The association of prolactin and gonadal hormones with cognition and symptoms in men with schizophrenia spectrum disorder: Divergent effects of testosterone and estrogen

.


Introduction
Antipsychotic medications, which primarily target dopamine D2 receptors, are commonly prescribed to alleviate symptoms in patients with schizophrenia spectrum disorders (SSD).However, these medications can lead to dysregulated prolactin secretion, since dopamine exerts tonic inhibitory control over prolactin secretion.Thereby, certain strong dopamine antagonists (i.e., haloperidol, amisulpride, paliperidone and risperidone) can profoundly increase prolactin secretion by reducing this tonic inhibitory control of dopamine (Leblanc et al., 1976;Peuskens et al., 2014).On the other hand, antipsychotics characterized as weaker dopamine antagonists or partial agonists (e.g., clozapine, quetiapine, aripiprazole), induce minimal or negligible elevations in prolactin levels (Peuskens et al., 2014).
Increases in prolactin secretion can result in hyperprolactinemia, which has been associated with several side effects such as sexual dysfunction and hypogonadism (Peuskens et al., 2014).In addition to the effects on general health, increased prolactin levels can exert negative effects on cognitive functioning and symptoms.In men with SSD, elevated prolactin has been associated with impaired working memory (Moore et al., 2013), reduced processing speed (Montalvo et al., 2018), and increased negative symptom scores (Ates et al., 2015).
This association between prolactin and side effects involves a complex interplay possibly involving gonadal hormones.Prolactin regulates the levels of gonadal hormones including testosterone and estrogen via the hypothalamic-pituitary-gonadal (HPG) axis (Dwyer and Quinton, 2019).More specifically, high levels of prolactin result in a decrease in the secretion of other pituitary hormones, like FSH and LH, subsequently leading to reduced production of gonadal hormones (Riecher-Rössler, 2017;van Rijn et al., 2011).In women with SSD, antipsychotic-induced prolactin elevations are associated with reduced estrogen levels.Furthermore, higher levels of prolactin were associated with increased symptom scores and reduced cognitive functioning, suggesting that prolactin elevations hinder the protective effects of estrogen (Brand et al., in press).In men with SSD, an inverse relationship between prolactin and testosterone levels has been established (Tasaki et al., 2021), and higher testosterone levels are linked to lower negative symptom scores (Ko et al., 2007;Sisek-Šprem et al., 2020).A previous study, using a subsample of the current study, found that higher testosterone was associated with better cognitive functioning (Moore et al., 2013), although contrasting findings have been reported in other studies, with some indicating an inverse relationship (Li et al., 2015), and others showing no significant associations (Halari et al., 2004).Regarding estrogen in men with SSD, negative associations with cognitive functioning (Moore et al., 2013) and negative symptoms (Kaneda and Ohmori, 2005) have been identified, however, positive effects of estradiol treatment have also been reported (Kulkarni et al., 2011).Both the involvement of antipsychotics, which tend to elevate prolactin levels, and the relatively small sample sizes of previous studies have made it challenging to delineate the precise effect of prolactin and gonadal hormones on symptoms and cognitive functioning in men with SSD.Therefore, a clear understanding of the interplay between antipsychoticinduced prolactin elevation, gonadal hormones, symptoms and cognitive functioning is needed by investigating a larger SSD sample.
The current study aimed to investigate whether the negative effects of increased prolactin are a result of prolactin's inhibitory effect on testosterone, similar to how this happens through estrogen in women with SSD.To increase our sample size, we combined baseline data of men with SSD from two clinical trials.We investigated the known prolactin-raising effects of antipsychotic medication, as well as how elevated prolactin levels relate to circulating testosterone and estrogen concentrations and to test how these may be associated with symptom scores and cognitive functioning.First, we compared prolactin levels among healthy controls, patients receiving prolactin-raising antipsychotics, and patients receiving prolactin-sparing antipsychotics.Second, we investigated the relationships among prolactin, testosterone and estrogen in each group.Third, we evaluated how these hormone levels relate to cognitive functioning and symptoms in men with SSD.Based on previous studies, we hypothesize that testosterone and prolactin levels are negatively correlated and that higher testosterone levels will positively relate to cognition and inversely relate to symptom scores.Additionally, we expect that higher estrogen levels may also correlate with cognition and symptoms in men.

Methods
This cross-sectional study combined pre-treatment data of the male participants with chronic schizophrenia from two different clinical trials (Brand et al., 2020;Weickert et al., 2015).The primary rationale for this approach was to increase the sample size, thereby enhancing statistical power and improving the generalizability of our findings.The first trial was conducted in Australia, including SSD patients and a comparator group consisting of healthy controls (Weickert et al., 2015: ACTRN 1260800461392).The second trial, conducted in Europe, only included SSD patients (Brand et al., 2020;NCT03043820).Both studies were approved by the associated medical ethical committees (Australia: University of New South Wales (07/121 and 09/187), South Eastern Sydney and Illawarra Area Health Service (07-259), Queen Elizabeth Hospital Ethics and Human Research Committee, Adelaide (2010188); Netherlands: University Medical Centre of Utrecht (16-103)) and followed the Declaration of Helsinki and Good Clinical Practice guidelines (64th WMA General Assembly, October 2013).Both trials shared the same study design, obtained similar study measures and used similar inclusion criteria, because of this, the combination of these samples was considered appropriate.

Participants
The inclusion criteria for patients were a current SSD diagnosis, confirmed by a structured clinical interview (First, 2007;Overbeek et al., 1999), and the use of a stable dose of antipsychotic medication for at least two weeks at the time of inclusion.Exclusion criteria were a history of pre-existing cardiovascular disease, uncontrolled diabetes or hypertension, history of substance abuse or dependence, any liver function or enzyme disorder, current pregnancy or breastfeeding, and the use of non-antipsychotic medication with prolactin-raising properties.For the healthy male control group, additional exclusion criteria consisted of a personal history or a family relative with a DSM-IV Axis I psychiatric diagnosis.A brief overview of the relevant study procedures is described below, extensive descriptions of the study procedures are described elsewhere (Brand et al., 2020;Weickert et al., 2015).

Clinical assessment
Demographic and clinical characteristics including age, duration of illness (DOI), and age of onset (AOO) were documented.Symptoms during the past seven days were assessed by a trained researcher or psychologist using the Positive and Negative Syndrome Scale (PANSS; Kay et al., 1987).In addition to the total score, 5 subscale scores were calculated using the PANSS 5-factor model: positive, negative, cognitive, excitement/hostility, and depression/anxiety symptoms (Citrome et al., 2011).Dosages of antipsychotic medications were converted into olanzapine equivalents according to Gardner et al. (2010) and Leucht et al. (2016).

Cognitive assessment
Cognition was assessed using either the Wechsler Adult Intelligence Scale-3rd edition (WAIS-III: Wechsler, 1997) for the Australian sample, and the Brief Assessment of Cognition in Schizophrenia (BACS: Keefe et al., 2004) for the European sample.Cognition scores were combined into three domains: processing speed, verbal fluency and working memory (Table A.1). Consequently, cognitive test scores were converted into age-and sex-corrected Z-scores to enable comparisons across the two studies.

Hormone levels
To obtain hormone levels, while minimizing variation induced by the circadian rhythm, fasting peripheral blood was collected between 8:30 am and 1:30 pm according to protocol.Hormone concentrations were determined using electrochemiluminescence enzyme immunoassay (ECLIA), using the Siemens Immulite 2000 for the Australian sample and the Beckman DXi and Atellica IM Analyzers for the European sample.The intra-and inter-assay coefficients of variability (CVs) for prolactin in the Australian sample were 3.4 % and 6.8 %, respectively.For the Australian sample, the interassay CV for estrogen was 13.5 %.In the European sample, the inter-assay CV for prolactin and estrogen were 3.1 % and 7.4 % respectively.The inter-assay CV for testosterone was 9.7 % for the Australian sample, and 3.9 % (Atellica) and 6.8 % (Beckman) for the European sample.The cut-off for defining hyperprolactinemia was set to 0.42 IU/L (20 ng/mL) which is considered as the upper limit for prolactin serum concentrations in men (Peuskens et al., 2014).

Prolactin-raising/sparing antipsychotics
Patients were divided into the prolactin-raising or prolactin-sparing group based on their antipsychotic treatment.The prolactin-sparing antipsychotics constituted aripiprazole, clozapine, asenapine, quetiapine, and flupentixol.Paliperidone, risperidone, haloperidol, olanzapine, zuclopenthixol, pimozide and amisulpride were classified as prolactin-raising antipsychotics (Huhn et al., 2019;Peuskens et al., 2014).As aripiprazole is known to counteract the prolactin-raising properties of other medications, patients receiving polytherapy with aripiprazole were placed into the prolactin-sparing group (Labad et al., 2020).All remaining patients receiving polytherapy with at least one prolactin-raising antipsychotic were allocated to the prolactin-raising group.

Statistical analyses
Statistical analyses were performed in R (v 4.3.0).For all variables, normality was checked using the Shapiro-Wilk test.If normality was violated, equivalent non-parametric tests were performed.For all tests, the significance level was set to 95 %, and in the case of multiple tests, Benjamin-Hochberg False Discovery Rate corrections were applied.Demographic and clinical variables were compared between the Australian and European samples, using a two-tailed independent samples t-test or a non-parametric Wilcoxon Test.
First, age, hormone serum levels, and cognition were compared between the prolactin-raising, prolactin-sparing and healthy control groups through a one-way ANOVA or non-parametric Kruskal-Wallis H test followed by post hoc comparisons (Tukey's HSD Test or Dunn's Test, respectively).Second, all clinical variables were compared between the prolactin-sparing and prolactin-raising groups, using independent samples t-tests or non-parametric Wilcoxon signed-rank tests.Third, Pearson's correlation analyses were performed to examine the relationships of prolactin to testosterone and estrogen in the healthy control, prolactin-raising and prolactin-sparing groups.When necessary, hormone levels were log-transformed to meet the normality assumption.
Finally, a series of backward linear regression analyses were conducted.The regression analyses were performed on the full patient sample to maintain sufficient statistical power.Separate regression models were constructed for each cognition domain (i.e., verbal fluency, working memory, processing speed) with the three hormones (prolactin, estrogen, and testosterone), age, years of education, and daily olanzapine equivalent dose as predictor variables.To account for the potential confounding effects of combining the two samples, site was also included as a predictor in the regression models.Regarding symptom scores, separate models were created for each subscale, with age of onset and illness duration as predictors, in addition to those noted above in relation to cognition.For the healthy control group, comparable models for each cognitive domain were established, except for the daily antipsychotic dose due to its inapplicability to this group.All models were checked to ensure the assumptions for linear regression were met.

Results
A total of 172 participants were included in this study, of which 128 men with SSD and 44 healthy men.Men with SSD had mild to moderate symptom severity (Table 1).Compared to the European sample (n = 71, 55.5 %), Australian men (n = 57, 44.5 %) had higher PANSS negative scores (Z = 1348.0,p adj = 0.012), and higher estrogen levels (Z = 1170.0,p adj = 0.004).Other demographic or clinical variables were not  significantly different between sites (Table A.2).

Group comparisons
The demographic and clinical information, together with the hormone levels of all groups are shown in Table 1.Compared to the men with SSD, the healthy men had a significantly lower age (H(2) = 15.23,p < .001)and more years of education (H(2) = 19.78,p < .001;Table 1).No significant differences were found in demographic characteristics, symptoms or cognition scores between the prolactin-raising and sparing SSD groups, apart from a higher daily antipsychotic dose in the prolactin-sparing group (Z = 2624.5,p = .023).
The three groups differed significantly in prolactin levels (H = 45.279,p < .001) as shown in Fig. 1A.As expected, post-hoc analyses revealed higher prolactin levels in the prolactin-raising group compared to both the prolactin sparing (p adj < 0.001) and the control group (p adj < 0.001) confirming that we classified most patients correctly.Furthermore, prolactin levels in the prolactin-sparing group were significantly lower compared to the healthy control group (p adj = 0.004).Both testosterone and estrogen did not significantly differ across the three groups (Fig. 1B/C).As expected, cognition scores on all domains were significantly lower in men with SSD as compared to healthy controls but did not differ between the two SSD groups (Table 1).

Correlations between prolactin, estrogen and testosterone
A moderate positive correlation between testosterone (T) and estrogen (E) levels was found in the healthy control group (r(42) = 0.409, p = .006)and a weak positive correlation was identified in the total patient group (r(126) = 0.240, p = .007;Fig. 2).In both groups, prolactin (PRL) levels were not associated with either estrogen (control p = .411,patients p = .536)or testosterone levels (control p = .363,patients p = .310).In the prolactin-raising subgroup, a significant negative correlation (r(73) = − 0.33, p = .004)was found between prolactin and testosterone levels (Fig. 3).Additionally, in line with the results for the whole patient group, a significant positive correlation was identified between testosterone and estrogen levels in the prolactin-raising group (r(73) = 0.30, p = .009).In the prolactin-sparing group, no statistically significant correlations were found among any of the three hormones (PRL ~ T p = .419,PRL-E p = .065,T ~ E p = .264).

Regression analysis of hormonal influences on symptoms and cognition
Regarding cognitive functioning, backward linear regression analyses revealed that both education and testosterone levels were significant positive predictors for working memory and verbal fluency in men with SSD (Table 2).For processing speed, education was a significant positive predictor, whereas estrogen was a significant negative predictor in men with SSD, indicating that increased levels of estrogen predicted slower processing speed (Table 2).In the control group, only testosterone was found to be a significant negative predictor (β = − 0.32, t(41) = − 2.2, p = .03)for processing speed (R 2 = 0.17, F(2, 41) = 4.08, p = .024),but no predictors were significantly associated with working memory or verbal fluency.
Regarding symptoms, both testosterone and estrogen were significant predictors for total and cognitive PANSS scores in men with SSD (Table 3), with testosterone also being a significant predictor for negative PANSS scores.Per the negative beta values, higher serum testosterone levels predicted lower PANSS scores.The opposite was true for estrogen, as positive beta values indicated that higher serum estrogen levels predicted higher PANSS scores.On the positive PANSS factor scale, the daily antipsychotic dose was a significant positive predictor (Table 3).Furthermore, education and was associated with cognitive symptoms, such that more years of education were associated with lower cognitive deficits (Table 3).None of the demographic or hormonal

Table 3
Summary of final multiple linear regression models examining the effects of testosterone, estrogen, prolactin and covariates (predictor variables) on PANSS factor scales in men with SSD.predictors were significantly associated with the PANSS subscales depression/anxiety or excitement/hostility. Site was not a significant predictor in any of the models for cognition or symptoms.

Discussion
This study investigated whether gonadal hormones are associated with cognitive functioning and symptoms in men with SSD while considering the influence of antipsychotic-induced prolactin elevations.Prolactin levels were highest in the patients using prolactin-raising antipsychotics, followed by the control group, and lowest in the prolactinsparing group.The degree of potential prolactin elevation by different antipsychotics did not appear to be associated with differences in testosterone and estrogen levels across the three groups.Nevertheless, prolactin levels were inversely correlated with testosterone only in the prolactin-raising group.In the total male SSD sample, serum testosterone emerged as a strong predictor for better cognitive functioning and reduced overall, negative, and cognitive symptom scores.Conversely, we found a negative effect of serum estrogen levels on certain domains of symptoms and cognitive functioning in men with SSD.
Our finding of increased prolactin levels and higher prevalence of hyperprolactinemia in the prolactin-raising group aligns with the known inducing effect of certain antipsychotic medications on prolactin secretion (Huhn et al., 2019;Peuskens et al., 2014).Interestingly, serum prolactin levels in the prolactin-sparing group were lower than the serum prolactin levels in healthy controls, suggesting that prolactinsparing antipsychotics may not only 'spare' prolactin levels, but may lower them beyond normal levels.This can partially be attributed to aripiprazole being the most frequently prescribed antipsychotic in the prolactin-sparing group.As a partial dopamine D2 agonist, aripiprazole stimulates the inhibitory control of dopamine on prolactin secretion, thereby reducing prolactin levels (Coward, 1992;Labad et al., 2020).While adverse effects associated with hyperprolactinemia are extensively documented, it should be noted that lower prolactin levels can also give rise to adverse effects, such as metabolic diseases and reproductive disorders (Rastrelli et al., 2015).Future studies should examine the potential adverse effects related to decreased prolactin levels to determine whether it is important to monitor prolactin levels and potential side effects in patients receiving prolactin-sparing medication.
Our findings indicated a negative association between prolactin and testosterone specifically in the prolactin-raising group, which is in line with previous research in patients receiving prolactin-raising antipsychotics (Tasaki et al., 2021).This negative correlation between prolactin and testosterone serum levels was specific to prolactin-raising antipsychotics, indicating that this association possibly requires a greater variation in prolactin levels to become evident.The interplay between prolactin and testosterone may therefore be contingent on higher prolactin levels triggered by specific dosages or types of antipsychotic medication.It is important to consider antipsychotic-induced prolactin elevations in the context of a potentially associated testosterone reduction and its downstream effects when deciding on treatment.
Higher testosterone levels were associated with better working memory and verbal fluency in men with SSD.In contrast to the previous study by Moore et al. (2013), we found an association of higher testosterone levels with faster processing speed in controls but not in patients.In healthy young men, other prior research has failed to establish a direct link between testosterone and cognition (Ulubaev et al., 2009), but in older men with lower testosterone levels, a negative association has been implied (Dong et al., 2021).The weaker relationship between testosterone and cognition in controls may suggest that the positive effects of testosterone may be more evident when testosterone levels are decreased (i.e., with aging or prolactin-raising drugs).This underscores the importance of maintaining healthy testosterone levels in men with SSD to mitigate cognitive impairment.Regarding symptoms, we replicated earlier findings of an association of increased testosterone with fewer negative symptoms (Ko et al., 2007), and extended these findings to overall symptom scores and symptoms related to cognition.This is in line with previous findings where increased testosterone has also been associated with reduced excitement/hostility symptoms (Moore et al., 2013).Combined, these findings accentuate the potential benefit of increasing testosterone levels in addressing both cognitive deficits and negative symptoms, which often persist in SSD patients despite antipsychotic treatment.
In contrast to our findings related to testosterone, higher estrogen levels in men with SSD were associated with decreased processing speed and higher scores on total and cognitive symptom scales.Previous studies showed a negative association between estrogen and negative symptoms (Kaneda and Ohmori, 2005) and processing speed (Moore et al., 2013) in men with SSD.We replicated the finding for processing speed in this larger sample.However, we did not identify an association between estrogen and negative symptoms, which may be due to these variables differing significantly between sites.These findings imply that the protective effects of higher circulating estrogen, especially on cognition, observed in women with SSD (Brand et al., in press;Riecher-Rössler, 2017), do not extend to men.
Collectively, these findings add to the growing body of evidence suggesting that prolactin may influence testosterone levels in men with SSD, particularly in individuals using prolactin-raising antipsychotic medication.The downstream consequences of lower testosterone levels may adversely impact symptoms and various domains of cognition in men with SSD.These findings have potential implications for both clinical management and cognitive outcomes in men with SSD and warrant the possible consideration of maintaining optimal testosterone and estrogen levels (i.e., higher testosterone and lower estrogen), particularly for men with cognitive symptoms.However, studies with hormonal augmentation therapy should monitor circulating hormone levels to prevent adverse effects.Furthermore, as estrogen is primarily synthesized from testosterone through aromatase (CYP19A1), reducing the action of aromatase might achieve more favourable serum levels (i.e., lower estrogen and higher testosterone), possibly leading to reduced symptoms and cognitive deficits.Although aromatase inhibitors have been shown to reduce estrogen levels and increase testosterone levels in healthy men (de Ronde and de Jong, 2011), the safety and efficacy of such an approach would need thorough investigation before clinical implementation for the treatment of men with SSD.Interestingly, a preclinical study showed that clozapine and haloperidol, but not olanzapine, can also increase the action of aromatase by increasing the expression of the gene encoding for aromatase (Bogus et al., 2019), suggesting that these antipsychotics might induce higher levels of estrogen, while potentially reducing testosterone levels.

Strengths and limitations
This study is one of the few cross-sectional studies investigating the association between prolactin, testosterone, and estrogen and cognition and symptoms in men with SSD while also considering antipsychotic medication.In addition, this study had a considerable sample size including patients from both Europe and Australia with broad inclusion criteria, adding to both the power and generalizability of the results.The lack of site as a significant predictor in any of the models, reinforces the validity of combining the samples.Nevertheless, the current study also had some limitations.Most importantly, the cross-sectional nature of the study makes it difficult to delineate causal relationships between hormones and cognition and symptoms.Nevertheless, backward linear regression models were built under the assumption that the underlying biological mechanisms would influence behavioural outcomes rather than vice versa.In addition, hormone levels were determined using immunoassays which brings disadvantages including standardization issues and differences in CVs between the two different labs.This divergence could potentially account for the observed differences in estrogen levels between sites.Furthermore, our control sample of men without SSD only consisted of men from Australia.Finally, the current study sample only included chronic SSD male patients with mild to moderate symptom severity and healthy men.While we identified significant associations, warranting further research in populations with more severe symptoms, our results may not be generalizable to the whole male SSD population (e.g., first-episode patients) or women with SSD.To achieve a more comprehensive understanding of the neuroendocrine systems affecting cognition and symptoms, future studies should also explore the influence of other factors such as aromatase activity.

Conclusion
This study provides valuable insights into the complex interplay between hormones, antipsychotic medication, cognition and symptoms in men with SSD.Although increased prolactin levels were associated with lower testosterone levels in men using prolactin-raising antipsychotics, no differences in circulating gonadal hormones were observed across the prolactin-raising, prolactin-sparing and healthy control group.Interestingly, we found that prolactin-sparing antipsychotics may reduce prolactin levels below normal levels.Higher peripheral testosterone levels were associated with better cognitive performance and reduced symptom scores in men with SSD, while peripheral estrogen may act antithetically.These associations support the importance of monitoring testosterone and estrogen levels in men with SSD, especially concerning cognition and negative symptoms, and underline the need for sex-specific research in SSD.Further research is needed to elucidate the underlying neuroendocrine mechanisms and their interplay with antipsychotic medication, as this may help contribute to the development of additional treatment targets for men with SSD.
-Spectrum Disorder; PANSS, Positive and Negative Syndrome Scale; SD, standard deviation.Significant group differences are indicated in bold.† one-way ANOVA, ‡ Kruskal-Wallis H test, § Wilcoxon signed-rank tests, ¶ Chi-squared test.a Significant difference between control and both prolactin groups (adjusted p < .05);b Significant difference between prolactin-sparing and prolactin-raising group (adjusted p < .05);c Significantly higher proportion as compared to prolactin-sparing group and the control group (adjusted p < .05).I.M.H. Hamers et al.

Fig. 1 .
Fig. 1.Group comparisons of hormone levels.A) Prolactin, B) testosterone, and C) estrogen levels with median and IQR of the control group (n = 44), the prolactinsparing group (n = 53) and the prolactin-raising group (n = 75), depicted separately for the two study samples from Europe (EU) and Australia (AUS).Red dotted line indicates hyperprolactinemia threshold ** adjusted p < .005,*** adjusted p < .001.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2 .
Fig. 2. Relationship between testosterone and estrogen levels.Spearman correlations between testosterone and estrogen in A) men with SSD and B) controls.

Fig. 3 .
Fig. 3. Relationship between prolactin and testosterone levels in the prolactinraising group.Pearson correlations between testosterone and prolactin in patients taking prolactin-raising antipsychotic medication.

Table 1
Demographics, clinical and cognition scores of the prolactin-raising group and prolactin-sparing group with corresponding significance levels from comparisons.

Table 2
Summary of final multiple linear regression models examining the effects of testosterone, estrogen, prolactin and covariates (predictor variables) on cognitive function in men with SSD.