Thymosin Alpha-1 research: what the published studies show

The Phase 3 Trial

A 2025 multicenter, double-blind, randomized, placebo-controlled phase 3 trial enrolled patients with sepsis across 22 centers, randomizing 1,106 and analyzing 1,089 in a modified intention-to-treat set.^[1]

All-cause mortality at 28 days was 23.4% on thymosin alpha 1 and 24.1% on placebo.^[1] The hazard ratio was 0.99, with a 95% confidence interval of 0.77 to 1.27 and a P value of 0.93.^[1] No secondary outcome and no safety outcome differed between the groups.^[1]

Prespecified subgroup analyses showed interaction by age and by diabetes status, and those are subgroup signals in a null trial that the authors present as such.^[1]

The authors' stated conclusion is that there is no clear evidence to suggest the compound decreases 28-day all-cause mortality, and that wording is reproduced rather than softened.^[1] This trial is registered as NCT02867267.^[1]

The Earlier Trial

A 2013 multicenter single-blind randomized controlled trial enrolled 361 patients with severe sepsis across six centers in China.^[2] All-cause mortality at 28 days was 26.0% in the treated arm and 35.0% in the control arm.^[2]

The significance of that difference depends on which test is read: the nonstratified P value was 0.062 and the log rank P value was 0.049.^[2] The relative risk was 0.74 with a 95% confidence interval of 0.54 to 1.02, and that interval crosses 1.^[2]Reporting only the log rank value would misstate the result.^[2]

Monocyte HLA-DR expression improved at day 3 and day 7.^[2] The authors conclude that the compound may be effective in a targeted population, and that hedge is theirs.^[2]

Pooled Analysis

A 2025 systematic review and meta-analysis pooled 11 randomized controlled trials, comparing 967 patients against 960.^[3] The overall odds ratio was 0.73, with a 95% confidence interval of 0.59 to 0.90 and a P value of 0.003.^[3]

Two prespecified subgroups reverse that picture.^[3] Restricted to high-quality trials the odds ratio was 0.82 with a P value of 0.09, and restricted to multicenter trials it was 0.86 with a P value of 0.20.^[3] Neither reaches significance.^[3]

The meta-analysis states that its heterogeneity analysis covered two large randomized trials representing 75% of its total patients.^[3] That figure matches reference 1's 1,089 analyzed plus reference 2's 361, which is 75.2% of the pooled 1,927, so reference 1 sits inside this pool rather than outside it.^[1] The meta-analysis's own trial sequential analysis states that the current sample size is inadequate.^[3]

Dendritic Cells

A 2004 study reported that thymosin alpha 1 induces maturation and interleukin-12 production by dendritic cells pulsed with Aspergillus.^[4] The effect ran through p38 MAPK and NF-kappaB, and through MyD88-dependent toll-like receptor signaling.^[4]

This is mechanism work in cells and in a fungal infection model.^[4] It is the only source on this page that offers a route by which the compound might act, and it does not connect to the mortality endpoints the trials measured.^[4]

Commercial and Regulatory History

A 2007 paper describes the compound's pharmaceutical development.^[5] It states that SciClone Pharmaceuticals licensed the commercial and patent rights for geographical regions of the world excluding the United States and Europe.^[5] It states that the compound had obtained approvals in more than 30 countries.^[5]

Both statements are nineteen years old and the author was affiliated with the company that commercialised the compound.^[5] Neither establishes current regulatory status anywhere.^[5] The current position is read from the FDA, the UK MHRA and the European Medicines Agency directly, and is stated on the About This Compound tab.^[5]

State of the Evidence

Three sources address the same clinical question and they do not agree.^[1] The largest, a double-blind phase 3 trial in 1,089 analyzed patients, found no difference on any outcome it measured.^[1] An earlier single-blind trial of 361 patients found a mortality difference whose significance changes with the test applied.^[2] A meta-analysis of eleven trials found a significant pooled effect that vanished in its own high-quality and multicenter subgroups.^[3]

This page does not adjudicate between them.^[1] The disagreement has a shape: the effect appears in smaller and less strictly blinded studies and is absent in the largest and most strictly blinded one.^[3]

Wu J is first author on both randomized trials, so the investigators who reported the earlier marginal result are the same ones who later reported null.^[1]

Around 1,850 records are indexed for this compound and five appear here.^[4] The mechanism work in reference 4 sits at a different level entirely and connects to none of the clinical endpoints.^[4] Nothing on this page describes what the compound does in any healthy population, because no source retrieved for this page studied one.^[5]