Does blocking IL-11 extend life? Mouse survival, function and human evidence

Our assessment: the mouse survival advantage has direct support, and the functional findings merit further testing; attribution of the lifespan mechanism and human life extension remain unestablished. We checked 236 mouse records, recalculated survival and functional estimates, and retained unresolved contradictions in dates, scores and missing observations. Main study Source tables

01Two interventions address different questions

Widjaja and colleagues' 2024 study contains two lifespan experiments:

The paper summarizes treatment initiation as 75 weeks. Individual records specify 74.14 weeks for females and 76 and 74 weeks for the two male batches. We calculate time since each mouse's recorded start, rather than subtracting 75 from every age. Lifespan animals received monthly injections until an endpoint; the separate functional experiment used injections every three weeks with planned tissue collection after approximately 25 weeks.

The survival endpoint includes both being found dead and humane euthanasia. Criteria covered severe morbidity and tumours exceeding permitted size or causing ulceration, infection or impaired mobility. Reducing tumours could therefore also reduce tumour-related euthanasia triggers. These are not exclusively natural deaths, and mouse weeks do not have a fixed conversion into human age or remaining lifespan.

02Survival improves, but median conventions matter

The following medians use R survival's convention: when survival remains exactly at 50% for a plateau, average its two ends. This may differ from the first time survival reaches or falls below 50%.

Male X203 survival reaches 50% at 155.57 weeks and next declines at 163.57 weeks. Their midpoint is the paper's 159.57 weeks. The lifelines implementation reports the first crossing, 155.57; a follow-up update is not needed to explain the discrepancy. Similarly, the pooled therapeutic control median is 120.57 by first crossing and 120.93 by the plateau convention. With one consistent definition, the therapeutic median increases are about 28.6% pooled, 25.0% in females and 22.5% in males. A pooled ratio need not equal the average of the sex-specific ratios.

Only 6/34 male knockout mice had reached a death endpoint. Their median is not estimable, not infinite. Heavy censoring also limits the treated males' tail: the upper 95% confidence limit for their median cannot be estimated from the available follow-up.

A sex-stratified Cox model for the genetic cohort gives a common HR of 0.185 (95% CI 0.109–0.312). For the therapeutic cohort, a model stratified by sex and male batch, using time since treatment initiation, gives 0.179 (0.092–0.346). These are proportional-hazards summaries, not percentages of extra life. The sex-interaction model's main treatment coefficients, approximately 0.18 and 0.23, represented the male reference group, not pooled treatment effects.

Sex-interaction p values are about 0.70 and 0.16; failure to detect interaction does not establish equivalent effects. The two male batches have log-rank p values of approximately 0.004 and 0.010, but only 5/12 treated mice in batch two died: its KM median is not reached. The ordinary median of observed event/censoring times, 147.9 weeks, is not a valid lifespan median. These batches also remain within one facility, not independent laboratory replications.

03Extra weeks depend on the observation window

Restricted mean survival time, or RMST, is the area under a survival curve up to a specified horizon. It estimates average time alive within that window without converting an HR into a lifespan percentage.

These are exploratory horizons selected during revision, not preregistered primary endpoints. At the longest common horizon, 18.81 weeks is approximately 39.6% of control RMST. Dividing by the window length instead yields 25.5%, which is not the percentage increase in remaining life. Medians and RMST describe different estimands. Only two controls remain at risk near the longest horizon, making tail estimation particularly consequential.

The intervals assume independent animals and noninformative censoring. They do not cover cage, litter, facility or all model uncertainty. A common calendar cutoff does not itself prove independent censoring: genetic groups have different birth-date distributions, and some knockout animals were administratively censored relatively young. A sensitivity model adding a linear birth-date covariate still supports an advantage, but cannot eliminate every cohort effect.

One source conflict is explicit: female WT mouse “WTF 46” has dates implying 148.14 weeks, but a reported death age of 131.86 weeks. We retain the reported age in the primary analysis and use dates in a sensitivity analysis. The genetic stratified HR changes from 0.185 to 0.189, with the same direction. We cannot determine which source field is wrong and have not silently corrected the original.

04Fewer visible tumours do not identify the mechanism of longer life

Among deaths with recorded assessments, the proportions are 49/80 versus 3/25 and 22/35 versus 3/19. The paper's all-death denominators, 49/84 and 22/36, include missing assessments, which cannot automatically be counted as negative. Living censored animals also cannot be classified as tumour-free.

These differences support different gross pathology at death. They are not cancer incidence, cancer mortality or the proportion of life extension mediated by cancer prevention. Treatment changes who has died and therefore who enters the autopsy comparison.

Deleting deaths with tumours while retaining all surviving animals and then comparing “tumour-free survival” selects the sample using a post-treatment outcome. It cannot rule out a benefit mediated mainly by cancer prevention or establish a second independent lifespan pathway. Better pathology, cause-of-death assessment and a design identifying mediation are needed. Having a tumour is not equivalent to dying from it.

05Functional improvement is supported in a separate cohort

We linked baseline and endpoint frailty scores by mouse identifier and checked both sexes. Functional source tables

These are exploratory Welch intervals for observed endpoints, not confirmatory intervals corrected for every endpoint in the study. Between-group differences in frailty change are −3.13 and −3.22 points; baseline-adjusted analyses agree in direction. Body mass is itself affected by treatment. The approximately 30% male difference in g/g therefore cannot be rewritten as a 30% increase in absolute force. The grip source table lacks mouse identifiers, so multiplication by body weights matched solely by column order would not establish a reliable reconstruction of raw force.

Source materials contain additional inconsistencies. Some male control baseline scores differ between the figure data and the item-level frailty table; 25 “change” cells in the latter actually equal endpoint scores. We use the verifiable sum of 27 items and calculate changes from linked identifiers. Female endpoint age is 96.14 weeks in the detailed table versus 100 weeks in the figure/text; this remains unresolved. Two of 13 female controls lack endpoint scores, with no clear reason. Even assigning both missing scores the scale minimum of zero gives a completed control mean of 4.85, above 2.62 in treated animals. That bounded check does not resolve missingness or measurement timing.

The results support functional improvement in these cohorts. They do not continuously measure an entire healthy lifespan or prove that functional improvement mediated survival in a different set of animals.

06Molecular mechanisms have experimental support; lifespan mediation does not

The proposed direction is IL-11 activation of ERK/p90RSK, inhibitory phosphorylation of LKB1, reduced AMPK activity and consequent release of mTORC1 inhibition. Describing every component as “activated” reverses part of that mechanism. The study includes pathway perturbations in human fibroblasts, hepatocytes and other experiments, so molecular support is not exclusively correlational. Mechanistic experiments

These findings connect with Schafer and colleagues' earlier fibrosis work. They do not quantify the share of lifespan change explained by mTOR, fibrosis, metabolism or tumours. Establishing a molecular causal relationship and showing that it mediates organismal life extension are separate tasks. Fibrosis study abstract

A 2026 ovarian study adds evidence concerning matrix stiffness and reproductive function. An Alport model shows kidney and survival improvements in a disease context. Both extend the biological evidence, but neither independently replicates whole-organism lifespan extension during natural ageing. Ovarian abstract Alport abstract

07Human disease outcomes are available; human life extension is not established

We checked six ClinicalTrials.gov records for LASN01 and 9MW3811 and a 2026 primary report. Development includes pulmonary fibrosis, thyroid eye disease, scars and advanced solid tumours. It is not restricted to fibrotic conditions, and a registry status of “completed” is not a finding of efficacy. Registries 9MW3811 records

The LASN01 thyroid eye disease randomized comparison contains 26 people: 17 across two active doses and nine receiving placebo. The registry total of 41 includes a separate open-label group of 15. At week 37, the registered primary efficacy endpoint, proptosis response, was 6/17 versus 3/9, without a clear advantage. The secondary clinical activity score, or CAS, response was 15/17 versus 4/9; the paper reports p=0.028. The primary result must accompany the positive secondary signal rather than relying only on the abstract's CAS finding. The paper's narrative of eight participants per arm differs from its actual tables; we use the tables and posted registry results. 2026 report

Treatment-emergent adverse events were reported in 4/9, 6/8 and 6/9 participants in the randomized arms. No serious events were reported there; the open-label arm reported one in 15. An event's occurrence does not establish drug causation. These small samples, approximately one year of follow-up and incomplete study completion do not establish long-term safety of pathway suppression in healthy older adults. Equally, prior haematopoietic uses of exogenous IL-11 do not by themselves prove a specific toxicity of blocking it; long-term safety needs direct observation.

We found no results establishing human life extension or broad healthspan benefits in the records and primary reports examined. This is a bounded search statement, not a claim that no ageing trial exists anywhere. Disease-specific trials cannot substitute for randomized evidence in healthy older people or on natural-ageing outcomes.

08What the evidence now supports

The central mouse finding survives reanalysis: genetic and antibody interventions shifted survival curves within one laboratory, with improvement in separate functional cohorts. The contribution of cancer prevention remains unresolved. Nonsignificant interactions do not establish equal effects, and molecular, functional and survival findings from different cohorts do not complete a demonstrated mediation chain.

The most informative next steps include replication across facilities and strains, adequate follow-up, cage and birth-cohort information, blinded pathology and cause-of-death assessment, and designs connecting function with lifespan. Human studies need explicit clinical endpoints and prolonged safety observation. This article describes evidence and provides no antibody-use recommendations.

Reanalysis scope: individual lifespan records, both-sex functional tables, KM/log-rank, correctly labelled Cox models, RMST and intervals, source-field sensitivities, and independent Python/R comparisons. We did not repeat the whole transcriptome, sequencing or every pathology analysis, and did not obtain human participant data. Intervals do not cover source errors, animal dependence or cross-species extrapolation.

Download the code, derived mouse data and reproducible results.

09Sources

• <a id="source-s1"></a>s1: Widjaja AA, et al. Inhibition of IL-11 signalling extends mammalian healthspan and lifespan. Nature 2024. Full text. Reread design, molecular perturbations, Figures 3/5, Extended Data Figure 8, statistics and humane endpoints.
• <a id="source-s2"></a>s2: Public supplementary materials. Lifespan Tables 6/7, frailty Tables 3/4 and Figure 3/Extended Data Figure 8 source data. All 236 lifespan records parsed; contradictions preserved. Large transcriptome table not reanalysed.
• <a id="source-s3"></a>s3: Miller RA. Blocking an inflammatory protein slows the pace of ageing. Nature 2024. Commentary. Contextual appraisal, not independent experimental replication.
• <a id="source-s4"></a>s4: ClinicalTrials.gov, NCT05331300 and NCT06226545. Current API records read, including participant flow, primary/secondary outcomes and safety results for the latter.
• <a id="source-s5"></a>s5: ClinicalTrials.gov, 9MW3811: scars, solid tumours, healthy-volunteer study one and study two. Retrieved 20 September 2026. Some records have unknown status and do not establish current recruitment.
• <a id="source-s6"></a>s6: Schafer S, et al. IL-11 is a crucial determinant of cardiovascular fibrosis. Nature 2017. Correct record and abstract. Abstract read; identified by the correct first author and PMID.
• <a id="source-s7"></a>s7: Wu M, et al. Modulating IL-11-dependent matrix stiffness to delay ovarian aging. 2026. Record and abstract. Abstract read; organ/reproductive outcomes, not whole-organism natural-ageing lifespan replication.
• <a id="source-s8"></a>s8: Widjaja AA, et al. A Neutralizing IL-11 Antibody Improves Renal Function and Increases Lifespan in a Mouse Model of Alport Syndrome. JASN 2022. Abstract. Disease-model boundaries; no independent numerical reanalysis.
• <a id="source-s9"></a>s9: King DJ, et al. IL-11 receptor is a novel target for drug development with pharmacological activity in fibro-inflammatory disease. Scientific Reports 2026. Primary report. Clinical design, results, safety and Tables 3/4 read and checked against registration. Authors include developer employees; the publication and registry are not independent studies.