EURURO Vol. 72 No. 6

921 1030

1. Introduction
Androgen deprivation therapy (ADT) for men with prostate
cancer is associatedwith adverse effects such as bone loss
with increased risk of fractures , metabolic aberrations
with a higher riskof cardiovascular disease (CVD) , and a
higher risk of diabetes mellitus type 2 .
It was recently reported that gonadotropin-releasing
hormone (GnRH) agonists are associated with a higher risk
of CVD compared to orchiectomy . However, the
difference in risk of CVD between GnRH agonists and
orchiectomy may have been confounded by indication for
treatment, since men who undergo orchiectomy are on
average older and have more advanced prostate cancer than
men treated with GnRH agonists, resulting in a shorter life
expectancy and less time at risk for CVD .
The aim of this study was to assess the association
between type of ADT and risk of CVD, while minimising
selection bias. We took advantage of a natural experiment
that took place in Sweden during the 1990s, when type of
ADT was often more influenced by the preference of the
healthcare provider than by a man’s prostate cancer
characteristics and comorbidity. We performed a semi-
ecologic study in which exposure to GnRH agonists was
assessed on a population level in experimental units defined
by healthcare provider, diagnostic time period, and age at
diagnosis, with outcomes assessed on an individual level
. We also analysed crude and net probability, with
exposure and outcome assessed on an individual level
2. Patients and methods
Prostate Cancer Data Base Sweden (PCBaSe) 3.0 contains information on
cancer characteristics and primary treatment from the National Prostate
Cancer Register (NPCR) of Sweden . Information on comorbidity
from the Patient Registry and data on educational level, income, and
marital status were obtained from the LISA database, and cause and date
of death were obtained from the Cause of Death Registry
The current study included men diagnosed with prostate cancer
during 1992–1999 who received GnRH agonists or bilateral orchiectomy
as primary [16_TD$DIFF]treatment (Supplementary Fig. 1).
No data on the date of treatment are available in the NPCR. Therefore,
we assessed the time from diagnosis to start of treatment using data
from a later calendar period (2006–2012) when these dates were
available from other sources. Information in the Prescribed Drug
Registry (which started in 2005) for date of first filled prescription for
GnRH agonist and data in the Patient Registry (which reached high
capture of orchiectomy procedures in the mid-2000s) showed that 90%
of[4_TD$DIFF] men had received their primary [16_TD$DIFF]treatment within 3 mo after the date
of diagnosis. Therefore, follow-up in the current study was started 3 mo
after the date of prostate cancer diagnosis, and the men were followed
until the event of interest, death, emigration, or end of the study period
(December 31, 2013), whichever event came first.
The CVD endpoint was identified as the first occurrence of a CVD
diagnosis (ICD-10 codes I00–I99), including hypertension (I10–I15),
ischaemic heart disease (I21–I25), stroke (I60–I64, G45), deep venous
thrombosis or pulmonary embolism (I80–I93, I26), and arterial
embolism (I74, K55), in the Patient Registry or Cause of Death Registry.
The associations between GnRH agonists or orchiectomy and
fractures (SX2) and diabetes (E10–E14) were also assessed.
The research ethics board at Umea˚ University Hospital approved the
study.
2.1. Statistical methods
Differences in characteristics between the treatment groups were tested
using the x2
[14_TD$DIFF] test for categorical variables and the Mann-Whitney U test
for continuous variables.
Threeanalyticalapproacheswereusedtocomparerisk of CVD between
men treatedwithGnRHagonistsandmen treatedwith orchiectomy. First, a
semi-ecologic study designwas applied to assess exposure to treatment on
agroup levelinanattempttominimise selectionbia . Exposurewas
measured as the proportion of men who received GnRH agonists in
experimental units defined by healthcare provider, 2-yr diagnostic time
period, and age at diagnosis (<70, 70–74, 75–79, 80+ yr). For each of the
580 experimentalunits, thenumberofeventsandperson-yearsatriskwere
calculated. A Poisson model with the logarithm of person-years at risk as
the offsetwas used toassess the associationbetween the proportionofmen
who received GnRH agonists, included as a restricted cubic spline, and the
risk of a CVD event, with prostate-specific antigen (PSA), T stage,
metastases, previous CVD, hypertension, and previous diabetes within
5 yr fromdiagnosis as covariates. Results are presented as relative risk (RR)
with 95% confidence interval (CI).
Using individual data on exposure and outcome, the crude and net
probability of CVD were estimated. The crude probability of death from
prostate canceranddeath fromcauses unrelated to CVDwere calculated in
a competing-risks analysis . The net probability of CVDwas estimated
using the Kaplan-Meiermethod, and the hazard ratio (HR) and95%CI were
estimated with multivariable Cox proportional hazards models using age
as the time scale and censoring observations at the time of the occurrence
of a competing event (ie, death from other causes) The multivariable
model included type of treatment (GnRH agonist vs orchiectomy), year of
diagnosis (continuous), PSA (categorical), stage (categorical), metastases
(categorical), and previous CVD (yes vs no), hypertension (yes vs no), and
diabetes (yes vs no) within 5 yr from diagnosis.
All tests were two-sided and the significance level was set to
p < 0.05. Statistical analysis was performed using R version 3.1.2 (R
Foundation for Statistical Computing, Vienna, Austria).
3. Results
The study population consisted of 6556 men who received
GnRH agonists and 3330 men who underwent orchiectomy
as primary [16_TD$DIFF]treatment. The median follow-up for men alive at
the end of follow-upwas 16 yr and therewas a total of 46 012
person-years of follow-up. There was a sevenfold difference
in use of GnRH agonists between experimental units with the
lowest and the highest use (14% vs 96%). The use of GnRH
agonists increased during the study period . Men
treated with GnRH agonists were younger and had a higher
proportion of nonmetastatic disease, lower serumPSA levels,
fewer previous CVD events. and higher educational level in
comparison to men who underwent orchiectomy (Supple-
mentary Table 1). These differences were smaller when
comparing units with high and low use of GnRH agonists, but
in the same direction as in the direct comparison between
men treated with GnRH agonists and orchiectomy ).
3.1. CVD risk according to type of ADT exposure in experimental
units
The CVD risk was similar for men treated in units with the
highest proportion of GnRH agonist use and men treated in
units with the lowest use (RR 1.01, 95% CI 0.93–1.11;
and ).
E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 9 2 0 – 9 2 8 921