FormalPara Main novel aspects

  • Single-fraction high-dose-rate brachytherapy (SFHDR) for localized prostate cancer shows favorable safety given the low incidence of toxic events.

  • Low-risk patients seem to be suited for SFHDR.

Introduction

Treatment options for definitive radiotherapy of prostate cancer can be performed with commonly used external-beam radiation therapy (EBRT) and low-dose-rate brachytherapy (LDRB). As a safe and effective method of dose escalation, high-dose-rate brachytherapy (HDRB) is also emerging, allowing for highly conformal coverage of the prostate while minimizing the dose to surrounding organs at risk (OAR) [1]. According to the German S3 guidelines, HDRB is recommended in combination with EBRT for the treatment of eligible patients with localized prostate cancer. However, mature results from clinical trials have demonstrated that multi-fraction HDRB as monotherapy in localized prostate cancer can achieve comparable disease control to LDRB and EBRT [2,3,4,5].

Over the past practice, some shortcomings of multi-fraction HDRB have been noticed. Frequent hospitalizations, multiple implants, and bed rest may be required during the treatment process. Due to resource consumption and logistic challenges, the appeal of multi-fraction HDRB tends to be detracted, especially compared to the permanent seed implant of LDRB [6, 7]. In addition, regarding radiobiological considerations, the low α/β ratio and high sensitivity to hypo-fractionated radiotherapy in prostate cancer have led to increased interest in single-fraction HDRB (SFHDR) [8,9,10]. Such an approach may make HDRB favorable to LDRB in that SFHDR reduces the need for multiple implants and is more attractive in terms of practicality, convenience, toxicity [11, 12].

Currently, however, SFHDR is not generally applied to localized prostate cancer patients and there are still controversies regarding the optimal dose regimen and selection of disease risk group. Assuming that SFHDR is a potential curative alternative for localized prostate cancer, we aim to better understand the safety and efficacy of the therapy through this latest systematic review and meta-analysis.

Materials and methods

Search strategy and study selection

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement was applied to this systematic review and meta-analysis [13]. We registered it in the International Prospective Register of Systematic Reviews (PROSPERO) of the Centre for Review and Dissemination (CRD42022319328). The participants, intervention(s), control, outcomes, study design (PICOS) approach [14, 15] was used to set the inclusion criteria for the literature (eTable 1 in the Supplement).

The inclusion criteria for the studies were as follows: 1) patients with localized prostate cancer who were primarily treated with HDRB radiotherapy; 2) prospective or retrospective randomized or non-randomized studies with a single group or multiple groups; 3) all patients in the treatment group received a single dose of HDRB (defined as more than 15 Gy per fraction), with or without androgen deprivation therapy (ADT); 4) at least one major endpoint measure was reported, including gastrointestinal (GI) toxicity, genitourinary (GU) toxicity, 3‑year bRFS, and 5‑year bRFS. Exclusion criteria included 1) non-human experimental studies, 2) patients who received adjuvant radiotherapy or had metastatic prostate cancer or developed disease relapse, 3) non-English articles, and 4) sample size fewer than 15.

Two investigators (M.H.W. and J.D.D.) conducted independent, comprehensive, and systematic searches in PubMed, Embase, and the Cochrane Library databases. The main retrieval strategies included the following different terms: “prostate” and “cancer or neoplasms” and “single or one” and “dose or fraction” and “high dose rate or HDR” and “brachytherapy.” In addition, potentially relevant literature was tracked through articles or reviews to supplement appropriate studies for the meta-analysis. Detailed search strategies are available at PROSPERO according to the CRD number. The searches were reconducted to complement the study before the final analysis and the final search was conducted in October 2022.

Data extraction

Data extraction was performed based on a pre-designed standardized form for the included studies. This process was independently conducted by two reviewers (N.W.X. and J.G.Z.). A third investigator (G.X.S.) was involved in resolving any differences between the two reviewers. Extracted variables included study design, country, time of the study, sample size, dose per fraction, HDRB image guiding method, dose constraints, median follow-up time, and patient characteristics before treatment (such as median age, prostate volume, PSA level, ADT usage, clinical tumor stage, Gleason score, and risk group), toxic events rate (GU and GI toxicity), 3‑year bRFS, and 5‑year bRFS. In the absence of specific bRFS data for articles, Plot Digitizer version 2.6.8 (SourceForge, Boston, Massachusetts, the United States) was used to extract values from Kaplan–Meier curves.

In addition, two researchers (Y.H.Z. and Y.D.X.) recorded the risk of bias assessment results for each study. The revised Cochrane risk-of-bias tool for randomized trials (RoB 2) was used to evaluate randomized trial studies. This consists of five domains, including randomization process, intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result [16]. For non-randomized experimental studies, methodological quality was assessed using the methodological index for non-randomized studies (MINORS) [17].

Outcomes

The primary outcome measure was safety, represented by GU and GI toxic event rates, and the secondary outcome measure was efficacy, represented by bRFS rates. Toxic event rates involve severe and grade 2 toxicity, where severe toxicity was defined as an event greater than or equal to grade 3, primarily based on the Radiation Therapy Oncology Group (RTOG) or the Common Terms for Adverse Events (CTCAE). Any toxic event requiring hospitalization or surgical intervention, or directly reporting “severe” toxicity, was considered as a grade ≥ 3 toxicity. Studies reporting toxicity events that were not distinguished as GU or GI toxicity were excluded. Nevertheless, if the toxic event was zero, both GU and GI toxic effects were included in the analysis at zero.

Efficacy was evaluated by 3‑year bRFS and 5‑year bRFS. Although different articles described biochemical failure in different ways, such as biochemical no evidence of disease (bNED), biochemical disease-free survival (bDFS), biochemical control rate (BCR), biochemical progression-free survival (bPFS), and biochemical failure-free survival (BFFS), they were all based on the Phoenix definition (a rise by 2 ng/mL or higher above the PSA nadir) to define biochemical failure (BF) [18]. Therefore, these outcomes were considered equivalent for bRFS.

Statistical analysis

The proportion rates of patients who experienced particular events were used as the effect measure for each study and were presented in a forest plot with corresponding 95% confidence intervals (CI) and related 95% prediction interval (PI). For any study with an event rate of 0 or 1, a continuity correction of 0.5 was applied. Restricted maximum-likelihood estimator (REML) and the Hartung–Knapp method were also used in the meta-analysis [19]. Due to possible heterogeneity in patient characteristics and study design, random-effects models with the inverse variance method were used to summarize effect measures. Cochran’s Q test and the I2 test were used to determine the heterogeneity between studies [20]. Since the precision of both the Q and the I2 statistic depend on sample size, we provided another estimate, τ, which is the square root of τ2, the between-study variance. An obvious advantage is that its estimates do not increase systematically with the number or size of studies [21]. Publication bias was detected by funnel plot and Egger’s regression test [22]. Subgroup analyses were conducted to assess the effect of different characteristic populations or dosages on the estimates. To explore possible factors associated with the combined effect measure, we performed a meta-regression analysis using pre-set covariables. In addition, sensitivity analysis was used to assess the robustness and reliability of the results.

All above statistical analyses were carried out in R, version 4.1.3. The packages “meta” (version 5.2.0) and “metafor” (version 3.4.0) were applied to carry out the meta-analysis [23, 24]. The R code to run these operations is provided in the eMethods of the Supplement. All tests were bilateral, and the significance level was set at P < 0.05.

Results

Study and patient characteristics

According to the search strategy, 4222 records were identified, of which 931 were duplicate records. Twenty-five studies including 1440 patients fit the inclusion criteria for quantitative analysis after screening full-text articles and abstracts to determine eligibility [6, 7, 10, 25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46]. The PRISMA flow diagram for the recognition and selection of studies is shown in Fig. 1. There were 5 retrospective studies, 14 prospective studies, and 6 randomized trials, all of which were published between 2014 and 2022. Three studies were counted twice because they included two groups using different single-fraction HDRB schemes [6, 7, 32]. Statistically, each treatment group was treated as one study for calculation, which results in a total of 28 studies. The assessment of bias risk results for each included study are listed in eTables 2–3 and eFigure 1 of the Supplement. The characteristics of the studies included in the meta-analysis are summarized in Table 1. Dose constraints for each of the included studies are provided in eTable 4. In these studies, the dose regimens of SFHDR for patients included 19 Gy, 19.5 Gy, 20 Gy, 20.5 Gy, and 21 Gy. The patient-level characteristics of each study are shown in Table 2. In general, the median age of patients was 66.9 years (range 62–73 years) and the median follow-up was 47.5 months (range 12–75 months).

Fig. 1
figure 1

Flow diagram of systematic reviews and meta-analyses based on PRISMA. HDR high-dose-rate

Full size image
Table 1 Summary of primary studies
Full size table
Table 2 Patient characteristics of primary studies
Full size table

Gastrointestinal toxicity

The incidence of grade 3–5 GI toxicity was very low, with only 3 cases reported, including 1 case of diarrhea and 2 cases of rectal fistula requiring colostomy [42, 44]. Based on the weighted random-effects model, the pooled cumulative occurrence was 0.1% (95% CI 0–0.2%; 95% PI 0–0.7%; Fig. 2a). The cumulative incidence of grade 2 GI toxicity was also fairly low, with an estimate of 1.6% (95% CI 0.1–4.7%; 95% PI 0–16.9%; eFigure 2a in the Supplement). Subgroup analysis showed no significant difference in the incidence of severe GI toxicity at different dosages, while statistical significance was found at grade 2 GI toxicity (P = 0.91 and P = 0.02; eFigure 3a and eFigure 3b; respectively). Note, however, that the incidence of grade 2 GI in the highest dosage subgroup was zero. eFigure 4 showed the estimated incidence of grade 3–5 and grade 2 GI toxic effects at different timepoints after receiving SFHDR, reaching the highest values at 6 months and 1 month postoperatively, respectively (0.03%, 95% CI 0–0.13%; 0.24%, 95% CI 0–1.04%).

Fig. 2
figure 2

Safety. a Based on the weighted random-effects model with inverse variance method, severe GI toxic effects for 17 individual groups were estimated. Egger’s regression test indicated the presence of publication bias (P < 0.01). b A weighted random-effects model with inverse variance method was applied to estimate the severe GU toxic effects for 17 individual groups. Egger’s regression test indicated the absence of publication bias (P = 0.48). DIL dominant intraprostatic lesion, fx fraction, GI gastrointestinal, GU genitourinary

Full size image

Genitourinary toxicity

Severe GU toxicity was also relatively rare, mainly manifested as acute hematuria, urethral stricture, and urinary retention [7, 26, 27, 31, 34, 44]. Based on the weighted random-effects model, the pooled cumulative occurrence was 0.4% (95% CI 0–1.2%; 95% PI 0–3.9%; Fig. 2b). For cumulative grade 2 GU toxicity, the pooled incidence was 17.1% (95% CI 5.4–33.5%; 95% PI 0–75.8%; eFigure 2b in the Supplement). Statistical differences were detected at different dosage subgroups for severe and grade 2 GU toxicity (P = 0.03 and P < 0.01; eFigure 3c and eFigure 3d; respectively). Again, note that the incidence of grade 2 GU in the highest dose subgroup was zero. The estimated incidence of grade 3–5 and grade 2 GU toxic events after SFHDR treatment peaked at 48 and 36 months, respectively (0.09%, 95% CI 0–0.59%; 6.31%, 95% CI 0.84–16.33%; respectively; eFigure 4). For further details of HDRB therapeutic schemes, grade 3–5 and grade 2 toxic effects were presented in Table 3 and eTable 5.

Table 3 Details of severe toxic effects and HDRB therapeutic schemes
Full size table

Three-year biochemical recurrence-free survival

Data from 17 treatment groups of 16 studies were included, with 3‑year bRFS ranging from 70.4% to 95.8% [7, 25, 26, 30, 31, 34, 37,38,39,40,41,42,43,44,45,46]. Based on the weighted random-effects model, the pooled rate was 87.5% (95% CI 84.4–90.3%; 95% PI 78.7–94.2%; Fig. 3a). Sensitivity analysis showed that the estimated value remained stable by eliminating studies one by one (eTable 5a). Statistical difference was found in different risk stratifications, while it was not detected under different dosage subgroups (P < 0.01; P = 0.33; Fig. 4a and eFigure 3e; respectively). Moreover, a statistically significant difference was also found between favorable intermediate-risk (FIR) and unfavorable intermediate-risk (UIR) subgroups (P = 0.01). Results of multivariable meta-regression analysis showed that no significant association with 3‑year bRFS was found for three predictors (P = 0.08), including ADT receipt, proportion of Gleason score ≥ 7, and proportion of clinical T stage ≥ T2b (P = 0.12, P = 0.05, and P = 0.36, respectively; eTable 6a in the Supplement).

Fig. 3
figure 3

Clinical benefit. a Based on the weighted random-effects model with inverse variance method, 3‑year bRFS rates for 17 individual groups were estimated. Egger’s regression test indicated the absence of publication bias (P = 0.63). b A weighted random-effects model with inverse variance method was applied to estimate the 5‑year bRFS for 11 individual groups. Egger’s regression test indicated the absence of publication bias (P = 0.46). In the study from Armstrong et al., a single 21 Gy dose was applied to the DIL of patients and two de-escalation prescription schedules based on V19Gy (percentage of volume receiving a dose of 19 Gy) for the remaining planning target volume (PTVnon-boost). bRFS biochemical recurrence-free survival, fx fraction, DIL dominant intraprostatic lesion

Full size image
Fig. 4
figure 4

Subgroup analyses based on risk group of bRFS. a Subgroup analysis forest plot of 3‑year bRFS. b Subgroup analysis forest plot of 5‑year bRFS. Studies with unavailable data were not included in the subgroup analyses. bRFS biochemical recurrence-free survival, fx fraction, DIL dominant intraprostatic lesion

Full size image

Five-year bRFS

Data from 11 treatment groups of 9 studies were included, with 5‑year bRFS ranging from 40.2% (actuarial 5‑year biochemical failure rate was 59.8% from the manuscript) to 88.0% [6, 7, 28, 30, 39, 40, 42, 43, 45]. The pooled rate was 71.0% (95% CI 63.0–78.3%; 95% PI 47.9–89.4%; Fig. 3b). Sensitivity analysis showed favorable robustness of the estimate (eFigure 5b). Statistical difference was detected at different risk stratifications and was not found at dosage subgroups (P = 0.04; P = 0.06; Fig. 4b and eFigure 3f; respectively). Only one study reported 5‑year bRFS of FIR and UIR subgroups, and no statistically significant difference was detected (P = 0.64) [6]. Results from multivariable meta-regression analysis showed that no significant association with 5‑year bRFS was found in the three predictors (P = 0.51), involving ADT receipt, proportion of Gleason score ≥ 7, and the proportion of clinical T stage ≥ T2b (P = 0.32; P = 0.89; P = 0.88, respectively; eTable 6b in the Supplement).

Discussion

The safety and efficacy of multi-fraction HDRB as monotherapy has been widely established, with studies reporting low toxicity rates and excellent bRFS rates [2, 47, 48]. However, due to the challenges of multi-fraction treatments such as resource utilization, cost control, and patient convenience, studies of SFHDR in localized prostate cancer are gradually emerging. To our knowledge, this is the first meta-analysis of the safety and efficacy associated with SFHDR for localized prostate cancer.

The results of our study provide evidence supporting the favorable tolerability of SFHDR, with cumulative severe GI and GU toxic estimates of 0.1% and 0.4%, respectively, and the corresponding grade 2 toxicity estimated at 1.6% and 17.1%, respectively. It is encouraging that a single fraction of HDRB can be comparable to fractionated HDRB, LDRB, stereotactic body radiotherapy (SBRT), and EBRT in terms of toxic effects [5, 49,50,51]. Prospective study results showed that the cumulative incidence of grade 2 or worse GI and GU was 6% and 33% in multi-fraction HDRB regimens [31]. A randomized study found no significant difference in GI and GU toxicity between SFHDR and LDRB [41]. The rate of ≥ grade 2 GI toxicity of SFHDR was found to be significantly lower than for fractionated SBRT (P < 0.05) [45]. Moreover, The CHHiP trial reported rates of 49% and 38% regarding acute ≥ grade 2 bladder and bowel toxicity after hypofractionated radiotherapy, respectively, while the crude estimates of our study remained below 20% or even lower at different timepoints after SFHDR [49]. Cumulative late ≥ grade 2 GI and GU toxicity after EBRT was 12% and 23% in the standard arm of the FLAME trial, indicating more frequent toxic events than with SFHDR [52]. In comparison, our findings suggest that the current single-fraction dosages of HDRB are acceptable, and the process of these studies is not significantly restricted by toxicity. Notably, some severe toxic effects may develop from grade 1 or 2, so early procedural intervention and close care are required to reduce the occurrence of increased toxicity. Further increase of the dose of a single fraction seems feasible, but the association between toxicity and clinical benefit should be fully considered.

Radiation-related toxicity can be prevented or mitigated in several ways. Hyaluronic acid injection through the perineum into the perirectal fat of patients could keep the rectal wall away from the radiation source and could help to avoid acute and late toxicity during the entire process, even when a single dose of 20.5 Gy HDRB was applied to the planning target volume (PTV) of the prostate [39, 40]. A recent meta-analysis evaluating seven studies showed that placement of a perirectal hydrogel spacer could significantly reduce rectal toxicity and improve bowel-related quality of life [53]. The application of these methods may facilitate further SFHDR studies under good control of radiation-related toxicity.

In terms of efficacy, a single dose of HDRB provided suboptimal biochemical control, with the estimate of 87.5% for 3‑year bRFS. Note that based on the principle of radiobiology, 19 Gy was already a not low dosage, which can achieve a similar biologically equivalent dose (BED) to 2 × 13 Gy and 3 × 10.5 Gy of HDRB for prostate cancer with an α/β ratio hypothesized to be 1.5 [8]. One prospective study demonstrated similar clinical benefits under these regimens with 4‑year bRFS approaching 90% (P > 0.05) [31]. Besides, SFHDR appeared to have comparable disease control effects with SBRT, given that a single 24 Gy SBRT achieved a 3-year bRFS of 86.3% and 4‑year bRFS of 77.1% [54]. In spite of this, patient selection may need to be well considered for SFHDR, with 3‑bRFS estimates of 99.0% and 5‑year bRFS estimates of 80.9% in low-risk patients, both of which were found to be statistically significant in risk stratification by bRFS (P < 0.05).

Nevertheless, in contrast, there was still a certain gap to other well-known experimental results. The POP-RT trial reported a 5-year BFFS of 81.2% (95% CI 71.6–87.8%) after 68 Gy in 25 fractions of prostate intensity-modulated radiotherapy (IMRT), but for high-risk patients with minimum 2 years of ADT [55]. The FLAME trial results showed that patients treated with 77 Gy in 35 fractions of EBRT acquired a 5-year bDFS of 85% (95% CI 80–89%) [52]. Besides, an estimated 5‑year failure-free survival of 84% (95% CI 80–87%) was reported from the HYPO-RT-PC study, with more than 80% intermediate-risk patients [56]. Current opinion supports the notion that most biochemical failures of SFHDR were local failures and occurred primarily at the site of initial disease [28, 37, 57]. This pattern of relapse provided researchers with a new therapeutic approach, i.e., to add a local boost to the lesion of interest based on SFHDR, although the studies failed to achieve the desired clinical benefit [6, 7, 25]. There are some possible radiobiological explanations. Because of the complex biochemical mechanism of malignant tumors, the very flexible DNA damage repair ability of tumor cells, especially cancer stem cells (CSCs), remains a concern, although the therapeutic effects can be observed at the macro level [58]. The absence of reoxygenation during SFHDR treatment may reduce the sensitivity of the tumor to radiation [59]. Moreover, due to the heterogeneity of tumors, some cancer cells with a higher α/β ratio may have relative resistance to single-fraction radiotherapy, which means that tumor cells cannot be sufficiently suppressed or killed [7].

Recently, investigators found a dose-response relationship in patients treated with SFHDR and favorable long-term disease control may be obtained by further increasing the dose of SFHDR [28, 40, 60]. The 5‑year bRFS at a dose up to 20.5 Gy reached 82.0%, better than the estimate of our study [40]. However, no significant correlation of dosage with bRFS was found for 19–21 Gy, possibly due to the narrow dose range that was not powered to detect the presence of a difference. Clinical trials at a dose up to 23 Gy and 25 Gy (NCT03424850) are underway, and the mature results will help us to better understand the toxicity and efficacy of SFHDR.

Although we conducted this meta-analysis strictly and carefully, some limitations should be recognized. First, due to the inconsistencies of control groups or the absence of a control group, our study only combined the incidence of a single treatment group of included studies, so a cautious attitude should be taken to evaluating the outcome. Second, our median follow-up was no more than 50 months, and the long-term follow-up data are still limited. Besides, as data from several studies are unavailable, the reliability of our subgroup analyses and meta-regression models needs to be further validated by including larger-scale studies.

Conclusion

Overall, SFHDR is well tolerated and associated with suboptimal clinical benefit in patients with localized prostate cancer. High-quality prospective studies of SFHDR are necessary to verify its safety and efficacy.