Clinical implementation: how can IMRT become routine?

The fact is, although some people may think that IMRT is fully mature and now represents the standard radiation therapy technique, this is not entirely the case. It does make sense, therefore, to try to identify the main causes underlying the delayed transition from three-dimensional conformal radiotherapy (3D-CRT) to IMRT. I will focus on the technical aspects here, leaving aside clinical and economical considerations, which merit a discussion of their own.

We should first remember that implementing IMRT following several years of 3D-CRT experience is a completely different challenge to carrying out the same development in a department where 2D radiotherapy is still the de facto standard. While this may be stating the obvious, we tend to forget one important reason why IMRT has not shown a faster adoption rate: when IMRT came into being, CRT (i.e., CT-based 3D treatment planning with conformal apertures, based on delineated contours for both the targets and organs-at-risk) was not the standard for every centre and/or treatment site, even in more affluent parts of the world.

Furthermore, for clinicians in particular, it may be harder to move from 2D radiotherapy to CRT than from CRT to IMRT. If we take these aspects into account, the slow pace of IMRT implementation suddenly becomes easier to understand.

Worlds apart
From a technical perspective, the difficulty in transitioning from CRT to IMRT may be explained by one simple observation. Namely, that for quite some time, CRT and IMRT were considered - by most clinics and by the companies producing radiotherapy equipment - as two different worlds that could not be connected via a smooth path.

Prostate IMRT treatment

Imagine, for example, that you want to treat a patient with prostate cancer and your choice is between a four-field CRT plan and an IMRT plan with 400 control points. Assuming you are not comfortable with dose escalation, the two plans can - in most cases - be designed to deliver the same dose to the target, with IMRT allowing you to reduce the volume of rectum receiving higher doses (65 Gy and above). But in this situation, would you be excited enough by the IMRT plan to immediately transition all of your prostate patients to IMRT?

I am stretching things a bit, but I'm sure you see the point: in the early days of IMRT, it was extremely difficult to envision it as a logical extension of CRT. People were faced with an entirely different planning procedure and brand new problems regarding dose calculation. And it was basically impossible to tune the complexity of the technique to the complexity of the clinical needs being addressed.

The radiotherapy community needed time to fully grasp the specifics of this new technique and to benefit from IMRT's additional degrees of freedom without making their lives unnecessarily complicated. In a way, people had to realize that even if you could deliver extremely complex segment patterns, this doesn't mean that you actually need to and/or should do for every patient from day one - given that this complexity usually comes at a cost.

But what do we mean exactly by "cost"? If we think about it, it's not about the treatment delivery per se. The multileaf collimator (MLC) has been criticized by some as a (highly) imperfect delivery tool for IMRT, linear accelerators need to be improved in some aspects (increasing output stability for small numbers of monitor units, for example), and the delivery itself might be quite slow. Despite this, it's not the hardware that's been the main obstacle to a widespread uptake of IMRT. The real cost lay in the complexity of the treatment-planning and verification process.

In my opinion, improvements in treatment-planning software have played an essential role in making IMRT less cumbersome than it was ten years ago. Additional improvements in planning packages, combined with more efficient verification techniques, will further simplify the technicalities of IMRT in the years to come.

Treatment planning
Getting used to the concepts that underlie treatment-plan optimization is, in itself, a fair obstacle in the transition from CRT to IMRT. If users of treatment-planning systems are faced with optimization software that makes it difficult to translate clinical requirements into a cost function and to "steer" the optimization towards an acceptable result, IMRT planning becomes a time consuming and frustrating task.

Nasopharynx IMRT treatment

If there's a single development in treatment planning worth mentioning here, it's the possibility of optimizing deliverable plans, as opposed to fluence distributions. The experience of the past years has taught us that this is a key step in improving the whole planning procedure, from the definition of the treatment objectives to the dose calculation.

What is still missing from the optimization modules of most commercial treatment-planning systems is a means to gain insight into the optimization results; for example, the ability to identify which planning objectives are most important in determining a given solution and to evaluate which adjustment to the cost functions could produce better results. When available, such tools will allow the planner to better trust the final plan, which will be not just "a" plan, but a dose distribution carefully chosen from several realistic alternatives.

Treatment verification
The increased complexity in segment patterns that characterizes IMRT was not matched, particularly in the early years, by an adequate increase in the accuracy of dose calculation. With CRT, tasks such as MLC quality assurance and treatment-planning-system commissioning were strictly decoupled from the evaluation of the dosimetric accuracy for a single treatment plan. With IMRT, however, this separation mostly disappeared.

Pre-treatment dose verification of each treatment plan thus became a routine part of the IMRT process - a clear indicator that the accuracy of the dose calculation was not fully trusted (and for good reasons) by most medical physicists, and that it was difficult to define a priori accuracy criteria for the treatment-planning system that, on the one hand, were achievable, and on the other, would provide full confidence in the accuracy of every single plan. In many clinics, the time required to carry out such pre-treatment dose verification is still the real bottleneck hindering large-scale IMRT implementation.

There are two possible solutions to this problem: make the verification process fast and efficient, or somehow avoid the need for it altogether. With regard to the first approach, the possibility of combining all treatment-verification steps, by performing EPID-based in vivo dosimetry during the first treatment session(s) is particularly interesting.

Such a scheme would enable the combined effects of dosimetric and anatomic discrepancies between planning and delivery to be detected in one measurement. Should differences larger than acceptable occur, a CT dataset acquired in the same session could then reveal whether these were caused by dose calculation and/or anatomy changes. These methodologies are accessible to a minority of institutions at the moment, but that situation may change (perhaps even significantly) in the next three to five years.

Others argue that there is no reason why the accuracy criteria and approaches deemed valid for CRT should not apply to IMRT too, and that dosimetric pre-treatment verification should be avoided by implementing better dose algorithms in commercial treatment-planning systems. In practice, this means that Monte Carlo dose calculations should become standard.

The large-scale availability of Monte Carlo-based dose calculations in clinical practice has been anticipated for years, but remained largely an unkept promise. One of the causes for this delay, was actually the interest in IMRT, which further shifted the focus from accuracy to speed in dose calculation. We are now at a time where high accuracy and satisfactory speed can be combined. Thus there is no reason why Monte Carlo should not become the standard "dose engine" for IMRT planning in the near future.