Risk Factors for Amisulpride-Associated Hyperprolactinemia: Gender-Specific Plasma Concentration Cutoffs Derived from ROC Analysis Authors: Hui Wang, Qing Shao, Miaomiao Yang, Yichen Chen, Wenjie Wang Published: 2026 ABSTRACT AbstractPurposeTo investigate independent risk factors for amisulpride-induced hyperprolactinemia (HPRL), establish gender-stratified diagnostic cutoff values for HPRL, and determine amisulpride plasma concentration cutoff values for different genders, thereby providing a basis for personalized medication.Patients and MethodsThis retrospective study included 209 patients treated with amisulpride. Significant univariate variables were entered into multivariable linear regression to identify independent predictors. We used receiver operating characteristic (ROC) curve analysis to determine the overall prolactin cutoff of 28 ng/mL for identifying HPRL. Sex-stratified ROC analyses were further conducted to establish gender-specific optimal amisulpride plasma concentration thresholds.ResultsDaily amisulpride dose was positively correlated with plasma drug concentration (p < 0.001); multivariable analysis identified sex, plasma amisulpride concentration, and concomitant idebenone or lorazepam use as independent risk factors for amisulpride-induced prolactin elevation (p < 0.05). ROC analysis identified a screening cutoff value of 28 ng/mL for elevated prolactin levels in the overall population. Plasma amisulpride concentration cutoffs were higher in women (317.8 ng/mL) than in men (294.2 ng/mL), suggesting greater tolerance to higher plasma concentrations among women and confirming significant sex differences in susceptibility to amisulpride-induced prolactin elevation.ConclusionClinically, 28 ng/mL can be adopted as the screening cutoff value for elevated prolactin levels in patients treated with amisulpride, we advocate for sex-dependent plasma concentration management for personalized therapeutic drug monitoring to effectively prevent and control amisulpride-associated HPRL.Keywords: amisulpride, hyperprolactinemia, receiver operating characteristic, cutoff value, therapeutic drug monitoring Introduction Schizophrenia is a common, severe mental disorder with a chronic, relapsing course. Some patients respond poorly to treatment, which markedly impairs their social functioning. Medication is the primary treatment modality.1 As a first-line atypical antipsychotic, amisulpride precisely modulates dopamine pathways, demonstrating proven efficacy against positive and negative symptoms of schizophrenia, with minimal extrapyramidal side effects, and is widely used in clinical practice. However, this drug tends to block pituitary dopamine receptors, inducing HPRL and leading to various endocrine adverse reactions. Untreated HPRL disrupts gonadal hormone homeostasis, causing menstrual disorders, galactorrhea and infertility in women, and sexual dysfunction and reduced libido in men. Persistent hyperprolactinemia further accelerates bone mineral loss and raises long-term risk of osteopenia, worsening the quality of life and potentially reducing long-term medication adherence.2 Furthermore, plasma amisulpride concentrations, efficacy, and prolactin-elevating responses vary significantly among individuals,3 and gender is a major factor influencing the occurrence of adverse reactions. Drug sensitivity and tolerability differ markedly by sex,4 and relying solely on dosage cannot ensure treatment safety; therefore, implementing individualized therapeutic drug monitoring holds significant clinical value.5 Therapeutic drug monitoring (TDM) is a key technical tool for achieving personalized treatment with psychotropic medications. Monitoring blood drug concentrations enables a precise correlation between dosage, clinical efficacy, and adverse reactions, thereby effectively preventing poor treatment outcomes due to underdosing or overdosing-triggered toxic reactions.6,7 Studies have examined sex, age, plasma drug concentrations, and concomitant medications as potential contributors to interindividual variability in amisulpride pharmacokinetics, but the findings remain inconsistent.8 Ding9 reported higher plasma amisulpride concentrations in female patients, whereas Müller10 et al found no significant gender differences. Although amisulpride plasma concentrations increase with age11–13 a TDM study by Bergemann et al found no significant association between age and plasma concentrations. Amisulpride is primarily excreted renally and metabolized almost entirely independently of CYP450 enzymes, so most studies suggest that concomitant medications do not significantly interfere with its pharmacokinetics.14,15 Regarding the association between plasma drug concentrations and HPRL risk, Smith et al suggested that plasma concentrations ≥100 ng/mL significantly increased the risk of prolactin elevation. However, the study’s small sample and lack of sex stratification limited its utility for individualized dosing. Overall, the existing studies still have significant limitations. Most studies employ simple correlation analyses or between-group comparisons, lack multivariate models to identify independent risk factors, and predominantly use a single population-wide prolactin cutoff value to diagnose HPRL. This approach overlooks physiological sex differences in baseline endocrine levels. Consequently, sex-specific thresholds for sharp prolactin increases cannot be clearly defined. In the present study, we performed a retrospective real-world investigation of 230 patients with schizophrenia treated with amisulpride, employing stratified stepwise analysis. First, independent risk factors for elevated prolactin levels were identified through correlation and multiple linear regression analyses.16 No unified clinical gold standard exists to define abnormal prolactin concentrations. Variations between laboratory testing platforms and research protocols have hindered consensus on a universal prolactin cutoff for identifying antipsychotic-associated hyperprolactinemia. A concentration of 25 ng/mL roughly corresponds to the upper physiological prolactin limit for non-pregnant adult females and is widely adopted for routine clinical screening, consistent with international endocrine practice guidelines.17 The 28 ng/mL cutoff applied in this study approximates the 28.3 ng/mL (600 mIU/L) cutoff specified in psychiatric consensus statements.18 We assessed both cutoffs to minimise diagnostic bias stemming from dependence on a single fixed value and to compare their predictive performance. Unlike earlier studies that examined risk factors and predictive cutoff values separately, this real-world investigation combines multivariable risk factor analysis with gender-stratified ROC-derived plasma concentration thresholds in a real-world setting. This approach yields sex-specific concentration targets instead of adopting a single uniform standard for all patients. We aim to generate robust real-world evidence to facilitate individualized risk surveillance and clinical safety management of amisulpride-induced HPRL. Materials and Methods Data Collection This study is a retrospective analysis based on real-world TDM data, enrolling patients with schizophrenia who were hospitalized at the Xi’an Mental Health Center and treated with amisulpride between January 1 and December 31, 2021. Since all data were extracted from pre-existing historical electronic medical and TDM archives, no a priori sample size calculation was performed before participant enrolment. The study data were sourced from the hospital’s TDM database and electronic medical record system. These patients were diagnosed with schizophrenia according to the International Classification of Diseases, 10th Revision (ICD-10). The inclusion criteria were as follows: receiving treatment with amisulpride, with or without concomitant use of other antipsychotic medications; taking a fixed dose of amisulpride for at least 7 days, with samples collected after reaching steady-state plasma concentrations; and for patients who underwent multiple TDM tests at the same dose, we included only the final test result. The exclusion criteria were as follows: amisulpride plasma concentrations outside the lower or upper limits of quantification; age < 18 years; and missing amisulpride dosage information; and missing serum prolactin data. Initially, 237 amisulpride TDM cases were identified. After eligibility screening, 230 cases were included in the preliminary analysis. After further excluding cases with missing serum prolactin data, 209 cases were ultimately included in the statistical analysis. Data on patient age, gender, daily dose, BMI and concomitant medications were extracted from electronic medical records. Univariate correlation analysis, multiple linear regression analysis, and ROC curve analysis were sequentially performed to identify independent risk factors for amisulpride-induced HPRL and to determine the optimal diagnostic cutoff value for serum prolactin. Figure 1 shows the study data selection process. The study also included the following concomitant medications: antipsychotics (eg, ziprasidone, Ziprasidone, risperidone, aripiprazole, quetiapine and clozapine), hypnotics (eg, zopiclone and lorazepam), antidepressants (eg, citalopram), anxiolytics (eg, buspirone), mood stabilizers (eg, sodium or magnesium valproate), and other medications (eg, idebenone and trihexyphenidyl). The study participants were divided into 12 subgroups based on the types of concomitant medications. The Ethics Committee of the Xi’an Mental Health Center (XAJWKY-2026005) approved the study protocol. Data selection flowchart. Bioanalysis After patients have taken a fixed dose of amisulpride for 7 days and plasma concentrations have reached steady state, 2–3 mL of peripheral venous blood is collected following standard clinical procedures and placed in an EDTA-containing anticoagulant tube for analysis.19 Analytical Determination of Amisulpride Plasma Concentration Steady-state amisulpride plasma concentrations were measured using a Shimadzu LC-MS/MS 8050 liquid chromatography–tandem mass spectrometry system (Kyoto, Japan). The bioanalytical method for TDM of amisulpride established by this laboratory has undergone FDA method validation. The plasma calibrators and quality control materials at high, medium, and low concentration levels used in the experiment were all purchased from Hunan Demite Instrument Co., Ltd., within their expiration dates. All plasma samples were stored at −20°C and analyzed simultaneously. Internal quality control was implemented throughout the experiment, and the analytical method accuracy and precision were regularly assessed. The inter-batch accuracy and precision for each batch of samples measured by the TDM method were within 15%. Statistical Analysis In this study, Origin Pro 2024 (Origin Lab, Inc., USA) and GraphPad Prism 9 (GraphPad Software, Inc., USA) were used for data plotting. All statistical analyses were performed using MedCalc version 19.6 (MedCalc Software, Inc., Ostend, Belgium). Data with non-normal distributions were presented as median (interquartile range), and comparisons between groups were performed using the Mann–Whitney U-test. Categorical variables were described in terms of frequency and percentage, and differences between groups were analyzed using the chi-square (χ2) test. Correlations between continuous variables were assessed using Spearman’s nonparametric rank correlation. Univariate and multiple linear regression analyses identified independent risk factors influencing elevated prolactin levels. Candidate predictors for multiple linear regression were predefined based on clinical plausibility and published pharmacokinetic evidence. All variables yielding p < 0.05 in univariate analyses were incorporated as candidate covariates, including concomitant use of idebenone and lorazepam. This variable selection approach has been widely adopted in retrospective observational and pharmacokinetic studies to adjust for potential confounding effects on outcome indicators. 14 Clinically critical covariates (age and BMI) were additionally forced into the model irrespective of univariate statistical significance to avoid omitting key confounders and minimize selection bias.15 ROC curves were used to evaluate the model’s predictive performance, with the area under the curve (AUC) serving as the evaluation metric: a higher AUC value indicates stronger predictive ability; an AUC of 1.0 indicates optimal predictive performance, while an AUC of 0.5 indicates poor predictive performance. All tests were two-sided; a p-value < 0.05 was considered statistically significant. Results Patient Characteristics This study included 230 patients. Table 1 lists the detailed baseline clinical characteristics. Male patients slightly outnumbered female patients—52.61% (121/230) versus 47.39% (109/230). The age range of the patients was 18–73 years, with a median age of 33.0 years (interquartile range: 27.0–48.0 years). Among them, young patients aged ≤44 years accounted for 69.13% (159/230), while those aged >44 years accounted for 30.87% (71/230). The median body weight was 65.0 kg (range: 39–100 kg), and the median body mass index was 23.45 kg/m2 (range: 15.6–40.3 kg/m2). The median daily amisulpride dose was 400 mg/day (range: 100–800 mg/day). Plasma drug concentrations showed significant interindividual variability (3.7–1732.68 ng/mL), with a median concentration of 339.80 ng/mL (interquartile range: 210.18–582.50 ng/mL) and a median dose-adjusted concentration of 1.00 ng mL−1 (mg d−1)−1 (interquartile range: 0.63–1.53 ng mL−1 (mg d−1)−1). Concomitant medication use was common in this patient group. The main concomitant medications included the mood stabilizer valproate (n = 99), antipsychotics (eg, ziprasidone, n = 77), anxiolytics and antidepressants, hypnotics and sedatives, as well as adjunctive medications such as idebenone and trihexyphenidyl. Spearman correlation analysis revealed a statistically significant positive correlation between the daily amisulpride dose and plasma concentration (rs = 0.337; p < 0.001) (Figure 2). Despite the significant correlation, wide data-point dispersion suggested substantial interindividual variability in plasma amisulpride concentrations at the same daily dose. Relying solely on the administered dose makes it difficult to accurately predict drug exposure levels. Additionally, Table 1 provides other demographic and dosage information. Table 1Patient Demographic and Clinical CharacteristicsCharacteristicsSamples nRangeMean±SDMedian (Q1,Q3)Gender Male121 Female109Age, year18–7338.86±13.7333.0 (27.0–48.0) ≤4415918–4328.93±6.7529.0 (22.0–34.0) >447145–7354.61±7.1254.0 (49.0–59.75)Weight, kg39–10066.26±12.8265.0 (57.0–75.0)BMI, kg/m215.6–40.323.66±4.2823.45 (21.0–25.7)Daily dosing, mg/d100–800391.74±158.80400 (300–400)Plasma concentration, ng/mL3.7–1732.68432.60±293.86339.80 (210.18–582.50)Dose-corrected concentration (C/D) ((ng/mL)/(mg/day))0.07–4.331.16±0.721.00 (0.63–1.53) Ziprasidone77 Aripiprazole32 Risperidol18 Clozapine35 Trihexyphenidyl89 Idebenone60 Zopiclone56 Lorazepam41 Quetiapine30Antidepressants Citalopram12 Buspirone18Mood stabilizers Valproate99Notes: M (Q1,Q3), median (interquartile range).Abbreviation: BMI, Body mass index; C/D, Concentration-to-dose ratio. Patient Demographic and Clinical Characteristics Notes: M (Q1,Q3), median (interquartile range). Abbreviation: BMI, Body mass index; C/D, Concentration-to-dose ratio. Scatter plot of the correlation between the plasma concentration and daily dose of amisulpride. The red solid line represents the fitted regression line, with the red shaded area indicating the 95% confidence band for regression. The light-pink shaded area denotes the 95% prediction band for individual observations. The cyan and magenta ellipses represent the 95% confidence ellipse of raw data and predicted values, respectively. Baseline Characteristics Were Compared Between the Normal Prolactin and HPRL Groups Univariate analyses compared baseline clinical characteristics between patients with normal prolactin levels and those with HPRL. No significant differences existed between the two groups in terms of age, BMI, or most concomitant medications (including risperidone, quetiapine, olanzapine, alprazolam, oxazepam, and buspirone, among others) (p > 0.05). However, significant differences were observed between the two groups in terms of gender, blood concentrations of aripiprazole, and certain concomitant medications (zilazepam, aripiprazole, clozapine, idebenone, trihexyphenidyl and lorazepam) (p < 0.05); see Table 2 for details. Table 2Comparison of Baseline Characteristics Between the Normal Prolactin Group and the Hyperprolactinemia GroupFactorsAllNormal ProlactinHyperprolactinemiaχ2/ZP Value(N=209)(N=24)(N=185)Gender, n (%)5.1920.022 Male111 (53.11%)18 (75.00%)93 (50.27%) Female98 (46.89%)6 (25.00%)92 (49.73%)Age, n (%)0.04710.828 ≤44144 (68.90%)17 (70.83%)127 (68.65%) >4465 (31.10%)7 (29.17%)58 (31.35%)Plasma concentration, ng/mL339.80 [210.18; 582.50]234.2 [164.13; 264.03]461.05 [334.90; 612.51]−9.138<0.001BMI, kg/m2, n (%)0.1170.732 <28179 (85.65%)20 (83.33%)159 (85.95%) ≥2830 (14.35%)4 (16.67%)26 (14.05%)Ziprasidone, n (%)5.910.015 Yes75 (35.89%)14 (58.33%)61 (32.97%) No134 (64.11%)10 (41.67%)124 (67.03%)Aripiprazole, n (%)11.024<0.001 Yes32 (15.31%)3 (12.50%)29 (15.68%) No177 (84.69%)21 (87.50%)156 (84.32%)Risperidol, n(%)0.0190.891 Yes19 (9.09%)2 (8.33%)17 (9.19%) No190 (90.91%)22 (91.67%)168 (90.81%)Clozapine, n(%)6.0160.014 Yes35 (16.75%)8 (33.33%)27 (14.59%) No174 (83.25%)16 (66.67%)158 (85.41%)Trihexyphenidyl, n (%)9.4730.002 Yes87 (41.63%)17 (70.83%)70 (37.84%) No122 (58.37%)7 (29.17%)115 (62.16%)Idebenone, n (%)20.381<0.001 Yes58 (27.75%)16 (66.67%)42 (22.70%) No151 (72.25%)8 (33.33%)143 (77.30%)Zopiclone, n (%)0.4180.518 Yes55 (26.32%)5 (20.83%)50 (27.03%) No154 (73.68%)19 (79.17%)135 (72.97%)Lorazepam, n (%)6.7690.009 Yes38 (18.18%)9 (37.50%)29 (15.68%) No171 (81.82%)15 (62.50%)156 (84.32%)Quetiapine, n (%)2.4880.115 Yes30 (14.35%)6 (25.00%)24 (12.97%) No179 (85.65%)18 (75.00%)161 (87.03%)Citalopram, n (%)0.3350.563 Yes12 (5.74%)2 (8.33%)10 (5.41%) No197 (94.26%)22 (91.67%)175 (94.59%)Buspirone, n (%)0.00260.959 Yes18 (8.61%)2 (8.33%)16 (8.65%) No191 (91.39%)22 (91.67%)169 (91.35%)Valproate, n (%)0.1050.746 Yes98 (46.89%)12 (50.00%)86 (46.49%) No111 (53.11%)12 (50.00%)99 (53.51%)Notes: χ2: Chi-square test, Z: Z-test. Bold P values indicate statistically significant differences (P < 0.05).Abbreviation: BMI, Body mass index. Comparison of Baseline Characteristics Between the Normal Prolactin Group and the Hyperprolactinemia Group Notes: χ2: Chi-square test, Z: Z-test. Bold P values indicate statistically significant differences (P < 0.05). Abbreviation: BMI, Body mass index. Multiple Linear Regression Analyses This study employed a multiple linear regression model to systematically analyze the independent effects of gender, amisulpride plasma concentration, and concomitant medication on prolactin levels. Multiple linear regression was constructed, with predefined clinically relevant covariates (age and BMI) forced into the model regardless of univariate statistical significance. All independent variables were included in a unified regression model for adjustment analysis. Table 3 presents the model fit results. Table 4 lists the model-fit results, including regression coefficients (effect sizes), 95% confidence intervals, and P-values for each variable. These findings are further visualized in the forest plot (Figure 3). The results of the forest plot visualization, the regression parameter table, and the regression equation results are fully consistent, collectively reflecting the independent predictive effects of each factor on prolactin levels. Multivariate regression results indicate that amisulpride plasma concentration, gender, concomitant use of idebenone, and concomitant use of lorazepam are independent risk factors influencing prolactin levels (all p < 0.05). The plasma amisulpride concentration was significantly positively correlated with prolactin levels (β = 0.06584, 95% CI: 0.0339–0.0978, p = 0.001), suggesting that higher plasma concentrations are associated with a more significant increase in prolactin. The male gender was significantly negatively correlated with prolactin levels (β = −41.4451, 95% CI: −60.641–−22.249, p < 0.001), suggesting that male patients had significantly lower prolactin levels than females. Co-administration of idebenone significantly increased prolactin levels (β = 23.7766, 95% CI: 2.683–44.87, p = 0.0294). Concomitant use of lorazepam significantly reduced prolactin levels (β = −25.309, 95% CI: −49.125–−1.493, p = 0.0394). Furthermore, the combination of BMI, age, ziprasidone, aripiprazole, clozapine, and trihexyphenidyl had no significant independent effect on prolactin levels (all p > 0.05). This regression model explained 22.6% of the total variance in prolactin levels (R2 = 0.226, Table 3). Table 3Effects of Covariates on Prolactin Levels: Multiple Regression Analysis*FactorsCoefficientβtPConstant72.626Sex (1 male, 0 female)−41.445−0.291−4.232<0.001Plasma concentration, ng/mL0.0660.2874.048<0.001Age (years)−0.0527−0.104−1.4780.1408BMI, kg/m2−1.346−0.093−1.3300.1851Ziprasidone−12.3280.091−1.2750.2037Aripiprazole−11.8860.063−0.8880.3754Clozapine−6.5730.038−0.5010.6168Trihexyphenidyl8.1990.0620.8670.3870Idebenone23.7770.1562.1940.0294Lorazepam−25.3090.147−2.0730.0394Notes: *R2=0.226, coefficient of determination for the multiple regression model. Bold P values indicate statistically significant differences (P < 0.05).Abbreviation: BMI, Body mass index. Effects of Covariates on Prolactin Levels: Multiple Regression Analysis* Notes: *R2=0.226, coefficient of determination for the multiple regression model. Bold P values indicate statistically significant differences (P < 0.05). Abbreviation: BMI, Body mass index. Table 4Regression Coefficients and 95% Confidence Intervals for Factors Associated with Serum Prolactin Levels (Corresponding to the Forest Plot in Figure 3)VariablesCoefficients (β)SELower (95% CI)Upper (95% CI)P-Value(Intercept)72.6263---0.001Plasma amisulpride concentration0.065840.016270.03390.09780.001Sex (Male vs Female)−41.44519.794−60.641−22.249<0.001Ziprasidone combined−12.32799.6676−31.2766.620.2037Aripiprazole combined−11.886313.3785−38.00814.2350.3754Clozapine combined−6.572813.1151−32.15819.0120.6168Trihexyphenidyl combined8.19959.4584−10.24926.6480.387Idebenone combined23.776610.83572.68344.870.0294Lorazepam combined−25.30912.208−49.125−1.4930.0394Notes: Bold P values indicate statistically significant differences (P < 0.05).Abbreviations: β, regression coefficient; SE, standard error; CI, confidence interval. Regression Coefficients and 95% Confidence Intervals for Factors Associated with Serum Prolactin Levels (Corresponding to the Forest Plot in Figure 3) Notes: Bold P values indicate statistically significant differences (P < 0.05). Abbreviations: β, regression coefficient; SE, standard error; CI, confidence interval. Forest plot of factors influencing prolactin in multiple linear regression analysis. The vertical dashed line represents β = 0, indicating the factor exerts no effect on the outcome. No overlap between the 95% confidence interval and the dashed line indicates statistical significance (P < 0.05). Cutoff Values and Gender-Specific Predictive Efficacy of Prolactin Levels for HPRL We evaluated the predictive performance of amisulpride plasma concentrations for the risk of HPRL using ROC curve analysis, focusing on commonly used cutoff values and gender-specific differences (see Figure 4 for the results). Among 209 patients, when using a cutoff of 25 ng/mL, the AUC for prolactin in predicting HPRL was 0.917 (95% CI: 0.870–0.950, p < 0.001), with 83.03% sensitivity, 97.73% specificity, and 0.8076 Youden’s index (Figure 4A). When using a cutoff value of 28 ng/mL, the AUC was 0.920 (95% CI: 0.875–0.953, p < 0.001), with 85.26% sensitivity, 96.23% specificity, and 0.8148 Youden index (Figure 4B). No statistical comparison was performed between the two AUC values; however, the combined comparison curve (Figure 4C) indicates that the discriminatory performance of the 28 ng/mL cutoff value shows a trend toward superiority over that of the 25 ng/mL cutoff. Analysis stratified by sex (using a cutoff value of 28 ng/mL for both): Among 108 male patients, the AUC was 0.892 (95% CI: 0.818–0.944, p < 0.001), with 81.25% sensitivity, 96.43% specificity, and 0.7768 Youden index (Figure 4D). Among 101 female patients, the AUC was 0.954 (95% CI: 0.893–0.986, p < 0.001), with 86.84% sensitivity, 100.00% specificity, and 0.8684 Youden index (Figure 4E). Figure 4F shows that the model had significantly better predictive performance in females than in males. Receiver operating characteristic (ROC) curves of prolactin (PRL) levels for predicting the risk of hyperprolactinemia. (A) ROC curve of PRL at the cut-off value of 25 ng/mL (AUC = 0.9165, P < 0.001). (B) ROC curve of PRL at the cut-off value of 28 ng/mL (AUC = 0.9204, P < 0.001). (C) Comparison of ROC curves for PRL at different cut-off values: 28 ng/mL (red line, AUC = 0.9204) and 25 ng/mL (green line, AUC = 0.9165). (D) ROC curve of PRL in male patients (AUC = 0.8924, P < 0.001). (E) ROC curve of PRL in female patients (AUC = 0.9540, P < 0.001). (F) Comparison of ROC curves between female (red line, AUC = 0.9540) and male (green line, AUC = 0.8924) patients. Discussion Previous studies have confirmed TDM significance for clinical amisulpride utility. It is an effective method for enhancing therapeutic efficacy and ensuring medication safety, as well as a core tool for developing individualized treatment regimens and preventing and controlling endocrine adverse reactions in patients with mental disorders.20 Currently, the dose–response relationship between amisulpride plasma concentrations and adverse reactions has been thoroughly validated, providing strong support for including this drug in routine psychiatric TDM systems.21 Existing evidence indicates that elevated plasma concentrations are a key factor in triggering excessive prolactin secretion and inducing drug-induced HPRL. Despite the established pharmacological link between the two, significant individual variations in plasma concentrations persist even after patients reach the target dosages in clinical practice. The risk of prolactin disorders varies among patients receiving the same dosage, substantially increasing the difficulty of preventing and managing adverse reactions.22 Ružić K et al23 also confirmed that patients with therapeutic drug concentrations have a significantly higher risk of developing HPRL than those with lower concentrations. Data from this study revealed that within the standard dosage range of 100–800 mg/day, patients’ blood drug concentrations fluctuate widely, and individual differences in drug exposure are significant. Relying solely on dosage cannot identify high-risk populations for adverse reactions. Related international cohort studies have also indicated that aripiprazole exhibits a dose–concentration decoupling phenomenon, which further exacerbates individual variability in endocrine adverse reactions and impedes accurate risk prediction. Currently, comprehensive studies analyzing the association between demographic characteristics, plasma drug concentrations, and concomitant medications with abnormal elevations in prolactin levels are lacking, and no gender-specific warning thresholds exist for prolactin levels. To address this research gap, this study employed univariate analysis and multiple linear regression to identify independent risk factors for elevated prolactin levels and established gender-specific diagnostic cutoff values using ROC curves, confirming that prolactin levels have good predictive value for drug-induced HPRL. Our findings provide real-world, evidence-based support for clinical personalized dosing, standardized TDM, and stratified management of prolactin-related adverse reactions. Daily dosage is a key factor influencing amisulpride plasma concentrations. This study demonstrates a significant positive correlation between the daily amisulpride dose and plasma concentrations (r = 0.337, p < 0.001), which corroborates the findings of previous studies.24 However, the relatively low correlation coefficient and the scattered distribution of data suggest nonlinear pharmacokinetic characteristics, indicating a limited association between dose and concentration.25 This study included patients with mental disorders, who exhibited a wide range of ages, weights, and BMIs, and in whom polypharmacy was common. The data revealed that, at the same dose level, plasma concentrations varied widely among patients (3.7–1732.68 ng/mL), indicating marked interindividual variability in drug exposure. These results indicate that the daily dose is merely one of several factors influencing plasma drug concentrations. Covariates such as age, weight, BMI, and concomitant medication can all significantly interfere with drug levels in the body, which is the primary reason for the weak positive correlation between dose and concentration. Therefore, clinicians cannot rely solely on the administered dose to predict plasma drug concentrations or assess therapeutic efficacy and safety risks.26 Personalized therapy emphasizes a comprehensive consideration of patient demographics, physical indicators, and concomitant medication regimens; drug exposure should not be managed solely through dose adjustment.27 The standardized implementation of TDM is an effective means to address this shortcoming. The weak dose–concentration correlation and interindividual variability observed in this study corroborate previous large-scale TDM studies,28 further confirming the prevalence of the dose–concentration decoupling phenomenon for amisulpride and highlighting the clinical value of routine TDM. Age can markedly affect amisulpride plasma concentration. In this study, the dose-adjusted concentrations in middle-aged and elderly patients aged >44 years were significantly higher than those in younger patients, which corroborates previous reports.29 Based on the literature, this difference likely results from age-related physiological changes: amisulpride is primarily excreted via the kidneys, and the physiological decline in renal function with age may lead to reduced drug clearance.30 Combined with changes in the volume of distribution and concomitant medication use, these factors ultimately result in increased drug exposure in elderly patients.31 However, this study did not find a significant effect of age on prolactin levels (p = 0.828), suggesting that the age-related increase in plasma drug concentrations did not concurrently increase HPRL risk. This finding may be related to drug concentrations not reaching the threshold for adverse reactions,32 differences in dopamine D2 receptor sensitivity across age groups,33 and the confounding effects of concomitant medications, gender, and other factors. These findings suggest that age does not independently predict amisulpride-associated prolactin abnormalities. Clinically, when conducting TDM for middle-aged and elderly patients, drug safety should be assessed comprehensively alongside other factors.34 BMI have been reported to influence the pharmacokinetic characteristics of amisulpride in previous studies.35 A higher BMI may increase the apparent volume of distribution and accelerate amisulpride clearance, theoretically lowering systemic drug exposure and the risk of HPRL.36 In the present analysis, clinically important confounders including age and BMI were forced into the multiple linear regression model irrespective of their univariate statistical significance to avoid omitting potential confounding variables.37 Even so, neither age nor BMI emerged as an independent predictor of prolactin levels in the multivariate model. No significant difference in BMI distribution was detected between patients with normal prolactin and those with HPRL (χ2 = 0.117, p = 0.732). This lack of a significant association may be attributable to confounding by polypharmacy, variable drug exposure, comparable baseline BMI across study groups, limitations of BMI for discriminating fat and muscle mass, and interindividual variability in receptor sensitivity.38 Together, these factors may mask underlying associations. Larger stratified analyses are warranted to clarify whether physiological indices including age and BMI exert indirect effects on the risk of amisulpride-related HPRL.39 This study also identified sex as a significant factor associated with amisulpride-induced prolactin elevation. Female patients had a higher risk of prolactin-related adverse reactions, consistent with previous reports.40 Women have higher baseline prolactin levels, and their dopamine pathways are more sensitive to the drug’s effects, making them more prone to endocrine abnormalities.41 Comparative analysis shows that gender has greater predictive value for prolactin levels than age.42 Clinically, priority should be given to monitoring prolactin levels in female patients and dynamically adjusting treatment regimens based on TDM results to ensure therapeutic efficacy while reducing adverse reaction incidence. Table 2 shows that amisulpride use alongside other medications is significantly associated with HPRL. Drugs with statistically significant differences between groups included ziprasidone (χ2 = 5.91, p = 0.015), aripiprazole (χ2 = 11.024, p < 0.001), idebenone (χ2 = 20.381, p < 0.01), clozapine (χ2=6.016, p=0.014), lorazepam (χ2=6.769, p=0.009), and trihexyphenidyl (χ2=9.473, p=0.002). Aripiprazole can antagonize the amisulpride blocking effect on dopamine D2 receptors, thereby reducing the risk of elevated prolactin levels.43 ziprasidone and clozapine indirectly influence prolactin secretion by modulating serotonin receptors.44 Trihexyphenidyl can increase prolactin levels through its anticholinergic effects.45 Lorazepam regimen influences prolactin release by affecting dopamine pathways through GABA-A receptor activation.46 Idebenone may alter prolactin expression by regulating central oxidative stress and neurotransmitter balance.47 Clinicians should be vigilant regarding the risks of drug interactions, monitor treatment regimens individually, and reduce HPRL incidence. Based on the results of the multiple linear regression analysis and forest plot in this study (Tables 3 and 4, Figure 3), the key factors influencing the increase in prolactin levels with amisulpride included gender, drug concentration, and concomitant medication. The analysis identified female sex (β = −0.291, p < 0.001) and idebenone use (β = 0.156, p = 0.0294) as independent predictors of prolactin levels, whereas lorazepam use was associated with lower prolactin levels (β = −0.147, p = 0.0394). The model was adjusted for multiple confounding factors with a goodness-of-fit R2 = 0.226, suggesting that other influencing variables not included in the analysis exist. Gender effect on amisulpride pharmacokinetics has been established. Women typically have higher blood concentrations due to differences in body fat distribution and liver enzyme activity. Idebenone may increase amisulpride exposure levels by reducing its metabolic clearance through CYP450 enzyme system inhibition or mitochondrial function improvement. Lorazepam may lower amisulpride-induced prolactin levels by competing for plasma protein binding sites or inducing liver enzyme activity.48 A forest plot visually presents the regression coefficients and 95% confidence intervals for each variable, clearly demonstrating the significant effects of idebenone and lorazepam. In clinical practice, individualized dosing should be implemented based on gender and concomitant medications, and TDM should be strengthened to optimize efficacy and reduce adverse reaction risk. This study systematically evaluated the predictive performance of prolactin levels for amisulpride-induced HPRL through ROC curve analysis (Figure 4). The results showed that when using cutoff values of 25 ng/mL and 28 ng/mL, the AUCs for the overall population were 0.917 and 0.920, respectively, both demonstrating excellent discriminatory performance. The 28 ng/mL cutoff value exhibited higher sensitivity and a higher Youden’s index, suggesting its greater advantages in clinical screening. A gender-stratified analysis further revealed that the predictive performance in female patients was significantly superior to that in male patients (AUC 0.954 vs 0.892), with specificity reaching 100%, indicating that prolactin levels have a greater predictive value for HPRL in women. This gender difference likely results from the regulatory role of estrogen on prolactin secretion and differences in the sensitivity of the female pituitary to dopamine-inhibitory signals. High estrogen levels can reduce dopaminergic tone, making prolactin-secreting cells more susceptible to the effects of antipsychotic medications, thereby amplifying the predictive performance of serum prolactin levels. Our study findings support the use of 28 ng/mL as the threshold for elevated prolactin levels in patients treated with amisulpride in clinical practice. Notably, female patients should be monitored more closely to enable early identification and intervention for HPRL risk, thereby facilitating personalized medication management. Several limitations of the present study should be acknowledged when interpreting our findings. First, this retrospective single-centre study enrolled patients over a one-year period without a priori sample size calculation. Several medication and gender-stratified subgroups had relatively small sample sizes, which may compromise statistical power and limit the generalizability of subgroup findings. Multiple statistical tests across different concomitant medication subgroups may also increase the risk of error stemming from multiple comparisons. Additionally, multiple clinically important confounding factors that potentially influence amisulpride plasma concentrations and serum prolactin levels were not routinely documented, including renal function, smoking status, baseline hormonal profiles, and drug-metabolizing genetic polymorphisms. Smoking alters dopaminergic signaling and circulating prolactin levels during long-term antipsychotic treatment, possibly introducing residual confounding bias.49 Since amisulpride is predominantly eliminated via renal excretion, missing renal function data may bias the assessment of systemic drug exposure. Furthermore, the absence of genetic testing data in this cohort precluded the evaluation of polymorphisms, which are known to modulate individual prolactin responses to amisulpride. These unmeasured covariates collectively account for the modest explanatory power of the regression model (R2 = 0.226). Patients missing core measurements of amisulpride plasma concentration or serum prolactin were excluded from the analysis. No imputation was applied for missing auxiliary variables, and sensitivity analyses to assess the robustness of our findings were not undertaken. Larger follow-up studies are needed to replicate these analyses and validate our conclusions. Accordingly, the gender-specific plasma concentration cutoffs identified in the present analysis cannot be directly generalized to routine clinical practice; external validation in independent multicentre cohorts and external datasets is essential before widespread clinical implementation.50 Despite these limitations—including insufficient subgroup sample sizes, unavailable metabolic genotyping data, low model explanatory capacity (R2 = 0.226), and restricted representativeness due to the single-centre retrospective design—this study retains several notable strengths. First, this study employed a precise and reliable LC-MS/MS method to measure plasma amisulpride concentrations. The method was stable, and the detection results were accurate, providing high-quality baseline data for subsequent analysis. Second, all samples were collected under steady-state conditions and strictly adhered to predefined inclusion and exclusion criteria, effectively controlling confounding factors. This ensured that only steady-state trough concentration data were analyzed, thereby guaranteeing the results’ validity and representativeness. Third, this study comprehensively and clearly documented concomitant medication information, allowing for a clear presentation of the effects of co-administered drugs such as idebenone and lorazepam on the plasma concentrations of amisulpride. Furthermore, this study systematically evaluated the predictive performance of prolactin levels for HPRL using ROC curves, established gender-specific risk thresholds, and provided direct evidence for individualized clinical monitoring. These findings provide practical guidance for psychiatrists and clinical pharmacists. The gender-specific plasma concentration thresholds determined in this analysis can support standardized therapeutic drug monitoring for patients receiving amisulpride. Clinicians may combine patient sex and measured drug concentrations to conduct stratified risk screening for the early detection of hyperprolactinemia. This strategy enables individualized dose adjustment and patient management to minimize endocrine adverse reactions without compromising antipsychotic efficacy. Adding further, because the study is retrospective and conducted at a single centre, the proposed cut-off values should be interpreted with caution and validated prospectively in larger, multicentre populations before being adopted in routine clinical practice.51 Conclusions This study used univariate analyses and multiple linear regression to identify independent predictors of prolactin elevation in patients treated with amisulpride. Using these risk factor data, we performed ROC curve analyses to establish gender-stratified plasma concentration thresholds and diagnostic cutoffs for predicting amisulpride-induced HPRL. This integrated analytical approach generates evidence to support therapeutic drug monitoring, enabling screening of patients at high risk of hyperprolactinemia and informing individualized treatment decisions. Nevertheless, these results require careful interpretation. As a retrospective single-centre investigation with a limited sample size, the thresholds derived in this work are preliminary. Larger multicentre prospective cohorts covering diverse patient populations are needed to externally validate these findings before widespread clinical adoption. Acknowledgments We would like to thank Xi’an Mental Health Center for providing the publicly available data. Data Sharing Statement The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Ethics Approval and Consent to Participate The retrospective study received ethical approval from the Xi’an Mental Health Center, and abided by the guidelines of the Declaration of Helsinki (Approval No. XAJWKY-2026005). The waiver of informed consent was approved by the institutional review board for this retrospective data analysis study, as the research poses no adverse impacts on participants’ rights or health. In addition, all patient data were anonymized to safeguard data confidentiality and individual privacy. Author Contributions All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. Hui Wang: Conceptualization, Investigation, Formal analysis, Writing – original draft, Writing – review & editing; Qing Shao: Methodology, Software, Formal analysis, Data curation; Miaomiao Yang: Methodology, Investigation, Formal analysis; Yichen Chen: Methodology, Formal analysis, Data curation; Wenjie Wang: Methodology, Investigation, Data curation, Visualization. Disclosure The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. References Associated Data Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.