targeting the right patients matters

If you’ve ever sat in a conference on inflammation and depression, you’ll recognise the scene: a researcher presents another trial testing anti-inflammatory drugs for depression. The audience leans forward in hope, thinking “maybe this is the one”, but then the results slide appears with “no significant effect”. The collective sigh is almost audible.

For the past three decades, immunopsychiatry researchers (myself included) have been chasing the same target: can we treat depression by calming the immune system? We’ve seen promising trials with compounds such as cytokine inhibitors, minocycline and celecoxib, but the results have been inconsistent and often underwhelming (Hasselman, 2014). It’s like throwing darts in the dark and wondering why we keep missing the bullseye.

The problem isn’t that the immune-hypothesis of depression is wrong. Depression is in fact a cluster of biological subtypes with around a quarter to a third of patients showing signs of low-grade inflammation, typically measured by elevated C-reactive protein (CRP) or interleukin-6 levels. This subgroup often presents with a distinct clinical profile: profound fatigue, psychomotor slowing, hyperphagia, hypersomnia and motivational anhedonia (i.e. loss of motivation or desire to engage in enjoyable or rewarding activities). Both cross-sectional and longitudinal evidence linking inflammation to this phenotype is strong and compelling (Osimo et al., 2020; Wessa et al., 2025).

The problem is that most trials have not been designed with this biology in mind. Too often, we’ve ignored the three core pillars of precision medicine: aiming at the right population, confirming target engagement and measuring the most relevant outcomes (Miller et al., 2025). Not every person with depression has immune activation, yet most trials treated them as if they did, leading to the fact that those who showed signs of inflammation were buried in the averages. And crucially, standard depression rating scales may not capture changes in motivation and reward processing: core symptoms that inflammation seems to affect most.

Inflammation plays a key role in depression, but most anti-inflammatory trials to date have failed to target the right patients and outcomes leading to null, conflicting, or confusing results.

Methods

Mac Giollabhui and colleagues conducted a systematic review and meta-analysis of randomised controlled trials (RCTs) that tested pharmacological anti-inflammatory treatments in adults with depression. Studies included adults with (sub)clinical depression and either (i) evidence of an inflammatory phenotype (CRP ≥2 mg/l), or (ii) information on outcomes for this high-inflammation subgroup. Eligible trials also had to measure depressive symptoms and/or anhedonia, using either clinician-rated or self-report scales.

A medical librarian designed and ran a comprehensive search across multiple databases with no restrictions on publication date. All stages of the review followed PRISMA standards, meaning study selection, data extraction and quality assessment were done transparently and systematically.

Results

In total, 19 RCTs met the inclusion criteria, with 14 studies contributing data to the quantitative meta-analysis. The authors contacted study authors twice to request additional information, and studies were only excluded when data could not be obtained after these attempts.

Across all included trials, more than 900 participants were studied. Trial durations ranged from 2 to 12 weeks and most studies tested anti-inflammatory drugs as an add-on to standard antidepressants, rather than as stand-alone treatments. When the analysis was restricted to studies using a CRP cut-off ≥2 mg/l, anti-inflammatory treatment was found to be associated with statistically significant improvements in both central outcomes:

Anhedonia: 4 studies, 163 participants; Hedges’ g = 0.40 (95% CI 0.08 to 0.71, p = 0.01).

Depression severity: 11 studies, 321 participants; Hedges’ g = 0.35 (95% CI 0.05 to 0.64, p = 0.02).

In plain terms, this means that people with depression and elevated inflammation experienced small-to-moderate symptom improvements when treated with anti-inflammatory medication compared with placebo. To put that in context, these effects are roughly in the same range as what we see with standard antidepressants such as SSRIs. However, the added benefit here was that this was observed in an inflammatory subgroup, which is often considered more difficult to treat.

Despite these symptom improvements, anti-inflammatory treatment did not significantly increase rates of full treatment response or remission:

Treatment response (≥50% symptom improvement): 49% on anti-inflammatory treatment vs 41% on placebo; RR = 1.27, p = 0.05.

Remission: 23% vs 21%; RR = 1.20, p = 0.46.

Interestingly, the significant improvements that were seen in anhedonia align with experimental evidence showing that inflammation interferes with the brain’s dopamine system, making it harder to feel motivated or rewarded. So while these treatments may not push everyone into full remission, they may help restore drive, energy and pleasure.

Analysis of 19 trials and >900 participants showed that targeting inflammation can reduce anhedonia and depressive symptoms, but participants did not reach full remission.

Conclusions

The mixed and often disappointing results of previous trials appear to reflect heterogeneity and imprecision in study design, rather than a failure of the immune hypothesis itself. The authors conclude:

anti-inflammatory agents are safe and effective in reducing depressive symptoms and anhedonia in inflammatory depression.

While the meta-analysis shows that anti-inflammatory drugs can reduce depressive symptoms and anhedonia in people with measurable inflammation, they’re not a cure-all. Their benefits appear strongest when:

The right patients are identified – those with evidence of inflammation.
Biological target engagement is confirmed – .i.e., the right drug is chosen for the task, successfully lowering inflammation via pathways known to cause depression.
The outcomes chosen to measure treatment success align with immune-related mechanisms.

Precision is not the future of immunopsychiatry, it is its prerequisite.

Precision is not the future of immunopsychiatry, it is its prerequisite.

Strengths and limitations

This meta-analysis is clearly grounded in a precision psychiatry framework, even if the authors do not explicitly label it as such. By restricting analyses to participants with biological evidence of inflammation, the review addresses an important limitation that has plagued the trials in the immunopsychiatry field: treating depression as a homogeneous condition. This review asks the right question in the right patient population, which already sets it apart from all previous meta-analyses in anti-inflammatories for depression.

Another strength is that the authors did not solely rely on what was reported in the published papers. When subgroup data were missing, they contacted authors twice and re-analysed data where possible.

That said, the sample sizes are small, particularly for anhedonia outcomes (only four trials, N=163), which is striking in itself given the importance of anhedonia in (immune-driven) depression. Additionally, most included trials did not prospectively select patients based on inflammation; instead, inflammatory subgroups were identified post hoc, often representing only a minority of the original sample.

In addition, the pooled effects for depressive symptom severity appear to be driven disproportionately by small trials with large effect sizes, while the largest studies reported minimal or negligible effects. This pattern raises concerns about small-study effects and potential publication bias, supported by some asymmetry in funnel plots for depressive symptoms.

Finally, by excluding trials involving inflammatory somatic conditions such as rheumatoid arthritis, the authors improved internal validity but at the cost of external validity. These excluded populations may in fact be those in whom anti-inflammatory treatments show the most robust antidepressant effects.

Overall, this is a carefully conducted and conceptually strong review. Its conclusions should be seen as promising rather than definitive. It convincingly shows how much methodological refinement is needed to fully capture the promise of immunopsychiatry treatment approaches.

This meta-analysis asks the right question in the right patients, but small samples and post-hoc analyses limit the promising but indefinite conclusions.

Implications for practice

Let’s go back to that conference room and the sigh-inducing “No significant effect” slide. This meta-analysis reinforces the idea that those moments of disappointment may not reflect a failure of the immune hypothesis of depression, but rather a failure of how we’ve been testing it. The conclusion resonates strongly with my own research experience in immunopsychiatry and with ongoing work I am involved in; arguing that progress in this field depends on improving precision in study design (Worrell et al. 2025).

For clinical practice, the message is not that anti-inflammatory drugs should suddenly replace antidepressants. Instead, it is a reminder that depression is biologically heterogeneous and that some patients are likely being treated with tools that don’t match their underlying biology. A subgroup of patients with elevated inflammation may respond differently to treatment. While CRP testing is not yet routine in mental health care, it is inexpensive, widely available and already used in other areas of medicine. Incorporating it into assessment could help clinicians begin to think more mechanistically, even if treatment options remain limited for now.

Importantly, strong and consistent effects were seen for anhedonia. Clinically, this matters. Improvements in motivational and reward responsiveness may not immediately translate into remission on standard rating scales, but they can meaningfully change how someone functions in daily life and how they engage with psychological and social interventions.

For researchers, the implications are clearer. Trials that continue to recruit unselected “all-comers” with depression, fail to demonstrate target engagement of the compounds being tested, and rely solely on broad symptom totals, are likely to keep producing the same disappointing conference slides.

Future studies need to:

recruit biologically enriched samples,
demonstrate target engagement,
prioritise mechanistically relevant outcomes.

Otherwise, we risk continuing to throw darts in the dark and mistaking methodological imprecision for biological failure.

Perhaps the most hopeful implication is this: the mixed results of the past 20-30 years were not wasted effort. They were necessary steps in learning where to look and how to look, paving the way toward precision psychiatry. We can now begin to make better use of the abundance of existing data, as demonstrated in this meta-analysis, to inform the design of new studies and more rigorous clinical trial recommendations, potentially extending beyond depression alone. This should give both clinicians and researchers reason to lean forward again with renewed optimism.

Future studies need to recruit biologically enriched samples, demonstrate target engagement, and prioritise mechanistically relevant outcomes. Otherwise, we risk mistaking methodological imprecision for biological failure.

Statement of interests

Céline Wessa is a PhD fellow of the Research Foundation, Flanders (FWO) (grant number 1S24925N) and conducts research in the field of immunopsychiatry and precision psychiatry. She has no personal involvement in the study reviewed in this blog and has no financial interests related to the interventions discussed. She has professional collaborations within the immunopsychiatry research community as reflected in cited work, but these did not influence the content of this blog.