Introduction

In addition to individual clinical trials, meta-analyses and pooled analyses are used for evidence-based evaluation of therapeutic interventions in depression to extend knowledge about treatment options and analyze specific data based on a larger sample size. In general, efficacy and tolerability of drugs are examined and confirmed by randomized controlled trials (RCTs) with narrowly defined patient collectives, under strict inclusion and exclusion criteria, and with a precisely defined therapeutic regimen. Therefore controlled trials may not fully represent real-world patients in daily practice. In psychiatric care, psychotropic co-medication also represents a risk for comorbid and elderly patients as a result of pharmacological interactions or possible effects on existing somatic diseases. Therefore non-interventional studies can provide relevant information in addition to existing evidence of RCTs, because heterogeneous patient populations are observed in everyday clinical practice.

In order to collect realistic data on agomelatine treatment of comorbid and co-medicated depressed patients, four non-interventional studies were performed in an ambulant setting in Germany between 2009 and 2013.

The antidepressant agomelatine is a melatonergic receptor agonist (MT1/MT2) and an antagonist at the postsynaptic serotonin receptor 5-HT2c [1, 2]. Its mode of action may be explained by the synergy of MT1/MT2 and 5-HT2C receptors, which increases noradrenergic and dopaminergic neurotransmission, and a specific release of noradrenaline and dopamine has been described in the prefrontal cortex without effect on the extracellular levels of serotonin [3,4,5]. In addition, an increase of neurotrophic factors (e.g., brain-derived neurotrophic factor, BDNF) and a decrease in stress-related glutamate elevation have also been described [2, 6, 7] as well as resynchronization of circadian rhythms [8,9,10].

After oral intake, agomelatine is rapidly and well absorbed (≥80%) with low absolute bioavailability (<5%) and substantial interindividual variability. Peak plasma concentration is reached after 1–2 h and mean plasma half-life is between 1 and 2 h. Agomelatine is rapidly metabolized, mainly by hepatic CYP1A2, CYP2C9, and CYP2C19, and inactive metabolites are eliminated in the urine. Agomelatine does not inhibit or induce CYP450 isoenzymes and will not modify exposure of medicinal products metabolized by CYP450. Additionally it is not a substrate, inducer, or inhibitor of P-glycoprotein (P-gp) [11,12,13]. Antidepressant efficacy and tolerability of agomelatine are confirmed in numerous clinical studies [10, 14,15,16,17,18,19,20,21,22,23,24,25,26,27].

The objective of this pooled analysis was an evaluation of agomelatine therapy in a large number of unselected patients in daily practice in order to gain information about the naturalistic treatment situation of patients with depression. This large sample size enables a valuable assessment of effectiveness, dosage, evolution of body weight, and rare or yet unknown adverse drug reactions (ADRs). Since monitoring of hepatic transaminases is recommended for agomelatine-treated patients, the presented data particularly focus on the analysis of incidence and course of hepatic transaminase elevation under daily treatment conditions.

Methods

Study Design and Population

The presented analysis was performed by pooling Germany-wide data collected by 1772 psychiatrists or specialized general practitioners (GPs) between 2009 and 2013. Overall 9601 outpatients of four prospective, multicentric, open-label, non-interventional studies with agomelatine were analyzed over a minimum of 12 to a maximum of 52 weeks (follow-up of VIVALDI study): VIVALDI (Valdoxan® Improves depressiVe symptoms And normaLizes circaDIan rhythms) and VIVALDI follow-up (2009/2010) [28,29,30,31] were performed exclusively in patients treated by psychiatrists, VIVALDI Praxis (2010) [32] only in GPs practices, VITAL (Valdoxan® Improves Treatment of depression and daytime Activity in real Life; 2011–2012) [33], and VIVRE (Valdoxan® ImproVes depRession with anxiEty symptoms; 2012–2013) [34] with psychiatrists and specialized GPs.

Data Collection

The pooled analysis includes all original data that had been collected in each of the individual studies in an identical or comparable way at a corresponding time of investigation. At the baseline visit (week 0), demographic data (age, sex, weight, smoking status), past psychiatric history (number of prior depressive episodes, duration of current episode, psychiatric comorbidities, previous suicide attempts), and form of therapy (pretreatment with antidepressants, psychotropic co-medication) were documented.

The antidepressant effect in all four included studies was assessed using the 7-point clinician rating scale Clinical Global Impression (CGI). Rater training had been offered to doctors in two studies to improve quality of assessment. Physicians evaluated severity of disease (CGI-S = Severity Scale; from 1 = “normal, not at all ill” to 7 = “most extremely ill”) and improvement or worsening of disease (CGI-I = Improvement Scale; from 1 = “very much improved” to 7 = “very much worse”); response (CGI-I ≤2) and remission rates (CGI-S = 1 or 2) were analyzed.

Tolerability data including the incidence and type of adverse drug reactions (ADR) were analyzed. Physicians were asked to document emergent ADR at each visit with standardized ADR documentation sheets according to German and European pharmacovigilance requirements. All ADRs were reported to BfArM (Bundesinstitut für Arzneimittel und Medizinprodukte) in accordance with applicable regulations. Hepatic transaminases alanine aminotransferase (ALT = GPT) and aspartate aminotransferase (AST = GOT) were documented, if available, at baseline and in week 6 (W6), 12, 24, and 52, and in accordance with the actual SmPC in 2012 (summary of product characteristics) additionally after 3 weeks in the VIVRE study. Recommended standard values for ALT and AST were used in cases where specified information was not available (50 U/I or 0.83 µmol/L for men; 35 U/I or 0.58 µmol/L for women). Transaminase values ALT and AST were classified as follows (ULN = upper limit of normal range):

Normal, within normal range

≤ULN

Abnormal, above normal range, but not clinically relevant

>ULN and ≤3× ULN

Abnormal, >3 times above normal range, clinically relevant

>3× ULN

The limit for clinical relevance of transaminase elevations (>3× ULN) was defined in accordance with FDA recommendations [35]. Transaminase values as well as ADRs were evaluated overall as well as separately for three treatment phases (W0–12, W12–24, and W24–52), considering acute, continuation, and maintenance treatment.

Data Analysis

Data of all four studies are included in the main analysis over 12 weeks. Patients with data at the inclusion visit and at least one further visit were eligible for statistical analysis of effectiveness (n = 9283; 96.2%). All 9601 patients with returned documentation forms were included in the safety analysis (ADR surveillance). Documented hepatic transaminase values (ALT/AST) were available for 9588 patients (at least one documented value), baseline and at least one follow-up value of ALT/AST for 6443/6282 patients (67%/65%), and values at every visit for 5061/4956 patients (53%/52%). Exclusion criteria for effectivity analysis were patients without intake of agomelatine or retrospective documentation.

Pooled analysis of CGI response and remission rates was performed for patients with valid data at each respective visit, resulting in different numbers of patients at each visit. A second approach of analysis was implemented to validate the results by evaluating only patients with available values at every visit.

Additional analyses of the treatment effect were performed in subpopulations over 24 weeks (VIVALDI follow-up/VITAL; n = 3610) and 52 weeks (VIVALDI follow-up; n = 605) as a result of the different duration of studies. Long-term tolerability was analyzed accordingly in the safety analysis for week 12–24 (VIVALDI follow-up/VITAL: n = 3915) and week 24–52 (VIVALDI follow-up: n = 605).

Evaluation of body weight and BMI (body mass index) are based on patients with a valid assessment of height and weight at W0 and weight at the respective follow-up visit (W12 or W24). Therefore body weight and BMI were analyzed over 12 weeks (VIVALDI/VIVALDI Praxis/VIVRE: n = 5273 and 5271) and 24 weeks (VITAL: n = 2687 and 2686, respectively).

The statistical analysis of pooled data was performed descriptively because of the non-interventional design of the individual studies using SAS® (version 9.2 for Microsoft Windows; SAS Institute Inc., Cary, NC, USA). Quantitative variables were evaluated using the basic statistical parameters (mean value ± standard deviation, median). For qualitative data (e.g., sex) and categorical variables (e.g., scale values), frequency distributions (absolute and relative frequency) were drawn up. Adverse events were coded according to MedDRA® and evaluated on the basis of the coding levels of “system organ class” and “preferred terms”.

Data management and statistical analysis were performed by the independent statistical institute GKM Gesellschaft für Therapieforschung mbH.

Compliance with Ethics Guidelines

This current article is based on previously conducted studies and does not involve any new studies of human or animal subjects performed by any of the authors.

All studies were conducted in accordance with recommendations of the German Medicines Act (AMG), the Federal Institute of Drugs and Medical Devices (BfArM), the Guidelines of the German Working Group for Epidemiology (DAE) ensuring Good Epidemiological Practice as well as recommendations of the Association of Research-Based Pharmaceutical Companies (VfA). All procedures followed were in accordance with ethical standards, all four individual studies were approved by the Free Ethics Committee Freiburg, Germany, submitted to German authorities (BfArM, KBV, GKV-Spitzenverband), included in the international study registry http://www.controlled-trials.com, and published [28,29,30,31,32,33,34]. All procedures also followed the ethical standards of the Helsinki Declaration of 1964, as revised in 2000 and 2008. Informed consent form was requested from all patients before being included in the original studies, but not mandatory because of the non-interventional design.

Results

Patient Characteristics

Patient characteristics and demographic data are shown in Table 1. The mean age of the pooled population was 50.5 ± 13.4 years (69% ≥45 years), 65.3% were female, and 60.2% of patients had a recurrent depression. Mean duration of the current episode was almost 11 weeks with the shortest duration reported by GPs (6 weeks) and the longest by psychiatrists (14 weeks). 72% of the pooled population (n = 6674) showed psychiatric comorbidities, most frequently sleep disorders (56.6%), anxiety/panic disorder (35.5%), post-traumatic stress disorder (7%), and others (14%), with multiple entries possible. The lowest incidence of psychiatric comorbidities (8.2%) was reported in VIVALDI Praxis (GP study) (see Table 1).

Table 1 Demographics of the pooled analysis dataset (n = 9283) and individual studies.
Full size table

After 12 weeks, 67.8% of patients were treated with agomelatine 25 mg/day (Table 2). In the subpopulation over 24 weeks (VITAL/VIVALDI follow-up) 72.6% of patients (n = 1914/2637) were treated with 1 tablet/day (25 mg). A dose increase was documented more frequently by psychiatrists in the VIVALDI study compared to GPs (VIVALDI Praxis) and following studies (VITAL and VIVRE).

Table 2 Patients (%) in the pooled dataset and individual studies according to agomelatine dosage over 12 weeks (W12).
Full size table

Psychotropic Co-medication

Overall, 60.1% of patients (n = 5581) were already pretreated with antidepressants, 70.2% in the VIVALDI study and 45.0% in VIVALDI Praxis. SSRI (selective serotonin reuptake inhibitors) were the most common medication in pretreatment (31.4%). In 18.9% of patients, previously prescribed antidepressants were continued in addition to agomelatine.

At baseline, co-medication with other psychotropic drugs was documented for 31.9% of patients, most frequently zolpidem or zopiclone (13.0%; n = 1204), benzodiazepines (10.1%; n = 933), antipsychotics/neuroleptics (5.7%; n = 531), anticonvulsant drugs (2.2%; n = 205), and 6.7% others (multiple entries were possible) (Table 1).

Change in Depressive Symptoms

The severity of depression (CGI-S) improved over 12 weeks (VIVALDI, VIVALDI Praxis, and VIVRE) with reduction of CGI-S (mean value) from 4.7 ± 0.8 at baseline (5 = patient is significantly ill) to 3.0 ± 1.3 after 12 weeks (3 = patient is only mildly ill). Overall, there was improvement of CGI-S in 81% of patients. As a result of longer observation periods in the studies VITAL and VIVALDI follow-up, the effect on the severity of depression was also analyzed after 6 months (W24). The pooled data (n = 3350) demonstrate a decrease in mean value of CGI-S from 4.7 ± 0.8 (W0) to 2.8 ± 1.3 (W24) (Fig. 1a), corresponding to an improvement of severity of depression (CGI-S) in 82.4% of patients after 24 weeks.

Fig. 1
figure 1

Modified with permission from [36]

Change of depressive symptoms according to CGI in the pooled dataset and individual studies over 12 weeks (W12) and 24/52 weeks (W24/52; subpopulations): a improvement of severity (CGI-S; mean value ± standard deviation). Asterisk indicates CGI-S values not available in VITAL study after 12 weeks (study duration 24 weeks), b response (CGI-I ≤2) and remission (CCI-S = 1 or 2) in %. CGI Clinical Global Impression scale, CGI-S (severity scale), CGI-I (improvement scale); assessment of all patients with available data at each respective visit; # in VIVRE study data were documented after 3 weeks

Full size image

On the basis of the CGI-I score, 78.7% of patients were classified as responders (CGI-I ≤2) after 12 weeks and 79.3%/75.9% after 24/52 weeks, respectively; 34.5% of patients met the criteria for remission (CGI-S = 1 or 2) after 12 weeks and 38.1%/47.5% after 24/52 weeks (Fig. 1b).

Tolerability

Frequency and type of ADRs, including serious ones (sADRs), were evaluated in the main analysis safety set (n = 9601) over 12 weeks: 5.32% of patients (511 of 9601) reported 859 ADRs (8.9%), most frequently headache, nausea, dizziness, and agitation; 0.49% of patients (n = 19/3915) reported 28 ADRs between weeks 12 and 24, six patients (n = 6/605; 0.99%) with seven ADRs between weeks 24 and 52. Headache, dizziness, and agitation were no longer reported after 24 weeks (Fig. 2a).

Fig. 2
figure 2

Tolerability according to most frequent ADRs in the course of treatment: a evaluation over 12 weeks, between 12–24 and 24–52 weeks, b incidence of ADR regarding psychotropic co-medication over 12 weeks. ADR adverse drug reactions, W week, ULN upper limit of normal range, >3 ULN clinically relevant elevation of transaminases; cutoff for incidence over 12 weeks >0.25%; asterisk indicates multiple entries were possible

Full size image

Eighteen patients (0.19%) reported 33 sADRs within the first 12 weeks. Three sADRs were documented in two patients (0.05%) between weeks 12 and 24; one of these patients (0.03%) had bradycardia and one patient (0.03%) had symptoms of hepatobiliary disorder starting after 10 weeks of treatment with agomelatine 50 mg/day. The hepatitis symptoms of this patient were completely reversible after discontinuation of treatment and the patient recovered without persisting impairment or sequelae. A detailed description of this patient has already been published [36]. No serious event occurred between weeks 24 and 52 (Table 3).

Table 3 Number (n) and percentage (%) of patients with serious adverse drug reactions (sADR) within different treatment periods (week 0–12/12–24/24–52), based on the safety analysis set (SAS) of each respective period.
Full size table

Regarding psychotropic co-medication, ADRs were documented for 5.28% (n = 161) of patients with (n = 3048) versus 5.34% (n = 350) of patients without co-medication (n = 6553) (Fig. 2b), showing no substantial difference in relation to psychotropic co-medication.

sADRs were reported for 0.3% of patients (n = 9) with psychotropic co-medication compared to 0.14% of patients (n = 9) without co-medication. The difference was mainly due to reports of headache, somnolence, aggression, dissociation, hallucination, insomnia and mania (one report for each symptom), and depression (n = 2) in co-medicated patients compared to no reports in patients without co-medication. For one patient of each group (with/without co-medication), transaminase elevations were assessed as sADRs (corresponding to 0.03%/0.02%). No deaths occurred and no long-term impairment was observed in the total population (n = 9601).

Change in Hepatic Values

Overall, 9588 patients with liver transaminase values (at least one value available) were included in this analysis. At baseline, hepatic transaminases (ALT and AST) were documented for 78.4% (n = 7517) and 76.4% (n = 7321) of all patients. Most of these patients showed ALT/AST values within the normal range (≤ULN; 88.4%/90.8%) or a mild increase without clinical relevance (≤3× ULN; 11.4%/9.0%). In 0.2% (ALT) and 0.1% (AST) of patients, a clinically relevant increase (>3× ULN) had already been documented before the start of treatment (Fig. 3).

Fig. 3
figure 3

Patients’ hepatic transaminase values (ALT and AST) during the course of treatment over 12 weeks (main analysis) and 24/52 weeks (subpopulations). ALT alanine aminotransferase, AST aspartate aminotransferase, >3 ULN >3 times above normal range; transaminase monitoring after 3 weeks was documented in VIVRE study only, according to the actual summary of product characteristics (3/2012) at time of study implementation

Full size image

In comparison of individual studies, values of liver function tests (ALT/AST) were available at every visit in 60.3% (ALT) and 58.9% (AST) of patients treated by GPs and 41.2% and 40.4% of patients treated by psychiatrists. Between 2009 and 2013, the proportion of documented values (ALT/AST) at every visit increased from 41.2%/40.4% up to 54.9%/53.9%. Accordingly, the proportion of documented values at baseline and at least one follow-up visit increased from 54.9%/53.5% to 76.4%/74.6% of patients.

Overall, clinically relevant transaminase elevations (>3× ULN) were documented in 0.5% of patients (n = 49/9601), 0.2% (n = 19) of whom showed clinically relevant ALT/AST elevations before the treatment with agomelatine at the study start (Table 4) and 30 patients (0.3%) an increase of values (>3× ULN) during treatment. In a calculation according to per protocol analysis (inclusion of patients with available values at every visit), 30 patients with elevations under treatment correspond to 0.6% and 23 patients with at least a possible causal link (ADR) to 0.46%. There was no difference in the incidence of transaminase elevations with respect to psychotropic co-medication (0.33%/0.32% of patients with/without psychotropic co-medication).

Table 4 Clinically relevant transaminase elevations (ALT and/or AST >3ULN) and documentation status of adverse drug reactions (ADR)
Full size table

Transaminase elevations were classified as ADRs in 23 patients (with at least possible causal relationship). Almost 80% of affected patients were between 25 and 65 years of age (4.4% >75 years). Detailed patient characteristics of affected patients are listed in Fig. 4.

Fig. 4
figure 4

Patient characteristics of the 23 patients with clinically relevant elevation of hepatic transaminases (ADR): a gender, age, used dosage, status of co-medication and reversibility; b patients (%) according to their duration of treatment until onset of ALT/AST elevation (>3 ULN). ADR adverse drug reaction, m male, f female, >3 ULN >3 times above the normal range; calculation based on pharmacovigilance data of patient reports; asterisk indicates no information available for dosage (n = 1; 4.4%), co-medication (n = 4; 17.4%), reversibility (n = 3; 13.0%)

Full size image

Over 12 weeks of treatment, 11 patients (0.2%) with normal baseline values developed ALT elevations and 7 patients (0.1%) AST elevations (>3× ULN). In patients with slight elevations at baseline, 0.1% showed clinically relevant elevations of ALT (n = 7) and/or AST (n = 5), respectively.

Evaluating liver function in long-term therapy, 0.04% of the patients with normal transaminase values at the beginning of treatment developed a relevant ALT and/or AST elevation (>3× ULN) over 24 weeks (n = 1/2437) and 0.3% (n = 1/288) over 52 weeks. Two patients with slightly elevated values (≤3× ULN) at the study start showed an ALT elevation (>3× ULN) over 24 weeks (0.1%) and no patient over 52 weeks (available values W0–W24, n = 2437/2387; W0–W52, n = 288/282). Additionally, no patient with normal or slightly elevated transaminase values at week 12 developed a clinically relevant increase in ALT or AST (>3× ULN) during further treatment.

Body Weight

Mean weight was 76.9 kg (±15.7) at the beginning of the observation period and 77.0 kg (±15.5) after 12 weeks. Corresponding values for the BMI were 26.4 ± 4.9 kg/m2 (W0) and 26.5 ± 4.8 kg/m2 (W12). Also after 24 weeks, the evaluated subpopulation (VITAL study) did not show any relevant changes in body weight or BMI (Fig. 5).

Fig. 5
figure 5

Evolution of a body weight (kg) and b BMI (kg/m2) over 12 weeks (VIVALDI, VIVALDI Praxis, VIVRE) and 24 weeks (subpopulation of VITAL). Evaluation of patients with valid data at baseline and W12 or W24, respectively

Full size image

Discussion

The presented analysis of naturalistic observational data provides important epidemiological data in a large number of non-selected patients (n = 9601). The findings demonstrate the effectiveness and good tolerability profile of agomelatine, which is in line with proven efficacy and tolerability in various RCTs [21, 26]. In the observational VIVALDI study [31] this effect was also demonstrated independent of psychotropic co-medication, which is confirmed in the presented mixed patient cohort.

Demographic data of the evaluated depressed population are representative of daily medical care with a high proportion of pretreated and comorbid patients [37]. Severity of depression was more pronounced in patients treated by specialists, reflected by the proportion of psychiatric comorbidities, co-medication, number and duration of depressive episodes as well as previous suicide attempts.

More than two-thirds of this large sample used agomelatine 25 mg/day (1 tablet) over 12 and 24 weeks, being in line with controlled data where about 75% of patients were medicated with 25 mg/day [38]. A dose increase in the presented patient cohort was documented more frequently by psychiatrists compared to GPs (VIVALDI Praxis), possibly as a result of more pronounced severity of depression and higher incidence of psychiatric comorbidities and co-medication in specialists’ practices.

In the present analysis, the CGI scale was used as an indicator for effectiveness, representing a general assessment of severity and improvement of depression by physicians. The CGI usually corresponds well with treatment effects evaluated with more specific scales (MADRS, Montgomery-Asberg Depression Rating Scale; Hamilton Depression Scale, HAM-D) in clinical and observational studies [31, 38, 39]. In the VIVALDI and VIVALDI Praxis trials, the short version MADRS (svMADRS) [40] was used in addition to CGI and confirmed CGI results, also being comparable to RCT data. Improvement according to CGI in our sample can therefore be seen as an indication for effectiveness of agomelatine with response and remission rates supporting the results of controlled data [38, 39]. Moreover, the antidepressant effect in this outpatient setting can also be evaluated in the continuation and maintenance phase of treatment, representing meaningful information in addition to clinical trials over 6–12 weeks, owing to a placebo control arm being included.

The overall incidence of ADRs in our unselected patient population is low with a known profile of adverse events (especially headache, nausea, dizziness). No new information was obtained about previously unknown ADRs compared to the existing database of controlled trials or non-interventional studies. Most ADRs/sADRs were documented within the first weeks of treatment and markedly decreased after 12 weeks. sADRs were no longer reported after week 24. Interestingly, ADRs and sADRs occurred nearly independent of psychotropic co-medication. Taken together, the presented results confirm data of clinical trials considering tolerability without effects on sexual function or cardiovascular parameters and a favorable interaction profile with emergence of ADRs primarily at the beginning of treatment [11].

In our large sample, no relevant changes of body weight occurred during short-term treatment or over 24 weeks. These results are in good agreement with possible positive metabolic effects of agomelatine described in several studies [41, 42]. A specific receptor profile without affinity to other receptors (especially no anti-histaminergic effect), normalization of circadian rhythms by melatonergic effects, and a positive influence on cortisol levels could serve as an explanation. These characteristics of the substance seem relevant, since weight gain is a frequently observed undesired effect with psychotropic medication [43].

In order to investigate the influence of agomelatine treatment on liver function, transaminase values were documented in all four studies. For the present pooled analysis, transaminase values (ALT/AST) were available for about half of the patients at every visit and for about two-thirds of all patients at baseline and at least one further follow-up visit. Comparison of the included individual studies confirms the assumption that transaminase monitoring is performed more often by GPs compared to psychiatrists. Besides, the increasing proportion of documented transaminase values over time between 2009 and 2013 is shown, presumably correlating with increasing awareness of physicians concerning transaminase monitoring in agomelatine-treated patients. Overall, clinically relevant elevations of ALT and/or AST (>3× ULN) were documented for 49 patients (0.5%), 19 (0.2%) of whom already showed elevations before the beginning of medication, hence not being classified as ADR. Transaminase values generally normalized after discontinuation of agomelatine, in some cases even during continuation of treatment, which is in line with RCT data [22, 23, 44]. One patient with symptoms of hepatitis and icterus recovered without any further impairment after discontinuation of treatment. A detailed case description has previously been published [36]. Before treatment initiation, only incomplete laboratory values were available for this patient, further supporting the importance of transaminase monitoring before medication, and thereby improving risk detection of preexisting impaired liver function. It is worth mentioning that no case of acute liver failure occurred under controlled conditions with strictly requested blood tests in RCTs in nearly 8000 patients [44].

Presented pooled data show a lower incidence rate of clinically relevant elevations compared to numerous controlled trials [11, 22, 23, 44]. Recruitment practice could possibly account for a higher incidence in RCTs, especially compared to American studies with more patients having preexisting risk factors (e.g., hepatobiliary disorder) [23, 45]. Strict clinical monitoring under controlled conditions could also explain higher reporting rates in RCTs compared to this naturalistic design. The demonstrated lower incidence in this pooled data set, however, is verified by an additional analysis based on patients with available values at every visit (ALT, n = 5061; AST, n = 4956). This additional calculation (comparable to “per protocol” analysis) confirms lower incidence rates of transaminase elevations (ADRs) in 23 patients corresponding to 0.46% and of all documented cases including preexisting baseline elevations with 0.98% (n = 49).

Various database analyses have been published so far, investigating spontaneous reports of adverse hepatic effects during agomelatine treatment [46,47,48], to explore quantitative signals about reporting frequencies without causality assessments, since reports listed in safety databases are “raw data” [49]. Gahr and colleagues describe that hepatic effects in agomelatine-treated patients (mostly asymptomatic transaminase increase) occur primarily in the initial phase of treatment. Polypharmacy, female gender, and age over 50 years were described as possible risk factors [46].

Our results confirm the emergence of transaminase elevations within the first weeks of treatment, reversibility, and increased occurrence in women and patients with co-medication. Older age, however, did not correlate with higher risk of transaminase abnormalities in our population with two-thirds of the patients being younger than 55 years.

This result is confirmed by RCT data, demonstrating good tolerability in older patients up to 6 months of treatment [50, 51] without the necessity for dose adjustment in elderly patients [11]. Higher incidence of transaminase elevations in women within this sample could possibly be explained by the high proportion of women in the depressed population in general. Nevertheless, even taking baseline data into account with a ratio of 2:1 (female/male), the presented results still show a trend for a higher frequency in women.

With the objective of ensuring therapeutic safety, the European Medicines Agency (EMA) has developed increasingly rigorous criteria regarding quality of studies and especially pharmacovigilance procedures. Regular monitoring of various blood tests, ECG (electrocardiogram), EEG (electroencephalography), or other controls are required or recommended for many antidepressants or psychotropic drugs in guidelines, standard references, and textbooks [37, 47, 52, 53].

In this context, monitoring of hepatic transaminases is listed in the SmPC of agomelatine and should be performed before the initiation of treatment, after approximately 3, 6, 12, and 24 weeks, after emergence of clinical symptoms of hepatic dysfunction, and after a dose increase [11].

Considering the general risk of possible drug-related liver damage during treatment with psychotropic drugs, monitoring of liver function tests is recommended in general [11, 37, 47, 52, 54, 55]. Thereby therapeutic safety as well as early detection of patients at risk with preexisting relevant diseases can be improved, especially in case of polypharmacy [37, 52, 54,55,56,57].

Limitations

An important limitation to be mentioned is the non-interventional, open-label, observational design of the four individual studies included in the present pooled analysis, with lack of randomization, blinding, and placebo control group. Therefore the observational design might lead to an overestimation of therapeutic effects compared to RCTs. Controlled studies with strict inclusion/exclusion criteria, however, do not represent naturalistic patient populations with comorbidities and co-medication, which underlines the relevance of non-interventional studies. Even though observational trials are not able to provide proof of efficacy because of methodological reasons, they nonetheless provide important information of everyday clinical practice. The notable strength of the presented pooled analysis is the large number of non-selected patients.

Potential underestimation of adverse drug reactions due to possible underreporting can not be completely ruled out. As a result of the non-interventional design, ADRs were not assessed systematically but were documented in the form of open questions at each visit. Considering data of controlled studies, however, type, severity, and time of emerging ADRs within our presented pooled data are in line with the tolerability profile demonstrated in RCTs.

Short duration of individual studies over 3 months is a further limitation worth being discussed. The main analysis of pooled data was performed after 12 weeks, thereby representing short-term treatment. However, this correlates well with the duration of RCTs between 6 and 12 weeks. Besides, a subgroup over 24 weeks with more than 3000 patients was additionally analyzed in this pooled dataset, providing relevant data for continuation treatment in unselected patients.

Another limitation to be mentioned is the mixed nature of data due to significant differences in investigators. In the presented pooled analysis the treatment effect of agomelatine has been evaluated rather generally by means of the CGI as a less specific scale compared to more detailed scales used in psychiatry. Besides, GPs in VIVALDI Praxis were offered a rater training (like psychiatrists in the VIVALDI study) via a video test version of the more complex svMADRS questionnaire to ensure a high quality of collected data. Concerning the analysis of tolerability, the specialization of doctors should probably generate only minor differences. According to the objective of gaining information about psychiatric care in daily routine, the strength of the actual pooled analysis is the large sample of comorbid and co-medicated patients of daily practice, as in German psychiatric care the intersectoral treatment situation between hospital care and outpatient treatment (by specialists and/or GPs) is an actual and important issue.

As a result of possible differences in the point of view on psychiatric disorders between the two groups of physicians, separate results for psychiatrists and GPs would have been informative. In conformity with the statistical analysis plan (SAP) and the stated objective of analyzing a large sample of naturalistic psychiatric patients as an overall group, stratified data analysis according to doctors’ specialization has not yet been performed. However, baseline data, dosage, and evolution of depressive symptoms by means of CGI are presented in the total population and on the basis of individual studies, thereby allowing a first insight into this interesting question. Further evaluation of this large database with a focus on stratification could offer an additional objective for a prospective data analysis.

Conclusion

In the presented pooled analysis of four prospective open-label studies, agomelatine demonstrated good clinical effectiveness and tolerability in a heterogeneous patient cohort (n = 9601) over 12 weeks and over 24/52 weeks (subpopulations). Overall improvement of depressive symptoms is demonstrated in CGI response and remission rates. The incidence of adverse drug reactions is low (5.3%), irrespective of psychotropic co-medication, and no new information about previously unknown or very rare ADR is obtained. Additionally, the results provide naturalistic information about potential effects of agomelatine on liver function, mainly observed in the form of reversible transaminase elevations and primarily observed within the first 12 weeks. Presented data confirm current recommendations for transaminase monitoring, especially during the first months of treatment.

In summary, the pooled data analysis reflects effectiveness and tolerability of agomelatine, supplementary to clinical trials, and independent of psychotropic co-medication in non-selected patients of daily clinical practice.