Two small trials tested borrowed drugs on depression, and neither claimed a win

Two randomised trials published in JAMA Psychiatry this summer gave depressed patients drugs developed for something else entirely — an interleukin-6 receptor antagonist used in rheumatology, and a GLP-1 receptor agonist used in diabetes and obesity. Both are small. Neither was designed to establish that its drug treats depression. Both are more interesting than that framing makes them sound, and both are easy to over-read.

The inflammation trial

The hypothesis that inflammation causes some depression has been circulating for years, largely on observational grounds: depressed patients have higher inflammatory markers on average. Observational association cannot establish direction, which is why a randomised test matters.

This was a 4-week, proof-of-concept, double-blind, parallel-arm, placebo-controlled trial [s1]. It recruited adults with moderate-to-severe ICD-10 depression, poor antidepressant response, low-grade systemic inflammation defined as high-sensitivity C-reactive protein of 0.3 mg/dL or above on two tests, and depression somatic symptoms scoring 7 or above on the Beck Depression Inventory II, between 2018 and 2022 [s1]. Participants received a single intravenous infusion of the IL-6 receptor antagonist tocilizumab at 8 mg/kg, to a maximum of 800 mg, or normal saline [s1].

Note the selection. Participants had to have measurable inflammation to enter. This is a trial of anti-inflammatory treatment in inflamed depressed patients, not in depressed patients generally, and its findings say nothing about the latter.

A total of 30 participants were randomised — mean age 41.1 years (SD 12.3), 24 of them female (80.0%) — 14 to tocilizumab and 16 to placebo, of whom 29 received the assigned infusion and completed follow-up [s1].

Thirty participants. The authors are explicit about what follows: as expected for a small proof-of-concept study, no results reached statistical significance [s1]. The primary outcome, somatic symptoms at day 14, showed little improvement, with an adjusted mean difference of −0.12 (95% CI, −2.51 to 2.28) [s1].

That confidence interval runs from meaningful benefit to meaningful harm. It is a study that did not answer its own primary question, which is what a proof-of-concept study of this size is expected to do.

What the authors report alongside it is a pattern rather than a result: greater stepwise improvement over time with tocilizumab in somatic symptoms, depression severity, fatigue, psychological symptoms, state anxiety and quality of life, with the largest effects at the final follow-up on day 28 [s1]. At that final follow-up, remission favoured tocilizumab — 7 participants (53.9%) against 5 (31.3%), number needed to treat 5 — as did response, 6 (46.2%) against 3 (18.8%), number needed to treat 4 [s1].

Those numbers look impressive and should be treated with real caution. They are secondary, they come from arms of 13 and 16 evaluable participants, and in a trial where nothing reached significance, a subset that appears favourable is exactly what chance produces. A number needed to treat of 4 calculated from six responders is a fragile quantity.

One finding is more portable than the efficacy signals. Baseline hs-CRP level, but not IL-6 level, tracked depression improvement, which the authors suggest means hs-CRP may better predict immunotherapy response in depression than drug-specific biomarkers [s1]. If that holds, it is a cheap, widely available test for selecting patients into the larger trial this one exists to justify.

Tocilizumab was well tolerated, with no serious adverse events or withdrawals [s1].

The GLP-1 trial

The second trial approaches depression through motivation rather than mood. Anhedonia and reduced motivation are core depressive symptoms that antidepressants often leave behind, and GLP-1 receptors are implicated in reward processing.

This was a 16-week, double-blind, placebo-controlled, parallel-group randomised trial of 72 participants with major depressive disorder and a body mass index of 25 or higher, randomised to oral semaglutide (n = 35) or placebo (n = 37), recruited from a university-based mood disorders programme in Toronto and enrolled between March 2022 and July 2024 [s2]. Semaglutide was given at 14 mg, initiated at 4 mg with a 4-week dose-escalation regimen, adjunctive to treatment as usual [s2].

The reported outcome here was the preregistered outcome of a secondary analysis: performance on the Effort-Expenditure for Rewards Task [s2]. That framing matters. This is not the trial's primary endpoint, and a preregistered secondary analysis is stronger than a fishing expedition but weaker than a primary result.

Semaglutide-treated participants showed increased willingness to exert physical effort as the expected reward rose, with a treatment × visit × expected value interaction of χ2 = 12.024 (P = .02) [s2]. Computational modelling attributed this to reduced effort discounting: sensitivity to effort was significantly reduced (β = −1.737; P = .03), while there was no treatment effect on sensitivity to probability (β = −0.776; P = .51) [s2].

The specificity is the most persuasive part. Effort sensitivity moved and probability sensitivity did not, which is harder to explain as a general improvement in mood, alertness or test-taking than a change in one parameter would be if everything had shifted together.

What the trial does not show is that patients felt better or functioned better. A laboratory task measuring willingness to squeeze a handle for money is a model of motivation, not motivation in a life. And the participants all had a body mass index of 25 or higher, so weight change is an obvious uncontrolled route to feeling more capable [s2].

What both trials have in common

Each takes a mechanism, selects patients in whom that mechanism should be active, and measures something close to the mechanism rather than the illness. That is good science and bad headline material, and the gap between the two is where these findings will be distorted.

Neither trial supports prescribing. Tocilizumab is an immunosuppressant with a serious infection risk profile, tested here as a single infusion in 30 people [s1]. The semaglutide finding is a secondary analysis of a laboratory task in 72 people [s2].

What to watch

Whether the large-scale efficacy trial of anti-IL-6 treatment in depression that the first paper calls for is funded and run, and whether it selects on hs-CRP as that paper's biomarker finding suggests [s1]. And whether the effort-discounting effect reproduces on clinical measures of anhedonia rather than on a behavioural task.

This article describes published trial results. It is not medical advice, and neither drug is indicated for depression.

Sources

• [s1] Interleukin 6 as a Treatment Target for Depression: A Proof-of-Concept Randomized Clinical Trial, JAMA Psychiatry, 2026;83(8):857–863.
• [s2] Semaglutide and Effort-Based Decision-Making in Major Depressive Disorder: A Randomized Clinical Trial, JAMA Psychiatry, 2026;83(7):751–754.