KEY POINTS
- Nine medical physicists from different cancer centers developed a generic external-beam RT process map and identified 63 potential failure modes relevant to photon IMRT/VMAT, including SBRT, in workflows requiring data transfer between the treatment planning and oncology information systems.
- Each failure mode was scored for occurrence, severity and detectability using TG-100 criteria. Investigators compared the traditional risk priority number (occurrence × severity × detectability) with the newer Action Priority framework adopted from automotive-industry FMEA, which gives greater weight to severity and places risks directly into high, medium or low action categories.
- Because scoring differed between institutions—particularly for detectability—the investigators evaluated average, worst-case and severity-focused scenarios. They ultimately emphasized the severity-focused approach, using the minimum occurrence, maximum severity and average detectability scores.
- Under that analysis, Action Priority selected only 7 high-risk failure modes: one high-priority and six medium-priority. Traditional TG-100-style selection identified 30 failure modes, consisting of the same seven plus 23 additional flags from the top risk-priority-number quartile and/or severity ≥8 criterion.
- The seven shared priorities included stray voxels in target contours, excessively small MLC gaps, planning on an outdated CT after weight loss, incorrect beam-parameter templates, using the wrong breathing-phase CT, unusually high monitor-unit complexity and modifying a plan after treatment had started.
- A structural problem with traditional RPN ranking was that different combinations of occurrence, severity and detectability could produce identical numerical scores despite very different clinical consequences. In addition, adding or removing failure modes changed the top-quartile cutoff, meaning the list of “high-risk” failures could change simply because the FMEA itself was expanded.
- Action Priority was reported as easier to apply, less sensitive to iteration of the failure-mode list and more explicit about whether action must, should or could be taken. However, this was a methodological exercise rather than validation against real incident rates, and both approaches ultimately depend on subjective expert scoring.
CLINICAL TAKEAWAY
The updated Action Priority approach may solve a familiar TG-100 problem: a long list of “high-risk” items is not useful if everything becomes a priority. In this exercise it preserved the same seven core risks while eliminating 23 additional flags, but the framework still needs validation against real-world incident and near-miss data before replacing conventional FMEA practice.