Risperidone is an atypical antipsychotic used to treat autism spectrum disorder (ASD). Its effect is mediated via dopamine D2 receptor and 5-hydroxytryptamine type 2A receptor antagonism (Corena-McLeod, 2015). The US Food and Drug Administration has approved risperidone for the treatment of irritability in children and adolescents aged 5–16 years with ASD (Goel et al., 2018). Risperidone has a more favorable safety and efficacy profile than typical antipsychotic drugs (Peuskens et al., 2014). Two clinical trials (McCracken et al., 2002; Shea et al., 2004) established the efficacy and tolerability of risperidone in patients with ASD and showed that risperidone significantly attenuated disruptive behaviors compared with the placebo over 8 weeks, as measured by a reduction in the irritability subscale of the aberrant behavior checklist (ABC) scores. The improvement in the risperidone treatment group was higher (56.9–64.0%) compared with that in the placebo group (14.1–31.0%) (McCracken et al., 2002; Shea et al., 2004).

Several studies (Baumann et al., 2004; Riedel et al., 2005) have explored the utility of plasma drug monitoring as a biomarker of treatment response. Therapeutic drug monitoring can improve the efficacy and safety of risperidone. Plasma levels of 20–60 ng/L of the active moiety lead to better clinical outcomes in adults with schizophrenia (Baumann et al., 2004; Nazirizadeh et al., 2010). Plasma levels of the active moiety in non-responders were significantly higher than in responders in a 6 weeks clinical trial of risperidone treatment in schizophrenia patients (Riedel et al., 2005). However, no therapeutic drug monitoring data are available for pediatric patients with ASD treated with risperidone.

There is considerable variation in the response to risperidone that may be explained, at least partly, by genetic variation in drug targets. A polymorphism in the dopamine D2 receptor (DRD2)/ankyrin repeat and kinase domain containing 1 (ANKK1) gene (often referred to as the Taq1A polymorphism; rs1800497) was associated with a clinical response in risperidone-treated ASD patients (Nuntamool et al., 2017). The DRD2/ANKK1 Taq1A A2 or C allele (Sukasem et al., 2016) have been associated with high dopamine receptor densities (Jonsson et al., 1999), which may contribute to blocking dopaminergic activity. DRD2 -241A>G (rs1799978), located in the 5′-promoter region of DRD2, may contribute to increased expression levels. (Nyman et al., 2009). DRD2 -141C insertion/insertion has been associated with high prolactin levels in antipsychotic-treated male schizophrenia (Zhang et al., 2011). Carriers of the dopamine D3 receptor Gly/Gly allele (rs6280) showed significantly better response rates compared with the Ser/Ser genotype in children with ASD (Firouzabadi et al., 2017). Genetic variation in the 5-hydroxytryptamine type 2A receptor (HTR2A), 5-hydroxytryptamine type 2C receptor (HTR2C), and ATP binding cassette subfamily B member 1 genes also contribute to clinical outcomes and are possible markers for predicting a positive response to risperidone therapy in ASD (Correia et al., 2010). Moreover, the dopamine transporter [DAT; also known as solute carrier family 6 member 3 (SLC6A3)] and serotonin transporter [5-hydroxytryptamine transporter-linked promoter region (5-HTTLPR) also known as solute carrier family 6 member 4] genes, although not directly targeted by antipsychotic medications, may influence neurotransmitter availability, and thus contribute to the variability in treatment response (Kirchheiner et al., 2007; Porcelli et al., 2012; Outhred et al., 2016).

The anterior pituitary hormone prolactin has essential physiological functions in the brain. Prolactin acts as a neuropeptide, regulating neuroendocrine and emotional stress responses (Torner, 2016). Serum prolactin levels may also reflect the antipsychotic treatment response. Several clinical studies (Zhang et al., 2002; Ates et al., 2015; Stern et al., 2018) revealed that prolactin may mediate effects on the neuropsychiatric response to risperidone. Zhang et al. (2002) observed a significant positive relationship between the reduction rate of positive subscale scores of the Positive and Negative Syndrome Scale (PANSS) and the change in prolactin levels before and after treatment in chronic schizophrenia. Ates et al. (2015) found that in patients with hyperprolactinemia, the PANSS negative symptom scores were significantly higher than in patients without hyperprolactinemia (p = 0.041). Furthermore, several moderators and mediators affecting risperidone response have been described (Stern et al., 2018), and lower baseline levels of prolactin predict responder status in autistic children. Additionally, genetic polymorphisms in the prolactin (PRL) (Lee et al., 2007; Ivanova et al., 2017) and prolactin receptor (PRLR) (Lee et al., 2007) genes may also contribute to increased prolactin levels. The strongest association was between a single SNP of PRL (A>T, rs2244502) and prolactin levels, which showed higher prolactin levels in T carriers than in A carriers. Patients with hyperprolactinemia carried the G/G genotype of −1149 G>T (rs1341239) in the PRL gene more frequently than patients without hyperprolactinemia (p = 0.009) (Ivanova et al., 2017). Therefore, prolactin is a promising candidate biomarker for risperidone response.

The impact of pharmacogenetics on increased prolactin levels has not been well investigated. Stern et al. (2018) reported the effect of hyperprolactinemia in pediatric ASD patients with disruptive behaviors undergoing risperidone therapy. Therefore, in this study we investigated the association between genetic variations in pharmacodynamic genes and risperidone-induced increases in prolactin levels in children and adolescents with ASD.

Several studies have reported that genetic variation (Young et al., 2004; Calarge et al., 2009; Yasui-Furukori et al., 2011) or plasma drug levels of risperidone and its metabolite (Knegtering et al., 2005; Melkersson, 2006; Ngamsamut et al., 2016) are related to hyperprolactinemia. However, there are no studies on the relationship between prolactin elevation and genetic variation, or whether controlling plasma drug levels of RIS C/D, 9-OH-RIS C/D, and active moiety C/D can prevent prolactin elevation. The findings of the present study suggest that the patients with the H1/H3 haplotype of DRD2 were at higher risk than patients with other diplotypes of presenting with significantly higher prolactin levels with increasing plasma RIS C/D, 9-OH-RIS C/D, and active moiety C/D. Furthermore, low prolactin levels were associated with response to risperidone treatment in Thai ASD children and adolescents.

Hyperprolactinemia is commonly observed in patients treated with antipsychotics. In our study, 44.35% (55 of 124) of patients had high prolactin levels, which is a concern for adverse drug reactions in children and adolescents treated with risperidone. The reported prevalence of hyperprolactinemia in patients on risperidone ranges between 44 and 61% (Hongkaew et al., 2015; An et al., 2016; Bonete Llacer et al., 2019), and abnormal prolactin levels occur in about 27% of patients taking risperidone (Lally et al., 2017). Hypothalamic dopamine inhibits prolactin secretion by acting on dopamine D2 receptors, which inhibits lactotroph cells (Ben-Jonathan and Hnasko, 2001). Hypothalamic dopamine inhibits pituitary prolactin secretion and proliferation of prolactin-producing lactotroph cells by activating lactotroph dopamine D2 receptors (Ben-Jonathan and Hnasko, 2001; Fitzgerald and Dinan, 2008). There are several hypotheses for why prolactin elevation occurs during treatment with certain antipsychotics. Risperidone may cause high prolactin levels because of its slow dissociation from dopamine D2 receptors, which increases central dopamine D2 receptor occupancy (Peuskens et al., 2014), resulting in a prolonged blockade and higher rates of prolactin release. Risperidone is also associated with weaker blood-brain barrier penetration (Peuskens et al., 2014) compared with olanzapine and quetiapine (Arakawa et al., 2010), resulting in greater dopamine D2 receptor occupancy in the pituitary area, which also increases prolactin levels.

We also investigated whether DRD2 haplotypes are more informative than each SNP on its own, and the H1/H3 diplotype had the highest serum prolactin levels. These findings are consistent with increased prolactin levels in patients carrying the DRD2/ANKK1 Taq1A A2 or C allele (Sukasem et al., 2016). These alleles are associated with high dopamine receptor densities (Jonsson et al., 1999), which may contribute to blocking dopaminergic activity. DRD2 -241A>G (rs1799978), which is located in the 5′-promoter region of DRD2, may contribute to increased expression levels. Data obtained with lymphoblastoid cell lines suggested that -241G carriers have higher DRD2 expression levels than non-carriers (Nyman et al., 2009), which is consistent with our conclusion that DRD2 -241A>G is involved in hyperprolactinemia. Our results are also consistent with previous findings. Calarge et al. (2009) found that the -241G allele had the largest effect on serum prolactin levels when combined with the Taq1A A1 allele, whereas Arinami et al. (1997) reported an association between the DRD2 -141C indel and high DRD2 expression levels in vitro. Stronger binding of risperidone to dopamine D2 receptors in the pituitary area is more likely to inhibit dopaminergic activity in prolactin secretion (Peuskens et al., 2014). Therefore, the findings of these studies support our observation that increased DRD2 expression levels, mediated by the DRD2 -141C indel, contribute to the increase in prolactin levels. Our finding is also consistent with a study describing antipsychotic-induced sexual dysfunction in male schizophrenia, which was attributed to high prolactin levels in the presence of the DRD2 -141C indel (Zhang et al., 2011). Therefore, the synergistic effects of the DRD2/ANKK1 Taq1A A2>A1, DRD2 -141C indel, and -241A>G SNPs may be better captured by the association of haplotypes, rather than SNPs, with serum prolactin levels. Based on our results, we suggest that clinicians should perform DNA analysis of the three SNPs of the DRD2 haplotype as pre-emptive genetic testing, and prolactin levels should be monitored before and after 3 months to prevent hyperprolactinemia, especially in patients with the H1/H3 diplotype.

We developed an algorithm to categorize DRD2 haplotypes simply according to their DRD2 expression (Arinami et al., 1997; Jonsson et al., 1999; Nyman et al., 2009). High-expression alleles (A2, Cin, G) were assigned a score of 1 and low-expression alleles (A1, Cdel, A) were assigned a score of 0.5. The score of each haplotype was the sum of the values of each allele. A high risk score was expected to indicate a high prolactin level. In this study, a higher DRD2 expression score was significantly related to a higher prolactin level. The increase in prolactin levels with risperidone can be explained by risperidone inhibiting dopamine binding to dopamine D2 receptors (Rosenbloom, 2010). The higher the number of dopamine receptors is, the greater the inhibition by risperidone, and the higher the increase in prolactin. Thus, high prolactin levels occurred at the highest risk score of 6. However, this study also observed the high prolactin level at the lowest risk score of 3.5. This could be explained by the upregulation of dopamine receptors in patients with low DRD2 expression and who received long-term risperidone treatment. Lidow and Goldman-Rakic (1997) found that 6 months treatment with antipsychotics upregulated prefrontal and temporal cortical dopamine D2 receptor mRNA expression in primates. However, there was no significant difference in expression due to the small sample size. The genetic risk score system is an easy-to-use tool for translating genotype data into DRD2 expression prediction in a clinical setting. However, further study is needed to validate this tool.

Because variations in the HTR2A gene (encoding the 5-hydroxytryptamine type 2A receptor) and HTR2C gene (encoding the 5-hydroxytryptamine type 2C receptor) influence the binding affinities of antipsychotic medication, in previous studies, two genetic polymorphisms related to hyperprolactinemia in schizophrenic patients treated with classical and/or atypical antipsychotic treatment have been analyzed. Several studies found contradictory associations between hyperprolactinemia and HTR2A (Correia et al., 2010; Ivanova et al., 2017) and HTR2C polymorphisms (Alladi et al., 2017; Koller et al., 2020). Our study was consistent with a study of risperidone in 289 Indian schizophrenia patients (Alladi et al., 2017). They reported no association between HTR2C (−759 C>T) genetic variants and prolactin levels during risperidone treatment. Genetic polymorphisms in the promoter region of the HTR2C gene (−759 C>T) have been investigated in antipsychotic drug-induced weight gain (Vanwong et al., 2021). Moreover, transporter and prolactin-related genes involving the neurotransmitter mechanism of prolactin secretion were studied. Osmanova et al. (2019) found an association between two variants in the DAT (SLC6A3) gene and hyperprolactinemia in the subgroup of patients taking risperidone/paliperidone in 446 Caucasian schizophrenia patients. Smith et al. (2004) reported that the short allele of the 5-HTTLPR indel polymorphism was associated with less of an increase in prolactin and cortisol than the long/long genotype in control participants. Lee et al. (2007) found a nominally significant association between PRL and PRLR tagSNPs and plasma prolactin levels in 95 advanced breast cancer cases. However, the results were inconsistent because of the different underlying disease groups and treatments. Thus, until our results are validated in a prospective study, they should not be applied to patients with other diseases or undergoing other treatments.

There were no significant differences between any genetic variants selected and responses to risperidone in this study. Thus, although the DRD2 haplotypes relate to prolactin increase, they are likely to be predictive for poor markers for treatment response. Our results agree with Sakumoto et al. (2007), who found that non-responders without the deletion allele of the DRD2 -141C indel showed a higher score for psychiatric, extrapyramidal, and total side-effects than those with the deletion allele after 3 weeks of treatment with dopamine antagonists. The small number of 19 patients in our study cohort may explain why there were no significant differences among the associations between genetic variants and treatment response.

Several studies have examined optimizing risperidone dosing in children and adolescents based on their clinical response (Research Units on Pediatric Psychopharmacology Autism, 2005; Haas et al., 2009; Pozzi et al., 2016). We found no significant association between plasma drug levels and risperidone response due to the small sample size, although non-responders had numerically higher drug plasma levels than responders, which is consistent with other published work (Wang et al., 2007; Lostia et al., 2009). However, our results contradict other studies that show a relationship between therapeutic drug levels and clinical efficacy (Baumann et al., 2004; Riedel et al., 2005). This discrepancy may be a result of the different underlying disease in patients with schizophrenia. In addition, there may be differences in the timing of the clinical outcome assessment and the sociodemographic profile of the patient population among studies. We performed our assessment at least 3 months into risperidone treatment rather than at 1–6 weeks because prolactin levels in children receiving long-term risperidone tend to peak within the first 2 months, and then steadily decline to values within or close to normal after 3–5 months (Findling et al., 2003). Although we did not discover an association between plasma drug concentration and efficacy, the association between plasma drug and serum prolactin levels remained significant. Therefore, monitoring risperidone plasma levels may help to prevent adverse drug reactions, such as hyperprolactinemia.

Although other studies have found that prolactin levels are inconsistently associated with efficacy (Wang et al., 2007; Lostia et al., 2009), our study revealed that prolactin levels may induce changes in neurogenesis, potentially affecting aberrant behaviors in ASD, consistent with several other studies (Zhang et al., 2002; Lee and Kim, 2006; Ates et al., 2015). Prolactin is an anterior pituitary peptide hormone, under inhibitory control by dopamine released from the tuberoinfundibular dopaminergic neurons (Grattan, 2015). Because of dopamine regulation of serum prolactin levels through dopamine D2 receptors in the hypothalamic-pituitary pathway, the serum prolactin levels may be a marker of central dopamine function. Risperidone could induce hyperprolactinemia in excessive dopamine receptor blockade (Oda et al., 2015) due to its high dopamine D2 receptor occupancy (68–70%) (Tsuboi et al., 2013), resulting in hyperprolactinemia in early treatment. Long-term treatment with antipsychotics could lead to increased dopamine D2 receptor density according to the dopamine supersensitivity psychosis hypothesis (Oda et al., 2015), and thus could cause adequate blockade, leading to behavioral improvement. This hypothesis may explain why some ASD patients had increased prolactin levels early in risperidone treatment and subsequently presented with improved symptoms.

The association between high prolactin and worse response could be explained by the brain/plasma concentration ratio of risperidone. Arakawa et al. (2010) reported that risperidone showed a slightly higher brain/plasma concentration ratio. The brain/plasma concentration ratio was calculated from dopamine D2 receptor occupancy in the temporal cortex and pituitary, which represented the permeability of antipsychotics into the brain. The brain/plasma concentration ratio of risperidone was 1.61 (Arakawa et al., 2010), indicating that the risperidone concentration in the temporal cortex was 1.5-fold greater than that in the pituitary. The permeability of risperidone into the brain resulted in a higher concentration in the brain than in plasma. Dopamine D2 receptor occupancy of risperidone in the temporal cortex was higher than that in the pituitary, causing an improvement in behaviors and reducing the risk of prolactin elevation. Therefore, our results suggest that low prolactin levels are an indicator of response to risperidone, and thus serum prolactin levels could be used as an indirect biomarker for risperidone response.

The present study is the first to report a cut-off value for prolactin-related poor behavioral responses of children treated with risperidone of less than 1 mg/day. The prediction of risperidone response in ASD had an area under the ROC curve of 0.833. The cut-off point of the prolactin level associated with sensitivity and specificity of 0.5 or higher was 10.25–18.85 ng/ml, and the optimal cut-off value for prolactin of 10.9 ng/ml had the highest accuracy of 78.95%. We applied the prolactin measurement as a biomarker for unusual behaviors in patients who did not respond to risperidone treatment. It might be used as a screening test to exclude false negatives. When NPV was 100%, meaning that when the test was negative (low prolactin value), the patient was responding to treatment, when the PPV was 60%, meaning that when the test was positive (high prolactin value) of 100 patients, only 60 patients were not responding to treatment. This information would be beneficial to health practitioners in clinical practice. However, further studies are needed to determine the optimal daily dose of risperidone for prolactin monitoring.

There are several limitations to this study. The small sample size in the cohort study prevented us from detecting significant genotype-phenotype associations, and thus the results should be viewed with caution. No relationship was observed for risperidone dose, weight-adjusted dose, and treatment duration between responders and non-responders. However, the study used cohorts under clinical conditions and did not control for risperidone dose and treatment duration.

We created a system for representing the expression of DRD2 haplotypes using genetic risk scores. Our findings suggest that patients with the H1/H3 diplotype, which had a DRD2 genetic risk score of 5.5, showed the highest serum prolactin levels, correlated with increasing plasma RIS C/D, 9-OH-risperidone C/D, and active moiety C/D. Lower levels of prolactin were detected in patients who responded to risperidone. These findings further our understanding of hyperprolactinemia, which is a commonly observed adverse effect of risperidone treatment. Preemptive pharmacogenetic testing may allow clinicians to identify patients at risk of developing high prolactin levels at the outset of therapy and use this information to guide therapeutic decision-making.