NRG/RTOG 9804 + E5194 combined analysis (tamoxifen after lumpectomy, DCIS)
ForGood-risk DCIS, lumpectomy without RT, low/int grade, ≤2.5 cm, margins ≥3 mm
TL;DR15-yr IBR 11.4% vs 19.0% with vs without tamoxifen in good-risk DCIS treated by lumpectomy without RT; invasive IBR HR 0.43.
The association was confined to invasive events (DCIS-IBR p=0.089), which made up 65 of 117 IBRs. For the RT-omission discussion, the 11.4% 15-yr IBR on tamoxifen has no RT arm in this analysis to benchmark against, and which good-risk subgroups drive it sits in Fig 3, with subgroup HRs not in source text.
In low/intermediate-grade DCIS ≤2.5 cm with margins ≥3 mm managed without RT after lumpectomy, this supports tamoxifen as an invasive-IBR risk-reduction option in the omission discussion; it does not extend to high-grade DCIS or pts receiving RT.
For RT omission in good-risk DCIS, tamoxifen alone still left 15-yr IBR at 11.4%, and this analysis has no RT arm to benchmark it against. The reduction tracked invasive IBR (HR 0.43), while DCIS-IBR did not separate (p=0.089).
In good-risk DCIS managed without RT, tamoxifen associated with lower invasive IBR (HR 0.43, 0.24-0.77) but not DCIS-IBR (p=0.089). Dose, duration and adherence are not reported, and contralateral breast event results, the other half of the endocrine rationale, are not in source text.
9 details 1 trial watching
Ancillary exploratory analysis combining the non-RT arm of NRG/RTOG 9804, which randomized good-risk DCIS to RT or no RT, with the good-risk E5194 cohort observed without RT. Tamoxifen was optional and not randomized in either trial. Univariate and multivariable Fine-Gray models; median follow-up 14.85 yr.
N=878 (317 RTOG 9804, 561 E5194); median age 59 (28-88). Good-risk: low- or intermediate-grade, ≤2.5 cm, excision margins ≥3 mm, lumpectomy without RT.
Tamoxifen, optional per trial; used by 43.1% overall (65.6% RTOG 9804 vs 30.3% E5194). Dose, duration and adherence not reported in source.
No RT in the analyzed population. The RTOG 9804 RT arm is not part of this analysis, so tamoxifen alone is not compared against RT here.
IBR (any), invasive IBR, DCIS-IBR, contralateral breast event and overall survival. CBE and OS results are not reported in the source text.
117 IBRs (65 invasive, 52 DCIS). The reduction tracked invasive IBR, not DCIS-IBR.
| Endpoint | Tamoxifen | No tamoxifen | Effect |
|---|---|---|---|
| Any IBR, 15-yr estimate | 11.4% (7.9-15.5) | 19.0% (15.3-22.9) | p=0.001 |
| Any IBR, multivariable | n/a | n/a | HR 0.54 (0.35-0.83), p=0.0045 |
| Invasive IBR | not reported | not reported | p=0.0048; MVA HR 0.43 (0.24-0.77), p=0.0042 |
| DCIS-IBR | not reported | not reported | p=0.089 |
| Characteristic | Tamoxifen | No tamoxifen | p |
|---|---|---|---|
| Enrolled on NRG/RTOG 9804 | 55.0% | 21.8% | <0.0001 |
| Non-white | 13.8% | 9.2% | 0.034 |
| Negative re-excision | 27.2% | 10.6% | <0.0001 |
| Pathologic size ≤5 mm | 57.0% | 41.3% | <0.0001 |
NSABP B-24 randomized tamoxifen after lumpectomy plus RT, and the UK/ANZ DCIS trial randomized tamoxifen and RT in a factorial design. Here tamoxifen is examined only in good-risk pts managed without RT, and without randomization.
Multivariable covariates are not listed in source, and trial membership also differs in follow-up (13.87 vs 16.15 yr). The DCIS-IBR result (p=0.089) rests on 52 events and does not exclude an effect.
Tamoxifen use tracked lower invasive recurrence in good-risk DCIS managed without RT, consistent in direction with randomized tamoxifen data. It does not settle whether tamoxifen alone approaches RT, or how much of the gap reflects the lower-risk profile of users.
Exploratory ancillary analysis; tamoxifen optional, not randomized, and imbalanced by trial, re-excision and size. Parent RCT tested RT, not tamoxifen, so residual confounding limits causal read.
- Does tamoxifen alone approximate RT for IBR control in good-risk DCIS? recruiting Omission of Adjuvant Radiotherapy in Low- and Intermediate-Risk Ductal Carcinoma In Situ After Breast-Conserving Surgery Phase 3n=636 · primary completion 2031-06 · ph3 NI RT omission in ER+ low/int-grade DCIS, tamoxifen
- Which good-risk DCIS subgroups gain enough to justify tamoxifen?
- Is the lack of DCIS-IBR reduction real or underpowered?
📚 Sources · 📄 1 paper
Abstract
The longer read
RTOG 9804 and E5194 are the two datasets that made RT omission a defensible conversation in good-risk DCIS: one randomized RT against observation, the other followed a comparable cohort without RT. Neither was built to ask what tamoxifen contributes once RT is dropped, because both left endocrine therapy to choice. This combined analysis turns that choice into an exposure, across 878 pts with a median follow-up of 14.85 years, and finds lower ipsilateral recurrence among users, with 15-year IBR of 11.4% versus 19.0% and an adjusted invasive IBR HR of 0.43.
Against the randomized tamoxifen evidence the direction is unsurprising. NSABP B-24 tested tamoxifen after lumpectomy plus RT, and the UK/ANZ DCIS trial examined tamoxifen and RT together in a factorial design. What this dataset adds is population and horizon rather than a new mechanism: a defined good-risk group (low or intermediate grade, size up to 2.5 cm, margins of at least 3 mm) managed without RT, followed long enough to capture late ipsilateral events. Agreement in direction with randomized data raises plausibility. It does not convert an observational contrast into a randomized estimate for this population.
The confidence a reader places in the magnitude should be tempered by who took the drug. Users were more often RTOG 9804 enrollees (55.0% vs 21.8%), more often had pathologic size ≤5 mm (57.0% vs 41.3%) and more often had a negative re-excision (27.2% vs 10.6%). Each of those plausibly lowers recurrence on its own. The Fine-Gray models adjust, but the source text does not list the covariates, and trial membership bundles differences a single term may not absorb, including follow-up length (13.87 years in RTOG 9804, 16.15 in E5194). Exposure is also coarse: dose, duration and adherence are not reported. Direction is more credible than size, and if residual confounding runs the expected way, the adjusted HRs sit on the favorable side of the true effect.
The split between recurrence types deserves attention. The association held for invasive IBR (p=0.0048) but not DCIS-IBR (p=0.089), with 65 and 52 events respectively. A non-significant result on 52 events is weak evidence of no effect, so the honest reading is that the analysis could not show a DCIS-IBR reduction, not that tamoxifen spares only invasive disease. If the invasive-only pattern is real, it is the more consequential direction for an omission discussion.
For a radiation oncologist, the decision this informs is how to frame RT omission. It quantifies what pts choosing tamoxifen without RT still carry, an 11.4% 15-year IBR estimate, but the analysis includes no RT arm, so it cannot say whether tamoxifen alone approaches RT or whether combining them would add further control. It also does not reach high-grade DCIS. The finding would be wrong in its magnitude if the lower-risk profile of users, not the drug, explains most of the gap; the subgroup forest plot (Fig 3, HRs not in source text) and the unreported contralateral and survival results are where that would show.