Comparison of toxicities between ultrahypofractionated radiotherapy versus brachytherapy with or without external beam radiotherapy for clinically localized prostate cancer

To compare gastrointestinal (GI) and genitourinary (GU) toxicities in patients with localized prostate cancer treated with ultrahypofractionated radiotherapy (UHF) or brachytherapy [BT; low dose rate, LDR or high dose rate (HDR) with or without external beam radiotherapy (EBRT)]. We compared 253 UHF and 1664 BT ± EBRT groups. The main outcomes were the incidence and severity of acute and late GU and GI toxicities. The secondary endpoint was biochemical control rate. Cumulative late actuarial GU toxicity did not differ for grade ≥ 2 (8.6% at 5-years in UHF and 13.3% in BT ± EBRT, hazard ratio [HR], 0.7066; 95% CI, 0.4093–1.22, p = 0.2127). Actuarial grade ≥ 2 late GI toxicity was higher in UHF (5.8% at 5-years, HR: 3.619; 95% CI, 1.774–7.383, p < 0.001) than in BT ± EBRT (1.1%). In detailed subgroup analyses, the high-dose UHF group (H-UHF) using BED ≥ 226 Gy1.5, showed higher GI toxicity profiles than the other subgroups (HDR + EBRT, LDR + EBRT, and LDR monotherapy, and L-UHF BED < 226 Gy1.5) with equivalent GU toxicity to other modalities. With a median follow-up period of 32 months and 75 months, the actuarial biochemical control rates were equivalent between the UHF and BT ± EBRT groups. UHF showed equivalent efficacy, higher GI and equivalent GU accumulated toxicity to BT ± EBRT, and the toxicity of UHF was largely dependent on the UHF schedule.

HDR-BT with EBRT. The multi-institution data were obtained from an open data source 18 , and the detailed method of applicator implantation has been described elsewhere 21 . All patients were treated with a combination of HDR and EBRT at various fractionations ( Table 2). The median dose of HDR used was 31.5 Gy (10.5-31.5 Gy) and that of EBRT was 30 Gy (30-51 Gy). The median fraction size of HDR was 6.3 Gy (5-11 Gy) and that of EBRT was 3 Gy (2-3 Gy). Patients who were administered EBRT comprised 1166 (98.2%) on 3D-CRT and 21 (1.8%) on IMRT.
UHF. The detailed method of this study has been described elsewhere 16,22 . The median dose of UHF used was 36 Gy (32-36.25 Gy) and the median fraction size of UHF was 7.25 Gy (7-9 Gy) ( Table 2).
Statistical analysis. The R stat package 23 was used for the statistical analyses. We analyzed percentages using chi-square tests. To compare medians or means, we used Mann-Whitney U-tests for skewed data and Student's t-tests for normally distributed data 23 . To analyze the biochemical control rate, overall survival, and toxicity, we used Kaplan-Meier method and log-rank tests including Bonferroni test in pos t-hoc p-value adjustment was used 23 . Univariate and multivariate analyses were made with Cox's proportional hazards model 23 . All analysis used statistical significance level set at p < 0.05. We divided the UHF group into two subgroups according to previous studies 16,17 : high (H-UHF) and low dose UHF (L-UHF) groups, using a cut-off value of BED of 226 Gy 1.5 ; BED = n × d × (1 + d/[α/β]) where d = dose per fraction in Gy, n = number of treatment fractions, α/β = 1.5.
Since the included patients were not randomized, unbalanced patients baseline characteristics could influence on the selection bias and, hence, influence the decision to undergo BT ± EBRT or BT. The propensity score was defined as the probability of allocation to the BT ± EBRT or UHF group, given the patient characteristics 23 . We used logistic regression model in the calculation of the propensity scores using the baseline covariates shown in Table 2.
We used a propensity score-matched pair analysis to reduce the bias for choice of treatment; the UHF or BT ± EBRT groups (total population and HDR + EBRT group). Five factors prescribed before were selected as www.nature.com/scientificreports/ the variables that would be significantly related to the decision to choose UHF or BT ± EBRT, and a 1:1 matched cohort was made. Same procedure was applied in comparison between UHF and HDR + EBTT.

Results
Patient and tumor characteristics. The baseline patient characteristics of the UHF and BT ± EBRT groups are shown in Table 1. The 1921 patients with stage T1-T3 N0M0 prostate cancers were treated using UHF or BT ± EBRT. The median patient age was 70 years (range, 42-86 years). The median follow-up duration for the entire cohort was 70 months (range, 22-177 months). BT ± EBRT was used to treat patients with advanced disease and hormonal therapy history with longer follow-up periods than those in the UHF group. Table 3 shows the incidence of maximal grade of early and late gastrointestinal (GI) and genitourinary (GU) toxicities. UHF showed higher maximal grade GI and lower maximal grade GU toxicity than the BT ± EBRT group. The 3-(and 5-year) cumulative incidence of grade ≥ 2 GI toxicities was 4.2% (5.8%) in the UHF group and 1.1% (1.8%) in the BT ± EBRT group (p < 0.0001; Fig. 2a), with a hazard ratio of 3.661 (95% CI: 1.799-7.454, p < 0.0001).

Discussion
UHF showed higher GI and equivalent GU toxicity to BT ± EBRT and was largely dependent on the UHF schedule. Additionally, we found an equivalent PSA control rate between UHF and BT ± EBRT, although this was inconclusive due to short follow-up periods. To our knowledge, this is one of the largest cohorts to compare the toxicity of UHF and BT ± EBRT. To reduce bias and amend short follow-up periods, we used the propensity score matched pair analysis, which is the best achievable statistical method and provides a direct comparison of BT ± EBRT and UHF. Recent advancements in radiotherapy for localized prostate cancer have enabled us to shorten the treatment period using hypofractionations and provide cost effectiveness and patient convenience. In addition to 2.3-3.4 Gy moderate hypofractionation, UHF gained attention for exploiting the low a/b ratio of this tumor and its high radiation fraction size sensitivity 1-6 . The recent HYPO-RT-PC phase 3 trial, which showed non-inferiority of ultrahypofractionation (42.7 Gy/7 fractions for 2.5 weeks) compared with conventional fractionation (78 Gy/39 fractions) 2 . It is anticipated that the efficacy of the UHF treatment schedule will be further validated when the PACE B trial outcome is consolidated and published 24 . Similarly, within our cohort of patients, an excellent biochemical control rate was achieved, which is comparable to HDR ± EBRT, although preliminary.
The HYPO-RT-PC phase 3 trial 2 reported 28% acute RTOG grade ≥ 2 GU toxicity, and grade ≥ 2 RTOG late GU toxicity was 5% at 5 years, while bowel toxicity was 1% at 5-years. The PACE-B trial reported that the worst acute RTOG toxicity grade ≥ 2 was 23% in GU and 10% in GI 24 . In our UHF data, the worst acute toxicity grade ≥ 2 was 13% in GU and 5% in GI, and accumulated late toxicity grade ≥ 2 was 6% in GU and 5.8% in GI, which concurred with their data. Jackson  www.nature.com/scientificreports/ grade ≥ 3 GU and GI toxicity rates were 2.0% (95% CI, 1.4-2.8%) and 1.1% (95% CI, 0.6-2.0%) after UHF using SBRT, respectively 4 , which also concurred with our cohort. In general, BT elevated GU toxicity and reduced GI toxicity compared to EBRT 7 . In addition, although the incidence of acute GU toxicity is tentatively elevated by BT, toxicity was ameliorated by time and cumulative late toxicity did not differ after a few years 7 . For GI toxicity, spacer (SpaceOAR etc.) insertion was found to reduce GI toxicity to almost negligible as no grade ≥ 2 GI events was found in spacer ( +) arms in a randomized controlled trial 25 . This technique could be applied not only in UHF but also in BT ± EBRT. Therefore, we hope that we could reduce GI toxicity in the near future, and the higher incidence of GI toxicity in H-UHF could be reduced with this technique.
As BT can achieve one of the best dose distributions among radiotherapy 7,26 , external beam radiotherapy has made efforts to improve dose distribution using SBRT, intensity-modulated radiotherapy, and image-guided radiotherapy techniques 26 . Several reviews 27,28 including three randomized controlled trials [8][9][10] have already indicated superiority of BT boost than eternal beam radiotherapy alone. However, BT boost did not show superiority to UHF, and only indicated similarity of BT boost to UHF in low to intermediate risk groups [11][12][13][14][15] . So far, UHF could achieve equivocal outcomes without elevation of toxicity than BT boost in low-to intermediate-risk groups.
In However, this does not apply for high risk and, most likely, a higher dose is needed 3,31 . Several groups seek better PSA control using higher prescribed doses, especially for intermediate-and high-risk groups 31,32 . In patients with high-risk disease, Royce et al. found that an EQD2 of 97 Gy (37.6 Gy in 5 fractions = 226 Gy1 .5 ) can achieve a TCP of 90% and an EQD2 of 102 Gy (38.7 Gy in 5 fractions = 238.4 Gy 1.5 ) can achieve a TCP of 95% 3 . Several studies used focal dose escalation with a boost of 38-50 Gy 31,32 . Although our cohort did not show the benefit of H-UHF (BED = 252 Gy 1.5 , with higher GI toxicity without improvement in biochemical control rate [88.3% at 3-years], although with short follow-up periods), further investigation could shed light on the dose escalation for high-risk prostate cancer.
Our study has several limitations. First, the lack of long-term follow-up and the small sample size limits its applicability, with only 25 (9.8%) patients with > 5 years of follow-up in the UHF group, especially in the high-risk group. Longer follow-up may reveal a divergence in toxicity or control rates in the UHF group. Next, the retrospective nature of this study led to an imbalance between the UHF and BT ± EBRT cohorts in terms of baseline characteristics. To mitigate this, we provided a comparative analysis and propensity score-matched analysis. Next, although using a free database is beneficial, its retrospective nature results in an ambiguous recording of the timing of toxicity and tumor control outcomes because of the heterogeneous follow-up periods depending on various institutions and physicians not restricted by protocol. Further studies should be conducted to validate our findings. Finally, for toxicity analysis, other predisposing factors are also important for prediction, including dosimetric factors for organs at risk 33 and non-dosimetric factors (preexisting symptoms or surgery, transurethral resection of the prostate, anticoagulant use, diabetes mellitus, etc.) 33 .

Conclusions
UHF showed equivalent efficacy, higher GI and equivalent GU accumulated toxicity to BT ± EBRT, and the toxicity of UHF was largely dependent on the UHF schedule.

Data availability
The data of UHF for this manuscript can be obtained from the public data base on reasonable request [19] and LDR data can be obtained from the author upon reasonable request.