Effect of anti-inflammatory treatment on depressive symptom severity and anhedonia in depressed individuals with elevated inflammation: Systematic review and meta-analysis of randomized controlled trials

Abstract

Studies evaluating the effect of anti-inflammatory treatment on depressive symptom severity and anhedonia in depressed individuals report mixed results. In this pre-registered systematic review and meta-analysis (Prospero: CRD42023472023), we evaluated whether anti-inflammatory treatments, compared to placebo, reduce anhedonia and depressive symptom severity in depressed individuals with an inflammatory phenotype. We included randomized controlled trials of pharmacological anti-inflammatory treatments that assessed anhedonia or depressive symptom severity and recruited depressed individuals with an inflammatory phenotype or measured baseline inflammatory biomarkers that permitted post-hoc analysis. A search was conducted in February 2025 of Ovid MEDLINE, Embase, Web of Science Core Collection, Cochrane Central Register of Controlled Trials, ClinicalTrials.gov and PsycINFO. Multiple reviewers independently applied criteria and discrepancies were resolved via consensus. Two reviewers independently extracted data and cross-checked for errors. In RCTs (k=11) using an established cut-off for elevated inflammation (CRP≥2mg/L), anhedonia (Hedge’s g=.40, 95%CI=[.08, .71], p=.013) and depressive symptoms (Hedge’s g=.35, 95%CI=[.05, .64], p=.022]) were reduced, but no differences in treatment response (Relative Risk (RR)=1.28, 95%CI=[.997, 1.64], p=.053) or remission rates (RR=1.18, 95%CI=[.71, 1.95], p=.52) were observed. Results did not vary by clinical, intervention, or demographic characteristics. Anti-inflammatory treatments may be safe and effective at reducing depressive symptoms and anhedonia in depressed individuals with heightened inflammation. Not accounting for inflammatory status may help explain prior mixed findings.

INTRODUCTION

Depression is the leading cause of the mental health-related global disease burden, affecting 400 million individuals globally and accounting for 40% of total days lost to poor mental health (1, 2). Approximately 65% of individuals will not achieve remission after a first-line antidepressant treatment and one third will not remit following multiple treatments (3). Novel efficacious antidepressant treatments are urgently needed.

Convergent evidence based on genetic and multiomics studies (4, 5), experimental research (6), and observational studies (7–11) suggest that a subtype of depression – inflammatory depression – is characterized by dysregulated immune physiology and occurs in ~25% of cases [for reviews, see (12, 13)]. Circulating levels of inflammatory biomarkers (e.g., C-reactive protein [CRP], an acute phase protein synthesized by the liver in response to tissue damage/pathogens) are: routinely elevated in depressed individuals when compared to healthy controls, highly associated with levels of other inflammatory mediators in the peripheral blood and cerebrospinal fluid (14), and predictive of future depression (7). Moreover, experimental induction of an innate immune response is reliably associated with onset of depressive symptoms (i.e., total depression score) (15) and inhibiting inflammatory signaling in patients with inflammatory disorders reduces depressive symptoms above and beyond treatment-related changes in physical health (16). There is emerging evidence that dysregulated inflammatory physiology is associated with a specific clinical presentation characterized by neurovegetative symptoms and anhedonia (12, 17).

Meta-analyses have previously found that anti-inflammatory treatments are safe and reduce depressive symptoms in individuals experiencing comorbid medical conditions (e.g., rheumatoid arthritis) (18–20). Considerably less is known about the efficacy of anti-inflammatory treatments in depressed individuals without comorbid medical conditions, and initial randomized controlled trials (RCTs) have produced mixed results (18–24). A recent review highlighted two potential causes of mixed results (25): 1) many RCTS have not recruited depressed individuals who exhibit an inflammatory phenotype (frequently indexed as CRP levels greater than 2 or 3mg/L) (26–28) and are therefore most likely to respond to anti-inflammatories (21, 23, 29); and 2) many RCTs do not assess a dimension of depressive symptomatology that is particularly associated with inflammation: anhedonia. Data support a cause-and-effect relationship between increased immune activation and anhedonia (12); however, null results are also observed (30) and focus on the effect of anti-inflammatory agents on anhedonia in RCTs for depression is relatively sparse (25). Experimental human research has demonstrated that a lipopolysaccharide inflammatory challenge differentially increases anhedonia in depressed individuals with CRP≥3mg (31). It should be noted that the association of inflammation and anhedonia is overwhelmingly based on measures of consummatory pleasure, and less is known about inflammation’s impact on motivational anhedonia [for reviews on: anhedonia in depression, see (32, 33); inflammation and anhedonia, see (34, 35)]. Thus, further work is needed to determine whether anti-inflammatory treatment is effective in inflammatory depression and whether we can observe a stronger effect of anti-inflammatory agents on anhedonia compared to overall depressive symptom severity.

This systematic review and meta-analysis is the first to examine the effect of anti-inflammatory medications, compared to placebo, on anhedonia and depressive symptom severity in depressed individuals exhibiting an inflammatory phenotype and without medical comorbidity. Secondary outcomes include: treatment response, remission status, mortality, and report of serious adverse events at study endpoint. Potential demographic and clinical/intervention moderators are also investigated.

METHODS

This protocol was pre-registered with Prospero (CRD42023472023) and adheres to PRISMA 2020 guidelines (see Supplementary Table 1 for adherence details) (36). The funders had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Eligibility Criteria

Studies were included if they: 1) were RCTs; 2) measured anhedonia or depressive symptoms via clinician/self-report assessment; 3) were available in English; 4) recruited for clinical or subclinical depression; 5) recruited for inflammatory phenotype (e.g., CRP≥2mg/L) or measured baseline inflammatory biomarkers that permitted post-hoc analysis in depressed individuals with an inflammatory phenotype; and 6) administered a pharmacological anti-inflammatory treatment either as monotherapy or adjunctive therapy. Studies were excluded if: 1) not conducted in human adults; 2) data were unavailable in manuscript and authors did not provide data after two contact attempts; 3) the control condition had anti-inflammatory potential (e.g., sustained exercise); 4) participants were recruited on the basis of somatic disease (e.g., rheumatoid arthritis); 5) patients were undergoing a medical treatment that may cause depression or anhedonia.

Minor deviations from our registered protocol occurred: studies including a subset of individuals diagnosed with bipolar I/II were permitted, and studies were included even if participants were not on a stable dose of psychiatric medication for at least four weeks because this was effectively controlled for by comparison of placebo to treatment. When multiple manuscripts reported on the same RCT, the manuscript that was most aligned with our question of interest was used.

Information Sources & Selection Process

A medical librarian conducted an electronic search, unrestricted by language or dates, in December 2023 (updated in February, 2025) using Medline (Ovid), Embase.com (Elsevier), Core Collection (Web of Science), Cochrane Central Register of Controlled Trials (Ovid), PsycINFO (Ovid), and ClinicalTrials.gov. The search strategy incorporated controlled vocabulary and free-text synonyms and was designed to identify RCTs where medications that target immune function were delivered to depressed individuals exhibiting an inflammatory phenotype, as indexed by immune biomarkers. The full database search strategies are documented in Supplementary Material. A validated RCT filter was applied and authors (AAM, ELQ, NMG) reviewed remaining titles, abstracts, and full text manuscripts (37). At each stage, each study received independent ratings from two authors; discrepancies were resolved via consensus.

Data Collection

Two authors (NMG, AAM) independently extracted data and the resultant dataset was quality checked (ELQ). Primary outcomes of interest were change in 1) anhedonia and 2) total depressive symptom severity from baseline to study endpoint in both the placebo and intervention arms in both the placebo and intervention arms. When a study featured multiple intervention/control groups, a comparison was selected prioritizing a clear contrast of anti-inflammatory treatment versus placebo, the most potent anti-inflammatory condition and the comparisons with the largest sample size. Additional outcomes were: depression remission and response rates at study endpoint, prevalence of serious adverse events post-randomization, and change in inflammatory biomarkers. Definitions of remission and response as defined by individual studies were used and when analysis of raw data was conducted, response was defined as ≥50% in depressive symptom severity and standard cut-off for remission were used for each measure (e.g., Hamilton Depression Rating Scale ≤7). Target engagement (reduction of inflammatory biomarkers) was estimated as: consistent target engagement (consistent decrease across inflammatory biomarkers); inconsistent target engagement (decrease not consistent across inflammatory biomarkers), or no target engagement (no decrease in biomarkers); given the small number of studies, consistent and inconsistent target engagement were combined in analyses. When raw data were available/provided, we undertook analyses following best practice guidelines (38). When requesting information, first and last authors were contacted twice via email and/or social media. When available, intent-to-treat data was preferred above per-protocol analyses. Critical information extracted included: author; year of publication; clinical trial registration number; anti-inflammatory agent/dose/duration; monotherapy/augmentation; analytic sample size; demographic (e.g., age/sex/race); and clinical information [e.g., body mass index (BMI)/treatment resistance status/diagnostic status]; outcome measurements; and enriched inflammatory sample/definition.

Bias Assessment

Authors (NMG, AAM) independently used the Revised Cochrane risk-of-bias (RoB) tool for randomized trials 2.0 (39) to assess study quality; discrepancies were resolved via consensus. See Supplementary Table 2 for RoB assessment. For studies in which analyses were re-conducted by the authors, analyses were considered as pre-specified (Item 5).

Statistical Analysis

We used the R package ‘meta’ for analyses and visualization (40). For placebo and intervention arms, change in symptom severity from baseline to study endpoint was used for continuous measures and standardized mean difference estimated using Hedge’s G; binary and percentage outcomes were converted to proportion and examined in term of relative risk. Given studies’ methodological heterogeneity, we used random effects models to examine treatment-related differences in the outcomes (e.g., anhedonia, depressive symptom severity). Eleven studies either recruited those who had CRP levels of at least 2 mg/L or had baseline CRP levels available so that we could conduct analyses only among the cohort with CRP≥2mg/L. We chose CRP≥2mg/L as a more inclusive cut-off than CRP≥3mg/L that still fulfilled the critical criterion of representing clinically significant inflammation, as indicated by its reference in updated American Heart Association guidelines (41) and widespread use in clinical studies of (26, 42, 43). Cochran’s Q and the inconsistency index estimated statistical heterogeneity. Publication bias was visually assessed with a funnel plot and statistically assessed with Egger’s regression intercept test. Missing descriptive statistics were, as appropriate, estimated using Cochrane review guidelines (38). Secondary analyses explored (meta-regression for binary variables and subgroup analyses for categorical variables) whether the treatment’s effect depended on intervention characteristics [drug class, target engagement (i.e., whether inflammation decreased), monotherapy vs. add-on therapy, RoB status (high versus low/some concerns)] and sociodemographic characteristics (age, sex, race) as well as clinical characteristics [treatment resistant depression (TRD) vs. non-TRD, BMI].

RESULTS

Search results are presented in a PRISMA diagram (Figure 1). From the initial 11,069 results, 5,532 unique studies were screened for eligibility. Following title and abstract review, 180 papers were included for full-text review. An additional 163 studies were subsequently removed; aggregate reasons for exclusions are detailed in Figure 1. One additional relevant study was added following review of reference lists (29). Additionally, one already-included manuscript featured results from two studies (44). Thus, 19 studies were included in the systematic review (44–47) and 14 of these studies were included in the meta-analysis (21, 23, 24, 29, 48–57) – analyses primarily focus on 11 of these studies that used an established clinical cut-off for low-grade inflammation (CRP≥2mg/L).

Figure 1: Flow diagram of search and study selection.

This flowchart shows the number of studies excluded at each stage of the systematic review as well as the rationale for exclusion.

Key characteristics of the 19 RCTs are described in Table 1. These 19 RCTs evaluated the efficacy of anti-inflammatory medications, including: interleukin-6 inhibitors (k=2), MAPK inhibitors (k=2), minocycline (k=5), nonsteroidal anti-inflammatories (NSAID) (k=5), recombinant interleukin-2 (k=1), and tumor necrosis factor (TNF) inhibitors (k=4). All RCTs compared anti-inflammatory drugs to placebo. Treatment duration ranged from two to 12 weeks. Overall, a majority of RCTs in the review did not recruit for TRD (k=14) and used anti-inflammatories as adjunctive treatments (k=13). Seven out of 14 RCTs in the meta-analysis recruited those with an inflammatory phenotype (i.e., enriched sample), although there was substantial variation in terms of how that was defined. Eleven of 14 RCTs permitted analysis of a subgroup with an established cut-off point (CRP≥2mg/L).

Table 1.

Study characteristics of all clinical trials included in the systematic review.

Author, Year	In Meta	Medication	Dose	Duration (weeks)	Adjunctive or Monotherapy	N Analytic Sample	n Intervention	n Placebo	Age	% Female	% White	BMI	TRD	Depression Measure	Anhedonia Measure	Enriched or Subgroup	Inflammatory Subtype Definition
Raison et al., 2013	Yes	Infliximab	5mg/kg	12	Monotherapy	22	13	9	43.40	67.00	77.00	31.95	Yes	HDRS		Subgroup	CRP≥5mg/L
Nettis et al., 2021	Yes	Minocycline	200mg	4	Adjunctive	18	6	12	45.22	56.41	84.58	31.32	Yes	HDRS	SHAPS	Subgroup	CRP≥3mg/L
Simon et al., 2021	Yes	Celecoxib	200mg	6	Adjunctive	11	5	6	40.45	36.00		25.36	No	MADRS		Subgroup	CRP≥3mg/L
Husain et al., 2017	Yes	Minocycline	200mg	12	Adjunctive	20	10	10	37.60	45.00	0.00	24.72	Yes	HDRS		Subgroup	CRP≥5mg/L
Treadway et a., 2024	Yes	Infliximab	5mg/kg	2	Monotherapy	38	20	18	39.03	88.10	42.90	33.46	No	HDRS	MDFI-RM	Enriched	CRP≥3mg/L
Abbasian et al., 2022	Yes	Adalimumab	5mg/kg	6	Adjunctive	30	15	15	35.24	56.66			No	HDRS		Enriched	CRP≥3mg/L
Hasebe et al., 2021	Yes	Minocycline	200mg	12	Adjunctive	27	14	13	49.48	66.21		29.70	No	MADRS		Subgroup	IL-6 Median Split
Inamdar et al. Study 1, 2014	No	Losmapimod	7.5mg	6	Monotherapy								No
Inamdar et al. Study 2, 2014	No	Losmapimod	7.5mg	6	Monotherapy								No
Kavakbasi et al., 2024	Yes	Celecoxib	400mg	6	Adjunctive	39	19	20	51.51	71.62		33.28	No	HDRS		Enriched	CRP≥3mg/L
Musil et al., 2011	No	Celecoxib	400mg	6	Adjunctive								No
Poletti et al., 2024	Yes	Aldesleukin	1 M IU	8.57	Adjunctive	11	4	7	43.64	63.40	100.00	26.56	No	MADRS		Subgroup	CRP≥2mg/L
Abbasi et al., 2012	No	Celecoxib	400mg	6	Adjunctive								No
Dodiya et al., 2022	No	Aceclofenac	200mg	12	Adjunctive								No	HDRS		Subgroup	IL-6≥6pg/ml
Miller et al., 2016	Yes	Infliximab	5mg/kg	2	Monotherapy	21	11	10	38.60	86.40	50.00		No	IDS-SR	SHAPS	Enriched sample	CRP≥3mg/L
Attwells et al., 2021	Yes	Minocycline	200mg	8	Adjunctive	21	12	9	51.43	71.42	76.19	24.87	Yes	HDRS		Enriched sample	TSPO in PFC, ACC, insula
Hamilton et al., 2017	Yes	Tocilizumab	162mg	4	Monotherapy	7	4	3					No	MADRS		Enriched sample	IL-6>1.5ng/L
Hellmann-Regen et al., 2022	Yes	Minocycline	200mg	6	Adjunctive	25	14	11	48.08	56.00	94.64	30.98	Yes	HDRS		Subgroup	CRP≥3mg/L
Janssen, 2015	Yes	Sirukumab	50mg	12	Adjunctive	86	42	44	44.70	77.20	89.60		No	HDRS	SHAPS	Enriched	CRP≥3mg/L

In Meta=Study included in meta-analysis;HDRS=Hamilton Depression Rating Scale; SHAPS=Snaith-Hamilton Pleasure Scale; MADRS=Montgomery-Asberg Depression Rating Scale; IDS-SR=Inventory for Depressive Symptomatology-Self Report; MDI-RM=Multidimensional Fatigue Inventory-Reduced Motivation subscale; CRP=C-reactive protein; IL-6=Interleukin-6; TPSO=Translocator protein; PFC=Prefrontal Cortex; ACC = Anterior Cingulate Cortex; sample size and demographic information only reported for studies included in quantitative analysis

In studies using a cut-off of at least CRP≥2mg/L (k=11), anti-inflammatory treatment led to a reduction in anhedonia (Figure 2A; k=4, n=163; Hedge’s g=.40, 95%CI=[.08, .71], p=.013) and depressive symptom severity (Figure 2B; k =11, n=321; Hedge’s g=.35, 95%CI=[.05, .64], p=.022) when compared to placebo. No significant difference in the occurrence of serious adverse events were reported in the inflammatory and placebo groups (k=10; RR=1.08, 95%CI =[.35, 3.39], p=.89; I²=0%) and mortality was not reported in any study in either group. Anti-inflammatory treatment compared to placebo was not significantly associated with greater likelihood of treatment response (average weighted treatment response rate: anti-inflammatory condition = 49%, placebo condition = 41%; k=8; RR=1.28, 95%CI=[.997, 1.64], p=.053; I²=0%) or remission (average weighted treatment remission rate: anti-inflammatory condition = 23%, placebo condition = 21%; k=7; RR=1.18, 95%CI=[.71, 1.95], p=.64; I²=0%). Rates of serious adverse events, mortality, and discontinuation rates due to adverse events are provided in Supplementary Table 3.

Figure 2: Effect of anti-inflammatory agents on anhedonia and depressive symptom severity.

Forest plots displaying effect of anti-inflammatory treatments on anhedonia (2a) and depressive symptoms (2b).

In studies (k=14) using a larger range of cut-offs to define an inflammatory phenotype (e.g., median split) no difference across results was observed when compared to studies using a stricter cut-off (i.e., CRP≥2mg/L), although the magnitude of reduction in depressive symptom severity was lower (k =14, n=376; Hedge’s g=.26, 95%CI=[.01, .51], p=.039; I²=23.3%). Similarly, when using a more conservative range to define an inflammatory phenotype (i.e., CRP≥3mg/L), no difference across results was observed when compared to studies using more liberal cut-offs (e.g., CRP≥2mg/L), although the magnitude of reduction in depressive symptom severity was lower (k =10, n=310; Hedge’s g=.28, 95%CI=[.01, .55], p=.04; I²=22.6%). All studies reporting on anhedonia used a cut-off of at least CRP≥3mg/L.

Complete meta-regression and sub-group analyses are reported as Supplementary Material and information on additional demographic/clinical information extracted from studies are provided in Supplementary Table 4. Treatment resistant status (k=11; Q =.99, p=.32, I²=34.56%: TRD Hedge’s g =.60 95%CI =[−.04, 1.25], I²=49.20%; non-TRD Hedge’s g=.23 95%CI =[−.08, .55], I²=23.50%); administration as adjunctive treatment (k=11; Q =.66, p=.42, I²=39.03%: adjunctive Hedge’s g =.15 95%CI =[−.29, .59],, I²=0.00%; monotherapy Hedge’s g=.47 95%CI =[.05, .88], I²=51.10%); and drug class (k=11; Q =1.59, p=.45, I²=39.80%): cytokine modulator Hedge’s g =.32 95%CI =[−.04, .68], I²=30.90%); minocycline Hedge’s g=.71 95%CI =[−.21, 1.64], I²=65.30%); NSAID Hedge’s g=.04 95%CI =[−.51, .60], I²=0.00%); or level of target engagement (k=10; Q =.40, p=.53, I²=40.11%): consistent/inconsistent target engagement =.49, 95%CI =[.09, .90], I²=25.10%); no target engagement Hedge’s g=.29 95%CI =[−.38, .97], I²=55.10%) did not modulate the treatment’s effect when compared to placebo.

RCTs exhibited ‘Low Risk’ (n=4), ‘Some Concerns’ (n=8), and ‘High Risk’ (n=7) when evaluated using version 2 of the Cochrane ROB tool. Visual inspection of funnel plots did not provide evidence of asymmetrical results for anhedonic symptoms; however, results appeared somewhat asymmetrical for depressive symptoms with two notably large effects from small trials – see Supplementary Figure 1. Egger’s regression test could not be performed for anhedonic symptoms and was not statistically significant in the sample using: a cut-off of CRP≥2mg/L, t(9)=2.11, p=0.065; CRP≥3mg/L, t(8)=1.38, p=0.21; or using a wider range of cut-off, t(12)=1.61, p=0.13.

DISCUSSION

This is the first known meta-analysis examining the effect of anti-inflammatory treatment, when compared to placebo, on anhedonia and depressive symptom severity among depressed individuals with high levels of inflammation and without somatic disease (i.e., inflammatory depression). We found that anti-inflammatory treatment, when compared to placebo, reduced anhedonia (Hedge’s g=.40, 95%CI=[.08, .71]) and depressive symptoms (Hedge’s g=.35, 95%CI=[.05, .64]) in studies that used an established clinical cut-off for inflammation (CRP≥2mg/L) – an effect size substantially higher than when ad hoc cut-offs for inflammation were employed (Hedge’s g=.26, 95%CI=[.01, .51]), but one potentially driven by large effect sizes observed across a modest number of small sample RCTs. No difference in the prevalence of serious adverse events for anti-inflammatory treatments versus placebo was observed. Treatment response and remission rates were not increased in individuals receiving anti-inflammatory treatment, although response rates were numerically and substantively higher. These results should be considered preliminary given the heterogeneity of effect sizes for depressive symptom severity and the small numbers of studies/sample sizes, but they indicate that anti-inflammatory agents are safe and effective at reducing depressive symptoms and anhedonia in inflammatory depression.

That anti-inflammatory treatment is effective at reducing symptoms of anhedonia and depressive symptom severity in depression is aligned with some results from prior RCTs in both unipolar and bipolar depression (29, 58). However, it is important to note that there was considerable heterogeneity in the effect of anti-inflammatory treatment on depressive symptom severity; indeed, the three largest studies (29, 52, 59) reported very small effect sizes (Cohen’s d <.12) and the observed effect size for depressive symptom severity was driven by a modest number of small sample RCTs. In contrast, the effect of anti-inflammatory treatment on anhedonia was larger and considerably more consistent when compared to overall depressive symptoms. A specific effect of anti-inflammatory treatment on anhedonia is consistent with data from animal and human studies across a range of modalities and research designs demonstrating the sensitivity of reward circuitry to peripheral pro-inflammatory immune signaling (34, 35, 60–62). However, most RCTs (~70%) did not examine the effect of anti-inflammatory treatment on anhedonia and the most common measure of anhedonia used (i.e., Snaith-Hamilton Pleasure Scale) primarily assesses consummatory pleasure to the exclusion of other important dimensions (i.e., anticipatory anhedonia, motivation) implicated in depression (63). Thus, further work is needed that: confirms the potential of anti-inflammatory treatment to target anhedonia in depression; captures the multi-dimensional nature of anhedonia and can meet regulatory requirements needed to serve as clinical outcome assessments in FDA-sponsored trials (64); better characterizes other symptoms likely associated with inflammatory depression (fatigue, psychomotor slowing) (17); and that characterizes the multiple pathways leading to anhedonia in depression (63). Further, this pattern of results suggest that future studies may benefit from enriching for anhedonia as well as an inflammatory phenotype, as the two may better represent a sub-population likely to respond more strongly to anti-inflammatory treatment.

The effect of anti-inflammatory treatment on anhedonia and depressive symptom severity was substantially stronger in studies that enriched for high levels of inflammation using an established clinical cut-off. This aligns with prior studies (21, 23, 29) and reinforces the need for clinical trial designs that include subjects with increased inflammation (25). These effects were observed across studies using a CRP cut-off of 2mg/L [although a cut-off of CRP≥3mg/L is also commonplace (57)] and support the utility of CRP as a biomarker to enrich clinical trials. CRP is an affordable, accessible and sensitive measure of peripheral inflammation with established clinical cut-offs, and it is strongly correlated with neuroinflammation (65). However, CRP exhibits undesirable qualities: 1) it may not be as stable as often assumed (66); 2) its clinical cut-offs were established using a small, non-representative dataset (see Mac Giollabhui et al. for review) (27); 3) CRP is associated with adiposity in a sex-specific manner (67); and 4) CRP is highly non-specific in nature, capturing systemic immune activation rather than identifying pathogenic processes causally related to depression (68). Thus, even though CRP may possess utility as a stratification tool, an ideal biomarker would be more reliable, less confounded, and serve as a specific marker of the underlying immune-based mechanism.

Overall, these results did not differ in relation to a variety of clinical/intervention and demographic characteristics; however, statistical power to detect moderation was limited (≤11 studies). Effect sizes were numerically and substantively higher for studies with: lower risk of bias (Hedge’s g=.46, 95%CI:[−.20, 1.13])); TRD (Hedge’s g=.60, 95%CI:[−.04, 1.25]); specific anti-inflammatory agents (cytokine modulators Hedge’s g=.32, 95%CI:[−.04, .68]; minocycline Hedge’s g=.71, 95%CI:[−.21, 1.64]); and some level of target engagement (inconsistent/consistent target engagement Hedge’s g=.49, 95%CI: [.09, .90]). Variability in results may be partially explained by differences in the specificity/potency of anti-inflammatory medications deployed and resultant capacity to modulate immune function in depressed individuals. In addition to more reliable biomarkers of inflammatory depression and better characterization of the underlying immune-based mechanism, a central focus of immunopsychiatry must be identifying the appropriate medication (as well as dose and duration) that can target dysregulated inflammatory physiology and deliver therapeutic benefit in inflammatory depression (69–71).

This systematic review suggests that anti-inflammatory treatments may effectively reduce depressive symptoms and anhedonia among those with an inflammatory phenotype. It reinforces the importance of considering a precision medicine approach targeting depressive symptoms (72). Clinically, the risk of a non-specific approach (e.g., failure to directly target inflammation in a patient with depression in which inflammation is a primary etiological factor) may lead to the mistaken label of “treatment-resistant” depression. Anti-inflammatory treatments have potential to augment existing treatment strategies; however, gaps in fundamental knowledge must first be addressed. In addition to a dearth of well-powered studies recruiting on the basis of elevated inflammation, this review highlights multiple sources of heterogeneity in terms of: medications used (as well as dose and duration), enrollment inclusion/exclusion criteria, and target engagement. Better understanding the requisite dose and duration of a specific medication that will meaningfully reduce levels of systemic inflammation and exert anti-depressant effects is a pre-requisite to their integration in clinical practice – see recent reviews on potential integration of inflammatory depression within clinical practice and design of future clinical trials (64, 73).

Interpretation of these results should be considered alongside limitations. First, confidence intervals were broad and individual study sample sizes were small (particularly for anhedonia and in moderation analyses); consequently, the findings of this meta-analysis require replication using better-powered RCTs (74). Second, many studies had a high risk of bias. A common reason is that trials did not recruit for an inflammatory phenotype and/or results were based on post-hoc analyses. Third, the number of studies that reported data on anhedonia as assessed using reliable and validated measures was small and more work is needed to confirm these results. Further, the included studies did not recruit patients for anhedonia regardless of depression status, so we are unable to speculate as to whether the observed result generalizes to people with anhedonia who do not have depression. We also did not compare the effect of anti-inflammatory treatment on those with CRP≥2mg/L compared to those with CRP<2mg/L to show that this effect is specific to an inflammatory subtype of depression; yet, the current mixed literature in this domain suggests that failing to target this inflammatory phenotype dilutes the observed effect. Lastly, there was considerable heterogeneity across studies based on inclusion criteria (e.g., inclusion of bipolar depression), study design (e.g., trial length, medication type and statistical analyses (e.g., use of intent-to-treat analyses). Thus, moderator analysis should be considered as exploratory and, further, this study is unable to assess bias in effect sizes attributable to participant non-adherence. These limitations further highlight the need for additional high-quality RCTs evaluating anti-inflammatory treatments in inflammatory depression.

Taken together, our findings suggest that anti-inflammatory treatments may effectively treat depression and anhedonia in depressed patients who exhibit an inflammatory phenotype (e.g., elevated CRP). It highlights the continued relevance of inflammatory physiology as a potential cause of depression as well as a treatment target, as conceived within a precision medicine approach. This systematic review also provides guidance for the design of RCTs that will provide high-quality evidence to inform more precise and effective depression treatment.

Data Sharing Statement:

Relevant data and statistical code will be made available upon reasonable written request to corresponding author.