This exploratory, post-hoc study is a secondary analysis of the VESPA study data. The VESPA study is a pragmatic, assessor-blinded, randomized, parallel-group, superiority multicentre trial. It was funded by the Italian Medicines Agency (Agenzia Italiana del Farmaco – AIFA) in the context of the 2016 call for Independent Research on Drugs (code: 2016-0234923). The study was designed according to the principles of the CONsolidated Standards of Reporting Trials (CONSORT) statement [16]. Full details on study protocol, design, and procedures are reported elsewhere [11, 17]. The study was conducted between February 2019 and February 2023 in inpatient and outpatient Mental Health Services and was run by 11 national academic centres, i.e. University of Verona, University of Catania, Magna Graecia University of Catanzaro, University of Ferrara, University of Genova, University of Chieti-Pescara, Insubria University of Varese and Como, University of L’Aquila, Sapienza University of Rome, University of Milan, and University of Milano-Bicocca. The study was first approved for the coordinating centre (University of Verona) by the Ethics Committee for Clinical Research of Verona and Rovigo (protocol 61211 of the 19/09/2018; protocol version 1.5 of the 09/06/2018) and thereafter by the local ethics committee of each recruiting centre. The study protocol was registered in ClinicalTrials.gov (NCT03779789) on 12 December 2018 and published in advance [17].

Participants were included in this study if a treatment with an antidepressant agent was considered clinically indicated. Individuals were eligible if they: (i) were aged 65 years or older, (ii) had a major depressive episode, according to the Diagnostic and Statistical Manual for Mental Disorders–fifth edition (DSM-5) criteria [18], and (iii) were willing to participate by signing an informed consent form. We excluded individuals with a formal diagnosis of dementia, bipolar or schizophrenia-spectrum disorders, as well as those treated with psychopharmacological agents that might influence depressive symptoms at the time of randomization, such as antidepressants, second-generation antipsychotics, and lithium. Moreover, for the specific purpose of the current study, individuals were included only if they had a baseline MADRS score ≥ 20 corresponding to the threshold for a moderate-to-severe depression [19, 20]. People with milder symptoms of depression were thus excluded.

Investigators consecutively enrolled participants from inpatient and outpatient services of each recruiting centre. Participants were randomly allocated to either vortioxetine or an SSRI chosen according to clinical judgment, with an allocation ratio of 1:1. A web-based application was used to manage the randomization process, employing a centralized procedure on the basis of a sequence of treatments randomly permuted in blocks of constant size (random even sizes between 2 and 8). This allocation sequence and the block size were concealed to study investigators. Allocation was stratified by the recruiting centre. For participants allocated to the SSRI group, the clinicians selected the SSRI (sertraline, citalopram, escitalopram, paroxetine, fluoxetine or fluvoxamine) that they deemed most appropriate according to each participant’s characteristics. A flexible dose of either vortioxetine or SSRI within the licensed dose range, as reported by AIFA, was allowed. The ratio between the prescribed daily dose (PDD) and the defined daily dose (DDD), according to the World Health Organization (WHO) ATC/DDD index [21], was used to estimate the equivalent daily doses of both vortioxetine and SSRIs. Drugs could be administered in tablets or drops. After randomization and throughout the trial, participants and clinicians were aware of treatment allocation, while outcome assessors operated under blinded conditions.

We collected clinical data and administered rating scales at baseline and at each follow-up visit, which took place at one (T1), three (T2), and six (T3) months after randomization. The MADRS was used to evaluate depressive symptoms at baseline and follow-up assessments. For the specific purpose of this post-hoc analysis, the primary outcome was the improvement of depressive symptoms as measured by change in MADRS score at follow-up assessments (T1, T2, and T3). Secondary outcomes at the study endpoint (T3) included; (i) clinical response (i.e. number of participants with a reduction in MADRS total score of ≥ 50%), (ii) clinical remission (i.e. number of participants with a MADRS score < 10) [22], (iii) any cause discontinuation (i.e. the number of participants discontinuing the assigned antidepressant), (iv) discontinuation due to side effects at T3, and (v) discontinuation due to inefficacy.

Analyses on the primary outcome (MADRS mean change score), as well as on clinical response and remission, were based on an intention-to-treat (ITT) approach, including all subjects regardless of drug discontinuation. A minimum of 141 participants per treatment arm was required for this post-hoc, exploratory analysis (alpha = 0.05; power = 80%), assuming a 3-point difference in MADRS mean change score between vortioxetine and SSRIs. To address missing data for participants with incomplete MADRS assessments or who discontinued the allocated treatment, the Last Observation Carried Forward (LOCF) method was employed. In addition, to more effectively handle missing data, a repeated-measures mixed-effects model analysis was performed, which accommodates unbalanced data under the assumption that data were missing at random.

In addition, subgroup analyses were carried out excluding individuals who either discontinued the drug or had missing information on discontinuation. For any cause and specific reason (due to side effects or inefficacy) discontinuation, a per-protocol (PP) analysis, excluding participants with missing information on this outcome, was carried out. Continuous variables were expressed as mean ± standard deviation (SD), while categorical variables as frequencies (%). Normality and heteroskedasticity assumptions for continuous data were assessed with Shapiro-Wilk and Levene’s tests, respectively. Continuous variables were compared using unpaired Student’s t-test or Mann–Whitney U test according to data distribution, while categorical variables were compared with chi-squared test or Fisher’s exact test. Moreover, survival analyses, estimating hazard ratio (HR) and related 95% confidence interval (CI), were used to test the differences in time of discontinuation between vortioxetine and SSRIs, using competing-risks regression to assess the occurrence of event discontinuation while taking into account death as a competing risk. Multiple regression analyses were used to test if PDD/DDD might influence endpoint outcomes (MADRS mean change score, response, remission, and discontinuation). The significance level was set at 5% and two-tailed tests were used. Statistical analyses were performed using Stata (StataCorp. 2023. Stata Statistical Software: Release 18. College Station, TX: StataCorp LLC).