Gary Middleton, MBBS, MD
Targeted therapies for non–small cell lung cancer (NSCLC) produced mixed results in a novel multiarm clinical trial that matched tumor mutations to drugs targeting the mutations.
An interim analysis of 19 mutation-defined patient cohorts identified several agents with promising activity. For example, crizotinib (Xalkori) had >99% probability of achieving a defined rate of durable clinical benefit (DCB) and objective response (OR) rate for patients with ROS1
fusions and MET
exon 14 skipping. Many of the match-ups in the National Lung Matrix Trial (NLMT), however, appeared unlikely to reach predefined goals for DCB, OR, or progression-free survival (PFS).
In a presentation at the 2019 World Conference on Lung Cancer, British cancer researcher Gary Middleton, MBBS, MD, suggested several explanations for the results and strategies to turn “apparently poor targets into potentially good targets.”
“One of the risks with the paradigm of precision medicine is that we focus on very high levels of single-agent activity in oncogene addiction,” said Middleton, of the University of Birmingham. “We might be missing a trait, or we might be missing a useful target in our patients.”
As an example, he cited palbociclib (Ibrance) and capivasertib (both chosen for the NLMT) for estrogen receptor–positive breast cancer. As single agents, the drugs had very modest activity, but studies subsequently showed that the drugs became much more effective in combination with fulvestrant.
“Critically, there is no evidence at all that the putative biomarkers for palbociclib or capivasertib are actually useful,” he said.
Tobacco-associated lung cancer poses a multiplicity of obstacles for therapeutic development: genetic instability, ongoing evolution, oncogene predominancy, and numerous pathways to resistance.
“That brings up the criticality of the preclinical models that we need,” said Middleton. “We need to make sure that the targets we are gunning for are tested in models that completely replicate the genomic complexity and chaos of tobacco-associated non–small cell lung cancer.”
The observations and comments followed his presentation of initial findings from the ongoing NLMT. The trial currently has 22 mutation-defined cohorts within four pathways of genomic alteration treated with one of seven targeted drugs: cell cycle progression (palbociclib), RAS
activation (palbociclib, vistusertib, and selumetinib/docetaxel), PI3K/PTEN/AKT/mTOR
(vistusertib, capivasertib), and RTK
signaling (FGFR inhibitor AZD4547, crizotinib, osimertinib [Tagrisso]).
Each cohort eventually will have 30 patients. The Bayesian adaptive design of the trial allows for interim analyses to predict outcomes when a cohort is fully accrued.
The principal outcomes of interest are OR and DCB (stable disease or better for ≥24 weeks), except for the RAS
activation category, which has endpoints of DCB and median PFS. The target estimates for OR and DCB are >30% for single-agent treatment and >40% for a combination, and the median PFS target is 3 months.
Investigators used Bayesian methodology to estimate true rates and medians and credible 95% confidence intervals. A treatment is considered a “go” for further evaluation if the posterior probability of exceeding a target estimate is >0.50.
Middleton reported findings from interim analyses (N = 15) for 19 of the 22 subgroups.
For the cell cycle progression category, median PFS with palbociclib exceeded the 3-month target for tumors with CDKN2A
loss (2 subgroups) and CCND1
amplification. Predictive probability of success (PPoS) ranged from 0.69 to 0.98. Median PFS remained below the target for tumors with CDK4
amplification, which was associated with a PPoS of 0.23. The DCB target had yet to be reached for any of the 4 subgroups.
In the RAS
activation category, the combination of selumetinib and docetaxel had a 50% DCB and PPoS of 0.89 for the subgroup with NF1
mutation, as well as a median PFS of 5.6 months. Palbociclib for tumors with KRAS
mutation resulted in a median PFS of 5.4 months, a predicted probability >0.99, and also had a DCB rate of 44%. Three other subgroups appeared unlikely to meet DCB or PFS targets.
In the PI3K/PTEN/AKT/mTOR
category, the DCB rates for the 5 subgroups ranged from 9% to 21%, associated with PPoS of 0.06 to 0.34. The PPoS for the response target was <0.20 for all of the subgroups.