G Model
CANEP-854; No. of Pages 12
Cancer Epidemiology xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Cancer Epidemiology
The International Journal of Cancer Epidemiology, Detection, and Prevention
journal homepage: www.cancerepidemiology.net
Optimal uptake rates for initial treatments for cervical cancer
in concordance with guidelines in Australia and Canada: Results
from two large cancer facilities
Yoon-Jung Kang a,b,c,*, Dianne L. O’Connell b,c,d,e, Jeffrey Tan f, Jie-Bin Lew a,b,
Alain Demers g,h, Robert Lotocki g,i, Erich V. Kliewer g,h,j, Neville F. Hacker k,
Michael Jackson l, Geoff P. Delaney m, Michael Barton m, Karen Canfell a,b
a
Prince of Wales Clinical School, UNSW, New South Wales, Australia1
Cancer Research Division, Cancer Council NSW, New South Wales, Australia2
c
School of Public Health, University of Sydney, New South Wales, Australia3
d
School of Public Health and Community Medicine, UNSW, New South Wales, Australia
e
School of Medicine and Public Health, University of Newcastle, New South Wales, Australia
f
Royal Women’s Hospital, Victoria, Australia
g
Community Health Sciences, University of Manitoba, Winnipeg, Manitoba, Canada
h
Epidemiology and Cancer Registry, CancerCare Manitoba, Winnipeg, Manitoba, Canada
i
Division of Gynecologic Oncology, CancerCare Manitoba, Winnipeg, Manitoba, Canada
j
Cancer Control Research, British Columbia Cancer Agency, Vancouver, British Columbia, Canada
k
Royal Hospital for Women, New South Wales, Australia
l
Prince of Wales Hospital, New South Wales, Australia
m
Collaboration for Cancer Outcomes Research & Evaluation, Ingham Institute for Applied Medical Research, UNSW Australia, Sydney, New South Wales,
Australia
b
A R T I C L E I N F O
A B S T R A C T
Article history:
Received 10 November 2014
Received in revised form 16 April 2015
Accepted 17 April 2015
Available online xxx
Background: Prior work estimating optimal treatment utilisation rates for cervical cancer has focused on
radiotherapy or chemotherapy, using proportions of patients with clinical indications for specific
treatment strategies which were obtained from the published literature.
Objectives: To estimate optimal uptake rates for surgery, radiotherapy, chemotherapy and chemoradiotherapy for cervical cancer treatment in Australia and Canada, and to quantify the differences in the
optimal and the observed treatment utilisation rates in a large cancer facility from each country.
Methods: A decision tree was constructed to reflect treatments according to guidelines and current
practice (in 1999–2008) in each setting. Detailed patterns of care data from a large cancer facility in each
country were obtained, and the observed stage distribution and proportions of patients with each clinical
indication were used as inputs.
Results: The estimated overall optimal treatment rates for cervical cancer in Australia and Canada
differed, largely due to the difference in the stage distribution at diagnosis in the two settings; 72% vs 54%
with FIGO IA-IIA disease, respectively. The estimated optimal rates for surgery, radiotherapy,
chemotherapy and chemo-radiotherapy in Australia were 63% (95% credible interval: 61–64%), 52%
(53–56%), 36% (35–38%) and 36% (35–38%), respectively. The corresponding rates in Canada were 38%
Keywords:
Cervical cancer
treatment
optimal rate
guidelines
decision tree
Abbreviations: FIGO, International Federation of Gynecology and Obstetrics; MCR, Manitoba Cancer Registry; HR, hazard ratio; CI, confidence interval; CrI, credible interval;
ECOG, Eastern Cooperative Oncology Group; NCI, National Cancer Institute; BCCA, British Columbia Cancer Agency; GMCT, Greater Metropolitan Clinical Taskforce; EBRT,
External beam radiotherapy; CCMB, CancerCare Manitoba; RWH, Royal Women’s Hospital; POWH, Prince of Wales Hospital; NSW, New South Wales.
* Corresponding author at: Cancer Screening Group, Cancer Research Division, Cancer Council NSW, 153 Dowling Street, Woolloomooloo NSW 2011, Australia.
Tel.: +61 2 9334 1632; fax: +61 2 8302 3550.
E-mail addresses: [email protected] (Y.J. Kang), [email protected] (D.L. O’Connell), [email protected]
(J. Tan), [email protected] (J.B. Lew), [email protected] (A. Demers), [email protected] (R. Lotocki), [email protected] (E.V. Kliewer),
[email protected] (N.F. Hacker), [email protected] (M. Jackson), [email protected] (G.P. Delaney),
[email protected] (M. Barton), [email protected] (K. Canfell).
1
Past affiliation of Kang, Lew and Canfell.
2
Current affiliation of Kang, Lew and Canfell.
3
Past affiliation of Kang.
http://dx.doi.org/10.1016/j.canep.2015.04.009
1877-7821/ß 2015 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Kang Y-J, et al. Optimal uptake rates for initial treatments for cervical cancer in concordance with
guidelines in Australia and Canada: Results from two large cancer facilities. Cancer Epidemiology (2015), http://dx.doi.org/10.1016/
j.canep.2015.04.009
G Model
CANEP-854; No. of Pages 12
2
Y.-J. Kang et al. / Cancer Epidemiology xxx (2015) xxx–xxx
(36–39%), 68% (68–71%), 51% (49–52%) and 50% (49–51%), respectively. The absolute differences
between the optimal and the observed rates were similar between the two settings; the absolute
differences for chemotherapy and chemo-radiotherapy uptake were more pronounced (9–15% less than
optimal) than those for surgery and radiotherapy uptake (within 5% of optimal).
Conclusions: This is the first study to use detailed patterns of care data in multiple settings to compare
optimal and observed rates for all cervical cancer treatment modalities. We found optimal treatment
rates were largely dependent on the overall stage distribution. In Australia and Canada, observed surgery
rates, as measured in the two large cancer facilities, were similar to the estimated optimal rates, whereas
radiotherapy, chemotherapy and chemo-radiotherapy appeared to be under-utilised.
ß 2015 Elsevier Ltd. All rights reserved.
1. Introduction
Optimal rates of cancer treatment utilisation in a particular
setting generally depend on the applicable stage-specific guidelines for treatment and the setting-specific stage-distribution of
cancer at diagnosis. Initial treatment for cervical cancer performed
with curative intent generally involves surgery and radiotherapy
for early stage disease, and radiotherapy, chemotherapy and/or
chemo-radiotherapy for locally advanced and regional disease. The
most current international recommendations are encapsulated in
guidelines published by the International Federation of Gynecology
and Obstetrics (FIGO) [1]. Prior studies have estimated the optimal
proportion of cervical cancer patients who should be treated with
radiation [2–4] or chemotherapy [5]. However, no study has taken
into account the full range of treatment types, nor explicitly
compared the findings for optimal treatment rates across
treatment modalities to observed data in different settings.
The aim of the current study was, firstly, to build on the prior
work to integrate existing clinical evidence to estimate the optimal
percentages of patients with cervical cancer who should, according
to the best available evidence, receive each specific treatment
modality as their initial therapy (i.e. within the first year after
diagnosis). The second aim was to quantify differences in the
optimal and the observed treatment utilisation rates, using data
from patterns of care studies in large cancer facilities. Our study
was performed in two settings, Melbourne, a metropolitan city in
the Australian state of Victoria and the Canadian province of
Manitoba. The results will assist health care decision makers to set
delivery targets for specific treatment modalities for cervical
cancer and to evaluate whether the level of current service is
adequate to meet demand [6].
2. Materials and methods
2.1. Recommended treatments for cervical cancer
National guidelines for cervical cancer treatment were not
identified for either Australia or Canada, but published treatment
recommendations from collaborative working groups [7] and
provincial guidelines [8], respectively, were available. These
available guidelines do not differ substantially from each other
and they are both similar to FIGO international guidelines for
cervical cancer treatment and recommendations from other
comparable countries [1,9,10]. Therefore, for the purpose of the
current analysis, synthesised international evidence-based guidelines were derived, which reflected both the standard of care in
each local setting and international standards since 1999 (Table 1).
2.2. Analysis of patterns of care data
For the purpose of the analysis, observed ‘‘current practice’’
was defined as treatment over the years 1999–2008. Patient
performance status was considered in the evaluation, and if
apparently sub-optimal treatments were observed for patients with
poor performance status (Eastern Cooperative Oncology Group
[ECOG] score 3–4) then treatment was assessed as being in
accordance with guideline treatments for the purposes of this
evaluation. This was done in order to take into account, as far as
possible, the potential impact of co-morbidities on the treatment
decision apparently deviating from the evidence-based guidelines.
Data from patterns of care studies performed in large cancer
facilities in Australian and Canadian jurisdictions were analysed to
provide information about actual treatment patterns and the
proportions of patients with clinical indications for various
treatment strategies in each setting. With respect to current
management practice in Australia, data from the Royal Women’s
Hospital (RWH) were used for patients who received their first
treatment for diagnosed cervical cancer during the period 1999–
2008. The RWH is an area-based tertiary referral centre in
Melbourne, a metropolitan city in the Australian State of Victoria.
During the period, 385 patients were treated at the centre,
representing 25% of all cervical cancer patients diagnosed in
Victoria during the same period. These data are not populationbased, but were the only FIGO stage-specific data available since
population-based cancer registries do not collect information on
FIGO stage.
Data on current practice in Canada were obtained for patients
diagnosed in the period 1999–2008 who were identified from the
population-based Manitoba Cancer Registry (MCR) [11]. Detailed
information on stage and treatment were obtained by combining
the MCR (for the period 1984–2008) and a database derived from
chart reviews (for the period 1984–1999); both the MCR and the
charts were maintained by CancerCare Manitoba. The Manitoba
data are population-based, and all cancer patients are referred to
CancerCare Manitoba for treatment. Of the 452 cervical cancer
patients who were identified from the MCR during the period
1999–2008, FIGO stage was recorded for 369 patients (75%). For
early stages (FIGO IA, FIGO IB-IIA with small lesions), the
proportions of patients with lymphovascular space invasion,
negative surgical margins or lymph node involvement and the
associated treatment patterns were only available for those who
were diagnosed in the period 1990–2003. The recommended
treatments, management practice and stage-specific survival for
these early stages of the disease did not significantly change over
the period 1990–2008. Therefore, it was expected that the use of
the data from the earlier period for early stage disease would not
affect the results substantially. Hence, for the current analysis, data
for 542 patients were used: these include 369 cases diagnosed in
1999–2008 who had complete records on FIGO stage, treatment
and lesion size (for FIGO IB-IIA stage disease) and 173 cases with
early stage disease (FIGO IA, FIGO IB-IIA with small lesions)
diagnosed in 1990–2003 who had complete records on lymphovascular space invasion, negative surgical margins or lymph node
involvement.
Please cite this article in press as: Kang Y-J, et al. Optimal uptake rates for initial treatments for cervical cancer in concordance with
guidelines in Australia and Canada: Results from two large cancer facilities. Cancer Epidemiology (2015), http://dx.doi.org/10.1016/
j.canep.2015.04.009
1
1.1
1.1.1
Clinical indications
Recommended treatment
Guidelines/recommendation
(published since 1999)
Observed proportions of patients
Baseline (range)b
(a) Australia
(b) Canada
Data sourcesd
FIGO stage IA1
IA1, medically operable
IA1, medically operable,
LVSI ve
IA1, medically operable,
LVSI ve, standard
–
–
–
–
–
–
0.22 (0.21–0.22)
0.98 (0.94–1.00)
1.00 (0.94–1.00)
0.14 (0.13–0.15)
1.00 (0.94–1.00)
0.92 (0.80–0.98)
RWH, CCMB
RWH, CCMB
RWH, CCMB
Total hysterectomy
0.40 (0.28–0.53)
0.78 (0.63–0.89)
RWH, CCMB
IA1, medically operable,
LVSI ve, margin ve/
fertility saving
IA1, medically operable,
LVSI +ve
Conisation
0.60 (0.72–0.47)
0.22 (0.11–0.37)
RWH, CCMB
0.00 (0.00–0.06)
0.08 (0.02–0.20)
RWH, CCMB
1.2
IA1, medically inoperable
Brachytherapy
FIGO Guidelines [1], NCI PDQ
statement [9], GMCT
Guidelines [7], BCCA
Guidelines [8], NCCN
Guidelines [10]
FIGO Guidelines [1], NCI PDQ
statement [9], BCCA
Guidelines [8]
GMCT Guidelines [7], BCCA
Guidelines [8], NCCN
Guidelines [10]
NCI PDQ statement [9], NCCN
Guidelines [10]
0.02 (0.00–0.06)
0.00 (0.00–0.06)
RWH, CCMB,
NSW Health
Survey [12]
2
2.1
2.1.1
FIGO stage IA2
IA2, medically operable
IA2, medically operable,
standard
–
–
(Modified) Radical
hysterectomy + PL
0.04 (0.03–0.06)
1.00 (0.94–1.00)
0.55 (0.55–1.00)
0.05 (0.04–0.06)
0.89 (0.67–0.99)
0.88 (0.55–1.00)
RWH, CCMB
RWH, CCMB
RWH, CCMB
2.1.2
IA2, medically operable,
consideration
IA2, medically operable,
consideration, LVSI ve
IA2, medically operable,
consideration, fertility
saving
IA2, Medically inoperable
–
–
–
FIGO Guidelines [1], NCI PDQ
statement [9], GMCT
Guidelines [7], BCCA
Guidelines [8], NCCN
Guidelines [10]
–
0.45 (0.00–0.45)
0.12 (0.00–0.45)
RWH, CCMB
Total hysterectomy + PL
FIGO Guidelines [1]
0.20 (0.03–1.00)
1.00 (0.03–1.00)
RWH, CCMB
Radical trachelectomy + PL
0.80 (0.00–0.98)
0.00 (0.00–0.98)
RWH, CCMB
Radical RT
FIGO Guidelines [1], GMCT
Guidelines [7], NCCN
Guidelines [10]
NCI PDQ statement [9], BCCA
Guidelines [8], NCCN
Guidelines [10]
0.00 (0.00–0.06)
0.11 (0.01–0.33)
RWH, CCMB,
NSW Health
Survey [12]
3
3.1
3.1.1
FIGO stage IB-IIA
IB-IIA, non bulky
IB-IIA, non bulky, not HR
–
–
–
–
–
–
IB-IIA, non bulky, low risk
(Modified) radical
hysterectomy + PL
3.1.1.2
IB-IIA, non bulky,
Intermediate risk
Radical RT
IB1: 0.00 (0.00–0.04)
IB2: 0.00 (0.00–0.27)c
IIA: 0.30 (0.30–0.30)
IB1: 0.33 (0.31–0.34)
IB2: 0.00 (0.00–0.27)c
IIA: 0.75 (0.41–0.97)
RWH, CCMB
3.1.1.3
IB-IIA, non bulky, not HR,
Radical
hysterectomy + PL + adjuvant
pelvic radiation
FIGO Guidelines [1], NCI PDQ
statement [9], GMCT
Guidelines [7], BCCA
Guidelines [8], NCCN
Guidelines [10]
FIGO Guidelines [1], NCI PDQ
statement [9], BCCA
Guidelines [8], NCCN
Guidelines [10]
FIGO Guidelines [1], GMCT
Guidelines, NCCN Guidelines
0.35 (0.34–0.36)
0.70 (0.64–0.75)
IB1: 0.94 (0.89–0.98)
IB2: 0.02 (0.01–0.05)
IIA: 0.03 (0.01–0.06)
IB1: 0.60 (0.55–0.64)
IB2: 1.00 (0.45–1.00)
IIA: 0.25 (0.03–0.34)
RWH, CCMB
RWH, CCMB
RWH, CCMB
3.1.1.1
0.46 (0.43–0.49)
0.74 (0.66–0.80)
IB1: 0.96 (0.88–0.99)
IB2: 0.03 (0.01–0.07)
IIA: 0.01 (0.00–0.05)
IB1: 0.77 (0.68–0.82)
IB2: 0.33 (0.08–0.35)
IIA: 0.40 (0.40–0.40)
IB1: 0.23 (0.18–0.28)
IB2: 0.67 (0.38–0.92)
IIA: 0.30 (0.30–0.30)
IB1: 0.08 (0.05–0.11)
IB2: 0.00 (0.00–0.27)
IIA: 0.00 (0.00–0.25)
RWH, CCMB
1.1.1.1
1.1.1.2
1.1.2
2.1.2.1
2.1.2.2
Modified radical
hysterectomy + PL
RWH, CCMB
G Model
CANEP-854; No. of Pages 12
Brancha
Y.-J. Kang et al. / Cancer Epidemiology xxx (2015) xxx–xxx
Please cite this article in press as: Kang Y-J, et al. Optimal uptake rates for initial treatments for cervical cancer in concordance with
guidelines in Australia and Canada: Results from two large cancer facilities. Cancer Epidemiology (2015), http://dx.doi.org/10.1016/
j.canep.2015.04.009
Table 1
Synthesised evidence-based guideline recommendation for invasive cervical cancer treatment according to stage and the corresponding proportions of patients diagnosed at each stage in (a) Australia and (b) Canada in 1999–2008.
3
Clinical indications
Recommended treatment
Guidelines/recommendation
(published since 1999)
3.1.2
IB-IIA, non bulky, HR
–
–
3.1.2.1
IB-IIA, non bulky, HR, good
PS
Radical
hysterectomy + PL + adjuvant
chemo-radiotherapy
3.1.2.2
IB-IIA, non bulky, HR, poor
PS
IB-IIA, bulky
IB-IIA, bulky, good PS
Radiotherapy
FIGO Guidelines [1], NCI PDQ
statement [9], GMCT
Guidelines [7], BCCA
Guidelines [8], NCCN
Guidelines [10]
–
3.2
3.2.1
–
–
Primary chemo-radiotherapy
3.2.2
IB-IIA, bulky, poor PS
4
4.1
4.2
4.3
4.4
4.1.1
FIGO stage IIB-IVA
FIGO stage IIB
FIGO stage IIIA
FIGO stage IIIB
FIGO stage IVA
Good PS (in each FIGO
sub-stage)
4.1.2
Poor PS (in each FIGO substage)
5
5.1
FIGO stage IVB
IVB, curative/palliative
Neoadjuvant CT + radical
hysterectomy + PL adjuvant
RT/chemo-radiotherapy
Radiotherapy
–
–
Chemo-radiotherapy
FIGO Guidelines [1], NCI PDQ
statement [9], BCCA
Guidelines [8]
Radiotherapy
Chemotherapy
Radiotherapy
Chemo-radiotherapy
FIGO Guidelines [1], NCI PDQ
statement [9], BCCA
Guidelines [8]
(a) Australia
(b) Canada
Data sourcesd
IB1: 1.00 (0.71–1.00)
IB2: 0.00 (0.00–0.15)c
IIA: 0.00 (0.00–0.15)c
1.00 (0.90–1.00)
IB1: 0.94 (0.66–1.00-)
IB2: 0.00 (0.00–0.14)
IIA: 0.06 (0.00–0.20)
1.00 (0.90–1.00)
RWH, CCMB
RWH, CCMB
0.00 (0.00–0.10)c
0.00 (0.00–0.10)c
RWH, CCMB
0.26 (0.20–0.34)
1.00 (1.00–0.90)
IB1: 0.00 (0.00–0.29)
IB2: 0.72 (0.58–0.81)
IIA: 1.00 (0.41–1.00)
0.30 (0.25–0.36)
1.00 (1.00–0.90)
IB1: 0.00 (0.00–0.33)
IB2: 0.69 (0.510.81-)
IIA: 0.93 (0.64–1.00)
RWH, CCMB
RWH, CCMB
RWH, CCMB
IB1: 1.00 (0.41–1.00)
IB2: 0.28 (0.19–0.33)
IIA: 0.00 (0.00–0.29)
IB1: 1.00 (0.34–1.00)
IB2: 0.31 (0.19–0.35)
IIA: 0.07 (0.00–0.22)
RWH, CCMB
IB1: 0.00 (0.00–0.29)
IB2: 0.00 (0.00–0.09)
IIA: 0.00 (0.00–0.29)
0.00 (0.00–0.10)c
IB1: 0.00 (0.00–0.33)c
IB2: 0.00 (0.00–0.14)c
IIA: 0.00 (0.00–0.15)c
0.00 (0.00–0.10)c
RWH, CCMB
0.27
0.51
0.10
0.27
0.12
1.00
0.40
0.48
0.04
0.43
0.05
1.00
(0.42–0.39)
(0.50–0.46)
(0.03–0.06)
(0.44–0.41)
(0.03–0.06)
(1.00–0.86)
RWH, CCMB
0.00 (0.00–0.14)
0.00 (0.00–0.14)
RWH, CCMB,
POWH
0.01
0.00
0.50
0.50
0.06
0.06
0.64
0.30
RWH, CCMB
RWH, CCMB
(0.27–0.27)
(0.45–0.58)
(0.07–0.13)
(0.27–0.28)
(0.09–0.15)
(0.86–1.00)
(0.01–0.03)
(0.00–0.24)c
(0.38–0.50)
(0.38–0.50)
(0.05–0.07)
(0.01–0.14)
(0.54–0.73)
(0.25–0.33)
RWH, CCMB,
POWH
RWH, CCMB
FIGO – International Federation of Gynecology and Obstetrics; NCI – National Cancer Institute; BCCA – British Columbia Cancer Agency; GMCT – Greater Metropolitan Clinical Taskforce.
LVSI ve – lymphovascular space invasion absent; LVSI +ve – lymphovascular space invasion present; Margin ve – margin negative; margin +ve – margin positive; PL – pelvic lymphadenectomy; RT – radiotherapy; EBRT – external
beam radiotherapy; CT – chemotherapy; Non bulky – tumour size in diameter 4 cm; Bulky – tumour size in diameter >4 cm; Not HR – not at high risk of recurrence (lymph node not present); HR – high risk of recurrence (LN
present, parametria positive, margin positive); Low risk – lymph node (LN) absent, parametria negative, margin negative, LVSI absent; Intermedaite risk – lymph node negative, large tumour volume or LVSI present or outer onethird invasion of the cervical stroma or GOG > 120; PS – performance status; Poor PS – poor performance status (Eastern Cooperative Oncology Group (ECOG) score 3).
a
This hierarchical branch represents detailed model structure.
b
Baseline is the point estimate for each parameter obtained from data analysis. The initial range was based on the binomial exact 95% confidence interval (CI)s around the point estimate of each parameter. Using the Dirichlet
distribution, the range was then re-calculated so that each of the lower confidence limits and the upper confidence limits in multiple branches sum to 1. This range was used in sensitivity/uncertainty analysis.
c
No data were available, and one-sided exact 97.5% CI was calculated.
d
Australian patterns of care data were obtained from the Royal Women’s Hospital (RWH, n = 385) in the state of Victoria and Canadian patterns of care data were obtained from CancerCare Manitoba in the Province of Manitoba
(CCMB, n = 542). Data from the New South Wales (NSW) Health Survey and the Prince of Wales Hospital (POWH) in the Australian state of NSW were used to inform the proportions of patients with FIGO IA1/IA2 disease who are
medically inoperable and the proportions of patients with FIGO IIB-IVA disease with poor performance status (ECOG score 3–4), respectively.
Y.-J. Kang et al. / Cancer Epidemiology xxx (2015) xxx–xxx
Radical
hysterectomy + PL adjuvant
RT
–‘
–
FIGO Guidelines [1], NCI PDQ
statement [9], GMCT
Guidelines [7], NCCN
Guidelines [10]
FIGO Guidelines [1], NCI PDQ
statement [9], GMCT
Guidelines [7], NCCN
Guidelines [10]
FIGO Guidelines [1]
Observed proportions of patients
Baseline (range)b
G Model
Brancha
CANEP-854; No. of Pages 12
4
Please cite this article in press as: Kang Y-J, et al. Optimal uptake rates for initial treatments for cervical cancer in concordance with
guidelines in Australia and Canada: Results from two large cancer facilities. Cancer Epidemiology (2015), http://dx.doi.org/10.1016/
j.canep.2015.04.009
Table 1 (Continued )
G Model
CANEP-854; No. of Pages 12
Y.-J. Kang et al. / Cancer Epidemiology xxx (2015) xxx–xxx
In terms of proportions of patients with poor performance
status, different data sources were used to inform the model for
each stage. The range of proportions of patients with FIGO IA1/IA2
stage disease who are medically inoperable was estimated from
the NSW Health Survey [12] using information on the proportion
with self-rated health status as poor or very poor, assuming
performance status for these patients was similar to that of the
general population (for the low end of the range), and also using
the estimated proportion of patients with FIGO IA1/IA2 stage
disease who received radiotherapy in the data analysis (for the
high end of the range). Because patient performance status was not
recorded in Manitoba, and was not complete in the RWH data, the
stage-specific proportions of patients (with FIGO IB-IIA or FIGO IIBIVA stage disease) with poor performance status (indicated as
ECOG score 3 or over) were estimated via an audit of hospital
records for 75 patients (31 cases with stage IB-IIA and 44 cases
with stage IIB-IVA disease) at the Prince of Wales Hospital (POWH),
New South Wales (NSW), Australia who were diagnosed with
cervical cancer in the period 2000–2007.
Table 1 describes the base-case values and ranges for the
observed proportions of patients with clinical indications as well as
the observed proportions of patients who received each guideline
treatment option for each clinical stage as per the synthesised
guidelines in Australia and Canada. In order to estimate optimal
uptake rates of initial treatments and to quantify differences from
the corresponding observed rates for current practice in the two
settings, proportions of patients receiving treatment in accordance
with the guidelines for each stage is described in Table 2.
2.3. Model structure
A decision tree was constructed to reflect the application of
treatments in accordance with observed current practice as
described in patterns of care studies in Melbourne and Manitoba,
5
as well as to alternately simulate the use of the synthesised
international guidelines for each FIGO stage (and sub-stage) at
diagnosis, if perfect compliance with guidelines could be achieved
(i.e. the optimal rate). The model was constructed to explicitly
reflect each recommended (and alternate) treatment option for
each FIGO stage as per the synthesised international guidelines.
The model considered initial therapy only, because good quality
data on the management of recurrence are not available. Also, in
general, treatment for recurrent disease is more heterogeneous
and more difficult to quantify than initial therapy. That is,
treatment of progressed or recurrent disease will depend on the
previous treatment given, the site or extent of recurrence, the
disease-free interval, and the patient’s performance status at
presentation. In addition, registries do not collect data on
recurrence, so it is impossible to know the population-based
recurrence rates which would be used as a denominator when
assessing the proportion of cases that adhered to guidelines. In the
model, external beam radiotherapy (EBRT) and brachytherapy
were grouped together since the synthesised guidelines recommend EBRT combined with brachytherapy for most cases except
patients with FIGO IA1 stage disease who are medically inoperable,
in whom brachytherapy only is recommended. For some disease
stages (FIGO IB-IIA with small lesions but with a high risk of
recurrence or bulky lesions), either surgery with adjuvant chemoradiotherapy or surgery with adjuvant radiotherapy is recommended. However, comprehensive data on adjuvant chemoradiotherapy were not available from the patterns of care studies
in the two countries, therefore the model did not consider adjuvant
chemo-radiotherapy explicitly and women receiving this treatment were considered together with those receiving surgery with
adjuvant radiotherapy.
Fig. 1 shows a schematic diagram of the decision tree, which
was structured by FIGO stage (IA1, IA2, IB-IIA, IIB-IVA and IVB) to
accommodate stage-specific treatment recommendations. In order
Table 2
Proportions of patients receiving treatment in accordance with the guidelines in (a) Australia and (b) Canada in 1998–2008.
Clinical indications
FIGO sub-stage
(a) Australia
Baseline (range)a
(b) Canada
Baseline (range)a
FIGO IA1
FIGO IA2
FIGO IB-IIA, non bulky, not HR
IA1
IA2
IB1
IB2
IIA
IB1
IB2
IIA
IB1
IB2
IIA
IB1
IB2
IIA
IB1
IB2
IIA
IIB
IIIA
IIIB
IVA
IIB
IIIA
IIIB
IVA
IVB
0.74
0.69
0.68
1.00
0.00
0.69
0.00
0.00
0.69
0.00
0.00
1.00
0.74
0.43
1.00
0.82
0.57
0.63
0.50
0.54
0.33
0.85
1.00
0.71
0.67
0.80
0.69
0.70
0.75
1.00
0.67
0.47
0.00
0.00
0.60
0.00
0.00
0.33
0.81
0.64
0.33
0.81
0.68
0.75
0.50
0.56
0.33
0.95
1.00
0.91
0.83
0.92
FIGO IB-IIA, non bulky, HR, good PS
FIGO IB-IIA, non bulky, HR, poor PS
FIGO IB-IIA, bulky, good PS
FIGO IB-IIA, bulky, poor PS
FIGO IIB-IVA, good PS
FIGO IIB-IVA, poor PS
FIGO IVB
(0.64–0.83)
(0.41–0.89)
(0.58–0.77)
(0.29–1.00)
(0.00–0.00)b
(0.41–0.89)
(0.00–0.00)b
(0.00–0.00)b
(0.32–0.84)
(0.00–0.00)b
(0.00–0.00)b
(0.29–1.00)
(0.56–0.87)
(0.10–0.82)
(0.29–1.00)
(0.65–0.93)
(0.18–0.90)
(0.49–0.76)
(0.19–0.81)
(0.34–0.72)
(0.10–0.65)
(0.72–0.93)
(0.69–1.00)
(0.51–0.87)
(0.35–0.90)
(0.28–0.99)
(0.57–0.79)
(0.50–0.86)
(0.68–0.82)
(0.40–1.00)
(0.22–0.96)
(0.21–0.73)
(0.00–0.00)b
(0.00–0.00)b
(0.32–0.84)
(0.00–0.00)b
(0.00–0.00)b
(0.01–0.91)
(0.54–0.96)
(0.41–0.83)
(0.01–0.91)
(0.54–0.96)
(0.45–0.86)
(0.64–0.85)
(0.07–0.93)
(0.45–0.67)
(0.04–0.78)
(0.87–0.98)
(0.40–1.00)
(0.83–0.96)
(0.36–1.00)
(0.78–0.98)
Non bulky – tumour size in diameter 4 cm; bulky – tumour size in diameter >4 cm; not HR – not at high risk of recurrence (lymph node not present); HR – high risk of
recurrence (lymph node present); PS – performance status; Poor PS – poor performance status (Eastern Cooperative Oncology Group (ECOG) score 3).
a
Baseline is the point estimate for each parameter obtained from patterns of care data analysis and the range was based on the binomial exact 95% confidence intervals
around the point estimate of each parameter. This range was used in sensitivity/uncertainty analysis.
b
No data were available as there were zero cases in this category.
Please cite this article in press as: Kang Y-J, et al. Optimal uptake rates for initial treatments for cervical cancer in concordance with
guidelines in Australia and Canada: Results from two large cancer facilities. Cancer Epidemiology (2015), http://dx.doi.org/10.1016/
j.canep.2015.04.009
G Model
CANEP-854; No. of Pages 12
6
Y.-J. Kang et al. / Cancer Epidemiology xxx (2015) xxx–xxx
Fig. 1. Schematic diagram of decision tree to estimate optimal rates for each type of initial treatment for invasive cervical cancer in Australia and Canada. FIGO – International
Federation of Gynecology and Obstetrics; Tx – treatment; Recommended Tx – treatment according to the guidelines; Other Tx – treatment not according to the guidelines;
LVSI – lymphovascular space invasion; LN – lymph node; HYST – hysterectomy; THYST – total hysterectomy; RHYST – radical hysterectomy; LEEP – Loop electrical excision
procedure; adj. – adjuvant; RT – radiotherapy; CT – chemotherapy; PS – performance status; ECOG – Eastern Cooperative Oncology Group. *The decision tree has seven
different ‘‘Payoffs’’ that represent different types of standard therapy for cervical cancer. These include surgery alone, radiotherapy alone, chemotherapy alone, concurrent
chemoradiotherapy, surgery with either adjuvant or pre-operative radiotherapy, surgery with either adjuvant or neo-adjuvant chemotherapy and no treatment. The payoff
for each terminal branch was set to either 1 (indicating therapy for the selected terminal branch corresponds to the treatment type specified at each payoff) or 0 (indicating
they do not correspond), and the probabilities for the seven different payoffs sum to 1.
to reflect treatment differentials according to the synthesised
international guidelines, the branch representing FIGO IB-IIA
disease was split by tumour size (bulky, non-bulky), and the
branch corresponding to non-bulky disease was further classified
by the risk of recurrence (high risk and not high risk). Although the
guidelines recommend that lymph node and parametrium
involvement as well as positive surgical margins should be used
to determine increased risk of recurrence, in the model lymph node
involvement alone was used as the indicator of high risk of
recurrence as data for the other indications were not available.
Each FIGO stage was constructed with two subsequent modelled
pathways based on the concordance with the recommended
practice (recommended treatment or alternate treatment) that end
with terminal branches specifying mutually exclusive therapies.
Patients with FIGO IB-IVA stage disease with poor performance
status are sub-optimally treated with radiotherapy alone rather
than chemo-radiotherapy. Therefore, the recommended treatment
for patients with FIGO IB-IIA stage disease with non-bulky lesions
(4 cm) and a high risk of recurrence, or with bulky lesions
(>4 cm) and FIGO IIB-IVA stage disease was further assigned
according to the patient’s performance status with branches for
good performance status (ECOG score 0–2) or poor performance
status (ECOG score 3–4) (see Table 1 for treatment types by FIGO
sub-stage).
The decision tree was constructed with seven different ‘‘payoffs’’ that represent different types of standard therapy for cervical
cancer. These include: (1) surgery alone; (2) radiotherapy alone;
(3) chemotherapy alone; (4) concurrent chemo-radiotherapy; (5)
surgery with either adjuvant or pre-operative radiotherapy; (6)
surgery with either adjuvant or neo-adjuvant chemotherapy; and
(7) no treatment. The payoff for each terminal branch was set to
either unity (indicating therapy for the selected terminal branch
corresponds to the treatment type specified at each payoff) or zero
(indicating they do not correspond), and the probabilities for the
seven different payoffs sum to unity. For example, if the treatment
type specified in the terminal branch is surgery alone, then the
terminal branch is set to unity for the payoff corresponding to
surgery alone and zero for the remaining six payoffs. This approach
was used to model multiple treatment types at the same time
including the situation where more than one treatment recommendation occurs for a selected disease stage. Values for the seven
different outcomes were then combined to estimate the overall
rates for: (1) surgery (surgery alone, surgery with either adjuvant
or pre-operative radiotherapy, surgery with either adjuvant or
neo-adjuvant chemotherapy); (2) radiotherapy (radiotherapy
alone, concurrent chemo-radiotherapy, surgery with either adjuvant or pre-operative radiotherapy); (3) chemotherapy (chemotherapy alone, concurrent chemo-radiotherapy, surgery with
either adjuvant or neo-adjuvant chemotherapy); and (4) concurrent chemo-radiotherapy. Therefore, these four treatment types
are not mutually exclusive.
The model was implemented in TreeAge Pro 2008 (Release
1.3.2, TreeAge Software, Inc., MA, USA).
2.4. Estimation of optimal treatment rate
The model was used to estimate the optimal treatment rate in
each setting. The optimal treatment rate is defined as the predicted
proportion of patients receiving a certain type of treatment
(surgery, chemotherapy, radiotherapy or chemo-radiotherapy) in
the patient population diagnosed with a particular cancer type, as a
proportion of all patients, if perfect compliance with the treatment
recommendations was achieved for all patients. The estimated
overall optimal rate for a specific treatment type represents a
Please cite this article in press as: Kang Y-J, et al. Optimal uptake rates for initial treatments for cervical cancer in concordance with
guidelines in Australia and Canada: Results from two large cancer facilities. Cancer Epidemiology (2015), http://dx.doi.org/10.1016/
j.canep.2015.04.009
G Model
CANEP-854; No. of Pages 12
Y.-J. Kang et al. / Cancer Epidemiology xxx (2015) xxx–xxx
weighted average of the recommended rate for that treatment
according to the stage distribution at diagnosis and other factors
important in treatment decisions. We estimated the optimal
treatment rate in Australia and Canada informed by patterns of
care data from a large cancer facility in each country.
7
cervical cancer were calculated by subtracting the estimated
optimal rates from the observed rates using the PSA.
3. Results
3.1. Optimal utilisation rates for initial treatments and comparison
with the observed rates in Melbourne, Australia
2.5. Sensitivity and uncertainty analysis
Both one-way sensitivity analysis and uncertainty analysis using
probabilistic sensitivity analyses (PSA) for selected parameters were
performed in order to examine the level of uncertainty associated
with the outputs from the model. Key clinical factors determining
treatment modality were included in the sensitivity analyses. These
included the overall FIGO stage distribution (IA1, IA2, IB-IIA, IIB-IVA
and IVB); the proportion with bulky IB-IIA tumours; the proportion
at risk of recurrence for FIGO IB-IIA stage disease (high risk or not at
high risk); the proportion of medically inoperable patients with IA1/
IA2 stage disease; and the proportion of patients with FIGO IB-IVA
stage disease and poor performance status.
The binomial exact 95% confidence intervals (CI) estimated
from the observational data were used as the range of uncertainty
for the parameters (when the observed number of cases
corresponding to a parameter was zero, a one-sided exact 97.5%
CI was calculated). These ranges were re-weighted to ensure that
existing dependencies were satisfied (e.g. the lower confidence
limits and the upper confidence limits in multiple branches
summed to 1) [13].
One-way sensitivity analysis was performed by varying one
parameter at a time while other parameters were held at their
baseline. Uncertainty analysis with PSA was performed involving
3000 scenarios, each one involving sampled parameter sets within
the range of uncertainty for the input parameters. Mean
differences and 95% credible intervals (CrI: the range within
which 95% of the estimates fell) for the differences between the
observed and the optimal uptake rates for each initial treatment for
The estimated overall proportions (and 95% CrI) of patients who
should optimally have received surgery, radiotherapy, chemotherapy or chemo-radiotherapy in Australia were 63% (61–64%), 52%
(53–56%), 36% (35–38%) and 36% (35–38%), respectively (Table 3).
The corresponding observed proportions of patients (and 95% CrI)
receiving each treatment in Melbourne were 63% (60–65%), 49%
(46–52%), 27% (24–30%) and 25% (21–28%), respectively (Table 3).
Detailed results for the seven different types of standard therapy
associated with payoffs are presented in the Appendix (Table A.1).
When the optimal rates were subtracted from the observed rates,
the differences (and 95% CrI) in the overall surgery, radiotherapy,
chemotherapy and chemo-radiotherapy rates were 1% ( 1% to 2%),
4% ( 7% to 2%), 9% ( 13% to 5%) and 11% ( 14% to 7%)
(Fig. 2). In general, uncertainty associated with the proportion of
patients with large IB-IIA lesions generated the greatest uncertainty in the estimated percentage of patients who should receive
each treatment (Fig. 3).
3.2. Optimal utilisation rates for initial treatments and comparison
with the observed rates in Manitoba, Canada
The estimated overall proportions (and 95% CrI) of the optimal
rate for surgery, radiotherapy, chemotherapy or chemo-radiotherapy
were 38% (36–39%), 68% (68–71%), 51% (49–52%) and 50% (49–51%),
respectively (Table 3). The corresponding proportions (and 95% CrI) of
patients receiving each treatment observed in Manitoba were 42%
(39–44%), 65% (64–68%), 37% (33–40%) and 36% (31–39%), respec-
Table 3
The estimated percentage of patients who should receive each treatment type according to the guidelines compared with the observed rates from patterns of care studies in
(a) Australia and (b) Canada in 1999–2008.
FIGO stage
Baseline % (95% credible intervala)
Surgeryb
(a) Australia
Overall
IA1
IA2
IB-IIA (Overall)
IB-IIA (4 cm)
IB-IIA (>4 cm)
IIB-IVA
IVB
(b) Canada
Overall
IA1
IA2
IB-IIA (Overall)
IB-IIA (4 cm)
IB-IIA (>4 cm)
IIB-IVA
IVB
Radiotherapyc
Chemotherapyd
Chemo-radiotherapy
Optimal rates
Observed rates
Optimal rates
Observed rates
Optimal rates
Observed rates
Optimal rates
Observed rates
63
98
100
81
100
28
0
0
(61–64)
(96–100)
(95–100)
(79–85)
(99–100)
(27–34)
(0–0)
(0–0)
63
99
100
77
89
43
9
0
(60–65)
(95–100)
(95–100)
(72–81)
(85–92)
(35–50)
(7–11)
(0–0)
52
2
0
52
35
100
100
100
(53–56)
(1–6)
(1–6)
(51–58)
(34–40)
(100–100)
(100–100)
(100–100)
49
1
0
52
40
84
87
80
(46–52)
(0–5)
(0–5)
(47–57)
(36–46)
(74–90)
(84–91)
(30–98)
36
0
0
19
0
72
100
50
(35–38)
(0–0)
(0–0)
(17–23)
(0–0)
(68–75)
(97–99)
(50–50)
27
0
0
22
13
48
60
40
(24–30)
(0–0)
(0–0)
(18–28)
(10–18)
(36–58)
(50–69)
(15–49)
36
0
0
19
0
72
100
50
(35–38)
(0–0)
(0–0)
(17–23)
(0–0)
(68–75)
(97–99)
(50–50)
25
0
0
20
11
48
56
40
(21–28)
(0–0)
(0–0)
(16–25)
(8–15)
(36–58)
(45–66)
(15–49)
38
100
89
56
69
23
0
0
(36–39)
(96–100)
(74–98)
(55–59)
(69–71)
(21–28)
(0–0)
(0–0)
42
100
93
59
66
41
6
6
(39–44)
(96–100)
(75–99)
(55–63)
(64–68)
(34–54)
(5–8)
(1–15)
68
0
11
62
45
100
100
94
(68–71)
(1–6)
(9–32)
(61–65)
(45–48)
(100–100)
(100–100)
(94–94)
65
0
19
57
45
84
98
89
(64–68)
(0–4)
(11–34)
(53–60)
(43–48)
(75–90)
(97–98)
(81–93)
51
0
0
23
0
77
100
36
(49–52)
(0–0)
(0–0)
(21–27)
(0–0)
(74–81)
(97–99)
(36–36)
37
0
0
24
11
56
65
36
(33–40)
(0–0)
(0–0)
(19–29)
(8–14)
(42–65)
(57–72)
(36–36)
50
0
0
23
0
77
100
30
(49–51)
(0–0)
(0–0)
(21–27)
(0–0)
(74–81)
(97–99)
(30–30)
36
0
0
23
10
54
64
28
(31–39)
(0–0)
(0–0)
(18–27)
(7–13)
(39–63)
(56–71)
(24–30)
Note: some of the numbers in the text are slightly different from the results presented in this table due to rounding. These four treatment types are not mutually exclusive.
a
95% credible interval: the range within which 95% of the estimates fell from the probabilistic sensitivity analysis (PSA) involving a total of 3000 scenarios, each one
involving sampled parameter sets within the range of uncertainty for the input parameters provided in Tables 1 and 2.
b
Surgery includes surgery alone, surgery with either adjuvant or pre-operative radiotherapy, surgery with either adjuvant or neo-adjuvant chemotherapy.
c
Radiotherapy includes radiotherapy alone, concurrent chemoradiotherapy, surgery with either adjuvant or pre-operative radiotherapy.
d
Chemotherapy includes chemotherapy alone, concurrent chemoradiotherapy, surgery with either adjuvant or neo-adjuvant chemotherapy.
Please cite this article in press as: Kang Y-J, et al. Optimal uptake rates for initial treatments for cervical cancer in concordance with
guidelines in Australia and Canada: Results from two large cancer facilities. Cancer Epidemiology (2015), http://dx.doi.org/10.1016/
j.canep.2015.04.009
G Model
CANEP-854; No. of Pages 12
Y.-J. Kang et al. / Cancer Epidemiology xxx (2015) xxx–xxx
8
50%
40%
30%
20%
10%
0%
-10%
-20%
-30%
-40%
-50%
(e) Surgery (Canada)
Mean differ
Rangeenc
of eth(PSA)
e differences (PSA)
95% credible interval (PSA)
Over-treated in current pracce
Difference
Difference
(a) Surgery (Australia)
Baselinedifference
difference
Baseline
Mean difference (PSA)
Under-treated in current pracce
Overall
IA1
IA2
IB-IIA
IB-IIA
(Overall) (NB)
FIGO stage
IB-IIA (B) IIB-IVA
50%
40%
30%
20%
10%
0%
-10%
-20%
-30%
-40%
-50%
Baselinediffer
difference
Baseline
ence
Mean difference (PSA)
Under-treated in current pracce
(b) Radiotherapy (Australia)
50%
40%
30%
20%
10%
0%
-10%
-20%
-30%
-40%
-50%
95% credible interval (PSA)
Mean difference (PSA)
Over-treated in current pracce
Under-treated in current pracce
Overall
IA1
IA2
IB-IIA
IB-IIA
(Overall) (NB)
FIGO stage
IB-IIA (B) IIB-IVA
50%
50%
40%
30%
20%
10%
0%
-10%
-20%
-30%
-40%
-50%
IVB
50%
95% credible interval (PSA)
30%
Over-treated in current pracce
20%
-10%
95% credible interval (PSA)
IA2
IB-IIA
IB-IIA
(Overall) (NB)
FIGO stage
IB-IIA (B) IIB-IVA
IVB
Mean Range
difference
of th(PSA)
e differences (PSA)
95% credible interval (PSA)
Mean difference (PSA)
Over-treated in current pracce
10%
0%
-10%
-20%
-20%
-30%
-30%
Under-treated in current pracce
-40%
Under-treated in current pracce
-50%
-50%
Overall
IA1
IA2
IB-IIA
IB-IIA
(Overall) (NB)
FIGO stage
IB-IIA (B) IIB-IVA
Overall
IVB
(d) Chemo-radiotherapy (Australia)
Baselinedifference
difference
Baseline
Mean difference (PSA)
Under-treated in current pracce
IA1
IA2
IB-IIA
IB-IIA
(Overall) (NB)
FIGO stage
IB-IIA (B) IIB-IVA
IA2
IB-IIA
IB-IIA
(Overall) (NB)
FIGO stage
Baselinediffer
difference
Baseline
ence
95% credible interval (PSA)
Over-treated in current pracce
Overall
IA1
IB-IIA (B) IIB-IVA
IVB
(h) Chemo-radiotherapy (Canada)
Mean Range
difference
of th(PSA)
e differences (PSA)
Difference
Difference
Mean differ
Rangeenc
of eth(PSA)
e differences (PSA)
40%
0%
50%
40%
30%
20%
10%
0%
-10%
-20%
-30%
-40%
-50%
IVB
(g) Chemotherapy (Canada)
10%
-40%
IA1
Baselinediffer
diffeence
rence
Baseline
Difference
Difference
20%
IB-IIA (B) IIB-IVA
Under-treated in current pracce
Overall
40%
30%
IB-IIA
IB-IIA
(Overall) (NB)
FIGO stage
Over-treated in current pracce
Mean Range
difference
of th(PSA)
e differences (PSA)
Mean difference (PSA)
IA2
Baselinediffer
diffeence
rence
Baseline
Mean difference (PSA)
(c) Chemotherapy (Australia)
Baselinediffer
diffeence
rence
Baseline
IA1
(f) Radiotherapy (Canada)
Mean Range
difference
of th(PSA)
e differences (PSA)
Difference
Difference
Baselinedifference
difference
Baseline
95% credible interval (PSA)
Over-treated in current pracce
Overall
IVB
Mean differ
Rangeenc
of eth(PSA)
e differences (PSA)
IVB
50%
40%
30%
20%
10%
0%
-10%
-20%
-30%
-40%
-50%
Mean Range
differenc
of eth(PSA)
e differences (PSA)
Mean difference (PSA)
95% credible interval (PSA)
Over-treated in current pracce
Under-treated in current pracce
Overall
IA1
IA2
IB-IIA
IB-IIA
(Overall) (NB)
FIGO stage
IB-IIA (B) IIB-IVA
IVB
Fig. 2. Differences in treatment rates between current practice and the estimated optimal rate in Australia and Canada. Mean differences and 95% credible intervals (the range
within which 95% of the estimates fell) for the differences between the observed and the optimal uptake rates for each initial treatment for cervical cancer were calculated by
subtracting the estimated optimal rates from the observed rates using the probabilistic sensitivity analysis (PSA). PSA involved 3000 scenarios, each one involving sampled
parameter sets within the range of uncertainty for the input parameters (refer to Tables 1 and 2 for the ranges of each parameters used in the PSA). The graphs illustrate
differences up to 50% and do not show the differences for radiotherapy, chemotherapy and chemoradiotherapy for FIGO stage IA1 and IA2; NB – non bulky (<4 cm); B – bulky
(4 cm). These four treatment types are not mutually exclusive.
Please cite this article in press as: Kang Y-J, et al. Optimal uptake rates for initial treatments for cervical cancer in concordance with
guidelines in Australia and Canada: Results from two large cancer facilities. Cancer Epidemiology (2015), http://dx.doi.org/10.1016/
j.canep.2015.04.009
G Model
CANEP-854; No. of Pages 12
Y.-J. Kang et al. / Cancer Epidemiology xxx (2015) xxx–xxx
(a) Opmal surgery rate (Australia)
IB-IIA (Bulky)
Stage distribuon (IB-IIA)
Stage distribuon (IVB)
Medically inoperable (IA1)
Stage distribuon (IA2)
Stage distribuon (Overall)
Stage distribuon (IA1)
Medically inoperable (IA2)
Stage distribuon (IIB-IVA)
Poor PS (IB-IIA_Bulky)
Poor PS (Overall)
IB-IIA (Nonbulky HR)
Poor PS (IIB-IVA)
Poor PS (IB-IIA_Nonbulky_HR)
55%
(e) Opmal surgery rate (Canada)
0.34
0.43
0.03
0.06
0.03
0.20
0.49
0.01
0.00
0.06
0.21
0.06
0.27
0.10
0.22
0.00
0.27
0.08
0.14
0.10
0.20
57%
59%
61%
63%
65%
67%
Stage distribuon (IIB-IVA)
Stage distribuon (Overall)
IB-IIA (Bulky)
Medically inoperable (IA2)
Stage distribuon (IA1)
Stage distribuon (IA2)
Medically inoperable (IA1)
Stage distribuon (IVB)
IB-IIA (Nonbulky HR)
Stage distribuon (IB-IIA)
Poor PS (Overall)
Poor PS (IB-IIA_Nonbulky_HR)
Poor PS (IIB-IVA)
Poor PS (IB-IIA_Bulky)
69%
(b) Opmal radiotherapy rate (Australia)
IB-IIA (Bulky)
IB-IIA (Nonbulky HR)
Stage distribuon (IA2)
Medically inoperable (IA1)
Stage distribuon (Overall)
Stage distribuon (IVB)
Stage distribuon (IA1)
Medically inoperable (IA2)
Stage distribuon (IIB-IVA)
Stage distribuon (IB-IIA)
Poor PS (IB-IIA_Bulky)
Poor PS (Overall)
Poor PS (IIB-IVA)
Poor PS (IB-IIA_Nonbulky_HR)
45%
0.20
0.20
0.06
0.06
0.01
0.21
0.00
0.57
0.43
0.03
0.22
0.06
0.27
0.49
0.10
51%
53%
55%
57%
59%
(c) Opmal chemotherapy rate (Australia)
IB-IIA (Bulky)
Stage distribuon (IB-IIA)
Poor PS (Overall)
Poor PS (IIB-IVA)
Stage distribuon (IA2)
Stage distribuon (IA1)
Stage distribuon (IVB)
Stage distribuon (Overall)
Stage distribuon (IIB-IVA)
Poor PS (IB-IIA_Bulky)
Poor PS (IB-IIA_Nonbulky_HR)
Medically inoperable (IA2)
Medically inoperable (IA1)
IB-IIA (Nonbulky HR)
30%
0.14
0.06
0.22
0.01
32%
0.27
0.00
0.00
0.20
34%
36%
38%
40%
42%
44%
(d) Opmal chemo-radiaon rate (Australia)
IB-IIA (Bulky)
Stage distribuon (IB-IIA)
Poor PS (Overall)
Poor PS (IIB-IVA)
Stage distribuon (IA2)
Stage distribuon (IA1)
Stage distribuon (IVB)
Stage distribuon (Overall)
Stage distribuon (IIB-IVA)
Poor PS (IB-IIA_Bulky)
Poor PS (IB-IIA_Nonbulky_HR)
Medically inoperable (IA2)
Medically inoperable (IA1)
IB-IIA (Nonbulky HR)
30%
0.14
0.06
0.22
0.01
32%
0.27
0.00
0.00
0.20
34%
36%
38%
40%
42%
45%
Stage distribuon (Overall)
IB-IIA (Bulky)
Stage distribuon (IIB-IVA)
Poor PS (Overall)
Poor PS (IIB-IVA)
Stage distribuon (IA2)
Stage distribuon (IB-IIA)
Stage distribuon (IA1)
Stage distribuon (IVB)
Poor PS (IB-IIA_Bulky)
Poor PS (IB-IIA_Nonbulky_HR)
Medically inoperable (IA2)
Medically inoperable (IA1)
IB-IIA (Nonbulky HR)
0.03
0.21
0.03
0.27
0.10
0.10
0.06
0.06
0.08
0.36
0.01
0.15
0.06
0.00
0.07
0.13
0.34
0.10
0.14
0.10
32%
34%
36%
38%
40%
42%
44%
0.36
0.42
0.33
0.13
0.13
0.04
0.06
0.07
0.34
0.10
0.10
0.14
0.25
0.39
0.01
0.05
0.15
0.06
0.00
0.05
0.36
62%
64%
66%
68%
70%
72%
74%
0.25
0.36
0.39
0.42
0.14
0.06
0.36
0.15
0.07
0.10
0.10
0.01
0.00
0.05
0.10
0.04
0.34
0.13
0.10
47%
0.33
0.06
0.13
49%
51%
53%
55%
(h) Opmal chemo-radiaon rate (Canada)
0.34
0.43
0.20
0.49
60%
IB-IIA (Bulky)
Stage distribuon (Overall)
Stage distribuon (IIB-IVA)
Poor PS (Overall)
Poor PS (IIB-IVA)
Stage distribuon (IA2)
Stage distribuon (IB-IIA)
Stage distribuon (IA1)
Stage distribuon (IVB)
Poor PS (IB-IIA_Bulky)
Poor PS (IB-IIA_Nonbulky_HR)
Medically inoperable (IA2)
Medically inoperable (IA1)
IB-IIA (Nonbulky HR)
0.03
0.21
0.03
0.27
0.10
0.10
0.06
0.06
0.08
0.25
0.33
0.13
0.04
0.06
0.05
0.05
0.36
(g) Opmal chemotherapy rate (Canada)
0.34
0.43
0.20
0.49
0.42
Stage distribuon (Overall)
IB-IIA (Bulky)
Stage distribuon (IIB-IVA)
Medically inoperable (IA2)
IB-IIA (Nonbulky HR)
Stage distribuon (IA1)
Stage distribuon (IA2)
Medically inoperable (IA1)
Stage distribuon (IVB)
Stage distribuon (IB-IIA)
Poor PS (IB-IIA_Nonbulky_HR)
Poor PS (IB-IIA_Bulky)
Poor PS (IIB-IVA)
Poor PS (Overall)
0.14
0.10
49%
30%
0.39
(f) Opmal radiotherapy rate (Canada)
0.34
0.08
0.03
0.00
47%
9
44%
45%
0.25
0.36
0.39
0.42
0.14
0.06
0.36
0.15
0.07
0.10
0.10
0.01
0.00
0.05
0.10
0.04
0.34
0.13
0.10
47%
0.33
0.06
0.13
49%
51%
53%
55%
Fig. 3. Tornado diagrams summarising the results from one-way sensitivity analysis on the overall percentage of patients who should receive each initial treatment in
Australia and Canada. HR – high risk of recurrence; PS – performance status. The tornado diagrams illustrate changes in the modelled estimates by varying the selected
parameter input values as shown in the graph one at a time, while the other parameters remained unchanged. The vertical lines in each of the graphs represent the baseline
results. Stage distribution (Overall) used the re-calculated proportions of each of the 95% lower and upper confidence limits for FIGO stage IA1, IA2, IB-IIA, IIB-IVA and IVB as
inputs so that each of the lower and upper confidence limits in multiple branches sum to 1. Poor PS (Overall) used the upper range of the proportions of patients with poor
performance status (FIGO stage IB-IVA) or who are medically inoperable (IA1/IA2).
These four treatment types are not mutually exclusive.
Please cite this article in press as: Kang Y-J, et al. Optimal uptake rates for initial treatments for cervical cancer in concordance with
guidelines in Australia and Canada: Results from two large cancer facilities. Cancer Epidemiology (2015), http://dx.doi.org/10.1016/
j.canep.2015.04.009
G Model
CANEP-854; No. of Pages 12
10
Y.-J. Kang et al. / Cancer Epidemiology xxx (2015) xxx–xxx
tively (Table 3). Detailed results for the seven different types of
standard therapy associated with payoffs are presented in the
Appendix (Table A.2). When the estimated optimal rates were
compared to the observed rates, the differences (and 95% CrI) in the
overall surgery, radiotherapy, chemotherapy and chemo-radiotherapy rates were 4% (3% to 6%), 3% ( 4% to 2%), 14% ( 17% to
10%) and 15% ( 18% to 11%), respectively (Fig. 2). In general,
uncertainty associated with the overall stage distribution, the
proportions of patients with FIGO IIB-IVA stage disease or FIGO IBIIA stage disease with large lesions had the greatest impact on
uncertainty in the estimated percentage of patients who should
receive each treatment (Fig. 3).
3.3. Similarities and differences in the two countries
In Australia and Canada, the estimated overall optimal
treatment rates for each type of treatment differed considerably
(Table 3). The larger proportion of patients diagnosed with early
stage disease in Melbourne compared to Manitoba (26% vs 19%
with FIGO IA disease and 46% vs 35% with FIGO IB-IIA disease)
resulted in the higher optimal surgery rate (63% vs 38%). By
contrast, the smaller proportion of patients diagnosed with
regional/metastatic disease in Melbourne (27% vs 40% with FIGO
IIB-IVA disease and 1% vs 6% with FIGO IVB disease) resulted in the
lower radiotherapy (52% vs 68%), chemotherapy (36% vs 51%) and
chemo-radiotherapy rates (36% vs 50%). However, the observed
utilisation rates within each disease stage were generally
comparable between the two settings (Table 3).
Although the absolute treatment rates varied, the relative
differences between the observed and the estimated optimal rates
were similar in the two settings. In both Melbourne and Manitoba,
the overall surgery rate in current practice was slightly higher than
the estimated optimal rate (1% and 4%), whereas the observed rates
for radiotherapy ( 4% and 3%), chemotherapy ( 9% and 14%)
and chemo-radiotherapy ( 11% and 15%) were lower than the
estimated optimal rates. The largest differences between the
observed and the estimated optimal radiotherapy rates were seen
for FIGO IVB stage disease in both Melbourne and Manitoba ( 20%
and 8%). For chemotherapy, the differences were largest for FIGO
IIB-IVA stage disease in the two settings ( 40% and 35%).
Similarly for chemo-radiotherapy, the differences were largest for
FIGO IIB-IVA stage disease in the two settings ( 44% and 36%)
(Fig. 2 and Table 3, note that some of the numbers in the text
presented here are slightly different from the results presented in
Table 3 due to rounding).
4. Discussion
This study estimated optimal rates for each type of initial
treatment for cervical cancer in Australia and Canada, and
quantified the differences between the estimated optimal and
the observed treatment utilisation rates in a large cancer facility in
each country. The estimated overall optimal rates for each initial
treatment type differed substantially in the two countries, but this
was driven mostly by differences in the stage distribution at
diagnosis. The estimated FIGO stage-specific optimal rates for each
treatment type in the two settings were generally comparable. The
extent of the differences between the two countries appeared to be
similar when the estimated optimal rates were compared to the
observed rates. In both cancer facilities, the observed surgery rates
were very close to, but slightly higher than the optimal rates,
whereas the observed rates for radiotherapy, chemotherapy and
chemo-radiotherapy rates were lower than the estimated optimal
rates. The reasons for the differences in stage distribution at
diagnosis between the two centres were not clear, as the agestandardised incidence rates and the screening rates are very
similar in the two settings. This could be the subject of future
research.
In general, the recommended treatment for locally advanced/
regional cervical cancer is radiotherapy with or without chemotherapy, while surgery is mostly given to patients diagnosed with
early stage disease. The larger proportion of patients diagnosed
with early stage disease in Melbourne compared to Manitoba
resulted in the higher optimal surgery rate, whereas the smaller
proportion of patients diagnosed with locally advanced/regional
stage disease in Melbourne resulted in the lower radiotherapy,
chemotherapy and chemo-radiotherapy rates.
In our study, the range of proportions of cervical cancer patients
treated with any of the possible multiple treatments was based on
the observed distribution obtained from a patterns of care study
in each setting, which represents treatment of choice in a given
setting when more than one treatment type is recommended for a
given sub-stage. As a result, there were some differences in the
estimated optimal rates for specific sub-stages between the two
settings. For example, for stage IB-IIA disease (4 cm without high
risk of recurrence) where both surgery and radiotherapy are
recommended, surgery (with or without adjuvant radiotherapy)
was the predominant treatment in Melbourne, while radiotherapy
was given to one-third of the patients in Manitoba. This resulted in
a higher estimated optimal surgery rate (100% vs 69%) and lower
radiotherapy rate (35% vs 45%) in Australia compared to Canada for
this sub-stage. Similarly, the estimated radiotherapy rate for stage
IA2 disease was lower (0% vs 11%) and the estimated chemoradiotherapy rate for stage IVB disease was higher (50% vs 30%) in
Melbourne compared to Manitoba.
Prior studies, which considered a single type of treatment only
and obtained data from the published literature, have estimated
that the optimal radiotherapy rate in Australia [2] and Canada [4]
were 56% and 63%, respectively. The estimated optimal radiotherapy rates from our model are not substantially different from the
previous studies in each setting (52% and 68%, respectively, as the
first line therapy). The small differences were due to different
assumptions about the proportions of key clinical indicators
(including FIGO stage distribution and the proportion of bulky
FIGO IB-IIA disease), as well as modelling single recommended
treatment in the prior studies versus our ability to model multiple
treatments.
Our study has some limitations due to data availability. We
considered initial therapy only, however it is expected that
treatment for recurrent disease would marginally increase the
overall optimal treatment rates in a given setting. For example,
recurrent disease or symptomatic control affects the number of
treatment episode of radiotherapy, but not the proportion of
patient treated. We used lymph node status alone as the indicator
for being at high risk of recurrence for FIGO IB-IIA stage disease due
to data availability; however, uncertainty associated with this
parameter did not affect the estimated percentage of patients who
should receive each treatment. When the proportion with positive
lymph nodes in FIGO IB-IIA stage disease (4 cm) was varied from
the lower to the upper limit of the range in one-way sensitivity
analysis, the estimated optimal rate for radiotherapy increased by
3% for Australia and 1% for Canada. In general, the estimated rates
for surgery, chemotherapy and chemo-radiotherapy remained
unchanged when the proportion was varied, with the exception of
a 1% increase in the estimated optimal surgery rate for Canada.
Data on adjuvant chemo-radiotherapy were not available, and
therefore this treatment type was not considered explicitly.
Despite some limitations due to data availability, the use of
patterns of care data is preferred over the use of information
obtained from the international literature, which does not take into
account the particular setting in which the optimal treatment rate
is being considered. A limitation of our study is the possibility that
Please cite this article in press as: Kang Y-J, et al. Optimal uptake rates for initial treatments for cervical cancer in concordance with
guidelines in Australia and Canada: Results from two large cancer facilities. Cancer Epidemiology (2015), http://dx.doi.org/10.1016/
j.canep.2015.04.009
G Model
CANEP-854; No. of Pages 12
Y.-J. Kang et al. / Cancer Epidemiology xxx (2015) xxx–xxx
the data sources used for the study, specifically in the Australian
setting, are not representative of the population-based practice
and the sample to which we had access was relatively small. The
Australian data came from one large tertiary cancer facility and
therefore may or may not be more broadly applicable. While the
vast majority of gynaecological oncology work in Australia does get
referred to tertiary centres, the data may not necessarily reflect
overall population-based patterns of care. Most patients who were
referred to this centre were likely to live in major cities; therefore,
the stage distribution observed in these patients as well as
treatment they received might differ from that observed elsewhere
in Victoria. However, population-based data were not available
since: (i) Australian State and Territory Cancer Registries do not
collect any staging information, except for New South Wales that
collects extent of disease at diagnosis; and (ii) detailed information
on radiotherapy and chemotherapy, usually administered on an
outpatient basis, is not routinely collected.
This is the first study to provide detailed modelled estimations
of optimal treatment rates for cervical cancer, that has used
patterns of care data in multiple settings. The study clearly
demonstrates that optimal treatment rates were largely dependent on the overall stage distribution in a given setting; therefore
observers in a different setting should expect different rates of
treatment utilisation. However, it appears that the differences in
the observed and the estimated optimal rates are similar for some
treatment types across different settings. A number of factors may
contribute to suboptimal radiotherapy, chemotherapy and
chemo-radiotherapy utilisation rates, such as limitations in
health care access, patient factors (such as patient’s age and
socio-economic status) and provider factors (including potentially suboptimal referral practices) [14]. We are uncertain of the
drivers for the suboptimal utilisation of treatments, and this
warrants further research. Our estimation of optimal treatment
rates and the gap between the optimal and the observed rates will
assist health care decision makers in evaluating whether a current
level of cancer care resources are adequate to meet patient
demand.
5. Conclusion
The proportions of patients receiving cervical cancer treatment
in accordance with international treatment guidelines were
similar in two large cancer facilities in Australia and Canada. In
both settings, the observed surgery rates were similar to the
estimated optimal rates, whereas radiotherapy, chemotherapy and
chemo-radiotherapy may be somewhat under-utilised compared
with optimal rates. Differences in stage distribution at diagnosis
explained most of the difference in the estimated overall optimal
treatment rates for specific treatment types in the two countries.
The results from this study will assist in setting clinical ‘targets’ for
various treatment modalities for cervical cancer.
Ethics approval
The study obtained human research ethics approval from NSW
Population and Health Services Research Ethics Committee, the
University of Manitoba Health Research Ethics Boards, the
University of Sydney Human Research Ethics Committee, the
Royal Women’s Hospital (RWH) Research Committee and RWH
Human Research Ethics Committee and Cancer Council NSW Ethics
Committee.
Conflict of interests
K.C. is co-PI of an investigator-initiated cervical screening trial
in Australia which is partly funded by Roche Molecular Systems
11
and Ventana Inc., USA. K.C. receives salary support from the
National Health and Medical Research Council Australia (CDF
1007994). E.K. and A.D. have received travel grants from Merck.
E.K. has consulted for Merck and GlaxoSmithKline. Other authors
have no competing interests to declare. Y-J.K. received funding
from Cancer Council NSW for her postgraduate study and has
received a travel grant from the Association of Canadian Studies in
Australia and New Zealand. Y-J.K., K.C. and D.O.’C received the
University of Sydney Health Data Linkage Funding to conduct
record linkage in NSW, Australia. The funding sources had no
involvement in study design, analysis, or interpretation of results,
writing of the manuscript or the decision to submit for
publication.
Authorship contribution
Y-J.K., D.L.C and K.C contributed to the conception and design of
the study and interpreted the findings. Y-J.K. designed and
constructed the decision tree and analysed the data. G.D. and
M.B. provided their radiotherapy model for cervical cancer, and
helped with the design of the study. J.T., N.H., M.J., A.D., R.L. and E.K.
provided the data and helped draft the manuscript. J-B.L. advised
on the model construction and simulation, and also developed a
platform to automate the simulation process and probabilistic
sensitivity analysis. Y-J.K. drafted the manuscript and all authors
read and approved the final manuscript.
Acknowledgements
We thank the NSW Centre for Health Record Linkage for
conducting record linkage. We also thank Mr Robert Walker
(Cancer Council New South Wales), Ms Margot Osinski (the Royal
Women’s Hospital), Ms Grace Musto (CancerCare Manitoba) and
Ms Sharon Lum (the Prince of Wales Hospital) for extracting the
data.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.canep.2015.04.
009.
References
[1] Denny L, Hacker NF, Gori J, Jones III HW, Ngan HYS, Pecorelli S. Staging
classifications and clinical practice guidelines for gynaecologic cancers.
FIGO Committee on Gynecologic Oncology. Int J Gynecol Obstet
2000;70:207–312.
[2] Delaney G, Jacob S, Barton M. Estimation of an optimal radiotherapy utilization
rate for gynecologic carcinoma: part I – malignancies of the cervix, ovary,
vagina and vulva. Cancer 2004;101(4):671–81.
[3] Thompson S, Delaney G, Gabriel GS, Jacob S, Das P, Barton M, et al. Estimation
of the optimal brachytherapy utilization rate in the treatment of carcinoma of
the uterine cervix: review of clinical practice guidelines and primary evidence.
Cancer 2006;107(12):2932–41.
[4] Usmani N, Foroudi F, Du J, Zakos C, Campbell H, Bryson P, et al. An evidencebased estimate of the appropriate rate of utilization of radiotherapy for cancer
of the cervix. Int J Radiat Oncol Biol Phys 2005;63(3):812–27.
[5] Ng W, Jacob S, Delaney G, Barton M. Optimal chemotherapy utilisation in
cervical cancer. Collaboration for cancer outcomes research and evaluation;
2010.
[6] Barton M, Jacos S, Shafiq J, Wong KHW, Thompson S, Delaney G, et al. Review of
optimal radiotherapy utilisation rates (prepared for Department of Health and
Ageing, Australian Government); 2013.
[7] Greater Metropolitan Clinical Taskforce (GMCT).. Best clinical practice: gynaecological cancer guidelines. NSW: NSW Department of Health, 2009.
[8] British Columbia Cancer Agency. Cancer management guidelines: cervical cancer.
Available from: http://www.bccancer.bc.ca/HPI/CancerManagementGuidelines/
Gynecology/UterineCervix1of2/Management/invasive.htm [13/10/2009].
[9] National Cancer Institute. PDQ1 cervical cancer treatment. Bethesda, MD:
National Cancer Institute. Available from: http://www.cancer.gov/
cancertopics/pdq/treatment/cervical/HealthProfessional [23/03/2010].
Please cite this article in press as: Kang Y-J, et al. Optimal uptake rates for initial treatments for cervical cancer in concordance with
guidelines in Australia and Canada: Results from two large cancer facilities. Cancer Epidemiology (2015), http://dx.doi.org/10.1016/
j.canep.2015.04.009
G Model
CANEP-854; No. of Pages 12
12
Y.-J. Kang et al. / Cancer Epidemiology xxx (2015) xxx–xxx
[10] National Comprehensive Cancer Network. Clinical practice guidelines in oncology – v.1.2010 – cervical cancer; 2009, Available at URL: http://www.nccn.
org/professionals/physician_gls/pdf/cervical.pdf.
[11] CancerCare Manitoba. Epidemiology & cancer registry. Winnipeg, Manitoba.
Available from: http://www.cancercare.mb.ca/home/cancer_research/
epidemiology_and_cancer_registry/ [24/12/2011].
[12] Centre for Epidemiology and Research. 2009 summary report on adult
health from the New South Wales Population Health Survey. Sydney: NSW
Department of Health, 2010, Available from: http://www.health.nsw.gov.au/
resources/publichealth/surveys/hsa_09summary.pdf.
[13] Gelman A, Carlin JB, Stern H, Rubin DB. Bayesian data analysis, 2nd ed.
Texts in statistical science series. Boca Raton, FL: Chapman & Hall/CRC,
2004 .
[14] Gillan C, Briggs K, Goytisolo Pazos A, Maurus M, Harnett N, Catton P, et al.
Barriers to accessing radiation therapy in Canada: a systematic review. Radiat
Oncol 2012;7:167.
Please cite this article in press as: Kang Y-J, et al. Optimal uptake rates for initial treatments for cervical cancer in concordance with
guidelines in Australia and Canada: Results from two large cancer facilities. Cancer Epidemiology (2015), http://dx.doi.org/10.1016/
j.canep.2015.04.009