Loading
Most radiotherapy errors come not from broken machines but from steps in the process: a wrong image, a mistyped number, a missed message. Departments therefore map every step of the treatment pathway, ask what could go wrong at each, score how likely, how harmful and how hard to spot each failure would be, and put their safety checks where the risk is highest.
AAPM Task Group 100 (2016) set out the reference method, starting from the observation that many errors are failures of workflow and process rather than of devices and software. A multidisciplinary team maps the process in detail, lists the failure modes at each step, scores occurrence (O), severity (S) and lack of detectability (D), usually from 1 to 10, and ranks them by the risk priority number (RPN = O x S x D). TG-100 looked at the top 20% to 25% by RPN and by severity, leaving each team to set its own threshold; studies commonly treat the top quartile of RPN and any severity of 8 or more as high risk. Fault tree analysis traces how failures propagate, and the quality management programme is then designed around the high-risk failure modes. MPPG 8.b (2023) asks every centre to run its own analysis, with published RPNs used only as a guide.
A 2025 systematic review found 62 prospective hazard analyses in external beam radiotherapy: FMEA in 75%, STPA in 10%; 579 high-risk failure modes, half in treatment delivery sessions and a quarter in contouring and planning; and mitigations that tended to add workload and sit low on the hierarchy of effective interventions. A 2026 Ontario multi-institution exercise by nine physicists found the automotive-industry 'action priority' table, which weights severity and occurrence over detectability, picked out the same critical failures with less scoring disagreement. Incident learning, checklists, training and staffing are the commonest other safety recommendations across seven international documents.
Why: Machine checks alone miss the commonest causes of error, such as wrong data passed between steps.
“Many errors that occur in radiation oncology are not due to failures in devices and software; rather they are failures in workflow and process.”
American Association of Physicists in Medicine (Huq et al., Medical Physics) 2016 · ungraded (task group report)
Why: Each profession sees different ways a step can go wrong.
“Disciplines involved in the studies were physicists (92%), physicians (75%), radiation therapists or dosimetrists (71%), external experts (38%), and facilitators (33%).”
Medical Physics (open access) 2025 · systematic review finding
Why: Writing every step down makes hidden weak points visible before they hurt anyone.
“involves generating a process map detailing the linear sequence of events involved in the process through analytical reduction, identifying the failure modes (FM) in each step of the process, assessing the effects of each FM in terms of their occurrence (O), detectability (D) and severity (S), and generating the risk priority number (RPN = O × S × D).”
Medical Physics (open access) 2025 · method as described (TG-100)
Why: A rare but catastrophic failure needs attention even when its overall score is modest.
“As suggested in TG-100, PFMs that were in the top quartile of RPN scores and/or with a severity score greater than or equal to eight were deemed to be high risk profile PFMs.”
Rangel et al., Journal of Applied Clinical Medical Physics 2026 · method as described (TG-100)
Why: Some harms happen only when two small failures line up; a fault tree shows where one check can block both.
“Of the included studies, 11 used fault tree analysis, however, only two studies used FTA alone, the rest of the studies used it alongside FMEA in the way described in TG-100.”
Medical Physics (open access) 2025 · systematic review finding
Why: Departments differ in equipment, staffing and workflow, so another centre's risk scores do not transfer.
“Each center should conduct their own risk analysis of their site-specific QA procedures.”
Why: Preventing an error is more reliable than hoping a busy person spots it later.
“The AP framework's emphasis on prevention over detection aligns with established hierarchies of intervention effectiveness.”
Rangel et al., Journal of Applied Clinical Medical Physics 2026 · study discussion (no grade)
Why: A fix can fail or create a new risk; checking again shows whether it worked.
“PHA involves the use of a formalized method of assessing and/or defining a system, technology, or technique, evaluating it for hazards, prioritizing the identified hazards and reassessment following the development of mitigation strategies to reduce the risk of the identified hazards.”
Medical Physics (open access) 2025 · method as described
Why: Risks can be found and designed out before anyone is exposed to them.
“A benefit of PHA is that it can be performed during development of a system, technology, or technique.”
Medical Physics (open access) 2025 · review statement
Why: Risk analysis is one of several layers; across international safety reports these others are named more often.
“They are, in order of most to least cited: training, staffing, documentation, incident learning, communication, check lists, quality control and preventive maintenance, dosimetric audit, accreditation, minimizing interruptions, prospective risk assessment, and safety culture.”
Dunscombe, Frontiers in Oncology 2012 · mapping of seven documents (no grade)
How care is usually carried out, read from the guidelines and standards cited. Your own team's protocol and their judgement come first; this page is for understanding it, not for doing it yourself.
AAPM TG-100 (2016) is the reference method: a multidisciplinary team maps the process, scores failure modes for occurrence, severity and detectability, ranks them, adds fault tree analysis, and builds the quality management programme around the high-risk failures; MPPG 8.b (2023) asks each centre to do this locally rather than borrow published scores. The 2025 systematic review shows FMEA in three-quarters of published analyses and warns that mitigations often add workload and sit low on the hierarchy of interventions. Where methods differ, both are shown: TG-100's RPN with a severity threshold, or the action priority table that weights severity and occurrence over detectability (Ontario multi-institution exercise, 2026). No trial shows that any risk analysis method reduces patient harm; the evidence is methodological and observational.
American Association of Physicists in Medicine (Huq et al., Medical Physics) 2016 · American Association of Physicists in Medicine (Krauss et al., Journal of Applied Clinical Medical Physics) 2023 · Medical Physics (open access) 2025 · Rangel et al., Journal of Applied Clinical Medical Physics 2026
Where it differs: Uses a look-up table from the automotive industry's updated FMEA that ranks failures High, Medium or Low, weighting occurrence over detectability, rather than multiplying O x S x D and applying a quartile and severity cut-off.
Evidence: In an exercise by nine physicists from different Ontario cancer centres on 63 failure modes, AP selected 7 high-risk failures and RPN 30; all 7 were among the RPN set; AP was less affected by scoring variability and easier to apply.
Rangel et al., Journal of Applied Clinical Medical Physics 2026
Where it differs: A top-down method based on system theory: it starts from the events to be prevented and works through how they could come about, focusing on control loops and feedback, instead of listing failure modes step by step.
Evidence: Used in 10% of the 62 studies in the 2025 systematic review.
Where it differs: A high-level assessment, developed in Australia, of the risks and benefits of a new technology or process before work to introduce it starts.
Evidence: Described in the 2025 systematic review; no comparative outcome data.
Where it differs: Following task group tolerance tables (TG-142, MPPG 8.b) as written, without a local FMEA. MPPG 8.b treats its list as the minimum and asks for local risk analysis on top.
Evidence: TG-100 argues prescriptive device-focused QA cannot keep pace with complex modern processes; MPPG 8.b keeps a minimum list and cites a survey in which physicists did not agree on the risk of common failures.
American Association of Physicists in Medicine (Huq et al., Medical Physics) 2016 · American Association of Physicists in Medicine (Krauss et al., Journal of Applied Clinical Medical Physics) 2023
| Recommendation | Grade, as printed | Source |
|---|---|---|
| TG-100 considered the top 20% to 25% of failure modes by RPN and severity, leaving the threshold to each team “TG-100 chose to consider the top 20%–25% of RPN and severity ranked failure modes but indicates that it is up to each individual team to determine the threshold appropriate for their system.” | method as reported | Medical Physics (open access) 2025 |
| Published mitigations tend to add workload “The mitigation strategies recommended by studies tended to add to the departmental workload.” | systematic review finding | Medical Physics (open access) 2025 |
| Use the MPPG 8.b tests as a minimum for any local risk analysis “We recommend using the tests and frequencies outlined in this report as a minimum for any risk analysis performed and that full consideration be given to the implication of removing any of the recommended minimum components.” | recommendation (committee) | American Association of Physicists in Medicine (Krauss et al., Journal of Applied Clinical Medical Physics) 2023 |
Shares Patient identification and pre-treatment time-out in radiotherapy, IMRT / IGRT (modern external beam).
Shares C-arm medical linear accelerators (TrueBeam, Versa HD and others), IMRT / IGRT (modern external beam).
Shares Linear accelerator quality assurance: daily, weekly, monthly and annual machine checks, C-arm medical linear accelerators (TrueBeam, Versa HD and others), IMRT / IGRT (modern external beam).
Shares C-arm medical linear accelerators (TrueBeam, Versa HD and others), IMRT / IGRT (modern external beam).
Shares Patient identification and pre-treatment time-out in radiotherapy, IMRT / IGRT (modern external beam).
Shares C-arm medical linear accelerators (TrueBeam, Versa HD and others), IMRT / IGRT (modern external beam).
Shares C-arm medical linear accelerators (TrueBeam, Versa HD and others), IMRT / IGRT (modern external beam).
Shares C-arm medical linear accelerators (TrueBeam, Versa HD and others), IMRT / IGRT (modern external beam).