Antipsychotic-Related Prolactin Changes: A Systematic Review and Dose–Response Meta-analysis

Abstract

Background

Prolactin increase is a common and potentially problematic adverse event of antipsychotics. We aimed to discover the relationship between antipsychotic doses and changes in prolactin levels.

Objective

To examine the relationship between antipsychotic doses and changes in prolactin levels in adults with acutely exacerbated schizophrenia.

Methods

We searched the Cochrane Schizophrenia Group register (last search 26 July 2024) and previous reviews for fixed-dose, randomized controlled trials (RCTs) that investigated monotherapy of 21 antipsychotics in adults with acutely exacerbated schizophrenia. The primary outcome was mean prolactin change from baseline to study endpoint adopting mean differences (MD) in ng/mL as the effect size measure. The dose–response curves were estimated by conducting random-effects dose–response meta-analyses using the restricted cubic spline method.

Results

Among 165 eligible studies, 68 studies with 238 dose arms (23,128 participants, 35% female) reported on prolactin and were meta-analyzed. Of these, 94% lasted ≤ 3 months, and 90% of the studies used oral formulations. Participants in one study experienced their first episode, while all other studies also included multiepisode participants. The dose–response curves indicated that with aripiprazole, higher doses were significantly associated with lower prolactin levels than lower doses. Brexpiprazole, cariprazine, lumateperone, and quetiapine carried negligible risks for prolactin increase across examined doses. During treatment with most other antipsychotics, i.e., asenapine, haloperidol, iloperidone, lurasidone, olanzapine, paliperidone, risperidone, and ziprasidone, prolactin levels rose with increasing doses and then continued to increase or plateaued. The shape of the dose–response curves was similar in males and females, with generally larger amplitudes of the curves in females.

Conclusions

The prolactin-increasing property varies among antipsychotics, is dose-related, and is greater in females. These findings in adults with acutely exacerbated schizophrenia can help clinicians titrate and adapt antipsychotic doses and consider patients’ sex in treatment decisions.

The protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO); registration no. CRD42020181467.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40263-025-01218-z.

Key Points

The prolactin-increasing effect varies among different antipsychotics and is dose-dependent in adults with acutely exacerbated schizophrenia.

This effect tends to be more pronounced in female patients.

Introduction

Prolactin increase is one of the most common and important side-effects of antipsychotic drugs [1, 2]. Hyperprolactinemia can result in numerous adverse effects, particularly adverse sexual and reproductive adverse effects [3], risk of fractures [4], and a potentially elevated risk of breast cancer in females [5]. For females, prolactin elevation can lead to acceleration of osteoporosis [6], amenorrhea, and galactorrhea [3]. For males, gynecomastia, decreased libido, and erectile dysfunction are also related to hyperprolactinemia [7]. Previous studies have reported elevated prolactin in 65% of female and 40–70% of male antipsychotic users [8, 9].

Prolactin production is inhibited in the hypothalamo-pituitary circuit by dopamine release and can be disinhibited by blocking dopamine 2 (D₂) receptors [10]. Antipsychotics vary in their propensity to produce prolactin changes. Significantly elevated prolactin levels versus placebo have been reported in descending order of magnitude with paliperidone, risperidone, amisulpride, haloperidol, sertindole, lurasidone, chlorpromazine, asenapine, iloperidone, and olanzapine in a network meta-analysis comparing 32 oral antipsychotics for acute treatment [11]. Similarly, risperidone and amisulpride increased prolactin most significantly in a meta-analysis of the long-term effectiveness of second-generation antipsychotics (SGAs) [12].

However, little is known to what degree and with which antipsychotics prolactin change is dose-related and whether this dose-related prolactin effect may differ between males and females. Furthermore, if such a relationship exists, its manifestation warrants examination. Specifically, it is unclear what would happen if the dose was further increased, whether a dose decrease can reduce prolactin levels, or whether switching to a prolactin-sparing antipsychotic is warranted for patients with abnormally increasing prolactin levels [13]. In addition, little is known whether there are differences between males and females. Thus, we conducted this dose–response meta-analysis to investigate the relationship between prolactin level changes and the doses of 20 SGAs and haloperidol, specifically investigating the following hypotheses: (1) prolactin levels increase with dose but eventually plateau at higher doses; (2) there are differences in the magnitude of dose-dependent effects among specific compounds; and (3) females are more sensitive, exhibiting greater prolactin increases at lower doses.

Methods

Search Strategy and Selection Criteria

We followed the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines (Supplementary Appendix 1: Checklist) and registered the protocol in the International Prospective Register of Systematic Reviews (PROSPERO; registration no. CRD42020181467) [14]. Any differences between the review and protocol are explained in Supplementary Appendix 2.

We searched the Cochrane Schizophrenia Group’s Study-Based Register (for a detailed introduction of it, please see Supplementary Appendix 3) until 26 July 2024 for randomized controlled trial (RCT) studies comparing SGAs (all formulations) with each other or placebo in adults with schizophrenia (diagnosed with schizophrenia or schizoaffective disorder by any criteria). We included only studies with fixed dosing regimens, because dose–response meta-analysis requires data that reflect the average treatment effect for a specific dose. Flexible-dose studies, by contrast, pool outcomes from patients receiving a range of doses, which prevents accurate estimation of dose-specific effects. Haloperidol is often regarded as a “gold standard” antipsychotic. Unlike most other first-generation antipsychotics, it is still widely used in clinical practice and frequently adopted as an active comparator in clinical trials; therefore, it was included in our analysis. All studies investigating haloperidol with fixed dosing were systematically retrieved and evaluated using the same eligibility criteria as for SGAs, including studies from before the introduction of SGAs, provided they met our inclusion criteria. Detailed search strategy and selection criteria are formulated in our previous publications [15–17] and Supplementary Appendices 3–4. We excluded patients in the maintenance phase of treatment or with stable psychopathology, in accordance with our pre-registered protocol [14] and previously published criteria [15–17]. Methodologically, maintenance-phase trials often involve substantially longer treatment durations (e.g., several months versus 6–8 weeks in acute-phase trials), and patients typically receive flexible or individualized doses. These factors, along with long-term prior antipsychotic exposure, may obscure accurate estimation of dose–response relationships. Furthermore, such trials frequently use enriched designs that selectively include patients who have previously tolerated the drug well, introducing additional selection bias.

Two reviewers (from among S.L., H.W., S.S., X.L., M.Q., or J.T.) independently screened the references, extracted data, and assessed risk of bias using the Cochrane Risk of Bias tool 1 [18].

Data Analysis

We evaluated three outcomes. The primary outcome was the mean prolactin change from baseline to endpoint adopting mean differences (MD) in ng/mL as the effect-size measure. All the prolactin data were transformed to the same unit (ng/mL) and if the change data was not available, we used the MD in endpoint values. Secondary outcomes were the prolactin changes for males and females separately (analyzed as MD), based only on studies that reported sex-specific results. Formal statistical testing for subgroup differences (e.g., sex interaction) was not conducted owing to limited statistical power in aggregate data meta-analyses. In addition, we assessed the number of participants with prolactin increase, either according to the original authors’ definition of hyperprolactinemia or when reported as an adverse event (analyzed with odds ratios).

Moreover, we planned to analyze the primary outcome separately in the following specific patient populations: (1) chronic with an acute exacerbation, (2) elderly, (3) first episode, and (4) predominant negative symptoms.

One-stage dose–response analysis was conducted in a frequentist framework applying restricted cubic splines developed by Crippa and Orsini [19, 20]. Knots at 25th, 50th, and 75th percentiles were used as in previous publications [15–17]. For drugs with available data, the Wald test was carried out to access the dose–response relationship, and p-values were reported. A two-sided alpha of 0.05 was set.

To aid interpretation, we briefly compare the one-stage model to the more familiar two-stage framework. In a two-stage approach, dose–response curves are first estimated separately within each study and then pooled across studies. However, many of the included studies reported only one active dose (e.g., drug versus placebo), making it difficult to fit study-level curves. The one-stage model overcomes this limitation by fitting a single dose–response curve using all available data points across studies. This allows for the inclusion of studies with limited dose contrasts and improves statistical power and consistency in estimation.

This approach is particularly useful in meta-analyses with heterogeneous dose reporting, where many studies provide only a single comparison (e.g., one fixed dose versus placebo), as it enables the inclusion of more studies in the dose–response analysis.

We pooled different formulations of one drug together by converting depot doses into corresponding oral equivalent doses (Supplementary Appendix 5: Supplementary Table 1). However, since the influence may vary depending on the formulation [21], sensitivity analyses were conducted to examine the effects of different forms of the same antipsychotics on prolactin levels.

Sensitivity analyses of primary outcomes were performed to test the robustness of the results by (1) excluding studies comparing one single dose of an antipsychotic with placebo, (2) excluding studies with treatment-resistant patients, and (3) analyzing different formulations separately.

Moreover, to facilitate comparison between specific antipsychotics, we transformed all doses of specific drugs to olanzapine equivalents (following instructions from Schneider-Thoma et al. [22]) and provided dose–response curves for all specific antipsychotics in one plot.

The variance partition coefficient (VPC), a multivariate extension of the I² value [19], was chosen for quantifying heterogeneity in this dose–response meta-analysis. For comparisons with more than ten trials, small-study effects/publication bias were explored with funnel plots and Egger’s tests [18]. R package dosresmeta was used for data analysis [19, 20], and R package meta version 4.12–0 [23] was used to investigate small-trial bias.

Results

We included 165 studies that met our inclusion criteria, and the 21 antipsychotics investigated were amisulpride, aripiprazole, asenapine, blonanserin, brexpiprazole, cariprazine, clozapine, haloperidol, iloperidone, lumateperone, lurasidone, olanzapine, olanzapine–samidorphan, paliperidone, perospirone, quetiapine, risperidone, sertindole, xanomeline, ziprasidone, and zotepine. Meta-analyzable data on prolactin were reported in 68 of these studies, yielding 238 drug arms (n = 23,128 participants). The median study duration was 6 weeks (range 4–24 weeks).

Study participants were patients with an acute exacerbation, almost exclusively of chronic/multiepisode schizophrenia; one study focused on participants with a first episode (evaluating haloperidol), and none was conducted solely in the elderly. Detailed descriptions of the included studies and their overall risk of bias are presented in Supplementary Appendix 6–7: Supplementary Table 2–4 and Supplementary Fig. 2.

Below, we present in alphabetical order the antipsychotic-specific results and characteristics, along with illustrations of the curves for the primary outcome (Fig. 1) and of the curves for males and females separately (Fig. 2). To more clearly illustrate the dose–response curves, the sex-specific plots used a different y-axis scale compared with the total population, but for the same drug, the male and female plots shared the same y-axis scale to allow for comparison between males and females. The all-in-one plot with olanzapine equivalencies is provided in Fig. 3.

Fig. 1.

Prolactin change of individual antipsychotics in different dosages in participants with acute exacerbations of chronic schizophrenia. The dose–response curve depicts the mean differences of prolactin (ng/mL) comparing a given dose of a drug with 0 mg/day or placebo. The area between 95% confidence intervals is shaded in gray. The knot locations are set at the 25th, 50th, and 75th percentiles to anchor the curve. Different formulations are pooled in aripiprazole, paliperidone, quetiapine, and risperidone studies. n. studies number of studies, n. arms number of arms, N number of participants

Fig. 2.

Prolactin change of individual antipsychotics at different dosages in participants with acute exacerbations of chronic schizophrenia in different sexes. The dose–response curve depicts the mean differences of prolactin (ng/mL) comparing a given dose of a drug with 0 mg/day or placebo. Orange lines represent females, and blue lines represent males. The area between the 95% confidence intervals is shaded in the corresponding color. The knot locations are set at the 25th, 50th, and 75th percentiles to anchor the curve. Different formulations are pooled in aripiprazole, paliperidone, quetiapine, and risperidone studies

Fig. 3.

Dose–response curves of individual antipsychotics using olanzapine dose equivalence

The dose–response curves of the binary outcome, that is, the number of participants with study-defined hyperprolactinemia or prolactin increase as an adverse effect are reported in Supplementary Appendix 8: Supplementary Table 7.

Amisulpride

No eligible study reported changes in prolactin.

Aripiprazole

Overall, nine placebo-controlled studies were included, which investigated oral aripiprazole (seven studies), aripiprazole lauroxil (one study) [24], and aripiprazole monohydrate (one study) [25]. Doses (oral equivalencies) within 2–30 mg/day were examined. The dose–response curve was hyperbolic and declined to a small extent, initially with increasing doses leading to lower prolactin levels, reaching a near-plateau at approximately 13.5 mg/day. The minimum MD of prolactin compared with placebo/0 mg dose was − 6.02 ng/mL (number of studies n = 9, number of participants N = 2474, Wald test p-value < 0.01).

When investigated separately by sex, the dose–response curve of males, which was similar to the total population, with a minimum MD − 6.98 ng/mL at 12.2 mg/day (n = 2, N = 513, p-value < 0.01). Conversely, the curve for females was a downward parabola, with the lowest point at − 14.08 ng/mL at a dose of 9.4 mg/day (n = 2, N = 254, p-value < 0.01) and an increase with higher doses, so that there was no significant net change in prolactin versus baseline with doses around 20 mg. However, because of high uncertainty and because there were no data for females using doses higher than 20 mg, the question of whether aripiprazole would lead to elevated prolactin levels at doses > 20 mg cannot be answered.

Asenapine

Three placebo-controlled studies were included. All were oral formulation studies, and doses ranged from 10–20 mg/day. The dose–response curve increased slowly until approximately 20 mg/day, and the maximum MD of prolactin was 10.44 ng/mL (n = 3, N = 983, p-value < 0.01). No data in males versus females were available.

Blonanserin

No eligible study reported changes in prolactin.

Brexpiprazole

Four oral formulation placebo-controlled studies were included with doses from 0.25 to 5 mg/day. The dose–response curve was flat at the peak with 1.7 mg/day, and the maximum MD of prolactin was 2.11 ng/mL (n = 4, N = 2079, p-value = 0.07).

For the female group, the dose–response curve was slightly bell-shaped, with a maximum MD of 6.43 ng/mL at 1.9 mg/day (n = 4, N = 848, p-value < 0.01). The male curve had a different shape, showing negligible risk of prolactin change as well, with a maximum MD of 0.29 ng/mL at 5 mg/day (n = 4, N = 1231, p-value = 0.73).

Cariprazine

Three studies with placebo-controlled oral doses between 1.5 and 12 mg/day were included. The dose–response curve was flat and bottomed out at approximately 8.7 mg/day, and the minimum MD of prolactin was − 3.40 ng/mL (n = 3, N = 1395, p-value = 0.17).

For sex groups, females had a similar dose–response curve as the total population, with a minimum MD of − 11.33 ng/mL at 4.5 mg/day (n = 1, N = 185, p-value = 0.24), whereas the curve for males had little fluctuation (maximum prolactin was 1.81 ng/mL at 4.5 mg/day, n = 1, N = 404, p-value = 0.37).

Clozapine

No eligible study reported changes in prolactin.

Haloperidol

A total of 12 studies investigated haloperidol doses within 2–20 mg/day. Two of them were dose-finding studies, meaning they included at least two arms for the antipsychotics being investigated. The dose–response curve rose gradually to the peak with 8.5 mg/day at which the MD of prolactin was 22.20 ng/mL and then fell slightly (n = 12, N = 1473, p-value < 0.01).

The dose–response curves of males and females were comparable, with a more bell-shaped appearance. The order of the amplitude across the groups was: females (maximum MD 58.10 ng/mL at 8 mg/day, n = 4, N = 68, p-value < 0.01) > (higher than) all patients > males (maximum MD of 18.30 ng/mL at 7.3 mg/day, n = 4, N = 261, p-value < 0.01).

Because of insufficient data, we could not build a dose–response curve for the only study with first-episode patients in the dataset [26], which investigated haloperidol.

Iloperidone

Three placebo-controlled studies with oral formulation were included. Dose range was 4–24 mg/day. The dose–response curve ascended gradually, and the maximum MD of prolactin at approximately 24 mg/day was 10.18 ng/mL (n = 3, N = 1035, p-value < 0.01). No data in males versus females were available.

Lumateperone

Three placebo-controlled studies were included. The oral doses were between 14–84 mg/day. The dose–response curve changed little with the minimum MD of prolactin − 3.60 ng/mL at approximately 84 mg/day (n = 3, N = 1009, p-value < 0.01). No data in males versus females were available.

Lurasidone

Ten placebo-controlled studies with oral doses between 20 and 240 mg/day were investigated. The dose–response curve increased moderately, and the maximum MD of prolactin was 17.91 ng/mL at 240 mg/day (n = 10, N = 2940, p-value < 0.01).

The order of the amplitude of the similarly shaped dose–response curves was: females (maximum MD was 5.05 ng/mL at 240 mg/day, n = 5, N = 466, p-value = 0.74) < (lower than) males (maximum MD 11.72 ng/mL at 240 mg/day, n = 5, N = 741, p-value = 0.03) < all patients, each with a similar trajectory.

Olanzapine

Overall, 11 placebo-controlled studies and 2 dose-finding studies were examined with an oral dose range of 1–40 mg/day. The dose–response curve increased constantly; the maximum MD of prolactin at 40 mg/day was 14.97 ng/mL (n = 13, N = 2963, p-value < 0.01).

The dose–response curve revealed the following order of the amplitude across the groups: females (maximum MD was 38.82 ng/mL at 40 mg/day, n = 7, N = 601, p-value < 0.01) > males (maximum MD 6.08 ng/mL at 40 mg/day, n = 7, N = 1123, p-value = 0.27) > all, each with similar trajectory.

Olanzapine–Samidorphan

No eligible study reported changes in prolactin.

Paliperidone

Eight placebo-controlled studies were included, with two examining long-acting injectable (LAI) paliperidone and six focusing on oral paliperidone. Oral equivalences ranged from 1.5 to 12 mg/day. The dose–response curve was hyperbolic, initially rising and then plateauing at approximately 12 mg/day, and the maximum MD of prolactin was 50.97 ng/mL (n = 8, N = 2574, p-value < 0.01).

Dose–response curve of males versus females remained the same as the total population. The order of the amplitude from large to small was: females (with a maximum MD 92.92 ng/mL at 15 mg/day, n = 9, N = 1095, p-value < 0.01) > all > males (with a maximum MD 30.96 ng/mL at 15 mg/day, n = 9, N = 1788, p-value < 0.01).

Perospirone

No eligible study reported changes in prolactin.

Quetiapine

Five placebo-controlled studies and one dose-finding oral study with doses between 75 and 1200 mg/day were examined. The dose–response curve of quetiapine appeared without much variation. The minimum MD was − 0.92 ng/mL at 445.1 mg/day, while the maximum was 1.36 ng/mL at 1144 mg/day (n = 6, N = 2048, p-value = 0.83). No data in males versus females were available.

Risperidone

Altogether, 15 studies were investigated, from which 12 were placebo-controlled trials and 3 were dose-finding studies. According to formulation, 3 were LAIs (one risperidone microspheres [27], one risperidone ISM, a new intramuscular injection [28], and one risperidone RBP-7000, a sustained-release subcutaneous injection [29]), and 12 were oral studies. Doses were 1–12 mg/day. The dose–response curve increased sharply until approximately 4 mg/day, then leveled off, with the maximum MD of prolactin reaching 41.03 ng/mL at 12 mg/day (n = 15, N = 3441, p-value < 0.01).

In the sex groups, the dose–response curves were similar to the total population. The order of the amplitudes was: females (with a maximum MD 50.70 ng/mL at 5.2 mg/day, n = 4, N = 329, p-value < 0.01) > all > males (with a maximum MD 19.00 ng/mL at 12 mg/day, n = 4, N = 714, p-value < 0.01).

Sertindole

One placebo-controlled oral study with doses within 20–24 mg/day was included. The curve for the dose–response relationship was flat at the beginning, then had a gentle upward slope, which ended with a maximum MD 20.85 ng/mL at 24 mg/day (n = 1, N = 49, p-value = 0.02). No data in males versus females were available.

Xanomeline

No eligible study reported changes in prolactin.

Ziprasidone

There were two oral formulation studies with doses between 40 and 200 mg/day, one was placebo-controlled and another one was a dose-finding study. The dose–response curve rose gradually to approximately 200 mg/day, and the maximum MD of prolactin was 8.07 ng/mL (n = 2, N = 495, p-value = 0.02). No data in males versus females were available.

Zotepine

No eligible study reported changes in prolactin.

Sensitivity Analyses

Most results were robust in the sensitivity analyses (Supplementary Appendix 9: Supplementary Figs. 3–5). Nevertheless, the height of the plateau of the haloperidol dose–response curve doubled when only studies aiming to investigate differences between different doses, i.e. studies having at least two arms with different doses of the same drug, were analyzed (22.44 versus 56.56 ng/mL). Regarding paliperidone, the amplitude of the curve for the LAI studies was nearly half of the one for oral drugs (30.32 versus 62.12 ng/mL).

Heterogeneity Assessments

For the primary outcome, no substantial heterogeneity, i.e., median VPC < 50%, was found across any of the antipsychotics investigated, except for paliperidone, risperidone, and sertindole (Supplementary Appendix 10).

Small-Study Effect/Publication Bias

Investigating small-study effects for haloperidol, olanzapine, and risperidone that had at least ten studies, no small-study effect/publications bias was found (details in Supplementary Appendix 11: Supplementary Figs. 6–7).

Discussion

To the best of our knowledge, this is the first dose–response meta-analysis exploring the relationships between antipsychotic-induced prolactin change and doses for 20 SGAs and haloperidol. The propensity to impact prolactin levels varied from negative risk (aripiprazole), negligible risk (brexpiprazole, cariprazine, lumateperone, and quetiapine), and moderate risk (asenapine, haloperidol, iloperidone, lurasidone, olanzapine, and ziprasidone) to high risk (paliperidone and risperidone). Notably, paliperidone and risperidone exhibited the steepest dose–response relationships and the highest predicted prolactin levels, making their prolactin-elevating potential especially prominent among all SGAs.

For most of the antipsychotics, prolactin values increased with higher dose. Some of the curves grew without an inflection point (asenapine, iloperidone, lurasidone, olanzapine, and ziprasidone), i.e., they were monotonic. In others, after reaching a peak, prolactin levels might continue to rise slowly (paliperidone and risperidone) or demonstrate a slight dip (haloperidol).

The results were generally consistent with the differences between antipsychotics found in the previous network meta-analysis for 32 oral antipsychotics that did not consider dose effects and that also included more studies without fixed doses [11].

The exact pharmacological mechanisms of prolactin alterations are not fully understood yet. Though diverse in their profiles of receptor-binding, all antipsychotics target D₂ receptors [30]. When D₂ receptor antagonists bind to the D₂ receptor on the prolactin-producing cells, they block the inhibitory effect of dopamine on the synthesis and release of prolactin, disinhibiting the synthesis and release of prolactin, resulting in prolactin increase [31]. Conversely, when D₂ receptor agonists bind to the D₂ receptor, they inhibit the synthesis and release of prolactin, resulting in decreased prolactin levels. Across certain doses, the increasing/decreasing dose–response curves of some antipsychotics in our analysis reached a plateau, as the dopamine receptors might have been entirely bound. This phenomenon is observed with dopamine high-affinity antagonists such as haloperidol, paliperidone, and risperidone, as well as with the high-affinity partial agonist aripiprazole, albeit with dopamine-reducing effects. In contrast, some antipsychotics in our study, including asenapine, iloperidone, lurasidone, olanzapine, and ziprasidone, displayed curves without a plateau but with constantly rising slopes. This pattern may be because the affinity of these drugs is not as high as that of the other antipsychotics, and at the doses used, not all receptors are fully bound to. Consequently, these antipsychotics remain in the more linear part of the binding curve. In addition, along with the dopaminergic pathway, antipsychotic affinities for serotonin and adrenergic and histamine receptors have been suggested to influence prolactin levels, at least to some degree [32]. Another factor that might influence the shape of the dose–response curve is the occurrence of premature study discontinuation by participants (dropouts), as higher doses likely lead to more dropouts, potentially occurring before the full effect on prolactin is exerted.

In this study, we also investigated the dose–response relationship in males versus females. The majority of the sex-specific curves had a similar shape as the whole population. Meanwhile, the amplitude of the curves was higher in the female group than in the male group in studies of haloperidol, olanzapine, paliperidone, and risperidone. Owing to estrogen leading to higher prolactin levels, females have higher baseline prolactin levels than males and are known to be more vulnerable to the prolactin variation effects of prolactin-rising antipsychotics [33, 34]. In addition, it is also possible that this increased vulnerability may result from higher antipsychotic plasma exposure levels in females versus males at the same antipsychotic dose [33, 35].

Aripiprazole, while generally decreasing prolactin levels across the entire population, showed a stronger decrease in females compared with males. Interestingly, the curve for females had a parabola-like shape, meaning that at doses around 10 mg/day, there was the greatest decrease in prolactin, while higher doses led to less decrease. For interpretation it needs to be noted that typically participants in the included trials used antipsychotics before the start and then switched to the study medication. Thus, it is likely that many patients had prolactin levels higher than normal at baseline and that a switch to aripiprazole leads to a reduction back to normal values. Taking this fact into account, it seems that low doses (less than 15 mg/day) carry the best prolactin profile in the female population. Notably, the absence of a clear prolactin-lowering effect at around 20 mg/day in females may be related to the combined influence of aripiprazole’s partial dopamine agonism and sex hormone-modulated prolactin regulation. Whether prolactin levels might even increase beyond 20 mg/day remains unclear. Therefore, careful dose adjustment is advisable when using higher doses in female patients.

With brexpiprazole, prolactin levels changed little with increasing doses in males. Similarly in females, although the dose–response curve appeared somewhat bell-shaped, it was still a flat curve (with a maximum increase of 8 ng/mL). Previous studies found that brexpiprazole was associated with a point estimate indicating some increase of prolactin, but the change was not statistically significant [36, 37]. Hence, brexpiprazole appears to have a negligible risk in prolactin level elevation, especially in males.

We could not expand the curve with regard to differences between males and females for cariprazine since there was only one study with data per sex and lack of high dose. For lurasidone, it appears that males might have greater prolactin increase with higher doses than females, but there was high uncertainty, particularly in the dose–response curve for females, preventing conclusions in this regard.

In the sensitivity analysis, a much higher prolactin plateau was observed with haloperidol (60 versus 20 ng/mL) when only dose-finding studies were investigated. This finding suggests differences due to study design, but it needs to be noted that in the 10 “non-dose-finding” studies excluded (out of 12) in this sensitivity analysis, haloperidol was used as an active comparator in a dose-finding study of another antipsychotic. The difference in the dose–response curves of paliperidone for oral and LAI formulations—specifically, that the amplitude of the curve in LAI studies was approximately half that of the oral ones—had been reported previously for prolactin levels and the finding of hyperprolactinemia [21], and is likely due to less peak-to-trough variations and more stable serum drug concentrations with LAI usage [38]. While this effect was not seen with the pharmacologically similar drug risperidone, fewer available studies may have reduced the power for these analyses. For aripiprazole, it showed a minor prolactin-lowering trend at lower LAI doses, likely reflecting both its partial D₂ agonism and steadier plasma levels with LAI use.

Limitations of this study cannot be ignored. First, we collated 165 studies, but roughly only 40% of them provided meta-analyzable data. We cannot be certain whether prolactin results were measured but not reported in the remaining studies. As a consequence, data were not available for all SGAs, for example, they were not available for amisulpride or clozapine. Second, all included data were derived from patients undergoing acute exacerbations of positive symptoms in the context of chronic or multiepisode schizophrenia. This was partly due to the characteristics of the available studies, in which prolactin levels were typically assessed during acute phases. In addition, we excluded patients with stable psychopathology or in maintenance treatment phases to reduce methodological heterogeneity associated with prior antipsychotic exposure and variable treatment durations. As a result, our findings may not be generalizable to populations with stable symptoms, predominant negative symptoms, or those receiving long-term maintenance treatment. Future studies specifically targeting these subgroups are warranted. Third, the findings based on RCT studies with highly selected patients cannot be straightly applied to real-world populations [39]. Fourth, most included studies were switch studies with short or no washout periods (46/68 studies had < 1 week washout). Thus, carry-over effects from prior antipsychotics may have confounded the observed prolactin levels, especially in short-duration trials. This limits the ability to attribute changes solely to the study drug, and future studies in antipsychotic-naïve populations are warranted. Fifth, some antipsychotics, such as haloperidol, lacked useable data for higher dose, and potentially higher risk of prolactin rising in higher doses should not be neglected. Sixth, results for the study-defined hyperprolactinemia were too sparse to yield dependable results. Hence, future studies should always also report categorical outcomes of prolactin elevations above a certain critical threshold. Seventh, our meta-analysis included studies with only one active drug arm (drug versus placebo). These non-dose-finding studies are designed to investigate proof of concept and thus may lead to heterogeneity, but a sensitivity analysis excluding non-dose-finding studies showed no materially different results. Eighth, we only focused on antipsychotic dose–response effects on prolactin levels and not on potentially related sexual/reproductive system adverse effects, which may or may not have a tight correlation with each other [40]. Ninth, while sex-specific differences in prolactin response were explored through stratified analyses, we did not perform formal statistical comparisons between males and females. Future individual participant data meta-analyses will allow more precise evaluation of such subgroup effects. Finally, we judged the curves by visual inspection; overinterpretation should be avoided.

Conclusions

Antipsychotic-induced prolactin changes are still not completely understood. Exploring its dose–response relationship is essential for clinical practice. As found here, the impact of antipsychotics on prolactin varies among antipsychotics, is dose-related, and is generally more pronounced in females than males. The strongest prolactin-elevating effects were seen with paliperidone and risperidone, underlining the relevance of individual drug profiles. Thus, monitoring of prolactin levels and prolactin-related side effects should be considered [31], especially for female patients and especially with prolactin-rising antipsychotics, and when using higher antipsychotic doses.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

Open Access funding enabled and organized by Projekt DEAL.

Declarations

Conflict of Interest

In the last 3 years, C.U. Correll has been a consultant and/or advisor to or has received honoraria from: AbbVie, Alkermes, Allergan, Angelini, Aristo, Boehringer-Ingelheim, Bristol-Meyers Squibb, Cardio Diagnostics, Cerevel, CNX Therapeutics, Compass Pathways, Darnitsa, Delpor, Denovo, Eli Lilly, Eumentis Therapeutics, Gedeon Richter, Hikma, Holmusk, IntraCellular Therapies, Jamjoom Pharma, Janssen/J&J, Karuna, LB Pharma, Lundbeck, MedInCell, MedLink, Merck, Mindpax, Mitsubishi Tanabe Pharma, Maplight, Mylan, Neumora Therapeutics, Neuraxpharm, Neurocrine, Neurelis, Newron, Noven, Novo Nordisk, Otsuka, PPD Biotech, Recordati, Relmada, Reviva, Rovi, Saladax, Sanofi, Seqirus, Servier, Sumitomo Pharma America, Sunovion, Sun Pharma, Supernus, Tabuk, Takeda, Teva, Terran, Tolmar, Vertex, Viatris and Xenon Pharmaceuticals; he provided expert testimony for Janssen, Lundbeck, and Otsuka. He served on a Data Safety Monitoring Board for Compass Pathways, IntraCellular Therapies, Relmada, Reviva, and Rovi; he has received grant support from Boehringer-Ingelheim, Janssen, and Takeda. He received royalties from UpToDate and is also a stock option holder of Cardio Diagnostics, Kuleon Biosciences, LB Pharma, Medlink, Mindpax, Quantic, and Terran. In the last 3 years, S.L. has received honoraria for advising/consulting and/or for lectures and/or for educational material from Angelini, Boehringer Ingelheim, Eisai, Ekademia, Gedeon Richter, Janssen, Karuna, Kynexis, Lundbeck, Medichem, Medscape, Mitsubishi, Otsuka, NovoNordisk, Recordati, Rovi, and Teva. H.W. is currently an employee of Elsevier in the role of senior editor for The Lancet Public Health. Her contribution happened before she was employed by Elsevier and she had no role in the editorial evaluation, peer-review process or decision to accept the article. All other authors have no conflicts of interest to declare.

Data Availability

The full dataset is available from the study authors upon reasonable request.

Ethics Approval

Not applicable.

Funding

Open Access funding enabled and organized by Projekt DEAL. This meta-analysis was conducted within the framework of a larger project funded by the Deutsche Forschungsgemeinschaft (German Research Foundation, grant no. 468853597). Xiao Lin receives support from The China Scholarship Council PhD program (file no. 202208310049). Publication fee is covered by institutional funding by Technical University of Munich. None of the funding institutions contributed to the study design, data analysis, manuscript writing, or the decision to submit the paper for publication.

Consent to Participate

Not applicable.

Consent for Publication

Not applicable.

Code Availability

Code is available from the corresponding author upon request.

Authors’ Contributions

S.L., S.S., J.S.T., and X.L. designed the review; S.L., H.W., S.S., X.L., J.T., and M.Q. were responsible for the literature search, article screening, and data extraction; X.L. conducted the statistical analysis under the guidance of J.S.T. and S.S; and X.L., J.S.T., and S.L. drafted the article. All authors critically revised the article for important intellectual content and approved the final version submitted for publication.