Searches, study and participant characteristics
For our meta-analysis individual patient data of three double-blind, randomised, placebo-controlled, multicentre phase III clinical trials in patients with subthreshold anxiety [14,15,16] were obtained from the manufacturer of Silexan. To identify any additional studies performed with Silexan in patients with subthreshold anxiety, we performed free-text searches of all fields of the MEDLINE database as well as of the ClinicalTrials.gov registry, the EMA Clinical Trials Register, and the ISRCTN registry for any records that included the search term ‘Silexan’ in combination with ‘anxiety’ and had been entered before 31 March 2017 (no further restrictions applied). In addition to the original publications presenting the results of the three trials above [14,15,16], the MEDLINE search retrieved a total of 14 other publications. Four described randomised, controlled trials with Silexan in other psychiatric indications [13, 17,18,19], and three covered non-clinical studies with Silexan [10, 12, 20]. Retrieved publications also included a contraceptives interaction study [21] and a case report [22]. The remaining five publications were review articles from which no additional studies with Silexan in patients with subthreshold anxiety could be identified [14, 23,24,25,26]. Searches performed in clinical trials registries also did not reveal any additional studies with Silexan performed in the indication of interest. The search results thus indicate that the three randomised, controlled trials performed in patients with psychiatric diagnoses characterised by subthreshold anxiety appear to be the only randomised, controlled studies performed with Silexan in the population of interest by the time when our searches were completed. The remaining identified trials covered different indications and were thus excluded.
Among the eligible trials, Study A assessed the efficacy of Silexan patients with SSAD (classified as ‘anxiety disorder not otherwise specified’ according to the DSM-IV 300.00 and as ‘anxiety disorder, unspecified’ according to ICD-10 F41.9) [14]. The participants of Study B suffered from restlessness and agitation (ICD-10 R45.1) as well as from disturbed sleep [16]. Study C was performed in patients diagnosed to be suffering from MADD (ICD-10 F41.2) [15].
The main characteristics and criteria for participant selection of the included trials are summarised in Table 1. The participants of all studies were female or male, adult out-patients who consulted a general practitioner or a psychiatrist. Each trial started with a 3–7 days qualification phase after which eligible patients were randomised and received Silexan or placebo for a scheduled period of 10 weeks, with assessments performed at baseline (randomisation) and at the end of weeks two, four, six, eight, and ten (Studies A and B), or at the end of weeks one, two, four, seven, and ten (Study C). For eligibility, patients of all trials had to meet the diagnostic criteria for the condition under investigation and had to present with a total score ≥ 18 points for the Hamilton Anxiety Rating Scale (HAMA) as well as with minimum scores for two individual HAMA items as shown in Table 1 at both inclusion and baseline.
Table 1 Main study characteristics and participant selection criteria Ethical conduct
All trials included into our meta-analysis were performed following the principles of Good Clinical Practice and the Declaration of Helsinki after obtaining approval from an Independent ethics committee. The studies were registered in the EudraCT database (all studies) as well as in the ISRCTN registry (studies A and C). All participants provided written informed consent.
Interventions
Silexan is a special active substance with an essential oil produced from Lavandula angustifolia flowers by steam distillation that complies with the monograph Lavender oil of the European Pharmacopoeia and exceeds the quality requirements of the monograph. Batch to batch consistency is assured by a well-defined, standardised manufacturing process. Immediate release soft gelatine capsules containing 80 mg of Silexan or identically matched placebo capsules were used. The smell of the investigational treatments was matched by flavouring the capsules containing placebo with 1/1000 of the amount of lavender oil contained in the Silexan capsules. In all studies randomised patients received Silexan monotherapy and had to administer one capsule per day unchewed in the morning. The daily dose was chosen in accordance with recommended dose of the marketed product.
Meta-analysis outcomes
The anxiolytic effect of Silexan was assessed by analysing HAMA total score change between baseline and treatment end, which was the pre-defined main outcome in all eligible studies. Moreover, we analysed the change of the HAMA psychic and somatic anxiety subscores [27] as well as of the individual item scores for anxious mood, tension, and sleep. Anxiety self-ratings were included into the meta-analysis using the change of the total score of the Zung Self-rating Anxiety Scale (SAS; studies A and B) or of the anxiety rating of the Hospital Anxiety and Depression Scale (HADS; study C). Sleep quality was assessed by analysing the total score change of the Pittsburgh Sleep Quality Index (PSQI; studies A and B only). Meta-analyses were also performed for the changes of the mental health and the physical health subscores of the SF-36 Health Survey Questionnaire which assesses health-related quality of life (studies A and C only), and of Items 1 and 2 of the Clinical Global Impressions scale (CGI).
Our meta-analysis also included assessments of treatment response and remission. We defined response as a decrease of the HAMA total score by at least 50% of the baseline value or as a score equal to or less than two points for CGI item 2 (i.e., much or very much improved compared to project admission), both assessed at treatment end. Remission was defined as a HAMA total score of less than ten points or of less than or equal to seven points at treatment end.
In all trials tolerability and safety were primarily assessed by monitoring adverse events (AEs).
Bias assessment
Bias assessment on the study level was performed by an independent assessor who was not involved in the planning, conduct, analysis or interpretation of any of the eligible trials, using the Cochrane Collaboration’s tool for assessing risk of bias [28]. Assessments were based upon the applicable publications, the patient raw data, and on the original protocols and the full integrated study reports made available to the authors and to the assessor.
Statistical methods
The meta-analysis of treatment efficacy was based on the original (raw) data of the included trials and was performed for the primary efficacy analysis data sets (full analysis set, FAS) of the original protocols. For comparability with the published results of the trials, missing data for efficacy outcomes were imputed by carrying forward the last valid observation.
Patient age, sex, and premature withdrawal rate were analysed using descriptive statistics. Within each trial continuous outcomes were analysed using analysis of covariance (ANCOVA) with treatment as a factor, the intraindividual difference between treatment end and baseline for the outcome of interest as the dependent variable, and the baseline value of the outcome as a covariate. Rating scales were analysed as continuous outcomes. For the analysis of CGI item 2 (‘Change from project admission’), which inherently includes change from baseline and which was thus assessed only at post-baseline visits, the baseline value of CGI item 1 (‘Severity of illness’) was used as a covariate.
Meta-analysis methods were pre-defined in a statistical analysis plan. We used a two-stage individual participant data (IPD) meta-analysis approach according to which the outcomes of interest were first analysed equally within each study and then combined using ‘traditional’ meta-analysis [29, 30]. For continuous outcomes, we used the marginal (adjusted) mean values and their estimated standard deviations as input for the meta-analysis and computed random effects models based on the treatment group mean value difference. Inverse variance weighting was used for combining the results of the single trials. We applied the DerSimonian–Laird method to calculate the variance between the trials. In the meta-analysis of self-rated anxiety, which combined ratings originating from the SAS and the HADS clinical questionnaires, the bias corrected Hedges’ G was calculated as an estimate for the combined, standardised mean value difference between the treatments in order to account for the different scales. For the HAMA total score difference and the clinical global impression of change from baseline (CGI item 2) the bias corrected Hedges’ G was calculated in addition in order to facilitate the comparison of results with other published work. Meta-analyses of binary outcomes (response, remission, and AE rates) were based on relative risk. For response and remission, random effects models were used, and trial results were combined according to the inverse variance method. Meta-analyses AE rates were performed using fixed effects models with Mantel–Haenszel weighting for combining the trial results. Additionally meta-analyses of binary outcomes based on risk differences were performed to calculate numbers needed to treat and numbers needed to harm.
For all analyses two-sided p values ≤ 0.05 were considered descriptively significant.
Heterogeneity between the trials was assessed using the I2 statistic in accordance with the criteria proposed in section 9.5.2 of the Cochrane Handbook for Systematic Reviews of Interventions [31].
Meta-analyses were computed with the R software package meta (version 4.3.2) using functions metacont and metabin for continuous and binary data, respectively. All other analyses were performed in SAS statistical software version 9.3.