VMAT delivers radiation by rotating the gantry of a linac through one or more arcs with the radiation continuously on. As it does so, a number of parameters can be varied. These include: i) the MLC aperture shape, ii) the fluence-output rate ("dose rate"), iii) the gantry rotation speed and iv) the MLC orientation. The genesis of the method was with intensity-modulated arc therapy (IMAT) from Cedric Yu back in 1995,1 but VMAT adds the variability of parameters ii, iii and iv, thus reducing the need to use as many arcs as there are maximum number of field components. The genesis of VMAT is reputed to be from Art Boyer in 2001.2

It is undisputed that VMAT can deliver highly conformal dose distributions similar to those created by other forms of intensity-modulated radiation therapy (IMRT), including the multiple-static field MLC technique, the dynamic MLC (DMLC) technique, static and helical tomotherapy, the CyberKnife, scanned-beam therapy and so on (hereafter referred to as "conventional" IMRT). As such, it becomes a valued member of the IMRT-delivery arsenal. That unmodulated VMAT fields can deliver the equivalent of modulated beams has recently been explained.3

Planning for VMAT is so far relatively embryonic, although several companies offer systems. These differ in their methodology and some are based on well-respected doctoral studies. In essence, most operate by creating some form of fixed-field modulated beams, decomposing these into MLC components, redistributing those over small arcs and re-optimizing the outcome. In doing so, VMAT can take advantage of the above-mentioned four available variable parameters, but must do so while respecting the physical constraints of the linac and MLC - such as the maximum gantry speed, maximum leaf speed, the MLC orientation constraints and the available subdivisions of fluence-output rate. The first and fourth of these are of course linked.

Provided that the gantry speed can be varied continuously, it does not require a continuous variation of fluence-output rate to obtain a continuous variability of fluence-output rate per degree. The minimum fluence-output rate and the maximum gantry speed determine the constraining minimum fluence-output rate per degree. Where there is a maximum fluence-output rate and minimum gantry speed, there will be a constraining maximum fluence-output rate per degree. Unlike the technologies for optimizing "conventional" IMRT techniques (i.e., those mentioned above), this planning technology is not yet considered mature and completed. However, a key observation is that both planning and delivery for VMAT are commercially available now. Their gestation did not go through the lengthy phase of some of the other methods with a vast collection of peer-review literature along the way.

Key considerations
So why is VMAT controversial? I submit several reasons. The first is that there have been early claims that VMAT can generate equivalently conformal dose distributions with fewer monitor units (MUs), i.e., in a faster time. To have that is clearly advantageous (shorter treatments; better for patients in discomfort; less susceptibility to intrafraction motion; possibly less induced secondary cancers; quicker overall treatment slots…). These claims have, not surprisingly, been bolstered by the companies marketing the technology. Lucrative sales are arising.

My "talking (debating) points" rest on the following statements:

  1. The majority of the VMAT peer-reviewed papers centre on planning comparisons between VMAT and other (usually fixed-field) IMRT methods. By definition, these compare a single implementation of each, which is valid, but only in that context. Improve the implementation on either side of the balancing scales and the conclusions may change.
  2. The literature is already showing that in some circumstances VMAT indeed "does better" than fixed-field IMRT, but that in other circumstances the reverse is true: VMAT "does worse". Any claims that VMAT is universally superior are not supportable.
  3. To date, the clinical benefit of VMAT is largely unquantified; even if it "does better", the benefit may be small. Those who have installed "conventional" IMRT equipment and are using it effectively in the clinic, and possibly in trials, should not somehow feel that they have outdated equipment.
  4. Finally, and as primarily a theoretician I am most interested in this, there is no universal theory that shows how the outcome of VMAT depends on the parameters available for variation. That is quite unlike some of the fixed-field IMRT techniques for which clear published theories exist. For example, the amount of modulation deliverable in a DMLC method can be clearly written down via the Stein equations worked out back in 1994 by Jorg Stein with Thomas Bortfeld and Wolfgang Schlegel.4 There are similar papers explaining how the conformality of step-and-shoot IMRT depends on the number of fields, their separation and the number of field components. There is no equivalent VMAT theory and sadly those who report comparative planning studies do not go on to say much more than that "the outcome depends on the nature and complexity of the problem". That is not a reason to feel that those papers are in some way unpublishable - quite the reverse. It is just that they fall short of an intellectually satisfying explanation of the observed outcomes. Whilst VMAT provides some indisputable advantages in some situations, it has unexplained features and should not at present be regarded as supplanting "conventional" IMRT techniques.

With colleagues, I have spent much of the last year attempting to understand the theory. Thomas Bortfeld and I made an overly simple 2D analysis.2 We welcomed the comments from two other groups on that Note and provided some response.5 We accepted most of the comments, but we ended both our Letters by pointing out the above lack of theory.

The new paper by Webb and McQuaid3 explains some key aspects of VMAT, including:
 • why an MLC on a moving carriage can still deliver IMRT via the DMLC method;
 • why what are essentially fields planned at discrete gantry angles can be delivered with a moving source and gantry (the "small-arc approximation");
 • why unmodulated VMAT fields can be considered equivalent to reorganized modulated fields;
 • why VMAT may offer some time advantage.
However, and it is a big "however", we make no claims that this is the universal theory of VMAT. These are "stepping stones" towards such. They are "jigsaw pieces" that must be part of the theory.

I should expect that as time passes, the methodology of VMAT - both in planning and delivery - will evolve and improve. There will be more papers comparing VMAT with "conventional" IMRT. The manufacturers will undoubtedly make further claims. What I hope is that this will take place in a spirit of friendly debate with acknowledgement that, until some general theory of VMAT is published, claims for the role of VMAT should be strongly contextualized.