Introduction

Major depressive disorder (MDD) is one of the most prevalent psychiatric diseases, affecting more than 300 million people globally [1]. MDD leads to cognitive impairment, which results in suicidal behaviors [2,3,4,5]. More than 800,000 people die from suicide each year [6]. For decades, medications for MDD followed clinical guidelines of MDD treatment, while numerous anti-depressant drugs went on the market and were considered to be “first-line” for comparable remission and/or response rates [7]. However, following the present anti-depressant treatment pathway, only 50% of patients achieve clinically significant responses within their first anti-depressant treatment; the number of patients achieving remission is even fewer [8]. Thus, the current standard treatment pathway is a trial-and-error approach until a relatively effective anti-depressant or combination treatment is found that provides full remission effects [7]. Current treatment approaches result in a prolonged duration of unremitted illness, which is related to worsened long-term prognosis, adverse changes in cognitive function and brain morphology, and increased side effects [9, 10]. Because of those problems with current MDD medications, better tools are urgently needed to help physicians and patients find effective treatment.

Pharmacogenomic testing offers a point-of-care, personalized, and ascendable tactic to guide clinical treatment in multiple diseases pre-emptively, including MDD [11]. A number of proprietary pharmacogenomic testing tools have been developed to guide drug selection or conversion by testing allelic gene variants of assigned genes mediating the pharmacodynamics and/or pharmacokinetics of anti-depressant medication [12]. Numerous randomized controlled trials (RCTs) have been conducted to evaluate the efficacy and safety outcomes of pharmacogenomic testing in MDD, but the results and conclusions are still conflicting. Several large-scale RCTs depicted significant improvement in response and remission rates, while a recent study by Oslin et al. showed no significant change between the pharmacogenetic testing guided treatment group and the usual treated group [13,14,15,16]. The key point of whether patients benefit from pharmacogenetic testing remains unclear.

Several previous systematic review and meta-analyses had been conducted based on this issue [12, 17, 18]. However, since the number of studies involved was limited, no sufficient data was provided to reach a reliable conclusion. On the other hand, cohort studies and RCTs were both included in the meta-analyses which could contribute to potential bias. Furthermore, plenty of RCTs regarding this topic have been published in recent two years [14, 15, 19], reflecting the necessity of an updated systematic review and meta-analysis. Here in our current study, we determined the effect of pharmacogenetic testing guided on clinical outcomes in MDD patients.

Methods

Protocol and registration

This is a protocol-driven systematic review and meta-analysis, prospectively registered with the International Prospective Register of Systematic Reviews (PROSPERO CRD42022360151). We performed this meta-analysis in accordance with the guideline of Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guideline [20], and the checklist is available in Supplemental Material (Method S1).

Literature research

We searched PubMed, Embase, and Cochrane Library of Clinical Trials from inception until August 2022. The key terms were ‘pharmacogenomic’ and ‘antidepressive’. The detail of the search terms is attached in Table S1. The reference lists of all included studies were also examined for relevant citations. No language restrictions were applied.

Study selection

Studies were considered for inclusion based on the following criteria: [1] participants were diagnosed with major depressive disorder; [2] the study included both pharmacogenomic testing guided group and usual care group; [3] randomized controlled trial (RCT). The screening and scanning for eligibility were performed manually by two independent reviewers (X. W. and C. W.) through Endnote (version 9.3.2). Any disagreements were resolved by discussion with a third reviewer (Z. A.).

Quality assessment

Six domains (random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, and selective reporting) were assessed by using the Cochrane Collaboration’s tool for assessing the risk of bias [21] by two reviewers (X. W. and C. W.) independently. Disagreements on study quality assessment were discussed with another reviewer (Y. Z.) until a consensus was reached.

Data extraction

Data were independently extracted by two investigators (X. W. and C. W.) for eligible studies. The characteristic data obtained for each study included the first author, year of publication, study design, inclusion criteria, target genes, baseline characteristics, and industry funding. Outcomes were response (≥ 50% decrease in HAM-D17/PHQ-9 score from baseline) rate and remission (a score of ≤ 7 for HAM-D17 or ≤ 5 for PHQ-9 or ≤ 2 for CGI score) rate at week 4, 8, 12, and 24, respectively. Another outcome was medication congruence in 30 days (participants that prescribed antidepressant medication that was categorized as having no drug-gene interaction or moderate drug-gene interaction). We gave preference to data from intent-to-treat analysis over pre-protocol analysis. Data with longer follow-up duration were chosen over the shorter one when data at several time points were provided within the period.

Statistical analysis

Odds ratios (RR) with 95% confidence intervals (95%CIs) were calculated with R (version 4.0.5). Heterogeneity across the studies was assessed using the Cochran’s Q test; the percentage of total variability attributable to heterogeneity was quantified by the I2 value. I2 less than 50% indicates low or moderate heterogeneity and more than 50% indicates high heterogeneity [21]. Random-effects model with inverse variance weighting was used for high heterogeneity, and the Mantel-Haenszel fixed-effects model was used for low or moderate heterogeneity of included studies. P values less than 0.05 were considered significant. Funnel plots were used to test the publication bias.

Subgroup analysis was done according to follow-up duration for less than 12 weeks (data with the longest follow-up duration through the period was extracted). Only a few studies included provided data with longer duration, so subgroup analysis for outcomes at week 24 was not conducted. Four further subgroup analyses were carried out according to study design (single-center or multi-center), sample size (less than 200 patients or more than 200 patients), population (the majority ethnicity of the population: Caucasian or Asian), diagnostic criteria (HAM-D17 or PHQ-9), and industry funding (fully funded, partially funded, or none funded). A difference between the estimates of these subgroups was considered significant for the P-value between subgroups < 0.10 [22].

To evaluate the stability of the results, we did a sensitivity analysis by sequentially removing every single study from the pooled effect estimates.

Results

1779 potentially relevant articles were screened. 1712 irrelevant articles were excluded based on screening of the titles and abstracts. 54 articles went through full-text review and 43 of them were excluded because they did not meet the inclusion criteria. Lastly, eleven RCT studies with a total of 5347 patients were included in the meta-analysis (Figure S1). The duration of follow-up was 8 weeks in four studies, 10 weeks in one study, 12 weeks in three studies, 24 weeks in two studies, and 52 weeks in one study. The specific genes tested differed between studies (Table 1).

Table 1 Baseline characteristics of the randomized controlled studies assessing the effect of pharmacogenomics testing guided in major depressive disorder patients
Full size table

Assessment of risk

The overall result of the risk assessment is summarized in Figure S2. Most of the studies were categorized as having a high bias in the other bias section since most of them were funded by industry. Only one study [23] received no industry funding. One study [15] was assigned high bias in the blinding of participants and personnel section since it was a single-blinded (rater-blinded) study; while others that assigned low bias were double-blind (patient- and rater-blinded) studies. Funnel plots of outcomes were presented in Figure S3.

Response rate

Response rates were significantly increased in the guided group compared with the usual group at week 8 (OR 1.32, 95%CI 1.15–1.53, 8 studies, 4328 participants) [13,14,15,16, 23,24,25,26] and week 12 (OR 1.36, 95%CI 1.15–1.62, 4 studies, 2814 participants) [14, 15, 24, 27]. However, no significant difference was observed between the guided group and the usual group at week 4 (OR 1.12, 95%CI 0.89–1.41, 2 studies, 2261 participants) [14, 15] or week 24 (OR 1.16, 95%CI 0.96–1.41, 2 studies, 2252 participants) [14, 15] (Fig. 1).

Fig. 1
figure 1

Forest plot of response rate at week 4, 8, 12, and 24 comparing pharmacogenomic testing guided treatment versus usual care treatment. OR, odds ratio; CI, confidence interval

Full size image

Remission rate

Compared with the usual group, the guided group was associated with an increased remission rate at week 8 (OR 1.58, 95%CI 1.31–1.92, 8 studies, 3971 participants) [13,14,15, 19, 23,24,25,26] and week 12 (OR 2.23, 95%CI 1.23–4.04, 5 studies, 2664 participants) [14, 15, 24, 27, 28]. Similarly, we found no significant difference between the guided group and the usual group at week 4 (OR 1.26, 95%CI 0.93–1.72, 2 studies, 2261 participants) [14, 15] and week 24 (OR 1.06, 95%CI 0.83–1.34, 2 studies, 2252 participants) [14, 15] (Fig. 2).

Fig. 2
figure 2

Forest plot of remission rate at week 4, 8, 12, and 24 comparing pharmacogenomic testing guided treatment versus usual care treatment. OR, odds ratio; CI, confidence interval

Full size image

Medication congruence in 30 days

Medication congruency in 30 days was significantly reduced in the pharmacogenomic guided group compared with the usual care group (OR 2.07, 95%CI 1.69–2.54, 3 studies, 2862 participants) [13,14,15](Fig. 3).

Fig. 3
figure 3

Forest plot of medication congruence in 30 days comparing pharmacogenomic testing guided treatment versus usual care treatment. OR, odds ratio; CI, confidence interval

Full size image

Subgroup analysis and sensitivity analysis

Table 2 shows the subgroup analysis according to several parameters. Specifically, we found significant differences between subgroups of the target population in both response (P-value between subgroup = 0.03) and remission rate (P-value between subgroup < 0.01). In addition, significant differences between subgroup in subgroup analyses according to study design, sample size, and industry funding of remission rate were also observed, while no significant difference between subgroups was found between different diagnostic criteria groups; however, the pharmacogenomic testing guided group compared with usual group were all associated with an increased rate of remission in subgroups.

Table 2 Subgroup analysis of remission and response rate in week 0–12 in studies comparing pharmacogenomics testing guided and usual care group in MDD patients
Full size table

No outcome changed when performing sensitivity analysis using the leave-one-out method (Figure S4).

Discussion

The current meta-analysis identified 11 RCTs that compared the effect of pharmacogenomic testing on target outcomes were identified. Quantitative pooled analysis showed that pharmacogenomic testing guided treatment contributed to improved response and remission rates compared to usual treatment at week 8 and 12, while no significantly higher response or remission rates were found at week 4 or 24. Pharmacogenomic testing guided treatment shortened the time to clinical remission and response to antidepressants. In addition, the pharmacogenomic testing guided treatment decreased medication congruence compared to usual care. In subgroup analysis, we found significant differences between the Asian subgroup and the Caucasian subgroup.

Overall, these data showed that pharmacogenomic testing guided treatment helped MDD patients achieve clinical remission or response in a shorter time compared to usual treatment while making no difference in final response or remission rates at the end of the pharmacogenomic testing guided treatment. Notably, our results were different from previous meta-analyses, which simply showed higher response and remission rates in pharmacogenomic testing guided MDD patients than usual treated patients [12, 18, 29]. When combining all the outcomes at any follow-up duration from included studies, we found similar results as the previous studies (Figure S5). However, when separating the data into week 4, 8, 12, and 24, we found significantly improved response and remission rates in week 8 and 12 rather than 4 and 24. No significant changes were found at week 4 may because of the long onset time of anti-depressants, and no significant changes were found at week 24 may because pharmacogenomic testing guidance only showed a ‘catalyst-like’ effect by accelerating the process of excluding the unsuitable anti-depressants for MDD patients rather than improving the MDD therapeutic effects.

This current meta-analysis found fewer gene-drug interactions in MDD patients receiving pharmacogenomic testing guided treatment than usual in those receiving usual treatment. These results showed an effect of pharmacogenomic testing guidance on decreasing severe adverse effects induced by gene-drug interactions during MDD medications. Pharmacogenomic testing guidance provided relevant clinical information on the effect of an anti-depressant, potential gene-drug interactions, and adverse effects for physicians and patients, helping to make more accurate decisions [15].

The genes that encode hepatic cytochrome P450 (CYP450) enzymes receive great concerns in pharmacogenomic testing, because of their roles in the regulations of psychotropic drug metabolism [30]. Previous studies revealed the critical role of CYP450 enzymes on the pharmacokinetics of anti-depressants, especially CYP2C19 and CYP2D6 [31, 32], with high-level evidence showing relevance to medications commonly used in psychiatry practice [32,33,34]. Briefly, findings showed that variants of CYP2C19 and CYP2D6 have been associated with blood concentrations of antidepressants, adverse drug reactions, and clinical outcomes of antidepressants [35, 36]. Moreover, variations of other genes were also thought to be closely associated to different responses to antidepressants, such as the human gene-encoding serotonin transporter (SCL6A4) gene, catechol-o-methyltransferase (COMT) gene, and serotonin-2 A receptor (HTR2A) gene [37,38,39,40].

In subgroup analysis, the current study showed a significant difference between the Asian subgroup and the Caucasian subgroup. The difference may be associated with the sub-genotype of allele frequencies of the gene variants. For example, more CYP2C19 and CYP2D6 variants were found in Asian people than in Caucasians. Almost 20% of Asians were CYP2C19 poor metabolizers while 3% of Caucasians were poor metabolizers [41, 42]; only 1% of healthy Asians were attributed to poor metabolizers of CYP2D6 while the proportion of Caucasians was 5–10% [43, 44]. Furthermore, allele frequencies of the SLC6A4 gene variant between Caucasians and Asians are different, the S allele being present in 42% of Caucasians but in 79% of Asians [45]. The difference in pharmacokinetics enzymes led to different responses to anti-depressants between Asian and Caucasian. Among those genes which were reported to have different responses to antidepressants, CYP450 enzyme genes, especially CYP2C19 and CYP2D6, may mask the effects of other genes because of the sensible effects on the regulation of antidepressant metabolism.

But meanwhile, it’s still challenging for pharmacogenomic testing guidance utilized in clinical MDD medications. Firstly, there is not enough evidence to support the use of pharmacogenomic testing guidelines in MDD clinical practice. The US Food and Drug Administration cautioned about the content validity and potentially detrimental impact of pharmacogenomic testing panels, while Dutch and French guidelines recommended pharmacogenomic testing guided treatment as a potentially useful tool of MDD medication [46, 47]. Although the current study showed the effects of pharmacogenomic testing guidance on shortening the time to identify an appropriate antidepressant, there are still difficulties in developing pharmacogenomic testing guidance on MDD medications. Secondly, the cost-effectiveness of pharmacogenomic testing guidance remains unclear. The average number of genes tested in studies included in the current meta-analysis was more than 10, which may lead to huge expenditure and result in low acceptance of MDD patients. Though pharmacogenomic testing may not lead to less direct cost than usual antidepressant treatment, it’s still a very profitable tool to decrease taking inappropriate antidepressants of MDD patients and help reduce indirect costs [48,49,50]. Moreover, several institutions provide different pharmacogenomic testing tools and services; all those tools have their unique pipelines, and results are not provided uniformly [51, 52]. Thus, clinicians should get support from genetic counselors, pharmacists, or companies for proper implementation [29]. Previous case reports highlighted the key role of pharmacist assessment during drug-gene interactions and drug‐induced pheno-conversion in MDD. Further open-label RCT about pharmacist-guided pharmacogenetic testing in antidepressant therapy is in progress [53, 54]. Pharmacogenomic testing guidance would make a huge difference in MDD medications if those problems could be resolved.

The current study included only RCTs with longer follow-up duration and a much larger sample size. We also analyzed data at multiple time points and conducted multiple comprehensive sub-group analyses. Although this study revealed the effect of pharmacogenomic testing guided treatment on MDD medications and indicated its striking value in Asian MDD patients, there are still several methodological limitations that may lead to interference and bias that emerged from this study. The reliability of the inclusion of unblinded clinicians was reduced in most of the reported pooled effect studies. In addition, patients were unblinded in the largest scale research included in this current analysis, which may also lead to detection bias of outcomes. Another significant limitation of this current study is on basis of data in per-protocol analysis rather than intention-to-treat analysis, which may lead to overestimating the outcomes of pharmacogenomic testing guided treatment on MDD medications. Nevertheless, this meta-analysis is a re-analysis based on current research so a few genotypes were involved. Previous reports focused on the effects of CYP2C19 and CYP2D6 on antidepressants, thus our research mostly mentioned the two variants [31, 32]. In the future, more genotypes would be involved in our prospective research.

Conclusions

Our study demonstrated the effect of pharmacogenomic testing guidance in shortening the process of reaching clinical remission and response to anti-depressant medications in MDD. Future well-designed multi-ethnic studies are needed to confirm the benefit of pharmacogenomic testing guidance in different populations.