This prospective clinical study to assess treatment of prostate cancer with 36.25-40 Gy in five fractions delivered on 1.5-T MR-linac demonstrated good patient tolerability of online ATS workflow, with moderate and transient acute GU toxicities and very mild acute GI toxicities.
With online adaptive contouring and planning, SBRT delivered on MR-linac is noninvasive, avoiding need for fiducial marker insertion; however, it is time and cost intensive. The time-consuming steps of ATS workflow include MR scan acquisition, online target contouring, online plan re-optimization, and treatment delivery, with mean times of 6 minutes, 10.5 minutes, 10.5 minutes, and 10.5 minutes, respectively, as per our data. The median on-couch time was 51 (24–95) minutes, which is close to the 53 (34–86) minutes reported by Alongi et al. , although the fractionation dose was higher in our study. The 1.5-T MR-linac offers two prespecified online MR scan sequences for prostate cancer: 0.6-mm slice thickness for 6-minute scan and 1-mm slice thickness for 2-minute scan. Since the resolution of the 2-minute scan was good enough for online registration and contour propagation, we switched to the 2-minute scan after the first 20 patients. This decreased the overall median on-couch time from 55 minutes to 39 minutes—a clinically significant reduction welcomed by the patients. Consistent with other series , around 9% of sessions needed another ATS or ATP workflow due to technical problems or organ motions. Although it is difficult to complete one fraction within 30 minutes with the current workflow, self-referrals are common with this noninvasive procedure and high-precision MRgRT, as was demonstrated by the high proportion of patients (98.4%) expressing satisfaction with the treatment (only two fractions were scored as “unsatisfactory or intolerable”).
Completion rate of CROM and PROM forms was almost 100%. A cumulative 33.3% and 8.3% of patients presented with grade 2 or above GU and GI toxic effects by CTCAE criteria, and 33.3% and 0% by the RTOG scale. Only two patients had RTOG grade 3 acute GU toxicity (urinary frequency, urgency, and hourly nocturia). To date, only small series on the acute toxicities of MRgRT have been published. There are marked differences in previous reports of toxicity rates. Bruynzeel et al.  and Ugurluer et al.  enrolled, respectively, 101 and 50 patients treated on MRIdian with 0.35-T MR and online daily adaptive planning. Bruynzeel et al.  reported 23.8% and 5% patients with grade 2 or worse GU and GI CTCAE toxicity, respectively, compared with 36% and 0% reported by Ugurluer et al. . Alongi et al. used 1.5-T MR-linac and reported grade 2 GU and GI acute toxicities in only 12% and 4% patients, respectively. It is difficult to compare our results directly with other studies investigating acute toxicity of MRgRT for several reasons. First, the prescription dose and normalization definition were different. The optimal SBRT schedule for prostate cancer is still unclear. NCCN guidelines, which are based on the evidence from PACE-B  and HYPO-RT-PC , recommend 36.25-40 Gy in five fractions or 42.7 Gy in seven fractions. We adopted a regimen of 36.25 Gy to PTV, with concomitant boost of 40 Gy to prostate, which is similar to PACE-B, and encouraged 100% prescription dose normalized to 95% PTV volume, which meant 36.25 Gy to 95% PTV and 40 Gy to 95% CTV4000. The margin was 3 mm in the studies by Bruynzeel et al.  and Ugurluer et al. , and 5 mm in the study by Alongi et al. , all of which assumed that 95% prescription dose received by PTV was acceptable without boost dose within prostate; this means that 34.4 Gy was mandatory for PTV in the study by Bryunzeel et al. , whereas only 33.2 Gy was required in the study by Alongi et al. . The higher dose in our study may explain the higher frequency of grade 2 or above GU toxicity in our cohort. Second, we did not use urethra sparing as in the other two studies, because it was hard to distinguish urethra online without an indwelling catheter. Even so, with Dmean of urethra lower than 42 Gy, most urinary symptoms resolved at 4 weeks post radiotherapy. Finally, we gathered data much more frequently: weekly during SBRT, at end of radiotherapy, two-weekly thereafter up to 8 weeks, and then at 12 weeks. In comparison, Alongi et al.  only collected toxicity data at baseline and end of treatment, Ugurluer et al.  at baseline and at 3 months after SBRT, and Bruynzeel et al.  at 6-week follow-up. As our study (Figs. 2 and 3) and PACE-B  show, the incidence of acute side effects peak at around 2 weeks after completion of SBRT and then declines slowly to baseline levels at around 4–6 weeks post radiotherapy. This change in toxicities over time was reflected in the patient-reported QoL analysis in our study. More frequent assessments, along with the nearly 100% completion of questionnaires, make our findings more reliable.
With prescription dose and the follow-up schedule similar to PACE-B , we found comparable acute grade 2 or above GU side effects (33.3% in our study vs. 30.8% in PACE-B) and less grade 2 or above GI toxicities (8.3% in our study vs. 15.7% in PACE-B). Actually, baseline urinary function was worse in our cohort, with 23% patients on α-blockers, 58% having prostate volume > 40 cc, and 46% presenting with moderate IPSS. Furthermore, we enrolled more intermediate-risk patients and high-risk patients, and 58% of our patients had proximal SV included in the PTV. Despite the less favorable patient profile, no grade 2 or above RTOG acute GI toxicities occurred in our study, as compared to the 10% reported by PACE-B . The interfraction variations in prostate and adjacent organ positions can affect the dose received by the target and OARs. Peng et al.  compared 486 CT scans from CT-on-rail of 20 prostate cancer patients (treated with conventionally fractionated radiotherapy) with their planning CT scans, and found that the actual dose received by rectum was higher than the planned dose, with increase of V45Gy by > 15% in 5.6% of fractions and increase of V70Gy by > 5% in 11.6% fractions. If intrafraction motions are considered, the dose variations of rectum might be larger. Moreover, Gunnlaugsson et al.  demonstrated significant prostate swelling during UHF-RT of prostate cancer, which can only be covered by adaptive radiotherapy with tight margin. Degree of reduction of rectal toxicity varies in different MRgRT series: Bruynzeel et al.  reported reduction of grade 2 and above GI toxicity by 5%, Alongi et al.  reported reduction of 4%, and Ugurluer et al. reported reduction of 0% . With daily online adaptive contouring and plan re-optimization it is possible to correct for any target or OAR deformation and motion, and deliver high-precision SBRT; this is probably the main reason for the lower incidence of rectal side effects.
Another issue with adaptive MRgRT is the intrafraction motion and the long on-couch time. Keizer et al.  showed that a margin of 5 mm is sufficient and might be further decreased. Our real-time dose analysis (unpublished data) also demonstrated relatively good target coverage during delivery with CTV-PTV margin of 3 mm, with mean of 98.5% ± 1.8% CTV receiving 36.25 Gy and 92.1% ± 0.6% CTV4000 receiving 40 Gy. With intrafraction plan adaptation by cine-MR tracking, it may be possible to safely decrease the margin even further [18, 22].
The main limitations of this study are the small sample size and the relatively short follow-up.