The research, led by investigators including Huiquan Liu and Dong Qian, enrolled 1,242 patients with inoperable oesophageal squamous cell carcinoma from eight large hospitals across China between June 2018 and December 2023. More than half of all new cases of this disease occur in China, where the squamous subtype accounts for roughly 85 percent of oesophageal cancers, and a considerable share of patients are diagnosed with disease that cannot be surgically removed. Because no randomised controlled trial has ever directly compared the two sequencing strategies, the team turned to an increasingly influential methodology: target trial emulation, a framework that allows researchers to extract causal insights from observational data by designing the analysis as if it were a randomised trial.

The emulation was meticulous. Time zero was defined as the start of initial treatment, and patients were conceptually randomised into two groups: an upfront radiotherapy group, in which radiation began within a 42-day grace period, and a deferred radiotherapy group, in which radiation was delivered at any clinically appropriate later time. To eliminate the biases that plague retrospective studies, particularly immortal time bias and selection bias, the researchers used the clone-censor-weight technique. Every patient was duplicated into two identical copies, one assigned to each strategy. When a patient’s actual treatment trajectory departed from the assigned strategy, that copy was artificially censored, and inverse probability of censoring weighting corrected for the resulting informative dropout. The weighted clones formed a pseudo-population in which the two strategies could be compared as though patients had truly been randomised.

The cohort was split into 400 patients in the upfront radiotherapy group and 842 in the deferred group. Baseline characteristics were balanced using stabilised inverse probability of treatment weighting, with standardised mean differences below 0.1 for every covariate, including age, sex, body mass index, tumour location, and clinical stage. Notably, the deferred group actually carried a heavier tumour burden at baseline, with more advanced stage disease, thicker tumours, and lower haemoglobin levels, making the eventual survival advantage all the more remarkable. Most patients received platinum-based doublet chemotherapy alongside PD-1 or PD-L1 inhibitors, with median radiation doses of 60.0 Gy in the upfront group and 50.4 Gy in the deferred group.

The survival results were unambiguous. After a median follow-up of 34.8 months, 51.8 percent of patients in the upfront radiotherapy group had died, compared with 40.3 percent of those in the deferred group. Median overall survival reached 44.1 months with deferred radiotherapy versus 33.4 months with upfront radiotherapy, an adjusted hazard ratio of 0.699. Progression-free survival told the same story: 21.1 months versus 16.0 months, with an adjusted hazard ratio of 0.760. Under the clone-censor-weight analysis, the three-year overall survival rate was 57.5 percent with the deferred strategy against 41.9 percent with upfront radiation, an absolute difference of 15.6 percentage points. Three-year progression-free survival was 33.8 percent versus 22.4 percent.

The robustness of these findings was tested through nine separate sensitivity analyses, and the conclusion held every time. The grace period was varied to 21 and 63 days; overlap weighting was applied to focus on clinically comparable patients; hospital-level variability was modelled with mixed effects; missing data were multiply imputed; and analyses were restricted to the 1,225 patients who received PD-1 inhibitors. Landmark analyses restricted to patients alive and progression-free at day 42, and a reclassification comparing patients who started with radiation against those who started with immunotherapy, all favoured the deferred approach. E-values of 2.460 for overall survival indicated that only very strong unmeasured confounding could plausibly explain away the association. Subgroup analyses across sex, age, residence, performance status, and tumour characteristics showed a consistent direction of benefit.

Safety data added a further dimension to the debate. Severe oesophageal fistula, a feared complication in which a hole forms in the irradiated oesophageal wall, occurred in 9.0 percent of the upfront radiotherapy group compared with 4.8 percent of the deferred group, and fatal fistulas claimed eight patients in the upfront arm versus seven in the deferred arm despite its smaller size. This pattern echoes prospective data: in the EC-CRT-001 trial, which combined chemoradiotherapy with concurrent immunotherapy, the fistula rate was 14.3 percent, whereas in EC-CRT-002, which used induction chemoimmunotherapy before deferred radiation, it was just 3.5 percent. The most likely explanation is that induction systemic therapy shrinks the tumour before radiation begins, reducing local invasion and making irradiation safer. Conversely, grade 4 leukopenia and neutropenia were somewhat more frequent in the deferred group, possibly reflecting depleted bone marrow reserve after prolonged systemic therapy.

The biological rationale for these findings is as intriguing as the statistics. Radiation is widely recognised to remodel anti-tumour immunity, and hypofractionated regimens can synergise with immunotherapy by releasing tumour antigens and activating immune cells. But definitive oesophageal chemoradiotherapy cannot routinely use hypofractionation because of fistula risk, so patients endure weeks of conventional fractionation. Experimental work has shown that X-ray irradiation dilates lymphatic vessels, stagnates lymph flow, and damages high endothelial venules, the specialised structures that allow immune cells to enter lymph nodes. A secondary analysis of pooled randomised trial data found that higher radiation dose-volume parameters in lymphatic drainage regions were associated with impaired T-cell activation and worse prognosis. In other words, early radiation may inadvertently sabotage the very lymph node machinery that checkpoint inhibitors rely upon to prime anti-tumour T cells.

The authors are careful about what their study can and cannot prove. Because radiation dose, chemotherapy cycles, and immunotherapy cycles are consequences of the assigned strategy rather than baseline characteristics, they were not adjusted for as confounders. Patients in the upfront group received a higher median radiation dose, while those in the deferred group completed more cycles of chemoimmunotherapy, so the observed survival gap may reflect differences in overall treatment intensity rather than sequence alone. The findings therefore represent the marginal total effect of two real-world strategies, not definitive evidence that radiation timing independently drives survival. Other limitations include the retrospective design, the absence of time-varying covariates, non-protocolised tumour assessments, and incomplete records of radiation target volumes and chemotherapy regimens.

Even with those caveats, the implications for clinical practice are immediate and potentially practice-changing. For the many patients with inoperable oesophageal squamous cell carcinoma who face this decision each year, the evidence now suggests that starting with chemoimmunotherapy and deferring radiation may extend both life and time without progression, while reducing the risk of a devastating oesophageal fistula. The survival rates observed in this retrospective cohort align closely with those reported in prospective phase II trials, lending credibility to the results. Ongoing randomised trials will ultimately determine whether sequence, intensity, or their interaction explains the difference, but for now, this study of more than 1,200 patients provides the strongest real-world signal yet that in the immunotherapy era, patience with radiation may be a virtue worth cultivating.

Subject of Research: Sequencing of radiotherapy with first-line chemoimmunotherapy in inoperable oesophageal squamous cell carcinoma

Article Title: Upfront or deferred radiotherapy combined with chemoimmunotherapy for inoperable oesophageal squamous cell carcinoma: a target trial emulation in China

Article References: Liu, H., Guan, J., Zhang, W., Li, D., Fu, X., Li, L., Zang, C., He, Y., Tian, J., Tong, Z., Huang, W., & Qian, D. (2026). Upfront or deferred radiotherapy combined with chemoimmunotherapy for inoperable oesophageal squamous cell carcinoma: a target trial emulation in China. eClinicalMedicine, 100, Article 104236. https://doi.org/10.1016/j.eclinm.2026.104236

Image Credits: AI Generated

DOI: 10.1016/j.eclinm.2026.104236

Keywords: oesophageal cancer, squamous cell carcinoma, radiotherapy, chemoimmunotherapy, immune checkpoint inhibitors, target trial emulation, clone-censor-weight, overall survival, oesophageal fistula, China, eClinicalMedicine, treatment sequencing