Study characteristics

As shown in Fig. 1, from 387 records screened on title and abstract, 20 full texts were analysed. One additional RCT was found on clinicaltrials.gov but this was an incomplete trial due to inability to reach the target sample size due to COVID-19, ⁴⁰ therefore this was not included in our final selection. The review process led to the selection of 14 studies ^{Reference Jahanbakhsh, Manteghi, Emami, Mahyari, Gholampour and Mohammadpour17,Reference Andrade, Aswath, Chaturvedi, Srinivasa and Raguram41–Reference Pandit, Srivastav, Sur, Chaudhuri, Wang and Biswas53} referring to 14 independent samples, corresponding to a total of 713 people (of which 360 treated with Withania somnifera and 353 controls), which were included in the quantitative synthesis (see also Supplementary Table 2 reporting the list of studies excluded after full-text review, along with the reason for exclusion).

Fig. 1 Preferred Reporting Items for Systematic Reviews and Meta-analyses flow diagram.

The trials were conducted in 4 countries: India (n = 8; 57.2%), Iran (n = 3; 21.4%), USA (n = 2; 14.3%) and Canada (n = 1; 7.1%). The overall percentage of females across the studies was 45.7%, mean age was 38.3 (s.d.: 10.9) years, with only one study (7.1%) involving underage participants. The most common diagnostic category was anxiety disorders (n = 5; 35.7%), followed by insomnia and chronic stress (each n = 2; 14.3%), major depressive disorder (MDD), bipolar disorder, schizophrenia and schizo-affective disorder, attention-deficit hyperactivity disorder (ADHD) and OCD, each n = 1 (7.1%). The clinical severity at the baseline was symptomatic or mild/moderate for most of the studies, except for one study performed on participants with moderate/high severity anxiety and one on participants with bipolar disorder in clinical remission. In 13 trials (92.9%) Withania somnifera was administered orally with daily dosing ranging from 10 to 12 000 mg (median 600 mg). One trial (7.1%) administered Withania somnifera with Shirodhara (oil dripping) therapy. Withanolides minimum concentration in the Withania somnifera extracts ranged from 1.5 to 8%, although this information was not reported by six studies. Withania somnifera was administered as add-on treatment in seven trials (50.0%), as the sole therapeutic agent in four trials (28.6%) and three trials (21.4%) did not report information on concurrent medication. For the studies administering Withania somnifera as add-on treatment, the concurrent treatment consisted of selective serotonin reuptake inhibitor (SSRI) antidepressants in two studies, mood stabilisers, antipsychotics, anti-ADHD medications without other specification and phytotherapeutic regimen with Nardostachys jatamansi and Lavandula stoechas, each n = 1. Most of the trials used placebo as the comparator (n = 12; 85.7%), one study (7.1%) used a combination of placebo and cognitive behavioural therapy (CBT) and one study (7.1%) considered the ongoing phytotherapeutic regimen alone as the control group. The median trial’s duration was 8 weeks, ranging from 6 to 12 weeks. The outcomes reported by the trials were: anxiety (n = 10), depression (n = 5), sleep (n = 4), stress (n = 3) and mania, psychotic symptoms and OCD symptoms (each n = 1). The main characteristics of the studies included in the review are summarised in Table 1 (see also Supplementary Table 3 for information on the composition of Withania somnifera extracts used in each study and Supplementary Table 4 for details on the outcome measures used in each study).

Table 1 Characteristics of the included studies

ADHD, attention-deficit hyperactivity disorder; AP, antipsychotics; C, control; CBT, cognitive–behavioural therapy; GAD, generalised anxiety disorder; MDD, major depressive disorder; mg, milligram; N, number; NA, not applicable; NOS, not otherwise specified; NR, information not reported; OCD, obsessive–compulsive disorder; PBO, placebo; po, per os (by mouth/orally); RCT, randomised controlled trial; SSRI, selective serotonin reuptake inhibitor; SZ, schizophrenia; T, treatment; WS, Withania somnifera.

Effect of Withania somnifera on anxiety symptoms

The meta-analysis of the effect of Withania somnifera on anxiety symptoms included ten studies and yielded significant evidence supporting the efficacy of the treatment (SMD: −1.62 (95% CI: −2.66; −0.57)), although with evidence of high between-study heterogeneity (I ²: 94%; p-value <0.001). Subgroup meta-analysis by diagnostic category revealed a significant between-group difference (p-value: 0.008); however, the pooled estimates indicated treatment effectiveness across most conditions, except for the single study conducted in individuals with bipolar disorder, which did not show a significant effect. The results are summarised in Fig. 2.

Fig. 2 Forest plot of anxiety among Withania somnifera and control groups: overall and by diagnostic subgroup. ADHD, attention-deficit and hyperactivity disorder; BD, bipolar disorder; IV, inverse variance; WS, Withania somnifera. The measures of anxiety across the studies were scaled as lower scores indicate lower anxiety symptoms. Reference to the included studies (in alphabetical order): Andrade et al, 2000 ^{Reference Andrade, Aswath, Chaturvedi, Srinivasa and Raguram41} ; Chengappa et al, 2013 ^{Reference Chengappa, Bowie, Schlicht, Fleet, Brar and Jindal42} ; Cooley et al, 2009 ^{Reference Cooley, Szczurko, Perri, Mills, Bernhardt and Zhou45} ; Fuladi et al, 2021 ^{Reference Fuladi, Emami, Mohammadpour, Karimani, Manteghi and Sahebkar46} ; Hosseini et al, 2019 ^{Reference Hosseini, Moharari, Soltanifar, Mohammad Poor, Hosseini and Zahedian48} ; Khyati et al, 2013 ^{Reference Khyati and Anup49} ; Langade et al, 2019 ^{Reference Langade, Kanchi, Salve, Debnath and Ambegaokar50} ; Langade et al 2021 ^{Reference Langade, Thakare, Kanchi and Kelgane51} ; Majeed et al, 2023 ^{Reference Majeed, Nagabhushanam, Murali, Vishwanathan, Mamidala and Mundkur52} ; Pandit et al, 2024. ^{Reference Pandit, Srivastav, Sur, Chaudhuri, Wang and Biswas53}

Our investigation found evidence of publication bias. Notably, most studies reported outcomes favouring Withania somnifera, one showed a null effect and none reported negative results. This can be seen in the funnel plot displayed in Supplementary Figure 1 and was confirmed by Egger’s test p-value <0.001. However, the estimate of the effect size obtained by inputting missing studies into the funnel plot until symmetry is reached using Duval and Tweedie’s trim-and-fill procedure was highly similar (SMD: −1.62 (95% CI: −2.67; −0.57)).

To investigate the source of heterogeneity, leave-one-out analysis and meta-regression were performed.

Leave-one-out analysis, in which the meta-analysis was serially repeated after the exclusion of each study, found that by excluding the study from Khyati et al ^{Reference Khyati and Anup49} the effect size decreased more than 25%, remaining statistically significant (SMD: −1.13 (95% CI: −1.65; −0.60); I ²: 85%). Interestingly, this study used the highest dose of Withania somnifera, which was more than ten times higher than the other trials. Instead, excluding other studies produced overall inconsequential changes (≤15%) in both the pooled estimate and heterogeneity. The results of leave-one-out analysis are available in Supplementary Table 5.

Meta-regression analyses were performed on the following variables, potentially associated with heterogeneity: (a) the mean age of participants; (b) the percentage of females in the total sample; (c) the country where the study was conducted; (d) Withania somnifera dose administered; (e) an interaction term of Withania somnifera dose × withanolides % in the extract used; (f) trial duration and (g) Withania somnifera use as monotherapy or add-on treatment. In the univariable meta-regression model, the variables resulting significantly associated with the estimate of the treatment effect was the Withania somnifera dose, with a larger anxiolytic effect observed in trials administering Withania somnifera at a higher dose (unstandardised regression coefficient (B: −0.004 (95% CI: −0.007; −0.001)) and the use as add-on treatment (B: −0.753 (95% CI: −1.47; −0.038)). It should be noted that this result might be strongly influenced by the outlier effect played by the study from Khyati et al ^{Reference Khyati and Anup49} which used a dose of Withania somnifera more than ten times higher than the median dose. Univariable meta-regression results are displayed in Supplementary Table 6.

The effect of Withania somnifera on anxiety symptoms was further examined using a Bayesian hierarchical meta-analysis, which confirmed a statistically significant effect of the treatment (SMD: −1.23 (95% CrI: −2.19; −0.12)). The estimated between-study heterogeneity (τ) was 1.71 (95% CrI: 1.04; 2.82). This model applied weakly informative priors: a normal(0,1) prior for the intercept and a cauchy(0,0.5) prior for the standard deviation of the random effects and converged at 4000 iterations. Results are shown in Fig. 3. For sensitivity analysis, the model was re-estimated using an alternative prior for τ (i.e. exponential(1)). Results were consistent across priors, with stable posterior distributions for both τ and the overall treatment effect (see Supplementary Figure 2). Model comparison using leave-one-out cross-validation indicated similar predictive performance (difference in Expected Log Pointwise Predictive Density (∆ELPD): −0.1). Given the conservative nature of the cauchy(0,0.5) prior, we retained it in the final model.

Fig. 3 Bayesian forest plot of study-level and pooled posterior estimates of Withania somnifera efficacy on anxiety. The measures of anxiety across the studies were scaled as lower scores indicate lower anxiety symptoms. Reference to the included studies (in alphabetical order): Andrade et al, 2000 ^{Reference Andrade, Aswath, Chaturvedi, Srinivasa and Raguram41} ; Chengappa et al, 2013 ^{Reference Chengappa, Bowie, Schlicht, Fleet, Brar and Jindal42} ; Cooley et al, 2009 ^{Reference Cooley, Szczurko, Perri, Mills, Bernhardt and Zhou45} ; Fuladi et al, 2021 ^{Reference Fuladi, Emami, Mohammadpour, Karimani, Manteghi and Sahebkar46} ; Hosseini et al, 2019 ^{Reference Hosseini, Moharari, Soltanifar, Mohammad Poor, Hosseini and Zahedian48} ; Khyati et al, 2013 ^{Reference Khyati and Anup49} ; Langade et al, 2019 ^{Reference Langade, Kanchi, Salve, Debnath and Ambegaokar50} ; Langade et al 2021 ^{Reference Langade, Thakare, Kanchi and Kelgane51} ; Majeed et al, 2023 ^{Reference Majeed, Nagabhushanam, Murali, Vishwanathan, Mamidala and Mundkur52} ; Pandit et al, 2024. ^{Reference Pandit, Srivastav, Sur, Chaudhuri, Wang and Biswas53}

Finally, we included a Withania somnifera dose as a moderator in the Bayesian model. To enhance interpretability, we log-transformed the dose variable and centred the intercept at the median dose of 600 mg/day. At this reference dose, the model estimated a significant treatment effect on anxiety symptoms (SMD: −1.27 (95% CrI: −2.17; −0.21)). However, we found no evidence of a statistically significant increase in treatment efficacy for doses above the median (B: −0.50 (95% CrI: −1.19; 0.16)).

Effect of Withania somnifera on depression symptoms

The meta-analysis of the effect of Withania somnifera on depression symptoms included five studies, and specifically for one of these the outcome data were extracted from a secondary publication on the same sample. ^{Reference Gannon, Brar, Rai and Chengappa54} The pooled estimate of the treatment effect favoured Withania somnifera (SMD: −1.28 (95% CI: −2.40; −0.16])), with high between-study heterogeneity (I ²: 93%; p-value <0.001). While generally recommended with a minimum of ten studies, we however conducted an exploratory subgroup analysis based on participants’ diagnoses to investigate potential sources of heterogeneity. The test for subgroup differences was statistically significant (p-value <0.001), with the largest effect of Withania somnifera on depression symptoms observed in studies involving participants with anxiety, followed by those with MDD. No significant effect was found in the single studies involving participants with bipolar disorder and schizophrenia. These results should be interpreted cautiously due to the limited number of studies within each subgroup. The results are summarised in Fig. 4.

Fig. 4 Forest plot of depression among Withania somnifera and control groups: overall and by diagnostic subgroup. BD, bipolar disorder; IV, inverse variance; MDD, major depressive disorder; SZ, schizophrenia; WS, Withania somnifera. The measures of depression across the studies were scaled as lower scores indicate lower depression symptoms. Reference to the included studies (in alphabetical order): Chengappa et al, 2013 ^{Reference Chengappa, Bowie, Schlicht, Fleet, Brar and Jindal42} ; Chengappa et al, 2018 ^{Reference Chengappa, Brar, Gannon and Schlicht43} ; Fulzele et al, 2014 ^{Reference Fulzele, Ingle, Huda and Mishra47} ; Majeed et al, 2023 ^{Reference Majeed, Nagabhushanam, Murali, Vishwanathan, Mamidala and Mundkur52} ; Pandit et al, 2024. ^{Reference Pandit, Srivastav, Sur, Chaudhuri, Wang and Biswas53}

Due to the limited number of studies, we did not conduct a test for publication bias and meta-regression.

Leave-one-out analysis showed that by excluding the study from Chengappa et al, ^{Reference Chengappa, Bowie, Schlicht, Fleet, Brar and Jindal42} the pooled estimate favouring Withania somnifera reached statistical significance (SMD: −1.28 (95% CI: −2.45; -0.10)), with high heterogeneity (I ²: 91%). In addition, by excluding the study from Majeed et al, ^{Reference Majeed, Nagabhushanam, Murali, Vishwanathan, Mamidala and Mundkur52} there was a marked decrease in the heterogeneity (I ²: 71%) and in the effect size (SMD: −0.39 (95% CI: −1.01; 0.24)). The results of the leave-one-out analysis are available in Supplementary Table 7.

To further examine the robustness of the findings in a context with high heterogeneity, we performed Bayesian hierarchical meta-analysis using weakly informative priors (i.e. normal(0,1) for the intercept, cauchy(0,0.5) and exponential(1) for τ). Both models converged at 4000 iterations and produced similar findings in terms of posterior distribution of τ and the overall treatment effect (see Supplementary Figure 3; ∆ELPD: −0.3). Given its conservative features, the model using a cauchy(0,0.5) prior was retained as the primary analysis. This model yielded a non-statistically significant treatment effect (SMD: −0.95 (95% CrI: −1.96; 0.24)) and high between-study heterogeneity (τ: 1.29 (95% CrI: 0.62; 2.66)). The results of the Bayesian meta-analysis are displayed in Fig. 5.

Fig. 5 Bayesian forest plot of study-level and pooled posterior estimates of Withania somnifera efficacy on depression. The measures of depression across the studies were scaled as lower scores indicate lower depression symptoms. Reference to the included studies (in alphabetical order): Chengappa et al, 2013 ^{Reference Chengappa, Bowie, Schlicht, Fleet, Brar and Jindal42} ; Chengappa et al, 2018 ^{Reference Chengappa, Brar, Gannon and Schlicht43} ; Fulzele et al, 2014 ^{Reference Fulzele, Ingle, Huda and Mishra47} ; Majeed et al, 2023 ^{Reference Majeed, Nagabhushanam, Murali, Vishwanathan, Mamidala and Mundkur52} ; Pandit et al, 2024. ^{Reference Pandit, Srivastav, Sur, Chaudhuri, Wang and Biswas53}

Effect of Withania somnifera on sleep quality

The meta-analysis of the effect of Withania somnifera on sleep included four studies. The pooled estimate favoured Withania somnifera (SMD: −1.35 (95% CI: −1.79; −0.91)) with moderate between-study heterogeneity (I ²: 53%, p-value: 0.097). The results are presented in Fig. 6.

Fig. 6 Forest plot of sleep quality among Withania somnifera and control groups. IV: inverse variance; WS, Withania somnifera. The measures of sleep quality across the studies were scaled as lower scores indicate better sleep quality. Reference to the included studies (in alphabetical order): Langade et al, 2019 ^{Reference Langade, Kanchi, Salve, Debnath and Ambegaokar50} ; Langade et al 2021 ^{Reference Langade, Thakare, Kanchi and Kelgane51} ; Majeed et al, 2023 ^{Reference Majeed, Nagabhushanam, Murali, Vishwanathan, Mamidala and Mundkur52} ; Pandit et al, 2024. ^{Reference Pandit, Srivastav, Sur, Chaudhuri, Wang and Biswas53}

Due to the limited number of studies included in the meta-analysis, sensitivity analyses could not be performed.

Effect of Withania somnifera on other mental health symptoms

Four other outcome measures were reported in the included studies. These consisted of schizophrenia symptoms, mania symptoms, OCD symptoms and perceived stress. Except for perceived stress, each outcome was reported by only one study, precluding meta-analysis. The meta-analysis for perceived stress was based on three studies, indicating a significant effect in favour of Withania somnifera (SMD: −0.95 (95% CI: −1.46; −0.43)), with moderate between-study heterogeneity (I ²: 58%, p-value: 0.090). The results are shown in Fig. 7.

Fig. 7 Forest plot of perceived stress among Withania somnifera and control groups. IV: inverse variance; WS: Withania somnifera. The measures of stress across the studies were scaled as lower scores indicate lower perceived stress. Reference to the included studies (in alphabetical order): Chengappa et al, 2018 ^{Reference Chengappa, Brar, Gannon and Schlicht43} ; Choudhary et al, 2017 ^{Reference Choudhary, Bhattacharyya and Joshi44} ; Pandit et al, 2024. ^{Reference Pandit, Srivastav, Sur, Chaudhuri, Wang and Biswas53}

Analysis of safety and tolerability

The analysis of the safety and tolerability of Withania somnifera treatment was made by assessing the rates of adverse events, drop-out due to any cause and drop out due to death or serious adverse events across the experimental and control groups. There were four studies ^{Reference Fuladi, Emami, Mohammadpour, Karimani, Manteghi and Sahebkar46–Reference Khyati and Anup49} not reporting information on the number of treatment’s adverse effects across the trial’s arms, although these stated in the respective papers that there were no significant differences between the experimental and control groups. Additionally, two studies ^{Reference Fulzele, Ingle, Huda and Mishra47,Reference Khyati and Anup49} did not report information on the number of drop-outs during the trial. Notably, no death or drop-out due to severe treatment adverse effects occurred both in the experimental and in the control groups of all the included trials. The meta-analysis of the drop-out due to any cause included ten studies, and the meta-analysis of the rate of any adverse event was based on five studies. Our investigation found similar rates of drop-out due to any cause and of any adverse effect across the experimental and control groups, supporting no significant differences between Withania somnifera and the placebo in terms of safety and tolerability (odds ratio 0.81 (95% CI: 0.42; 1.58) and odds ratio 0.99 (95% CI: 0.54; 1.82) for the rate of drop-out due to any cause and any adverse effect, respectively), with no evidence of between-study heterogeneity. The most reported adverse events were gastro-intestinal disturbances (e.g. diarrhoea or dyspepsia), which were transitory and not severe. The results are displayed in Fig. 8.

Fig. 8 Forest plots of drop-out due to any cause (Panel A) and any adverse effect (Panel B) among Withania somnifera and control groups. MH: Mantel-Haenszel; WS: Withania somnifera. Lower rates indicate better safety/tolerability. Reference to the included studies (in alphabetical order): Andrade et al, 2000 ^{Reference Andrade, Aswath, Chaturvedi, Srinivasa and Raguram41} ; Chengappa et al, 2013 ^{Reference Chengappa, Bowie, Schlicht, Fleet, Brar and Jindal42} ; Chengappa et al, 2018 ^{Reference Chengappa, Brar, Gannon and Schlicht43} ; Choudhary et al, 2017 ^{Reference Choudhary, Bhattacharyya and Joshi44} ; Cooley et al, 2009 ^{Reference Cooley, Szczurko, Perri, Mills, Bernhardt and Zhou45} ; Fuladi et al, 2021 ^{Reference Fuladi, Emami, Mohammadpour, Karimani, Manteghi and Sahebkar46} ; Hosseini et al, 2019 ^{Reference Hosseini, Moharari, Soltanifar, Mohammad Poor, Hosseini and Zahedian48} ; Jahanbakhsh et al, 2016 ^{Reference Jahanbakhsh, Manteghi, Emami, Mahyari, Gholampour and Mohammadpour17} ; Langade et al, 2019 ^{Reference Langade, Kanchi, Salve, Debnath and Ambegaokar50} ; Langade et al 2021 ^{Reference Langade, Thakare, Kanchi and Kelgane51} ; Majeed et al, 2023. ^{Reference Majeed, Nagabhushanam, Murali, Vishwanathan, Mamidala and Mundkur52}

GRADE of the evidence

A detailed summary on the risk of bias in all 14 trials is reported in Supplementary Figs 4 and 5, along with an assessment of the quality of the evidence (see Supplementary Table 8). In the GRADE system, the evidence from RCTs is initially set to high; there are then criteria that can be used either to downgrade or upgrade (see further information in the Supplementary Material). The quality of the evidence ranged from very low for depression symptoms to low for anxiety symptoms, sleep quality and stress. The main reasons for downgrading the certainty of evidence were the limited number of studies contributing to each pooled estimate, the high level of heterogeneity between the studies and bias threats related to the blinding procedures.