The objective of this study was to report the rates of disease-free survival (DFS), cause-specific survival (CSS), and overall survival after low-dose-rate (LDR) prostate brachytherapy (PB).
The objective of this study was to report the rates of disease-free survival (DFS), cause-specific survival (CSS), and overall survival after low-dose-rate (LDR) prostate brachytherapy (PB).
Data from 1006 consecutive patients with prostate cancer who received LDR-PB and underwent implantation on or before October 23, 2003 were extracted from a prospective database on November 11, 2011. The selected patients had low-risk (58%) or intermediate-risk (42%) disease according to National Comprehensive Cancer Network criteria. The Phoenix threshold was used to define biochemical relapse. Sixty-five percent of patients received 3 months of neoadjuvant androgen-deprivation therapy (ADT) and 3 months of concomitant ADT. Univariate and multivariate analyses are reported in relation to patient, tumor, and treatment variables.
The median follow-up was 7.5 years. By using Fine and Gray competing risks analysis, the 5-year and 10-year actuarial DFS rates were 96.7% (95% confidence interval, 95.2%-97.7%) and 94.1% (95% confidence interval, 92%-95.6%), respectively. When applied to the whole cohort, none of the usual prognostic variables, including dose metrics, were correlated with DFS. However, in both univariate and multivariate models, increasing dose was the only covariate that correlated with improved DFS for the subset of men (N = 348) who did not receive ADT (P = .043). The actuarial 10-year CSS rate was 99.1% (95% confidence interval, 97.3%-99.7%). The overall survival rate was 93.8% at 5 years (95% confidence interval, 92%-95.1%) and 83.5% at 10 years (95% confidence interval, 79.8%-86.6%). Only age at implantation (P = .0001) was correlated with overall survival in multivariate analysis.
In a consecutive cohort of 1006 men with National Comprehensive Cancer Network low-risk and intermediate-risk prostate cancer, the actuarial rate of recurrent disease after LDR-PB was approximately 3% at 5 years and 6% at 10 years. Cancer 2013. © 2012 American Cancer Society.
Comparative health outcomes research can help inform public debate, as well as individual decision makers, by revealing the real-world utility of medical interventions; however, it requires large sample sizes, adequate follow-up, and patients whose demographics and pretreatment prognostic factors are known and relevant to those at risk.1 Comprehensive population-based outcomes databases are ideal resources for this type of research. Herein, we describe the results from such a database in which patients with low-risk and intermediate-risk prostate cancer were treated in a universal access, publicly funded, low-dose-rate (LDR) prostate brachytherapy (PB) program founded in 1997 by radiation oncologists and medical physicists at the British Columbia Cancer Agency (BCCA).
The BCCA LDR-PB program is provincial in scope, with multiple practitioners working in 5 regional cancer centers using uniform, but evolving, evidence-based eligibility criteria and treatment protocols based on the Seattle “preplan” experience combined with a manual planning algorithm that was developed in-house. From its onset, the program incorporated comprehensive prospective data entry, conducted regular peer-review and quality-assurance activities, and later developed formal training and certification procedures.2
In a previous article,3 we presented the disease-free survival (DFS) and overall survival (OS) endpoints for the same cohort. Although excellent outcomes were reported in that study, the median follow-up was relatively short at 4.5 years from the date of implantation. In this update, the median follow-up was 7.5 years (maximum, 13 years).
The BCCA PB database contains patient, tumor, treatment, and outcomes data on more than 4000 consecutive LDR-PB implantations. The current analysis consists of all patients who received LDR-PB on or before October 23, 2003 (N = 1006). All patients received 0.424 U iodine-125 (NIST99, model 6711; Oncura, Arlington Heights, Ill) prescribed to 144 gray (Gy) as a minimum peripheral dose. RAPIDStrand (Oncura) was used for all peripheral needles, incorporating the majority of seeds (>85%) in 679 of implantations (67.5%); the remainder employed exclusively loose sources, a practice that was discontinued after 361 implantations. No patients received supplemental external-beam radiation therapy (EBRT).
Although 42% of this cohort had intermediate-risk disease according to National Comprehensive Cancer Network (NCCN) stratification guidelines, at the time these implantations were performed, our eligibility criteria excluded men with pretreatment prostate-specific antigen (iPSA) levels >15 ng/mL as well as men with both an iPSA >10 ng/mL and a Gleason sum of 7. Moreover, all men with either a Gleason sum of 7 or an iPSA >10 ng/mL received 3 months of neoadjuvant androgen-deprivation therapy (ADT) and 3 months of concomitant ADT, as previously described.3, 4 In addition, 272 of 586 low-risk patients (46.4%) received the same regimen of ADT either for size reduction or because they had started on ADT before referral to radiation oncology. For the sake of consistency, when possible, the latter group received the same regimen of neoadjuvant and concomitant ADT.
All patients were offered cancer center-based follow-up with clinic visits at 6 weeks, every 6 months for 4 or 5 years, and every 12 to 24 months thereafter. Nevertheless, according to protocol, PSA and testosterone values continued to be measured and recorded every 6 months.
The “nadir + 2 ng/mL” PSA threshold (Phoenix definition) was use to define biochemical relapse. DFS was defined as the absence of biochemical, clinical, histologic, or imaging evidence of recurrent or persistent prostate cancer and not having received any secondary treatment for prostate cancer at any time after implantation. Like in our previous reports, patients with early rises in PSA (PSA “bounces”) that triggered the Phoenix definition were not considered relapses in the final analysis if subsequent PSA values declined to <0.5 ng/mL without intervention.3, 4
Mortality data were derived from the provincial death registry. Patients who died were scored as “dead of/with disease” (DO/WD) if they were not disease-free (as defined above) on the date of death regardless of the cause(s) of death listed in the registry. Note that all deaths that were categorized as DO/WD were recorded as events in estimating cause-specific survival (CSS); this method of scoring CSS tends to overestimate cause-specific mortality, but it insures that disease-related mortality is not underestimated. Patients who died and were considered “disease-free” (as defined above) at the time of death were scored “dead no evidence of disease” (DNED) and were censored at date of death for actuarial analyses of DFS and CSS.
Actuarial DFS and CSS rates and their respective 95% confidence intervals (CIs) were estimated using the competing risks method of Fine and Gray.5 Actuarial rates of OS ware estimated using the Kaplan-Meier method.
The Fine and Gray method5 also was used in univariate analysis for DFS. The following variables were included: patient age at implantation, NCCN risk group, Gleason sum, iPSA (with log10 transform to obtain a more Gaussian distribution of iPSA values), receipt of ADT, clinical stage, the percentage of positive cores (PPC), the consecutive order of implantation, the minimum dose received by 90% of the postimplantation prostate volume (D90), and the percentage of the postimplantation prostate volume that received ≥144 Gy (V100).
A multivariate Fine and Gray model (backward:conditional) was used for DFS. The model included the following categorical variables: receipt of ADT, Gleason sum, clinical stage, and the following continuous variables: age at implantation, consecutive implantation number, log10 iPSA, PPC, and D90 or V100 (the dose metrics were tested separately, because they are strongly correlated). For OS, multivariate analysis using a Cox model (backward:conditional) included the same variables with the addition of disease recurrence. Competing risk analyses of DFS and CSS were performed using the R 2.11.1 statistical package (R Foundation for Statistical Computing, Vienna, Austria; available at: http://www.r-project.org [accessed May 2012] ), and the OS analysis was performed using the SAS statistical software package (SAS version 9.3; SAS Institute Inc., Cary, NC).
Forty-nine men (4.9%) have developed recurrent disease, resulting in actuarial DFS estimates of 96.7% at 5 years (95% CI, 95.2%-97.7%) and 94.1% at 10 years (95% CI, 92%-95.6%) (Table 1, Fig. 1). At the time of data extraction, the median follow-up was 7.5 years; 392 men (39%) had been followed for ≥8 years, and 145 men (14%) had been followed for ≥10 years. With 1 exception (see Patterns of Failure, below), all patients who developed clinical, pathologic, or imaging evidence of recurrent cancer also had antecedent or simultaneous biochemical recurrence using the Phoenix definition. It is noteworthy that post-treatment PSA values were exceptionally low by EBRT standards; for example, at follow-up times ≥96 months, the median PSA for nonrelapsing patients was 0.02 ng/mL (mean, 0.05 ng/mL; range, 0.01-0.3 ng/mL).
|Oncologic Endpoint, %|
|Actuarial Time Point||DFS (95% CI)ab||CSS (95% CI)ac||OS (92-95.6)d|
|5 Years||96.7 (95.2-97.7)||99.8 (99.1-99.9)||93.8 (92-95.2)|
|7 Years||95.1 (93.3-96.4)||99.8 (99.1-99.9)||90.1 (87.8-91.9)|
|10 Years||94.1 (92-95.6)||99.1 (97.3-99.7)||83.5 (79.8-86.6)|
Table 2 lists the prognostic factors and dose metrics for the entire cohort and compares the ADT subgroup (N = 658) with the non-ADT subgroup (N = 348). Despite having significantly worse prognostic factors (higher iPSA, Gleason sum, and PPC) as well as significantly lower D90 and V100 values (Table 2), the ADT subgroup had the same actuarial DFS as the non-ADT subgroup (P = .550), as illustrated in Figure 2.
|No. of Patients (%)|
|Variable||Entire Cohort, N = 1006||ADT Subgroup, N = 658||Non-ADT Subgroup, N = 348||Pa|
|Pretreatment PSA, ng/mLb|
|≤6||766 (76)||419 (63)||347 (99)||< .0001d|
|7||239 (24)||239 (37)||1 (<1)|
|T1||450 (45)||292 (44)||158 (45)||.0105c|
|T2||556 (55)||366 (56)||190 (55)|
|Low||586 (58)||272 (41)||314 (90)||< .0001d|
|Intermediate||419 (42)||386 (59)||34 (10)|
|Percentage positive coresb|
|<50%||640 (64)||389 (59)||251 (72)||.0022d|
|≥50%||269 (27)||195 (30)||78 (22)|
|Missing||97 (10)||74 (11)||1(6)|
|V100, %be||< .0001d|
Analyzing the entire cohort, none of the prognostic variables entered in the univariate or multivariate models were predictive of DFS, although log iPSA had a strong trend (P = .057) (Table 3). However, when the analysis was restricted to the non-ADT subgroup, decreasing V100 and D90 values were each predictive of an increased risk of disease recurrence in both univariate and multivariate models (P = .043) (Table 3).
|Entire Cohort, N = 1006||ADT Subgroup, N = 658||Non-ADT Subgroup, N = 348|
|Variable||HR (95% CI)||P||HR (95% CI)||P||HR (95% CI)||P|
|NCCN risk stratum: Intermediate vs low||1.013 (0.579-1.774)||.9600||1.320 (0.630-2.766)||.4600||0.478 (0.067-3.398)||.4600|
|Gleason sum: 7 vs ≤6||1.225 (0.649-2.312)||.5300||1.439 (0.704-2.942)||.3200|
|Clinical T-classification: T2b/c vs T1c/T2a||1.001 (0.436-2.299)||1.000||1.281 (0.506-3.245)||.6000||0.501 (0.071-3.566)||.4900|
|ADT: Yes vs no||0.837 (0.469-1.495)||.5500|
|Order of implantation: Unit = 1a||1.000 (0.999-1.001)||.9400||1.000 (0.999-1.002)||.6700||0.999 (0.997-1.001)||.3900|
|Age at implantation: Unit = 1 ya||1.021 (0.978-1.066)||.3400||1.014 (0.962-1.068)||.6100||1.040 (0.961-1.124)||.3300|
|Log iPSA: Unit = 10 ng/mLa||3.200 (0.965-10.619)||.0570b||2.831(0.536-14.955)||.2200||11.617 (0.601-224.627)||.1000|
|PPC: Unit = 1%a||1.799 (0.507-6.382)||.3600||1.496 (0.311-7.189)||.6100||2.957 (0.355-24.662)||.3200|
|D90Gy: Unit = 1 Gya||0.994 (0.981-1.008)||.3900||1.003 (0.986-1.020)||.7400||0.977 (0.955-0.999)||.0430c|
|V100: Unit = 1%a||0.982 (0.945-1.020)||.3400||0.995 (0.944-1.049)||.8600||0.938 (0.882-0.998)||.0430c|
To further explore the dose response, D90 cutoff points of 130 Gy, 140 Gy, 150 Gy, 160 Gy, 170 Gy, and 180 Gy were tested using a Fine and Gray competing risks analysis. In the non-ADT subgroup, all of the tested dose cutoff points demonstrated trends that correlated increased D90 values with improved DFS, but none reached statistical significance. The lowest P value (P = .1200) was observed using a cutoff point of 130 Gy (see Fig. 3a). No significant or nonsignificant trends were observed linking D90 to DFS for the ADT subset or the entire cohort. Figure 3b illustrates the Fine and Gray analysis of the ADT subgroup using a D90 cutoff point of 130 Gy (P = .9100).
Seven men were scored as DO/WD, resulting in an actuarial CSS rate of 99.1% (95% CI, 97.3%-99.7%) at 10 years (Table 1, Fig. 1). Proportionally, there have been a similar number of deaths among relapsed versus disease-free patients (7 of 49 [14.3%] vs 106 of 957 [11.1%], respectively) (Table 4), and disease recurrence was not predictive of OS in univariate or multivariate analysis.
|No. of Patients/Total No. (%)|
|Variable||All Patients, N = 1006||ADT Subgroup, N = 658||Non-ADT Subgroup, N = 348|
|Status at lastest follow-up|
|ANED||851 (84.6)||550 (83.6)||301 (86.5)|
|DNED||106 (10.5)||77 (11.7)||29 (8.3)|
|AWD||42 (4.2)||27 (4.1)||15 (4.3)|
|DO/WD||7 (0.7)||4 (0.4)||3 (0.9)|
|DF (ANED + DNED)||957 (95.1)||627 (95.3)||330 (94.8)|
|Relapsed (AWD + DO/WD)||49 (4.9)||31 (4.7)||18 (5.1)|
|Alive (ANED + AWD)||893 (88.8)||577 (87.7)||316 (90.8)|
|Dead (DNED + DO/WD)||113 (11.2)||81 (12.3)||32 (9.2)|
|Deaths among patients with recurrent disease: DO/WD ÷ (AWD + DO/WD)||7/49 (14.3)||4/31 (12.9)||3/18 (16.6)|
|Deaths among disease-free patients: DNED ÷ (ANED + DNED)||106/957 (11.1)||77/627 (12.3)||29/330 (8.8)|
|Proportion of recurrences at <36 mo||8/49 (16.3)||4/31 (18.2)||4/18 (22.2)|
|Proportion of recurrences at >96 mo||5/49 (10.2)||4/31 (18.2)||1/18 (5.6)|
There have been 113 deaths in total, and the actuarial OS rate was 93.8% at 5 years (95% CI, 92%-95.1%) and 83.5% at 10 years (95% CI, 79.8%-86.6%) (Table 1, Fig. 1). The projected median survival is >20 years; and, at the date of the current analysis, patients who died were an average of 4.2 years older at the time of implantation (69.1 years) than the patients who remained alive (64.9 years). In univariate analysis (Table 5), age at implantation (P < .0001) was joined by log iPSA (P = .0180) and the consecutive order of implantation (P = .0457) as significant predictors of OS, whereas both clinical stage (P = .0580) and PPC (P = .0503) demonstrated strong trends. However, in the multivariate Cox model (Table 6), only age at implantation (P = .0001) retained statistical significance when applied to the whole cohort, whereas log iPSA (P = .0282) was added to age at implantation (P = .0007) in the ADT subgroup. None of the covariates were significant for OS in the non-ADT subgroup in univariate or multivariate models. Similar to our previous studies,3, 4 we did not detect a statistically significant correlation linking protocol ADT and differences in all-cause mortality (see Table 5, Fig. 4).
|All Patients||ADT Group||No ADT Group|
|Variable||HR (95% CI)||P||HR (95% CI)||P||HR (95% CI)||P|
|NCCN risk stratum: Intermediate risk vs low risk||1.339 (0.926-1.936)||.1213||1.301 (0.828-2.043)||.2533||1.151 (0.401-3.306)||.7942|
|Gleason sum: 7 vs ≤6||0.990 (0.635-1.545)||.9658||0.891 (0.555-1.430)||.6325||—||—|
|Clinical T-classification: T2b/c vs T1c/T2a||1.598 (0.984-2.594)||.0580a||1.760 (1.018-3.042)||.0429a||1.209 (0.421-3.476)||.7244|
|ADT: Yes vs no||1.229 (0.815-1.853)||.3248||—||—||—||—|
|Order of implantation: Unit = 1b||0.999 (0.999-1.000)||.0457a||1.000 (0.999-1.000)||.2314||0.999 (0.997-1.000)||.0983|
|Age at implantation: Unit = 1 yb||1.067 (1.035-1.100)||< .0001a||1.075 (1.035-1.117)||.0002a||1.046 (0.992-1.103)||.0982|
|Log iPSA: Unit = 10 ng/mLb||2.699 (1.185-6.144)||.0180a||5.313 (1.585-17.803)||.0068a||1.151 (0.339-3.910)||.8221|
|PPC: Unit = 1%b||2.356 (0.999-5.558)||.0503a||2.036 (0.747-5.549)||.1645||2.975 (0.547-16.172)||.2069|
|D90Gy: Unit = 1Gyb||1.005 (0.995-1.015)||.2930||1.007 (0.995-1.019)||.2433||1.005 (0.985-1.025)||.6451|
|V100: Unit = 1%b||1.027 (0.994-1.061)||.1076||1.029 (0.992-1.068)||.1208||1.029 (0.950-1.114)||.4856|
|Disease relapse: No vs yes||0.935 (0.434-2.015)||.8646||0.722 (0.264-1.976)||.5256||1.745 (0.527-5.774)||.3617|
|All Patients||ADT Subgroupa|
|Variable||HR (95% CI)||P||HR (95% CI)||P|
|Age: Unit = 1 y||1.062 (1.030-1.095)||.0001b||1.069 (1.029-1.110)||.0007b|
|Log iPSA: Unit = 10 ng/mL||1.984 (0.873-4.509)||.1020||3.900 (1.157-13.150)||.0282b|
The site of first recurrence was established in 19 of 49 patients, including 8 who had clinical and/or histologic evidence of local relapse and 11 who had lymph node or distant metastatic relapse. The site of first recurrence was unknown in 30 patients who had biochemical relapse (61%), and all 30 had a normal digital rectal examination and either a normal postimplantation biopsy (N = 12) or no biopsy (N = 18, including 7 who also had no imaging studies). One individual underwent transurethral prostatic resection (TUPR) to relieve a stricture >10 years postimplantation, and the operative specimen revealed persistent/recurrent prostate cancer. Before the TUPR, that patient's PSA was unusually high at 0.83 ng/mL, but it was well short of triggering the Phoenix threshold, and recurrent cancer was not suspected preoperatively.
Overall compliance with follow-up has been excellent. Only 20 patients (2%) had no follow-up PSA values available, and the median number of postimplantation PSA values was 11 (mean ± standard deviation, 10.6 ± 4.5 PSA values; range, 1-36 PSA values). Figure 5 compares the actual number of postimplantation PSA values with the expected number if compliance with institutional protocol (every 6 months) had been perfect. By 5 years of follow-up, the expected number of PSA values is 10, and an average of 8 values were recorded. Compliance diminished somewhat with length follow-up, such that, by 10 years, an average of 14 PSA values per patient were recorded, whereas the expected number is 20 PSA values.
Compared with disease-free patients, relapsed patients have disproportionately numerous follow-up PSA values as a result of surveillance bias, whereby patients who have postimplantation PSA values that appear worrisome for recurrence tend to experience a sharply increased frequency of PSA measurements, and those with established relapse often have frequent PSA measurements to monitor the effects of secondary treatment. For these reasons patients with relapse were excluded from this part of the analysis.
The current data are unusual in presenting a consecutive cohort of uniformly treated men who received LDR-PB in a universal access, nonfee-for-service model of health care delivery. Most important, the data demonstrate excellent rates of DFS at 5 years and 10 years and unusually low follow-up PSA values among nonrelapsing patients, suggesting that long-term outcomes will be similar to the 10-year results and those reported by Seattle, which now has follow-up exceeding 15 years.6 However, these observations lead to other questions.
No single-arm study can hope to answer this question, but it is worth exploring its implications. In this cohort, the crude rate of disease recurrence was 4.9% at a median follow-up of 7.5 years, and the actuarial estimate of disease recurrence at 10 years was just 5.9%. Given these low rates after a relatively long follow-up, we conclude that the patients in this cohort had a very low (<5%) incidence of occult lymph node or distant metastatic disease before implantation. In this sense, the patients were well selected to benefit from effective local therapy. Conversely, it seems increasingly unlikely that early radical intervention improves OS for men with favorable-risk prostate cancer,7 and new strategies are emerging based on active surveillance or on focal therapies that promise less treatment-related morbidity.8, 9
Although new approaches and changes are welcome in oncology, men seeking guidance today need reliable benchmarks for conventional therapies by which to judge variations and alternatives. Therefore, we contend that the results of this cohort are relevant to newly diagnosed patients and to practitioners who provide therapy advice in today's era of widespread PSA screening. It is noteworthy that only 30% of this cohort would have been eligible for active surveillance using contemporary criteria established by the international Prostate Cancer Research International: Active Surveillance (PRIAS) protocol.10 Their median age was 66 years; and, with rare exceptions, operating room protocol restricted implantation recipients to those with American Society of Anesthesiologists risk class 1 and class 2 disease; this careful selection has resulted in a median survival that is predicted to exceed 20 years from the date of implantation. Furthermore, 24% of patients in the cohort had a Gleason sum of 7, 14% had an iPSA >10 ng/mL, 55% had T2 disease, and 27% had ≥50% positive cores (Table 2). Therefore, it is reasonable to speculate that some men in this cohort may have been spared metastatic disease along with the associated need for prolonged systemic therapy or potentially morbid radical salvage procedures after less effective forms of local therapy.
Because most recurrence events in this cohort (61%) were defined by biochemical parameters only, this is a relevant question. In our view, biochemical endpoints are reliable surrogates for disease recurrence provided the data are derived from large samples with regular PSA testing and long post-treatment follow-up. Furthermore, the Phoenix definition provides a multiply validated failure endpoint that is correlated with diminished OS in both intermediate-risk and high-risk patients11 after EBRT. The current study also confirms the sensitivity of the Phoenix definition, because there was only 1 patient with a known clinical or pathologic cancer recurrence in this cohort who did not also trigger the Phoenix definition of biochemical recurrence (see Patterns of Failure, above).
In our previous reports,3, 4 receipt of ADT and iPSA were predictive of DFS in multivariate analysis, but neither of these prognostic variables reached statistical significance in the updated univariate and multivariate models. In contrast, this update confirms our previously reported3, 4 lack of an obvious dose response for the widely accepted whole prostate dose metrics D90 and V100. The absence of an apparent dose response probably is explained by the existence of a threshold dose above which local tumor eradication is almost certain. Apparently, as discussed in detail in a previous report,12 this threshold dose is not well captured by the traditional dose metrics D90 and V100 when applied to our data set, which can be explained in part by 4 factors: 1) the consistent receipt of neoadjuvant and concomitant ADT by all patients with a Gleason sum of 7 or an iPSA >10 ng/mL; 2) relatively uniform, high-quality implants from the onset of the program (just 13% of all implantations in this cohort had D90 values <130 Gy; data not shown); 3) a planning philosophy that delivers a confluent 150% prescription dose to the periphery of the prostate; and 4) the deliberate use of extraprostatic seeds to provide dose coverage beyond the prostate boundary without increasing the central dose.12
The value of ADT for high-risk disease is established13; however the use, duration, and timing of ADT in men with low-risk and intermediate-risk disease remain controversial.14, 15 At the time our initial protocol was developed (1997), the purported value of adjuvant ADT was in ascendancy, and clinicians were skeptical about the efficacy of LDR-PB, particularly in men with intermediate-risk disease. Accordingly, mandatory, short-course ADT was built into our protocol, and nearly 66% of the men in this cohort received it, including all those with a Gleason sum of 7 or an iPSA >10 ng/mL (Table 2). Consequently, this cohort has no meaningful comparison group of intermediate-risk patients without ADT, because only 3% of intermediate-risk patients did not receive ADT (Table 2). Recognizing this important limitation, the data presented here suggests that the use of protocol ADT may have improved the DFS outcomes by partially compensating for dose deficiencies, because increasing D90 and V100 were only predictive of improved DFS in the non-ADT subgroup (see Table 3, Fig. 3a,b). It is interesting to note that the use of protocol ADT in this population has not resulted in a significant decrement in OS that some studies have reported.16 Note that our LDR-PB program discontinued mandatory ADT as of February 2005; therefore, future analyses of larger data sets will provide an opportunity to explore the therapeutic value of short-course ADT when combined with LDR-PB as well as the long-term effect of short-term ADT on OS.
By using matched pair analysis, we previously demonstrated that LDR-PB is much more effective than standard-dose EBRT (median dose, 68 Gy), although it carries no greater risk of serious complications.17 Comparisons with the literature suggest that PB also is more effective than dose-escalated EBRT.18 Finally, the success of a local therapy must be judged by its inverse: namely, local failure. Unless the false-negative rate is very high in postimplantation biopsies, our results suggest that the rate of local persistent/recurrent disease in this cohort must be less than 3% and possibly much less than 3% (see Patterns of Failure, above).
However, these results need to be interpreted with caution, because they apply only to a specific subtype of permanent LDR-PB and may or may not generalize to other brachytherapy techniques and protocols, which comprise a broad menu. Furthermore, oncologic outcomes, even confining comparisons to patients who receive iodine-125 LDR-PB, appear to vary substantially among centers.19-26
Ideally, we would either compare these data with population-based DFS after radical prostatectomy in clinically staged patients, or there would be adequately powered randomized trials to address the issue of which treatment provides the best disease control. Unfortunately, such data do not exist. The single published randomized trial comparing LDR-PB with radical prostatectomy27 included only 200 men and produced equivalent results in terms of DFS at 5 years, although a somewhat greater incidence of adverse effects after radical prostatectomy was reported. A recent systematic literature review of biochemical outcomes suggests that LDR-PB may be more effective than surgery in terms of biochemical endpoints.28
Surgery and radiotherapy tend to use different definitions of biochemical relapse, which compromises comparison. But this is not always the case. In a previous study,29 using a >0.4-ng/mL threshold for both interventions, we demonstrated that the actual biochemical relapse rate after LDR-PB in a large, consecutive cohort treated at BCCA was only half that predicted by the widely used Kattan nomogram if the same men had undergone surgery by experienced surgeons operating in high-volume centers (7% vs 13% actuarial at 5 years; P < .003).
During the era in which these men received treatment, only NCCN low-risk and selected (so-called “low-tier”) intermediate-risk patients were eligible for LDR-PB by BCCA policy, as mentioned above. “High-tier” intermediate-risk patients only became eligible outside of a trial setting in February 2009. We expect that this high-tier subgroup will experience more distant recurrence events after brachytherapy, but the effect on local control, in our opinion, is less predictable.
Despite its status as a population-based, consecutive cohort, our study carries some of the limitations of single-institution series, in that patients are not necessarily representative of other regions in which differences in population risk profile, comorbidities, Gleason scoring, and screening behavior may influence outcomes as much or more than treatment. Finally, the adverse effects of prostate cancer treatment are vitally important in treatment recommendations. In terms of prevalence and incidence, the treatment-related toxicity in the BCCA LDR-PB program has been comparable to that reported by other centers for urinary bladder, lower bowel, and erectile function.30-33
In conclusion, these population-based data provide credible evidence that LDR-PB is capable of delivering unsurpassed DFS and local control for the risk strata examined when used as a single modality or in combination with short-course ADT. It is perhaps noteworthy that these results were achieved in a nonfee-for-service, universal access medical environment in which multiple practitioners receive specific training; adhere to consistent, evidence-based eligibility criteria; and apply uniform, but evolving, treatment protocols and planning algorithms.
We acknowledge the contributions of Alexander Agranovich, MD, FRCPC, Eric Bethelet, MD, FRCPC, and Jonn Wu, MD, FRCPC; of Vince Lapointe, BSc (Medical Physics) for database design and maintenance; and of Jeremy Hamm, MSc (Department of Population Oncology, British Columbia Cancer Agency) for all statistical analyses.
No specific funding was disclosed.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.