Definitive radiotherapy (RT) is a vital multimodal treatment option for prostate cancer, the most common malignant tumor in men (1, 2). Definitive RT demonstrates excellent local control rates from 85–90%, and randomized clinical trials have indicated that dose escalation further improves the local control rate (3–6). Consequently, increasing focus has been laid on hypofractionated RT or stereotactic body radiotherapy (SBRT) to treat prostate cancer (7). However, the risk of treatment-related toxicity increases with dose escalation and hypofractionation (8, 9).

A higher RT dose is delivered at each fraction in hypofractionated RT or SBRT; therefore, accurate and precise target localization is necessary. Minor deviations may lead to undercoverage of target volumes and higher toxicity rates (10). Several studies have attempted to assess the optimal target margin in prostate SBRT using image-guided RT (IGRT); however, they obtained limited information on prostate movement owing to the lack of full-time continuous monitoring or the use of X-ray-based image guidance (11, 12).

MRI-guided radiotherapy (MRIgRT) facilitates improved soft tissue contrast and better visualization of the target and surrounding tissues compared with that using X-ray-based IGRT. The most recent versions of MRIgRT systems, such as the ViewRay and Elekta MR-LINAC systems, combine intensity-modulated radiotherapy (IMRT) and real-time MRI-guidance with motion-monitoring capability, which allows online adaptive planning (ART) (13–15). MRIgRT with online adaptation for the prostate is expected to better address the daily changes of the target and organs at risk (OAR), such as prostate swelling, bladder filling, or rectal gas movement, and provide improved tumor control and reduced toxicities (16).

Furthermore, target volume can be reduced by accurately assessing the inter- and intrafractional variabilities. Previous studies have reported a significant reduction of the planning target volume (PTV) with adaptive planning using MRIgRT (17, 18). Interobserver variability is another important source of uncertainty in fractionated radiotherapy with online adaptation. Concerns for interobserver variability arise when recontouring the target and OARs for ART, even with advanced imaging techniques (19).

Accordingly, we aimed to assess interfractional and intrafractional variability during 1.5 T MRIgRT for the prostate and suggest the optimal PTV margin for definitive prostate RT. We delineated 400 prostate contours on daily pretreatment MRI images for analysis and presented the results of the comprehensive analysis.

Patient and tumor characteristics are presented in Table 1 . The median age was 77 years (interquartile range [IQR], 72–83), and 11 patients (55%) had stage T2 cancer. Regarding risk classification, one patient (5%) was low risk, five (25%) were unfavorable intermediate risk, four (20%) were high risk, and 10 (50%) were very high risk. The median volumes of the prostate, bladder, and rectum were 42.5, 182.6, and 62.8 mL, respectively. Regarding the radiation field, six patients (30%) received RT only to the prostate, two (10%) received RT to the prostate and the seminal vesicles, and 12 (60%) received RT to the prostate and pelvic lymph node area. All patients completed their RT course. Nineteen patients (95%) received hormone therapy concurrently with RT, and 14 (70%) continued to receive hormone therapy after RT.

A total of 100 PRE-MRIs were collected, and 400 prostate contours were delineated. Regarding intrafractional variability assessment, 25 MM-MRIs were analyzed.

In this study, we used daily pretreatment MRI images of 20 patients receiving RT to evaluate the interfractional and intrafractional variation of the prostate. The variation in the SI direction was the largest in SM, IM, and IO, and that in the LR direction was the smallest. The PTV margin is a determining factor for local control and toxicity in definitive prostate RT; therefore, an accurate assessment of the PTV margin is necessary, particularly with the implementation of hypofractionation and SBRT (22–24).

In our study, we observed that the interfractional and intrafractional variability of the prostate was greater in the SI and AP directions compared with that in the LR direction. This finding is in parallel with many others. A study by M.D. Anderson Cancer Center quantified the prostate and seminal vesicle interfractional movement by contouring the target and OARs on 369 CT scans and observed that the dominant variation was in the AP and SI directions (25). Huang et al. measured the intrafractional prostate motion of 20 patients with prostate cancer and observed that intrafractional movement predominantly occurred in the anterior and superior directions (26). Mah et al. also measured intrafractional prostate motion with cine-MRIs and observed that prostate movement mostly occurred anteriorly and superiorly (27). Wong et al. reviewed CT images of 329 patients with prostate cancer receiving image-guided RT to analyze the interfractional prostate shifts and reported that the shift in the AP direction was higher than that in the LR or SI directions (28).

Interfractional variability based on bone anatomy and prostate registration revealed a strong correlation in the AP direction. In contrast, the correlation was moderate in the SI direction and weak in the LR direction. Our finding differs from those of a previous Mayo Clinic study in which Beltran et al. observed a strong correlation in the LR direction (Pearson r = 0.89), a moderate correlation in the AP direction (Pearson r = 0.59), and a weak correlation in the SI direction (Pearson r = 0.35) (29). However, the method of prostate-based registration by these authors differed from ours: we used the outer contours of the prostate, whereas Beltran et al. used intraprostatic gold seeds, which may have resulted in such a difference. Nonetheless, the two studies are concordant in that bone registration cannot be a perfect surrogate for prostate registration; therefore, ART is necessary for accurate treatment delivery.

Interobserver variability was also a substantial source of variation in prostate RT. Previous studies also highlighted the clinical significance of interobserver variability. In a study from Princess Margaret Hospital, Toronto, Canada, five genitourinary oncologists analyzed prostate contours on five cone beam CT images and identified considerable disagreement (30). Variability decreased with improved soft tissue contrast imaging. Lütgendorf-Caucig et al. reported interobserver variability of the prostate target volume delineation and observed larger variability for cone beam CT-based contouring than that for CT and MRI (31). Villeirs et al. compared the interobserver variation of prostate and seminal vesicle contours using CT alone and CT plus MRI and observed that CT plus MRI reduced CTV volume and the standard deviation values (32). Nevertheless, our findings revealed that interobserver variability should still be accounted for when assessing the prostate PTV margin. In our study, the observers did not discuss how to contour the prostate prior to contouring. One way to reduce interobserver variability may be to use the same consensus guideline (33).

In this study, we assessed SM, IM, and IO and suggested a PTV margin of 2.2, 5.2, and 5.4 mm in the LR, AP, and SI directions, respectively. Beltran et al. compared four localization methods of skin marks using tattoos, pelvic bony anatomy, and intraprostatic gold seeds using a 5 mm action threshold and no threshold and suggested that localization using gold fiducials can largely reduce the PTV margin compared to skin marks or bony anatomy for prostate RT. The suggested PTV margin when using intraprostatic gold seeds was 4.8, 5.2, and 5.4 mm in LR, AP, and SI directions, respectively, which was twice larger than our result in the LR direction but identical to our results in the AP and SI directions (29). MRIgRT may be preferred for patient comfort and safety, considering that fiducial insertion is invasive and carries the risk of infection, albeit low (34, 35). We also assessed the PTV margin for patients with clinical T1-2, N0 disease, which are potential candidates for prostate SBRT, and observed that a smaller margin was necessary for these cases. This implies that smaller margins can be employed for selected SBRT cases; however, further validation with a larger patient cohort is necessary. Moreover, the PTV margin calculated using the van Herk formula based on the systematic (Ʃ) and random (σ) error values obtained in this study can be further reduced to 1.4, 4.5, and 5.2 mm in the LR, AP, and SI directions, respectively, when IO was neglected. Recently, hydrogel rectal spacer insertion has been attempted to further reduce PTV margin for prostate SBRT. Cuccia et al. analyzed the translational and rotational shifts of patients who underwent MRI-guided prostate SBRT with or without hydrogel spacer, and observed significantly minimized rotational shifts in the antero-posterior axis with hydrogel spacer (36). Similarly, Mazzola et al. analyzed the impact of hydrogel spacer on seminal vesicle motion in patients undergoing MRI-guided prostate SBRT and found significant reduction of translational shift in the cranio-caudal axis with hydrogel spacer (37). While the method of intrafractional variability assessment of the previously mentioned studies differed from the one used in this study—the former using pre- and post-treatment MRI images and the latter using real time MM images—all highlight the importance of PTV margin minimization and appropriate patient selection for prostate SBRT.

The strengths of this study include the implementation of MRIgRT, a non-invasive method with superior soft tissue contrast, and the collection of daily MRIs for image guidance and real-time motion monitoring to assess all potential sources of variability during RT (SM, IM, and IO) and suggest an optimal PTV margin using a well-known formula. However, this study also had some limitations, mainly attributed to its retrospective design. Image registration was performed by one radiation oncologist, which may have introduced bias. Patients with various risk types and RT fields were included, giving rise to a heterogeneous population. However, 400 prostate contours were delineated and analyzed together in this study, strengthening statistical power. Moreover, all three sets of axial, coronal, and sagittal images of each patient were used for analysis, allowing an unimpaired assessment.

In conclusion, movements in the SI direction were a major source of variability in definitive prostate RT. Moreover, interobserver variability was a non-negligible source of margin. The optimal PTV margin should also consider the internal margin, particularly in the SI direction. Further studies with dosimetric evaluation should be performed to better assess the PTV margin for definitive prostate RT.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by Gangnam Severance Hospital institutional review board. The studies were conducted in accordance with the local legislation and institutional requirements. Informed consent was waived due to the retrospective study design.

JK: Data curation, Formal analysis, Methodology, Visualization, Writing – original draft, Writing – review & editing. JS: Data curation, Methodology, Writing – review & editing. SJL: Data curation, Writing – review & editing. KSC: Data curation, Writing – review & editing. BHC: Data curation, Writing – review & editing. DY: Data curation, Writing – review & editing. JHK: Methodology, Writing – review & editing. JWK: Conceptualization, Funding acquisition, Project administration, Supervision, Writing – review & editing.