Registered, Pre-Specified, and Still Retracted: What LungTIME-C01 Reveals About Trial Integrity

Randomised controlled trials are presented as the gold standard of medical research precisely because their methodology is designed to be auditable. Auditability, however, relies on the integrity of the records being audited. In February 2026, Nature Medicine published LungTIME-C01, a phase 3 trial reporting that early time-of-day immunochemotherapy reduced the risk of progression or death by about 60% in advanced non-small cell lung cancer [1]. As reported, the trial had prospective registration, a written protocol, a pre-specified primary endpoint, blinded central review and a CONSORT flow diagram [1]. In June 2026, it was retracted, with a note stating that the editors no longer had confidence in the integrity of the results [2]. This is a case study in how reproducibility infrastructure can be present, formally complete, and still fail.

Before discussing areas for improvement, it is worth crediting what the researchers did meet. The study was registered on ClinicalTrials.gov (NCT05549037) before enrolment, specified progression-free survival as the primary endpoint with a sample-size calculation, assessed the primary endpoint and objective response through a blinded independent review committee using RECIST 1.1, and tabulated adverse events by grade and arm [1]. On paper, the trial carried the full formal toolkit of reproducibility: registration, protocol, pre-specification, blinded assessment, transparent flow, and structured safety reporting. The retraction note nonetheless highlights four unmet requirements: two documentary, one statistical, and one procedural [2].

First, registration integrity. The editors reported that endpoints, eligibility criteria, sample size, and study design appeared to have been “inconsistently reported and modified over time” on the registry [2]. The authors largely attributed these changes to “administrative errors rather than prospectively documented amendments” [2]. The record could therefore no longer serve as evidence of what was pre-specified before the data were seen.

Second, protocol integrity. The English translation published with the manuscript was dated 2022 yet cited studies published in 2023 and 2024. Further discrepancies were identified between the original Chinese protocol and the English translations supplied to the journal [2]. A protocol that cannot be placed chronologically cannot anchor reproducibility.

Third, data-pattern anomalies. The progression-free survival curve was smoother than is typical of similar large phase 3 trials. There was no censoring in the first year, and no adverse event led to treatment discontinuation in either arm. Immune-related adverse event rates were similar in the two arms despite markedly different efficacy [2]. These are red flags rather than proof, since the note did not benchmark them against comparable trials, and one oncologist quoted in press coverage considered similar toxicity across arms biologically plausible. It is the combination of anomalies, together with the registration and protocol problems, that undermined confidence.

Fourth, procedural deviation. Source data showed that randomisation was performed on the day of treatment for almost all patients, which is uncommon [2] and appears difficult to reconcile with the independent, envelope-based allocation described in the Methods [1]. Day-of-treatment randomisation matters because it can allow treating clinicians to influence who enters the trial, which erodes allocation concealment even when envelopes are nominally used. The authors also acknowledged that scans were scheduled by treatment cycle rather than calendar date, deviating from RECIST timing recommendations [2]. Scan timing then depends on when each patient’s cycles fall and can differ between patients and arms in ways that affect when progression is detected. The blinded reviewers could only judge the images they were given. They could not correct a problem that arose earlier, in how and when those images were scheduled and collected.

LungTIME-C01 had a registration, a written protocol, and pre-specified analyses. What it lacked was a record that could be trusted to say what had been planned before the data were seen.

The lessons reach well beyond one trial. In biomedical research, results shape treatment regimens and the questions pursued in future research. When a finding cannot be reproduced, resources are wasted on misdirected research and patients may be harmed. Journals decide which results enter the scientific record, so they are well placed to close these gaps: by checking registration histories, requiring time-stamped protocols with certified translations, and running basic checks on patient-level data in high-impact trials. Single-centre studies should be treated as hypothesis-generating until independently verified, and this retraction does not settle whether treatment timing matters. By publishing replication and validation work and holding submissions to transparent standards, journals such as RevisX can make verification a routine part of scientific publishing.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. To view a copy of this licence, visit https://creativecommons.org/licenses/by/4.0/.