In new work, researchers at The Netherlands Cancer Institute (NKI) in Amsterdam have made the first detailed assessment of the dosimetric impact of inter-fractional anatomical variations on pencil-beam proton therapy. Variations include those resulting from tumour regression, systematic shifts in tumour position and changes in the extent of atelectasis or lung parenchyma collapse. The retrospective planning study revealed significant deviations from planned dose distributions due to the variations (Radiother. Oncol. doi: 10.1016/j.radonc.2016.04.002).

"We demonstrated that inter-fractional anatomical variations can considerably reduce the dose to the tumour and increase the dose to organs-at-risk," said Yenny Szeto, first author and PhD candidate. The deviations could impact the likelihood of disease control and the risk of side effects. Consequently, there is a strong need for treatments that are robust to the variations, the authors argue.

Led by Marcel van Herk and Jan-Jakob Sonke, the researchers studied a subgroup of patients with locally advanced non-small cell lung cancer (NSCLC) who had received intensity-modulated radiotherapy (IMRT). The patients had not received the full, standard 66 Gy photon prescription used at NKI due to unacceptably high organ-at-risk (OAR) doses.

Treatment plans were optimized on an image representing the average anatomy over the respiratory cycle, derived from a planning 4DCT scan. To improve robustness to respiratory motion, an internal gross tumour volume (GTV), representing all GTV positions over the respiratory cycle in the 4DCT, was assigned the density of soft tissue during dose calculation. Planning software also optimized the robustness of the treatment plans for a ±3% range uncertainty and a ±3 mm setup uncertainty.

Anatomical variations over each patient's treatment course had been captured using daily cone-beam CT (CBCT) scans. The researchers used the images to estimate the cumulative, "delivered" dose distributions that they compared with the planned dose distributions.

The cumulative distributions were reconstructed using deformable registration, first to simulate a 4D CBCT from each treatment image, exploiting respiratory motion data from the planning 4DCT. Using this, the dose "delivered" in each respiratory phase was mapped onto a single image for each fraction. In turn, deformable registration was used to map the dose from each fraction onto the planning CT scan.

When the researchers examined the combined effect of anatomical variations and respiratory motion, they found that tumours were underdosed by more than 2 Gy equivalent in eight of 16 patients. The largest target underdose was 14 Gy equivalent.

Some OARs received a higher dose, though to a smaller extent than the target underdoses. The third quartile dose of the mediastinal OARs, which demonstrated the biggest deviations, corresponded to an overdose of 2 Gy equivalent. Further analyses revealed deviations from the planned dose distributions due to anatomical variations were an order of magnitude greater than those due to respiratory motion.

Based on their findings, the researchers have begun to develop a strategy for generating treatment plans that are robust to anatomical variations. The group is using daily CBCT scans acquired in individuals with NSCLC receiving IMRT to build a statistical model describing anatomical variations across the patient population. In parallel, they are exploring ways to incorporate the model into treatment planning.

In cases with large changes in anatomy, however, robust treatment plans alone will be inadequate to deliver clinically acceptable dose distributions. These cases can be identified using daily imaging and dose recalculation and will require more labour-intensive adaptive treatments, Szeto told medicalphysicsweb.

"In patients where the dosimetric changes exceed clinically defined tolerance limits, the existing treatment plan must be re-optimized following re-contouring of the target and organs-at-risk," said Szeto.

Related stories

Adaptive treatments are a must for protons
CBCT works well in adaptive proton therapy
Improving proton therapy for lung tumours
Robustness analysis improves proton plans