Vous êtes sur la page 1sur 7

Cancer Epidemiology 39 (2015) 126132

Contents lists available at ScienceDirect

Cancer Epidemiology
The International Journal of Cancer Epidemiology, Detection, and Prevention
journal homepage: www.cancerepidemiology.net

Comparison of expected health impacts for major cancers: Integration


of incidence rate and loss of quality-adjusted life expectancy
Mei-Chuan Hung a, Wu-Wei Lai b, Helen H.W. Chen c, Wu-Chou Su d, Jung-Der Wang a,d,e,*
a

Department of Public Health, National Cheng Kung University College of Medicine, Tainan 704, Taiwan
Department of Surgery, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, Tainan 704, Taiwan
c
Department of Radiation Oncology, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, Tainan 704, Taiwan
d
Department of Internal Medicine, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, Tainan 704, Taiwan
e
Department of Occupational and Environmental Medicine, National Cheng Kung University Hospital, Tainan 704, Taiwan
b

A R T I C L E I N F O

A B S T R A C T

Article history:
Received 26 July 2014
Received in revised form 14 November 2014
Accepted 8 December 2014
Available online 30 December 2014

Purpose: The study aims to quantify the expected impacts of different cancers through multiplying the
incidence rate by loss-of-QALE (quality-adjusted life expectancy), with QALY (quality-adjusted life year)
as the common unit, to aid prevention policy decisions.
Methods: 464,722 patients with pathologically veried cancer registered in the Taiwan Cancer Registry
during 19982009 were used to estimate lifetime survival through KaplanMeier estimation combined
with a semi-parametric method. A convenience sample for measuring the utility value with EQ-5D was
conducted with 11,453 cancer patients, with the results then multiplied by the survival functions to
estimate QALE. The loss-of-QALE was calculated by subtracting the QALE of each cancer cohort from the
life expectancy of the corresponding age- and gender-matched reference population. The cumulative
incidence rates from age 20 to 79 (CIR2079) were calculated to estimate the lifetime risk of cancer for
each organ-system.
Results: Liver and lung cancer were found the highest expected lifetime health impacts in males and
females, or expected lifetime losses of 0.97 and 0.41 QALYs that could be averted, respectively. While the
priority changes for prevention based on expected health impacts were slightly different for females
based on standardized mortality rates, those of males involve a broader spectrum, including oral,
colorectal, esophageal and stomach cancer.
Conclusion: The integration of incidence rate with loss-of-QALE could be used to represent the expected
losses that could be averted by prevention, which may be useful in prioritizing strategies for cancer
control.
2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

Keywords:
Health impacts
Incidence rate
Loss-of-QALE (quality-adjusted life
expectancy)
Cancer prevention

1. Introduction
While applications of epidemiologic methods to health policy
decisions were once popular before the 19th century, they have
become less widely used since the beginning of the second half of
the 20th century. Although estimates of standardized mortality
rates can still provide useful information on the impacts of diseases
or technologies, which can then aid health policy decision, such

* Corresponding author at: Department of Public Health, National Cheng Kung


University College of Medicine, No. 1, University Road, Tainan 701, Taiwan.
Tel.: +886 6 2353535x5600; fax: +886 6 2359033.
E-mail address: jdwang121@gmail.com (J.-D. Wang).

comparisons for different illnesses have become less efcient,


especially because for most cancers patients usually survive for
more than 510 years, and thus one must wait for a long period of
time for mortality to occur [16]. The current coding of one
underlying cause of death might also underestimate the impact of
other causes, including cancer. At the same time, the sustainability
of current healthcare systems has become questionable under the
nancial pressures of an aging population and development of new
technologies [7]. From 1971 to 2012, the number of cancer
survivors in the United States increased from 3.0 million to
13.4 million, representing 4.6% of the population [8,9]. As the
population of cancer survivors increases, the economic impact of
cancer for patients, caregivers, employers, the health-care system
and society overall is expected to grow [10,11]. There is thus a need

http://dx.doi.org/10.1016/j.canep.2014.12.004
1877-7821/ 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/
3.0/).

M.-C. Hung et al. / Cancer Epidemiology 39 (2015) 126132

2. Method
2.1. Study population and datasets
The study commenced after gaining approval of the Institutional Review Board of the National Cheng Kung University
Hospital (NCKUH, IRB number: ER-102-034). Written informed
consent was obtained from every interviewed cancer patient. A
total of 464,722 patients with pathologically diagnosed cancer
aged 2079 years at diagnosis were abstracted and veried from
the Taiwan Cancer Registry for 19982009, which contains data on
the patients demographics, date of diagnosis, cancer site, histology
and vital status, and followed up until 2011 [19]. The identication
numbers of all individuals were encrypted to protect their privacy.
The nine major cancers included lung (ICD-9-CM code: 162),
esophagus (ICD-9-CM code: 150), liver (ICD-9-CM code: 155),
stomach (ICD-9-CM code: 151), colorectum (ICD-9-CM code: 153
154), oral (ICD-9-CM code: 140141), nasopharynx (ICD-9-CM
code: 147), cervix (ICD-9-CM code: 180), and breast (ICD-9-CM
code: 174).
2.2. Survival analysis and extrapolation to estimate life expectancy for
different cancers
All of the above patients were linked to the Taiwan Mortality
Registry during 19982011 to obtain their survival functions via
the KaplanMeier estimation method up to the end of the followup period. These were further extrapolated to lifetime based on a
semi-parametric method using the age- and sex-matched referents
simulated from the life tables of the Taiwan Vital Statistics, which
only requires an assumption of constant excess hazards after the
end of follow-up [20,21]. The computation of these estimates were
carried out using iSQoL software [22]. Detailed methods
and mathematical proofs are described in our previous studies
[14,20,21].

2.3. Measurements of quality of life (QOL) via EQ-5D for estimation of


quality-adjusted life expectancies (QALE) and loss-of-QALE
To estimate the utility value of QOL for cancer patients at
different duration-to-dates (namely, from diagnosis of disease up
to the date of interview), we recruited a convenient, cross-sectional
sample with 11,453 measurements aged 2079 years at diagnosis
between 2011 and 2014 at the Oncology Center of NCKUH. Patients
were not included if they had disturbances in consciousness,
moderate to severe dementia or aphasia, or an inability to
communicate with research assistants. We assessed their QOL
by administering the Taiwanese version of EQ-5D [23] through
direct face-to-face interviews. EQ-5D is a preference-based,
generic instrument [24,25] that provides a utility value based
on survey carried out in Taiwan [26]. A kernel-type smoothing
method (using a moving average of the nearby 10%) was performed
to estimate the mean QOL across time [27,28]. The lifetime survival
probabilities over the course of a cancer were multiplied (or
adjusted) with the QOL utility values to obtain a quality-adjusted
survival curve. The loss-of-QALE for these patients was calculated
by subtracting the QALE of cancer patients from the QALE of the
age- and gender- matched referents simulated from the hazard
functions of vital statistics with the EQ-5D utility values borrowed
from those of the 2009 National Health Interview Survey in
Taiwan, as illustrated in Fig. 1 with colorectal cancer as an
example.
2.4. Estimation of cumulative incidence rates (CIR) of major cancers
We used data from the Catastrophic Illnesses Registry database
of Taiwan for the period 20082010 [29] to calculate the incidence
rates of major cancers. The CIR was calculated from age 20 to 79
(CIR2079) to estimate the lifetime risk of a specic cancer, as
follows
i
h X
CIR 1  exp  i IRi Dt i
where IRi represents the age-specic incidence rate and Dti
indicates the range of each age stratum.
2.5. Expected health impacts for cancer
The health impact of cancer at each organ-system was
estimated by multiplying the CIR with loss-of-QALE of nine major
cancers. We rst estimated the lifetime risks by CIR2079, or age
from 20 to 79, followed by CIR2044, and CIR4564, and CIR6579, and
1.0

Age, sex-matched referents ____


0.8

Utility

and opportunity to consider alternative outcome measures when


making policy decisions.
In clinical medicine, QALY (quality-adjusted life years) is a
measure that incorporates both quality of life and length of survival
into one common unit, and this has been widely applied in costutility and/or cost-effectiveness analysis over the last few decades
[1214]. Recently, the authors of this study have developed
estimates of the expected years of life lost and loss-of-QALE that
quantify the consequences of illness with life-year and QALY,
respectively, which make possible direct comparisons of prevention
with clinical care [14,15] if incidence rates can also be incorporated
together [16]. It may thus be possible to quantify both incidence
rates and how much utility is lost from developing a specic illness,
rather than simply making policy decisions based on mortality rates.
Unfortunately, randomized controlled trials are generally
expensive, and the nancial costs for those focused on prevention
are even higher because of the low occurrence rates of most
diseases [17,18]. Therefore, we attempted to use the big data and/
or databases of electronic medical records available in Taiwan to
integrate the likelihood of an event, or incidence rate and
consequence of an event, or loss-of-QALE, to estimate the lifetime
health impacts of major cancers to aid prevention policy
decisions, and further compare the results of this with the
standardized mortality rates in order to decide priorities for
cancer prevention efforts. In other words, we quantied the
lifetime risk (or, cumulative incidence rate for age 2079) and
then multiplied this by the loss-of-QALE to estimate the health
impacts for different cancers, and compared the results with the
related standardized mortality rates.

127

Colorectal cancer

0.6

------

Loss-of-QALE=5.19 QALY

0.4
0.2
0.0

100

200

300

400

500

600

Time in month after diagnosis


Fig. 1. Estimation of loss-of-QALE (quality-adjusted life expectancy): the dashed
line represents the QALE of male patients with colorectal cancer, while the solid line
represents the QALE of corresponding age and sex-matched referents. The area
between solid and dashed lines would be the loss-of-QALE, or 5.19 QALY (qualityadjusted life years).

M.-C. Hung et al. / Cancer Epidemiology 39 (2015) 126132

128

Table 1
Comparison of frequency distributions of cancer patients registered in the Cancer
Registry of Taiwan and a cross-sectional sample at oncology clinics.

Liver, no. case


Sex (% male)
Mean age (years, SD)
Lung, no. case
Sex (% male)*
Mean age (years, SD)*
Oral, no. case
Sex (% male)
Mean age (years, SD)
Colorectum, No. case
Sex (% male)
Mean age (years, SD)
Esophagus, no. case
Sex (% male)
Mean age (years, SD)
Stomach, no. case
Sex (% male)
Mean age (years, SD)
Nasopharynx, no. case
Sex (% male)
Mean age (years, SD)
Cervix, no. case
Mean age (years, SD)
Female breast, no. case
Mean age (years, SD)
*

Cancer
patients (19982009)

Cross-sectional
sample (20112014)

91,747
67,461 (73.5)
60.8 (11.8)
74,720
49,598 (66.4)
64.4 (11.0)
38,662
35,222 (91.1)
52 (11.4)
89,276
51,190 (57.3)
62 (12.2)
15,924
14,931 (93.7)
58 (11.2)
35,720
22,978 (64.3)
62.8 (12.5)
17,118
12,750 (74.5)
49.6 (12.4)
24,525
53.9 (12.8)
77,030
50.9 (11.2)

1612
1182 (73.3)
58.7 (10.4)
3073
1567 (51)
60.9 (10.3)
1559
1445 (92.7)
52 (9.2)
2763
1552 (56.2)
60 (11.3)
284
266 (93.7)
53.4 (9.1)
336
198 (58.9)
58.3 (12.3)
284
214 (75.3)
48 (10.9)
689
50.1 (10)
991
48.7 (9.5)

p < 0.05.

the results were then multiplied by the corresponding expected


loss-of-QALE within different age ranges.
2.6. Validation of the extrapolation method and sensitivity analysis
for loss-of-QALE
To show that our extrapolation method is valid, we selected
sub-cohorts of cancer patients between 1998 and 2004,
extrapolated these results to the end of 2011, and compared
the extrapolated results with the KaplanMeier (KM) estimates
of the actual 14 years of follow-up. Assuming that the KM

estimates are the gold standard, this study calculated the relative
biases for sub-cohorts with different cancers. The relative bias
(RB) is dened as follows: RB = (estimate from extrapolation  K
M estimate)/KM estimate. In addition, we applied a sensitivity
analysis by assuming a uniform utility value of 1 for referents to
test if the nal conclusions of decision making are affected.
2.7. Other statistical analysis
The differences in the frequency distributions of cancer patients
between the cohort and cross-sectional samples at clinics were
tested with the chi-squared and two-sample t test, with
p < 0.05 regarded as signicant. The analysis was carried out
using SAS software (ver. 9.4). The standardized mortality rates
were calculated by adjusting the 2000 world standard population
for nine major cancers for different age periods with the age
specic cancer rates of Taiwan for the period 20082010, stratied
by gender [30].
3. Results
The basic demographic details of the cancer patients are
categorized by cancer sites in Table 1, which shows that the sex
distributions and mean ages at diagnosis of the cross-sectional
sample interviewed at the Oncology Center were similar to those in
the national data, except for lung cancer.
The average estimated loss-of-QALE were the highest for
esophageal cancer in males (16.69 QALYs) and lung cancer in
females (13.11 QALYs) followed by other cancers, while those of
male colorectal cancer and female cervical cancer had the lowest
estimates, at 5.19 and 3.17 QALYs, respectively (Table 2). After
multiplication with the incidence rates in Taiwan for the specic
cancer sites during 20082010, we found the highest expected
health impacts in males and females were liver and lung cancer,
with expected losses of 0.97 and 0.41 QALYs, respectively. As
summarized in Fig. 2, the priorities for prevention based on
expected health impacts are different from those based on
standardized mortality rates. While the priority changes in females
may be less affected except breast and colorectal cancers, those of
males involve a broader spectrum, including oral, colorectal,

Table 2
Comparison of expected lifetime impacts and standardized mortality rates adjusted by 2000 world standard population for major cancers in Taiwan stratied by gender:
expected lifetime risk (CIR2079)a, loss-of-QALE (quality-adjusted life expectancy) under unit of QALY (quality-adjusted life year).
Sites of
cancer

Gender

20082010
(CIR2079)a (%)

Loss-of-QALE

Loss-of-QALE
under lifetime risk

Loss-of-QALE
(with referents
utility as 1)

Loss-of-QALE under
lifetime risk
(with referents utility as 1)

20082010 standardized
mortality rates2079b

Liver

Male
Female

6.80
3.12

14.22 (0.03)
11.44 (0.07)

0.97
0.36

17.2 (0.04)
16.01 (0.07)

1.17
0.50

37.53  105
14.33  105

Lung

Male
Female

6.05
3.15

11.66 (0.03)
13.11 (0.05)

0.71
0.41

14.29 (0.03)
18.14 (0.04)

0.86
0.57

35.22  105
16.49  105

Oral

Male
Female

2.99
0.38

13.54 (0.12)
5.07 (0.51)

0.40
0.02

16.68 (0.1)
10.68 (0.53)

0.50
0.04

14.56  105
0.96  105

Colorectum

Male
Female

6.30
4.36

5.19 (0.08)
5 (0.19)

0.33
0.22

8.05 (0.08)
10.1 (0.16)

0.51
0.44

17.04  105
11.84  105

Esophagus

Male
Female

1.55
0.12

16.69 (0.06)
11.41 (0.78)

0.26
0.01

19.64 (0.07)
16.39 (0.66)

0.30
0.02

9.35  105

Stomach

Male
Female

2.08
1.13

8.26 (0.16)
8.14 (0.19)

0.17
0.09

11.05 (0.15)
13.33 (0.19)

0.23
0.15

9.49  105
4.98  105

Nasopharynx

Male
Female

0.88
0.29

13.75 (0.27)
11.54 (1.67)

0.12
0.03

17.1 (0.31)
17.28 (1.49)

0.15
0.05

3.73  105

Female breast
Cervix

Female
Female

6.52
1.37

5.22 (0.16)
3.17 (0.26)

0.34
0.04

10.8 (0.16)
8.76 (0.24)

0.70
0.12

10.61  105
4.39  105

a
b

CIR2079, cumulative incidence rates from age 20 to 79.


Cited from: https://cris.hpa.gov.tw/pagepub/Home.aspx?itemNo=cr.q.50.

M.-C. Hung et al. / Cancer Epidemiology 39 (2015) 126132

129

Fig. 2. Comparison of priority ratings for cancer prevention policy decisions between lifetime expected loss-of-QALE (quality-adjusted life expectancy) using the QALY
(quality-adjusted life years) as the common unit and standardized mortality rate adjusted for the 2000 world standard population stratied by gender.

esophageal and stomach cancers (see also Table 2). The above
trends remained unchanged for the age groups of 4564 after
stratication. Lung cancer became the leading expected loss of
0.272 QALYs in the age group of 6579 (Table 3).
The results of the validation of the extrapolation method for
estimates of survival for each cancer subcohort showed that all the
relative biases were less than 6.82%, despite the high censoring
rates (more than 80% for breast cancer), as summarized in the
Appendix 1. We expect that the possible bias would be even less
after 14 years of follow-up, because the life expectancy for each
cancer was usually below 15 years, except breast and cervical
cancer. If we assume a uniform utility value of one for referents in
the sensitivity analysis of loss-of-QALE, the priority setting for
prevention based on expected health impacts would be quite
different, with breast cancer becoming the top priority for cancer
control in females (Table 2).

4. Discussion
This study is the rst that integrates incidence rates with lossof-QALE to estimate the health impacts and cancer risks to aid
cancer prevention policy decisions using QALY (quality-adjusted
life years) as the common unit. The results show that liver and lung
cancer had the highest expected health impacts in males and
females, or expected losses of 0.97 and 0.41 QALY, respectively. The
priorities for prevention based on expected health impacts are
different from those based on standardized mortality rates, with
oral, colorectal, esophageal and stomach cancer becoming the third
to sixth priority in cancer control for males (Table 2 and Fig. 2).
However, we must carefully examine the accuracy of our estimates
before making further inferences. First, since we only included
patients with the nine cancers veried with histopathological
evidence (except liver cancer) [31], the diagnoses were all valid.

M.-C. Hung et al. / Cancer Epidemiology 39 (2015) 126132

130

Table 3
Estimation of expected risk, loss-of-QALE (quality-adjusted life expectancy) for an average case and expected loss-of-QALE for different major cancers stratied by age
periods.
Sites of cancer

20082010
(CIR2044)

Loss-of-QALE
(2044 years)

Expected loss-ofQALE for


CIR2044

20082010
(CIR4564)

Lossof-QALE
(4564 years)

Expected lossof-QALE for


CIR4564

20082010
(CIR6579)

Loss-ofQALE
(6579 years)

Expected lossof-QALE for


CIR6579

Liver
Lung
Oral
Colorectum
Esophagus
Stomach
Nasopharynx
Female breast
Cervix

0.0020
0.0011
0.0027
0.0021
0.0006
0.0006
0.0014
0.0118
0.0019

28.31 (0.14)
29.53 (0.17)
21.42 (0.29)
16.03 (0.33)
29.3 (0.21)
21.59 (0.54)
18.08 (0.87)
13.49 (0.44)
5.22 (0.23)

0.057
0.032
0.058
0.034
0.018
0.013
0.025
0.159
0.010

0.0142
0.0104
0.0086
0.0143
0.0039
0.0046
0.0027
0.0352
0.0059

17.21
17.85
12.91
7.33
19.33
11.42
12.62
4.98
5.82

0.244
0.186
0.111
0.0143
0.075
0.053
0.034
0.175
0.034

0.0322
0.0326
0.0057
0.0355
0.0037
0.0147
0.0017
0.0211
0.0060

8.15
8.34
5.04
3.27
8.34
5.69
6.42
0.92
2.42

0.262
0.272
0.029
0.116
0.031
0.084
0.011
0.019
0.015

(0.05)
(0.05)
(0.12)
(0.12)
(0.09)
(0.12)
(0.44)
(0.19)
(0.32)

(0.04)
(0.03)
(0.14)
(0.07)
(0.07)
(0.2)
(0.38)
(0.44)
(0.57)

CIR, cumulative incidence rates.

Second, all the extrapolations of survival functions are based on the


validated assumption of constant excess hazard after the end of
follow-up, which can be obtained by showing a straight line after
taking the logit transform of the survival ratio between the index
and age- and gender-matched referents [20,21]. In addition, since
our validation study of this method showed that the relative biases
were all below 6.8% after seven years of follow-up, it is reasonable
to assume that those of the estimation of life expectancies based on
14 years of follow-up would be even smaller. Third, we collected
the QOL measure using the EQ-5D from all patients attending the
oncology clinic of NCKUH, and obtained their utility values at
different time points after diagnosis to integrate with the survival
function at time t. In total we excluded approximately 1% of the
invited patients who were unable to communicate with research
assistants, plus 9% who refused to be interviewed, and the gender
and age distributions at diagnosis seemed similar to the national
data (Table 1). Therefore, our convenience samples might not be
biased too much, or at least are more accurate than simply making
inferences based on the mean values. Finally, CIR is an indicator
that is inherently standardized for the specied age period, and
there is no need to invoke any outside weights for comparison. In
addition, everyones QALY would be the same across different
illnesses and technologies, and could also be directly compared
with clinical care services [32]. The combination of these two
indicators would thus be more efcient in quantifying the loss of
health after the occurrence of a specic illness. We thus tentatively
conclude that our results are relatively accurate and useful for
assisting policy decisions related to cancer prevention.
Based on the above analysis, oral and esophageal cancer became
the third and fth priority in cancer control for males followed by
colorectal and stomach cancer, especially for the age group of 45
64, owing to its high loss-of-QALE; breast cancer became the third
priority for females followed by colorectal cancer, because of its
high CIR and loss-of-QALE. Had we assumed a uniform utility value
of 1 for the referents simulated from the general population, the
priority for prevention based on expected health impacts would
have been quite different, with breast cancer becoming the top
priority for cancer control in females because of their long
survivorship (Table 2). Because cancer patients may die of other
causes, the health effects attributed to cancer would be underestimated based on standardized mortality rate. In the U.S., an
estimated 64% of cancer survivors live for more than ve years
after their diagnosis [33], and this usually requires continual
monitoring with regard to recurrence and long-term care for the
late effects of cancer [11,33], issues that must be addressed when
developing of cancer control policies. Therefore, counting the
potential savings with regard to loss-of-QALE, costs of lifetime
healthcare and long term care, and productivity loss for society
[34] when estimating the total health impact would be more
comprehensive and efcient than considering the standardized

mortality rate or CIR alone, and it would be better if future


decisions cancer prevention policies would take all these potential
impacts into consideration.
5. Limitations
Our study has following limitations: rst, the lifetime
extrapolation is based on current and prior experiences, especially
the national life tables. However, such an ex post approach could
easily underestimate the actual survival of future cancer populations, because of the development and adoption of newer
technologies for cancer care. Therefore, our estimation of the
lifetime survival of cancer patients may be a conservative one,
while the loss-of-QALE and lifetime health impacts might be
overestimated. Second, during the lifetime extrapolation of the
QOL function, it was assumed that the patients continued to stay at
the same level of QOL as near the end of follow-up. This
assumption could have resulted in an overestimation, because
the real QOL might gradually decline as the patient ages [35]. Third,
as most of interviewed subjects were recruited from clinics, their
general conditions were probably better than those conned at
home or institutions, which might have led to over-estimating the
QOL and under-estimating the prevention effects.
6. Conclusion
The integration of incidence rates with loss-of-QALE could be
used to represent the expected losses that could be averted by
prevention. This estimate can become a summary index of cost
effectiveness for cancer prevention, and may be useful in
prioritizing strategies for cancer control after considering other
related parameters.
Competing interests
The authors declare that they have no competing interests.
Authorship contribution
Conceived and designed the experiments: J.-D.W., M.-C.H.,
W.W.L., H.H.W.C. and W.C.S. Performed the data analysis: M.-C.H.
Wrote the paper: M.-C.H., J.-D.W. All authors read and approved
the nal manuscript.
Acknowledgements
This work was supported by grants from the National Science
Council of Taiwan (Grant Nos. NSC 102-2314-B-006-029-MY2, NSC
102-2811-B-006-018 and NSC 101-314-Y-006-001), Ministry of

M.-C. Hung et al. / Cancer Epidemiology 39 (2015) 126132

Health and Welfare of Taiwan (DOH 102-TD-C-111-003) and the


Top University Project to the National Cheng Kung University from
the Ministry of Education of Taiwan. The funding bodies had no
role in the study design, data collection and analysis, decision to

131

publish, or preparation of the manuscript. We are grateful to all the


patients who participated in this study, and to Drs. Jenq-Chang Lee,
Yih-Jyh Lin, Jenn-Ren Hsiao, Ya-Min Cheng and Yan-Shen Shan for
helping us in the collection of QOL data from their clinics.

Appendix 1
Estimates of mean survival after 14 years of follow-up using the semi-parametric method of extrapolation based on the rst seven years
of follow-up data with a high censored rate and compared with KaplanMeier (KM) estimates of 14 years of follow-up for patients with
major cancers

14-year survival
based on KM estimate
(SE) (month)

Extrapolation from
rst 7 years of follow
up (SE) (month)

Relative bias (%)a

Cancer

Gender

Cohort size

Age at diagnosis (SD)

Censored rate (%)

Lung

Male
Female

30,231
13,756

67.92 (11.39)
64.59 (13.1)

23.03
29.68

1.98 (0.02)
2.57 (0.04)

2.08 (0.01)
2.52 (0.02)

4.83
2.07

Esophagus

Male
Female

7515
624

60.29 (12.6)
68.36 (13.24)

26.88
33.33

2.25 (0.04)
3.53 (0.21)

2.4 (0.02)
3.61 (0.08)

6.82
2.19

Liver

Male
Female

38,770
14,275

60.27 (13.39)
65.59 (12.63)

30.15
33.22

2.71 (0.02)
2.92 (0.03)

2.65 (0.02)
2.83 (0.02)

2.12
3.23

Stomach

Male
Female

16,160
8418

67.23 (13.17)
63.58 (15.2)

42.57
47.96

4.53 (0.05)
5.43 (0.06)

4.56 (0.03)
5.41 (0.03)

0.74
0.43

Colorectum

Male
Female

29,017
21,856

64.96 (13.29)
64.03 (14.18)

63.14
65.67

6.83 (0.03)
7.44 (0.04)

6.82 (0.02)
7.32 (0.02)

0.18
1.56

Oral

Male
Female

17,310
1839

51.57 (11.99)
58.74 (15.63)

63.26
68.24

6.95 (0.05)
8.02 (0.14)

7.01 (0.02)
8.17 (0.07)

0.86
1.90

Nasopharynx

Male
Female

7329
2553

49.63 (13.21)
48.77 (13.72)

69.04
75.52

7.85 (0.07)
9.27 (0.11)

7.71 (0.04)
8.88 (0.05)

1.71
4.14

Breast
Cervix

Female
Female

3,6874
17,269

50.63 (12.16)
55.26 (14.28)

87.61
78.82

11.03 (0.03)
9.96 (0.04)

10.54 (0.01)
10.01 (0.02)

4.39
0.51

SD, standard deviation; SE, standard error of mean.


a
Relative bias = (estimate from extrapolation  KM estimate)/KM estimate.

References
[1] Liu PH, Wang JD, Keating NL. Expected years of life lost for six potentially
preventable cancers in the United States. Prev Med 2013. http://dx.doi.org/
10.1016/j.ypmed.
[2] Chu PC, Wang JD, Hwang JS, Chang YY. Estimation of life expectancy and the
expected years of life lost in patients with major cancers: extrapolation of
survival curves under high-censored rates. Value Health 2008;11:11029.
[3] Hung MC, Liu MT, Cheng YM, Wang JD. Estimation of savings of life-years and
cost from early detection of cervical cancer: a follow-up study using nationwide
databases for the period 20022009. BMC Cancer 2014;14:505.
[4] Yang SC, Lai WW, Su WC, Wu SY, Chen Helen HW, Wu YL, et al. Estimating the
lifelong health impact and nancial burdens of different types of lung cancer.
BMC Cancer 2013;13:579.
[5] Arrington AK, Goldstein L, Kruper L, Vito C, Yim J, Chen SL. Life expectancy after
curative-intent treatment of breast cancer: impact on long-term follow-up
care. Am Surg 2014;80(6):6049.
[6] Gross CP, McAvay GJ, Krumholz HM, Paltiel AD, Bhasin D, Tinetti ME. The effect
of age and chronic illness on life expectancy after a diagnosis of colorectal
cancer: implications for screening. Ann Intern Med 2006;145(9):64653.
[7] Black N. Can Englands NHS survive? N Engl J Med 2013;369:1.
[8] Howlander N, Noone AM, Krapcho M, Garshell J, Miller D, Altekruse SF, et al.,
eds. SEER cancer statistics review, 19752011. Bethesda, MD: National Cancer
Institute, 2014. Available at http://seer.cancer.gov/statfacts/html/all.html.
[9] De Moor JS, Mariotto AB, Parry C, Alfano CM, Padgett L, Kent EE, et al. Cancer
survivors in the United States: prevalence across the survivorship trajectory
and implications for care. Cancer Epidemiol Biomark Prev 2013;22:56170.
[10] Institute of Medicine Delivering Affordable Cancer Care in the 21st Century:
Workshop Summary; 2014, Available at: http://www.nap.edu/catalog.
php?record_id=18273 [cited 16.06.14.].
[11] Ekwueme DU, Yabroff KR, Guy Jr GP, Banegas MP, de Moor JS, Li C, et al.
Division of Cancer Prevention and Control, National Center for Chronic Disease
Prevention and Health Promotion, CDC. MMWR Morb Mortal Wkly Rep
2014;63(23):50510.
[12] Weinstein MC, Torrance G, McGuire A. QALYs: the basics. Value Health
2009;12(Suppl. 1):S59.
[13] Lipscomb J, Drummond M, Fryback D, Gold M, Revicki D. Retaining and
enhancing the QALY. Value Health 2009;12(Suppl. 1):S1826.

[14] Kind P, Lafata JE, Matuszewski K, Raisch D. The use of QALYs in clinical and
patient decision-making: issues and prospects. Value Health 2009;12(Suppl.
1):S2730.
[15] Burnet NG, Jefferies SJ, Benson RJ, Hunt DP, Treasure FP. Years of life lost (YLL)
from cancer is an important measure of population burden and should be
considered when allocating research funds. Br J Cancer 2005;92:2415.
[16] Raiffa H. Decision analysis introductory lectures on choices under uncertainty.
America: Addison-Wesley Publishing Company, 1968.
[17] Bridges CB, Thompson WW, Meltzer MI, Reeve GR, Talamonti WJ, Cox NJ, et al.
Effectiveness and cost-benet of inuenza vaccination of healthy working
adults: a randomized controlled trial. JAMA 2000;284(13):165563.
[18] Herrero R, Quint W, Hildesheim A, Gonzalez P, Struijk L, Katki HA, et al.
Reduced prevalence of oral human papillomavirus (HPV) 4 years after bivalent
HPV vaccination in a randomized clinical trial in Costa Rica. PLoS ONE
2013;8(7):e68329.
[19] Taiwan Cancer Registry Cancer Registry Dataset; 2014, Available at: http://tcr.
cph.ntu.edu.tw/main.php?Page=N1 [accessed 11.07.14.].
[20] Hwang JS, Wang JD. Monte Carlo estimation of extrapolation of qualityadjusted survival for follow-up studies. Stat Med 1999;18:162740.
[21] Fang CT, Chang YY, Hsu HM, Twu SJ, Chen KT, Lin CC, et al. Life expectancy of
patients with newly diagnosed HIV infection in the era of highly active
antiretroviral therapy. QJM 2007;100:97105.
[22] iSQoL Package; 2014, Available at: http://www.stat.sinica.edu.tw/isqol/
[accessed 11.07.14.].
[23] Chang TJ, Tarn YH, Hsieh CL, Liou WS, Shaw JW, Chiou XG. Taiwanese version of
the EQ-5D: validation in a representative sample of the Taiwanese population.
J Formos Med Assoc 2007;106(12):102331.
[24] EuroQol Group. EuroQol-a new facility for the measurement of health-related
quality of life. Health Policy 1990;16(3):199208.
[25] Szende A, Oppe M, Devlin N. EQ-5D value sets: inventory, comparative review
and user guide. Dordrecht, the Netherlands: Springer, 2007.
[26] Lee HY, Hung MC, Hu FC, Chang YY, Hsieh CL, Wang J-D, et al. Estimating
quality weights for EQ-5D health states with the time trade-off method in
Taiwan. J Formos Med Assoc 2013;112(11):699702.
[27] Hwang JS, Tsauo JY, Wang JD. Estimation of expected quality adjusted survival
by cross sectional survey. Stat Med 1996;15(1):93102.
[28] Hwang JS, Wang JD. Integrating health prole with survival for quality of life
assessment. Qual Life Res 2004;13(1):110.

132

M.-C. Hung et al. / Cancer Epidemiology 39 (2015) 126132

[29] National Health Research Institutes in Taiwan: National Health Insurance


Research database; 2003, Available at: http://www.nhri.org.tw/nhird/
date_01.html#_edn1 [accessed 11.07.14.].
[30] Health Promotion Administration, Ministry of Health and Welfare of Taiwan.
Standardized mortality rates of cancer; 2010, Available at: https://cris.hpa.
gov.tw/pagepub/Home.aspx?itemNo=cr.q.50 [in Chinese].
[31] Bruix J, Sherman M. Management of hepatocellular carcinoma: an update.
Hepatology 2011;53:10202.
[32] Weinstein MC. A QALY is a QALY is a QALY or is it. J Health Econ 1988;728990.

[33] Centers for Disease Control and Prevention. Cancer Survivors United States,
2007. MMWR Morb Mortal Wkly Rep 2011;60:26972.
[34] Neiger BL, Thackeray R, Fagen MC. Basic priority rating model 2.0: current
applications for priority setting in health promotion practice. Health Promot
Pract 2011;12(4):630.
[35] Fryback DG, Dunham NC, Palta M. US norms for six generic health-related
quality-of-life indexes from the National Health Measurement Study. Med
Care 2007;45:116270.