This study found that large shifts (> 1 cm) routinely occurred during the course of radiotherapy, even when 5-point masks are used. In general, there was no trend with time in the magnitude of shifts in the RL, AP, or SI directions. Although some patients did show either an increasing or decreasing trend in the size of shoulder shift as treatment progressed, equal numbers of patients had shifts that were larger or smaller and the vast majority had no trend in any direction. While large shoulder shifts could be expected to occur late in treatment due to patient relaxation or weight loss, we observed such shifts to occur equally both early and late in treatment.
In this study, we observed the largest shoulder shifts in the AP and SI directions, which is different from results reported in the literature where the largest shifts were found in the RL direction. However, our study examined the position after correct isocenter setup, whereas the others examined the position prior to correct setup [3, 4] and evaluated how much the shoulders or the shoulder region had to be moved to properly align isocenter. These studies were also limited by visibility on megavoltage images. However, the magnitudes of the shoulder displacements were comparable between our study and the others, being typically less than 6 mm.
The observed shoulder motion is consistent with previous studies that showed lower-neck structures experience more setup variability when aligning to C2 [7, 8]. If the patient were setup using a lower target for alignment, such as C7, there may be a reduction in the size of the observed shoulder shift as medial lower-neck targets would show less variability. However, this was not observed in this study. Only one patient was aligned to vertebral bodies in the low neck (C7-T3) and this patient showed a great deal of shoulder variability (Figure 3). The authors believe that the shoulders may still show large displacements because the shoulders are far from mid-line and can move independently of the low neck vertebral bodies, and are often ignored in setup.
One drawback to our method of aligning the patient is that is only occurs in the 3 linear directions without any rotational component. Therefore, it was not possible to reduce or minimize shoulder shifts that were the result of rotated anatomy. Therefore, the size of observed shoulder shifts may be reduced if the IGRT process includes rotational as well as translational shifts. However, based on the findings throughout this work that the shoulders move largely independently, it is likely that large shoulder displacements could still occur, even with rotation accounted for.
The greatest change in lower neck target coverage was found for superior shifts because these brought shoulder tissue into a region where it was previously absent, thereby changing the depth and beam attenuation to the target. This was observed even when the inferior jaws of lateral fields are closed above the shoulder, as seen in Patient 2 (Table 2). A similar result was found for large posterior shifts, which resulted in target coverage loss for IMRT plans. VMAT plans did not show a similar loss of coverage for posterior shifts. This likely resulted from the Monitor Unit (MU) distribution; even though beams were evenly spaced around the patient, the IMRT plans had ~50% of the MU from posterior or posterior oblique beams. In contrast, VMAT plans had a relatively even distribution of MUs with gantry angle.
It is important to note that the coverage loss from superior and posterior shifts was not compensated for by an equivalent increase in coverage from inferior or anterior shifts. That is, the effect of the shift does not average out over the course of treatment. This can be understood by considering that a superior shift will cause attenuation and loss of coverage to a transverse section of the neck. A subsequent inferior shift will increase the dose slightly, but to a different transverse section of the neck (an inferior section), thereby not compensating for the dose loss associated with the superior shift. The position of the shoulder each day has an impact on coverage, and a mean shoulder position will not represent the total effect of the shoulder movement over the course of treatment. Documentation of the number of superior and large posterior shifts will give the best information about loss of coverage to the target.
In a clinical setting, the most important impact of shoulder motion is the loss of target coverage (~1 Gy to 99% of a lower neck target). This may be important, particularly if there is primary disease in that region. Moreover, shoulder shifts may be an important consideration for Stereotactic Spine Radiosurgery (SSRS) or for other patients undergoing hypofractionated therapy with lesions near C6. These IMRT plans are delivered in few (or one) fractions, so large errors are not mitigated by subsequent fractions. It is important to consider that this positioning study found no trend with time for large shoulder shifts; large shifts (> 1 cm) were seen in the first few days of treatment for many patients. In addition to these considerations, treatments that use predominantly posterior beams may suffer coverage loss worse than that predicted in this study. Also, systematic shoulder shifts are more likely to cause substantial dose losses similar to those shown in Table 2. While most patients in this study had small systematic shifts, 2 patients demonstrated large (8-10 mm) systematic shifts.
When we evaluated dosimetric impact, no clinically important change was seen in dose to the spinal cord because it had been avoided in the treatment plans and it was always associated with low photon fluences. However, the brachial plexus was located close to the targets, so changes in shoulder position affected beam attenuation and dose to this structure. The overall 72 cGy increase in dose to 0.1 cm3 of the brachial plexus is not likely to cause harm because the max dose to 0.1 cm3 of the brachial plexus is not always in the same location within the structure, depending on shoulder position. The daily increase in dose to 0.1 cm3 of the brachial plexus was a few cGy; therefore, the dose escalation required to receive a TD5/5 dose of over 60 Gy  on a single day was not observed.