For clinical development leads designing immuno-oncology trials

Responsible disclosure for trial-design vulnerabilities in immuno-oncology.

Encounter reads phase 2 and phase 3 immuno-oncology trials before they read out, and notices structural problems the design teams cannot see from inside their own assumptions. Sponsors get a dated private letter first. If the private path closes, the read becomes a public, timestamped entry on the scorecard below.

How it works

Private disclosure first. Public scorecard if the private path closes.

Read the programme →

Same model that made security research a profession

Prospective predictions — published before readout

Structural calls timestamped on Zenodo before each trial reports. When a trial reads out, the card moves to retrospective with the score attached.

Published prediction Reads April 2026

VOLGA · NCT04960709

Durvalumab + tremelimumab + enfortumab vedotin — MIBC, cisplatin-ineligible

Encounter call WIN ~10–15pp

pCR triplet ~40.6%, pCR control ~25–30%, Δ ~10–15pp, ctDNA clearance ≥70%. Flagged as likely optimistic on magnitude; direction call unchanged.

pCR / EFS FAIL on pCR, weak on EFS → pending

Phase 3 n=677 ~$33M est.^*

Purchase →

Published prediction Reads ASCO 2026 (target)

ENERGIZE · NCT03661320

Gemcitabine/cisplatin ± nivolumab ± linrodostat (IDO1) — Cisplatin-eligible MIBC, perioperative

Encounter call FAIL (IDO1 arm)

Arm B (gem/cis + nivo) > Arm A (gem/cis); Arm C (gem/cis + nivo + linrodostat) ≈ Arm B. Critical test = (B > A) AND (C ≈ B).

pCR / EFS FAIL on primary → pending

Phase 3 n=1200 ~$48M est.^*

Purchase →

Published prediction Reads Expected 2026

IMbrave251 · NCT04770896

Atezolizumab + (lenvatinib or sorafenib) vs (lenvatinib or sorafenib) alone — 2L HCC after atezolizumab + bevacizumab

Encounter call FAIL

OS HR 0.90–1.10 (not significant); ORR Δ 0–5pp.

OS HR FAIL vs lenvatinib/sorafenib → pending

Phase 3 n=554 ~$22M est.^*

Purchase →

Published prediction Reads July 2026 (target)

KEYLYNK-012 · NCT04380636

Pembro + CRT → pembro ± olaparib vs CRT → durvalumab — Stage III NSCLC, unresectable

Encounter call FAIL (olaparib arm)

No PFS or OS benefit from adding olaparib to pembro consolidation in unselected population. Possible signal in HRD-positive subgroup if reported.

OS HR FAIL — no OS benefit over pembro+chemo → pending

Phase 3 n=870 ~$35M est.^*

Purchase →

Published prediction Reads Primary OS readout expected 2028–2029

DeLLphi-312

Tarlatamab + durvalumab + chemo vs durvalumab + chemo — 1L extensive-stage SCLC

Encounter call EQUIVALENT (~70% ORR)

Adding tarlatamab to a durvalumab + platinum/etoposide backbone in 1L ES-SCLC is a minor structural increment in an already-saturated first-line window. Framework predicts equivalent ORR, delta within ±5 points.

OS HR Within [0.80, 1.25] → pending

3 n=330

Purchase →

Published prediction Reads H2 2026

Harmony Melanoma · NCT05352672

Fianlimab + cemiplimab vs pembrolizumab vs cemiplimab monotherapy — 1L unresectable / metastatic melanoma

Encounter call WIN

ORR experimental ~58% vs control ~33.7%, Δ ~24pp; cemiplimab monotherapy arm clinically equivalent to pembrolizumab.

PFS HR WIN — HR < 0.75 → pending

Phase 3 n=1590 ~$64M est.^*

Purchase →

* Trial cost estimate: enrollment × $40K median per patient (oncology Phase 3, US sites, 2015–2017 cross-sectional analysis). Order-of-magnitude only.

Retrospective reports — confirmed predictions

Re-analyses of published immunotherapy trials. The framework's structural call versus what the trial measured. Methodology demonstrations, fully open after one form.

Retrospective Discontinued for futility

STAR-221 · NCT05568095

Domvanalimab + zimberelimab + chemo — 1L gastric/GEJ

Encounter call FAIL

Class-level structural call: TIGIT addition does not address the limiting step in 1L gastric. Predicted FAIL.

Observed Discontinued for futility

OS / PFS FAIL → pending

Phase 3

Read →

✓ Confirmed hit Published 2019–2022

EXPRESSO

Anti-PD-1 monotherapy — Metastatic melanoma

Encounter call Held 21% · Past 48%

Encounter predicts a 27pp gap between Activation-held and Past-Activation patients, where PD-L1 separates only ~14pp.

Observed Held 21.4% · Past 48.2%

+26.8 pp gap (p = 0.0002)

ORR gap (Held vs Past) Held 21% · Past 48% → Held 21.4% · Past 48.2%

Pooled retrospective n=240

Read →