OCEANUS
ForAdvanced or refractory NSCLC receiving both RT and an ICI
20.3 vs 16.0 mo
aHR 0.68, 95% CI 0.47-0.99, P=.045 (sequential vs concurrent)
TL;DRSequential iRT beat concurrent for OS in newly diagnosed advanced NSCLC (20.3 vs 16.0 mo, aHR 0.68, P=.045); territory-wide cohort.
For an RT reader the actionable variable is timing, and the only signal here favors giving RT sequentially rather than concurrently with ICI (20.3 vs 16.0 mo, aHR 0.68). But the source reports no dose, fractionation, target volume or pneumonitis rate, so the parameter that would let you transfer this to a plan is absent.
In newly diagnosed advanced NSCLC already going on an ICI who also need thoracic or palliative RT, this weakly supports separating RT from the ICI start rather than overlapping them; it says nothing about stage III unresectable chemoRT-plus-durvalumab, where the concurrent-then-consolidation standard is randomized.
The only actionable variable is timing: sequential rather than concurrent RT with ICI carried longer OS (20.3 vs 16.0 mo, aHR 0.68). Dose, fractionation, target volume and pneumonitis rates are absent from the source, so the parameters that would let you build a plan around this are missing.
Chemotherapy alongside iRT was associated with longer OS in newly diagnosed advanced disease but not in refractory disease, and ICI maintenance after RT in refractory pts was not significant (11.2 vs 6.7 mo, P=.20). This argues against dropping the chemo backbone in the newly diagnosed setting when RT is added.
13 details 3 trials watching
Territory-wide retrospective cohort (OCEANUS) using the Hong Kong Hospital Authority CDARS, covering more than 90% of the population. Propensity score overlap weighting was the primary method with IPTW for sensitivity; analysis ran December 2024 to April 2025. Landmark analysis was applied, and refractory pts had to survive at least 90 days.
NSCLC diagnosed January 1, 2010 to December 31, 2021 who subsequently received iRT for advanced or refractory disease. Of 3522 pts who received ICIs, 335 received RT: 155 newly diagnosed advanced and 180 refractory. 247 (73.7%) male, median age 64 (range 34-90).
The exposure is RT timing relative to ICI, sequential vs concurrent, in newly diagnosed disease, and RT with vs without ICI maintenance in refractory disease. Dose, fractionation, modality, target volume and irradiated site are not reported in the source, which is the gap that limits transfer to a specific plan.
Primary: real-world OS after landmark, estimated with weighted Kaplan-Meier and Cox models. Where proportional hazards were violated per Schoenfeld residuals, effects were summarized with restricted mean survival time.
Sequential iRT carried longer OS than concurrent in newly diagnosed advanced disease, aHR 0.68 (0.47-0.99), P=.045. In refractory disease the ICI-maintenance difference was not significant (P=.20), and added chemotherapy showed no significant OS association there.
PACIFIC established consolidation durvalumab after concurrent chemoRT in stage III unresectable disease, and PEMBRO-RT and MDACC randomized data tested RT added to pembrolizumab in metastatic disease, but none of these randomize sequential against concurrent iRT in advanced NSCLC. That question has no randomized answer, which is why a 155-pt weighted cohort is currently among the larger reads on it.
Timing was clinician-assigned, so pts pushed to concurrent RT plausibly had more urgent, symptomatic or bulky disease, an indication bias that overlap weighting on recorded covariates cannot remove. The 2010-2021 window also mixes ICI eras, agents and lines, and the source reports no toxicity, so the pneumonitis risk that motivates avoiding concurrency is unmeasured here.
The result argues that separating RT from ICI administration does not cost survival and may favor it, which is the opposite of the abscopal-synergy rationale often used to justify concurrency. It does not establish causality, define an interval, or identify which pts the timing matters for.
Retrospective propensity-weighted registry cohort, 155 pts in the primary comparison, P=.045 with CI touching 1.0. Authors themselves call it hypothesis generating.
- Optimal interval between RT and ICI administration active Concurrent or Sequential Immunotherapy and Radiation Therapy in Patients With Metastatic Lung Cancer Phase 1n=78 · primary completion 2026-12 · randomises seq vs concurrent SBRT + nivo/ipi in stage IV
- Whether concurrent iRT increases pneumonitis in advanced NSCLC n=150 · primary completion 2027-02 · biomarker cohort tracking pneumonitis after CRT then ICI
- Which pts, if any, benefit from concurrent rather than sequential timing active Concurrent or Sequential Immunotherapy and Radiation Therapy in Patients With Metastatic Lung Cancer Phase 1n=78 · primary completion 2026-12 · same SBRT dose levels in each timing arm, stage IV NSCLC
📚 Sources · 📄 1 paper
Abstract
The longer read
The question OCEANUS takes on is one of the few in thoracic oncology where practice is genuinely unanchored. Radiation and checkpoint blockade are both routine in advanced NSCLC, they are frequently given to the same patient, and there is no randomized evidence telling anyone whether to overlap them. The mechanistic argument for concurrency, that irradiated tumor releases antigen into an ICI-primed system, has been attractive enough to shape practice patterns despite never having been tested against the alternative in a powered trial. So a weighted cohort of 155 newly diagnosed patients is not competing against better evidence; it is arriving where there is close to none.
What it finds runs against the synergy story. Sequential delivery was associated with longer survival, 20.3 versus 16.0 months, adjusted HR 0.68 with the upper confidence bound at 0.99 and P=.045. A reader should treat that estimate as directional at best. An interval that ends at 0.99 in a subset of this size is one reclassified patient away from crossing, and the point estimate itself was produced by a weighting model whose covariates are whatever a territory-wide administrative dataset happened to record. The refractory comparison, where ICI maintenance after RT gave 11.2 versus 6.7 months at P=.20, is the more honest illustration of what this data can support: a difference of that visual magnitude failing to separate tells you how thin the strata are.
The confounding that matters most is not one weighting can fix. Sequencing was a clinical decision, and the clinical decision to irradiate a patient at the same time as starting immunotherapy usually reflects urgency, a symptomatic lesion, cord compression, hemoptysis, bulk that will not wait. That is a worse-prognosis population selected by the exposure itself, and it produces exactly the direction of effect the paper reports. Propensity methods adjust on recorded covariates; the urgency of a treating oncologist's judgment is not a recorded covariate. Until a trial randomizes timing, the sequential advantage and the indication bias remain observationally indistinguishable.
For a radiation oncologist there is a second and more practical limit. The paper's exposure is timing, but the source gives no dose, no fractionation, no target volume, no irradiated site, and no toxicity data. The result therefore cannot be carried into a plan. It also cannot speak to the mechanism most clinicians actually worry about when contemplating concurrency, which is pneumonitis. If concurrent thoracic RT with an ICI drives more lung toxicity, that would be a plausible route from timing to survival, and it is precisely the analysis this source does not contain. Absent it, the survival difference has no proposed pathway.
What the study should change is small and specific. It weakens the case for deliberately overlapping RT with ICI in advanced NSCLC on synergy grounds, and it strengthens the argument that a prospective randomized trial of timing is worth running. It does not touch stage III unresectable disease, where PACIFIC established concurrent chemoradiation followed by consolidation durvalumab through randomization and where this cohort's population, endpoints, and intent do not apply.