Skip to main content
Premium Trial:

Request an Annual Quote

Real-World, Umbrella Studies Highlight Challenges of Evaluating Precision Oncology in NSCLC


NEW YORK – Biomarker-driven approaches have transformed the treatment of lung cancer, and advancing knowledge about the genetic underpinnings of the disease has redefined it from a monolithic category to multiple tumor types characterized by the genetic alterations driving them.

Meanwhile, the US Food and Drug Administration has approved a variety of molecularly targeted lung cancer drugs, and treatment guidelines now recommend testing for these biomarkers as part of the standard of care.

But, despite these advances, some experts caution that just because precision oncology has proven successful for some lung cancer patients, that doesn't mean it will benefit all lung cancer patients, or even all cancer patients. "Two things can be true simultaneously," said Vinay Prasad of the University of California, San Francisco, who has argued that precision oncology hasn't yet proven to benefit most patients with cancer, despite the outcomes gains in very particular subgroups.

"A few drug [and] tumor mutation pairings can be really beneficial," he acknowledged, "but simultaneously, the whole strategy might be detrimental … The average person might see a detrimental effect from broad genomic sequencing and unproven drugs."

Assessing the benefit of precision oncology has proved challenging due to the heterogeneity of cancer. Targetable biomarkers are often rare, making it difficult to enroll patients in clinical trials. Moreover, delays in test results, difficulties in access to off-label drugs, and advanced, heavily pretreated populations too sick to benefit from treatment, make it even more difficult to evaluate the effect of these treatments.

While Prasad is a crusader for the randomized-controlled trial when it comes to assessing whether precision oncology is truly helping patients live longer and better overall, many researchers, drugmakers, and even the FDA, have said that such studies are difficult to enroll, may be unethical given the broad acceptance of genomic profiling in cancer care, and will take too long to complete. As such, the field has largely been willing to accept evidence from small, single-arm studies showing robust response rates to precision oncology drugs and real-world data.

In one study, published in The Oncologist in early July and sponsored by Roche, researchers conducted a retrospective analysis of real-world data to suggest that advanced NSCLC patients who receive companion diagnostic testing for genetic alterations for which there are FDA-approved treatments live longer than those who receive treatments uninformed by testing. Based on the data, the authors believe that advanced NSCLC patients should be routinely offered CDx testing to profile their tumors for targetable biomarkers and identify opportunities for precision treatment.

Roche investigators assessed data from thousands of advanced non-squamous NSCLC patients treated at oncology practices within its subsidiary Flatiron Health's network and compared survival outcomes between those who received CDx testing and those who did not. They defined CDx testing as any test to identify alterations in ALK, BRAF, EGFR, and ROS1, and PD-L1 expression levels in patients' tumors. These are all extensively studied predictive markers in NSCLC with associated FDA-approved drugs that are part of the standard of care in NSCLC.

The researchers analyzed data from 17,555 advanced NSCLC patients and found that the 14,732 patients who received CDx testing lived an average of 13 months versus six months for the 2,823 patients who did not receive CDx testing or for whom testing status was unknown.

They also looked at the factors associated with which patients were more likely to receive CDx testing and concluded that many patients who should be tested according to guidelines are not being tested.

Indeed, patients who received CDx testing were those who were more likely to test positive for targetable oncogenic driver mutations, such as patients with light tobacco exposure and patients who were younger, female, and Asian. Additionally, the study reported access disparities; nearly 92 percent of patients who received CDx testing were insured, and the tests were given to a greater population of patients treated in the Northeast and Western regions of the United States than in the Midwest. The results, accordingly, did not account for all patients with advanced NSCLC so much as those with the ability to access the tests and those who were more likely to have the markers of interest.

Based on the two-fold survival increase for patients who received CDx testing compared to those who didn't, however, the researchers concluded that the data from this study support the value of precision oncology in NSCLC, and that lung cancer patients should be broadly offered CDx testing. 

"Precision medicine … is an expanding treatment approach that can help match patients to the most effective treatments sooner based on an individual's characteristics, including specific molecular targets likely to respond to biomarker-driven therapies," the researchers wrote. "Thus, it is imperative that healthcare providers and payors prioritize timely patient access to CDx testing, incorporate it as a critical first step to inform individualized treatment strategies, and use it as an opportunity to leverage multidisciplinary molecular tumor boards to improve management and timely delivery of precision medicine for patients with non-squamous advanced NSCLC."

The problem, of course, is that in the real world, patients are receiving information about validated CDx markers with associated FDA approved drugs alongside markers with scant evidence on their ability to predict response to treatments. With the increasing use of large next-generation sequencing panels that gauge mutations in hundreds of genes, patients are receiving drugs off-label and within clinical trials based on biomarkers with limited evidence suggesting that the strategy may benefit patients.

Shortly after Roche published this study, the results from the National Lung Matrix Trial of Personalized Therapy in Lung Cancer were published in Nature, highlighting some of the pitfalls in this regard. In this large umbrella trial, researchers assessed the benefit of drug-biomarker strategies based mostly on promising preclinical data, but many of the treatment arms failed to accrue, and only a few showed benefit. By the end of the trial, the cohorts that ended up yielding positive results were those that consisted of patients with molecularly defined indications who received treatments already known to work. This study recently became a flashpoint for controversy among oncologists, with precision oncology skeptics using it as an example of why the field is overhyped.

"Ninety-nine percent of oncologists believe precision oncology is the greatest thing since sliced bread," Prasad said, arguing that the National Lung Matrix Trial results suggest otherwise. "When they're in a difficult clinical situation, the first thing they do is run broad genetic panels and then start trying to figure out if there are some mutations they can drug. The idea that you're going to find these mutations and drug them is so seductive that it dominates the field to the point where oncologists in the real world will skip giving a proven drug in the hopes that they can give some of these genomic drugs that have no data at all."

Indeed, in the past decade, top cancer centers around the country have invested in internal sequencing capabilities, hired genetics experts, and organized molecular tumor boards all with the goal of making sense of complex genetic testing data on rare markers, identifying marketed or investigational drugs that can interrogate those markers, and bring precision medicine to more patients. There are efforts now to decentralize these precision oncology frameworks so cancer patients in the community setting can also have access to precision oncology. 

But these efforts don't always benefit patients, as seen in the National Lung Matrix Trial. In this study, many of the drug matches lacked substantial clinical data, and the guiding preclinical data did not translate into patient responses. Despite negative findings in this and other studies, Prasad believes most oncologists are still holding out hope that all biomarker-guided treatments will benefit their patients in the way imatinib (Novartis' Gleevec) transformed treatment for Philadelphia chromosome-positive chronic myeloid leukemia, or shrink tumors in 80 percent of their patients, like the ALK-targeting alectinib (Genentech's Alecensa).

"The way the mind works is that we generalize a little bit too quickly," Prasad said. "We [oncologists] think that because the best precision oncology drugs are actually good, that whenever you apply precision oncology, that it's also good. That's the logical leap where people go wrong."

Value of a negative trial

On the other hand, there are experts who argue that when arms of studies like the National Lung Matrix Trial are negative, that shouldn't be a black mark against precision oncology as a discipline but as a necessary reality of advancing science. Roy Herbst at Yale Cancer Center is one oncologist who believes more umbrella studies should be done because their design makes it possible to test out matched treatments for very rare subsets of patients, such as the roughly 1 percent of NSCLC patients with NTRK fusions or the 2 percent who have ROS1 rearrangements.

"All of these [rare oncogenic drivers], you would never find them if you didn't have some sort of master protocol," said Herbst, who is the study chair of the ongoing US-based Lung Cancer Master Protocol (Lung-MAP), which is another umbrella trial assessing precision oncology strategies in lung cancer.

He acknowledged, however, that large umbrella trials can be challenging territory for investigators when it comes to matching patients to drugs based on limited biomarker evidence. "Sometimes you don't have a lot of data for a drug in this tumor type," he said. "Maybe the data for the drug is in another tumor type [or] maybe you have some preclinical data, but that's always challenging. Lung-MAP has been a very important learning experience for me in that area."

The National Lung Matrix Trial, considered the largest umbrella trial in NSCLC, also faced these same challenges. Researchers led by Gary Middleton of the University of Birmingham reported results from 19 out of 22 different biomarker cohorts to which patients were assigned based on the results of next-generation sequencing. Patients in each cohort received one of eight targeted drugs or a checkpoint inhibitor, and the trial's adaptive design allowed investigators to continue or stop arms based on periodic assessments.

Three of the original 22 cohorts were closed because they did not accrue enough patients, and among patients in the 19 remaining cohorts, response rates were mostly poor. In 15 of these cohorts, investigators observed either zero or single-digit response rates. There were a few cohorts that saw higher, encouraging response rates, but in several cases, these occurred in molecularly defined indications for which the FDA approved treatments during the study. This is positive for patients and reconfirms these agents' activity, but it biased the evaluation in these cohorts.

"For the approved drugs, these types of trials give us post-marketing experience," Herbst said, arguing that there is still value to glean from the responses seen in these cohorts, which others may be quick to write off as irrelevant due to their already known outcomes. "Many times, these drugs are getting approved [based on data from] small cohorts, so large biomarker-based studies like this can increase the denominator."

After excluding instances in which "actionable aberrations had become apparent during the timeframe of NLMT," Middleton and colleagues pointed out the sobering fact that out of all remaining patients screened in the study, there were only nine confirmed overall responses. Middleton didn't reply to a request for comment for this story.

"Despite preclinical data supporting the drug-biomarker combinations [used in this study], evidence shows that a limited number of combinations demonstrate clinically relevant benefits," the authors wrote. In other words, the National Lung Matrix Trial was, for the most part, a negative trial.

In Herbst's view, negative studies like this are important for advancing precision medicine. "Even if a trial like this doesn't hit the mark … a lot of these outcomes are important negatives," he said. "If you have a biomarker and a drug for it, and you don't see activity, that tells you that [that biomarker] might not be the driver. Or you might need combinations."

The nature of precision oncology

Geoffrey Oxnard, who was not directly involved in the Roche study but who recently joined Foundation Medicine, a Roche subsidiary, as VP and global medical lead for its liquid biopsy portfolio, argued that the very nature of precision medicine makes it challenging to assess its value for NSCLC in a single study. Because precision oncology research is progressing rapidly, studies like the National Lung Matrix Trial — designed to clinically test out multiple biomarker-guided treatment strategies often based on early data — can be negatively impacted as more mature evidence comes to light as the trial progresses.

Instead of looking to one positive or negative trial to support the value of precision oncology, Oxnard places more emphasis on the survival and quality-of-life improvements he's seen among patients successfully treated with biomarker-guided approaches. "I don't know if a one-size-fits-all clinical trial solution to say, 'precision medicine: yes or no' exists," Oxnard said. "But when I see the remarkable story, for example, with MET exon splice site mutations, which just a few years ago was a crazy genomic discovery but [now have led to] the FDA approval of potent oral MET inhibitors … that, to me, is the story of precision cancer medicine. And stories like that make me believe."

In May, the FDA approved capmatinib (Novartis' Tabrecta) for metastatic NSCLC patients with MET exon 14 skipping mutations alongside Foundation Medicine's FoundationOne CDx to identify best responders. 

With regard to the Roche paper specifically, Oxnard called the overall survival improvement associated with CDx testing "only part of the picture" and noted that the value of precision medicine in NSCLC depends on multiple factors: whether patients receive testing in the first place; whether the testing they receive is comprehensive and identifies a targetable mutation driving their cancer; whether an effective drug exists to target that mutation; and then, ultimately, whether the patient is able to access and receive that drug. 

The real-world data analysis showed there are many factors that influence whether patients receive testing. "Certain kinds of patients get access to precision cancer medicine and others don't," Oxnard acknowledged. He added that Foundation Medicine is actively working to expand access to FoundationOne CDx, which features multiple FDA-approved CDx indications, and assesses more than 300 genes and several genomic signatures, so that all patients with advanced cancer, not just lung cancer patients, can benefit from a precision medicine approach. 

"We are rapidly growing out the utility associated with the test," he said. "And while the value [of precision medicine] will be related to the likelihood of finding [an actionable mutation], that likelihood seems to be growing every day."

Arguing for randomization

Prasad, meanwhile, has maintained that broader utilization of genomic profiling and precision medicine is ill-advised and may even harm patients in the absence of randomized-controlled trials objectively showing that such an approach benefits patients over the standard of care. Analyses like Roche's using real-world data provide a biased view of the value of precision oncology, he noted.

"As a general rule for any technology that is gradually being utilized over time, people who gain access to it early versus the people who gain access to it later are going to be fundamentally different groups of people. … You can never look at [the technology's value] retrospectively," he said. "It's just not a useful way to do it. The bias will always go in one direction. It's always likely to inflate the benefit."

There will always be factors — including socioeconomic status or interactions that occur between physicians and patients on an individual level — that are not possible to fully control for, making the analyses inherently biased, he argued.

For example, Roche's real-world data analysis suggested that patients' clinical factors, such as non-smoking status, appeared to be associated with the likelihood that they would gain access to CDx testing. And while the investigators used a modified Lung Cancer Prognostic Index in their statistical analysis to control for the possibility that some of these factors might have impacted survival outcomes, the adjustments were limited to data collected within the Flatiron Health Database. Scenarios in which doctors choose not to offer testing to patients based on more nuanced factors, such as individual conversations, aren't always captured in medical records or patient data, nor are they possible to measure. As such, in Prasad's view, retrospective analyses of real-world outcomes can never completely control for every variable that can influence the results.

"Authors always say they adjusted for factors to minimize bias, but some of these methods have baked-in biases that won't go away," he said. "To test the strategy of whether every cancer patient should be sequenced, you need a randomized trial … every other design is misleading."

In contrast, despite the challenges faced by the non-randomized, multi-cohort National Lung Matrix Trial, Prasad felt the study achieved what it had set out to. "To be fair to the National Lung Matrix Trial authors, their goal was not to prove precision oncology is beneficial, but to say which of these drugs were worth considering further," he said. The mostly disappointing results, however, did show that the benefit of precision oncology in NSCLC is still limited to select patients with known targetable mutations.

Despite Prasad's arguments in favor of randomized-controlled trials in precision oncology, he doesn't expect them to occur any time soon, not because they're too difficult to accomplish, as some in the field have maintained, but because the outcomes would be too disappointing.

"People know they're not going to win if they do that type of study," he said. "If the National Lung Matrix Trial is as negative as it is, the prospect of a randomized study has a very low probability of success." Prasad pointed to the disappointing results of the randomized controlled SHIVA study, published several years ago, to illustrate why this type of study is not being done.

SHIVA found that patients matched to targeted therapy based on biomarkers did not experience a progression-free survival benefit compared to a control group, and some experienced worse toxicities. To be sure, precision oncology proponents have pointed out a number of flaws in SHIVA, but Prasad said the results nonetheless deterred investigators from randomization in this setting.

"If you repeated SHIVA three more times and all three trials are negative, the results would be catastrophic," he said, highlighting the immense resources that have been funneled into precision oncology on the belief that the approach will continue to benefit more and more patients.

There are plenty of experts who reject Prasad's view that investigators are not doing randomized studies to avoid findings that would surely be negative. These experts instead argue that given the well-known success stories in precision oncology, patients may not be willing to be randomized to an arm where they would not receive tumor-matched treatment.

Notably, within the MD Anderson Cancer Center's IMPACT II study of targeted, tumor-matched therapy, investigators are trying to navigate the challenges of randomization in the precision medicine era by allowing patients to choose whether they'd like to be matched to a treatment or receive best available care. The study — which is not limited to NSCLC, but enrolling all patients with metastatic, previously treated cancer — has an expected completion date of December 2024.

"[Molecular] testing is now routine at so many centers, and so many oncologists view it as standard-of-care that such a randomized trial would be hard to implement," Oxnard said. "You would effectively be withholding a tool that many already see as standard of care [in the control arm] and so I would envision such a trial might not be ethical."

Oxnard's views must be considered in light of his new position at Foundation Medicine, a leader in NGS testing. However, he said his experiences over a decade treating patients at the Dana-Farber Cancer Institute made him a stalwart believer in precision oncology. "I don't need a randomized trial to convince me that precision oncology is a compelling option for patients when the alternative is toxic empiric therapy," he said.

Instead of going the randomized route, Oxnard believes that companies like Foundation Medicine will "leverage real-world evidence in an unbiased fashion" to demonstrate the value of tests like FoundationOne CDx and its blood-based counterpart, FoundationOne Liquid, in identifying precision medicine options for patients.

Meanwhile, Herbst, whose Lung-MAP umbrella trial features single-arm experiments and secondary randomized arms, believes randomization has a role in precision oncology in specific contexts. "You don't want to randomize if you don't have any activity," he said. "And if a single arm is extremely positive, you may not need randomization. You don't want to randomize patients to [chemotherapy] or a standard of care that's underwhelming unless you really have to."

Where randomization becomes necessary, Herbst continued, is in cases where an agent appears to have moderate activity compared to available options. In the Lung-MAP trial, matched treatment arms that demonstrate efficacy within a certain pre-specified range are expanded into randomization, with patients assigned to either the targeted agent or standard of care, which is usually chemotherapy or a combination of chemotherapy and a VEGFR2 inhibitor.

Ultimately, Herbst believes researchers will likely continue to rely on umbrella trials with master protocols when evaluating precision oncology approaches, as in the National Lung Matrix Trial and Lung-MAP, given their efficiency. "These master protocols have set the standard," he said. "You biopsy patients, you sequence them, you do broad-based next-generation sequencing, you look for the best marker you can possibly find, and you try to match those with the best drugs."

Going forward, Herbst anticipates the field will embrace the "next iteration" of master protocols in lung cancer precision medicine and test combination regimens with immunotherapy or multiple targeted agents in the hopes of addressing precision medicine's Achilles heel: acquired resistance.