Introduction
Major depressive disorder is a psychiatric condition characterised by depressed mood, feelings of hopelessness and lack of interest or motivation (American Psychiatric Association, 2013). Major depressive disorder is associated with an increased risk of suicidal behaviour, reduced quality of life and impaired cognition (Chen and Dilsaver, Reference Chen and Dilsaver1996; Ishak et al., Reference Ishak, Mirocha, James, Tobia, Vilhauer, Fakhry, Pi, Hanson, Nashawati, Peselow and Cohen2015; Kessler et al., Reference Kessler, Borges and Walters1999; Pan et al., Reference Pan, Park, Brietzke, Zuckerman, Rong, Mansur, Fus, Subramaniapillai, Lee and Mcintyre2019; Qin, Reference Qin2011; Saragoussi et al., Reference Saragoussi, Christensen, Hammer-Helmich, Rive, Touya and Haro2018). Major depressive disorder affects approximately 280 million people globally, causing a severe burden on patients and societies (World Health Organization, 2023). The serotonin-norepinephrine reuptake inhibitor, venlafaxine, is approved for the treatment of major depressive disorder in several countries, including the United States, EU countries and the United Kingdom (European Medicines Agency, 2008; National Health Service, 2018; Singh and Saadabadi, Reference Singh and Saadabadi2021). Venlafaxine is often recommended as a second-line treatment for major depressive disorder (Malone, Reference Malone2007; National Health Service, 2018; NHS East and North Hertfordshire Clinical Commissioning Group and NHS Herts Valleys Clinical Commissioning Group, 2018; Singh and Saadabadi, Reference Singh and Saadabadi2021).
Previous reviews have demonstrated that venlafaxine lowers scores on depression rating scales, producing statistically significant results but the relevance to patients remains uncertain. (Cipriani et al., Reference Cipriani, Furukawa, Salanti, Chaimani, Atkinson, Ogawa, Leucht, Ruhe, Turner, Higgins, Egger, Takeshima, Hayasaka, Imai, Shinohara, Tajika, Ioannidis and Geddes2018; Hengartner and Ploderl, Reference Hengartner and Ploderl2022; Jakobsen et al., Reference Jakobsen, Gluud and Kirsch2020; Schueler et al., Reference Schueler, Koesters, Wieseler, Grouven, Kromp, Kerekes, Kreis, Kaiser, Becker and Weinmann2011). Previously, understanding of the adverse effects of venlafaxine has primarily been based on results of non-randomised studies, which may under- or overestimate the occurrence of adverse effects due to confounding and biased reporting (Vandenbroucke, Reference Vandenbroucke2006). Previous reviews have not systematically assessed the risks of all adverse events using data from randomised clinical trials (Cipriani et al., Reference Cipriani, Furukawa, Salanti, Chaimani, Atkinson, Ogawa, Leucht, Ruhe, Turner, Higgins, Egger, Takeshima, Hayasaka, Imai, Shinohara, Tajika, Ioannidis and Geddes2018; Schueler et al., Reference Schueler, Koesters, Wieseler, Grouven, Kromp, Kerekes, Kreis, Kaiser, Becker and Weinmann2011).
Our primary objective was to assess the risks of adverse events with venlafaxine versus placebo or ‘active placebo’ in the treatment of adults with major depressive disorder.
Methods
We report this systematic review based on the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines (Supplementary File 1) (Liberati et al., Reference Liberati, Altman, Tetzlaff, Mulrow, Gotzsche, Ioannidis, Clarke, Devereaux, Kleijnen and Moher2009; Moher et al., Reference Moher, Shamseer, Clarke, Ghersi, Liberati, Petticrew, Shekelle and Stewart2015). The methodology was predefined in our protocol based on the recommendations of the Cochrane Handbook of Systematic Reviews of Interventions (Higgins et al., Reference Higgins, Thomas, Chandler, Cumpston, Li, Page and Welch2022; Jorgensen et al., Reference Jorgensen, Juul, Siddiqui, Horowitz, Moncrieff, Munkholm, Hengartner, Kirsch, Gluud and Jakobsen2023). Our protocol was also preregistered in the PROSPERO database (ID: CRD42022315395). This systematic review is a part of a larger project investigating the beneficial and harmful effects of all antidepressants for major depressive disorder (Jorgensen et al., Reference Jorgensen, Juul, Siddiqui, Barbateskovic, Munkholm, Hengartner, Kirsch, Gluud and Jakobsen2021; Juul et al., Reference Juul, Siddiqui, Barbateskovic, Jorgensen, Hengartner, Kirsch, Gluud and Jakobsen2021b; Kamp et al., Reference Kamp, Petersen, Faltermeier, Juul, Siddiqui, Barbateskovic, Kristensen, Moncrieff, Horowitz, Hengartner, Kirsch, Gluud and Jakobsen2024; Siddiqui et al., Reference Siddiqui, Barbateskovic, Juul, Katakam, Munkholm, Gluud and Jakobsen2021).
Search strategy
An information specialist searched the Cochrane Central Register of Controlled Trials (CENTRAL), Medical Literature Analysis and Retrieval System Online (MEDLINE), Excerpta Medica Database (Embase), Latin American and Caribbean Health Sciences Literature (LILACS), PsycINFO, Science Citation Index Expanded (SCI-EXPANDED), Social Sciences Citation Index (SSCI) and Conference Proceedings Citation Index–Social Science & Humanities (CPCI-SSH) to identify relevant trials. We searched all databases from their inception to 7 March 2024. For a detailed search strategy for all electronic databases, see Supplementary File 2. We searched clinical trial registers and websites of pharmaceutical companies to identify unpublished data. We requested clinical study reports from the U.S. Food and Drug Administration, European Medicines Agency and national medicines agencies (Sharma et al., Reference Sharma, Guski, Freund and Gotzsche2016).
Selection criteria
Our eligibility criteria included randomised clinical trials of adults with a primary diagnosis of major depressive disorder as defined by standardised diagnostic criteria, such as the Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association, 2013) or the International Classification of Diseases (World Health Organization, 2021). Major depressive disorder had to be the primary diagnosis, and we did, therefore, not include trials randomising participants with a non-psychiatric primary diagnosis and comorbid major depressive disorder. If a trial reported data where only a subset of participants was eligible (e.g. a combination of adolescents and adults), we only included those that fulfilled our inclusion criteria, so it required that data could be obtained for that specific group. The experimental intervention was venlafaxine, and the control interventions were placebo, ‘active placebo’ or no intervention.
Data extraction and risk of bias assessment
Three review authors (CBK, SJ and FS) screened abstracts and articles using Covidence to identify relevant trials (Covidence, 2024). Three review authors (JJP, PF and CBK) extracted data and assessed risk of bias based on the Cochrane Risk of Bias tool, version 2 (RoB 2) (Higgins et al., Reference Higgins, Thomas, Chandler, Cumpston, Li, Page and Welch2022; Sterne et al., Reference Sterne, Savovic, Page, Elbers, Blencowe, Boutron, Cates, Cheng, Corbett, Eldridge, Emberson, Hernan, Hopewell, Hrobjartsson, Junqueira, Juni, Kirkham, Lasserson, Li, Mcaleenan, Reeves, Shepperd, Shrier, Stewart, Tilling, White, Whiting and Higgins2019). Screening, data extraction and risk of bias assessments were performed independently by at least two review authors. Any discrepancies were resolved through internal discussion or, if required, through discussion with the last author (JCJ). We emailed all corresponding authors with available contact information to confirm data and risk of bias assessments.
Outcomes and subgroup analyses
The primary outcomes were suicides or suicide attempts, serious adverse events (International Council on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use, 2015) and non-serious adverse events (Jorgensen et al., Reference Jorgensen, Juul, Siddiqui, Horowitz, Moncrieff, Munkholm, Hengartner, Kirsch, Gluud and Jakobsen2023). Exploratory outcomes were depressive symptoms measured on the 17-item Hamilton Depression Rating Scale (HDRS-17), quality of life, all adverse events, suicidal ideation, level of functioning, depressive symptoms measured on the Montgomery–Asberg Depression Rating Scale (MADRS) (Montgomery and Åsberg, Reference Montgomery and Åsberg1979), the Beck’s Depression Inventory (BDI) (Beck et al., Reference Beck, Steer and Brown1996) or HDRS-6 (López-Pina et al., Reference López-Pina, Sánchez-Meca and Rosa-Alcázar2009; Timmerby et al., Reference Timmerby, Andersen, Søndergaard, Østergaard and Bech2017), withdrawal symptoms and proportion of participants that guessed their treatment allocation (Jorgensen et al., Reference Jorgensen, Juul, Siddiqui, Horowitz, Moncrieff, Munkholm, Hengartner, Kirsch, Gluud and Jakobsen2023). Short-term follow-up was defined as the assessment closest to 3 months after randomisation, and long-term follow-up was defined as 6 months or more after randomisation. We also planned several subgroup analyses (Jorgensen et al., Reference Jorgensen, Juul, Siddiqui, Horowitz, Moncrieff, Munkholm, Hengartner, Kirsch, Gluud and Jakobsen2023).
Assessment of statistical and clinical significance
We performed meta-analyses according to the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al., Reference Higgins, Thomas, Chandler, Cumpston, Li, Page and Welch2022), Keus et al. (Reference Keus, Wetterslev, Gluud and Van Laarhoven2010) and the eight-step procedure by Jakobsen et al. (Reference Jakobsen, Wetterslev, Winkel, Lange and Gluud2014). We adjusted the threshold for statistical significance by the number of primary outcomes and therefore used a p-value of 0.025 or less as threshold (Jakobsen et al., Reference Jakobsen, Wetterslev, Winkel, Lange and Gluud2014). We analysed data using the software Stata version 17 (Statacorp, 2019). We used both random-effects (Hartung–Knapp–Sidik–Jonkman) (Inthout et al., Reference Inthout, Ioannidis and Borm2014) and fixed-effect model meta-analyses (Mantel–Haenszel for dichotomous outcomes and inverse variance for continuous outcomes) to assess intervention effects (DeMets, Reference DeMets1987; Higgins et al., Reference Higgins, Thomas, Chandler, Cumpston, Li, Page and Welch2022). We primarily reported the most conservative result (highest p-value), and considered the less conservative result a sensitivity analysis (Jakobsen et al., Reference Jakobsen, Wetterslev, Winkel, Lange and Gluud2014). Trial Sequential Analysis was used to control for random errors by estimating the diversity-adjusted required information size (Brok et al., Reference Brok, Thorlund, Gluud and Wetterslev2008a, Reference Brok, Thorlund, Wetterslev and Gluud2008b; Copenhagen Trial Unit – Centre for Clinical Intervention Research, 2023; Imberger et al., Reference Imberger, Thorlund, Gluud and Wetterslev2016; Thorlund et al., Reference Thorlund, Anema and Mills2010, Reference Thorlund, Devereaux, Wetterslev, Guyatt, Ioannidis, Thabane, Gluud, Als-Nielsen and Gluud2008, Reference Thorlund, Engstrøm, Wetterslev, Brok, Imberger and Gluud2017; Wetterslev et al., Reference Wetterslev, Thorlund, Brok and Gluud2008, Reference Wetterslev, Thorlund, Brok and Gluud2009). We estimated the required information size based on the observed proportion of participants with an outcome in the control group, a relative risk reduction or a relative risk increase of 20%, an alpha of 2.5%, a beta of 10% and the observed diversity. The 20% relative risk reduction or increase was chosen as it is a common effect size when an intervention is beneficial and may be important to patients. We calculated Bayes factor for all primary outcomes. We used Grading Recommendations Assessment Development Evaluation (GRADE) to assess the certainty of evidence for the primary outcomes (Guyatt et al., Reference Guyatt, Oxman, Schünemann, Tugwell and Knottnerus2011, Reference Guyatt, Oxman, Vist, Kunz, Falck-Ytter, Alonso-Coello and Schünemann2008; Schünemann et al., Reference Schünemann, Best, Vist and Oxman2003). Please see the published protocol for a detailed description of the statistical analysis plan (Jorgensen et al., Reference Jorgensen, Juul, Siddiqui, Horowitz, Moncrieff, Munkholm, Hengartner, Kirsch, Gluud and Jakobsen2023).
Results
A total of 28 trials randomising 6,253 participants were included (Alvarez et al., Reference Alvarez, Perez, Dragheim, Loft and Artigas2012; Cunningham, Reference Cunningham1997; Cunningham et al., Reference Cunningham, Borison, Carman, Chouinard, Crowder, Diamond, Fischer and Hearst1994; Guelfi et al., Reference Guelfi, White, Hackett, Guichoux and Magni1995; Hewett et al., Reference Hewett, Chrzanowski, Schmitz, Savela, Milanova, Gee, Krishen, Millen, Leary and Modell2009, Reference Hewett, Gee, Krishen, Wunderlich, Le Clus, Evoniuk and Modell2010; Higuchi et al., Reference Higuchi, Kamijima, Nakagome, Itamura, Asami, Kuribayashi and Imaeda2016; Hopkins et al., Reference Hopkins, Reasner and Koblan2013; Hunter et al., Reference Hunter, Muthen, Cook and Leuchter2010; Khan et al., Reference Khan, Fabre and Rudolph1991, Reference Khan, Upton, Rudolph, Entsuah and Leventer1998; Learned et al., Reference Learned, Graff, Roychowdhury, Moate, Krishnan, Archer, Modell, Alexander, Zamuner, Lavergne, Evoniuk and Ratti2012; Lieberman et al., Reference Lieberman, Montgomery, Tourian, Brisard, Rosas, Padmanabhan, Germain and Pitrosky2008; Luthringer et al., Reference Luthringer, Toussaint, Schaltenbrand, Bailey, Danjou, Hackett, Guichoux and Macher1996; Mendels et al., Reference Mendels, Johnston, Mattes and Riesenberg1993; Nemeroff et al., Reference Nemeroff, Thase and Group2007; Rudolph et al., Reference Rudolph, Fabre, Feighner, Rickels, Entsuah and Derivan1998; Rudolph and Feiger, Reference Rudolph and Feiger1999; Schatzberg and Roose, Reference Schatzberg and Roose2006; Schweizer et al., Reference Schweizer, Feighner, Mandos and Rickels1994, Reference Schweizer, Weise, Clary, Fox and Rickels1991; Sharma et al., Reference Sharma, Guski, Freund and Gotzsche2016; Sheehan et al., Reference Sheehan, Nemeroff, Thase, Entsuah and Group2009; Silverstone and Ravindran, Reference Silverstone and Ravindran1999; Thase, Reference Thase1997; Wyeth-Ayerst, 1991a, 1991b, 1992, 1991c; Wyeth-Ayerst Research, 1996) (Fig. 1). We identified unpublished data through clinical study reports for 11 trials (Alvarez et al., Reference Alvarez, Perez, Dragheim, Loft and Artigas2012; Eudract 2004-000562-13, 2016; Eudract Number 2007-007025-51, 2010; Hewett et al., Reference Hewett, Chrzanowski, Schmitz, Savela, Milanova, Gee, Krishen, Millen, Leary and Modell2009; Learned et al., Reference Learned, Graff, Roychowdhury, Moate, Krishnan, Archer, Modell, Alexander, Zamuner, Lavergne, Evoniuk and Ratti2012; Sharma et al., Reference Sharma, Guski, Freund and Gotzsche2016; Wyeth-Ayerst, 1991a, 1991b, 1992, 1991c; Wyeth-Ayerst Research, 1996). Most trials included both men and women aged 18 years or older with a primary diagnosis of major depressive disorder (Supplementary Table 1). One trial only included participants aged 65 years or older (Schatzberg and Roose, Reference Schatzberg and Roose2006). The mean HDRS entry scores ranged from 22.4 to 30.6 (Supplementary Table 1). All trials involved an inert placebo control, i.e. we did not identify any trials using an ‘active placebo’ or no intervention as control interventions. All trials assessed outcomes at a maximum of 12 weeks after randomisation. Most trials did not adequately report the proportion of participants with missing data at follow-up, and it was, therefore, not possible to perform ‘best-worst/worst-best’ sensitivity analyses. All trials were assessed at overall high risk of bias, particularly due to lack of information, missing data, lack of information on blinding, risk of unblinding due to adverse events or other noticeable changes, inappropriate analysis methods and poor reporting (Supplementary Fig. 1). Three trials did not report any relevant data (Hopkins et al., Reference Hopkins, Reasner and Koblan2013; Hunter et al., Reference Hunter, Muthen, Cook and Leuchter2010).
Primary outcomes
Suicides or suicide attempts
Only 7 of 28 trials reported results on suicides or suicide attempts (Hewett et al., Reference Hewett, Chrzanowski, Schmitz, Savela, Milanova, Gee, Krishen, Millen, Leary and Modell2009; Higuchi et al., Reference Higuchi, Kamijima, Nakagome, Itamura, Asami, Kuribayashi and Imaeda2016; Sharma et al., Reference Sharma, Guski, Freund and Gotzsche2016; Sheehan et al., Reference Sheehan, Nemeroff, Thase, Entsuah and Group2009; Wyeth-Ayerst, 1991b, 1992; Wyeth-Ayerst Research, 1996). All trials only assessed outcomes at the end of the treatment period, i.e. 6–8 weeks after randomisation. A total of 6/1,127 (0.5%) experimental participants attempted or committed suicide compared with 6/780 (0.8%) control participants. Meta-analysis showed no evidence of a difference between venlafaxine versus placebo on suicides or suicide attempts (odds ratio: 0.65; 95% confidence interval (CI): 0.25–1.71; p = 0.38; 7 trials; Bayes factor: 0.74) (Fig. 2). Visual inspection of the forest plot and statistical tests (I 2 = 0.0%) indicated no clear signs of heterogeneity. Trial Sequential Analysis showed that we did not have enough information to confirm or reject the hypothesis that venlafaxine influenced the risk of suicides or suicide attempts with a relative risk reduction of 20% (no graph produced). This outcome result was assessed as overall high risk of bias, and the certainty of the evidence was very low (Supplementary Tables 2–3).
Tests of interaction comparing the effects of using a placebo washout period prior to randomisation showed no evidence of a difference (p = 1.00) (Supplementary Fig. 2). The remaining predefined subgroup analyses could not be performed due to a lack of relevant data.
Serious adverse events
Twenty-two trials reported results on serious adverse events (Alvarez et al., Reference Alvarez, Perez, Dragheim, Loft and Artigas2012; Cunningham, Reference Cunningham1997; Eudract 2004-000562-13, 2016; Eudract Number 2007-007025-51, 2010; Hewett et al., Reference Hewett, Chrzanowski, Schmitz, Savela, Milanova, Gee, Krishen, Millen, Leary and Modell2009, Reference Hewett, Gee, Krishen, Wunderlich, Le Clus, Evoniuk and Modell2010; Higuchi et al., Reference Higuchi, Kamijima, Nakagome, Itamura, Asami, Kuribayashi and Imaeda2016; Learned et al., Reference Learned, Graff, Roychowdhury, Moate, Krishnan, Archer, Modell, Alexander, Zamuner, Lavergne, Evoniuk and Ratti2012; Lieberman et al., Reference Lieberman, Montgomery, Tourian, Brisard, Rosas, Padmanabhan, Germain and Pitrosky2008; Mendels et al., Reference Mendels, Johnston, Mattes and Riesenberg1993; Nemeroff et al., Reference Nemeroff, Thase and Group2007; Rudolph et al., Reference Rudolph, Fabre, Feighner, Rickels, Entsuah and Derivan1998; Rudolph and Feiger, Reference Rudolph and Feiger1999; Schatzberg and Roose, Reference Schatzberg and Roose2006; Sharma et al., Reference Sharma, Guski, Freund and Gotzsche2016; Sheehan et al., Reference Sheehan, Nemeroff, Thase, Entsuah and Group2009; Silverstone and Ravindran, Reference Silverstone and Ravindran1999; Thase, Reference Thase1997; Wyeth-Ayerst, 1991a, 1991b, 1992, 1991c; Wyeth-Ayerst Research, 1996). All trials only assessed outcomes at the end of the treatment period, i.e. 4–12 weeks after randomisation. A total of 224/3,164 (7.1%) experimental participants had one or more serious adverse event compared with 58/2,362 (2.5%) control participants. Meta-analysis showed evidence of a harmful effect of venlafaxine versus placebo on serious adverse events (risk ratio (RR): 2.66; 95% CI: 1.67–4.25; p < 0.01; 22 trials; Bayes factor: 0.06) (Fig. 3). Visual inspection of the forest plot and statistical tests (I 2 = 52.0%) indicated heterogeneity that could not be resolved. Trial Sequential Analysis showed that we did not have enough information to confirm or reject the hypothesis that venlafaxine influenced the risk of serious adverse events with a relative risk reduction or increase of 20% (Supplementary Fig. 3). This outcome result was assessed as overall high risk of bias, and the certainty of the evidence was very low (Supplementary Tables 2–3).
Tests of interaction comparing the effects of using a placebo washout period prior to randomisation showed no evidence of a difference (p = 0.87) (Supplementary Fig. 4). The remaining predefined subgroup analyses could not be performed due to a lack of relevant data.
When each specific serious adverse event was analysed separately, 2/11 meta-analyses showed evidence of a harmful effect of venlafaxine versus placebo: sexual dysfunction (RR: 6.49; 95% CI: 3.02–13.93; p < 0.01; I 2 = 1.9%; 8 trials; number needed to harm (NNH): 12) (Supplementary Fig. 5) and anorexia (RR: 3.23; 95% CI: 1.75–5.97; p < 0.01; I 2 = 44.7%; 9 trials; NNH: 14) (Supplementary Fig. 6). The remaining meta-analyses showed no evidence of differences (Supplementary Tables 4–5, Supplementary Fig. 7–15).
Non-serious adverse events
Twenty-four trials reported results on non-serious adverse events (Alvarez et al., Reference Alvarez, Perez, Dragheim, Loft and Artigas2012; Cunningham, Reference Cunningham1997; Eudract 2004-000562-13, 2016; Eudract Number 2007-007025-51, 2010; Guelfi et al., Reference Guelfi, White, Hackett, Guichoux and Magni1995; Hewett et al., Reference Hewett, Chrzanowski, Schmitz, Savela, Milanova, Gee, Krishen, Millen, Leary and Modell2009, Reference Hewett, Gee, Krishen, Wunderlich, Le Clus, Evoniuk and Modell2010; Higuchi et al., Reference Higuchi, Kamijima, Nakagome, Itamura, Asami, Kuribayashi and Imaeda2016; Khan et al., Reference Khan, Fabre and Rudolph1991, Reference Khan, Upton, Rudolph, Entsuah and Leventer1998; Learned et al., Reference Learned, Graff, Roychowdhury, Moate, Krishnan, Archer, Modell, Alexander, Zamuner, Lavergne, Evoniuk and Ratti2012; Lieberman et al., Reference Lieberman, Montgomery, Tourian, Brisard, Rosas, Padmanabhan, Germain and Pitrosky2008; Nemeroff et al., Reference Nemeroff, Thase and Group2007; Rudolph et al., Reference Rudolph, Fabre, Feighner, Rickels, Entsuah and Derivan1998; Rudolph and Feiger, Reference Rudolph and Feiger1999; Schatzberg and Roose, Reference Schatzberg and Roose2006; Schweizer et al., Reference Schweizer, Feighner, Mandos and Rickels1994, Reference Schweizer, Weise, Clary, Fox and Rickels1991; Sheehan et al., Reference Sheehan, Nemeroff, Thase, Entsuah and Group2009; Silverstone and Ravindran, Reference Silverstone and Ravindran1999; Thase, Reference Thase1997; Wyeth-Ayerst, 1991b, 1992, 1991c; Wyeth-Ayerst Research, 1996). All trials only assessed outcomes at the end of the treatment period, i.e. 4–12 weeks after randomisation. A total of 1,804/3,127 (57.7%) experimental participants had one or more non-serious adverse events compared with 1,111/2,356 (47.2%) control participants. Meta-analysis showed evidence of a harmful effect of venlafaxine versus placebo on non-serious adverse events (RR: 1.43; 95% CI: 1.21–1.69; p < 0.01; 24 trials; Bayes factor: 0.001) (Fig. 4). Visual inspection of the forest plot and statistical tests (I 2 = 92.9%) indicated heterogeneity that could not be resolved. Trial Sequential Analysis showed that we did not have enough information to confirm or reject the hypothesis that venlafaxine influenced the risk of non-serious adverse events with a relative risk reduction or increase of 20% (Supplementary Fig. 16). This outcome result was assessed as overall high risk of bias, and the certainty of the evidence was very low (Supplementary Tables 2–3).
Tests of interaction comparing the effects of using a placebo washout period prior to randomisation showed evidence of a difference (p < 0.01) (Supplementary Fig. 17). When the subgroup of trials with a placebo washout period was analysed separately, meta-analysis showed evidence of a harmful effect of venlafaxine (RR: 1.63; 95% CI: 1.30–2.05; p < 0.01; 15 trials). When the subgroup of trials without a placebo washout period was analysed separately, meta-analysis showed evidence of a harmful effect (RR: 1.13; 95% CI: 1.04–1.23; p < 0.01; 9 trials). The remaining predefined subgroup analyses could not be performed due to a lack of relevant data.
When each specific non-serious adverse event was analysed separately, 11/47 meta-analyses showed evidence of a harmful effect of venlafaxine versus placebo: nausea (RR: 2.72; 95% CI: 2.26–3.28; p < 0.01; I 2 = 46.4%; 23 trials; NNH: 5) (Supplementary Fig. 18), dry mouth (RR: 2.16; 95% CI: 1.71–2.74; p < 0.01; I 2 = 40.7%; 21 trials; NNH: 10) (Supplementary Fig. 19), dizziness (RR: 2.49; 95% CI: 1.90–3.26; p < 0.01; I 2 = 37.9%; 20 trials; NNH: 11) (Supplementary Fig. 20), sweating (RR: 3.99; 95% CI: 2.88–5.54; p < 0.01; I 2 = 20.5%; 20 trials; NNH: 11) (Supplementary Fig. 21), somnolence (RR: 2.23; 95% CI: 1.78–2.78; p < 0.01; I2 = 16.9%; 18 trials; NNH: 11) (Supplementary Fig. 22), constipation (RR: 2.24; 95% CI: 1.64–3.04; p < 0.01; I 2 = 38.3%; 18 trials; NNH: 14) (Supplementary Fig. 23), nervousness (RR: 2.20; 95% CI: 1.43–3.40; p < 0.01; I 2 = 33.4%; 11 trials; NNH: 15) (Supplementary Fig. 24), insomnia (RR: 1.73; 95% CI: 1.37–2.19; p < 0.01; I 2 = 26.9%; 19 trials; NNH: 19) (Supplementary Fig. 25), asthenia (RR: 1.78; 95% CI: 1.30–2.43; p < 0.01; I 2 = 19.7%; 16 trials; NNH: 27) (Supplementary Fig. 26), tremor (RR: 2.30; 95% CI: 1.22–4.32; p = 0.01; I 2 = 37.0%; 11 trials; NNH: 29) (Supplementary Fig. 27) and decreased appetite (RR: 2.52; 95% CI: 1.04–6.09; p < 0.01; I 2 = 1.0%; 3 trials; NNH: 47) (Supplementary Fig. 28). The remaining meta-analyses showed no evidence of differences (Supplementary Table 6, Supplementary Fig. 29–64).
Exploratory outcomes and sensitivity analyses
HDRS-17
Only two trials reported results on HDRS-17 (Higuchi et al., Reference Higuchi, Kamijima, Nakagome, Itamura, Asami, Kuribayashi and Imaeda2016; Learned et al., Reference Learned, Graff, Roychowdhury, Moate, Krishnan, Archer, Modell, Alexander, Zamuner, Lavergne, Evoniuk and Ratti2012). Both trials only assessed outcomes at the end of the treatment period, i.e. 8–10 weeks after randomisation. Meta-analysis showed evidence of a beneficial effect of venlafaxine (mean difference (MD): −1.50 points; 95% CI: −2.48 to −0.53; p < 0.01; 2 trials) (Supplementary Fig. 65), however, the effect size was below proposed minimal important differences (Jakobsen et al., Reference Jakobsen, Gluud and Kirsch2020). Visual inspection of the forest plot and statistical tests (I 2 = 2.5%) indicated no clear signs of heterogeneity. This outcome result was assessed as overall high risk of bias.
Suicidal ideation
Four trials reported results on suicidal ideation (Hewett et al., Reference Hewett, Gee, Krishen, Wunderlich, Le Clus, Evoniuk and Modell2010; Higuchi et al., Reference Higuchi, Kamijima, Nakagome, Itamura, Asami, Kuribayashi and Imaeda2016; Sheehan et al., Reference Sheehan, Nemeroff, Thase, Entsuah and Group2009; Wyeth-Ayerst Research, 1996). All trials only assessed outcomes at the end of the treatment period, i.e. 6–8 weeks after randomisation. A total of 91/812 (11.2%) experimental participants had suicidal ideation compared with 42/548 (7.7%) control participants. Meta-analysis showed no evidence of a difference between venlafaxine and placebo (RR: 1.13; 95% CI: 0.74–1.73; p = 0.58; 4 trials) (Supplementary Fig. 66). Visual inspection of the forest plot and statistical tests (I 2 = 15.8%) indicated no clear signs of heterogeneity. This outcome result was assessed as overall high risk of bias.
MADRS, BDI and HDRS-6
Nine trials reported results on MADRS (Alvarez et al., Reference Alvarez, Perez, Dragheim, Loft and Artigas2012; Guelfi et al., Reference Guelfi, White, Hackett, Guichoux and Magni1995; Hewett et al., Reference Hewett, Gee, Krishen, Wunderlich, Le Clus, Evoniuk and Modell2010; Higuchi et al., Reference Higuchi, Kamijima, Nakagome, Itamura, Asami, Kuribayashi and Imaeda2016; Khan et al., Reference Khan, Fabre and Rudolph1991; Mendels et al., Reference Mendels, Johnston, Mattes and Riesenberg1993; Schweizer et al., Reference Schweizer, Feighner, Mandos and Rickels1994; Sheehan et al., Reference Sheehan, Nemeroff, Thase, Entsuah and Group2009; Thase, Reference Thase1997; Wyeth-Ayerst Research, 1996). No trials reported results on BDI or HDRS-6. All trials only assessed outcomes at the end of the treatment period, i.e. 4–8 weeks after randomisation. Meta-analysis showed evidence of a beneficial effect of venlafaxine versus placebo (MD: −4.03 MADRS points; 95% CI: −5.30 to −2.75; p < 0.01; 9 trials) (Supplementary Fig. 67), however, the effect size was below proposed minimal important differences (Leucht et al., Reference Leucht, Fennema, Engel, Kaspers-Janssen, Lepping and Szegedi2017). Visual inspection of the forest plot and statistical tests (I 2 = 60.0%) indicated heterogeneity. This outcome result was assessed as overall high risk of bias.
Remaining results
It was not possible to analyse the remaining exploratory outcomes due to a lack of relevant data. We performed all meta-analyses as both fixed-effect and random-effects meta-analyses and reported the most conservative results as the main results. For the less conservative results, please see Supplementary Fig. 68–131.
Discussion
We conducted a systematic review assessing the risks of adverse events with venlafaxine for adults with major depressive disorder. We included 28 trials randomising 6,253 participants to venlafaxine versus placebo. All results were at high risk of bias, and the certainty of the evidence was very low. Data were limited to a maximum of 12 weeks after randomisation. Meta-analysis and Trial Sequential Analysis showed insufficient information to assess the effects of venlafaxine on risks of suicides or suicide attempts. Meta-analysis showed evidence of a harmful effect of venlafaxine on serious adverse events, mainly due to higher risks of sexual dysfunction and anorexia. Meta-analysis showed that venlafaxine increased the risk of several non-serious adverse events: nausea, dry mouth, dizziness, sweating, somnolence, constipation, nervousness, insomnia, asthenia, tremor and decreased appetite. Our results contribute important information on the risks of adverse events, since previous reviews have not systematically assessed all adverse effects (Cipriani et al., Reference Cipriani, Furukawa, Salanti, Chaimani, Atkinson, Ogawa, Leucht, Ruhe, Turner, Higgins, Egger, Takeshima, Hayasaka, Imai, Shinohara, Tajika, Ioannidis and Geddes2018; Schueler et al., Reference Schueler, Koesters, Wieseler, Grouven, Kromp, Kerekes, Kreis, Kaiser, Becker and Weinmann2011). We confirmed the previously shown statistically significant effects of antidepressants on depressive symptom rating scales (Cipriani et al., Reference Cipriani, Furukawa, Salanti, Chaimani, Atkinson, Ogawa, Leucht, Ruhe, Turner, Higgins, Egger, Takeshima, Hayasaka, Imai, Shinohara, Tajika, Ioannidis and Geddes2018; Hengartner and Ploderl, Reference Hengartner and Ploderl2022; Jakobsen et al., Reference Jakobsen, Gluud and Kirsch2020; Kamp et al., Reference Kamp, Petersen, Faltermeier, Juul, Siddiqui, Barbateskovic, Kristensen, Moncrieff, Horowitz, Hengartner, Kirsch, Gluud and Jakobsen2024; Schueler et al., Reference Schueler, Koesters, Wieseler, Grouven, Kromp, Kerekes, Kreis, Kaiser, Becker and Weinmann2011), but with effect sizes below proposed minimal important differences (3 HDRS points and 7–9 MADRS points) (Jakobsen et al., Reference Jakobsen, Gluud and Kirsch2020; Leucht et al., Reference Leucht, Fennema, Engel, Kaspers-Janssen, Lepping and Szegedi2017).
Our systematic review has several strengths. Previously, understanding of the adverse effects of venlafaxine has primarily been based on results of non-randomised studies. Accordingly, this is the first systematic review to assess all adverse effects of venlafaxine in adults with major depressive disorder. Data on adverse effects are essential for enabling patients and clinicians to make informed decisions about the use of any treatment. Our predefined methodology was based on the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al., Reference Higgins, Thomas, Chandler, Cumpston, Li, Page and Welch2022), PRISMA (Liberati et al., Reference Liberati, Altman, Tetzlaff, Mulrow, Gotzsche, Ioannidis, Clarke, Devereaux, Kleijnen and Moher2009), Trial Sequential Analysis (Copenhagen Trial Unit – Centre for Clinical Intervention Research, 2023; Thorlund et al., Reference Thorlund, Engstrøm, Wetterslev, Brok, Imberger and Gluud2017), the eight-step procedure by Jakobsen et al. (Reference Jakobsen, Wetterslev, Winkel, Lange and Gluud2014b), the GRADE approach (Guyatt et al., Reference Guyatt, Oxman, Schünemann, Tugwell and Knottnerus2011), and risks of systematic and random errors, external validity, publication bias and heterogeneity were taken into account. Furthermore, unpublished data were included in the analyses to increase the validity of our results (Golder et al., Reference Golder, Loke, Wright and Norman2016; Higgins et al., Reference Higgins, Thomas, Chandler, Cumpston, Li, Page and Welch2022; Maund et al., Reference Maund, Tendal, Hróbjartsson, Jørgensen, Lundh, Schroll and Gøtzsche2014; Tang et al., Reference Tang, Ravaud, Riveros, Perrodeau and Dechartres2015).
Our systematic review also has limitations. First, the included trials only reported results at the end of treatment at a maximum of 12 weeks, so the long-term effects of venlafaxine are unknown. There is a need for trials with long-term follow-up to assess the benefits and harms since, for example, half of patients on antidepressants in the United Kingdom and 70% of patients in the USA have used them for more than 2 years (Johnson et al., Reference Johnson, Macdonald, Atkinson, Buchanan, Downes and Dougall2012; Mojtabai and Olfson, Reference Mojtabai and Olfson2014). This is particularly pertinent for medications that are associated with tolerance and withdrawal effects, which tend to show diminishing beneficial effects over time (Kinrys et al., Reference Kinrys, Gold, Pisano, Freeman, Papakostas, Mischoulon, Nierenberg and Fava2019). Second, all included trials were assessed at overall high risk of bias mainly driven by risks of bias due to missing data, lack of information on blinding (Juul et al., Reference Juul, Gluud, Simonsen, Frandsen, Kirsch and Jakobsen2021a), risk of unblinding due to adverse events, inappropriate analysis methods and poor reporting. The reporting and assessment of adverse events were especially inadequate, and a major limitation is that data on adverse events are rarely collected systematically, even in randomised trials. It usually depends on spontaneous reporting, and our analyses are likely to underestimate serious adverse event incidences. Third, the certainty of the evidence was very low for all outcome results. It is, therefore, likely that trials conducted with higher methodological quality will show different results and possibly reveal more severe adverse events. Fourth, only six of the included trials reported on suicides or suicide attempts, and there was not enough information to confirm or reject the effects of venlafaxine on suicides or suicide attempts. This is particularly problematic since major depressive disorder is associated with increased risks of suicidal behaviour (Chen and Dilsaver, Reference Chen and Dilsaver1996; Kessler et al., Reference Kessler, Borges and Walters1999; Qin, Reference Qin2011), and because antidepressants have been linked with an increased risk of suicidal behaviour, especially in young people (Sharma et al., Reference Sharma, Guski, Freund and Gotzsche2016). There is a need for larger trials at low risk of bias to assess the risks of suicides and suicide attempts. Fifth, only eight trials had publicly available protocols or trial registrations, making it difficult to assess risks of bias, particularly whether selective reporting had occurred. Sixth, we included three primary outcomes, which increased the risk of type I errors. To control the risks of random errors, we adjusted our threshold for significance according to the number of primary outcomes, but we did not adjust the thresholds for significance according to the total number of comparisons, including exploratory outcomes, subgroup analyses and sensitivity analyses. Seventh, trials comparing antidepressants with ‘active placebo’ indicate that the beneficial effects may be inflated due to the unblinding effects of the drug when compared with an inert placebo (Moncrieff et al., Reference Moncrieff, Wessely and Hardy2004). We did not identify any trials comparing venlafaxine to ‘active placebo’, which leaves open the possibility that efficacy results are impacted by unblinding caused by adverse effects or other noticeable changes produced by medication, as shown in trials of other antidepressants (Jureidini et al., Reference Jureidini, Moncrieff, Klau, Aboustate and Raven2024). Lastly, withdrawal effects of venlafaxine were only assessed in two of the included trials (Wyeth-Ayerst, 1991b, 1991c), emphasising the lack of data on a harm of treatment that has increasingly been highlighted (Davies and Read, Reference Davies and Read2019) and is particularly relevant to venlafaxine, given its short half-life (Campagne, Reference Campagne2005; Gastaldon et al., Reference Gastaldon, Schoretsanitis, Arzenton, Raschi, Papola, Ostuzzi, Moretti, Seifritz, Kane, Trifiro and Barbui2022). These limitations need to be considered when interpreting our results.
Conclusions
Short-term results show that venlafaxine has uncertain effects on the risks of suicides but increases the risks of serious adverse events (especially sexual dysfunction and anorexia) and many non-serious adverse events. The long-term effects of venlafaxine for major depressive disorder are unknown. More information on adverse effects is critical if patients and clinicians should make informed decisions about the costs and benefits of using venlafaxine. It is a particular cause for concern that there are no data on the long-term adverse effects of venlafaxine given that so many people use these drugs for several years.
Differences between the protocol and the review
There were no differences between the protocol and the review.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S2045796024000520.
Availability of data and materials
All data generated or analysed during this study are included in this article and its supplementary material files.
Author contributions
CBK, SJ, FS, JM, MAH, MPH, IK, CG and JCJ contributed to the conceptualisation and design of the systematic review. CBK, SF and FS screened studies for inclusion. JJP, PF and CBK extracted data. CBK analysed data. CBK and JCJ wrote the original draft. All authors commented and approved the final manuscript. CBK and JCJ are the guarantors. The guarantors had full access to all the data in the systematic review, take responsibility for the integrity of the data and the accuracy of the data analyses and had final responsibility for the decision to submit for publication. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.
Financial support
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Competing interests
We received no support from any organisation for the submitted work. MAH is a co-applicant and member of the DSMB of the RELEASE trial in Australia, funded by the Medical Research Future Fund (MRFF). MAH is co-founder and consultant for Outro Health, a digital clinic helping patients to stop unnecessary antidepressant medication. MAH has been paid honoraria by several NHS Trusts for grand rounds presentations, and by Salomon’s University, the University of Arizona and the University of Washington. JM is a co-investigator on REDUCE (programme grant studying discontinuation of antidepressants) and Chief Investigator on RADAR (programme grant to explore antipsychotic reduction and discontinuation). JM has been paid honoraria by University of Basel, Alberta Psychiatric Association and Case Western University. JM receives royalties from Palgrave Macmillan and PCCS Books for three books about psychiatric drugs. JM is a co-chair person (unfunded position) of Critical Psychiatry Network. MPH receives royalties from Palgrave Macmillan for a book about antidepressants. We report no other relationships or activities that could appear to have influenced the submitted work.
Ethical standards
Ethics approval and written informed consent were not required for this study.