How tightly do the RTB101 trials constrain excess adverse-event risk?

Quantify the safety precision of two public RTB101 trials and produce explicit evidence constraints for advancing a clinically tested immunometabolism candidate. Determine which prespecified excess-risk margins are excluded by aggregate data and which remain unresolved. “No clear difference” must not be treated as demonstrated safety or longevity benefit.

The OpenLabs immunometabolism track seeks safe oral regulators from clinically tested compounds. This bounty supplies a reproducible quantitative due-diligence input for that choice, using RTB101 as a fixed worked candidate. It does not rank untested molecules or infer human lifespan effects.

Analyze three patient-level adverse-event categories in each of the phase 2b and phase 3 trials: any adverse event, serious adverse event, and adverse event leading to study-drug discontinuation. Use the RTB101 10 mg once-daily versus placebo columns in Tables 3–4 of the fixed paper. Preserve trial-specific denominators and follow-up. These six comparisons are overlapping outcomes and must not be added into a total burden or treated as independent observations.

For each, calculate treatment-minus-placebo risk difference. Use Wilson score intervals for each arm and Newcombe's unpaired score combination: if arm point estimates are pT,pC and Wilson limits LT,UT,LC,UC, the difference limits are d−sqrt((pT−LT)^2+(UC−pC)^2) and d+sqrt((UT−pT)^2+(pC−LC)^2). Report ordinary 95% two-sided limits using z0.975, an unadjusted one-sided 95% upper limit using z0.95 in the upper-limit expression, and a six-comparison Bonferroni one-sided upper limit using z(1−0.05/6). These are aggregate-data compatibility calculations, not a recreation of the original trial's analysis plan.

Check whether each one-sided upper limit is strictly below excess absolute-risk margins 0.01, 0.03 and 0.05. These margins are illustrative selection constraints, not clinically validated tolerances. For each comparison/margin, provide an approximate equal-allocation future-trial planning count per arm assuming equal true event rates p in both arms, p fixed to that trial's observed placebo rate: ceil(2*p*(1−p)*(z(1−alpha)+z0.80)^2/margin^2), for alpha=0.05 and 0.05/6. Treat this as a normal-approximation sensitivity calculation requiring prospective refinement, not a complete protocol or guaranteed power claim.

• trial-inputs.csv: counts/denominators for the six safety comparisons, plus the phase 3 primary clinically symptomatic respiratory-illness endpoint and its published reported effect/interval. Every input needs exact source text and table/paragraph locator. Include trial population, treatment duration, endpoint definitions and safety-analysis population metadata in README.md.
• analysis.py or analysis.R, risk-results.csv, and planning.csv: the specified interval, margin and recruitment calculations, with full-precision normal quantiles and integer ceilings. Code must handle counts faithfully and label absolute risks as proportions or percentage points consistently.
• risk-precision.svg or risk-precision.png: a readable plot of the six risk differences and interval/upper-bound results against the three hypothetical margins, clearly separating phases and correction assumptions.
• decision.md: state the advancement constraint supported by each outcome and where safety remains unresolved; explain dependence between categories, short follow-up, event definitions and trial-population transportability. Compare the safety conclusions with the phase 3 primary efficacy outcome and its gatekeeping rule. Identify what future endpoint/follow-up data would be needed before calling the candidate “safe” for an aging indication. No generic recommendation to take RTB101, mTOR inhibition or any other drug is requested.
• README.md: provenance, environment, a command regenerating the results and known approximation limitations. Include all required file bytes in the Submission; links are citations rather than artifact substitutes.

Fixed input: Mannick et al. (2021), *Targeting the biology of ageing with mTOR inhibitors to improve immune function in older adults: phase 2b and phase 3 randomised trials*, PMC8102040, DOI 10.1016/S2666-7568(21)00062-3, public PMC version accessible on 17 September 2026: https://pmc.ncbi.nlm.nih.gov/articles/PMC8102040/ . Required numerical sources are Tables 3–4 and the phase 3 primary-efficacy results paragraph; trial methods and discussion supply design context. No participant-level data or supplemental-file download is required.

Project context: https://openlabs-git-codex-openlabs-elgora-adapter-bio-xyz.vercel.app/projects/b7883dfd-8b74-484f-be5c-9827cedff06f . Independent research contribution without project-owner endorsement.

Only submissions satisfying all criteria are eligible. Prefer fewer material scientific/numerical errors, then the clearest defensible advancement constraint grounded in calculated precision, then stronger reproducibility and source traceability. Optimistic conclusions do not earn preference. Remaining ties go to earlier on-chain submission timestamp, then lower numeric submission ID. No winner is required if none qualifies.