# Section 3 (Part B) + Section 7: Trial Results & Failure/Success Analysis

*What each program showed in humans, and — the most transferable lesson for EoE
— why it advanced, halted, or pivoted. Endpoint outcomes are visualized in
Figure "Pivotal-trial outcomes matrix"; trial-level design fields are in
trials_master.csv.*

---

## 3B.1 Tzield / teplizumab — the one that reached the market

**The arc.** Teplizumab's clinical history is a 20-year lesson in *endpoint
choice*. Early new-onset trials (Herold 2002; AbATE, NCT00129259, n=83) consistently
showed **preserved C-peptide** — a mechanistic win — but the pivotal Protégé
Phase 3 (NCT00385697, n=554) **missed its composite primary endpoint** of
insulin dose + HbA1c in 2011. That miss nearly killed the program.

**The pivot that worked.** Rather than chase the crowded new-onset space,
investigators (TrialNet) tested teplizumab in a fundamentally different
population: **at-risk relatives who had not yet developed clinical disease**
(Stage 2 T1D — autoantibody-positive + dysglycemia). The TN-10 trial
(NCT01030861, n=76; NEJM 2019) showed a **median ~2-year delay in progression
to clinical diabetes**, extended to ~3 years on follow-up (Sims 2021). This
"delay of onset in a defined at-risk population" is exactly what earned FDA
approval in November 2022. The later PROTECT Phase 3 (NCT03875729, n=328, 2023)
then confirmed C-peptide preservation in newly-diagnosed patients.

**Success factors:** (i) a robust mechanistic biomarker (C-peptide) that
survived even endpoint failures; (ii) willingness to redefine the *population*
and the *clinical endpoint* (delay-of-onset, not cure); (iii) an at-risk
screening infrastructure (TrialNet) that made a prevention trial feasible.

---

## 3B.2 Nexvax2 — the most instructive failure in the field

**Result.** The RESET CeD Phase 2 (NCT03644069, n=~150) was **halted in 2019**.
At interim analysis Nexvax2 provided **no protection** against gluten-challenge
symptoms versus placebo; in fact treated patients experienced acute
gluten-like GI symptoms. ImmusanT wound down shortly after.

**Why it failed — the transferable lessons:**
1. **The endpoint was symptom protection under gluten challenge, not a
   mechanistic tolerance marker.** The program leapt to a demanding clinical
   efficacy endpoint before a validated pharmacodynamic biomarker of tolerance
   was locked.
2. **On-target reactogenicity confounded the readout.** Because the vaccine *is*
   the pathogenic peptide, dosing itself provoked gluten-like symptoms —
   contaminating a symptom-based endpoint.
3. **Possible antigenic incompleteness.** Three immunodominant peptides may not
   have covered enough of the polyclonal, epitope-spreading response in
   established disease.
4. **Treating established, epitope-spread disease is harder than preventing or
   treating early disease** — a theme teplizumab's at-risk pivot reinforces.

**EoE relevance:** an EoE peptide/pMHC program must (a) define a
pharmacodynamic tolerance biomarker *before* betting on symptom endpoints, and
(b) anticipate that delivering allergen epitopes could itself provoke symptoms —
a real safety and trial-design concern given anaphylaxis risk with food
allergens.

---

## 3B.3 TAK-101 / TIMP-GLIA — the mechanistic proof-of-concept that (partly) worked

**Result.** The Phase 2a (NCT03738475, n=34; Kelly 2021, *Gastroenterology*)
**met its mechanistic endpoint**: TAK-101 significantly reduced the
gluten-induced surge in circulating IFN-γ+ gliadin-specific T cells after a
gluten challenge, and attenuated other markers of immune activation — the first
clear human demonstration that a tolerogenic nanoparticle can blunt an
antigen-specific T-cell response. A subsequent larger study (NCT04530123) gave
more mixed clinical/histologic signals.

**Why it advanced where Nexvax2 failed:** TAK-101 led with a **pharmacodynamic
biomarker** (antigen-specific T-cell activation) rather than a symptom endpoint,
and delivered antigen in a *tolerogenic context* (apoptotic-mimic nanoparticle)
rather than as bare peptide — decoupling dosing from reactogenicity.

---

## 3B.4 KAN-101 — the current celiac frontrunner

**Result.** Phase 1 ACeD (NCT04248855, n=41) showed KAN-101 was **well
tolerated with dose-dependent modulation of immune markers** (including
IL-2/cytokine responses after gluten exposure). It holds FDA Fast Track.
Phase 2 SynCeD (NCT06001177) is delivering efficacy/mechanistic readouts; a
linked earlier Phase 1/2 (NCT05574010) was terminated for a sponsor/portfolio
decision, not a safety or efficacy signal. Liver-targeted delivery is the
mechanistic differentiator.

---

## 3B.5 GAD-alum / Diamyd — the precision-medicine rescue of a failed antigen therapy

**Result.** The original European Phase 3 (NCT00723411 / NCT00751842)
**missed its C-peptide primary endpoint** and was terminated around 2011 — a
high-profile failure paralleling Nexvax2. But Diamyd did not abandon the
antigen; it re-analyzed by genotype and found benefit concentrated in
**HLA-DR3-DQ2** patients, then changed the *delivery route* to
**intralymphatic** (direct lymph-node injection) to improve tolerogenic
targeting. DIAGNODE-2 (2021) **missed its overall endpoint but showed a
significant C-peptide benefit in the HLA-DR3-DQ2 subgroup**, and the Phase 3
DIAGNODE-3 now enrolls that genetically-defined responder population.

**EoE relevance — arguably the single richest lesson:** the *same antigen* went
from failure to a viable Phase 3 program by (i) **HLA-stratified patient
selection** and (ii) **optimizing the delivery route/niche**. An EoE pMHC
program should build HLA/biomarker stratification in from the start.

---

## 3B.6 Navacim / Parvus — validated mechanism awaiting clinical translation

**Result.** The strongest *preclinical* dataset in the field (Nature 2016):
pMHC-II nanoparticles reversed established disease across multiple autoimmune
mouse models via TR1 induction. This is the true pMHC platform, but it has not
yet produced a pivotal human efficacy readout — the clinical whitespace an EoE
pMHC vaccine would enter.

---

## 3B.7 Proinsulin peptide / MonoPepT1De — safety-first peptide tolerance

**Result.** MonoPepT1De Phase 1a (2017, n=27) demonstrated **safety with no
acceleration of beta-cell loss** and an IL-10-biased regulatory signature — an
important safety proof that a single immunodominant peptide could be given to
new-onset patients without harm. It remained early-phase.

---

## 3B.8 ATX-MS-1467 (MS) — signal in a small, uncontrolled design

**Result.** Phase 2 (Chataway 2018, n=37) showed a **reduction in new
gadolinium-enhancing MRI lesions** versus baseline, but the small, largely
open-label/baseline-controlled design limited interpretability, and Merck KGaA
did not advance it to a pivotal program.

---

## 7. Cross-program failure-mode & success-factor synthesis

### Failure modes (what killed or stalled programs)

| Failure mode | Programs affected | Mechanism |
|---|---|---|
| Symptom/clinical endpoint chosen before validated PD biomarker | Nexvax2, GAD Ph3 | Efficacy signal too noisy/demanding to detect |
| On-target reactogenicity (the drug is the antigen) | Nexvax2 | Dosing provokes disease-like symptoms, confounds endpoint |
| Treating established/epitope-spread disease | Nexvax2, GAD new-onset | Polyclonal response outruns a defined-antigen therapy |
| Wrong (unselected) population | GAD Ph3, Tzield Protégé | Benefit diluted across non-responders |
| Small/uncontrolled design | ATX-MS-1467 | Signal not confirmable |
| Portfolio/business decision | KAN-101 (one arm), some Tzield | Non-scientific attrition |

### Success factors (what let programs advance/approve)

| Success factor | Exemplar | Lesson for EoE |
|---|---|---|
| Robust mechanistic biomarker that survives endpoint misses | Tzield (C-peptide, incl. AbATE NCT00129259), TAK-101 (antigen-specific T cells) | Define & validate an EoE tolerance PD marker early |
| Redefining population toward prevention / early / at-risk | Tzield TN-10 | Consider early/at-risk EoE, not only established fibrostenotic disease |
| Delivering antigen in a tolerogenic context (not bare peptide) | TAK-101, KAN-101, Navacim | Choose delivery niche deliberately (NP, liver, pMHC scaffold) |
| HLA / genetic responder stratification | GAD-alum DIAGNODE pivot | Build HLA stratification into EoE trial design |
| Delay-of-progression as a valid regulatory endpoint | Tzield approval | A disease-modifying (not curative) endpoint can win approval |

### The single sentence
**Programs that led with a validated pharmacodynamic tolerance biomarker, a
tolerogenic delivery context, and a genetically/clinically enriched population
advanced; programs that bet a bare antigen against a symptom endpoint in
established disease failed.**