A high prevalence of polypharmacy is seen among psychiatric patients (Hefner et al., 2020). Polypharmacy, the use of five or more drugs, is often associated with drug-drug interactions and the risk of side effects. Preventing these side effects is important to ensure optimal psychopharmacotherapy and therapeutic adherence (Bousman et al., 2021).

The use of pharmacogenetics (PGx) contributes to individual patient treatment and can play an important role in preventing side effects (Sharp et al., 2019). PGx can distinct the different genetic variants of genes encoding for cytochrome P450 enzymes (CYP) such as CYP2C19 and CYP2D6 with a different metabolic capacity (KNMP, 2022a; KNMP, 2022b). A patient’s genotype can be translated into the following predicted phenotypes: normal metabolizer (NM), intermediate metabolizer (IM), poor metabolizer (PM) or ultrarapid metabolizer (UM) (Brouwer et al., 2022). Different consortia such as the Dutch Pharmacogenetics Working Group (DPWG) or the Clinical Pharmacogenetics Implementation Consortium (CPIC) have written guidelines regarding dose and pharmacotherapeutic recommendation for each genotype with an actionable drug-gene interaction (DGI) (Abdullah-Koolmees et al., 2021).

CYP2C19 and CYP2D6 are responsible for the metabolism of many psychiatric drugs (Abdullah-Koolmees et al., 2021; Brouwer et al., 2022). In addition, these are also highly polymorphic enzymes and therefore pharmacogenetic advise on these DGIs are widely available (Bousman and Dunlop, 2018; Hahn and Roll, 2021). Furthermore, PGx can help improve and optimize pharmacotherapy for individual patients using psychiatric drugs. Other CYP-enzymes such as CYP1A2, CYP2C9 and CYP3A4 can also play a role in the metabolism of psychiatric drugs. However, the impact of their genotypes on the pharmacokinetics of commonly used psychiatric drugs is less distinct as compared to the effect of CYP2C19 and CYP2D6 genotypes. Previous studies have reported that patients with a predicted phenotype of PM for either CYP2C19 or CYP2D6 have a higher risk of side effects (Chou et al., 2000; Kobylecki et al., 2009; Mrazek et al., 2011). However, in other studies no association between the genotype and the development of side effects has been observed (Peters et al., 2008; Hodgson et al., 2015).

Other non-genetic factors, such as co-medication, also influence the patient’s phenotype, i.e., a patient’s metabolic capacity (Klomp et al., 2020). This can have a significant impact, especially on patients with polypharmacy. This phenomenon, where the predicted metabolic capacity shifts because of the use of co-medication or other non-genetic factors, is called phenoconversion (Hahn and Roll, 2021). In this study, phenoconversion by co-medication will be taken into account.

Research shows that CYP-inhibition or CYP-induction by co-medication often has the greatest influence on NMs and IMs, causing a change in the drug exposure (AUC) (Bahar et al., 2017). For instance, it has been shown that patients using (es)citalopram are more prone to a dose reduction or switching to another antidepressant when there is a drug-drug-interaction combined with a DGI with CYP2C19 (MuhA et al., 2020). However, the relationship between side effects, pharmacogenetic profile and phenoconversion remains to be studied. The aim was to study side effects experienced by psychiatric patients and to identify risk factors including but, not limited to, pharmacogenetic profile of CYP2C19 and CYP2D6 and phenoconversion.

In total, 117 LL-patients were eligible for this study, of which 104 (88.9%) LL-patients filled in the UKU-questionnaire and could be included in the analysis (Figure 1). Of the 104 LL-patients who were included in the analysis, 15 (14,4%) LL-patients concomitantly used a CYP2C19-substrate and 49 (47,1%) a CYP2D6-substrate. Table 2 shows the baseline characteristics of the study population. Figure 2 compares the distribution of the different genotype-predicted phenotypes of CYP2C19 and CYP2D6. The proportion of UMs in CYP2C19 was significantly larger for the LL-patients (8.5%) compared to the CAD-patients (4.0%).

Approximately 10% of the LL-patients underwent phenoconversion for either CYP2C19 or CYP2D6 (Table 2). The comparison of the distributions of CYP2D6 before and after phenoconversion, revealed a significant difference between the proportions of PMs (Figure 2). In the genotype-predicted phenotypes, 6.8% of the patients were a PM of CYP2D6, which was 16.2% of the P-CYP2D6 patients (p < 0.05). No significant differences were found within CYP2C19 phenotype groups.

Comparing the total UKU-score, significant differences were found for the main diagnosis when comparing a psychotic diagnosis category (median score 13) to non-psychotic category (median score 22, p < 0.05) (Table 3). There was also a significant difference in the total UKU-score considering polypharmacy (patients without polypharmacy had a median score of 11, patients with polypharmacy a median score of 20, p < 0.05). The UKU-score was higher for patients with BMI over 40, but this was not a significant difference (p = 0.064; median BMI < 30 is 13, median BMI 30–40 is 15 and median BMI > 40 is 23.5). No significant differences were found for age and gender.

Patients with a non-psychotic main diagnosis had an odds ratio (OR) of 2.60 (95% CI: 1.04–6.54; adjusted OR 2.43; 95% CI: 0.90–6.55, Table 4) for higher UKU-score. Patients with polypharmacy had an OR of 4.47 (95% CI: 1.90–10.53; adjusted OR 4.26; 95% CI: 1.76–10.32) for higher UKU-score. No other statistical differences between UKU-score and covariates were found.

There was no increase of the total UKU-score associated with the genotype-predicted phenotypes of CYP2C19 and CYP2D6 (Table 3; Table 4). For P-CYP2D6, there was a significant difference between the total UKU-score of the NM (median 14) and UM (median 43.5, p < 0.05). There were no significant differences for IM (median 19.5) and PM (median 15). For P-CYP2C19, no increase in total UKU-score was seen considering the different phenotypes.

No significant differences were seen in the prevalence of certain side effects comparing the different genotype-predicted phenotypes of CYP2C19 (Table 5) and CYP2D6 (Table 6). Considering phenoconversion, several differences were noted. For P-CYP2C19, there was a higher prevalence of nausea and/or vomiting in the IM and PM group (75.0%) compared to the NM and UM group (9.1%, p < 0.05, Table 5). There were no further side effects with significant between-group differences. Comparing the prevalence of specific side effects and P-CYP2D6, there was a significant difference in the prevalence of depression (Table 6). 30.4% of the NMs and UMs experienced this side effect compared to 64.0% of the IMs and PMs. The IMs and PMs also experienced a higher rate of increased dream activity (48.0%) compared to the NMs and UMs (21.7%), but this result was not statistically significant (p = 0.057). The same is true for sleepiness, which 76.0% of the IMs and PMs experienced compared to 52.2% of the NMs and UMs (p = 0.085), and nausea and/or vomiting, which 36,0% of the IMs and PMs experienced compared to 13,0% for the NMs and UMs (p = 0.067). More patients in the NM and UM group of P-CYP2D6 (13.6%) experienced gynecomastia, compared to 0% of the IMs and PMs. However, this was also not statistically significant (p = 0.095). Other side effects did also not show any significant differences.

For the comparison of the CD-ratio, the number of patients was sufficient for the analysis of CYP2D6-substrates aripiprazole, risperidone, haloperidol and venlafaxine. From the data in Figure 3, it was apparent that the concentration-dose ratio of aripiprazole is significantly higher for IMs and PMs than NMs with (7.55 ± 3.46 μg/l/mg versus 16.03 ± 4.92 μg/l/mg) and without (8.98 ± 5.16 μg/l/mg versus 15.43 ± 5.25 μg/l/mg) considering phenoconversion. For risperidone, haloperidol and venlafaxine no statistic significant differences were seen.

This study shows that in 10% of the psychiatric patients any form of phenoconversion, where the predicted phenotype shifts based on genotype and co-medication, occurred. It also plays a role in the side effects experienced by these patients. This study shows that specific side effects such as nausea and depression are more prevalent in patients with an IM or PM phenoconverted phenotype of CYP2C19 and CYP2D6.

In this study population, there were significantly more CYP2C19 UM patients compared to the control population consisting of CAD-patients. In the Dutch Caribbean population, in which the phenotype distribution is comparable to Caucasians, a study found no differences at all in the prevalence of specific CYP2D6 or CYP2C19 phenotypes in psychiatric patients (Koopmans et al., 2017). This difference may be explained by the fact that the UMC Utrecht is a tertiary care center, to which patients are only referred if the treatment in the first or secondary line of care was not adequate. In a previous study in an American tertiary psychiatric hospital, a higher prevalence of genetic variants leading to a phenotype other than NM was seen (Ruaño et al., 2008). Looking at the distribution of phenotypes before and after phenoconversion, there were specifically more P-CYP2D6 PM patients. These findings seem to be consistent with the existing literature (Preskorn et al., 2013; Mostafa et al., 2019). This shows it is important to consider phenoconversion when predicting a patient’s phenotype. For CYP2C19 and CYP2D6, the phenotype influences the efficacy and tolerability of antidepressants and consequently there are different pharmacotherapeutic recommendations for each specific genotype-predicted phenotype (Gressier et al., 2015; Brouwer et al., 2022; Campos et al., 2022). There are also pharmacotherapeutic recommendations available for antipsychotics (Beunk et al., 2023). So, if the genotype-predicted phenotype shifts because of phenoconversion, it is possible that other recommendations are given.

Based on the total UKU-score, no statistically significant associations were found between the genotype-predicted or phenoconverted phenotype and amount and/or severity of side effects. However, the OR of the CYP2D6 phenoconverted phenotype was 1.31 and higher than 0.74, the OR of the CYP2D6 genotype-predicted phenotype. No comparison could be made for CYP2C19 because no logistic regression could be performed in the genotype-predicted group. In other studies, it has also been seen that there is a stronger association between phenoconverted phenotype and antidepressant efficacy and not necessarily between genotype-predicted phenotype and the antidepressant efficacy (Gressier et al., 2015).

This study found that there are also some non-genetic factors that influence the UKU-score of psychiatric patients. Patients who have polypharmacy have an OR of 4.26 for a moderate or high total UKU-score and therefore experience more or more severe side effects. These results are consistent with other studies, the higher the number of drugs a patient uses, the higher the chance of side effects and drug-drug interactions (Maher et al., 2014; Malki and Pearson, 2020).

A non-psychotic diagnosis also seemed to have an influence on the amount of side effects experienced. LL-patients with a non-psychotic diagnosis had a crude OR of 2.60 compared to LL-patients with a psychotic diagnosis. However, this result was not statistically significant when it was corrected for other covariables. The relationship between diagnosis and side effects is not clear, because several other underlying factors can play a role. The diagnosis of patients not only tells us something about the psychiatric disease, but also about the possible pharmacotherapy with associated side effects. Moreover, patients enrolled in the LL-clinic typically have more complex and more persistent mental disorders and therefore the current pharmacotherapy may lead to more side effects.

BMI may also play a role in the amount and/or severity of the side effects experienced. Patients with a BMI > 40 had a median UKU-score of 23.5, which was significantly higher than patients with a BMI between 30 and 40 or lower than 30, who respectively had scores of 15 and 13. However, this result was not seen when looking at the association between the UKU-score and the BMI as the OR was not statistically significant. Patients with a higher BMI often have a different response to antidepressants or antipsychotics, most of the time needing higher doses and thus experiencing more side effects (Warrings et al., 2021). Obesity and psychiatric disease have a complex bidirectional relationship (Holt and Peveler, 2009; Woo et al., 2016). Therefore, it is important to also consider the weight effects of a certain drug when choosing effective treatment for the mental disorder (McElroy, 2009). These aspects are already incorporated into the LL-clinic at the UMC Utrecht (UMC Utrecht, 2022a).

In this study it was also possible to look at the prevalence of CYP-specific side effects. It was found that 75.0% of the P-CYP2C19 IMs and PMs experience nausea and vomiting compared to 9.1% of the NMs and UMs. Another study also found CYP2C19 PMs have a higher chance of gastro-intestinal side effects when using an SSRI (Fabbri et al., 2018). Side effects that occurred more often in P-CYP2D6 IMs and PMs were depression, increased dream activity and sleepiness. 64.0% P-CYP2D6 IMs and PMs experienced the side effect depression compared to 30.4% of the NMs and UMs. The side effect depression is associated with antipsychotic use, possibly related to hyperprolactinemia (Milano et al., 2017). However, the relationship between hyperprolactinemia and CYP2D6 phenotype is unclear (Calafato et al., 2020). For the sleep-related side effects, some research indicates that these side effects are CYP-related, but different factors may influence this. For example, patients using a SSRI metabolized by CYP2D6 more often report nightmares as a side effect (Eugene, 2019). However, it also seems that psychiatric patients in general more often have more vivid dreams (Schredl and Schredl, 2018). The same goes for the relationship between sleep and psychiatric disorders (Krystal, 2012). Gynecomastia was a side effect more often seen in P-CYP2D6 NM + UM group. However, this side effect occurred specifically by the two UM patients. These two patients had a very complex background, having multiple comorbidities and/or switching drugs when the questionnaire was filled out, and therefore it is unsure if the occurrence of gynecomastia is due to the phenotype of these patients.

Lastly, next to the UKU-results, there was also an analysis of the drug concentration in plasma, corrected for the dose and phenotype. A two times higher CD-ratio was seen for (P-)CYP2D6 IMs and PMs, which was statistically significant. These results are consistent with the findings of previous work (Jukic et al., 2019; Kiss et al., 2020).

This study has several strengths. First, general genotyping (i.e., without a specific reason) is not standard clinical practice for psychiatric patients. However, for each patient enrolled in the LL-clinic, a psychiatric PGx panel was performed. Moreover, this study considered phenoconversion, specifically by co-medication, and its effect on the phenotype. Most studies in the field currently focus on the genotype-predicted phenotype and do not take the effect of co-medication into account (Shah et al., 2016). Specifically in the psychiatric population there is a lot of drugs that are known CYP-modulators, so it is an important factor (Flockhart et al., 2022). For future research it is important to look at other factors that influence phenoconversion such as smoking, alcohol consumption and disease state (Klomp et al., 2020). Third, the UKU is a validated standardized questionnaire, specifically for psychiatric drug-users (Lingjærde et al., 1987).

The current study was, however, limited by the sample size. Only one patient per week was enrolled in the LL-clinic. This could be a limitation specifically for phenotypes like UM and PM because of their lower prevalence. Moreover, not every patient used a CYP-substrate drug and therefore not all patients could be included in the analysis of CYP-substrate users and phenotype and phenotype groups had to be combined for some of the analyses. To include enough PMs and UMs, large-sample prospective trials need to be conducted. Another limitation is the data collection. Genotypes and drug use had to be obtained manually. To minimize the chance of errors, this was done in a standardized manner. Lastly, extrapyramidal symptoms were not included in the adapted version of the UKU-questionnaire. Also, although side effects were collected through the validated UKU-questionnaire, some side effects may need to be objectified using laboratory tests or physical examinations (Lingjærde et al., 1987).

In conclusion, this study shows that phenoconversion is important to consider when looking at a patient’s genotype. In the psychiatric population, where a difference in genotype distribution is observed, this phenomenon causes a shift from one phenotype to another. Although no significant associations were found between the phenotype and side effects experienced, there was a difference in the occurrence of specific side effects for the different phenoconverted phenotypes. More research on this topic is important to take the next step towards better prediction of a patient’s phenotype and possible prevention of side effects, contributing to personalized medicine.