Article Text

Download PDFPDF

Research paper
Efficacy and safety of anti-inflammatory agents for the treatment of major depressive disorder: a systematic review and meta-analysis of randomised controlled trials
  1. Shuang Bai1,
  2. Wenliang Guo2,
  3. Yangyang Feng1,
  4. Hong Deng1,
  5. Gaigai Li1,
  6. Hao Nie1,
  7. Guangyu Guo1,
  8. Haihan Yu1,
  9. Yang Ma1,
  10. Jiahui Wang1,
  11. Shiling Chen1,
  12. Jie Jing1,
  13. Jingfei Yang1,
  14. Yingxin Tang1,
  15. Zhouping Tang1
  1. 1 Department of Neurology, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China
  2. 2 Department of Neurology, Liyuan Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China
  1. Correspondence to Professor Zhouping Tang, Department of Neurology, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan 430030, China; ddjtzp{at}; Dr Yingxin Tang; yingxintang{at}


Objectives To systematically review the efficacy and safety of anti-inflammatory agents for patients with major depressive disorders.

Methods We searched the literature to identify potentially relevant randomised controlled trials (RCTs) up to 1 January 2019. The primary outcome was efficacy, measured by mean changes in depression score from baseline to endpoint. Secondary outcomes included response and remission rates and quality of life (QoL). Safety was evaluated by incidence of classified adverse events. Heterogeneity was examined using the I2 and Q statistic. Pooled standard mean differences (SMDs) and risk ratios (RRs) were calculated. Subgroup meta-analyses were conducted based on type of treatment, type of anti-inflammatory agents, sex, sponsor type and quality of studies.

Results Thirty RCTs with 1610 participants were included in the quantitative analysis. The overall analysis pooling from 26 of the RCTs suggested that anti-inflammatory agents reduced depressive symptoms (SMD −0.55, 95% CI −0.75 to −0.35, I2=71%) compared with placebo. Higher response (RR 1.52, 95% CI 1.30 to 1.79, I2=29%) and remission rates (RR 1.79, 95% CI 1.29 to 2.49, I2=41%) were seen in the group receiving anti-inflammatory agents than in those receiving placebo. Subgroup analysis showed a greater reduction in symptom severity in both the monotherapy and adjunctive treatment groups. Subgroup analysis of non-steroidal anti-inflammatory drugs, omega-3 fatty acids, statins and minocyclines, respectively, disclosed significant antidepressant effects for major depressive disorder (MDD). For women-only trials, no difference in changes of depression severity was found between groups. Subanalysis stratified by sponsor type and study quality led to the same outcomes in favour of anti-inflammatory agents in both subgroups. Changes of QoL showed no difference between the groups. Gastrointestinal events were the only significant differences between groups in the treatment periods.

Conclusions Results of this systematic review suggest that anti-inflammatory agents play an antidepressant role in patients with MDD and are reasonably safe.

  • antidepressant effects
  • anti-inflammatory agents
  • major depressive disorder
  • meta-analysis
  • systematic review
View Full Text

Statistics from


Major depressive disorder (MDD) is one of the most common mental diseases. It affects approximately 6% of adults worldwide every year.1 MDD, with typical symptoms of depressive mood, anhedonia and lack of interest, is the leading contributor to disability.2 Despite a wide variety of pharmacological and psychological treatments, around 30% of patients do not experience remission, and even treatment-responsive patients have high relapse and recurrence rates.3 Available medication plays antidepressive roles mainly by regulating levels of neurotransmitters in the brain. With a single therapeutic target, these drugs are often ineffective and have side effects, such as nausea, insomnia, weight gain, metabolic disorders, cardiovascular toxicity and even suicide, which are poorly tolerated.4 Therefore, new treatments to improve efficacy and safety are urgently needed.

Recently, inflammation has been recognised as being involved in the pathogenesis of MDD. Several observational studies have detected raised circulating levels of inflammatory markers in patients with MDD, such as interleukin-6 (IL-6), soluble interleukin-2 receptor, C-reactive protein and tumour necrosis factor-α.5 6 A meta-analysis of 22 studies found that IL-1β and IL-6 levels decreased in patients who responded to treatment, together with an improvement in depressive symptoms.7 Research has suggested that cell-mediated immune response (CMI) also plays a role in MDD. Maes et al report an increased percentage of T cells and stimulated production of interferon-γ.8 Another study found an increasing level of CMI-related cytokines, such as plasma soluble IL-2 receptor (sIL-2R) and soluble CD8 molecules.9 These findings suggest that inflammation contributes to depressive symptoms in patients with MDD, and that anti-inflammatory agents may provide antidepressive effects.

Over the past few decades, a number of randomised controlled trials (RCTs) have examined antidepressive effects of anti-inflammatory agents for MDD.10–40 However, these studies have conflicting results and no clear conclusions. One meta-analysis found that anti-inflammatory agents had a beneficial effect on both depressive and manic symptoms for mood disorders.41 However, the authors of that study included both patients with MDD and those with bipolar disorders, which can be considered heterogeneous in both pathogenesis and prognosis. A meta-analysis by Köhler found that anti-inflammatory drugs decreased depression and depressive symptoms. However, the drugs tested were non-steroidal anti-inflammatory drugs (NSAIDs) and cytokine inhibitors, and high heterogeneity was detected among studies.42

We therefore conducted a systematic review and meta-analysis to compare the efficacy and safety of anti-inflammatory agents against a placebo as monotherapy or adjunctive treatment for patients with MDD. Anti-inflammatory treatments were defined as NSAIDs, omega-3 fatty acid (omega-3 FA), cytokine inhibitors, statins, corticosteroids, minocycline, pioglitazone, modafinil and N-acetylcysteine (NAC). The systematic review was registered at the PROSPERO (registration number: CRD42019129499).


Search strategy

A computerised literature search was carried out using PubMed, Embase, the Cochrane Library and Web of Science to identify relevant articles through 31 December 2018. Search strategies are reported in online supplementary table S1. To locate unpublished literature, meeting abstracts were searched in the ISI Proceedings database and we also conducted a search using the Clinical Trial Register and International Clinical Trials Registry Platform reference lists of the retrieved articles, and recent reviews on this topic were also reviewed manually to find additional eligible studies.

Supplementary data

Inclusion and exclusion criteria

All identified articles were assessed to determine whether they met the following criteria: (1) randomised, placebo-controlled, trial investigating the efficacy and safety of anti-inflammatory agents as monotherapy (anti-inflammatory agents vs placebo) or as add-on therapy (antidepressants with anti-inflammatory agents vs antidepressants with placebo) for MDD in humans; (2) diagnosis of MDD based the Diagnostic and Statistical Manual for Mental Disorders, fourth edition (DSM-IV), the Diagnostic and Statistical Manual for Mental Disorders, fourth edition, text revision (DSM-IV-TR) or the Diagnostic and Statistical Manual for Mental Disorders, fifth edition (DSM-5); (3) participants older than 18; (4) standardised and validated scales to assess depression severity, such as the Hamilton Rating Scale for Depression (HAMD), the Montgomery-Asberg Depression Rating Scale (MADRS), Beck’s Depression Inventory (BDI), Beck’s Depression Inventory-II (BDI-II) or the Geriatric Depression Scale (GDS); (5) standardised and validated scales to assess quality of life (QoL), such as Quality of Life Enjoyment and Satisfaction Questionnaire (Q-LES-Q), Quality of Life Enjoyment and Satisfaction Questionnaire–Short Form (Q-LES-Q-SF), EuroQoL (EQ-5D) scale and the 36-item Short Form Health Survey.

Exclusion criteria were (1) studies published only as abstracts from conference proceedings of scientific meetings; (2) studies including patients with bipolar depressive disorder or minor depression; (3) randomised trials without a placebo or treatment group; (4) if multiple studies reported data from the same population, only the study with largest sample size or most information was included.

Data extraction

Data were extracted independently by three authors using data extraction forms. We gave priority to using intention-to-treat data for the meta-analysis. We obtained the following information from each study: the first author’s surname, publication year, study period, study design, mean ages of participants, percentage of female participants and number of participants in each group, description of intervention, methods for diagnosing MDD, methods for measuring depression severity, sponsor type, effect estimates: changes in depression severity between the baseline and endpoint (mean±SD), depression severity at baseline and endpoint (mean±SD), rates or number of responses and remission, changes in QoL between baseline and endpoint (mean±SD), rates or number of adverse effects (AEs) in each group. The response to treatment was defined as at least 50% reduction from baseline in depression scales. Remission was defined as HAMD ≤7, MADRS ≤7, BDI-II ≤8 or GDS ≤11 at the endpoint. When data were insufficient or missing, we contacted the authors for further information.

Assessment of risks of bias

The risk of bias was assessed for all RCTs included in this meta-analysis. Following the Cochrane Handbook for Systematic Review of Interventions,43 the bias risk for these studies was assessed based on the following seven domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting and other bias. The included trials were graded as low, moderate or high quality based on the following criteria: (1) trials were considered low quality if either randomisation or allocation concealment was assessed as a high risk of bias, regardless of the risk of other items; (2) trials were considered high quality when both randomisation and allocation concealment were assessed as a low risk of bias and all other items were assessed as low or unclear risk of bias in a trial; (3) trials were considered moderate quality if they met neither the criteria for high nor low risk.

Statistical methods and assessment of heterogeneity

Meta-analyses were conducted if data from three or more studies were available for each main comparison. Trials without necessary data were excluded from the quantitative analysis. We assessed heterogeneity among studies using Cochrane's Q test and I2 statistics. For the Q test, a p value <0.10 was considered to indicate significant heterogeneity, while for I2, a value of 0–25% represented insignificant heterogeneity, a value >25% but ≤50% represented low heterogeneity, a value >50% but ≤75% represented moderate heterogeneity, and a value >75% represented high heterogeneity.44 For high heterogeneity, quantitative analysis was not conducted. In other cases, we decided a priori to use a random-effects analysis owing to the expected heterogeneity resulting from different treatment regimens and patient populations.

The main comparison was the efficacy and safety of anti-inflammatory treatment with/without antidepressants versus placebo with/without antidepressants. The primary outcome for efficacy was the change in depression score from baseline to endpoint for both groups. If the trials did not provide changes in depression scores, values were calculated from depression scores at baseline and endpoint using the method recommended in the Cochrane Handbook.43 Standard mean differences (SMDs) and 95% CIs were calculated using the inverse variance statistical method. A study including two intervention groups was split into two studies with half the number of patients in their control groups. Other outcomes included the rates of response and remission, the score changes in quality of life from baseline to endpoint for both groups. We performed a meta-analysis to calculate risk ratios (RRs) and 95% CIs using the Mantel-Haenszel statistical method. If no events were reported for one group in a comparison, a value of 0.5 was added to both groups for that study. Based on the recommendation of the Cochrane Handbook, trials with no events in both intervention and control groups were not included in the calculation of RRs. The incidence of AEs between groups was used to evaluate the safety. Pooled RRs and 95% CIs were also calculated using the Mantel-Haenszel statistical method, as above.

We conducted subgroup meta-analyses based on the type of treatment (monotherapy or adjunctive therapy), type of anti-inflammatory agents, sex (women-only trials or trials that included both men and women), sponsor type (non-profit organisations or commercial industries) and the quality of studies. Sensitivity analyses were performed, excluding trials not providing changes in HAMD.

We assessed publication bias by visual inspection of a funnel plot when at least 10 trials reported the primary outcomes. Publication bias was also assessed by both the Begg test and the Egger test, with p<0.05 indicating statistical significance. We performed the Duval and Tweedie non-parametric 'trim and fill' procedure to further assess the possible effect of publication bias in our meta-analysis.

All meta-analyses were performed using Revman (version 5.3) or STATA (version 14.0). All tests were two-tailed, and p<0.05 was considered statistically significant.


Search results

We identified 2331 records through the computerised retrieval and two additional articles through the manual search. After reviewing the titles and abstracts, 2252 were excluded. The full texts of 81 articles were reviewed for inclusion, and 31 RCTs met the inclusion criteria for systematic review. Details of the study selection process are shown in figure 1.

Figure 1

Flow diagram representing the selection process. BP: bipolar disorder; MD: minor depression.

Study characteristics

We included 31 randomised control trials in the meta-analysis.10–40 The study by Meyer et al did not provide necessary data and was excluded from quantitative analysis.27 Of these 31 trials, 26 reported changes in depression score from baseline to endpoint directly or indirectly, involving a total of 1610 participants. Nineteen studies showed the response rates, while 16 reported the remission rates between groups. Five studies reported the score changes of QoL, and a total of 20 studies reported the incidence of AEs between groups. Among these, four studies looked at NSAIDs,10–13 17 at omega-3,14–30 3 at statins,31–33 3 at minocycline,34–36 1 at pioglitazone,37 2 at modafinil38 39 and 1 at N-acetylcysteine.40 Twenty-two trials evaluated the the effects of anti-inflammatory agents in addition to conventional treatment for MDD, while eight investigated their antidepressant effects in monotherapy versus placebo. In addition, studies by da Silva et al examined both monotherapy and adjunctive therapy.21 Most studies used the DSM-IV criteria for MDD diagnosis. Seven studies used the DSM-IV-TR and two the DSM-5. Of these, three studies defined participants as pregnant and postpartum women.17–19 Except for three studies using GDS24 and MADRS,30 40 all others provided scores using the HAMD scale for evaluation of depressive changes, while three studies on perinatal depression also provided data using the Edinburgh postpartum depression scale. Of the 31 studies, 13 declared a sponsorship from non-profit organisations (NPOs), 14 from commercial industries (CIs), and 4 did not declare. Follow-up periods ranged from 4 to 16 weeks. Details of study characteristics are presented in table 1.

Table 1

Characteristics of the included trials and participants

Assessment of bias

Online supplementary figures S1 and S2 show the assessment of the risk of bias. All studies were randomised placebo-controlled trials; 30 claimed to be double blind, 19 of which described a reliable method; 19 trials described an adequate random sequence generation process; 19 trials described the methods used for allocation concealment. According to the criteria, 16 studies were of a high quality while the others were of a moderate quality.

The funnel plot displayed an asymmetrical funnel shape (figure 2A). The Egger test (z=3.34, p=0.001) and Begg test (p=0.000) also raised the possibility of publication bias. Thus, we undertook a sensitivity analysis using the trim and fill method. In this analysis, no trimming was performed and the outcome of the meta-analysis was unchanged, which revealed the robustness of the outcome with no publication bias (figure 2B).

Figure 2

Funnel plots without and with trim and fill. SMD, standard mean difference.

Pooled antidepressant effect of anti-inflammatory agents

A total of 26 studies were included in the meta-analysis for the primary outcome. Of these, 15 reported the mean changes of depression scores from baseline to the endpoint, while 11 gave the depression scores at baseline and endpoint, allowing mean changes of depression score to be calculated. The pooling results of these studies suggested that anti-inflammatory agents had a statistically significant antidepressant effect (a larger reduction in depression scale score) with a pooled effect estimate (SMD) of −0.55 (95% CI −0.75 to −0.35, p<0.00001) (figure 3). Heterogeneity among studies was found to be moderate (X2=92.65, df=27, p<0.00001, I2=71%).

Figure 3

Forest plot of primary outcome in overall analysis and subgroup analysis stratified according to the type of treatment.

Nineteen studies reported the response rates. As shown in figure 4, pooling analysis suggested a significant difference in response rates between anti-inflammatory agents and placebo with a pooled RR of 1.52 (95% CI 1.30 to 1.79, p<0.00001). The heterogeneity among studies was found to be low (X2=25.41, df=18, p=0.11, I2=29%).

Figure 4

Forest plot of secondary outcome (response rate) in overall analysis.

Seventeen studies provided the remission rates. However, one study was excluded from the meta-analysis owing to no events in both intervention and control groups. Pooling the other 16 trials, the anti-inflammatory agent group demonstrated a significantly larger remission rate than placebo (RR 1.79, 95% CI 1.29 to 2.49, p=0.0005). Heterogeneity among studies was found to be low (X2=25.52, df=15, p=0.04, I2=41%) (figure 5).

Figure 5

Forest plot of secondary outcome (remission rate) in overall analysis.

Subgroup analysis

Monotherapy of anti-inflammatory agents for MDD

Eight studies investigated the effect of anti-inflammatory agents as monotherapy for MDD treatment, including a total of 560 patients. Subanalysis indicated that anti-inflammatory agents had an antidepressant action when the primary outcome of changes in symptom severity was considered (eight RCTs, SMD −0.30, 95% CI: −0.58 to −0.02, p=0.03). Moderate heterogeneity was detected among studies (X2=22.30, df=9, p=0.008, I2=60%) (figure 3).

Adjunctive anti-inflammatory agents for MDD

Antidepressant effects of anti-inflammatory agents as adjunctive treatment for MDD were reported in 21 of the studies. For changes in the depressive symptom score from baseline to endpoint, pooled results from 18 trials showed a significant antidepressant effect with SMD of −0.70 (95% CI: −0.97 to −0.43, p<0.00001). Moderate heterogeneity was found among studies (X2=66.25, df=17, p<0.00001, I2=74%) (figure 3).

Antidepressant effect of NSAIDs in MDD

Four studies explored the antidepressant effects of NSAIDs in addition to traditional antidepressants. In all four studies, celecoxib was added at the same dose (400 mg/day). Traditional antidepressants used were sertraline, fluoxetine and reboxetine.

We combined these studies for a subgroup meta-analysis, finding that antidepressants combined with celecoxib were better than antidepressants alone, leading to a larger reduction in depressive score (four RCTs, SMD −0.76, 95% CI −1.14 to −0.39, p<0.0001). Heterogeneity among studies was insignificant (table 2).

Table 2

Subgroup analyses based on therapy type, type of anti-inflammatory agents, sex, sponsor type and the quality of studies

Omega-3 fatty acid for MDD

Seventeen studies examining the efficacy of omega-3 unsaturated fatty acid for MDD were included, with 12 reporting the primary outcome, nine reporting response rates and eight remission rates. Of these studies, eight demonstrated significant effects in antidepression in either adjunctive use or monotherapy. The study by Meyer et al not included in quantitative analysis show no detectable benefit of omega-3 compared with placebo when added to a standard outpatient treatment.

Subanalysis found that omega-3 FAs had an antidepressant action, leading to a larger reduction in depression symptoms (12 RCTs, SMD −0.35, 95% CI −0.60 to −0.09, p=0.008). Heterogeneity was moderate among studies (X²=33.67, df=13, p=0.001, I²=61%) (table 2).

Antidepressant effect of statin for MDD

Three studies were included which investigated the effects of adding statins to fluoxetine or citalopram for MDD. Subgroup meta-analysis found that antidepressants combined with statins resulted in a superior antidepressive effect compared with antidepressants alone, with a larger score reduction (three RCTs, SMD −0.65, 95% CI −0.96 to −0.33, p<0.0001). Heterogeneity among studies was insignificant (table 2).

Antidepressant effect of minocycline in MDD

Three studies assessing antidepressant effects of adjunctive minocycline in patients in MDD were identified. Pooling of effect sizes for these three RCTs revealed an SMD of −0.79 (95% CI −1.29 to −0.28, p=0.002), which was indicative of a significant reduction in depression scores in the minocycline group compared with the placebo group. Moderate heterogeneity were detected among studies respectively (table 2).

Pioglitazone, modafinil and N-acetylcysteine for MDD

No adequate studies were identified to carry out a subgroup meta-analysis for pioglitazone and NAC, while heterogeneity was too high (X2=13.40, df=1, p=0.0003, I2=93%) among the modafinil group to conduct a quantitative analysis. Sepanjnia et al found that patients in the pioglitazone (added to citalopram) group had significantly lower scores at all time points and higher response and remission rates than the placebo group.37 The frequency of side effects was similar between the two groups. Another study examining the effects of modafinil augmented to fluoxetine demonstrated the same outcomes.39 However, two studies for modafinil38and NAC40 added to ongoing treatment did not find a difference between anti-inflammatory drugs and placebo. However, response and remission rates were greater in the NAC group at measurement after discontinuation.

Antidepressant effect of anti-inflammatory agents in women with MDD

Five trials focused on women with MDD. Of these, one assessed the effect of celecoxib added to sertraline, while the other four examined the antidepressive effect of omege-3 FA as monotherapy. Combining the five studies, no difference in score changes was found between groups (SMD −0.49, 95% CI −1.01 to 0.03, p=0.06), with moderate heterogeneity detected among studies (table 2).

Subgroup analysis based on sponsor type

Of the trials included in quantitative analysis, 11 declared a sponsorship from non-profit organisations 12 from commercial industries, and three did not declare. Subgroup meta-analysis for primary outcome suggested that anti-inflammatory agents showed a greater antidepressive effect compared with placebo in all the subgroups (table 2).

Subgroup analysis based on study quality

Of the trials included in quantitative analysis, 15 were high quality while the other 15 were moderate. Subgroup meta-analysis suggested that anti-inflammatory agents showed a superior antidepressive effect than placebo, with a greater score reduction in both high and moderate quality groups (table 2).

Sensitivity analyses of the primary outcome

A sensitivity analysis including only trials with HAMD scores showed no change in primary outcome (online supplementary figure S3 in online supplementary file).

Supplementary data

Quality of life

Five studies reported the changes in quality of life. The pooling results of these studies suggested that no difference was found for improvement in quality of life between the anti-inflammatory agent and the placebo groups, with a pooled effect estimate (SMD) of 0.17 (95% CI −0.10 to 0.43, p=0.22). Heterogeneity among studies was found to be low (X2=9.73, df=5, p=0.08, I2=49%).

Safety of anti-inflammatory agents for MDD treatment

Some 27 studies reported the incidence of AEs. Of these, Berk et al reported five serious AEs in the NAC group and four in placebo group, with no significant difference between groups.40 In a study by Bot et al, one person assigned to the eicosapentaenoic acid group had an allergic reaction, which consisted of rashes and itching.25 No serious AEs were reported in other studies.

Quantitative analysis was conducted on data from 20 studies. As shown in table 3, where both the main and subgroup analyses were stratified by drug type, there were no differences in event incidence for nervous system, psychiatric, cardiovascular, respiratory, genitourinary, musculoskeletal or cutaneous events between groups, with one exception: patients in the modafinil group had a higher incidence of psychiatric AEs than the control group. The incidences of gastrointestinal AEs were found to be higher for participants taking anti-inflammatory drugs, especially statin and NAC (table 3).

Table 3

Meta-analysis of adverse effects


This meta-analysis focused on RCTs to examine the efficacy and safety of anti-inflammatory agents as adjunctive treatment and monotherapy in patients with major depressive disorder. Inclusion criteria were strictly set and applied according to recognised diagnosis criteria, ensuring that only patients with MDD were included. According to available experimental and clinical trials, we included NSAIDs, polyunsaturated fatty acids, cytokine inhibitors, statins, corticosteroids, minocycline, pioglitazone, modafinil and NAC as anti-inflammatory agents.

The overall analysis of 30 RCTs suggested a significant antidepressant effect of anti-inflammatory agents in comparison with placebo for patients with MDD in various efficacy outcomes. A significantly larger reduction in depressive rating scales was found in the intervention group, with moderate heterogeneity among studies. Similarly, the response and remission rates were significantly higher for patients taking anti-inflammatory drugs with low heterogeneity. The heterogeneity among studies may be associated with difference in depression assessment scales, demographic characteristic of participants, drug type and treatment course. HAMD and MADRS are both widely used scales for measuring the severity of depression, which have high reliability and validity, yet MADRS was found to be more sensitive to changes of depressive symptoms.45 While the HAMD scale was used in most studies and SMD was chosen as the effect size, the difference of scale sensibility might affect the effect size, contributing to heterogeneity. Most studies recruited patients from the general adult population, while some focused on women, old people or those with a certain disease. The overall analysis synthetically evaluated seven types of anti-inflammatory drugs for courses of 4 to 12 weeks, which might have led to diverse therapeutic effects.

In the subgroup analysis, we examined the effects of adjunctive treatments and monotherapy, respectively. Anti-inflammatory agents demonstrated significant antidepressive effects in adjunctive treatments, while in monotherapy a larger reduction was found only for the intervention group. Furthermore, adjunctive treatment showed a larger effect size than monotherapy. The reason for the subgroup difference is unknown, though may be associated with the synergy between drugs, considering that the antidepressants, NSAIDs, statins, pioglitazone and modafinil, were all metabolised by liver Kodachrome P450 enzyme. Subgroup analysis of NSAIDs, omega-3 FAs, statins and minocyclines, respectively, revealed significant antidepressant effects for MDD for the primary outcome. However, only three studies was included in the statin and minocycline group, while pioglitazone, modafinil and NAC were not analysed for inadequate study and high heterogeneity. In the women-only trials, no difference in change of depression scores was found between groups. Subanalysis stratified by sponsor type disclosed a result favouring anti-inflammatory agents in both the NPO and CI groups; the former reached a larger effect size, which suggested that no sponsor bias was detected. Subanalysis stratified by study quality led to the same outcomes, favouring anti-inflammatory agents in both subgroups.

Quality of life was an important measure for treatment of MDD. This meta-analysis found that anti-inflammatory agents did not improve the quality of life, possibly owing to the small number of studies. In addition, MDD is a chronic disease that requires long-term treatment, so it is difficult to obtain improvement in quality of life measures in several weeks.

The safety of anti-inflammatory agents was evaluated by the incidence of AEs between groups. Among the included studies, two studies reported serious AEs, with no difference in incidence rates between groups. AEs were classified into eight categories and evaluated accordingly. The only significant differences found between groups in the treatment periods from 4 to 12 weeks was for gastrointestinal events, and differences were found only for statins and NACs. No major cardiovascular or gastrointestinal events—the biggest concern reported in previous study for NSAIDs—were seen in the studies. Previous studies found that omega-3 FA could affect coagulation and lead to bleeding, but this was not reported in the included studies. Gastrointestinal events—the most common AEs for statins—were detected in the meta-analysis, while no study reported rhabdomyolysis—one of most serious AEs for statins. One of three studies of statins reported levels of alanine transaminase and aspartate transaminase, finding no difference between the groups. Similarly, incidence of typical AEs such as vertigo, photosensitivity and skin hyperpigmentation revealed no difference between minocycline and placebo. Few studies were included which investigated pioglitazone, modafinil and NACs. With pioglitazone, hypoglycaemia is rare when used alone and is often limited to diabetic patients receiving combination therapy, whereas oedema is a well-known side effect of the drug.46 According to previous studies, common AEs for modafinil include insomnia and decreased appetite, while the most common AEs for NACs are gastrointestinal events.

Our meta-analysis has several limitations. First, several studies did not report the changes in depression scores between baseline and endpoint, instead providing the scores at baseline and at the endpoint, respectively. For these studies, we calculated the changes from the provided baseline and endpoint scores according to the recommendation in the Cochrane Handbook for Systematic Reviews of Interventions.43 This method might have introduced bias into the meta-analysis, although the sensitivity analysis excluding these studies led to the same outcome.

Second, while most studies assessed depression severity using the HAMD scale, three studies used the GDS or MADRS. The Cochrane Handbook for Systematic Reviews of Interventions suggests pooling change in symptoms scores into one MD only if the same rating scale is used in all studies. Therefore, we used the SMD as a pooling estimate in this meta-analysis. However, an SMD may not be indicative of the true effect size of anti-inflammatory agents for MDD treatment.

The third limitation was the small number of studies and small sample sizes in the statin and minocycline groups, yielding a total of just 166 and 151 participants respectively. As a result, these subgroup results might be less robust and therefore insufficient to guide clinical treatments.

Fourth, the asymmetry of the funnel plot and the Egger test suggested the possibility of publication bias. Four studies identified were published only as abstracts from conference proceedings of scientific meetings, of which two reported negative outcomes. However, the trim and fill analysis did not change the general result, suggesting that the effect is not an artefact of unpublished negative studies.

A final limitation was the relatively short treatment durations, mostly ranging from 6 to 12 weeks. This meant that the long-term efficacy and AEs could not be evaluated.


Our systematic review and meta-analysis suggests that anti-inflammatory agents exert an antidepressant effect in the treatment of MDD and were generally safe with rates of adverse effects similar to those of placebo. Adjunctive treatment shows a more remarkable effect with NSAIDs, omega-3 FAs, statins and minocyclines showing significant antidepressant effects for MDD.

Anti-inflammatory agents show promising effects for MDD. However, owing to the chronic course of MDD, quality of life and adverse effects should be further investigated in high-quality randomised clinical trials with long-term follow-up.

Supplementary data

Supplementary data


We thank Jie Jia for his invaluable assistance in preparing the figures of the paper.


View Abstract


  • Contributors SB, WG, YF, HD, GL, ZT and YT were responsible for study concept and design. SB, WG, YF, HD, GL, HN, GG, HY, YM, JW, SC, JJ, JY, ZT and YT were responsible for the acquisition, analysis and interpretation of data. SB, WG, SC, JJ and JY drafted the manuscript. SB, WG, YF, HD, ZT and YT critically revised the manuscript for important intellectual content. ZT obtained the funding and was responsible for study supervision.

  • Funding This study was supported by funds from the National Natural Science Foundation of China, grant number: 81873750, URL:, recipient: ZPT.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.