STELLAR NCT02533271
ForDistal/middle-third rectal adeno, cT3-4 and/or cN+, age 18-70, ECOG 0-1
64.5% v 62.3%
HR 0.883, 1-sided 95% CI to 1.11, P<.001 for noninferiority
TL;DR3yr DFS 64.5% v 62.3% (HR 0.883, NI margin 1.43, P<.001 NI): 5x5Gy then CAPOX non-inferior to 50Gy/25f CRT in LARC.
Every pt got IMRT, unlike RAPIDO, Polish II or PRODIGE 23, and 3yr LRR was 8.4% v 11.0% with 25Gy/5fx to a full elective pelvic CTV, so the short-course arm did not trade local control for convenience. Compliance was the mechanism: 100% completed 5x5Gy with no dose reduction. This supports offering 5x5Gy over 50Gy/25f when the systemic phase is the priority.
In cT3-4 or node-positive mid/low rectal cancer where getting full-dose neoadjuvant chemo delivered is the constraint, this supports 25Gy/5fx followed by CAPOX instead of long-course chemoradiation; it does not extend to upper rectal tumors or pts over 70, both excluded.
All pts received IMRT with full elective pelvic CTV coverage, unlike RAPIDO, Polish II or PRODIGE 23, and 3yr LRR was 8.4% v 11.0% with 25Gy/5fx. Every TNT pt completed all five fractions with no dose reduction. This supports 5x5Gy over 50Gy/25f without conceding local control.
Moving CAPOX preoperatively raised full-dose preop completion of the systemic phase but cut it the other way on delivery overall, 74.8% v 93.2%. Four preop cycles produced no distant-metastasis benefit (3yr DM 22.8% v 24.7%), unlike the heavier RAPIDO and PRODIGE 23 programs, so cycle number may matter for distant control.
TME at 6-8 weeks after short-course RT plus chemotherapy gave R0 resection in 91.5% v 87.8% (P=.189) and grade III+ complications of 14.0% v 15.7% (P=.625), so the compressed schedule did not degrade the operation. Sustained cCR was higher after TNT (21.8% v 12.3% for pCR plus sustained cCR), enlarging the nonoperative-management pool to 9.4% v 3.4%.
12 details 5 trials watching
Multicenter open-label randomized phase III noninferiority trial, 16 hospitals across 11 provinces of China, accrual August 2015 to August 2018, N=599 randomly assigned 1:1 (TNT 302, CRT 297). Stratified by tumor location, clinical stage and MRF status; median follow-up 35.0 months (range 8.3-63.9).
Age 18-70, ECOG 0-1, rectal adenocarcinoma of the distal or middle third (0-10cm from anal verge), cT3-4 and/or node-positive, no distant metastases, no prior anticancer treatment. Arms were balanced: cT4 15.9% v 12.8%, MRF involvement 56.3% v 56.2%, clinical stage III 85.8% v 83.5%.
TNT arm received 25Gy in 5 fractions over one week; CRT arm 50Gy in 25 fractions over 5 weeks with concurrent capecitabine 825mg/m2 twice daily. All patients received IMRT, a departure from RAPIDO, Polish II and PRODIGE 23. CTV covered mesorectum, presacral space, internal iliac, obturator nodes and ischiorectal fossa; superior border at the sacral promontory, inferior 2-3cm distal to the tumor, external iliacs added only for cT4b; CTV-to-PTV expansion 0.5-1.0cm.
TNT arm: 4 cycles CAPOX (oxaliplatin 130mg/m2 day 1, capecitabine 1,000mg/m2 twice daily days 1-14) starting 7-14 days after radiotherapy, then TME, then 2 further cycles. CRT arm: concurrent capecitabine during radiotherapy, then TME, then 6 cycles CAPOX. Total mesorectal excision was scheduled 6-8 weeks after preoperative treatment in both arms; a watch-and-wait pathway was permitted for clinical complete responders.
Primary: 3-year disease-free survival, noninferiority claimed if the upper bound of the 95% CI of the HR was at or below 1.43 (an 11% absolute margin against an assumed 65% CRT rate). Secondary endpoints were overall survival, metastasis-free survival, locoregional recurrence and surgical complications. Target accrual 600 with at least 194 DFS events for 80% power at one-sided alpha 0.05.
Acute grade III-V toxicity during preoperative treatment was 26.5% with TNT versus 12.6% with CRT (P<.001), driven almost entirely by hematologic events (grade 3-4 15.8% v 2.0%, P<.001) rather than by anything the radiotherapy added. Radiotherapy delivery itself was cleaner in the short-course arm, with no dose reductions at all, and grade III+ surgical complications were comparable (14.0% v 15.7%, P=.625). Grade III-IV toxicity during adjuvant chemotherapy ran lower after TNT (3.3% v 11.8%, P=.003).
STELLAR is the third randomized trial of short-course radiotherapy plus neoadjuvant chemotherapy against long-course CRT, after Polish II and RAPIDO, with PRODIGE 23 addressing the adjacent question of chemotherapy sequencing around long-course CRT. Its 3-year LRR of 8.4% with TNT sits alongside RAPIDO and PRODIGE 23 (4%-8.3%) and well below Polish II (21%-22%), which enrolled a heavier fixed cT3/cT4 population. Unlike RAPIDO and PRODIGE 23, STELLAR showed no reduction in distant metastasis (3-year DM 22.8% v 24.7%), and its OS advantage mirrors the early Polish II signal that later disappeared at 8 years.
The 11% noninferiority margin is wide, and the authors concede in retrospect that a narrower margin or larger sample would have been defensible given how close the arms proved. Distant metastasis, the endpoint short-course TNT was designed to move, did not separate at all (22.8% v 24.7%), which weakens the mechanistic case for the OS difference. Adjuvant chemotherapy completion also differed between arms (60.0% v 48.3%, P=.009), so postoperative treatment intensity is not held constant across the comparison.
The defensible read is the primary endpoint: 25Gy in 5 fractions followed by CAPOX delivers the same 3-year DFS and the same locoregional control as 50Gy in 25 fractions, in one week of radiotherapy instead of five. The 11.4-point OS separation is the number that will be quoted and the one least supported, since neither MFS nor LRR moved and the same pattern in Polish II did not survive longer follow-up. What transfers most reliably is the technique: full elective pelvic coverage with IMRT at 5x5Gy, delivered without a single dose reduction.
| Endpoint (3yr) | TNT | CRT | Effect size |
|---|---|---|---|
| DFS (1°) | 64.5% (58.3-70.7) | 62.3% (56.1-68.5) | HR 0.883, 1-sided 95% CI to 1.11, P<.001 NI |
| OS | 86.5% (82.1-90.8) | 75.1% (69.4-80.8) | HR 0.67 (0.46-0.97), P=.033 |
| MFS | 77.1% (71.7-82.6) | 75.3% (70.0-80.7) | HR 0.88 (0.63-1.24), P=.475 |
| LRR | 8.4% (4.6-12.2) | 11.0% (6.5-15.5) | HR 0.80 (0.45-1.44), P=.461 |
| Subgroup | DFS HR (95% CI), P | OS HR (95% CI), P |
|---|---|---|
| cT4 | 0.621 (0.328 to 1.177), .144 | 0.362 (0.152 to 0.859), .021 |
| Distance to anal verge ≤5cm | 0.706 (0.485 to 1.028), .070 | 0.540 (0.318 to 0.916), .022 |
| cT2-3 | 0.916 (0.674 to 1.245), .575 | 0.752 (0.493 to 1.149), .187 |
| Distance >5cm | 1.120 (0.744 to 1.687), .587 | 0.808 (0.468 to 1.394), .443 |
CONSORT flow
Prespecified noninferiority met on 3yr DFS with adequate accrual, but open-label, 35mo follow-up only, and the OS gain is a secondary endpoint unsupported by MFS or LRR.
- Does the 3yr OS advantage survive 5-10 year follow-up active Short RT Versus RCT,Followed by Chemo.and Organ Preservation for Interm and High-risk Rectal Cancer Patients Phase 3n=702 · primary completion 2023-09 · ACO/ARO/AIO-18.1: 5x5Gy vs 54Gy CRT, n=702
- Optimal number of preoperative chemotherapy cycles with short-course RT n=42 · primary completion 2026-11 · 5x5Gy then 9 cycles mFOLFOX6, CR rate 1° EPactive Preoperative Sequential Short-course Radiation Therapy and FOLFOX for Locally Advanced Rectal Cancer Phase 2n=364 · primary completion 2028-12 · SCRT then 4 cycles FOLFOX vs CRT, n=364n=608 · primary completion 2029-12 · phase 3 SCRT+CAPOX vs SCRT+CAPOXIRI, n=608
- Late toxicity and quality of life, not yet reported recruiting Short Course Radiation Therapy and Combination Chemotherapy for the Treatment of Stage II-III Rectal Cancer Phase 1n=25 · primary completion 2026-10 · SCRT then chemo with QoL assessment as endpoint
📚 Sources · 📄 1 paper
Abstract
The longer read
STELLAR answers a narrower question than its OS curve suggests, and the narrow answer is the durable one. On the prespecified endpoint, 25Gy in 5 fractions followed by four cycles of CAPOX was noninferior to 50Gy in 25 fractions with concurrent capecitabine for 3-year DFS, with locoregional recurrence at 8.4% versus 11.0% and no separation in metastasis-free survival. For a radiation oncologist that is the transferable result: a week of radiotherapy bought the same disease control as five weeks, in a population that was overwhelmingly node-positive and more than half MRF-involved.
The overall survival difference deserves more skepticism than it will receive. It was a secondary endpoint, it was not accompanied by movement in either of the two failure endpoints that would have to change for survival to change, and the trial's own subgroup analysis found no consistent effect across prognostic strata. The comparison that should govern interpretation is Polish II, which reported the same early survival advantage for short-course TNT and saw it dissolve by 8 years. When an OS benefit appears without a corresponding shift in distant metastasis or local failure, the likeliest explanations are competing causes of death, imbalance in post-progression therapy, or chance, none of which generalize. STELLAR's 3-year distant metastasis rates were 22.8% and 24.7%, essentially identical, which is precisely the pattern that argues against a real anticancer mechanism for the survival gap.
That null on distant metastasis is also the most interesting divergence from RAPIDO, which reported a reduction in disease-related treatment failure driven by distant control, and from PRODIGE 23, which improved DFS with a substantially heavier neoadjuvant chemotherapy program. STELLAR used four cycles of CAPOX preoperatively where RAPIDO used six cycles of CAPOX or nine of FOLFOX and PRODIGE 23 used six of FOLFIRINOX. If the distant-control benefit of total neoadjuvant therapy is dose-dependent, STELLAR sits below the threshold, and its result is better understood as demonstrating equivalence of two radiotherapy schedules within a TNT framework than as validating TNT's systemic promise.
The toxicity comparison should be read carefully before it is used as an argument against short-course treatment. The 26.5% versus 12.6% grade III-V figure is almost entirely hematologic, 15.8% versus 2.0%, and it reflects full-dose oxaliplatin doublet delivered preoperatively rather than anything the radiotherapy contributed. The radiotherapy itself was the better-tolerated component: every patient in the short-course arm completed all five fractions without dose reduction, against dose reduction or interruption in a small fraction of the long-course arm. The honest framing is that this trial moved chemotherapy toxicity earlier, not that short-course radiotherapy is more toxic.
Two features limit how far the result travels. Every patient received IMRT with a defined elective pelvic CTV, which is not what RAPIDO or Polish II delivered, so the low locoregional recurrence rate should not be assumed for a 3D-conformal short-course program. And adjuvant chemotherapy completion differed markedly between arms, 60.0% versus 48.3%, so the comparison is not of two radiotherapy schedules alone but of two complete treatment programs with different realized systemic intensity. For a reader deciding between schedules in cT3-4 or node-positive mid to low rectal cancer, the case for 5x5Gy rests on equivalent local control and better delivery of the systemic phase, not on the survival curve.