EURURO Vol. 72 No. 4

535 656

1. Introduction
Active surveillance (AS) is a common management strategy
for men with low-risk prostate cancer (PCa). AS, with the
intention to start curative treatment at the time of
progression, may preserve some of the benefits of screening
by minimising the harm caused by overtreatment of
indolent cancer .
AS has been used for management of men with low-risk
PCa for more than a decade . AS ends when curative
treatment is started, when changing to watchful waiting
(WW), or when death occurs . WW, also called
symptom-guided treatment, is a management option for
men with low-risk PCa who have a limited life expectancy
and for whom curative treatment at the time of progression
is not deemed to be beneficial. Men on WW typically start
on androgen deprivation therapy (ADT) when symptomatic
progression occurs .
The PCa among a large proportion of men on AS will not
progress, and these men will thus remain on AS for a long
period of time. Their life expectancy will at some point
become so short that curative treatment is no longer
indicated, leaving WW as the remaining option. This change
to WW has an impact on health care resource allocation, as
it is associated with less intense follow-up (eg, no more
biopsies). However, the precise time point for this change is
rarely defined and documented in medical records. Only a
few studies to date have investigated the change from AS to
WW, as most studies on AS focus on the change to curative
treatment .
The aim of the current study was to assess the change
from AS to WW among men with very low-risk PCa. We
adopted a state transition model using comprehensive
data on cancer characteristics and comorbidity to estimate
the proportion of men who change from AS to WW and
when this occurs . We also investigated how this
change from AS to WW is affected by age and comorbidity.
2. Patients and methods
2.1. Study population and data collection
In 2012, the National Prostate Cancer Register (NPCR) of Sweden was
record linked to a number of other population-based registers using the
Swedish personal identity number . The linkage was updated in
2014. The NPCR became nationwide in 1998 and captures more than 98%
of all newly diagnosed PCa compared to the Swedish Cancer Registry.
NPCR keeps detailed information on tumour characteristics and primary
treatment at the time of diagnosis. Combining data from the National
Patient Register and the National Prescribed Drug Register regarding
subsequent treatment changes with the NPCR and other health care
registers and demographic databases resulted in PCBaSeTraject, which is
the basis for this study . To assess the burden of concomitant disease,
all changes in Charlson comorbidity index (CCI) during the course of
follow-up were retrieved from the National Patient Register and the
Swedish Cancer Register. The date of prostate biopsy as registered in
the National Patient Register was used as an indication of ongoing AS.
More specifically, WW and AS were both registered in the NPCR as
deferred treatment up to 2007, after which they were distinguished.
Their definitions are consistent with international guidelines AS is
defined as a strategy in which the man is followed closely and curative
treatment is initiated if there is evidence of disease progression. By
contrast, WW is usually performed for men with a limited life
expectancy, and ADT is initiated if symptomatic progression occurs .
We selected men with very low-risk PCa who started AS. Very low-
risk PCa was defined as Gleason score 6 with prostate-specific antigen
(PSA) 10 ng/ml, prostate volume <90 cm 3, PSA density <0.2, 6–12 core
biopsies performed, T1c or T2, positive cores 33%, and cancer length
6 mm as defined by Loeb et al in a previous PCBaSe study. We
added an additional age restriction (age 40–75 yr) in accordance with
the Swedish recommendation for AS Biopsy-related information
was not available for all men, and therefore we used multiple
imputation based on chained equations to create five imputation
data sets in which it was possible to define whether the above criteria
were fulfilled or not. Data were missing for cancer length in biopsies
(51%), percentage positive cores (40%), and prostate volume (37%);
complete data were available for 46% of men. The imputationwas based
on all data presented in as well as time-to-event data.
Furthermore, since the NPCR did not make a distinction between WW
and AS before 2007, this group of men ( n = 856) are classified as
unknown deferred treatment (DT). To determine whether these men
were on AS or WW, we treated this as a missing data problem
(Supplementary material). Finally, men on WW as their primary
treatment were included in the study to facilitate the estimation of
changes from AS to WW.
To enhance the precision of our estimates, we used the above
methods after inclusion of men with low-risk PCa (Gleason score 6,
stage T1c/T2, PSA <10 ng/ml) who did not fulfil the above-mentioned
definition of very low-risk PCa. In Sweden, a large proportion of men
with low-risk PCa are also on AS. For instance, 74% of men with low-risk,
but not very low-risk, PCa were assigned to AS in 2014 More
specifically, we used information for men with low-risk PCa to obtain
additional information on treatment and comorbidity changes, which
was then included in our models. Detailed results for men with low-risk
PCa who did not fulfil the definition of very low-risk PCa are presented in
the Supplementary material.
The Research Ethics Board at Umea˚ University approved this study.
2.2. Statistical methods
The change from AS to WW involves a competing risks problem. In
addition to the two competing risks described above (ie, the change to
curative treatment [radical prostatectomy or radiotherapy] or death), we
also considered AS failure as a competing risk. AS failure was defined as
initiation of ADT for men on AS without any signs of disease progression.
Since the change from AS to WW was not documented in PCBaSe, we
could not calculate cumulative incidence proportions as defined by
Kalbfleisch and Prentice nor could we use standard imputation
techniques.
We therefore investigated changes from AS using a state transition
model . We estimated the probability of treatment changes
(transition probabilities) between predefined treatments (states), as
illustrated by the arrows (transitions) and circles (states) in . Each
patient was followed from the initial state (yellow boxes) until ending up
in an absorbing state (orange boxes). All state transition models were
developed in accordance with the good practice guidelines of the ISPOR
Modeling Task Force and the Society for Medical Decision Making .
A shows the intended study. However, as no data are
available to directly investigate the change from AS to WW, we also
included men onWWas a primary treatment in our analyses B).
Apart from transitions to curative treatment, WW, and AS failure, we also
added the possibility of a transition from AS to ADT (AS ! WW ! ADT).
As our main aim was to study the transition from AS to WW, secondary
transitions (eg, radical prostatectomy ! death) are not shown in
E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 5 3 4 – 5 4 1 535