While most of what the FDA laid out in a recent draft guidance on the assessment of overall survival in oncology clinical trials was expected, the document still raised one concern for Eli Lilly’s oncology chief, Jake Van Naarden, and other experts.
In the draft guidance, released in August, the FDA suggested that “in general,” the use of crossovers in oncology clinical trials “should be limited.”
Crossovers allow patients in the control arm to receive the drug used in the experimental arm upon disease progression. Doing so may mask the drug’s efficacy in the tested disease setting, sometimes leading to unfavorable overall survival data. As the FDA duly noted, “[c]rossover can impact the interpretation of overall survival results because it can confound treatment effect estimates.”
The overall theme of the guidance is to emphasize the importance of the evaluation of overall survival as a key trial endpoint.
Simultaneously, the FDA is pushing for more U.S. patient representation in oncology trials.
“Unfortunately, if you take a look at all the oncology trials that come to us, only about 20% of the population is derived from the United States,” Richard Pazdur, M.D., director of the FDA’s Oncology Center of Excellence, said during an advisory committee meeting in May. “We’d like to see robust increased enrollment in the United States.”
To Van Naarden, these two policies are not compatible with each other.
“I don’t think you can have both of those,” Van Naarden said in a recent interview with Fierce. “We need to pick a lane here as a country. Which one do we care more about? Proving survival, which means no crossover; or getting U.S. participation, where crossover is a pretty important incentive.”
Lilly’s chief medical officer, David Hyman, M.D., echoed that tension.
“We struggle every day to accrue our patients from the U.S., and it’s a major focus of the organization, and this just makes it that much harder, unfortunately,” he said in the joint interview, calling crossover and its impact on trial enrollment the main topic Lilly had in its discussions of the draft guidance.
The FDA acknowledged in its draft guidance, too, that “[i]nclusion of crossover in clinical trials can improve enrollment and help with retention of enrolled participants.”
Related
Lilly has been caught by both policies in the past. In a high-profile case in 2022, the FDA declined to approve Innovent Biologics’ then Lilly-partnered PD-1 inhibitor Tyvyt because the phase 3 trial supporting the application was conducted in China without U.S. participation and featured a weak comparator arm. Lilly later terminated its collaboration with Innovent on the drug in countries outside of China.
Then last year, the phase 3 Bruin CLL-32 trial of Lilly’s novel BTK inhibitor Jaypirca in second-line chronic lymphocytic leukemia or small lymphocytic lymphoma missed its OS endpoint after a high, 76% crossover rate.
More recently, Lilly’s partner Lantheus called it quits on their PSMA-directed radioligand candidate PNT2002 after an 85% crossover rate confounded the phase 3 Splash trial’s overall survival readout in metastatic castration-resistant prostate cancer. After the trial reached the prespecified number of deaths for a final analysis, the OS results did not turn in favor of PNT2002.
With the OS readout, Lantheus has decided not to pursue an approval with the FDA because the agency has not been lenient with trial results that suggest any hint of potential detriment to patients’ survival.
Related
While OS data from a crossover design could scupper—or cause major delays, such as in the case of Novartis’ rival PSMA radiotherapeutic Pluvicto—a phase 3 readout, it can also sometimes put one trial’s data at a disadvantage compared with a rival study that doesn’t feature crossovers.
At the 2025 European Society for Medical Oncology (ESMO) Congress this past weekend, AstraZeneca and Daiichi Sankyo’s Datroway showed in the Tropion-Breast02 trial that it can reduce the risk of death versus chemotherapy by 21% in first-line triple-negative breast cancer patients who are not candidates for immunotherapy.
At the same time, Gilead Sciences’ rival TROP2 antibody-drug conjugate Trodelvy is not yet able to show a survival benefit against chemo in the same disease setting, with a mere 2% trend in favor of the drug, albeit from a shorter follow-up time compared with the AZ/Daiichi trial, according to results published at ESMO 2025.
Because Trodelvy is FDA-approved as a later-line therapy for triple-negative breast cancer (TNBC), Gilead’s ASCENT-03 trial allowed crossover; among control-arm patients who initiated subsequent therapy, 82% received Trodelvy. In contrast, only about 30% of patients in the Datroway trial received any ADC in subsequent treatments. Datroway is not approved as a later-line TNBC treatment, and its Tropion-Breast02 trial does not feature a crossover design.
The Lilly execs aren’t the only ones finding the FDA’s crossover policy problematic. In an article published last week in The Lancet Oncology, three oncologists led by Bishal Gyawali from Queen’s University in Canada argued that the FDA should generally “encourage, rather than discourage, crossover in all trials testing an already approved drug.”
In cases where a drug has already been shown to improve OS in later lines, crossover should be mandated in an earlier-line study involving the drug, “because patients in the control group should receive the best proven therapy at progression,” the experts wrote in the article.
In a 2023 Nature Reviews Clinical Oncology article, Gyawali argued that even for new “me-too” drugs following the approval of an agent in the same drug class, crossover should also be mandatory. Such was the case with Lantheus’ PNT2002 trial, but not AZ and Daiichi’s Datroway study.
The FDA appeared to agree with the scenarios Gyawali proposed, although such strong words as “mandatory” were not used.
Writing in an article published last fall in the journal Clinical Cancer Research, FDA staffers, including Pazdur, suggested that while crossover may complicate OS analysis, it “may be appropriate under certain circumstances.” These include “the investigational therapy is approved for a later line of therapy; drugs similar to the investigational therapy are available off-study,” and “challenging patient recruitment is expected without crossover,” among others.
In that article, the FDA also recognized that a crossover design “may appeal to potential trial participants by increasing their chances of receiving the investigational therapy,” but also said it reflects “the unproven assumption that an investigational therapy is superior to [standard of care].”
In the current draft guidance, the FDA only said that crossover “would be most appropriate” in disease settings that “have no other or very limited therapeutic options.”
Comments on the draft guidance closed yesterday, Oct. 20. The U.S. government’s website shows the document received 33 comments. In one of five comment documents displayed, three doctors from Italy, Kenya and Morocco recommended that the FDA expand the discussion on crossover “to better reflect real-world complexities.”
In addition to raising the same scenario where a later-line approval already exists, the doctors argued that even if such an approval isn’t already in place in a particular country, offering ex-protocol access to experimental drugs is also important where regional disparities in access exist. This would create a quasi-crossover effect even if crossover is not part of the trial’s protocol.
“Ensuring equitable post-protocol treatment access, especially when the investigational drug is already available and manufactured by the sponsor, reflects not only sound scientific design but also ethical imperatives,” the doctors wrote.
FDA guidance documents are not binding, yet they outline important development pathways for drugmakers.
“We can agree or disagree on the finer points,” Lilly’s Hyman said. “But at least if these things are in guidances […] the roads are very clear and written, and I think that’s a step in the right direction.”
