentration of at least one drug fell within the recommended therapeutic reference range; in 41 samples, drug concentrations were below the lower limit of the therapeutic window; and in 12 samples, drug concentrations exceeded the potential toxicity threshold.

Table 4.Basic characteristics of the clinical validation cohort patients (N = 150).

Figure 2.Distribution of blood concentration monitoring results from 150 clinical samples.

Based on TDM results combined with comprehensive clinical assessment, drug therapy regimens were adjusted for 35 patients. Subsequent follow-up showed that among these, 28 patients showed improved Clinical Global Impression (CGI) scores or reduced drug-related adverse reactions after regimen adjustment, preliminarily demonstrating the potential value of this method in guiding clinical individualized medication.