Phase II Single-Tumor Example:

Go / No-Go Decision for Phase III

Case Study: IMvigor210 → Decision to Proceed to Phase III

Indication: Metastatic urothelial carcinoma (mUC)
Line of therapy: Post-platinum (2L+)
Investigational drug: Atezolizumab (anti–PD-L1)
Study phase: Phase II
Design: Single-arm, open-label
Assessment: RECIST 1.1 with BICR

This is a classic real-world example of a Phase II study that directly triggered the question:

“Is this strong enough to justify a Phase III trial?”

---

Key Phase II Results (at decision time)

IMvigor210 – Cohort 2

• ORR: 15% (95% CI: 11–19%)
• CR rate: ~5%
• Median DoR: Not reached
• 12-month DoR rate: ~84%
• Grade ≥3 TRAE: ~16%
• Historical ORR (2L mUC): ~10%

---

Go / No-Go Evaluation Using a Structured Checklist

A. Efficacy Signal

Criterion	Assessment	Rationale
ORR magnitude	Borderline	15% vs ~10% historical control; modest absolute improvement
CI lower bound	Risk / Borderline	Lower bound close to historical response
DoR durability	Go (key strength)	Median DoR not reached; responses were long-lasting
CR presence	Go	Presence of CR supports biological activity
Consistency across endpoints	Borderline	ORR modest; durability strong

Module conclusion: Borderline, supported primarily by DoR

---

B. Consistency and Robustness

Criterion	Assessment	Rationale
Multicenter consistency	Go	Responses observed across sites
Subgroup consistency	Borderline	Higher PD-L1 expression associated with better response
Independent review (BICR)	Go	Reduced assessment bias
Sensitivity analyses	Borderline	ORR sensitive to assumptions

Module conclusion: Borderline → Go

---

C. Biological Plausibility

Criterion	Assessment	Rationale
Mechanism of action	Go	PD-L1 pathway well-established in mUC
Biomarker support	Borderline	PD-L1 enrichment helpful but not definitive
Exposure–response	Not critical	Platform antibody with prior knowledge

Module conclusion: Go

---

D. Safety and Risk–Benefit

Criterion	Assessment	Rationale
Grade ≥3 TRAE	Go	Manageable and expected for class
Serious safety signals	Go	No unexpected safety concerns
Discontinuation due to AE	Go	Acceptable
Overall risk–benefit	Borderline → Go	Benefit driven mainly by durable responses

Module conclusion: Go

---

E. Phase III Feasibility (Critical Gate)

Criterion	Assessment	Rationale
Control arm	Go	Chemotherapy standard available
Primary endpoint	Go	Overall survival feasible
Sample size	Go	Achievable
Recruitment	Go	High unmet need population

Module conclusion: Go

---

F. Regulatory and Strategic Fit

Criterion	Assessment	Rationale
Accelerated approval potential	Go	ORR + durable DoR acceptable to regulators
Confirmatory pathway	Go	Phase III OS study defined
Portfolio priority	Go	Core oncology asset

Module conclusion: Go

---

Final Go / No-Go Decision (at the time)

Decision: GO → Proceed to Phase III (IMvigor211)

Rationale:

• ORR was modest but durability was compelling
• Safety acceptable
• Phase III design feasible
• Immunotherapy long-tail benefit hypothesis

---

Post-hoc Outcome (Critical Learning)

• Phase III IMvigor211 failed to meet the primary OS endpoint
• Retrospective interpretation:
• Phase II ORR improvement was too small
• Phase II signal relied heavily on DoR alone
• Biomarker stratification assumptions were optimistic

---

Key Takeaway for Decision-Making

A Phase II study can justify a Phase III trial even when the decision is “reasonable but risky.”
Strong durability cannot always compensate for a marginal response rate.

---

One-Sentence Summary

In single-tumor Phase II trials, the decision to proceed to Phase III hinges not on whether there is a highlight, but on whether the weakest link in the evidence chain can support a large, confirmatory trial.

Phase II: More focused on ORR / DoR (tumor reduction and duration)

Phase III: More focused on OS / PFS (survival outcomes)

Go/No-Go Decision:

✔ The framework must be defined in advance.

✔ The metrics must be defined in advance.

✔ Thresholds can be defined as ranges/levels.

❌ Rules cannot be "invented" after the data is available.

(Summary of Experience)

If the Phase I sample size is small but the signal is strong and the mechanism is clear, prioritize a single-dose main cohort plus a small exploratory cohort. This represents a compromise between efficiency and risk.

If Phase I shows significant non-linear PK or PD that has not reached a plateau, do not rush to determine a single dose; it is safer to include dose-finding in Phase II.

For immunotherapy or drugs with long-term administration, chronic toxicity is particularly important. The recommended Phase 2 dose (RP2D) is often chosen at a dose lower than the maximum tolerated dose (MTD); continue monitoring and retain room for adjustment in Phase II.

All "if...then..." trigger conditions must be actionable, auditable, and quantified as much as possible (e.g., increase in ORR, proportion of Grade ≥3 TRAEs, PK Ctrough differences, etc.).

Phase II Dose Strategy vs Risk Reduction Matrix

Phase II Dose Strategy	Description	Dose Uncertainty Reduced	Safety Risk Reduced	Efficacy Risk Reduced	Cost / Operational Complexity	Typical Use Case
Single RP2D Only	One RP2D used for all patients	●	●	●	★ (Lowest)	Phase I shows clear PK/PD plateau, strong efficacy signal, wide safety margin
RP2D + Exploratory Cohort	Main cohort at RP2D plus small cohort at adjacent dose	●●	●●	●●	★★	Mild residual uncertainty around optimal dose; low-cost confirmation needed
Randomized Dose Selection	Two doses randomized (e.g., 300 vs 350 mg)	●●●	●●●	●●●	★★★	Dose–response uncertainty could materially impact Phase III success
Adaptive / Bayesian Dose Optimization	Dynamic allocation to better-performing dose(s)	●●●●	●●●●	●●●●	★★★★	Complex MoA, narrow therapeutic window, very high Phase III stakes
Schedule / Dose Strategy Evaluation	Same dose, different schedules or dose-modification rules	●●	●●●	●	★★	Primary concern is chronic toxicity or tolerability rather than peak exposure
Seamless Phase I/II Dose Expansion	Continued multi-dose learning during expansion	●●●	●●	●●	★★–★★★	Strong early signal but remaining uncertainty in optimal dose

Legend

• ● = Low
• ●● = Moderate
• ●●● = High
• ●●●● = Very high
• ★ = Relative cost / complexity (more ★ = higher)

---

How to Use This Table for Decision-Making

1️⃣ Assess Residual Dose Uncertainty

Ask:

• Is PK/PD fully saturated at the RP2D?
• Is there evidence of cumulative or dose-sensitive toxicity?
• Could a suboptimal dose materially affect Phase III outcomes?

---

2️⃣ Match Strategy to Risk Profile

• Low uncertainty → Single RP2D
• Moderate uncertainty → RP2D + exploratory cohort
• High uncertainty + high Phase III cost → Randomized or adaptive dose selection

---

3️⃣ Avoid Over-Engineering

Not every program needs maximal risk reduction—
only enough to make Phase III a rational investment.

---

Typical Industry Pattern (Oncology, Rule of Thumb)

• ~30%: Single RP2D
• ~40–50%: RP2D + exploratory cohort
• ~15–20%: Randomized or adaptive dose selection
• <10%: Highly complex dose optimization

---

One-Sentence Takeaway

Phase II dose strategy should be chosen based on how much dose uncertainty remains and how costly it would be to be wrong in Phase III.

Oncology Phase I Go / No-Go Example

Study Type

First-in-Human (FIH) Phase I, dose escalation with expansion

---

Background

• Drug class: Novel targeted small-molecule (first-in-class)
• Indication: Advanced solid tumors (pan-tumor)
• Design:
• Dose escalation: BOIN design
• DLT window: 28 days
• Dose expansion at RP2D
• Primary objectives:
• Safety and tolerability
• MTD / RP2D determination
• Secondary objectives:
• PK
• Preliminary anti-tumor activity (RECIST 1.1)

---

Key Phase I Data at Decision Point

1. Safety and Tolerability

• 6 dose levels evaluated
• MTD not reached
• RP2D selected based on PK + safety
• Grade ≥3 TRAEs: 18%
• DLTs: 2 / 24 patients (8%)
• DLTs were predictable and manageable (hematologic)

---

2. Pharmacokinetics

• Dose-proportional exposure
• RP2D exposure:
• Exceeds preclinical IC90
• Achieves steady-state
• No unexpected accumulation

---

3. Preliminary Anti-Tumor Activity

(Not a primary Phase I objective, but critical for Go / No-Go)

• Evaluable patients: 28
• Partial responses (PR): 4 patients (14%)
• Stable disease ≥6 months: 6 patients (21%)
• Disease control rate (DCR): 36%
• 2 PRs ongoing ≥9 months
• Responses enriched in biomarker-positive tumors

---

Phase I Go / No-Go Evaluation

---

A. Safety (Hard Gate)

Criterion	Outcome	Decision
DLT rate	8%	Go
Predictability of toxicity	Yes	Go
Irreversible toxicity	None	Go
Treatment-related deaths	None	Go

Conclusion: Go

---

B. PK and Dose Justification (Core Phase I Requirement)

Criterion	Outcome	Decision
Dose–exposure relationship	Linear	Go
RP2D scientifically justified	Yes	Go
Target coverage at RP2D	IC90 covered	Go
Dosing schedule feasible	Yes	Go

Conclusion: Go

---

C. Preliminary Anti-Tumor Activity (Value Driver)

Criterion	Outcome	Decision
Objective responses observed	Yes	Go
Durability of responses	≥9 months	Strong Go
Biological plausibility	Yes	Go
Activity driven only by SD	No	Go

Note: Objective responses are not required in Phase I,
but durable PRs represent a strong Go signal.

Conclusion: Strong Go

---

D. Biological and Mechanistic Support

Criterion	Outcome	Decision
MoA consistent with responses	Yes	Go
Biomarker–response relationship	Clear enrichment	Go
Exposure–response trend	Observed	Go

Conclusion: Go

---

E. Readiness for Phase II

Criterion	Outcome	Decision
RP2D defined	Yes	Go
Target population identifiable	Yes	Go
Phase II design feasible	Yes	Go
Competitive landscape	Acceptable	Go

Conclusion: Go

---

Final Go / No-Go Decision

Decision: GO → Proceed to Phase II (biomarker-enriched design)

---

What Would Have Led to a No-Go?

Common oncology Phase I No-Go scenarios:

• RP2D exposure fails to reach biologically active levels
• DLTs are unpredictable or cumulative
• No objective responses or durable disease control
• Lack of mechanistic coherence
• Safety acceptable but no evidence of clinical activity

---

Key Phase I Decision Principle (Oncology)

In oncology Phase I, safety alone is not sufficient.
The question is whether acceptable safety reveals any biologically or clinically meaningful signal worth validating.

Example: How RP2D Is Determined in an Oncology Phase I Study

Study Background

• Drug class: First-in-class targeted small-molecule
• Indication: Advanced solid tumors
• Design: BOIN dose-escalation with expansion
• Dosing schedule: QD continuous dosing
• DLT window: 28 days

---

Key Principle (Critical)

RP2D is not necessarily the MTD.
RP2D is the dose that provides the optimal balance of safety, PK, PD, and preliminary efficacy.

---

Dose-Escalation Summary

Dose Level	Dose (mg QD)	DLT (n/N)	Grade ≥3 TRAE	PK Exposure (AUC)	PD Target Inhibition	PR
DL1	50	0/3	0	Low	30%	0
DL2	100	0/3	0	↑	45%	0
DL3	200	1/6	1	↑↑	65%	0
DL4	300	1/6	2	↑↑↑	80%	1
DL5	400	2/6	3	↑↑↑	85%	1
DL6	500	3/6	4	Plateau	90%	0

---

Step-by-Step RP2D Determination

---

Step 1: Identify the MTD Range

• 500 mg: DLTs in 3/6 patients (50%) → unacceptable
• 400 mg: DLTs in 2/6 patients (33%) → borderline
• 300 mg: DLTs in 1/6 patients (17%) → acceptable

MTD estimated around 300–400 mg.

---

Step 2: Evaluate PK Saturation

• Exposure increased from 300 → 400 mg, but incremental gain was modest
• From 400 → 500 mg, no meaningful increase in exposure

PK approached a plateau at ≥300 mg.

---

Step 3: Assess Pharmacodynamic (PD) Target Engagement

• 300 mg: ~80% target inhibition
• 400 mg: ~85%
• 500 mg: ~90%

Near-maximal biological activity achieved at 300 mg.

---

Step 4: Review Preliminary Anti-Tumor Activity

• Partial responses observed at:
• 300 mg
• 400 mg
• No additional responses at 500 mg

Higher doses did not improve efficacy.

---

Step 5: Examine Safety Trends Beyond DLTs

• Grade ≥3 TRAEs increased with dose:
• 300 mg: 2 patients
• 400 mg: 3 patients
• 500 mg: 4 patients
• Evidence of cumulative toxicity at higher doses

---

Integrated Dose-Selection Decision

Decision Framework

Select the lowest dose that achieves:

• Adequate target engagement
• Demonstrable clinical activity
• Acceptable and manageable safety for chronic dosing

---

Final RP2D Selection

RP2D = 300 mg QD

Rationale

• Below or at the lower end of the MTD range
• PK exposure sufficient and near saturation
• PD shows near-maximal target inhibition
• Objective responses observed
• Superior safety margin compared with higher doses

---

Common Pitfalls to Avoid

❌ Equating MTD with RP2D

• Often leads to excessive toxicity in Phase II

❌ Basing RP2D solely on DLTs

• Ignores chronic toxicity, PK saturation, and PD data

---

One-Sentence Takeaway

In oncology Phase I studies, RP2D is the dose most likely to deliver a favorable long-term risk–benefit profile—not simply the highest tolerable dose.

If there is sufficient PK/PD/efficacy and safety evidence for 300 mg, a single RP2D should be prioritized to accelerate development; if there are modeling or biological concerns regarding 350 mg, a small exploratory cohort should be conducted first, and then a decision should be made based on pre-defined trigger rules whether to proceed to a randomized dose comparison.