Jin 2022 — STELLAR Trial

Citation: Jin J, Tang Y, Hu C, et al. Multicenter, Randomized, Phase III Trial of Short-Term Radiotherapy Plus Chemotherapy Versus Long-Term Chemoradiotherapy in Locally Advanced Rectal Cancer (STELLAR). J Clin Oncol. 2022;40(15):1681-1692. doi:10.1200/JCO.21.01667.

Study design

• Multicentre (16 hospitals, 11 provinces, China), randomised, open-label, phase III trial (NCT02533271), designed by the National Cancer Center/Cancer Hospital, CAMS and PUMC, Beijing. 599 patients randomised (302 TNT / 297 CRT), August 2015–August 2018.
• Population: cT3-4 and/or regional-lymph-node-positive, distal- or middle-third rectal cancer, age 18–70, ECOG 0-1.
• TNT arm: short-term radiotherapy (SCRT, 25 Gy/5 fractions over 1 week) → 4 cycles CAPOX → TME → 2 cycles adjuvant CAPOX.
• CRT arm: long-course chemoradiotherapy (50 Gy/25 fractions over 5 weeks with concurrent capecitabine) → TME → 6 cycles adjuvant CAPOX.
• Primary endpoint: 3-year DFS, noninferiority (prespecified margin HR<1.43).
• Median follow-up 35.0 months (range 8.3–63.9) — notably shorter than RAPIDO’s 4.6 years or PRODIGE 23’s 6.8 years.

Key findings

Outcome (3-year)	TNT	CRT	HR (95% CI)	p
Disease-free survival	64.5%	62.3%	0.883 (one-sided 95% CI, NA to 1.11)	<.001 for noninferiority
Overall survival	86.5%	75.1%	0.67 (0.46–0.97)	.033 (log-rank)
Metastasis-free survival	77.1%	75.3%	—	.475 (not significant)
Locoregional recurrence	8.4%	11.0%	—	.461 (not significant)
pCR + sustained clinical CR	21.8%	12.3%	—	.002

• Primary endpoint met: TNT was noninferior to CRT for 3-year DFS.
• Overall survival numerically favored TNT (86.5% vs 75.1%), but this was a secondary, non-powered endpoint — the authors explicitly caution this OS finding needs longer follow-up before being treated as confirmatory, unlike PRODIGE 23’s mature, purpose-built long-term OS analysis.
• No locoregional-recurrence trade-off — LRR was numerically lower with TNT (8.4% vs 11.0%), the opposite direction from RAPIDO’s finding, despite STELLAR also using an SCRT-based TNT regimen.
• Toxicity was clearly asymmetric: acute grade III–V toxicity roughly doubled with TNT (26.5% vs 12.6%, p<.001), driven mainly by hematologic toxicity (grade 3-4 leukopenia 5.7% vs 1.7%, thrombocytopenia 11.1% vs 0.7%). Full-dose completion rate was also lower in the TNT arm (74.8% vs 93.2%).
• Adjuvant chemotherapy completion (of those who received any): 60.0% (TNT, 2 planned cycles) vs 48.3% (CRT, 6 planned cycles) completed their full course — note the TNT arm’s adjuvant burden is much smaller to begin with, so this isn’t a like-for-like compliance comparison.

Limitations

• Follow-up (median 35 months) is considerably shorter than RAPIDO or PRODIGE 23 — the OS finding in particular should be treated as preliminary/hypothesis-generating pending longer follow-up, per the authors’ own discussion.
• Higher acute toxicity with TNT (roughly double the grade ≥3 rate) is a real practical consideration not fully captured by the efficacy headline numbers.
• Single-country (China) population — cross-population generalizability to Western TNT trial populations (RAPIDO’s Europe/USA cohort, PRODIGE 23’s France-only cohort) is untested.
• IMRT was used uniformly in this trial (unlike some comparator trials), which the authors note as a strength for locoregional control but a difference in radiotherapy technique from RAPIDO/PRODIGE 23 that complicates direct cross-trial comparison of local-recurrence rates.

Relevance

Closes part of the primary-literature gap flagged in Total Neoadjuvant Therapy (TNT) and Rectal Cancer’s Open items. STELLAR is the wiki’s third SCRT-or-chemotherapy-based TNT trial alongside Bahadoer 2021 - RAPIDO Trial and Conroy 2024 - PRODIGE 23 Long-Term Results, and its results complicate rather than confirm RAPIDO’s locoregional-recurrence trade-off finding: STELLAR also uses SCRT-based TNT but shows numerically better, not worse, local control versus CRT. This is exactly the kind of cross-source divergence this wiki is meant to flag explicitly rather than silently resolve — plausible explanations (uniform IMRT use, different population, shorter follow-up masking a later-emerging difference, or simple play of chance in two moderately-sized trials) are not adjudicated here and would need dedicated comparative literature to resolve.