onc brain

About · curated by Nick Boehling, MD · @nb2276

RT + Systemic Therapy: What to Continue vs Hold (Speers)

TL;DRASCO 2026 education session slides triaging concurrent systemic agents with breast/CW + RNI into continue, caution, and hold/sequence.

Trials discussed

HERANCCTG N9831APHINITYKATHERINEATEMPTDESTINY-Breast05COMBARTKEYNOTE-522

Why it mattersRadiation oncology

The operational content is the PK column: talazoparib 5t½ 19 days and pembrolizumab 5t½ ~110 days mean a brief hold buys nothing, so the mitigation is field size, lung dose and monitoring rather than a washout. Field size, not agent class, gates the CDK4/6i and olaparib calls.

Monday clinic

In a pt starting breast/CW + RNI while on T-DXd, this supports concurrent treatment with lung-dose scrutiny rather than a hold; it does not extend to baseline ILD or active pulmonary disease, where sequencing is advised.

RT + Systemic Therapy: What to Continue vs Hold (Speers)
+3 more figures
RT + Systemic Therapy: What to Continue vs Hold (Speers)
T-DXd ILD ~12% at 5.4 mg/kg, fatal ~0.9%. DESTINY-Breast05 ILD 9.6% T-DXd vs 1.6% T-DM1. RT timing 10.7% seq vs 9.6% concurrent. COMBART 40 pts, acute tox 20%.
T-DXd ILD ~12% at 5.4 mg/kg, fatal ~0.9%. DESTINY-Breast05 ILD 9.6% T-DXd vs 1.6% T-DM1. RT timing 10.7% seq vs 9.6% concurrent. COMBART 40 pts, acute tox 20%.
KEYNOTE-522 post-hoc: 1,174 pts, 715 received RT. Pembrolizumab t½ 22 days, 5t½ ~110 days. P-RAD preop RT 0 / 9 / 24 Gy.
KEYNOTE-522 post-hoc: 1,174 pts, 715 received RT. Pembrolizumab t½ 22 days, 5t½ ~110 days. P-RAD preop RT 0 / 9 / 24 Gy.
14 details 4 trials watching

ASCO 2026 education session slide set from Corey W. Speers, adapted from Wong, Speers, Schaverien, ASCO Educational Book 2026, Table 5. It is an allocation framework, not a trial: agents are sorted into continue, caution, and hold/sequence for concurrent use with breast/chest wall + RNI.

The RT context throughout is breast/CW + regional nodal irradiation. The modifiers that move an agent between buckets are plan-level, not drug-level: large lung volumes, IMN coverage, bolus, reconstruction, CNS SRS for T-DM1, high lung-dose plans and thoracic/lung RT for T-DXd, and field size for CDK4/6 inhibitors and olaparib.

Each agent is anchored to its half-life and five-half-life washout: olaparib 15 hr / 3 days, veliparib 5.5 hr / 1 day, talazoparib 90 hr / 19 days, capecitabine ~45 min / 4 hrs, palbociclib 28.8 hr / 6 days, abemaciclib-ribociclib 24-55 hr / 5-11 days, T-DM1 4 days / 20 days, T-DXd 6 days / 30 days, pembrolizumab 22 days / ~110 days. The stated default outside protocol is PK-based washout → RT → resume.

T-DXd's dominant toxicity is ILD: ~12% at 5.4 mg/kg with ~0.9% fatal, and 9.6% vs 1.6% for T-DM1 in DESTINY-Breast05; RT timing within the T-DXd arm showed 10.7% sequential vs 9.6% concurrent. Veliparib carries dose-limiting moist desquamation and fibrosis with concurrent RT (TBCRC 024). T-DM1 signals are dermatitis (possibly underreported), a small pneumonitis signal, and CNS radionecrosis with SRS.

The HER2 mAb call rests on trial precedent rather than new data: HERA started trastuzumab after chemotherapy and XRT, NCCTG N9831 gave XRT concurrently with trastuzumab, and APHINITY gave trastuzumab plus pertuzumab concurrently with RT. The immunotherapy call rests on KEYNOTE-522, whose V1 protocol required restarting pembrolizumab ≥2 wks post-RT and whose V2 amendment permitted concurrency; the post-hoc of 1,174 pts (715 irradiated) found numerically fewer G3-5 and immune AEs in the concurrent group.

pts receiving breast/chest wall plus regional nodal irradiation while on concurrent systemic therapy
Does not represent other disease sites, definitive thoracic RT, or the CNS SRS setting except as a named caution.

The evidence tiers behind the buckets are uneven and the slide says so: the KEYNOTE-522 concurrency read is post-hoc with no adjustment for why pts were irradiated concurrently versus sequentially, and the CDK4/6i call rests on limited prospective data, most evidence retrospective, with a single trial cited (NCT05996107). P-RAD's endpoint is a biomarker (T-cell infiltration, tertiary lymphoid structures), not a clinical outcome.

The organizing logic is that PK sets whether a hold is even available, and plan geometry sets whether it is needed. Where 5t½ is short (capecitabine 4 hrs, veliparib 1 day, olaparib 3 days) the framework holds because it costs almost nothing; where 5t½ is long (talazoparib 19 days, T-DXd 30 days, pembrolizumab ~110 days) it concedes that a hold is theatre and shifts to lung dose, field size and surveillance. The unresolved item it flags rather than settles is T-DM1 vs T-DXd, where the mAbs are described as settled and the ADCs are not.

AgentCallBasis given
Endocrine therapyContinueMinimal radiosensitization
Trastuzumab ± pertuzumabContinueGenerally safe; modern heart-sparing
T-DM1ContinueDermatitis + pneumonitis vigilance; caution with CNS SRS
PembrolizumabContinuePneumonitis vigilance + immune-toxicity workflows
T-DXdCautionSequence/hold for high lung-dose or active pulmonary disease
CDK4/6 inhibitorsCautionUsually hold for large fields; concurrent only in protocol
OlaparibCautionReasonable with limited fields; protocol settings only
Cytotoxic chemo (anthracycline/taxane/platinum)Hold / sequenceSequence, do not give concurrently
CapecitabineHold / sequenceAdjuvant paradigm non-concurrent
Veliparib / talazoparibHold / sequenceVeliparib + RT severe acute/late tox
mTOR / PI3K agentsHold / sequenceMucosal, skin, metabolic, inflammatory risk; ESMO-ESTRO advises caution
📚 Sources · 🐦 1 tweet

The longer read

The useful move in this slide set is that it does not treat concurrency as a single question with a single answer. It sorts agents by two independent axes and then lets the axes disagree. The first is pharmacokinetic: how long is five half-lives, and therefore is a hold something a clinic can actually execute between the decision and the first fraction. The second is geometric: how much lung, marrow, mucosa or brain the plan will irradiate. Where both axes point the same way the call is easy, and where they conflict the framework picks the geometric one. That is the part worth carrying, because it is the opposite of how concurrency is usually reasoned about in a busy clinic, where the drug's name does the deciding.

Pembrolizumab is the clearest case. At 22 days, five half-lives is roughly 110 days, so the only honest reading is that a pt starting RT on pembrolizumab is receiving concurrent treatment whether or not anyone writes a hold order. The slide states this plainly and moves the mitigation to volume, dose constraints and monitoring. The supporting data are weaker than the conclusion is confident, and it is worth being explicit about why: the KEYNOTE-522 concurrency comparison is post-hoc, drawn from 1,174 pts of whom 715 were irradiated, and the direction of the finding (numerically fewer G3-5 and immune AEs in the concurrent group) is exactly the direction confounding by indication would produce if sicker pts were the ones whose RT got deferred. The further observation that the highest AE rate sat in the pembro-alone no-RT arm is presented on the slide with a question mark attached, and the question mark belongs there. None of that undermines the practical conclusion, which follows from PK rather than from the post-hoc; it does mean the post-hoc should not be cited as evidence that concurrency is protective.

The PARPi and CDK4/6i sections carry the opposite structure: short or moderate half-lives make a hold cheap, so the framework holds by default and asks for evidence before conceding concurrency. Veliparib is the anchor for that caution, with dose-limiting moist desquamation and fibrosis in TBCRC 024, and it is doing a lot of work as the class exemplar even though talazoparib's 19-day washout and olaparib's 3-day washout put them in different practical positions. RadioPARP is cited as suggesting concurrent olaparib safety while the early-stage paradigm on OlympiA stays sequential; the framework resolves that tension by field size rather than by picking a winner, which is a defensible way to hold an unsettled question open. For CDK4/6 inhibitors the honest statement is on the slide: limited prospective data, most evidence retrospective, one trial cited. A reader should treat the hold-for-large-fields default as a marrow and GI exposure argument, not as an observed toxicity signal.

What the framework does not settle, and names as unsettled, is T-DM1 versus T-DXd. T-DM1 sits in the continue column on KATHERINE and ATEMPT precedent; T-DXd sits in caution on ILD, with the incidence figures the slide reports and a within-arm comparison of RT timing that showed essentially no separation. That last number is the one that will get quoted, and it deserves a caveat the slide does not supply: a timing comparison inside a single arm is not randomised, and an ILD rate near 10% in both groups is compatible with a real timing effect that the study was not built to detect. The reasonable read is that RT timing is not the dominant driver of T-DXd ILD, not that timing is irrelevant.