Nature Medicine Retracted the Lung Cancer Trial That Made Infusion Timing Look Like a Free Win

I wanted this one to be true, and I suspect you did too. In February, Nature Medicine published a lung cancer trial claiming that patients who got their first four rounds of immunochemotherapy before 3 p.m. went nearly twice as long before their cancer advanced. It felt obviously right to me. Your immune cells keep a schedule, so why wouldn’t a drug built to unleash them care when it shows up? Then people started reading the paperwork, and according to Applied Clinical Trials one of the first things they found was a trial protocol dated January 2022 that cited papers published in 2022, 2023 and 2024. On June 24 the journal retracted the paper, saying its editors “no longer have confidence in the integrity of the results.”

The trial, LungTIME-C01, ran at Hunan Cancer Hospital in Changsha. Nature Medicine published it on February 2 as a randomized phase 3 trial. EMJ’s results summary describes 210 untreated patients with advanced non-small cell lung cancer and no driver mutations, split between getting their first four cycles of anti-PD-1 immunochemotherapy before or after 3 p.m. Median progression-free survival came in at 11.3 months for the early group and 5.7 months for the late group, and median overall survival at 28.0 months versus 16.8. Morning blood draws showed circulating CD8+ killer T cells climbing in the early group and falling in the late one. That’s a mechanism, a huge effect and an intervention that costs nothing but a scheduling change, all in one paper. Eric Topol promoted it on X as “probably the best proof of the importance of timing of therapy that we’ve seen in medicine to date.” Honestly, I’d have shared it too.

A 2022 protocol that cited 2024

The trouble surfaced fast. Seventeen days after publication, on February 19, the editors posted a note warning readers about inconsistencies “between the registration record of this trial on clinicaltrials.gov and published version the study protocol, as well as with some of the findings in this study.” Applied Clinical Trials’ account of the four-month investigation fills in the rest. The registry record shows a trial launched in 2022 whose study type switched between interventional and observational in 2024, whose primary and secondary endpoints were altered, and whose enrollment criteria were nearly overhauled. The original Chinese-language protocol didn’t match the translated versions supplied for peer review and publication. The authors blamed administrative errors, and the editors found that explanation insufficient.

The data had its own oddities. The progression-free survival curve was smoother than typical phase 3 curves, which usually drop in a staircase because patients get scanned at fixed intervals. No patients were censored in the first year. No adverse events led anyone to discontinue treatment in either arm. Immune-related side effects showed up at similar rates in both arms despite the enormous efficacy gap, and that drew scrutiny because the authors had explained the survival gap with a shift in T cells. Nearly every patient was randomized on the day of treatment, which the retraction note called uncommon. The authors acknowledged COVID-era deviations from fixed imaging schedules, which the trade outlet read as non-adherence to standard RECIST timing. The corresponding authors agreed to the retraction, and several other authors never answered the journal.

None of this was buried. The protocol date and its citations sat in a supplement that went to peer review, and the registry’s history was public. Even so, a top journal’s reviewers and editors cleared the paper, a famous physician-scientist amplified it, and it took a formal investigation to pull it back. This week Nature Medicine ran a feature by Natalie Healey on chronotherapy moving into clinical trials, and its standfirst says “a high-profile retraction reveals the ups and downs of circadian medicine.” That’s a gentle way to put it. The retracted paper was the journal’s own, and it got through the journal’s own peer review.

Wait, why would a weeks-long drug care about the hour?

I should have asked this in February. The PD-1 drugs in the trial were pembrolizumab or sintilimab, and pembrolizumab stays in the body for more than 20 days. Daina Graybosch, an analyst at Leerink Partners, pointed to that long half-life as a reason large timing effects are hard to accept. She had called the study hypothesis-generating from the start because it came from a single center. If the antibody is still sitting in your blood at 3 a.m. on day nine, why would the hour of the drip matter?

The biology does offer a possible answer, and it’s the part I find most fun to think about. Immune function and immune cell traffic run on daily rhythms, so the first hours after the antibody arrives could matter even if the drug lingers for weeks: which T cells happen to be in which tissue when the drug first reaches them. LungTIME’s CD8 data was supposed to show exactly that. Now it can’t be trusted, and we’re back to a plausible idea without proof.

A commentary in Cancer Immunology Research by Sam Kleeman, Tobias Janowitz and colleagues, published in June before the retraction, raised the point I keep coming back to. Every patient in the early arm got their chemotherapy early too. Even taking the results at face value, the authors argued, the effect could come from circadian variation in how chemotherapy works, from circadian immune regulation, or from plain healthcare-delivery logistics, not from immunotherapy timing alone. They called for factorial trials that separate chemotherapy timing from immunotherapy timing, validated across multiple centers, before anyone reschedules infusion chairs. Some retrospective support for circadian immunology still stands, including a March 2026 analysis in Blood that linked morning CAR T-cell infusion to better one-year progression-free survival in large B-cell lymphoma. But that’s a look back at existing records, not a trial.

The counterweight carries a company’s name

The biggest dataset pointing the other way is the i-TIMES analysis presented by Solange Peters. It pooled 3,060 patients with advanced lung cancer from 8 international phase 2/3 randomized trials of checkpoint inhibitors. In a propensity-matched group of 1,550 patients, median overall survival was 17.3 months for those treated before noon and 16 months for those treated after, a hazard ratio of 1.039 that met the analysis’s predefined non-inferiority criterion. “Treatment timing within the day is unlikely to be a critical determinant of outcomes,” Peters said.

Look at the label, though. The study is cited as “ETOP-Roche i-TIMES,” and Roche sells checkpoint inhibitors. The trials it pooled randomized patients to drugs, not to hours, so its timing comparison is observational even though its source trials weren’t. It cut the day at noon rather than 3 p.m. It matched patients on just four variables (gender, PD-L1 status, liver metastases and brain metastases) and left the 852 patients with mixed timing out of the matched comparison. The coverage reports no confidence interval, and passing a non-inferiority test isn’t the same as showing the outcomes were identical. So the evidence on cancer looks like this: one retracted randomized trial, one industry-partnered null analysis, and a scattering of retrospective signals. No one gets to call that settled, in either direction.

Cardiology went through this same cycle with blood pressure pills. Spain’s Hygia trial put 19,084 patients on bedtime or wake-up dosing and reported a hazard ratio of 0.44 for cardiovascular death with bedtime pills, an effect that looked too good to be true. Then Britain’s TIME trial, funded by the British Heart Foundation, randomized 21,104 people, and vascular death, heart attack or stroke occurred in 3.4% of the evening group and 3.7% of the morning group, not a significant difference. Canada’s BedMed trial in JAMA followed 3,357 patients for a median of 4.6 years, tracking death from any cause plus hospital or emergency visits for stroke, acute coronary syndrome or heart failure, and found an adjusted hazard ratio of 0.96 (95% CI 0.77 to 1.19). The dramatic early benefit didn’t survive two big trials built to check it.

What makes my head spin is what happened inside TIME. In a chronotype sub-study, 5,358 participants answered a questionnaire about their natural sleep timing. For hospitalization with a nonfatal heart attack or stroke, every hour later a person’s internal clock ran carried a hazard ratio of 1.46 (95% CI 1.14 to 1.86) when they took their pills in the morning, against 0.96 (95% CI 0.70 to 1.30) when they took them in the evening. Intermediate chronotypes, about half the group, showed no effect of dosing time at all. The researchers reported fewer heart attacks when dosing matched the body clock, evening pills for night owls and morning pills for early risers, and they said new randomized trials of personalized timing have to confirm it before anyone changes a routine.

So I wonder whether all these trials have been reading the wrong clock. Three in the afternoon on the wall isn’t three in the afternoon for your liver, your lymph nodes or your T cells. For a lark, it’s late in the day. For an owl, it’s barely morning. If infusion timing matters for cancer at all, it might depend on each patient’s internal clock, and a trial that sorts people by the hour on the wall could blur the signal or invent one.

If I were starting immunotherapy tomorrow, I’d take a morning slot if one were open and I’d know that no surviving trial shows it helps, but I wouldn’t push back a single infusion to get one.

Sources

1. Nature Medicine: Huang et al., “Time-of-day immunochemotherapy in non-small cell lung cancer: a randomized phase 3 trial” (retracted June 24, 2026)
2. Applied Clinical Trials: Nature Medicine retracts viral lung cancer timing study after four-month investigation
3. EMJ: LungTIME-C01 results summary
4. Cancer Immunology Research: Kleeman, Janowitz et al., “Cancer chemo-immunotherapy: time will tell” (June 2026)
5. HMP Global: ETOP-Roche i-TIMES pooled analysis of checkpoint inhibitor timing
6. Nature Medicine: Healey, “Only time will tell” (October 16, 2026)
7. eClinicalMedicine: TIME chronotype sub-study (2024)
8. JAMA: BedMed randomized trial of bedtime vs morning antihypertensive dosing (2025)
9. American College of Cardiology: TIME trial summary (2022)
10. University of Dundee: Chronotype and the timing of blood pressure medication
11. Wiki Journal Club: Hygia Chronotherapy Trial summary