NRG-HN009 (p16-negative phase II toxicity)
Forp16-negative locoregionally advanced SCCHN, 70 Gy RT + concurrent cisplatin
T-score ratio 1.11
90% upper confidence bound 1.33, p=0.77 · phase III criteria not met
TL;DRT-score ratio (weekly/Q3) 1.11, p=0.77: weekly cisplatin not less toxic with 70 Gy in p16-negative locoregionally advanced SCCHN, no phase III.
6-mo LRF was 4.6% (Q3) vs 9.6% (weekly), absolute difference 5.0% (95% CI 1.0, 8.9), despite worse RT QA on Q3 (81% vs 94% per protocol or acceptable). RT quality does not explain the gap. With 70 Gy fixed, schedule choice rests on toxicity type, not total G3-4 burden.
In p16-negative locoregionally advanced SCCHN receiving 70 Gy with concurrent cisplatin, this questions choosing weekly 40 mg/m2 on tolerability grounds, since G3-4 event counts were not lower; it does not extend to the p16-positive cohort, reported separately.
6-mo LRF was 4.6% (Q3) vs 9.6% (weekly), absolute difference 5.0% (95% CI 1.0, 8.9), despite worse RT QA on Q3 (81% vs 94% per protocol or acceptable). RT quality does not explain the gap. With 70 Gy fixed, the concurrent schedule is the variable to weigh for locoregional control.
Weekly 40 mg/m2 traded toxicities instead of lowering them: G3-4 AKI 11% vs 3% and nausea 12% vs 4% on Q3, WBC decrease 28% vs 14% and hypomagnesemia 8% vs 2% on weekly. Chemotherapy delivery was per protocol or acceptable in 80% on Q3 vs 98% on weekly; cumulative dose not reported.
12 details 1 trial watching
Randomized phase II/III; this is the phase II acute toxicity report for the p16-negative cohort. 234 randomized (118 Q3, 116 Q1) between Oct 2021 and Sept 2025; 91% completed all follow-up up to 6 months. QA modality reviews were done on a random sample of cases.
p16-negative locoregionally advanced SCCHN. Median age 60; 76% male; 58% Zubrod 0; 87% former or current smokers, 70% >10 pack-years. 57% larynx, 28% oropharynx, 15% hypopharynx; 83% T1-3 and 74% node-positive.
Cisplatin 100 mg/m2 Q3 weeks (Q3) vs cisplatin 40 mg/m2 weekly (Q1), each with RT. Chemotherapy delivery scored per protocol or acceptable variation in 80% on Q3 vs 98% on Q1. Cumulative cisplatin dose delivered is not reported in source.
70 Gy in both arms; fractionation, technique and target volumes are not reported in source. RT reviews were completed for 91% of sampled pts, scored per protocol or acceptable variation in 81% on Q3 vs 94% on Q1.
acute toxicity by T-score, the number of all treatment-related grade 3-4 AEs during treatment or within six months. To proceed to phase III, the mean T-score ratio (Q1/Q3) had to be below 1 at 1-sided p<0.1, together with a 6-month locoregional failure difference criterion (printed as '= 8%' absolute in the portal text).
The cohort did not meet criteria for continuing to phase III; arm-level values are in the tables.
| Endpoint | Q1 Cis (weekly) | Q3 Cis | Effect |
|---|---|---|---|
| Mean T-score | 2.48 (95% CI 1.97, 2.99) | 2.23 (95% CI 1.81, 2.65) | Ratio 1.11 (90% upper bound 1.33), p=0.77 |
| 6-month locoregional failure | 9.6% | 4.6% | Absolute difference 5.0% (95% CI 1.0, 8.9) |
| AE term | Q3 Cis | Q1 Cis |
|---|---|---|
| Nausea | 12 | 4 |
| WBC decreased | 14 | 28 |
| Dehydration | 9 | 1 |
| Hypomagnesemia | 2 | 8 |
| Acute kidney injury | 11 | 3 |
Total G3-4 burden was similar, but the profile differed by schedule. Q3 carried more G3-4 nausea (12% vs 4%), dehydration (9% vs 1%) and acute kidney injury (11% vs 3%); weekly carried more WBC decrease (28% vs 14%) and hypomagnesemia (8% vs 2%). Late toxicity is not reported in source.
The Tata Memorial randomized trial (Noronha et al.) favored Q3 over a lower weekly dose for locoregional control in a mostly postoperative population, while JCOG1008 supported weekly dosing after surgery. This cohort tests the schedule question with 70 Gy RT in p16-negative disease and gives no toxicity support for weekly; the source draws no comparison to either trial.
The T-score weighs every G3-4 event equally, so marrow and renal events are interchangeable in the primary. QA was scored on a random sample, not all cases, and per-arm cumulative cisplatin dose is missing. The portal text has lost the inequality in the LRF criterion, so the abstract alone does not show whether that half of the rule passed.
The tolerability premise for weekly cisplatin did not hold here: toxicity was redistributed, not reduced. The LRF difference favors Q3 but rests on a 6-month read in a toxicity-driven phase II, so it is a direction, not an efficacy result. Pooled results for all 475 pts were deferred to the presentation.
CONSORT flow
Randomized, prespecified phase II go/no-go analysis; weekly failed its toxicity criterion, leaving Q3 as the reference schedule. Abstract only; 6-month LRF is immature and not an efficacy test.
- Does the 6-month LRF gap persist in longer-term locoregional control and survival?
- Do pooled p16+ and p16- results (475 pts) show the same toxicity pattern?
- Late toxicity and hearing outcomes by cisplatin schedule n=100 · primary completion 2028-09 · randomized weekly LD vs HD cisplatin CRT, p16-stratified
📚 Sources · 📄 1 paper
Abstract
The longer read
Weekly cisplatin spread through head and neck practice on one assumption: that it buys tolerability at little cost in disease control. This phase II was built to test the first half of that assumption with a count of high-grade events, and the first half failed. The ratio did not merely miss its threshold, it pointed the other way (1.11, with a 90% upper bound of 1.33). What the arm-level rates describe is a trade, not a saving. High-dose Q3 dosing costs nausea, dehydration and acute kidney injury, weekly dosing costs white cell count and magnesium.
That trade is where the T-score stops being informative. It treats every grade 3-4 event as one unit, so a transient WBC decrease and an episode of acute kidney injury carry identical weight. A reader who regards renal injury as the more consequential event in a pt who may need salvage systemic therapy later will read the same data as mildly favoring weekly on safety, and a reader worried about marrow suppression interrupting RT will read it the other way. The trial answers whether weekly lowers the event count. It does not answer which schedule is easier on the pt, and the abstract reports no late toxicity to help.
The locoregional failure figures deserve more caution than their confidence interval invites. Six months mostly captures persistent disease and very early progression, the phase II was sized around toxicity, and event counts are not given. The portal text has also lost the inequality sign in the go criterion, so the abstract cannot show whether the control half of the rule was satisfied on its own. Still, the direction is worth noting because of what sits beside it: the Q3 arm had the lower per protocol rates for both RT (81% vs 94%) and chemotherapy (80% vs 98%) on sampled review. A quality imbalance of that kind would be expected to work against Q3, and Q3 still had less failure. The plausible reading of the chemotherapy deviations is dose modification after toxicity, but that is inference, and the missing variable is the cumulative cisplatin dose each arm received.
Against prior randomized data, this fits the older concern more than it overturns anything. The Tata Memorial trial favored Q3 for locoregional control against a lower weekly dose in a largely postoperative population, and JCOG1008 found weekly dosing acceptable after surgery. Neither addressed intact p16-negative disease treated to full dose, which is the population here, and the larynx-dominant, heavy-smoking case mix is the one where locoregional failure drives outcome.
What changes is the justification, not necessarily the prescription. Selecting weekly because it is assumed gentler overall is no longer supported in this population. Selecting it for a specific pt with limited renal reserve remains coherent with the AE pattern. The read would weaken if longer follow-up closes the failure gap, if delivered cisplatin dose turns out similar between arms, or if the pooled analysis of all 475 pts shows a different toxicity pattern. The p16-positive cohort was reported separately and its result is not in this source.