NSAIDs act by inhibiting COX-1 and COX-2. These cyclooxygenase enzymes are necessary for the synthesis of some of the prostaglandins involved in inflammation. COX-2 inhibitors have direct effects on the CNS serotonergic system and can also affect it indirectly via immune processes in the CNS. A study in rats showed higher serotonin levels in both the frontal and temporoparietal cortices after administration of rofecoxib, a COX-2 selective NSAID (12). Consequently, the authors hypothesized that COX-2 inhibitors may have antidepressant effects. In a study in bulbectomized rats (a model for depression), chronic administration of the COX-2 inhibitor celecoxib decreased cytokine levels and changed the animals’ behavior (13). Brunello et al. (14) studied ASA in rats with the chronic escape deficit model and found that ASA accelerated the antidepressant effect of fluoxetine. In people with MD, a randomized, double-blind pilot study compared reboxetine plus celecoxib with reboxetine plus placebo and found a significant therapeutic effect of celecoxib (15). One interesting finding of this study was that the kynurenine/tryptophan ratio, which reflects the activity of indoleamine 2,3-dioxygenase (IDO), a pro-inflammatory, cytokine-driven enzyme, predicted the antidepressant response to celecoxib; that is, celecoxib had better effects on patients with a high level of IDO activity (16). A double-blind, randomized controlled trial (RCT) in MD (n = 50 patients) compared fluoxetine plus celecoxib with fluoxetine plus placebo and found a significantly better outcome in the group receiving adjunctive celecoxib (17). Similar results were found in two studies of sertraline plus celecoxib or placebo in MD (n = 40 and n = 30) (18, 19), where Hamilton Depression Rating Scale scores decreased significantly more in the celecoxib group; in one of the studies (18), serum IL-6, a pleiotropic immune-activating cytokine that primarily promotes innate and B- and T-cellular immunity and plays an important role in inflammation, correlated with the decrease in the depression rating score.

The efficacy of adjunctive treatment with an NSAID in MD was evaluated in a meta-analysis of four celecoxib studies in a total of 150 patients (20). The analysis concluded that celecoxib may be a potential treatment in this disorder, although the authors stated the benefit and safety of celecoxib and other NSAIDs need to be confirmed in larger studies of longer duration (20).

The findings of another meta-analysis on inflammation-related therapeutic approaches in MD are also of great interest (21). This analysis evaluated data from 14 studies (10 on NSAIDs, n = 4,258; 4 on cytokine inhibitors, n = 2,004) and found that the anti-inflammatory treatments had positive effects compared with placebo [standardized mean difference (SMD), −0.34; 95% confidence interval (CI), −0.57 to −0.11; I ² = 90%], both in depression (SMD, −0.54; 95% CI, −1.08 to −0.01; I ² = 68%) and in depressive symptoms (SMD, −0.27; 95% CI, −0.53 to −0.01; I ² = 68%). The type of depression (clinical depression vs. depressive symptoms) or the agent (NSAID vs. cytokine inhibitor) did not explain the heterogeneity of the studies with respect to differences in treatment regimens and patient populations. Sub-analyses provided support for the positive effects of celecoxib (SMD, −0.29; 95% CI, −0.49 to −0.08; I ² = 73%) on both remission [odds ratio (OR), 7.89; 95% CI, 2.94 to 21.17; I ² = 0%] and response (OR, 6.59; 95% CI, 2.24 to 19.42; I ² = 0%). Six of the studies reported adverse effects but found no difference in the rate of gastrointestinal or cardiovascular side effects at 6 weeks or infections at 12 weeks between the active treatments and placebo. The meta-analysis suggested that treatment with an anti-inflammatory agent, particularly celecoxib, can ameliorate symptoms of depression but does not carry a higher risk of side effects (21). Other studies, however, reported higher rates of adverse cardiovascular events with COX-2 inhibitors (22).

ASA has also been studied in depression. In an 8-week RCT in MD, patients were assigned to 160 mg of aspirin add-on to sertraline (n = 50) or placebo add-on (n = 50); the groups were matched for age, gender, and severity of depression (23). After 4 and 8 weeks of treatment, depression scores were significantly lower than at baseline only in the sertraline plus aspirin group. These results indicate that aspirin has positive effects on depression, although the unusually low responder rate in the sertraline (plus placebo) group may indicate that treatment resistance affected the results. A complex, placebo-controlled study of patients with bipolar depression found a significantly higher response rate in the aspirin group; however, the comparison of the mean values found no significant difference between aspirin and placebo (24).

Another recent meta-analysis included additional studies on NSAIDs and differentiated between studies on patients with a diagnosis of MD and patients with “classical” inflammatory diseases, such as arthritis or psoriasis, who also had depressive symptoms (25). The analysis found highly significant effects on both MD and depressive symptoms and a highly significant overall effect on the combined analysis of both indications. In the analysis of different substance classes and mechanisms of action of the anti-inflammatory agents, classical NSAIDs were significantly superior to placebo. Other anti-inflammatory compounds, such as cytokine inhibitors, glucocorticoids (two RCTs), and minocycline, also showed a significant advantage than did placebo. The overall effect of all anti-inflammatory substances was p = 0.00001.

Table 1 gives an overview of studies on selective COX-2 inhibitors and ASA in MD. Despite the limitations of the studies described above, they provide important information on the effects of anti-inflammatory treatment and, in particular, COX-2 inhibition in MD. Additional studies are needed in larger samples. Furthermore, studies need to consider the high placebo response, particularly in add-on studies with an effective antidepressant, and the severity of depression.

Tumor necrosis factor-alpha (TNF-a) promotes the activation of the innate and adaptive immune response and is a key molecule in inflammation. The anti-TNF-a antibody infliximab prevents the cytokine TNF-a from interacting with receptors on the cell surface and has an anti-inflammatory effect. TNF-a was initially designed as a treatment for inflammatory joint disorders and psoriasis but was then found to significantly improve symptoms of depression in patients with psoriasis (30).

In a random-effect meta-analysis of seven placebo-controlled RCTs, anti-cytokine treatment had significantly greater effects on depressive symptoms than had placebo (anti-cytokine drug: n = 1,309; placebo: n = 1,061; SMD = 0.40, 95% CI, 0.22 to 0.59) (31). Five of the seven studies were on anti-TNF-a agents, that is, adalimumab, etanercept, and infliximab (SMD = 0.33; 95% CI, 0.06 to 0.60). Similar, small-to-medium effect sizes for anti-cytokine therapy were obtained in the analyses of the two RCTs on adjunctive anti-cytokine treatment (SMD = 0.19; 95% CI, 0.00 to 0.37) and the eight non-randomized and/or non-placebo studies (SMD = 0.51; 95% CI, 0.34 to 0.67). The three anti-TNF-a drugs and the anti-IL-6 antibody (tocilizumab) significantly improved symptoms of depression. A meta-regression showed that baseline symptom severity was a predictor of antidepressant effect (p = 0.018), but sex, age, study duration, and improvement in the primary physical illness were not. An important limitation of these studies was that besides having an inflammatory disease, such as atopic dermatitis, psoriasis, Crohn’s disease, or rheumatoid arthritis, the patients had concomitant symptoms of anxiety or depression or both. In addition, their overall symptoms of depression were mild to moderate, and they did not have a formal diagnosis of depression.

To my knowledge, only one 12-week, placebo-controlled study has evaluated an anti-TNF-a antibody in treatment-resistant MD (n = 60). The participants, who were either taking an antidepressant (n = 37) or partly medication free (n = 23), received three infusions of infliximab (n = 30) or placebo (n = 30). Infliximab was not superior to placebo, but the study found a significant interaction between time, treatment, and baseline levels of CRP (≤5 mg/L); that is, the response rate to infliximab (62%) was higher than that to placebo (33%) in patients who had higher CRP levels at baseline. In addition, participants who responded to infliximab had significantly higher baseline concentrations of TNF-a, sTNFR1, and sTNFR2 (p ≤ 0.01) and significantly larger decreases in CRP (p ≤ 0.01) than had those who did not respond (32). This study is encouraging, in particular because treatment-resistant patients with MD represent a negative selection for treatment outcome, and it indicates that CRP—the most widespread clinical marker for inflammation—may be a biomarker of anti-TNF-a antibody treatment outcome.

As mentioned above, studies have found that in patients with depression, IL-6 levels in the peripheral blood and cerebrospinal fluid (CSF) are higher. Consequently, the IL-6 complex has been proposed as a target for anti-cytokine treatment. The above-mentioned random-effect meta-analysis of seven RCTs (31) included two open studies of the anti-IL-6 antibody tocilizumab (33, 34), both of which showed improvements in patients with concomitant anxiety and depression. Despite these findings, we still need valid, reliable data on the effects of anti-IL-6 treatment in MD and the correct target for such treatment. Because higher levels of IL-6 are found in the CSF than in the blood and in patients with depression than in controls, IL-6 in the CSF was hypothesized to be the most promising therapeutic target. Further support for this hypothesis was given by the finding that IL-6 is overexpressed in the pre-clinical model of chronically stressed rats (35). However, IL-6 levels in the periphery are not completely independent of those in the CSF or brain but are closely connected to them. For this reason, researchers have proposed that targeting peripheral IL-6 may be a potential treatment approach in depression (35). This proposal is supported by the findings in an animal model of depression that significantly higher serum IL-6 levels were found in learned helplessness rats, which were susceptible to chronic inescapable electric stress, than in control and non-learned helplessness rats, which were resilient to the same stress (36). Supporting evidence was also provided by the finding that in depression, serum levels of IL-6 predict patients’ response to ketamine (37). As an alternative, more promising strategy to inhibit IL-6, Maes et al. (38) proposed that administering soluble glycoprotein 130 (sgp 130), a component of the IL-6 complex, would increase inhibition of IL-6 trans-signaling and consequently maintain IL-6 receptor (IL-6R) signaling. In addition to tocilizumab, sirukumab, another monoclonal antibody to IL-6, has also been proposed as a potential treatment for depression (39). It acts on the signaling pathway of IL-6 and can inhibit its pro-inflammatory and anti-inflammatory effects; studies have shown beneficial effects of sirukumab in the inflammatory diseases lupus erythematosus and rheumatoid arthritis, among others (39).

Microglia cells act as macrophages in the brain and are important in inflammatory processes in the CNS. Positron emission tomography (PET) studies showed activation of microglia in MD (40, 41). Minocycline, an antibiotic, can cross the blood–brain barrier into the CNS, where it inhibits the activation of microglia. Consequently, it was proposed as a potential treatment in MD. A recent meta-analysis identified 18 clinical studies (a case report, RCTs, and open and ongoing trials) on minocycline in depression; however, only three RCTs (in a total of 158 participants) were suitable for analysis (42). The analysis observed a large, statistically significant antidepressant effect of minocycline compared with placebo, although limitations of the study included the small number and sample size of the trials included, their heterogeneity, and potential publication bias.

Although statins are primarily used to lower cholesterol, they also have direct anti-inflammatory effects (43, 44). For example, a population-based study in users of selective serotonin reuptake inhibitors (SSRIs; N = 872,216), n = 113,108 of whom were also taking a statin, found that the risk for (a relapse to) depression was clearly lower in participants using both SSRIs and statins than in those using an SSRI alone (43).

Perhaps the most convincing evidence for an involvement of inflammation in schizophrenia is provided by the finding that anti-inflammatory medication is useful in schizophrenia (see Table 2 ). Noteworthy in this context is the paper by Sommer et al. (50), which reviews the effects of various anti-inflammatory treatments, including celecoxib, in schizophrenia.

My group performed a 6-week double-blind RCT in acute schizophrenia to compare risperidone plus celecoxib with risperidone alone and risperidone plus placebo (54). Outcome was significantly better in the celecoxib add-on group (n = 25) than in the group receiving risperidone alone (n = 25) (54), particularly regarding cognition (58). Data from this study and another 6-week risperidone/celecoxib study were pooled (total N = 90), and the analysis found a benefit of celecoxib add-on if the illness duration was 24 months or less but no such benefit in more chronic cases. A large study that included a broad spectrum of schizophrenia patients found no advantage of celecoxib over placebo (57), whereby the disease duration of up to 10 years in many patients, some of whom had chronic schizophrenia, may have contributed to the negative result. Another study in chronic schizophrenia showed no advantage of add-on celecoxib in patients with chronic schizophrenia (56), providing further support for differential effects of COX-2 inhibitors in acute and chronic phases of the disease. To further test this hypothesis, my group performed a double-blind study of amisulpride with adjunct celecoxib or placebo in first-episode schizophrenia (55); the celecoxib group showed more improvement than did the placebo group on the positive, negative, total, and general psychopathology scores of the Positive and Negative Syndrome Scale (PANSS) (55, 59).

Similar benefits have been found for ASA in schizophrenia spectrum disorders. A double-blind RCT studied 70 inpatients and outpatients with schizophrenia spectrum disorders treated with antipsychotics and add-on aspirin 1 g/day or placebo (53). The total and positive PANSS scores decreased significantly more in patients taking aspirin add-on than in those taking placebo, and the more altered the immune function, the greater the effect on the total PANSS score. Another double-blind, placebo-controlled study evaluated 6-week adjunctive treatment with ASA in 60 inpatients with schizophrenia (52): In addition to their antipsychotic, one group (n = 20) received aspirin 325 mg/day; one (n = 20), aspirin 500 mg/day; and one (n = 20), placebo. At the end of the study, the PANSS positive and negative symptom scores and general psychopathology score were significantly lower in the aspirin groups than in the placebo group.

A meta-analysis of data from n = 264 patients from five double-blind, randomized, placebo-controlled studies on NSAIDs (celecoxib, four studies; ASA, one study) in schizophrenia found that the drugs had significant positive effects on total symptom severity and positive and negative symptom severity (60). A meta-analysis of data from n = 774 patients from eight studies (celecoxib, six; ASA, two) found that the drugs had significant effects only in patients with a first manifestation of schizophrenia and not in those with chronic schizophrenia and in inpatients but not outpatients (61).

PET studies have shown that microglia cells are also activated in schizophrenia (62). As mentioned in the section Other Immune-Related Substances, the antibiotic minocycline can inhibit microglia activation, and studies in animal models of schizophrenia found that minocycline improves cognition (63). Double-blind, placebo-controlled, add-on studies of minocycline in schizophrenia patients found positive effects on negative and cognitive symptoms (64), and case reports described benefits for symptoms overall (65).

Other potential anti-inflammatory agents have shown some benefit in schizophrenia. For example, a meta-analysis of 26 double-blind RCTs found significant effects for N-acetylcysteine and estrogen (50). Positive effects were also found for interferon-gamma (IFN-γ), a cytokine responsible for the monocytic type 1 immune response (66). IFN-γ can have severe side effects, such as hallucinations, seizures, and immunosuppression, so patients must be monitored carefully.

The initial studies on monoclonal antibodies against pro-inflammatory cytokines indicate that they may also be a potential treatment approach in schizophrenia, but more studies are needed (67).

Two interesting case reports shed light on a theoretical concept that is interesting but will probably never become an alternative for schizophrenia treatment: In one case, schizophrenia seems to have transferred by bone marrow transplantation to a leukemia patient (68); and in another contrasting case, schizophrenia was remitted by bone marrow transplantation to a patient with schizophrenia (69).