EURURO Vol. 72 No. 5

Introduction

Prostate cancer (PCa) remained the second leading cause

of cancer death among men in 2015

[1] .

This is largely

attributable to patients with unfavorable PCa (ie, inter-

mediate- and high-risk disease), which is associated with

10-yr survival as low as 25% without treatment

[2]

. Multi-

ple prospective trials of men with unfavorable PCa have

shown that the combination of radiation therapy and

androgen suppression (AS) improves overall survival (OS)

when compared to either treatment alone

[3–12]

. Howev-

er, these studies did not use dose-escalated external-beam

radiation therapy (DE-EBRT), which has been shown to

improve biochemical control, local progression, and

distant metastases in contemporary randomized con-

trolled trials

[13–18]

. Brachytherapy is a treatment

modality that has been used to escalate radiation doses

beyond those that can be delivered routinely with EBRT.

Until recently, there were only limited data comparing

DE-EBRT and brachytherapy among men with unfavorable

PCa

[19,20] .

The ASCENDE-RT trial is a randomized phase 3 study

comparing AS and whole-pelvis EBRT with either an EBRT

boost or a low-dose rate brachytherapy (LDR-B) boost for

men with intermediate- and high-risk PCa. Preliminary

results suggest improved biochemical control for those

receiving LDR-B boost. There was no difference in 7-yr OS,

but there was a trend favoring LDR-B boost (85.7% vs 81.5%)

[21]

. At median follow-up of 6.5 yr, the study had not

reached median survival, and if a difference exists, the

results may not be available for many years. Therefore, we

analyzed the National Cancer Data Base (NCDB) to compare

survival outcomes between those treated with AS and EBRT

followed by either EBRT alone or LDR-B boost in a large

national cohort.

Patients and methods

2.1.

Data source

We performed an analysis of the NCDB, a clinical oncology database

containing hospital registry data from

1500 Commission on Cancer

(CoC)–accredited centers

[22]

. Detailed information on OS, first

treatment, clinical characteristics, demographics, radiation type, site,

dose, and with AS are included in the database. The CoC of the American

College of Surgeons has not validated the database and is not responsible

for the conclusions of this study.

2.2.

Cohort identification

Men diagnosed between 2004 and 2012 with intermediate- or high-risk

PCa, defined according to National Comprehensive Cancer Network

Guidelines

[23] ,

and treated with definitive radiotherapy were identified

in the database. Inclusion and exclusion criteria for our study population

are summarized in Supplementary Figure 1 and are based on the

ASCENDE-RT trial

[21]

. Administration of neoadjuvant AS was limited to

those who started AS within 8 mo before EBRT followed by LDR-B or DE-

EBRT. Patients in the DE-EBRT arm were limited to those who received a

dose between 75.6 and 86.4 Gy. Patients who did not receive radiation to

the prostate and/or pelvis were excluded.

2.3.

Patient covariates and outcomes

Treatment modality following AS and EBRT (DE-EBRT vs LDR-B boost)

was the primary independent variable. Covariates identified in the

database were age, race (black, non-Hispanic white, other), insurance

(Medicare, uninsured, private, Medicaid or government/unknown),

geography (Northeast, South, Midwest, West), facility (academic or

non-academic), Charlson comorbidity score (CCS) (0, 1, or 2), risk group

(intermediate or high), Gleason score (GS; 6, 7, or 8–10), prostate specific

antigen (PSA;

10, 10 to

20, or 20 to

40 ng/ml), and clinical T stage

(T1c–T2a, T2b–T2c, or T3a). OS was the primary outcome of interest.

2.4.

Statistical methods

Descriptive statistics for patient characteristics were compared between

the DE-EBRT and LDR-B arms using

and

tests for categorical and

continuous variables, respectively. To identify covariates associated with

LDR-B treatment selection, logistic regression was used. Treatment

groups were compared by evaluating the proportion of patients treated

each year with LDR-B boost compared to DE-EBRT. Univariate analyses

(UVA) were carried out using Cox proportional hazards regression, the

log-rank test, and Kaplan-Meier survival analysis. OS time was

calculated from the time of diagnosis to date of death. Multivariate

analyses (MVA) and subset survival analyses were carried out using Cox

proportional hazards to control for covariates found to be significant

(

0.05) on UVA, including treatment type, age, race, insurance type,

geographic region, facility type, CCS, GS, PSA, and clinical stage. To

further adjust for unbalanced variables, propensity score matching was

carried out using covariates associated with treatment selection on

logistic regression. Covariates were matched between treatment groups

using one-to-one nearest-neighbor matching without replacement. A

matched sensitivity analysis was conducted by varying GS from 7 to

8–10 in 5% increments for those receiving DE-EBRT to investigate the

degree of underadjustment that would result in a null result. Subset

survival analyses were performed using log-rank testing, Kaplan-Meier

survival analysis, and testing of interactions on Cox proportional hazards

regression. The first subset analysis was limited to men aged

60 yr with

no comorbidities to exclude patients more likely to have non-PCa causes

of death. This age was chosen as it reflects a large proportion of patients

with no comorbidities

[24]

. Interactions between treatment (DE-EBRT vs

LDR-B boost) and age, treatment and risk group (high and intermediate

risk), treatment and dose, and risk group and dose were tested. Statistical

analyses were performed using Stata version 13.1 (StataCorp, College

Station, TX, USA) and statistical tests were two-tailed with

= 0.05.

Results

3.1.

Patient characteristics

For median follow-up of 63 mo (interquartile range

37–88 mo) we identified 25 038 patients treated with AS

and definitive radiation, of whom20 522 (82%) were treated

with DE-EBRT and 4516 (18%) with EBRT followed by LDR-B

boost. A comparison of patient characteristics is shown in

Table 1

3.2.

Treatment selection and care patterns

The relative odds of receiving LDR-B compared to DE-EBRT

are reported in Supplementary Table 1. Between 2004 and

2012, the proportion of patients receiving LDR-B boost

decreased from 29% to 14%, while patients receiving

E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 7 3 8 – 7 4 4

739