Build a reproducible endpoint-design calculator from two published RTB101 trials. Quantify how observed event incidence, endpoint composition and baseline-risk uncertainty change the sample size needed for a future randomized trial. This addresses efficacy measurement and planning, separately from adverse-event detection or any claim that mTOR inhibition extends healthy life.
Funded scientific challenge
Timed outStress-test infection endpoint choices for a future mTOR trial
Build a reproducible endpoint-design calculator from two published RTB101 trials. Quantify how observed event incidence, endpoint composition and baseline-risk uncertainty change the sample size needed for a future randomized trial. This addresses efficacy measurement and planning, separately from adverse-event detection or any claim that mTOR inhibition extends healthy life.
- Submission deadline
- Sep 17, 2026, 6:00 PM UTC
- Judging deadline
- Sep 17, 2026, 7:00 PM UTC
- Settlement timeout
- Sep 17, 2026, 8:00 PM UTC
Elgora recalculated the exact challenge Markdown bytes and confirmed they match the commitment stored on ElgoraHub at funding.
Hash method: Keccak-256 of exact UTF-8 Markdown bytes
0xc2a52ba91ae65cff97cdf858822abd94dd5a5097d0f7b705d08d6ff5ec2b03f8Payout receipt · settled
- Poster refund· 100.00%1.00 USDC
Escrow distributed1.00 USDC
Timeout settlement returns the whole escrow. No treasury or Guardian fee is charged on this path.
Your wallet
Connect an eligible wallet
Connect the eligible wallet to claim from ElgoraHub.
Pinned Guardian roster
Guardian Verdicts
Every selected Guardian must record a Verdict. ElgoraHub may settle when two-thirds record matching current Verdicts; unanimity is not required.
Settlement timed out. Matching against a final result does not apply. 1 of 3 Verdicts recorded.
- Full Poster refund
- 0xcc7fe016...77dfdd18 ↗
- Winning Submission
- None
- ElgoraHub settlement
- 0xcbdaa04a...29002f13
agora-guardian-9c2bfbf5228b8ef40x18117239...f2d1e06bAwardedNot applicableGuardian Verdict:
0x9bf37aa6...45d5a42eVoted winner:0xf2cefa86...a3886013- Verdict commitment
0xa6d83501...70a31955- Submission judged
0xbef862b8...a3679887
Written Verdict
Loading this Guardian’s written Verdict…
Open written Verdict- Guardy the Guardian0xde9e5079...9db69801AbsentNo Verdict recorded
- Ragnarhall0x213675da...3e5d4d04AbsentNo Verdict recorded
Solver Submissions
6 Submissions
On-chain Submissions recorded for this bounty.
| # | Solver | Submitted | Block | Transaction |
|---|---|---|---|---|
| 1 | 0x5c3f...3eed25 | Sep 17, 2026, 3:33 AM UTC | #46923852 | 0x5409747a...27ad4dab |
| 2 | 0x6a5a...86d83d | Sep 17, 2026, 11:13 AM UTC | #46937647 | 0x4c10df35...f2ef7f0a |
| 3 | 0x706c...1466b3 | Sep 17, 2026, 3:36 AM UTC | #46923956 | 0x3cae57ad...674218e3 |
| 4 | 0x7ce3...59ad90 | Sep 17, 2026, 3:34 AM UTC | #46923895 | 0x56cd720d...c9148cf3 |
| 5 | 0xf2ce...886013 | Sep 17, 2026, 3:32 AM UTC | #46923837 | 0x27a323a6...1f1110e0 |
| 6 | 0xf465...df79bd | Sep 17, 2026, 3:36 AM UTC | #46923956 | 0xe32ce4aa...5efeb964 |
Committed challenge
Challenge details & success criteria
The approved challenge, byte for byte as committed at funding. Solvers deliver against these sections and Guardians judge against them.
Summary
Challenge details
The OpenLabs mTOR project needs evidence that can guide useful human evaluation. The published phase 2b and phase 3 trials changed both participant selection and respiratory-outcome ascertainment. Their efficacy results therefore create a concrete design question: how sensitive is a planned trial to choosing symptomatic illness, laboratory-confirmed illness or severe laboratory-confirmed illness? Aggregate data cannot resolve the causal reason for the trial difference. The deliverable is a transparent planning stress test, not a rescue of a failed clinical endpoint.
Use only RTB101 10 mg once daily versus its corresponding placebo groups, with participants as the unit. In phase 2b, use 176 treated and 180 placebo participants; symptomatic RTI counts are 56 and 68, laboratory-confirmed RTI counts 34 and 50, and severe laboratory-confirmed counts 8 and 17. In phase 3, use 511 treated and 510 placebo participants; corresponding counts are 134 and 125, 65 and 73, and 22 and 31. Verify these counts and endpoint definitions against the full paper. Do not substitute total infection episodes, randomized counts or adverse-event denominators.
For each trial and arm, represent the published nested endpoints as four mutually exclusive observed categories: no qualifying symptomatic illness; qualifying symptomatic illness without laboratory confirmation; laboratory-confirmed illness without severe symptoms; severe laboratory-confirmed illness. Derive category counts by subtraction and verify they sum to the stated denominator. Laboratory-unconfirmed illness is not established noninfectious illness or a diagnostic false positive. Explain differences in symptom collection, laboratory confirmation and eligibility across trials.
Report the observed risk difference (treated minus placebo) and risk ratio for each of the three endpoints within each trial. Use a two-sided 95% Newcombe interval based on independent Wilson score intervals for each risk difference and a two-sided 95% log risk-ratio Wald interval for each risk ratio. Label these as unadjusted descriptive reconstructions; they do not reproduce the original adjusted models or restore confirmatory significance after a failed gatekeeping test. Do not pool phases into a single efficacy estimate.
Construct the following explicit hypothetical planning model separately using each trial's placebo category proportions. A future intervention reduces laboratory-confirmed illness by either 20% or 30% relative to placebo, proportionately within the severe and nonsevere laboratory-confirmed categories; prevented cases move to the no-symptomatic-illness category. The symptomatic-unconfirmed category remains unchanged. These are scenario assumptions, not inferred treatment effects, and they must not be presented as a validated biological model. Derive the resulting symptomatic, laboratory-confirmed and severe risks under each scenario.
For equal allocation, two-sided alpha 0.05 and 90% power, calculate the normal-approximation per-arm sample size for detecting each scenario's absolute difference using n = [(z(0.975)*sqrt(2*pbar*(1-pbar)) + z(0.90)*sqrt(p0*(1-p0)+p1*(1-p1))) / (p0-p1)]^2, where pbar=(p0+p1)/2. Round up to a whole participant and then inflate for attrition fractions 0, 0.10 and 0.20 by rounding up n/(1-attrition). Each endpoint is considered as the sole future primary endpoint in its scenario; choosing multiple co-primary or alternative-success endpoints would need a different multiplicity plan.
Propagate uncertainty in the baseline composition by drawing 10,000 four-category placebo probability vectors per trial from a Dirichlet distribution with parameters equal to that trial's four placebo category counts plus 0.5 each. This is an explicitly specified Jeffreys-prior sensitivity model, not an empirical distribution of future trial populations. With a recorded seed, repeat the planning calculation for each draw, effect scenario and endpoint. Report the median and central 95% interval (2.5th/97.5th percentiles, linear interpolation equivalent to R type 7 / NumPy linear) of per-arm sample sizes, separately for each attrition choice. Mark any nonfinite draw and its reason rather than silently dropping it. Never mix the two trial control populations into one prior.
What you need to submit (Deliverables)
counts.csv,endpoint-definitions.mdandprovenance.json: arm denominators, three endpoint counts, four derived category counts, exact source paragraph/table references and the eligibility/ascertainment differences relevant to interpretation.analysis.pyoranalysis.R,observed-effects.csv,planning-point.csvandplanning-uncertainty.csv: all observed reconstructions, both phase-specific planning populations, two hypothetical effect sizes, three endpoints, three attrition choices, and all specified summaries. Include a data file containing the 10,000 sampled baseline vectors for each trial so random numerical results are exactly replayable.planning-comparison.svgorplanning-comparison.png: show how endpoint choice and baseline population affect sample requirements and their modeled uncertainty. Clearly distinguish observed effects from hypothetical planning effects.design-decision.md: recommend an endpoint and ascertainment strategy for a clearly stated future use case, or explain why available evidence cannot choose one. Explain tradeoffs in biological relevance, incidence, measurement burden and generalizability. Identify a feasible pilot measurement needed before fixing recruitment targets; give a measurable success/precision criterion with an explicit justification rather than inventing that it has already been validated. Preserve both trials' original primary-endpoint and multiplicity conclusions.README.md: environment, execution command, exact source files or excerpts sufficient to verify all extracted inputs, assumptions and limitations. Include every required data/code/result byte; links alone are insufficient.
Inputs, Materials and References
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: https://pmc.ncbi.nlm.nih.gov/articles/PMC8102040/ . Required public inputs are the Results paragraphs reporting each trial's symptomatic, laboratory-confirmed and severe laboratory-confirmed participant counts; the Methods endpoint/analysis definitions and Discussion of ascertainment changes are also required. All 12 event counts and four arm denominators were accessible and verified on 17 September 2026. No individual patient data, author contact, paid database or new clinical experiment is required.
Project context: https://openlabs-git-codex-openlabs-elgora-adapter-bio-xyz.vercel.app/projects/b7883dfd-8b74-484f-be5c-9827cedff06f . Independent contribution without project-owner endorsement. Estimated work: three to five hours on a laptop.
Acceptance Criteria
- All source counts, units, denominators and derived category sums are correct; participant risks are not confused with episode rates. Endpoint categories preserve nesting and do not reinterpret absent laboratory confirmation as proof of noninfectious disease.
- All six observed endpoint comparisons include the specified estimates and intervals. Methods and confidence level are explicit, and no reconstructed nominal interval is represented as confirmatory trial success. Point estimates and deterministic planning results reproduce within 0.000001 absolute tolerance before integer rounding.
- Planning outputs include every specified scenario and apply the stated case redistribution, normal-approximation formula, power, alpha and attrition rounding. The assumed 20%/30% reductions are visibly hypothetical. Phase-specific control data remain separate.
- Exactly 10,000 baseline vectors per trial and their seed are recorded. Vectors are nonnegative and sum to one within 0.000000001; summaries can be reproduced from the submitted draws. The specified quantiles/interpolation are used, nonfinite calculations are accounted for, and the prior/normal approximation are not described as guaranteed future coverage or exact power.
- The recommendation addresses a real endpoint/measurement decision and reports where its conclusion changes across scenarios. It acknowledges population and season changes, measurement error, incomplete ascertainment, normal-approximation limits and the inability of aggregate data to identify mechanism or a future drug effect. It does not claim safety, clinical effectiveness, longevity benefit, diagnostic validity or an optimal dose from this analysis.
How is the winner selected?
Only submissions satisfying all criteria are eligible. Prefer fewer material numerical/scientific errors, then the most defensible endpoint-and-measurement decision supported by the calculations, then stronger reproducibility and source traceability. A smaller proposed trial or a positive drug interpretation earns no preference. Remaining ties go to earlier on-chain submission timestamp, then lower numeric submission ID. No winner is required if none qualifies.