Breast
Both reads are about what RT competes with or costs: tamoxifen as the alternative omission lever in good-risk DCIS, and reconstruction-modality interaction as a PMRT toxicity counterweight to the cardiac/pulmonary dosimetry case for protons.
Proton vs Photon PMRT Capsular Contracture
ForPost-mastectomy breast cancer, implant-based reconstruction (TE/I or DTI), receiving PMRT
TL;DR2yr CC 50% with proton+DTI vs 12% photon+TE/I; proton HR 2.3 univariate, 1.76 (0.93-3.32) multivariable, ns.
The actionable variable is reconstruction type, not beam: DTI carried HR 3.0 (1.7-5.5) for CC independent of modality, and proton+DTI was the worst cell at 50% at 2 years vs 12% for photon+TE/I. That reframes the pre-RT conversation with plastic surgery toward staged TE/I when protons are planned, rather than toward declining protons outright.
In a woman heading to PMRT after mastectomy with implant reconstruction, this informs the timing and type of reconstruction discussed with plastic surgery when proton is on the table; it says nothing about autologous reconstruction or prepectoral placement, neither of which was studied.
The proton CC signal did not survive adjustment (HR 1.76, 0.93-3.32, P=.083), so this is not on its own grounds to decline protons where cardiac sparing is the indication. It is grounds to know the reconstruction plan before simulation: proton with DTI hit 50% CC at 2 years, and every TE/I patient here was irradiated with the expander in place.
Reconstruction type outweighed beam modality: DTI carried HR 3.0 (1.7-5.5) for CC versus staged TE/I on multivariable analysis. When PMRT is planned, particularly proton PMRT, that argues for staging the reconstruction rather than direct-to-implant, in subpectoral placement at least, which is all this cohort covers.
9 details
IRB-approved retrospective cohort at 2 centers within one institution, PMRT delivered 2017-2023. N=175 (89 PBS proton, 86 IMRT photon). CC estimated by Kaplan-Meier, with Cox proportional hazards for covariates and a binary logistic model as verification.
Breast cancer pts who underwent subpectoral 2-stage TE/I or DTI reconstruction and then PMRT. Median age 49 (range 24-78), 63% Hispanic. Groups were balanced except on tumor laterality (P < .001) and reconstruction type (P < .001), the two axes the analysis turns on.
Pencil beam scanning proton PMRT vs IMRT photon PMRT. All TE/I pts had the tissue expander in place and irradiated, so this cohort speaks to expander-in-situ RT, not to post-exchange irradiation of a permanent implant. Dose and fractionation are not reported in source.
Proton was associated with CC on univariate analysis (HR 2.3, 1.26-4.30, P=.007) but the association did not hold after adjustment (HR 1.76, 0.93-3.32, P=.083). DTI vs TE/I carried HR 3.0 (1.7-5.5), P < .001 in the multivariable model. No other factor was significantly associated with CC.
| Group | n | 2yr CC rate |
|---|---|---|
| Proton + DTI | 36 | 50% |
| Photon + DTI | 15 | 35% |
| Proton + TE/I | 53 | 23% |
| Photon + TE/I | 71 | 12% |
Modality was assigned by practice pattern, not randomized, so the residual proton association could be confounding the model did not capture. CC is clinician-graded on unblinded chart review, and the DTI cells are thin (36 proton, 15 photon), which is where the widest rate gap sits.
The paper set out to test a prespecified suspicion that protons increase CC and returned a trend that did not clear significance once reconstruction type entered the model. What it does establish is the interaction cell worth counseling on: proton + DTI at 50% CC at 2 years.
Retrospective, non-randomized modality assignment with baseline imbalance in reconstruction type and laterality; the proton signal loses significance once adjusted.
- Does prepectoral placement change the proton CC signal?
- Expander-in-situ vs post-exchange RT sequencing for implant reconstruction
- Proton PMRT reconstruction toxicity in a prospective randomized comparison
📚 Sources · 📄 2 papers
Abstract
NRG/RTOG 9804 + E5194 combined analysis
ForGood-risk DCIS post-lumpectomy, low/int grade, ≤2.5 cm, margins ≥3 mm, no RT
TL;DR15yr IBR 11.4% vs 19.0% with tamoxifen after lumpectomy alone in good-risk DCIS; MVA HR 0.54 for any IBR.
For the RT-omission conversation this is the other half of the ledger: in the same good-risk cohort 9804 randomized to RT, endocrine therapy alone carried 15-yr IBR 11.4% vs 19.0%, and the benefit sits entirely on invasive IBR (HR 0.43) rather than DCIS-IBR (P=.089). No RT-vs-tamoxifen comparison is reported in source, so this sizes the alternative, it does not substitute for it.
For the patient with low/intermediate-grade DCIS ≤2.5 cm and ≥3 mm margins who has already declined radiotherapy, this quantifies what endocrine therapy adds over surveillance alone at 15 years; it does not inform high-grade, larger, or close-margin DCIS, nor patients receiving RT.
This sizes the non-RT alternative in the very cohort 9804 randomized to RT or observation: 15-yr IBR 11.4% vs 19.0%, with the effect concentrated on invasive IBR (HR 0.43) and absent for DCIS-IBR (P=.089). No RT-versus-tamoxifen comparison exists in source, so it informs the omission discussion without answering it.
Endocrine therapy in good-risk DCIS buys a 46% relative reduction in any IBR (HR 0.54) and 57% in invasive IBR at a median 14.85 years, but the source gives no ER status, duration or adherence data, so the number cannot be gated to a receptor-defined subgroup. Uptake was only 43.1%, itself non-random.
Margin and excision quality track with the exposure rather than being controlled for it: tamoxifen users more often had a negative re-excision (27.2% vs 10.6%) and size ≤5 mm (57.0% vs 41.3%). For a surgeon this reinforces that the ≥3 mm margin plus small low/intermediate-grade lesion is the substrate on which these 15-year numbers rest.
9 details 3 trials watching
Ancillary exploratory combined analysis of two cooperative-group datasets, not a new randomization. Tamoxifen use was optional in both parent trials, so the tamoxifen comparison is observational; Fine-Gray univariate and multivariable models were used for the competing-risk endpoints.
N=878: the non-RT arm of NRG/RTOG 9804 (n=317) plus the good-risk cohort of E5194 (n=561). Good-risk was defined identically across both: low- or intermediate-grade DCIS, ≤2.5 cm, margins ≥3 mm. Median age 59 (28-88).
Lumpectomy alone without radiotherapy, with or without tamoxifen by patient/physician choice. Overall uptake 43.1%, but lopsided by trial: 65.6% in NRG/RTOG 9804 versus 30.3% in E5194.
No radiotherapy in any analyzed patient by design: the 9804 contribution is its observation arm and E5194 was a non-RT cohort. The result therefore describes the population in which a reader has already elected RT omission.
IBR, invasive IBR, DCIS-IBR, contralateral breast event and overall survival, compared between tamoxifen groups. Median follow-up 14.85 years overall (13.87 in 9804, 16.15 in E5194); 15.92 years among those still alive.
117 IBR events (65 invasive, 52 DCIS). 15-yr IBR 11.4% (7.9-15.5) with tamoxifen vs 19.0% (15.3-22.9) without, P=.001. On multivariable analysis HR 0.54 (0.35-0.83) for any IBR and HR 0.43 (0.24-0.77) for invasive IBR; DCIS-IBR was not significantly reduced (P=.089).
Beyond the non-randomized exposure, the tamoxifen groups differ on the axes that predict recurrence: trial of origin (55.0% vs 21.8% from 9804), negative re-excision (27.2% vs 10.6%) and size ≤5 mm (57.0% vs 41.3%). Duration of tamoxifen, adherence and receptor status are not given in the source, so a dose- or ER-defined read is unavailable.
The split between a significant invasive-IBR reduction and a null DCIS-IBR effect is the part worth carrying: it argues the drug is acting on the events that carry downstream consequence rather than uniformly suppressing recurrence. Whether that separation is biology or a power artifact of 52 DCIS events is not settled here.
NRG/RTOG 9804 itself established that RT reduces IBR in this same good-risk cohort, so the field now has magnitudes for both omission levers in one population. The source reports no direct RT-versus-tamoxifen comparison, and the pooled cohort cannot supply one.
Tamoxifen was optional and unrandomized in both parent trials; users differed on trial of origin, re-excision status and pathologic size, so confounding by indication is unadjustable away.
- Whether the invasive-only benefit reflects biology or DCIS-event power
- Optimal tamoxifen duration in RT-omitted good-risk DCIS active Trial of Low Dose Tamoxifen in Women With Breast Intraepithelial Neoplasia - Long Term Follow-up Phase 3n=500 · primary completion 2022-12 · tamoxifen 5mg/d vs placebo in ER+ DIN, long-term
- Whether ER status selects who benefits in DCIS recruiting DCIS: RECAST Trial Ductal Carcinoma In Situ: Re-Evaluating Conditions for Active Surveillance Suitability as Treatment Phase 2n=400 · primary completion 2028-11 · randomised endocrine agents in HR+ DCIS onlynot yet Avoiding Surgery in Estrogen Receptor Positive Atypical Ductal Hyperplasia and In-situ Carcinoma Treated With Endocrine Treatment Trial Phase NAn=340 · primary completion 2032-12 · ER+ DCIS/ADH, 5y invasive IBC on endocrine alone