Elevated prolactin and association with treatment resistance indicators in first-hospitalized schizophrenia spectrum disorders: a real-world cohort study Authors: Dan Qi, Wenjie Sun, Meijuan Li, Tongxin Li, Jing Ma, Di Liu, Minghui Li, Chenghao Lu, Xinxu Wang, Yeqing Dong, Shaobing Li, Yanzhe Li, Jie Li, Shen Li, Nannan Liu Published: 2025 ABSTRACT BackgroundHyperprolactinemia frequently occurs during antipsychotic treatment but is also observed in antipsychotic-naïve schizophrenia patients. The relationship between prolactin (PRL) and treatment resistance remains underexplored in real-world populations.ObjectiveTo characterize age-/sex-stratified PRL distribution in first-hospitalized schizophrenia spectrum disorder (SSD) patients and examine associations with treatment resistance proxies (clozapine use/rehospitalization).MethodsWe analyzed 4,103 first-hospitalized SSD patients using real-world data. PRL levels were stratified by age/sex. Binary logistic regression evaluated PRL-clozapine/rehospitalization associations with confounder adjustment. ROC analysis assessed PRL’s discriminative capacity for clozapine use.ResultsElevated PRL prevalence was highest in females aged 18–45 (71.5%). PRL negatively correlated with clozapine use in males (P < 0.001), with lower median levels in users vs. non-users (374.40 vs. 529.10 mIU/L, P < 0.001). This persisted in males aged 18–45 (434.90 vs. 558.60 mIU/L) and ≥ 55 years (308.60 vs. 583.30 mIU/L) (both P < 0.01). In males ≥ 55, each 1 mIU/L PRL increase reduced clozapine use probability by 0.2% (aOR = 0.998, 95%CI:0.997-1.000). ROC analysis showed moderate discriminative capacity (AUC = 0.72, sensitivity 78.6%, specificity 39.4%). No significant associations occurred in females or for rehospitalization.ConclusionThe PRL demonstrates inverse, age-specific associations with clozapine use in male SSD patients, suggesting potential as a stratification biomarker for treatment resistance.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12888-025-07726-y. Introduction Antipsychotic medications remain the cornerstone of treatment for schizophrenia spectrum disorders (SSD) but commonly induce hyperprolactinemia [1]. Notably, 30–40% of antipsychotic-naïve first-episode patients exhibit elevated serum prolactin (PRL) levels [2, 3] – significantly exceeding general population prevalence (2.5%) [4]. This treatment-independent dysregulation implicates PRL in SSD pathogenesis through hypothalamic-pituitary-gonadal axis [5] and dopaminergic pathway disruptions [6]. Despite implications for PRL’s role in disease mechanisms, its clinical significance remains contested. While some studies report associations with symptom severity and subtypes (e.g., reduced levels in paranoid schizophrenia) [7], others demonstrate no correlations [8, 9] or suggest neuroprotective effects [10]. Critically, existing evidence derives from small cohorts (< 100 patients) with heterogeneous methodologies, precluding definitive conclusions. Consequently, a pivotal question remains unresolved: whether serum PRL predicts treatment-resistant schizophrenia (TRS) – a severe condition affecting 30% of SSD patients with substantial morbidity and healthcare burden [11]. This knowledge gap is clinically significant. No large-scale studies have systematically evaluated associations between PRL levels and treatment resistance outcomes in real-world cohorts [12]. Early biomarkers predicting TRS could enable personalized interventions and improve long-term outcomes. To address this gap, we leveraged large-scale real-world data to (1) characterize the distribution of serum PRL levels, stratified by age and sex, within a substantial cohort of first-hospitalized SSD patients; with explicit consideration of the temporal relationship between PRL sampling and hospital admission; and (2) investigate the association between PRL levels and clinically relevant proxies for treatment resistance. These proxies included the requirement for clozapine treatment (the gold-standard intervention for TRS) [13] and psychiatric rehospitalization within a defined follow-up period, serving as indicators of inadequate response to standard antipsychotic therapy and disease chronicity or relapse. A secondary objective was to evaluate the potential clinical utility of PRL as a stratification biomarker for these resistance indicators using receiver operating characteristic (ROC) curve analysis. We hypothesized that elevated PRL levels would be associated with an increased likelihood of clozapine use and rehospitalization, reflecting treatment resistance. Confirming this hypothesis would provide compelling evidence supporting PRL’s potential role not only in schizophrenia pathogenesis but also as a readily measurable biomarker for early risk stratification. Furthermore, elucidating any age- or sex-specific patterns in these associations could significantly inform the development of personalized treatment algorithms, aligning with the goals of precision psychiatry. Methods Data source Data for this study were obtained from the Electronic Health Records (EHR) database of Tianjin Anding Hospital, developed by the Beijing-Tianjin-Hebei Mental Health Alliance using the Observational Medical Outcomes Partnership (OMOP) Common Data Model [14]. Structured data were extracted through Structured Query Language (SQL). For unstructured data, natural language processing (NLP) models were applied to convert clinical text into a structured format. A sample of extracted results was cross-verified with the original medical records, and the process was deemed successful if the accuracy exceeded 90%; otherwise, manual annotation and adjustments were performed until the threshold was met. Study design and patient selection This retrospective cohort study utilized real-world data to investigate patients treated at Tianjin Anding Hospital between January 1, 2013, and December 31, 2019. The selection process followed the inclusion and exclusion criteria outlined in Fig. 1. The inclusion criteria were as follows: (1) First hospitalization with no prior outpatient or medication records; (2) A single diagnosis of SSD (the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM), codes F20.X-F29.X) at discharge. Exclusion criteria included: (1) A history or any records of diagnosed somatic diseases; (2) Pregnancy or nursing; (3) Prior use of medications that could affect PRL levels; (4) Modified electroconvulsive therapy (MECT) within 1 week prior to PRL testing (see Supplementary Materials for detailed definitions). Fig. 1Study population flowchart. EHR: the Electronic Health Records. SSD: schizophrenia spectrum disorders. PRL: prolactin. MECT: Modified electroconvulsive therapy Study population flowchart. EHR: the Electronic Health Records. SSD: schizophrenia spectrum disorders. PRL: prolactin. MECT: Modified electroconvulsive therapy Among 6,612 eligible patients, 2,509 (37.9%) had missing prolactin (PRL) data. Comparison between patients with and without PRL data showed differences in age and marital status but no significant difference in the rates of clozapine use or rehospitalization (primary outcomes). Patient characteristics extraction Demographic data (sex, age, marital status, ethnicity) and the first available serum PRL measurement with its corresponding test date were extracted from EHR. Sex-specific reference ranges defined normal PRL as 106–424 mIU/L for males and 148.4–530 mIU/L for females [15]. Two treatment resistance indicators were evaluated: (1) whether the patient was treated with clozapine during hospitalization, and (2) whether the patient was readmitted to the hospital after discharge (Detailed definitions in Supplementary Materials). Statistical analysis To address missing PRL data and preserve statistical power, Multiple Imputation was employed under the Missing at Random (MAR) assumption. Five imputed datasets were generated using a fully conditional specification model in SPSS 26.0, incorporating variables related to the probability of PRL measurement or its value (sex, age, marital status, ethnicity). Pooled estimates from these datasets form the basis of the primary analyses. Given known variations in PRL secretion by sex and age, all analyses were stratified accordingly, with patients categorized into four mutually exclusive age groups: <18, 18–44, 45–54, and ≥ 55 years [16]. Descriptive statistics for categorical variables are presented as counts and percentages, with between-group differences assessed using Chi-square tests. The normality of continuous variables was evaluated with the Kolmogorov-Smirnov test. As these variables were non-normally distributed, they are summarized as medians with interquartile ranges (IQR) and compared using the Mann-Whitney U test for two-group comparisons. To evaluate associations, binary logistic regression models were employed to assess the relationship between PRL levels and each treatment resistance proxy. All regression models included adjustment for the covariate “Days from admission to PRL test.” Results are reported as adjusted odds ratios (aOR) with 95% confidence intervals (CI). For specific subgroups identified in the regression analyses, the discriminative capacity of PRL for clozapine use was further evaluated using Receiver Operating Characteristic (ROC) curve analysis. The area under the curve (AUC), along with sensitivity and specificity at the optimal cutoff (determined by Youden’s index), are reported. Finally, to test the robustness of the primary findings, two sensitivity analyses were conducted: (1) repeating the regression analyses using only observed cases (n = 4,103), and (2) restricting the analysis to the Early Collection sub-cohort, defined as patients with PRL measured within 1 day of admission (n = 1,592). A two-tailed P-value of < 0.05 was defined as the threshold for statistical significance. Results Overall distribution of prolactin by sex, age, and collection time The distribution of PRL levels across sex and age groups is presented in Fig. 2, with patients stratified by PRL collection time relative to admission. Overall, a high prevalence of elevated PRL was observed. Among females with PRL data (n = 1027), 64.26% had levels above the sex-specific upper limit (530 mIU/L), with the highest proportion (69.01%) and median level (1,439.39 mIU/L) in the 18–44 age group (n = 652). Among males with PRL data (n = 565), 57.9% had levels above 424 mIU/L, with the highest proportion (60.7%) and median level (661.38 mIU/L) also in the 18–44 age group (n = 389). The proportion of patients with PRL levels below the lower limit of normal was low in both sexes (females: 4.6%; males: 2.7%). Fig. 2Overall distribution of prolactin Levels Overall distribution of prolactin Levels Differences in readmission and clozapine groups after sex stratification Demographic characteristics Table 1 shows that among the 4,103 patients with SSD included in the primary analysis, the age distribution of male patients (n = 1,567) showed significant differences. The median age of the clozapine use group (n = 218) was 39 years (IQR 21.00), significantly higher than that of the non-use group (n = 1349) at 35 years (IQR 20.00) (P = 0.002). The median age of readmitted patients (n = 418) was 34 years (IQR 21.00), significantly lower than that of non-readmitted patients (n = 1149) at 36 years (IQR 19.50) (P = 0.029). In contrast, female patients (n = 2,536) did not show such differences (Supplementary Table S1). However, in terms of marital status, the proportion of single individuals in the clozapine use group (n = 396) was 54.0%, significantly higher than that of the non-use group (37.9%) (P < 0.001), and there were also significant differences in the distribution of marital status in the readmitted group (P = 0.012) (Supplementary Table S1). Prolactin level characteristics among male patients Among male patients, the PRL level of the clozapine use group (n = 218) was significantly lower than that of the non-use group (n = 1349) (374.40 mIU/L, IQR 513.95 vs. 529.10 mIU/L, IQR 593.55) (P < 0.001) (Table 1). This difference was maintained in the 18-45-year-old group and the ≥ 55-year-old group. Specifically, for the 18-45-year-old group (n = 132 vs. n = 893), the levels were 434.90 mIU/L vs. 558.60 mIU/L, and for the ≥ 55-year-old group (n = 25 vs. n = 125), the levels were 308.60 mIU/L vs. 583.30 mIU/L, both with P < 0.01, indicating statistically significant differences (Table 2). However, there was no difference between the readmitted group (n = 418) (508.60 mIU/L, IQR 581.25) and the non-readmitted group (n = 1149) (506.10 mIU/L, IQR 598.95) (P = 0.152) (Table 1). In the female group, there were no significant differences in PRL levels regardless of whether they were stratified by clozapine use or readmission (Supplementary Table S1). Regarding the time of PRL collection, significant differences were observed in both male and female patients. The median collection time for the clozapine use group (n = 218) was significantly longer than that of the non-use group (n = 1349) (1,391.50 days, IQR 3,176.25 vs. 4.00 days, IQR 1,430.50) (P < 0.001) (Table 1 and Supplementary Table S1). Similarly, the readmitted group (n = 418) had a significantly longer median collection time than the non-readmitted group (n = 1149) (98.50 days, IQR 2,754.50 vs. 4.00 days, IQR 1,459.00) (P < 0.001) (Table 1). This difference was also observed in female patients (Supplementary Table S1). Further analysis of the differences in PRL collection time among male patients in different age groups and clozapine use groups showed significant differences in the 18-45-year-old group (n = 132 vs. n = 893) (1119.0 days vs. 3.0 days) and the 45-55-year-old group (n = 58 vs. n = 280) (1586.5 days vs. 135.0 days) (P < 0.01) (Table 2). Relationship between clozapine use and prolactin levels Further analysis has revealed a correlation between PRL levels and the use of clozapine (Table 3). Subsequent regression analysis indicated that this association is significantly influenced by age. Among patients aged 55 years and older, for each 1 mIU/L increase in PRL levels, there was a 0.2% decrease in the probability of clozapine use (adjusted odds ratio [aOR] = 0.998, 95% confidence interval [CI]: 0.997–1.000, P = 0.004). No significant associations were observed in other age groups, such as those under 18 years old and those between 18 and 45 years and 45–55 years (P > 0.05). To prevent false-positive results due to the interference of PRL collection time, we included the timing of PRL collection as a mediating variable to explore the mediating effects of early versus late PRL collection on the prediction of clozapine use (Supplementary Table S2). The results showed no statistically significant findings, indicating that our results are robust. ROC curve analysis in the male ≥ 55 years subgroup showed that PRL had modest discriminative capacity for clozapine use (area under the curve [AUC] = 0.721, 95% CI: 0.682–0.760). At the optimal cutoff, sensitivity was 78.6% and specificity was 39.4% (Fig. 3). The AUC values for other subgroups ranged from 0.57 to 0.59. Fig. 3ROC curve of prolactin levels and clozapine use ROC curve of prolactin levels and clozapine use However, when analysis was restricted to the Early Collection sub-cohort (PRL measured ≤ 1 day from admission, n = 1,592), the sample size in the male ≥ 55 years subgroup became very small (n = 39). In this analysis, the association between PRL and clozapine use was no longer statistically significant (P > 0.05; see Supplementary Table S3). Discussion This study is the first large-scale, real-world investigation to characterize PRL distribution and examine its association with proxies for treatment resistance in first-hospitalized patients with SSD. The key findings are as follows: (1) The prevalence of hyperprolactinemia in first-hospitalized SSD patients exceeded normative levels; (2) Significant differences in PRL levels were observed only in male patients stratified by clozapine use (P < 0.001), with no such differences detected in rehospitalization subgroups or female patients; (3) A robust negative correlation between PRL levels and clozapine use was identified, particularly prominent in male patients aged ≥ 55 years (adjusted OR = 0.998, 95% CI: 0.997–1.000); (4) Demographic analyses revealed that male clozapine users were older than non-users, whereas female patients exhibited significant marital status disparities without age-related differences. These findings provide novel insights into the physiological and clinical significance of PRL in SSD and support its potential utility as a biomarker for clinical decision-making. The baseline prevalence of hyperprolactinemia in our cohort(Fig. 2) (males: 57.88%; females: 64.26%) was higher than in healthy populations [17], consistent with prior reports [18, 19]. While some studies attribute PRL dysregulation to stress [20], childhood adversity [21, 22], emotional disturbances [23], or metabolic comorbidities [18], others found no significant associations [24]. Notably, our age- and sex-stratified analyses demonstrated that despite higher hyperprolactinemia rates in females, PRL levels showed no significant divergence between clozapine users and non-users in this subgroup. Potential explanations include: (1) Natural PRL fluctuations in females due to hormonal cycles [25], which may obscure clozapine-related differences; (2) Elevated estrogen levels in females, which modulate PRL secretion and protect dopaminergic function [25], potentially decoupling PRL variations from disease severity; (3) Real-world clinical adjustments for clozapine intolerance, more prevalent in females [26], leading clinicians to prioritize tolerability over disease severity and thereby attenuating PRL disparities between treatment groups. The core finding—an inverse association between PRL levels and clozapine use specific to older male patients—requires careful interpretation. This association was robust in our primary analysis using multiply imputed data and adjusted for sampling time. However, it was not statistically significant in a sensitivity analysis restricted to the ‘Early collection’ subgroup (PRL sampled ≤ 1 day from admission), likely due to a substantially reduced sample size. This discrepancy underscores the critical impact of sampling time heterogeneity in real-world research. Although clozapine is known to have relatively minor effects on PRL [27], and previous studies have shown that antipsychotic-induced hyperprolactinemia is more pronounced in female patients [28], our study did not observe significant differences in PRL levels between the female clozapine group and the rehospitalization group. To directly assess the role of medication timing, we compared PRL levels between the ‘Early collection’ and ‘Late collection’ (> 1 day) groups. The absence of a significant difference (Mann-Whitney U test, P > 0.05; Supplementary Table S2) argues against a major systematic effect of sampling time or acute medication exposure on PRL variation in this cohort. This supports the interpretation that the observed PRL differences may be more intrinsically linked to the disease state. Consequently, within the framework where clozapine use is a guideline-recommended proxy for treatment resistance [28, 29], our results suggest that in male patients—particularly older males—lower PRL levels may be indicative of greater disease severity and a higher likelihood of meeting criteria for TRS. The potential mechanisms underlying the observed sex- and age-specific association warrant consideration within established and emerging neurobiological frameworks. Our finding of lower PRL levels associated with clozapine use in older males may reflect a distinct pathophysiological subtype. One explanatory model posits that the illness involves excessive subcortical D2 receptor activation concurrent with cortical dopaminergic hypoactivity [30]. In this context, diminished PRL could indicate a more profound disruption of the tuberoinfundibular dopamine system’s inhibitory tone. This interpretation aligns with prior observations in antipsychotic-naïve males, where lower PRL has been linked to more severe positive symptoms [31] and other studies associating PRL abnormalities with emotional [32] and social deficits [33]. Although evidence across studies is not uniform [7], some longitudinal data suggest that higher PRL may predict better functional outcomes, while PRL receptor dysfunction could exacerbate socio-cognitive impairments [34]. This positions PRL as a multifunctional marker, potentially intersecting with both metabolic comorbidity and core disease processes. The pronounced association in older males may be further clarified by endocrine interactions. The dynamic interplay between PRL and testosterone [35] might render the hypothalamic-pituitary axis in males a more sensitive indicator of disease-related dopaminergic dysregulation. Furthermore, the age-specific finding in males ≥ 55 years could be partly attributable to the attenuation of sex-hormone confounders with age, potentially allowing clearer delineation of disease-specific associations [36]. This underscores the critical importance of age-stratified analyses in biomarker research for schizophrenia. From a broader, systems-level perspective, peripheral PRL abnormalities may not occur in isolation but could reflect interconnected systemic dysregulation. Emerging interdisciplinary research suggests that alterations in peripheral biomarkers can mirror central nervous system pathophysiology. For instance, the recent report of glymphatic system dysfunction in early psychosis [37]illustrates that disturbances in fundamental brain clearance mechanisms are part of the illness biology. This provides a novel conceptual context for interpreting PRL alterations, suggesting they may be one component of a wider systemic pathological process. Our study has several limitations inherent to its retrospective, real-world design. First, the reliance on EHR data precluded controlling for important unmeasured confounders (e.g., body mass index, smoking, detailed hormonal status, and acute stress levels), which may influence both PRL levels and clinical outcomes. Second, the single-timepoint measurement of PRL limits our ability to assess longitudinal dynamics or causality. Third, our proxies for treatment resistance—clozapine use and rehospitalization—while clinically relevant, do not capture granular symptom severity or functional outcomes. Fourth, the primary finding of an inverse association between PRL and clozapine use in males ≥ 55 years, though robust in the primary analysis, was not statistically significant in the sensitivity analysis restricted to the Early Collection sub-cohort. This discrepancy highlights the critical influence of sampling time heterogeneity and the reduced statistical power in subgroup analyses. Finally, the generalizability of our findings may be constrained by the single-center design and the specific demographic characteristics of our cohort. Conclusions This large-scale real-world study demonstrates a high prevalence of hyperprolactinemia in first-hospitalized SSD patients, extending beyond antipsychotic exposure. We identified a significant, inverse association between serum PRL levels and clozapine use specifically in older male patients (≥ 55 years), suggesting that lower PRL may be linked to a clinical profile requiring advanced treatment. No such association was observed with rehospitalization or in female patients. These findings underscore the importance of age- and sex-stratified analyses in biomarker research for schizophrenia. While serum PRL shows potential as a stratifying biomarker, particularly in older males, its clinical utility requires further validation in prospective studies with standardized sampling protocols, longitudinal PRL assessments, and integration with comprehensive clinical and biomarker profiles. Table 1Comparison of demographics and prolactin levels between male patients in clozapine group and rehospitalization group [median (IQR)/N (%)]CharacteristicsClozapine (N = 1567/4103)z/χ2P-valueRehospitalization (N = 1567/4103)z/χ2P-valueYes (N = 218)No (N = 1349)Yes (N = 418)No (N = 1149)Age (years) (median, IQR)39(21.00)35.00 (20.00)-3.1080.002*34.00(21.00)36.00(19.50)-2.1770.029*Marital status, N (%)6.2190.1833.6360.457Single147/218839/1349264/418722/1149Married42/218355/1349104/418293/1149Divorced18/21898/134936/41880/1149Losing spouse11/21852/134914/41849/1149Other0/2185/13490/4185/1149Ethnicity, N (%)0.6760.8791.3070.727Han216/2181331/1349413/4181134/1149Manchu0/2182/13491/4181/1149Hui1/21811/13492/41810/1149Other1/2185/13492/4184/1149Prolactin (mIU/L, median, IQR)374.40 (513.95)529.10 (593.55)-4.724 < 0.001** 508.60 (581.25)506.10 (598.95)-1.4320.152Time of Prolactin Collection(days) (median, IQR)1391.50 (3176.25)4.00 (1430.50)-7.336 < 0.001** 98.50(2754.50)4.00 (1459.00)-3.630 < 0.001** Unless otherwise indicated, continuous variables are presented as median, interquartile rangeAbbreviations: IQR, interquartile range; mIU/L, milli-International Units per liter; χ², chi-square value; Z, Z-score; Time of Prolactin Collection: The number of days from prolactin collection to hospital admission. *P < 0.05, **P < 0.01 Table 1Comparison of demographics and prolactin levels between male patients in clozapine group and rehospitalization group [median (IQR)/N (%)]CharacteristicsClozapine (N = 1567/4103)z/χ2P-valueRehospitalization (N = 1567/4103)z/χ2P-valueYes (N = 218)No (N = 1349)Yes (N = 418)No (N = 1149)Age (years) (median, IQR)39(21.00)35.00 (20.00)-3.1080.002*34.00(21.00)36.00(19.50)-2.1770.029*Marital status, N (%)6.2190.1833.6360.457Single147/218839/1349264/418722/1149Married42/218355/1349104/418293/1149Divorced18/21898/134936/41880/1149Losing spouse11/21852/134914/41849/1149Other0/2185/13490/4185/1149Ethnicity, N (%)0.6760.8791.3070.727Han216/2181331/1349413/4181134/1149Manchu0/2182/13491/4181/1149Hui1/21811/13492/41810/1149Other1/2185/13492/4184/1149Prolactin (mIU/L, median, IQR)374.40 (513.95)529.10 (593.55)-4.724 < 0.001** 508.60 (581.25)506.10 (598.95)-1.4320.152Time of Prolactin Collection(days) (median, IQR)1391.50 (3176.25)4.00 (1430.50)-7.336 < 0.001** 98.50(2754.50)4.00 (1459.00)-3.630 < 0.001** Unless otherwise indicated, continuous variables are presented as median, interquartile rangeAbbreviations: IQR, interquartile range; mIU/L, milli-International Units per liter; χ², chi-square value; Z, Z-score; Time of Prolactin Collection: The number of days from prolactin collection to hospital admission. *P < 0.05, **P < 0.01 Comparison of demographics and prolactin levels between male patients in clozapine group and rehospitalization group [median (IQR)/N (%)] Unless otherwise indicated, continuous variables are presented as median, interquartile range Abbreviations: IQR, interquartile range; mIU/L, milli-International Units per liter; χ², chi-square value; Z, Z-score; Time of Prolactin Collection: The number of days from prolactin collection to hospital admission. *P < 0.05, **P < 0.01 Table 2Comparison of demographics and prolactin levels between male patients in different age cohorts [median (IQR)/N (%)]CharacteristicsMale patients in clozapine group (N = 1567)YES (N = 218)NO (N = 1349)< 18(N = 3)18–45(N = 132)45–55(N = 58)≥ 55(N = 25)< 18(N = 51)18–45(N = 893)45–55(N = 280)≥ 55(N = 125)Age (years) (median, IQR) 39.00(21.00) * 35.00 (20.00) * Marital status, N (%)Single3/3108/13227/589/2551/51661/89391/28036/125Married0/314/13218/5810/250/51162/893129/28064/125Divorced0/33/13210/585/250/5135/89346/28017/125Losing spouse0/37/1323/581/250/5130/89314/2808/125Other0/30/1320/580/250/515/8930/2800/125Ethnicity, N (%)Han3/3132/13258/5823/25*50/51879/893277/280125/125*Manchu0/30/1320/580/25*0/512/8930/2800/125*Hui0/30/1320/581/25*0/519/8932/2800/125*Other0/30/1320/581/25*1/513/8931/2800/125*Prolactin (mIU/L, median, IQR)934.60(--) 434.90(537.88) ** 342.30(481.20) 308.60(215.30) ** 475.90(475.30) 558.6(618.05) ** 458.70(516.12) 583.30(647.15) ** Time of Prolactin Collection(days) (median, IQR)3.00(--) 1119.00(3409.00) ** 1586.5(2953.25) ** 1742.00(2487.00)1.00(3.00) 3.00(938.5.00)** 135.00(2848.25) ** 184.00(2307.50)Unless otherwise indicated, continuous variables are presented as median, interquartile rangeAbbreviations: IQR, interquartile range; mIU/L, milli-International Units per liter; Time of Prolactin Collection: The number of days from prolactin collection to hospital admission. *P < 0.05, **P < 0.01 Table 2Comparison of demographics and prolactin levels between male patients in different age cohorts [median (IQR)/N (%)]CharacteristicsMale patients in clozapine group (N = 1567)YES (N = 218)NO (N = 1349)< 18(N = 3)18–45(N = 132)45–55(N = 58)≥ 55(N = 25)< 18(N = 51)18–45(N = 893)45–55(N = 280)≥ 55(N = 125)Age (years) (median, IQR) 39.00(21.00) * 35.00 (20.00) * Marital status, N (%)Single3/3108/13227/589/2551/51661/89391/28036/125Married0/314/13218/5810/250/51162/893129/28064/125Divorced0/33/13210/585/250/5135/89346/28017/125Losing spouse0/37/1323/581/250/5130/89314/2808/125Other0/30/1320/580/250/515/8930/2800/125Ethnicity, N (%)Han3/3132/13258/5823/25*50/51879/893277/280125/125*Manchu0/30/1320/580/25*0/512/8930/2800/125*Hui0/30/1320/581/25*0/519/8932/2800/125*Other0/30/1320/581/25*1/513/8931/2800/125*Prolactin (mIU/L, median, IQR)934.60(--) 434.90(537.88) ** 342.30(481.20) 308.60(215.30) ** 475.90(475.30) 558.6(618.05) ** 458.70(516.12) 583.30(647.15) ** Time of Prolactin Collection(days) (median, IQR)3.00(--) 1119.00(3409.00) ** 1586.5(2953.25) ** 1742.00(2487.00)1.00(3.00) 3.00(938.5.00)** 135.00(2848.25) ** 184.00(2307.50)Unless otherwise indicated, continuous variables are presented as median, interquartile rangeAbbreviations: IQR, interquartile range; mIU/L, milli-International Units per liter; Time of Prolactin Collection: The number of days from prolactin collection to hospital admission. *P < 0.05, **P < 0.01 Comparison of demographics and prolactin levels between male patients in different age cohorts [median (IQR)/N (%)] 558.6(618.05) ** Unless otherwise indicated, continuous variables are presented as median, interquartile range Abbreviations: IQR, interquartile range; mIU/L, milli-International Units per liter; Time of Prolactin Collection: The number of days from prolactin collection to hospital admission. *P < 0.05, **P < 0.01 Table 3Correlation of serum prolactin with clozapine use stratified by age or prolactin collection time in malesVariableΒWaldc2P-valueOR (95% CI)Age stratification (years)< 180.0010.2830.5951.001 (0.999, 1.003)18–45< 0.0012.8660.0901.000(1.000, 1.000)45–55< 0.0010.5300.4661.000(1.000, 1.000)≥ 55-0.0028.136 0.004** 0.998 (0.997, 1.000)Abbreviations: OR = odds ratio; CI = confidence interval the age stratification, controlling for Time of Prolactin Collection(days) Table 3Correlation of serum prolactin with clozapine use stratified by age or prolactin collection time in malesVariableΒWaldc2P-valueOR (95% CI)Age stratification (years)< 180.0010.2830.5951.001 (0.999, 1.003)18–45< 0.0012.8660.0901.000(1.000, 1.000)45–55< 0.0010.5300.4661.000(1.000, 1.000)≥ 55-0.0028.136 0.004** 0.998 (0.997, 1.000)Abbreviations: OR = odds ratio; CI = confidence interval the age stratification, controlling for Time of Prolactin Collection(days) Correlation of serum prolactin with clozapine use stratified by age or prolactin collection time in males Abbreviations: OR = odds ratio; CI = confidence interval the age stratification, controlling for Time of Prolactin Collection(days) Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1