PART I — DEFINITION, RATIONALE, AND HOW ONLINE ART WAS BUILT
Quick Definition: What Is Online ART?
Online adaptive radiotherapy is the process of adjusting the daily treatment plan in response to observed changes in target and/or organ-at-risk geometry while the patient remains on the treatment table. In practical terms, it condenses what would traditionally be 1-2 weeks of planning work into a single session lasting roughly 15-90 minutes.
Vocabulary That Gets Tested
Term
Meaning
Board-style distinction
Offline ART
Replanning between fractions after interval imaging or clinical change
Useful for slower changes such as tumor shrinkage, weight loss, or anatomic drift over weeks
Online ART
Same-day contour review, dose prediction, re-optimization, QA, and treatment while the patient is on table
Best for day-to-day OAR / target geometry changes
Scheduled ART
Planned adaptation at set time points
Common concept for predictable changes during longer courses
Triggered / ad hoc ART
Adapt only when anatomy or dose metrics cross a threshold
Most intuitive model for daily SBRT OAR rescue
Plan-of-the-day / library
Select a pre-existing plan that best matches daily anatomy
Adaptive, but not necessarily same-day re-optimization
Adapt-to-position
Update plan or plan choice mainly for daily target/OAR position
Less anatomy editing; closer to advanced image guidance
Adapt-to-shape
Re-contour or deform anatomy and re-optimize for changed shape / volume
More powerful, but more contouring and QA burden
Why ART Emerged from the Limitations of Standard IGRT
Static RT plans cannot account for substantial day-to-day anatomic variation.
Upper abdominal and pelvic sites are especially vulnerable because OAR position may change enough to invalidate an otherwise acceptable plan.
ART therefore aims to preserve target coverage while avoiding unplanned OAR overdosing.
Three Core Pieces Required to Make ART Possible
Requirement
Why it matters
Board-style takeaway
1. On-board imaging of adequate quality
Must visualize target and OAR changes reliably enough to justify re-planning
No meaningful adaptation without anatomy you can trust
2. Paired TPS for rapid re-contouring and re-optimization
Adaptation has to happen in real time while the patient remains on the table
Speed and contouring logic are part of the technology, not an afterthought
3. Online pre-treatment QA
Every adapted plan still needs to be safe to deliver
New workflows still require old-fashioned rigor
How Platforms Were Vetted: Stepwise Development Path
Step
MR-guided example
Key result
Imaging study
Original low-field 0.35T MRgRT platform
No imaging dose, 4 or 8 f/s, automatic gating, TruFISP-based imaging; MRI visualized soft tissue better than standard C-arm CBCT
In silico workflow / TPS testing
Simulated delivery of 50 Gy / 5 fx SMART
Large daily anatomic shifts observed; predicted 63% daily ART need to protect OARs; adaptation predicted to reverse 100% of those OAR violations
Pilot / Phase I trial
Adaptive upper-GI SMART pilot
Demonstrated feasibility and safety before broader efficacy testing
Board pearl:ART platforms should be developed in a repetitive, stepwise fashion: imaging validation, dosimetric / in silico testing, then feasibility and safety trials. That sequence is itself testable.
Current Clinical Platform Landscape
Platform type
Clinical use timeline
Practical note
MR-guided ART
In clinical use since 2014
Best soft-tissue visualization and integrated motion management
CT-guided ART
In clinical use since roughly 2018-2019
Typically higher throughput and easier integration into conventional IGRT-heavy workflows
C-arm linac adaptive
FDA 510(k) cleared in January 2026
Important because C-arm linacs are the most globally common treatment platform
Four Clinical Reasons to Adapt
Paradigm
Clinical problem
Canonical examples
Safe dose escalation
Ablative dose is desirable but bowel, stomach, duodenum, or liver constraints are fragile
Pancreas SMART, HCC / liver SBRT
Margin reduction
Standard PTV margins irradiate avoidable normal tissue
DARTBOARD head-and-neck, prostate SBRT
Extreme hypofractionation
Very high dose per fraction leaves little room for anatomy misses
SMART ONE, HERMES, ARCHER
Workflow compression
Simulation, planning, and treatment delays reduce access or palliation speed
FAST METS, ONE-STOP
Do-not-adapt pearl:online ART should be used when it changes a clinically meaningful decision. If the predicted delivered plan already meets target and OAR goals, and adaptation only adds time or uncertainty, the correct choice may be to treat with the scheduled plan.
PART II — CLINICAL PILLAR #1: SAFE DOSE ESCALATION
Phase I Adaptive SMART: Upper GI / LAPC Feasibility
Phase I adaptive upper-GI SMART pilot (example LAPC, N = 10) established that online adaptive SBRT was safe and feasible. The key result was that the initial plan violated OAR constraints in 70 of 95 fractions, and online ART resolved all of those violations, with 0 grade 3+ acute GI toxicities.
SMART Pancreas Phase II Trial
Feature
Details
Population
Borderline resectable / locally advanced pancreas
Platform / concept
MRI-guided online ART
Dose
50 Gy / 5 fx (BED10 = 100 Gy), subject to strict OAR constraints
Primary endpoint
CTCAE grade 3+ definitely related GI toxicity within 90 days
Primary result
Zero definitely treatment-related grade 3+ toxicities; primary endpoint met
2-year local control
Nearly 80% overall; about 70% in LAPC; >90% in BR disease that proceeded to resection
2-year OS context
BR SMART about 60.1%; LAPC SMART about 34.3%
Why SMART matters:it is one of the clearest examples that ART is not just about convenience; it is about making an otherwise aggressive, ablative prescription deliverable without unacceptable GI toxicity.
Current and Ongoing GI / Hepatobiliary Adaptive Trials
In the HCC HELIO-RT example, a day-1 plan met stomach and bowel constraints, but by day 2 the large bowel constraint of < 33 Gy to 0.5 cc was violated, with parts of bowel exceeding 40 Gy. Post-adaptation, bowel was protected while target coverage was maintained. This is the board-style illustration of why ART matters for mobile upper-abdominal anatomy.
1 mm PTV margin with daily ART instead of 5 mm standard IGRT margin
Dosimetric effect
Reduced parotid and submandibular gland dose, smaller targets
Primary endpoint
Xerostomia Questionnaire (XQ) at 1 year
Workflow timing
CBCT to end of RT: 33 min; door-in to door-out: 39.4 min; MD contour time: 12.6 min; MD at console: 22 min
DARTBOARD Toxicity and Oncologic Outcomes
Outcome
IGRT
DART
Comment
Dermatitis grade 2+
31%
8%
Statistically lower with adaptive treatment
Mucositis grade 2+
92%
75%
Trend favoring ART
Dysphagia grade 2+
81%
75%
No clear difference
Dysphagia grade 3+
19%
8%
Numerically lower with ART
Safety signal: median follow-up was about 18 months, there were no marginal recurrences, and the single local recurrence was in-field and salvaged. That is exactly the kind of disease-control reassurance required for margin-reduction strategies.
PART IV — CLINICAL PILLAR #3: EXTREME HYPOFRACTIONATION, SAFELY
The prospective MOMENTUM registry across 1.5T MRgART sites now contains over 8,000 patients, about 98,000 MRI scans, and 33,800 dose plans. One registry-generated hypothesis is that MRgART may lower toxicity in 5-fraction prostate regimens relative to historical standards, whereas that advantage is less obvious with conventional fractionation.
Intermediate-risk / lower high-risk prostate cancer
Adaptive strategy
Daily ART for all patients, 3 mm PTV for all
Dose
36.25 Gy / 5 fx
24 Gy / 2 fx
Boost
CTV to 40 Gy
GTV SIB to 27 Gy / 2 fx
Primary endpoint
Acute grade 2+ GU toxicity, benchmarked against a 62% historical rate (PACE-B based)
Board pearl: HERMES is not just a prostate trial; it is a proof-of-concept that ART may help push SBRT toward fewer fractions while maintaining acceptable toxicity. The larger SABR-DUAL phase II/III effort is the natural next step.
Can 5-fraction adaptive SBRT be non-inferior to 20-fraction hypofractionated RT for bladder-intact event-free survival?
ART requirement
Mandatory ART, CT- or MR-guided
Systemic therapy
Concurrent chemo allowed: weekly cisplatin, gemcitabine, or 5-FU + mitomycin-C
Primary endpoint
3-year BI-EFS with a 10% non-inferiority margin (HR < 1.32)
Sample size
N = 486
Example PTV margins
Adaptive arm approximately 7 mm vs conventional arm 1 to 1.5 cm
PART V — IMPLEMENTATION: HOW TO BUILD AN ART PROGRAM
ART Has Reached the Mainstream
Online ART is no longer a niche physics project. There are now published safety white paper guidelines addressing quality, personnel, and workflow, and ART has already entered phase III cooperative-group trials.
Daily Online ART Workflow and Safety Failure Modes
Step
What must happen
Common failure mode
Daily image
Acquire anatomy good enough to see the target and priority OARs
Poor image quality creates false confidence
Registration
Align daily anatomy to planning anatomy with the correct clinical priority
Registering to target may worsen an adjacent OAR problem
Contour propagation / edit
Review propagated target and OAR contours; edit high-priority structures
Contour error becomes dose-calculation error
Dose prediction
Calculate what the scheduled plan would deliver on today's anatomy
Adapting without first knowing whether adaptation is needed
Adapt decision
Compare scheduled plan vs adapted plan using pre-specified priorities
Chasing tiny dosimetric improvements at the cost of long couch time
Re-optimization
Generate the adapted plan while preserving target/OAR hierarchy
Target coverage recovered by silently exceeding a serial OAR constraint
Online QA
Perform independent calculation, delivery checks, and approval before treatment
Assuming fast planning eliminates conventional safety duties
Motion management
Confirm breath-hold, gating, tracking, or intrafraction monitoring strategy
Perfect pre-treatment adaptation followed by intrafraction miss
Documentation
Record contours, chosen plan, delivered dose, overrides, and responsible approvals
Cumulative dose and decision history become unreconstructable
Adapt trigger pearl: adapt for priority OAR violation, compromised target coverage, major interfraction geometry change, or a meaningful therapeutic-index gain. Do not adapt simply because an adaptive platform is available.
The Roadmap to Launch
Phase
Core tasks
Pre-work
Choose platform and supporting technologies; identify champions; align with department strategy
Team preparation
Train the whole team; build workflows and procedures by disease site / indication
Go-live and beyond
Peer support, quality assurance, continuing education, and program growth
Slower throughput and setup limitations for standard IGRT-style workflow
CT-guided ART
Built on high-throughput CBCT platforms; broader day-to-day clinical versatility; lower capital cost
Motion management is less intrinsically robust and may require paired technologies (ABC, surface guidance, etc.)
C-arm linac adaptive
Potentially brings adaptive workflows to the most common linac architecture
Clinical evidence and operational models are still emerging
Proton online ART
Highly relevant because proton range is sensitive to anatomy and density changes
Requires robust range-aware QA and remains early clinically
Functional MR / BgART
May adapt to biology or function rather than geometry alone
Promising but still investigational for routine board-style practice
Where to Start: Choose High-Yield Champion Sites
The implementation lesson is simple: volume does not appear from thin air; ART programs start where there is real clinical pull and committed local expertise.
Who Is a Good ART Candidate?
Favors ART
Argues against routine ART
Mobile upper-abdominal targets near stomach, bowel, duodenum, or liver reserve constraints
Anatomy is stable and the scheduled plan robustly meets goals every day
Pelvic targets with large bladder, rectal, or bowel variability
Target / OAR visibility is poor on the adaptive image set
Very tight margins where marginal miss would be unacceptable
Patient cannot tolerate longer couch time, breath-hold, gating, or immobilization
High dose per fraction or serial OAR adjacency
No staff coverage model for physician, physicist, therapist, and QA approvals
Sim-free / urgent palliative workflows where treatment access is the bottleneck
Adaptive plan would not meaningfully improve target coverage, OAR dose, or access
Operational Pieces That Need to Exist Before Go-Live
Coverage model: choose it deliberately and expect it to evolve.
Team-wide training: ASTRO task-force guidance recommends broad multidisciplinary training, not just physician and physics education.
Adaptive school curriculum: system overview, clinical rationale, planning templates, imaging, case-based ART, documentation, billing, and even retention quizzes.
Handoff design: communicate goals of ART, what to re-contour, what constraints matter most, key anatomy issues, setup needs, and whether the patient may need analgesics or anxiolytics.
Documentation: the ASTRO white paper includes practical templates.
Ongoing quality culture: regular peer review and continuing education remain essential after launch.
PART VI — FUTURE DIRECTIONS: SIM-FREE WORKFLOWS AND NEW PLATFORMS
FAST METS: Direct-to-Unit / Sim-Free Palliative ART
FAST METS treated 47 patients using a simulation-free ETHOS-based adaptive workflow for palliative treatments. Diagnostic CT was used to pre-plan, symptom review was performed before arrival, and final plan recalculation / optimization occurred on the treatment unit. Average total treatment time, including consult, was 85 minutes, with average on-couch adaptive time of 30 minutes.
Simulation-free, one-fraction CT-guided stereotactic adaptive treatment for early-stage NSCLC or oligometastatic lung lesions
Primary endpoint
Feasibility of the sim-free workflow in at least 70% of selected patients
Prescription target
Low-risk peripheral lung lesions suitable for 34 Gy x 1
Inclusion highlights
Target <5 cm, >2 cm from proximal bronchial tree / mediastinum, estimated SI motion ≤1 cm
Accrual status
Listed as closed to accrual and intervention; early report noted 10 / 10 patients successfully accrued and treated
Board pearl: for the purpose of ONE-STOP, a backup conventional sim CT was still obtained for dosimetric comparison and as a rescue plan if direct-to-unit treatment could not be completed. "Sim-free" does not mean "no safety net" in early clinical development.
Adaptive on Additional Systems
C-arm linac adaptive RT: first commercial platform announced in 2024, CE mark granted 9/16/24, FDA 510(k) cleared1/16/2026.
Proton online ART: one early online ART proton case was treated at WashU in June 2024, within the PARTy clinical trial framework.
Functional MR / biologic guidance: MRgART and BgART are emerging as the next layer of adaptation beyond purely geometric replanning.
Where ART Is Likely to Land Long-Term
The long-term view is that ART becomes a versatile, common solution rather than a niche technique: not used in every case, but applicable in most modalities and treatment styles, much like IMRT, VMAT, and SBRT eventually became.
CROSS-CUTTING HIGH-YIELD POINTS
Definition: online ART = same-day re-contouring / re-optimization on the treatment table.
Offline ART replans between fractions; online ART replans while the patient remains on the treatment table.
Plan-of-the-day means selecting from a prepared library; it is adaptive but not the same as full online re-optimization.
Three prerequisites: adequate on-board imaging, fast paired TPS, and online pretreatment QA.
Contour uncertainty is often the dominant failure mode; an adapted plan is only as trustworthy as the daily target and OAR contours.
Development logic matters: imaging study -> in silico workflow testing -> pilot / phase I -> larger prospective trials.
MRgRT low-field platform: 0.35T, no imaging dose, 4 or 8 f/s, automatic gating, better soft-tissue visualization than standard C-arm CBCT.
In silico pancreas SMART: daily ART predicted to be needed in 63% of fractions to protect OARs.
Phase I adaptive upper-GI SMART: OAR violations in 70/95 fractions, all resolved by ART, with 0 grade 3+ acute GI toxicity.
SMART pancreas phase II:50 Gy / 5 fx, BED10 100 Gy, with zero definitely related grade 3+ GI toxicity.
HELIO-RT liver example: large bowel constraint was < 33 Gy to 0.5 cc; adaptation corrected a day-to-day violation.
DARTBOARD: margin reduction from 5 mm to 1 mm lowered dermatitis and produced no marginal recurrences.
SMART ONE: 1-fraction SBRT on 0.35T MR-Linac; ART used in 53% of sites; 10% acute grade 3+, 0% late grade 3+, and zero definitely/probably related events.
SMART ONE key doses: lung 30-34 Gy x 1, liver 35-40 Gy x 1, adrenal / kidney / pancreas / abdominal-pelvic LN 25 Gy x 1.
HERMES: prostate daily ART with 36.25 Gy / 5 or 24 Gy / 2 regimens, all with 3 mm PTV.
ARCHER: phase III bladder trial; mandatory ART; 5-fx SBRT vs 20-fx hypofractionated RT; N = 486.
ART is now mainstream: safety white paper exists, and cooperative-group phase III trials are open.
Platform tradeoff: MR offers superior visualization and motion management; CT-guided adaptive offers higher throughput and versatility.
Cumulative dose is not automatically solved by online adaptation; documentation and dose tracking remain essential.
Best ART candidates have mobile anatomy, tight OAR constraints, high dose per fraction, or a real workflow-access advantage.
Implementation lesson: start with high-volume disease sites and true local champions.
Sim-free future: FAST METS and ONE-STOP show that ART can compress consult + planning + treatment into a single visit for selected indications.
C-arm adaptive and proton adaptive mark the expansion of ART beyond ring-gantry MR/CT adaptive systems.
CONSOLIDATED CONCEPT, DOSE, AND WORKFLOW TABLE
Use case / trial
Concept / dose / margin / workflow
Why it matters
Offline ART
Replan between fractions
Classic approach for slower anatomic change
Online ART
Same-day contour / optimize / QA / treat
Needed for day-to-day OAR rescue
Plan-of-the-day
Select from a prebuilt plan library
Adaptive without necessarily re-optimizing online
Upper-GI / pancreas SMART development
50 Gy / 5 fx
Canonical ablative adaptive GI example
SMART pancreas phase II
50 Gy / 5 fx
Ablation with strict OAR governance and favorable early safety
