A Nature Medicine paper landed on Feb. 2 like a stone in a pond, sending out ripples that quickly turned into tidal waves across biomedical circles. By midday, renowned physician-scientist Eric Topol, M.D., had signaled its importance to his more than 700K followers on X, then hailing it as “probably the best proof of the importance of timing of therapy that we’ve seen in medicine to date.” By the time the Super Bowl kicked off on Feb. 8, the study’s 3 p.m. threshold has become a punchline, as longtime biotech investor and TV host Brad Loncar jested that the Patriots’ sluggish offense was “like immunotherapy administered after 3:00 PM.”
However, by Feb. 19, the journal would put up an editor’s note, alerting readers to several concerns around the article, titled “Time-of-day immunotherapy in non-small cell lung cancer: a randomized phase 3 trial.” Topol apparently deleted his comment at some point as well.
At the time, concerns raised against the article ran the gamut from major revisions to trial design and inconsistent trial protocol, to unusual efficacy and safety data patterns.
Now, after a four-month investigation, Nature has decided to pull the study.
“[D]ue to the amount and nature of the problems identified, the editors no longer have confidence in the integrity of the results,” according to a retraction note obtained by Fierce.
The 3 p.m. divide
The Chinese study, a randomized phase 3 trial known as LungTIME-C01, found that patients with advanced non-small cell lung cancer who received their first four cycles of a standard PD-1 inhibitor combined with chemotherapy before 3 p.m. experienced significantly better outcomes than those treated later in the day. The PD-1 drugs used were either Merck & Co.’s Keytruda (pembrolizumab) or Innovent Biologics’ Tyvyt (sintilimab).
The early treatment group had a massive, statistically significant 60% lower risk of disease progression or death. In a nearly two-fold difference, the early-treatment group enjoyed a median progression-free survival (PFS) of 11.3 months, compared with 5.7 months for the late-afternoon group.
Overall survival (OS) was similarly improved substantially. Early treatment of immunochemotherapy led to a 58% reduction in the risk of death, as the median OS was 28 months, versus 16.8 months for those treated after 3 p.m. in a day. The team said it was conducting additional follow-up to evaluate long-term survival outcomes. For both PFS and OS, the p-value was below 0.001.
Many experts adopted a cautious view toward the results. Before Nature’s editorial alert, Charu Aggarwal, M.D., who leads the head and neck and thoracic cancers practice at Penn Medicine, called the magnitude of benefit associated with earlier-day PD-1/chemo treatment “somewhat surprising,” and noted that it “warrants careful consideration.”
“These results underscore the need for international, multicenter prospective studies to validate and extend the time-of-day findings across diverse populations—and an emphasis on controlling for factors other than demographics and tumor related factors alone,” Aggarwal said in an email interview with Fierce in early February.
Separately, in an interview with Fierce before Nature launched its investigation, Leerink Partners analyst Daina Graybosch, Ph.D., called the study “hypothesis-generating” because it’s just one study conducted at one single center, Hunan Cancer Hospital in China. Yet its prospective design, randomized nature and effect size lent it weight to a degree that “every trial with an immuno-oncology drug designed should be thinking about how you take this into consideration,” Graybosch said at the time.
Merck already collects time-of-treatment information in all its oncology trials, Cathy Pietanza, M.D., vice president of global clinical development at Merck Research Laboratories, told Fierce in February in response to a question about the Nature Medicine article.
Graybosch also argued that any new study with similar time-of-day designs should collect translational data to help better understand the findings’ underlying biological explanations. If verified, the findings could be practice-changing, she said.
There’s a biological rationale behind the findings. Growing evidence shows that the immune system also operates on circadian rhythms. Just as humans wake up in the morning and rest in the evening, immune cells have been found to be more active during daylight hours and are primed to mount stronger inflammatory and cytotoxic responses.
Multiple retrospective studies have supported circadian immunology in cancer treatment, although those analyses are inherently subject to confounding factors. An international retrospective study involving more than 1,000 patients with relapsed or refractory large B-cell lymphoma linked morning infusion of CD19-directed CAR-T therapies to a higher one-year PFS rate than afternoon administration, although OS was similar, according to a study published March 2026 in the journal Blood.
The now-retracted Nature Medicine study also found that early-day infusions were associated with increased levels of circulating CD8+ T cells and a higher ratio of activated-to-exhausted CD8+ T cells, which the authors suggested “may partly explain” the results. But as Graybosch noted, it’s hard to parse correlation from causation.
Because anti-PD-1 antibodies and CAR-T cells persist in the body long after administration—Keytruda has a half-life of more than 20 days—many experts remain deeply skeptical that the time of day a treatment is given could drive meaningfully different treatment outcomes.
Acknowledging that argument, Graybosch wondered if a first-impression phenomenon is at play. Biochemically speaking, Keytruda has a slow off-rate, as it takes a long time for the antibody to dissociate from its target receptor. That could translate into diverging treatment effects if the antibody meets and binds to a different set of immune cells first when infused in the morning, she suggested. But again, this explanation is merely a hypothesis that needs to be validated.
Myriad questions
For LungTIME-C01, the jaw-dropping treatment effect and high statistical significance generated from a trial of just 210 randomized patients initially swept some people off their feet. To others, however, the results were simply too good to be true. Documentation discrepancies and certain abnormalities within the findings quickly drew scrutiny.
For starters, the study’s ClinicalTrials.gov record shows multiple major mid-trial design changes following its 2022 launch. These included switching the study type between interventional and observational in 2024, altering the primary and secondary endpoints, and making a near-complete revamp of the enrollment inclusion and exclusion criteria.
While such modifications do not inherently invalidate a trial’s results, they “introduce potential concerns regarding bias and interpretability,” Penn Medicine’s Aggarwal noted.
What’s more, the trial protocol presented as a supplement to the Nature Medicine paper bore the date Jan. 2, 2022, but it contained multiple references to articles published later in 2022, 2023 and 2024.
During its investigation, Nature uncovered other discrepancies between the original study protocol in Chinese and the translated versions provided to the journal during the peer review process and for publication.
The authors provided a time-stamped version of the original protocol, attributing the issues largely to “administrative errors rather than prospectively documented amendments,” according to the retraction note.
Multiple concerns were also raised surrounding “unexpected data patterns” in the study, Nature flagged in its note.
For example, the PFS curve appeared smoother than what’s expected from similar phase 3 trials. In prospective oncology trials, patients typically schedule follow-ups in rigid intervals. In this case, patients undergo imaging scans to assess changes in tumor burden every two 21-day treatment cycles. Because patients are only evaluated at these designated timepoints, a standard PFS curve often drops in discrete stair steps.
In response to Nature, the study’s authors acknowledged that, due to COVID-19, the trial deviated from fixed-calendar imaging schedules and cycle-based assessment, the retraction note stated.
In another unusual phenomenon, the trial recorded zero censoring in the first year, indicating that every single patient was successfully tracked and remained on treatment throughout the period.
The trial also notably recorded no adverse events leading to treatment discontinuation in either group.
“This is not surprising; with proactive toxicity management, many patients are able to remain on therapy without requiring discontinuation,” Aggarwal said.
For context, another recent phase 3 trial conducted in China, Akeso’s Harmoni-6 trial in first-line squamous NSCLC, recorded treatment-related adverse events leading to discontinuation in just 4.5% of patients receiving BeOne Medicines’ PD-1 inhibitor Tevimbra (tislelizumab) and chemotherapy. That study is about 2.5 times the size of LungTIME-C01.
Critics of the paper also took issue with the finding of no significant differences in immune-related adverse events between the two arms, even as the authors partly attributed the improved efficacy to shifts in T-cell subpopulations.
“While somewhat counterintuitive, this finding is biologically plausible,” Aggarwal said. “Immune efficacy and immune toxicity do not always correlate directly, and circadian modulation may preferentially enhance antitumor immune responses without proportionally increasing immune-related adverse events.”
Further compounding those concerns, Nature’s investigation into the source data provided by the research team indicated that “randomization was performed on the day of treatment for almost all patients, which is uncommon,” according to the retraction note. The lack of adherence to standard tumor progression analysis timing was also cited as a red flag.
Ultimately, the Nature Medicine editors lost confidence in the findings due to the compounding volume—and character—of the identified issues.
According to the retraction note, the study’s authors either agree with the retraction—including all corresponding authors—or didn’t respond to correspondence from Nature. The research team did not respond to Fierce’s request for comment.