EURURO Vol. 72 No. 3

346 476

1. Introduction
Disseminated prostate cancer is the second leading cause of
cancer death in Western men, and median survival remains
at only a few years . Radiation therapy or surgery
(radical therapy) is well established in managing localized
and locally advanced prostate cancer , and local
treatments have been shown to be of benefit in dissemi-
nated cancers of other histologic subtypes .
A recent US Surveillance Epidemiology and End Results
(SEER) study of 8185 men with disseminated prostate
cancer found that those treated with surgery or radiation
had higher 5-yr overall and cancer-specific survivals than
those treated without radical therapy . However, this
study was highly limited and further high-quality epidemi-
ologic data are needed.
We conducted a nationwide population-based cohort
study in Sweden, and compared survival among prostate-
cancer cases at high risk of dissemination that were initially
treated with radiation or radical prostatectomy with those
that had androgen deprivation therapy (ADT) as their first
treatment. We hypothesized that cases treated by initial
radical therapy would have improved survival compared
with those treated by initial ADT.
2. Patients and methods
This study is based on the PCBaSe Sweden, which has been described
previously . In brief, it is a population-based dataset of the National
Prostate Cancer Registry of Sweden, the Swedish Cancer Register, and
seven other national registers, using the unique 10-digit personal identity
number assigned to every resident in Sweden. Cause of death is abstracted
using the Cause of Death Register, which is assigned by those not directly
involved in the patients’ care. The PCBaSe dataset covers 98% of all
prostate-cancer cases in Sweden diagnosed since 1998 (with limited
coverage from 1996 to 1997), and has virtually complete, prospectively
collected data on year of diagnosis, age, clinical tumor/node/metastasis
(TNM) stage, tumor grade, presenting prostate-specific antigen (PSA),
planned primary treatment within 6 mo of diagnosis, Charlson
comorbidity index (CCI), marital status, employment status, county of
treatment, and mortality events during follow-up. CCI was estimated
from registrations in the In-patient Register and has been shown in a prior
PCBaSe study to have an impact on management and survival
From 109 333 men diagnosed from 1996 to 2010, we identified those
at high risk of disseminated prostate as defined by PSA > 50 ng/ml
(n= 22 132); this definition has been used in other PCBaSe Sweden
studies After excluding men with unknown treatment or death
before treatment (n= 973), no active treatment (n = 1477), and
incomplete data on follow-up or characteristics (n = 1330), we compared
the remaining 17 602 cases treated with ADT with the 750 radically
treated (radiation therapy, n = 630, or radical prostatectomy, n = 120)
patients. Owing to gross disparities in characteristics between the
‘‘average’’ patient receiving radical therapy and the ‘‘average’’ patient
who did not, we matched the two cohorts of cases using 1:1 nearest-
neighbor matching with a caliper width of 0.5 standard deviations; the
ADT patient of the exactly same grade, T stage, M stage (M0 vs Mx/M1),
and Charlson score, and who lay at most 0.5 standard deviations from
the corresponding radical therapy patient on year of diagnosis, age, and
ln[PSA + 1] was included. PSA was logarithmized due to its skewed
distribution so as to allow for more reasonable matching intervals. With
these criteria, we found ADT patient matches for 575 of the radical therapy
patients.
World Health Organization (WHO) grade 1 tumors were merged with
Gleason scores 2–6, WHO grade 2 tumors with Gleason score 7, and
WHO grade 3 tumors with Gleason scores 8–10, as consistent with other
PCBaSe Sweden studies .
The primary outcome of interest was prostate-cancer mortality.
Survival time was defined as the interval between the date of diagnosis
of prostate cancer and the date of death, emigration, or end of follow-up
on December 31, 2010.
2.1. Statistical analysis
Fisher’s exact and Wilcoxon–Mann–Whitney’s tests investigated differ-
ences in the distributions of patient characteristics by treatment groups.
Cumulative incidence curves were plotted to visualize cause-specific
mortality, and differences in each cause of mortality (prostate cancer/
other cause) were assessed using Fine-and-Gray competing risk
subdistribution hazard ratios (sHR) To address remaining
imbalances in the distribution of covariates among treatment groups,
multivariable model–adjusted estimates of sHRs were produced,
adjusting for the clinical and socioeconomic characteristics listed above.
Additional analyses were performed to interrogate the validity of the
results: (1) subgroup analyses on matched PSA > 100, M0, and M1/Mx
cohorts; (2) cumulative incidence curves for prostate-cancer and other-
cause mortality stratified by PSA, tumor grade, TNM stage, age, CCI, year
of treatment, marital status, employment status, and county of
treatment; and (3) owing to selective staging of nodal status in prostate
cancer, which could affect treatment assignment, analyses to compare
characteristics and received treatment of Nx and N0/N1 cases. We then
reassessed outcomes of radical treatment versus ADT in the subcohort of
men without known N stage before treatment (Nx). Finally, to avoid
issues arising from potential differences in time from diagnosis to
treatment assignment (‘‘immortal-time bias’’), we performed a land-
mark analysis where we reset the start of follow-up to 6 mo after
diagnosis and reassessed survival outcomes.
All tests performed were two sided at the 5% significance level.
Statistical analyses were performed using R software (version 3.0.1; R
Foundation for Statistical Computing, Vienna, Austria) using the cmprsk,
survival, and Matching packages.
3. Results
Median follow-up times for the living individuals and the
baseline patient–tumor characteristics for the treatment
cohorts are shown in . In the unmatched cohort, the
ADT group had worse prognostic factors on average than the
radically treated cases. In the matched cohort, the ADT
group had marginally higher PSAs on average, and a greater
proportion of N1 and M1 patients, but other differences
disappeared.
By the end of the study, the total number of prostate-
cancer (other-cause) deaths in the unmatched ADT and
radical therapy arms were 9062/17 602 and 86/750,
respectively (4388/17 602 and68/750, respectively); corre-
sponding figures for the matched cohort were 177/575 and
71/575 (62/575 and 58/575; ). In the unmatched
cohort, there appeared to be significant differences in both
prostate-cancer and other-cause deaths ); however, in
the matched cohort, prostate-cancer mortality reached over
50% by 14 yr in the ADT group compared with 25% in the
corresponding radically treated group, with minimal
differences in other-cause mortality and ).
Multivariable analyses showed similar sHR of circa 3 for
E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 3 4 5 – 3 5 1 346