OCEANUS
ForAdvanced NSCLC receiving RT with ICI, newly diagnosed or ICI-refractory
TL;DRSequential RT+ICI mOS 20.3 vs 16.0 mo for concurrent, aHR 0.68 (0.47-0.99), in newly diagnosed advanced NSCLC; retrospective registry.
The RT-relevant gap is the exposure itself: the source text neither defines the sequential window nor reports dose, fractionation or target, so the timing signal cannot yet guide SBRT versus palliative-course scheduling around ICI. In refractory disease, continuing ICI after RT gave 11.2 vs 6.7 mo rwOS (P=.20), limited to 90-day survivors.
In newly diagnosed advanced NSCLC starting ICI with an RT indication, this adds observational support for sequencing RT rather than overlapping it; it does not extend to unresectable locally advanced disease, and in ICI-refractory disease the maintenance signal is non-significant.
Timing is the RT variable: sequencing RT around ICI rather than overlapping it tracked with longer rwOS in newly diagnosed advanced disease. Dose, fractionation, target and the sequential window are not reported in source text, so the signal cannot yet separate SBRT from palliative-course scheduling.
In ICI-refractory NSCLC, ICI maintenance after RT gave 11.2 vs 6.7 mo rwOS (P=.20), non-significant and limited to 90-day survivors. Chemotherapy tracked with longer rwOS only in newly diagnosed disease, magnitude not reported in source, so the systemic partner question stays open.
9 details 2 trials watching
Territory-wide retrospective cohort from the Hong Kong Hospital Authority CDARS database (more than 90% population coverage), NSCLC diagnosed 2010 to 2021. Overlap weighting was the primary propensity method, IPTW the sensitivity analysis; RMST was specified where proportional hazards failed.
Of 3522 ICI-treated pts, 335 received RT: 155 newly diagnosed advanced and 180 refractory NSCLC. 247 (73.7%) male, median age 64 (range 34-90).
RT dose, fractionation, modality, target site and the interval separating sequential from concurrent delivery are not reported in source text.
Primary: real-world OS after landmark, via weighted Kaplan-Meier and Cox models. Exposures: sequential vs concurrent RT+ICI (newly diagnosed), RT with vs without ICI maintenance (refractory), and receipt of chemotherapy.
Chemotherapy tracked with longer rwOS in newly diagnosed disease but not in refractory disease; no chemotherapy effect size in source text. IPTW estimates were similar to the primary analysis.
| Cohort | Exposure arm | Reference arm | Effect |
|---|---|---|---|
| Newly diagnosed advanced (n=155) | Sequential RT+ICI: 20.3 mo (13.3-NR) | Concurrent RT+ICI: 16.0 mo (8.3-30.0) | aHR 0.68 (0.47-0.99), P=.045 |
| Refractory (n=180) | RT + ICI maintenance: 11.2 mo (7.9-20.6) | RT without ICI maintenance: 6.7 mo (4.4-17.4) | P=.20 |
The source notes randomized data on RT+ICI sequencing are limited. PEMBRO-RT tested SBRT before pembrolizumab in advanced NSCLC, and PACIFIC placed ICI after concurrent chemoRT in unresectable locally advanced disease; neither randomized timing as this cohort compares it.
Undefined timing window and unreported RT parameters make the exposure heterogeneous, possibly mixing consolidative and palliative courses. The 2010 to 2021 window spans shifting first-line ICI practice, and any RMST estimates are not reported in source text.
A sequential advantage cuts against the common assumption that overlap maximises RT-ICI interaction, but urgent symptomatic RT given during ICI could explain it equally well. Whether continuing ICI after RT helps refractory disease remains open.
Retrospective propensity-weighted registry cohort; residual confounding by indication likely, upper CI bound 0.99 in n=155, and authors frame it as hypothesis generating.
- Does sequential vs concurrent RT+ICI timing change OS when randomized?
- Does continuing ICI after RT benefit refractory NSCLC? recruiting A Study Evaluating The Combination of Immunotherapy With Radiotherapy in Non-Small Cell Lung Cancer Phase 2n=39 · primary completion 2030-06 · post-IO progression, ICI + ablative hypofractionated RT
- Which RT dose and target drive any timing effect? n=366 · primary completion 2027-06 · abscopal control across RT fractionation arms + ICI
📚 Sources · 📄 1 paper
Abstract
The longer read
The question OCEANUS addresses is timing, and the answer it offers is the less intuitive one. Much of the enthusiasm for pairing RT with checkpoint blockade in metastatic NSCLC rests on the idea that overlap is where the interaction lives: radiation exposes antigen, the checkpoint inhibitor acts on the primed response, so giving them together should do more than giving them apart. A propensity-weighted adjusted HR of 0.68 favouring sequential delivery in newly diagnosed advanced disease runs the other way. That makes the result worth attention, and it also makes it the kind of result most exposed to selection.
The central methodological worry is confounding by indication, and it is not symmetrical. In routine practice, RT delivered during an active ICI course is often RT that could not wait: symptomatic bone or brain disease, cord compression, airway compromise, haemoptysis. Pts in that position carry a worse prognosis for reasons a registry covariate set rarely captures. RT delivered before or after an ICI course is more often elective. Overlap weighting balances what was measured; it cannot balance disease tempo or the clinical urgency that drove the scheduling. An upper confidence bound of 0.99 with P=.045 in 155 pts leaves very little margin, so a modest amount of unmeasured confounding in that direction would move the estimate across the null. The source also does not define sequential or concurrent, and does not report dose, fractionation or target, so the exposure being compared may blend consolidative stereotactic treatment with short palliative courses whose biology and intent differ.
The landmark approach is the right answer to immortal time bias, and the refractory analysis was restricted to pts surviving at least 90 days. That choice guards against the obvious artefact, where pts who live longer simply have more opportunity to receive maintenance, but it also defines a selected group that had already shown something about its disease course. Within it, 11.2 vs 6.7 mo with wide, overlapping intervals and P=.20 is not evidence that continuing ICI after RT helps; it is evidence the question is unanswered. The chemotherapy associations are harder still to read, since no magnitude is given and receipt of chemotherapy is itself a marker of fitness.
Against the randomized literature, OCEANUS sits in a gap rather than in conflict. PEMBRO-RT irradiated a lesion before starting pembrolizumab, a sequenced design, and PACIFIC placed ICI after concurrent chemoRT in unresectable locally advanced disease, a different setting. Neither randomized the timing variable this cohort compares. The practical read is that this cohort does not justify delaying urgently needed palliative RT to avoid ICI overlap, and it does not establish that overlap causes harm. It offers a hypothesis with a plausible biological counterargument and a plausible confounding explanation, which is exactly the situation a trial randomizing RT timing around ICI would resolve. High-coverage territory-wide data is a strength for representativeness within the Hong Kong system and a limit when extrapolating to populations whose ICI access evolved differently across 2010 to 2021.