The PTV is costly and seems old-fashioned, but it suffers unjustly from the prevention paradox: nobody notices when it works. Its cost is particularly obvious when it is applied for daily (so-called real-time) adaptive treatments planned on high-contrast MR images, which constitute the peak of WYSIWYG in radiotherapy. So, margins seem redundant, but what if this is only an illusion?
Recall that PTV margins predominantly compensate for systematic errors, whose nature it is to be hard to quantify if they are known, or be rather not known at all. In addition, random errors quickly become as relevant as systematic ones in hypofractionation. The beauty of a margin is that it is not specific to the source of uncertainty it compensates, so that unknown systematic errors may be compensated along with the known ones. Hence, the only remedy against PTV margins is neither MR imaging nor real-time planning, but scrupulous analysis of the workflow and profound quality assurance of the equipment.
There are plenty of known systematic errors that carry over from conventional radiotherapy: contouring of targets and organs, alignment of imaging and irradiation coordinates, linac and MLC accuracy, image quality for dose computation. In addition, there is the question of how much anatomical information is lost in “real time” between imaging and irradiation. Due to the complexity of the workflow and the tools involved, unknown systematic errors can hide anywhere, but especially where QA procedures deviate from the workflow employed on patients. Under-complex QA routines and naive end-to-end-tests are the best argument in favour of PTV margins.
Being non-specific is both the greatest strength and the greatest weakness of margins: a little bit of safety goes a long way even for a combination or errors. However, margin reduction, selective, individualized or just blunt, requires quantitative knowledge about error amplitudes that is hard to come by in practice. This is also a severe impediment should the margin concept be replaced with robust dose optimization approaches.
Ultimately, it is only clinical outcome that can measure the sufficiency of a margin, even though it offers only delayed, perilous and cumbersome feedback. However, it is instructive to observe the effect of errors on clinical DVHs. Those obtained from zero margin treatment plans tend to be much more sensitive than those obtained from small margins. Thus, the realistic benefit of margin reduction quickly vanishes for margins smaller than 3 mm, say.
Concluding, daily adaptive MR-Linac treatments per se are a rather weak argument for margin reduction. QA that parallels the clinical workflow and quantifies uncertainties, as opposed to simplistic end-to-end tests with a binary outcome, could yield the relevant information to make an educated guess about an expedient, selective, non-zero PTV margin.