In vitro cell survival experiments on the effects of proton therapy have suggested that there is an increase in relative biological effectiveness (RBE) towards the end of range – going from values of around 1.0–1.1 in the entrance region to around 1.6 in the falloff region – brought about by a similar increase in LET.

In contrast, proton treatment planning and dose reporting are traditionally based on physical dose, assuming a constant RBE. This issue is particularly problematic for IMPT, whose highly modulated fields have the potential to deliver very non-uniform LET distributions, even with a uniform physical dose. Any high LET values in critical structures within, or near, the target volume results in an increased risk of side-effects.

While it has previously been proposed that treatments might be optimized based on biological, rather than physical, dose to compensate for this, such approaches risk partial target under-dosage if the RBE is overestimated.

In their new study, Jan Unkelbach, a medical physicist at Massachusetts General Hospital, and colleagues have focused instead on reducing the risk of normal tissue complications through a hybrid physical dose/LET-based approach that does not require knowledge of RBE.

The hybrid process is implemented through GPU-based Monte Carlo code that provides dose-averaged LET for each pencil beam and has two steps. First – as with current clinical practice – an IMPT plan is developed based on physical dose objectives. Subsequently, this plan is subjected to a prioritized optimization scheme that modifies the LET distribution, shifting LET hotspots to the periphery of the target while maintaining the physical dose distribution.

The researchers demonstrated their treatment re-optimization approach on data from five patients who were previously treated for various intra-cranial tumours (including high-grade meningiomas, base-of-skull chordomas and ependymomas) where the target volume overlaid the brainstem, optic nerve, chiasm or pituitary gland.

"LET hot spots in [these] critical structures can often be avoided at almost no cost regarding physical dose. In that sense, IMPT treatments become safer," says Unkelbach. The success of the optimization method was found to be dependent on patient geometry and beam arrangement, however. For example, high LET values in critical structures overlapping the target volume could only be avoided if the structures were positioned within the entrance region of some pencil beams.

The new optimization method could be easily implemented into commercial treatment planning systems and into the clinical planning workflow, Unkelbach says. "From a technical perspective, the planning system would need the functionality to calculate LET in addition to dose. For Monte-Carlo algorithms, this information comes for free," he explains, adding: "But also, pencil beam algorithms can be extended to provide estimates of LET."

With their initial study complete, the researchers are now looking to develop their optimization approach to increase LET in the gross tumour volume, along with investigating how the LET optimization method relates to robust optimization methods for handling range uncertainty.

Related stories

• Proton RBE: is 1.1 just a myth?
• Robustness analysis improves proton plans
• LET guides proton planning
• LET mapping: a treatment planning tool