Chapter 11: Case Studies of Basket and Umbrella Trials#
Basket trial - one or more targeted therapies are evaulated on multiple diseases that share common risk factors
Umbrella trial - multiple targeted therapies are evalueated for a single diseases
Case 11.1 Trial of Ado-Trastuzumab Emtansine for Pateinets with HER2 Amplified or Mutant Cancers#
Background#
HER2 (Human epidermal factor receptor 2) is an emerging target in multiple cancers
Ado-trastuzumab emtansine is an anti-HER2 monoclonal antibody that was approved for metastatic breast cancer by the US FDA in 2019
In 2018, results based on phase ⅡA basket trial that evaluated whether Ado-trastuzumab emtansine could achieve anti-tumor response in HER2 amplified or mutant cancers of multiple histologies are published
Design and Methods#
Inclusion criteria: HER2 amplification or mutation identified via molecular profiling (ex. lung, endometrial, salivary gland, biliary tract, ovarian, bladder, colorectal, and other cancers)
Dose&route: Ado-trastuzumab emtansine 3.6 mg/kg IV every 21 days
Primary Endpoint: ORR(Overall Response Rate)- CR(Complete Response) or PR(Partial response)(RECIST v1.1)
Study Design: Simon’s two-stage design (Two stage design that allows for early futlilty stopping, Most commonly used in basket trial)
Stage 1: minimum 7 patients per cohort
Stage 2: expand to maximum 18(7+11) patients if stage 1 response is ≥ 1
The trial was powered to detect a true objective response rate (ORR) of at least 40% (considered desirable) versus a null hypothesis ORR of 10% or less
This design achieved 89% statistical power with a one-sided type I error rate of 2.7%
Findings#
HER2‑mutant lung cancer (n=18)
ORR 44%
8 PR,no CR
39% had stable disease up to 11 months
HER2‑amplified tumors (n=53,8 cancer types)
Overall ORR 26%
Salivary cancer 100%, Lung 50%, Endometrial 22%, Ovarian, biliary 17%
Case 11.2 NCI-MATCH (The Molecular Analysis for Therapy Choice of the National Cancer Institute)#
Background#
NCI-MATCH is a phase ⅡA basket trial for advanced refractory solid tumor, lymphoma, or multiple myeloma patients who have progressed on their previous treatment (NCT02465060) Started in August 2015 with 10 treatment arms, NCI-MATCH is one of the largest biomarker-guided trials As of Aprill 22, 2022, accroding to NCT02465060, the trial expanded to 1,425 active sites, enrolling across 50 different tumor types
Design and Methods#
Patients undergo a mandatory biopsy, followed by centralized NGS-based molecular screening.
Based on results, patients are assigned to one of 38 predefined treatment sub-protocols.
Follow-up schedule:
Every 3 months for 2 years
Then every 6 months for 1 additional year
Single-stage design per sub-protocol
Accrual target: 35 patients → aiming for 31 eligible (10% ineligibility assumed)
Primary Endpoint: Objective Response Rate (ORR)
Treatment considered promising if:
ORR ≥ 16% (≥ 5 responders out of 31)
Statistical assumptions:
91.8% power if true ORR = 25%
1.8% one-sided type I error if true ORR = 5%
Findings:#
Patients enrolled: 70
48 patients (69%) treated and eligible for analysis
Overall ORR: 8% (4/48; 95% CI: 3–18%)
Mutation breakdown:
FGFR1 or 2 amplification → 20 patients
FGFR2 or 3 SNV → 19 patients
FGFR1 or 3 fusion → 9 patients
Higher ORR (22%) observed in fusion group (2/9), but too small to confirm subgroup effect
AZD4547 showed limited efficacy in FGFR1–3-altered tumors (ex. lung, gastric, GE junction cancers)
Sub-protocol W did not meet the threshold for a promising response
NCI-MATCH continues to evaluate other targeted therapies across different molecular cohorts
Case 11.3 NCI-MPACT (The NCI’s Molecular Profiling-Based Assignment of Cancer Therapy)#
Background#
NCI-MPACT is a randomized basket trial designed to evaluate the effectiveness of tumour-agnostic, biomarker-guided targeted therapy selection in patients with advanced refractory solid tumors.
In contrast to the non-randomized NCI-MATCH trial, NCI-MPACT randomizes patients to either:
Targeted therapy matched to their tumor’s actionable mutation (experimental arm)
The same therapies assigned without reference to mutation status (control arm)
Genetic Focus to 20 genes across three pathways:
RAS/RAF/MEK (5 genes)
PI3K/mTOR/AKT (5 genes)
DNA repair (10 genes)
Evaluated Regimens:
Trametinib (MEK inhibitor) – RAS
Everolimus (mTOR inhibitor) – PI3K
Veliparib (PARP inhibitor) + Temozolomide – DNA repair
Adavosertib (WEE1 inhibitor) + Carboplatin – DNA repair
Design and Methods#
Randomization: Patients with actionable mutations randomly assigned (2:1 ratio) to targeted or non-targeted arms.
Patient Recruitment: Aimed for 180 evaluable patients (120 experimental, 60 control).
Cohort Structure:
For each targeted arm regimen, up to 30 patients per matching mutation cohort.
Two-stage response-adaptive design:
Stage 1: If no confirmed responses (PR/CR) in first 12 patients per mutation group, cohort terminated for futility.
Stage 2: If ≥4 responses observed among 30 patients (ORR ≥13%), regimen considered promising for that group.
Statistical Power:
88% to detect ORR difference (20% vs. 5%) with a 4% one-sided type I error (objective response).
90% to detect an 80% increase in median progression-free survival with a 1% one-sided type I error.
Findings#
Enrollment: 96 patients with target mutations (2014–2018)
64 to treatment group, 32 to control group.
49/64 in treatment and 16/32 in control received assigned therapy.
Efficacy Results (ORR):
Treatment group: 2% (1/49; 95% CI: 0–10.9%)
Control group: 0%
Regimen-Specific Results:
Trametinib (RAS): 20 treated, 1 partial response (5% ORR)
Everolimus (PI3K): 8 treated, no responses
Veliparib + TMZ (DNA repair): 3 treated, no responses
Adavosertib + Carboplatin (DNA repair): 18 treated, no confirmed responses
Challenges:
Dropout: Higher in controls (22%, 7/32) vs. experimental (6%, 4/64), despite blinding.
Preference: Patients and physicians favored targeted therapy, complicating randomization.
Mutation Prevalence: Two of four regimens failed to meet recruitment goals due to rarity of actionable mutations.
Sample Logistics: High biopsy success rate (~90%), median <10 days for NGS results.
Case Study 11.4: plasmaMATCH (NCT03182634)#
Background#
plasmaMATCH is an umbrella phase IIA trial that evaluated five different therapies for advanced breast cancer.
This trial used plasma-test to detect circulating tumour DNA (ctDNA) on targeted genetic mutations to decide treatments.
ctDNA is found in over 90% of patients with advanced breast cancer
Design and Methods#
Design:Phase IIA, non-randomised, umbrella trial.
Patient Assignment:
Patients were assigned to one of the targeted therapies based on their genetic mutations.
Based on mutation status, they were assigned into one of five treatment groups.
Group A: ESR1 mutation
Group B: HER2 mutation
Group C: AKT mutation
Group D: AKT activation
Group E: Triple-negative (ESR1, HER2, AKT negative)
Interventions:
Group A: Fulvestrant 500mg every 2 weeks
Group B: Neratinib ± Fulvestrant (if oestrogen receptor co-mutated)
Group C: AZD5364 + Fulvestrant
Group D: AZD5364 only
Group E: Olaparib + AZD5364
Statistical Design:
Single-stage design for all groups, aiming for 80% power at a 5% two-sided type I error.
Group A: Target of 78 evaluable patients, with ≥13 objective responses (ORR ≥17%) to be considered promising.
Groups B, C, D: Each assumed a target response rate of 25% and a 5% response rate was unaceepatable.
The regimen was considered to be promising and warrant further clinical investigation if it demonstrated 3 or more responses from 16 evaluable patients.
Findings#
Enrollment and Assignment:
1,051 patients registered.
1,044 underwent ctDNA testing.
357 (34%) had target mutations.
136 patients were assigned to different groups.
Group |
Evaluable Patients |
Target Responses |
Observed Responses |
ORR (%, 95% CI) |
|---|---|---|---|---|
A (ESR1) |
74 |
13 |
6 |
8% (3–17%) |
B (HER2) |
20 |
3 |
5 |
25% (9–49%) |
C (AKT mutation) |
18 |
3 |
4 |
22% (6–48%) |
D (AKT activation) |
19 |
3 |
2 |
11% (1–33%) |
E (Triple-negative) |
1 |
— |
— |
— |
Case Study 11.5: ALCHEMIST Trial (Adjuvant Lung Cancer Enrichment Marker Identification and Sequencing Trial)#
Background#
ALCHEMIST is an umbrella trial, started in 2014, for patients with operable early-stage (IB–IIIA) lung adenocarcinoma (a common sub type: NSCLC(non-squamous non-small cell lung cancer)).
Evaluating two important genetic mutations; EGFR & ALK mutations
Design and Methods#
Multiple sub-protocols:
ALCHEMIST Screening (A151216; NCT02194738)
ALCHEMIST-EGFR (A081105; NCT02193282)
ALCHEMIST-ALK sub protocols (E4512; NCT02201992)
ALCHEMIST-Immunotherapy Treatment Trial(ANVIL)(EA5142; NCT02595944)
ALCHEMIST Chemo-IO Study (A081801; NCT04267848, added in 2020)
Screening and Assignment
Screening is centralized: up to 8,000 adenocarcinoma patients, genomic analysis of EGFR and ALK status, before or after surgery.
Patient assignments based on genomic results:
EGFR mutation: enrolled in ALCHEMIST-EGFR (erlotinib trial)
ALK rearrangement: enrolled in ALCHEMIST-ALK (crizotinib trial)
EGFR/ALK wild-type (regardless of PD-L1): enrolled in ANVIL (nivolumab trial)
Sub-Protocol Designs
All three main therapeutic protocols (ALCHEMIST-EGFR, ALCHEMIST-ALK, ANVIL) are phase Ⅲ randomized controlled trials with Overall Survival (OS) as the Primary Endpoint.
Sub-Protocol |
Treatment |
Design |
Target Accrual |
Events for Analysis |
Statistical Power (α) |
Key Assumptions |
|---|---|---|---|---|---|---|
ALCHEMIST-EGFR |
Erlotinib vs Placebo |
Equal randomization |
410 |
183 deaths |
85% power at 5% 1-sided α |
HR=0.67, median OS control=60mo |
ALCHEMIST-ALK |
Crizotinib vs Placebo |
Equal randomization |
360 (180/arm) |
164 deaths |
80% power at 5% 1-sided α |
HR=0.67, median OS control=66mo |
ANVIL |
Nivolumab vs Observation |
Phase 3 design |
714 |
— |
30% OS improvement or 33% DFS reduction |
— |
Chemo-IO (A081801) |
pembrolizumab during & after chemo) |
Design details not reported |
— |
— |
— |
— |
sequential chemo → pembrolizumab |
||||||
chemo alone |
Findings#
As of January 2020:
5,362 patients were screened in the ALCHEMIST Screening study.
Trial assignments:**
352 randomized to ALCHEMIST-EGFR
99 randomized to ALCHEMIST-ALK
935 randomized to ANVIL
No formal efficacy results yet for ALCHEMIST-EGFR, ALCHEMIST-ALK, or ANVIL.
Data from the screening protocol was reported in May 2020 and March 2022.
Centralized genomics-based screening** allows broad patient assignment to new personalized therapy trials within one umbrella protocol.
The ALCHEMIST platform facilitates efficient accrual and flexible research expansion (including new immunotherapy/chemo-immunotherapy arms).
While efficacy data are not yet mature for primary therapeutic sub-protocols, the infrastructure positions ALCHEMIST to address multiple cutting-edge research questions in early-stage lung adenocarcinoma.
Case Study 11.6: Lung-MAP(Lung Cancer Master Protocol) (NCT02154490)#
Background#
Lung-MAP is one of the largest ongoing umbrella trials in oncology since June 2014 for patients with advanced squamous non-small cell lung cancer (NSCLC).
Initially included five sub-studies (S1400A–S1400E) and later expanded to include:
Non-match sub-studies: S1400F, S1400I
Additional biomarker-driven sub-studies: S1400G, S1400
In January 2019, Lung-MAP was expanded to include all histological NSCLC types under a new screening protocol S1900 (NCT03851445).
It operates across over 800 sites in the United States using a centralized infrastructure.
Design and Methods#
Master protocol approach combining central genomic screening with multiple biomarker-based therapeutic studies.
Includes both randomized and non-randomized sub-studies:
S1400A: Single-arm, non-match study using durvalumab (PD-L1 inhibitor).
S1400B: PI3K inhibitor taselisibn for PIK3CA mutations.
S1400C: CDK4/6 inhibitor palbociclib for CDK4/CCND1/2/3 amplifications.
S1400D: FGFR inhibitor AZD4547 for FGFR1–3 mutation/fusion/amplifications.
S1400E: Rilotumumab vs. erlotinib in MET-mutated patients (closed early due to toxicity concerns).
Most biomarker-driven sub-studies used docetaxel as the control arm.
Seamless phase II/III randomized trial designs applied in biomarker sub-studies.
Findings#
S1400 chapter (June 2014 – Jan 2019):
1,864 patients with squamous NSCLC enrolled.
99% (1,841) submitted tissue for biomarker screening.
91% (1,674) received biomarker results.
84% (1,404) of those with biomarker results were assigned to a sub-study.
47% (655) of assigned patients enrolled in a corresponding sub-study.
Outcomes and results from all S1400 sub-studies are not reported here due to the high number and variation in design; however, several have been published.
Lung-MAP evolved beyond squamous NSCLC and now includes broad NSCLC histologies and new therapies, facilitated by screening protocol S1900 and newer sub-studies (ex, S1800 series).
Debate exists around whether Lung-MAP qualifies as a platform trial due to continuous incorporation of new therapies—but functionally, it is an adaptive umbrella trial.
Lung-MAP set a precedent for multi-arm, multi-target precision oncology trials using centralized infrastructure across a national trial network.
Demonstrated high patient engagement in biomarker screening (submission and analysis) and sub-study assignment.