Cancer Epidemiology 43 (2016) 76–86

Contents lists available at ScienceDirect

Cancer Epidemiology
The International Journal of Cancer Epidemiology, Detection, and Prevention

journal homepage: www.cancerepidemiology.net

Data quality at the Singapore Cancer Registry: An overview of

comparability, completeness, validity and timeliness
Janice Wing Mei Funga,* , Sandra Bee Lay Lima , Huili Zhenga , William Ying Tat Hoa ,
Bee Guat Leea , Khuan Yew Chowa , Hin Peng Leeb
a
National Registry of Diseases Office, Singapore Cancer Registry, Health Promotion Board, 3 Second Hospital Avenue, Singapore 168937, Singapore
b
Saw Swee Hock School of Public Health, National University of Singapore, 12 Science Drive 2 #10-01 Singapore 117549, Singapore

A R T I C L E I N F O A B S T R A C T

Article history: Aim: To provide a comprehensive evaluation of the quality of the data at the Singapore Cancer Registry
Received 4 March 2016 (SCR).
Received in revised form 21 June 2016 Methods: Quantitative and semi-quantitative methods were used to assess the comparability,
Accepted 23 June 2016
completeness, accuracy and timeliness of data for the period of 1968–2013, with focus on the period
Available online 9 July 2016
2008–2012.
Results: The SCR coding and classification systems follow international standards. The overall
Keywords:
completeness was estimated at 98.1% using the flow method and 97.5% using the capture-recapture
Cancer registry
Data quality
method, for the period of 2008–2012. For the same period, 91.9% of the cases were morphologically
Neoplasms verified (site-specific range: 40.4–100%) with 1.1% DCO cases. The under-reporting in 2011 and 2012 due
Comparability to timely publication was estimated at 0.03% and 0.51% respectively.
Completeness Conclusion: This review shows that the processes in place at the SCR yields data which are internationally
Validity comparable, relatively complete, valid, and timely, allowing for greater confidence in the use of quality
Accuracy data in the areas of cancer prevention, treatment and control.
Timeliness ã 2016 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
2. Material . . . . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Sources of notification .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3. Methods . . . . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.1. Comparability . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.2. Completeness . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.3. Validity . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.4. Timeliness . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4. Results . . . . . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.1. Comparability . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.2. Completeness . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.3. Validity . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.4. Timeliness . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5. Discussion . . . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
6. Conclusion . . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Conflict of interest . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Authorship contribution . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
References . . . . . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

* Corresponding author.
E-mail addresses: janice_fung@hpb.gov.sg, wingmei.fung@gmail.com
(J.W.M. Fung).

http://dx.doi.org/10.1016/j.canep.2016.06.006
1877-7821/ã 2016 Elsevier Ltd. All rights reserved.
 J.W.M. Fung et al. / Cancer Epidemiology 43 (2016) 76–86
1. Introduction
Established in 1968, the Singapore Cancer Registry (SCR) is a
population-based national registry with a repository of informa-
tion on borderline, in-situ and malignant tumours and benign
neoplasms which were diagnosed or treated in Singapore, with
data collected by all available sources nation-wide [1]. Singapore is
a multi-ethnic country with the Chinese, the Malays and the
Indians being the three main ethnic groups in its resident
population, which includes Singapore citizens and permanent
residents. Residents comprised 94.5% of the total population in
1980, 81.2% in 2000, and 71.2% in 2013 due to a large influx of
foreign workers in the past decades [2,3]. The SCR registers all
cases of cancer diagnosed in Singapore, but for this study, the
incidence rates pertain only to Singapore’s resident population.
The registry data is used for international and national bench-
marking, guiding policy direction and in research [4]. The high
quality of data is hence critical for accurate decision making [5,6].
The reporting of neoplasms has been mandatory since 2009,
when the Ministry of Health enacted the National Registry of
Diseases (NRD) Act [7] to provide statutory regulations on
reporting. All hospitals, laboratories, and healthcare institutions
in Singapore were required to report cancer cases [1,7,8]. Prior to
this, cancer notifications were voluntary.
This study presents the first comprehensive evaluation of the
data quality of SCR in line with the evaluation guidelines published
by the International Agency for Research on Cancer (IARC) [9] and
techniques covered by Bray and Parkin [10,11]. The quality of the
data collected by the SCR will be assessed in the four indicative
areas of comparability, completeness, validity and timeliness.
2. Material
Sources of notification
The sources of notification for the SCR had grown increasingly
comprehensive since its establishment in 1968. Registrations were
initially based on voluntary notifications received from clinicians
[12]. From 2001 onwards, the pool of sources was enlarged through
data obtained from pathology records, hospital discharge summa-
ries and death certificates [2]. This was enhanced by the enactment
of the NRD Act in 2009 which mandated medical professionals to
report all histologically verified and clinically diagnosed cancer
incidents within three months from the date of diagnosis, using a
structured notification form [1,8].
Modes of notification include usage of hardcopy forms or via an
online national healthcare portal [4], with monthly tracking done
by registry staff. A secure electronic system also facilitates the
information transfer of the Hospital Inpatients Discharge Summary
(HIDS) listings. As a majority of cancer cases are diagnosed
histologically, the SCR receives monthly pathology reports from all
public and private laboratories. Haematological, radiotherapy, and
nuclear medicine information from treatment centres are also sent
to the SCR on a regular basis. The workflow of the notification
process is illustrated in Fig. 1.
Patient data are collected using the unique national identity
number issued to Singapore residents. For non-residents, their
passport numbers or foreign identification numbers (FIN) are used
[1]. These unique numbers minimize the risk of potential
duplication of records in the SCR database amidst multiple
notification sources.
Data are captured via electronic transmission from hospitals to
the registry using the National Registry of Diseases System (NRDS)
since 2004, and additional information is collected manually from
hardcopy medical records. During the data collection process,
verification of information for accuracy and completeness is
77
performed manually by a team of nine registry staff and a visiting
consultant pathologist [13]. Retrospectively, a separate team of
staff perform annual audits on the data collected to ensure that the
level of data accuracy is at least 95%.
Mortality data of all patients are obtained from the National
Death Registry on a two-monthly basis via secured file transfer
channels. The certification of death is virtually complete in
Singapore [13].
3. Methods
3.1. Comparability
Comparability is the extent of which coding and collection
practices adhere to international guidelines [14]. In this study,
comparability is examined by reviewing the registry’s standards,
definition and practices in incidence date and basis of diagnosis,
topography and morphology, coding of multiple primaries, and
staging [15].
3.2. Completeness
Completeness is the extent to which all diagnosed cancer cases
in Singapore are captured in the registry database [10,14–16]. This
allows us to determine whether incidence rates and survival
proportions reflect their true values. Taking reference from Parkin
and Bray [10], we used the historic data approach (stability of
incidence rates over time (1968–2013), shape of age-specific
curves for selected cancers (2008–2012), and age-specific inci-
dence rates of childhood cancer (2008–2012) compared with the
reference deciles for childhood cancer published in the Cancer
Incidence in Five Continents (CI5), Volume X [9]). Segi’s world
population was used for direct standardisation in calculating age-
standardised rates. Other semi-quantitative methods used include
the mortality/incidence (M:I) ratio (2008–2012) versus one minus
five years relative survival (2003–2012), and the number of
notifications per case (2008–2012) [10,15,17].
Quantitative methods included the capture-recapture method
(2008–2012) and the flow method for cases diagnosed in 2008 and
followed up until 31 December, 2012. The capture-recapture
method assumes that all the notification sources are independent
of each other and all patients have the same probability of being
captured in the Registry. However, these assumptions are likely to
be violated in cancer registration. The notification sources can be
broadly classified into: clinical sources, pathological sources and
death certificates. When substantially two of three groups are
dependent on each other, it is possible to correct for the
dependency between the sources by pooling the two groups into
one broad group and compare it with the remainder group by a
two-way capture-recapture method [18,19].
The flow method models the flow of individuals through the
case ascertainment process from diagnosis to registration, taking
into account of the time since diagnosis [20]. Two datasets were
used to run the flow method: cases diagnosed in 2008 and cancer
cases who died in 2012. Using the flow method, we computed the
probability of patient being registered, missing or lost, and we
plotted the probability of patient being registered over the years
since diagnosis.
All statistical analysis were performed using Stata SE Version
13.
3.3. Validity
Validity refers to the accuracy of the data, and is defined by Bray
& Parkin as the proportion of cases in the registry with a given
characteristic which truly have this attribute [11]. The proportions of
78 J.W.M. Fung et al. / Cancer Epidemiology 43 (2016) 76–86

Cancer notifications HIDS / Cancer listings Listings from Death

by physicians Pathology reports from HCIs e.g. NCC and registry / MHA
/nurses or listings other treatment centres

Match Patient’s Identification

number / Name against
existing records in NRDS

NO
New Primary?

YES

Create a new
Update existing case record in
records NRDS

need 1 NO
additional Notification from Wait for
Information? respective data Notification
YES source received?

YES
NO Abstract data at
EMR/NEHR/MRO
case record NO
to complete the
‘complete’?
registration.

YES

collect data at
Registration completed another HCI?
NO
YES

Legends: Go to 1
HCI Healthcare Institute
HIDS Hospital Inpatient Discharge Summary
MHA Ministry of Home Affairs
MRO Medical record office in hospital
NCC National Cancer Centre
NRDS National Registry of Diseases system
EMR Electronic Medical Records
NEHR National Electronic Health Records

Fig. 1. Workflow for cancer registration.

cases with morphological verification (MV%), death certificate-
only cases (DCO%), and cases with other and unspecified sites (O&U
%) were analysed. Taking reference from CI5, Volume X [9], SCR
data were compared with seven other regional registries which
were selected based on geographical and ethnicity comparability,
for the period 2003–2007. This allows for the assessment of
statistical differences using the detailed methods described in CI5,
Volume VIII [22].
3.4. Timeliness
Timeliness can be interpreted using the median time taken
from diagnosis to registration and the time from registration to the
reporting of incidence in SCR’s annual reports [14,15].
4. Results
4.1. Comparability
The SCR registers incident cases of malignant and in-situ
neoplasms. From 1968 to 1992, the topographical codes of the
cancer site were coded according to International Classification of
Diseases (ICD) 9th revision. Since 1993, the basis for coding has
been changed to ICD-O-2 and subsequently to ICD-O-3 [23] from
2003 to present. The tumour morphology was coded according to
the Manual Of Tumor Nomenclature And Coding (MOTNAC) and
ICD-O-2 in the early years and converted to ICD-O-3 [23] from
2003 to present. For coding of haematopoietic and lymphoid
tissues, SCR follows the rules of the World Health Organisation
(WHO) Classification of Tumours [24], for cases from 2010 and
forward. Fig. 2 gives a graphic overview of the international
standards that the SCR has been following for the classification and
coding of neoplasms.
The incidence date and basis of diagnosis has been recorded
according to the IARC guidelines [25] since 2003. Prior to that, the
basis of diagnosis was determined based on a hierarchy of sources
with the date of histology examination accorded the highest
priority, to clinical confirmation [2]. The incidence date was based
on notifications by medical professionals.
The reporting of multiple primary tumours follows the IARC
multiple primary rules [26]. A primary tumour or multiple primary
tumours at an organ or paired organ with the same morphology or
 J.W.M. Fung et al. / Cancer Epidemiology 43 (2016) 76–86 79

Fig. 2. The standards of classification and coding of neoplasms, Singapore, 1968–2013.

within the same morphology group is counted as one incidence. 4.2. Completeness
However, for data collection, the SCR also registers two tumours of
different laterality but of the same morphology in the same paired Annual trends in incidence rates of cancer did not display
organ, and two tumours at different sub-sites of the same organ of unusual or sudden systematic fluctuations for the period 1968–
the same morphology. The coding of morphology and recording of 2013 for all sites combined and top cancers (Fig. 3).
multiple primary cancers at the same organ or paired organ are The shapes of the age-specific incidence curves for males and
guided by the Multiple Primary and Histology Coding Rules Manual females in Singapore are typical for the major cancers, indicating
2007 from the Surveillance, Epidemiology, and End Results (SEER) no variability in completeness by age (Fig. 4). To retain similar
Program [27]. ethnic demographics, international comparisons were made with
The Registry follows the staging rules from the American Joint Eastern Asia and South Eastern Asia countries using data from
Committee on Cancer (AJCC) 6th edition [28] for cases diagnosed GLOBOCAN 2012 [31]. Countries with high quality national or
between 2003 and 2009. For cases diagnosed from 2010, the regional data were selected, namely Japan, Philippines, and
staging is assigned according to AJCC 7th edition [29]. Thailand for the year 2012, to compare with Singapore during
Classification and coding of cancers can be challenging. To the period of 2008–2012.
ensure accuracy and consistency, the SCR consults an in-house The age specific incidence rates for childhood cancer in
pathologist for reviewing of complex cases. The SCR also houses Singapore, 2008–2012, was compared with the upper and lower
two Certified Tumor Registrars (CTR) certified by the National deciles for childhood cancer published in CI5 Volume X [9]. The
Cancer Registrars Association (NCRA) [30], who ensure that the values were within range except for boys and girls aged 0–4 years
SCR coding practices adhere to international guidelines. which exceeded the upper limit of the reference interval (Table 1).

Annual trends in age-standardised incidence rates, Males Annual trends in age-standardised incidence rates for all sites
300.0 combined, and for Top 5 cancers, Females, 1968-2013, Singapore
300.0
Age-standardized rate (per 100,000 per year)

Age-standardized rate (per 100,000 per year)

250.0
250.0

200.0
All sites combined 200.0
All sites combined
Colo-rectum
150.0 Colo-rectum
150.0
Lung
Enactment of NRD Act Lung
Enactment of NRD Act
Prostate
100.0 100.0 Female Breast
Liver
Corpus uteri
Stomach
50.0 50.0 Ovary, etc

0.0 0.0
1968
1971
1974
1977
1980
1983
1986
1989
1992
1995
1998
2001
2004
2007
2010
2013
1968
1971
1974
1977
1980
1983
1986
1989
1992
1995
1998
2001
2004
2007
2010
2013

Year of diagnosis Year of diagnosis

Fig. 3. Annual trends in age-standardised incidence rates for all sites combined, for top cancers, Males and Females, 1968–2013, Singapore.
80 J.W.M. Fung et al. / Cancer Epidemiology 43 (2016) 76–86

Colorectum, Male Colorectum, Female

500 400
AGE-SPECIFIC RATE (per 100,000 per year)

450

AGE-SPECIFIC RATE (per 100,000 per year)

350
400
300
350
300 250
Singapore (2008-2012)
250 200
200 Japan
150
150 Philippines
100
100 Thailand
50
50
0 0

0-14

15-39

40-44

45-49

50-54

55-59

60-64

65-69

70-74

75+
0-14

15-39

40-44

45-49

50-54

55-59

60-64

65-69

70-74

75+
AGE AGE
Thyroid, Females
Thyroid, Male 25

AGE-SPECIFIC RATE (per 100,000 per year)

25
AGE-SPECIFIC RATE (per 100,000 per year)

20
20 Singapore (2008-2012)

15 15
Japan

10 10
Philippines

5 5
Thailand

0 0
0-14

15-39

40-44

45-49

50-54

55-59

60-64

65-69

70-74

75+

0-14

15-39

40-44

45-49

50-54

55-59

60-64

65-69

70-74

75+
AGE AGE

Breast, Females
250
AGE-SPECIFIC RATE (per 100,000 per year)

200

150
Singapore (2008-2012)

100 Japan
Philippines
50
Thailand

0
0-14

15-39

40-44

45-49

50-54

55-59

60-64

65-69

70-74

75+

AGE

Fig. 4. Age-specific incidence curves for colorectal, thyroid and breast cancers, Singapore 2008–2012, compared with three Asian registries using data from GLOBOCAN 2012.

Table 1
Age-specific incidence rates per million of childhood cancers by gender, Singapore, 2008–2012.

Age Boys Reference1 Girls Reference1

0 4 27.6 <13.7 to >25.6 24.0 <11.3 to >23.2
5–9 11.8 <8.9 to >16.5 10.9 <7.0 to >12.7
10–14 16.3 <9.2 to >16.3 11.6 <7.9 to >14.9
1
The upper and lower deciles for childhood cancer incidence rates in CI5 Vol IX.
 J.W.M. Fung et al. / Cancer Epidemiology 43 (2016) 76–86 81

Males Females
1.0 1.0 Liver Oesophagus
Pancreas
Oesophagus Pancreas
Liver Lung Lung
0.8 0.8

Melanoma
0.6 0.6

M:I Rao
M:I Rao

All sites
0.4 Colorectum 0.4 Cervix uteri Colorectum
Melanoma Ovary, etc.
Lymphoma Lymphoma
Prostate Female
0.2 0.2 All sites
breast

Corpus uteri
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
1-survival 1-survival

Fig. 5. Mortality:Incidence ratios (2008–2012) versus 1 minus 5 yrs relative survival (2003–2012), Singapore.

The mortality/incidence (M:I) ratio is a comparison of mortality
statistics obtained from an independent source and the number of
new cancer cases in the same time period [10,14,32]. If the
mortality data is reliable and accurate, with a steady trend of
incidence and survival, the M:I ratio will be approximately 1 minus
5 years survival probability [10,33,34]. Any distortions in the M:I
ratio may suggest completeness issues [34].
The mortality/incidence (M:I) ratios which had a strong
coherence with 1-survival were pancreas, colorectum and lung
cancers for males and females (Fig. 5). For prostate cancer and
breast, liver, oesophagus and melanoma cancer for females, the M:I
ratio was higher than (1-survival), whereas lymphoma for males,
corpus uteri and ovary cancer had a M:I ratio that was lower than
(1-survival). The M:I ratios for most cancer sites were similar to
those of the region, with the exception of oesophagus cancer which
had a M:I ratio higher than the region.
The average number of notifications per case for all sites
combined was 2.3 (Table 2). The site with the highest rate of
notifications per case was cervical uteri with 3.1 notifications per
case, compared to leukaemia with 1.4 notifications per case. A high
number of notifications per case reduces the possibility of cancer
diagnoses going unreported, thus increasing the completeness of
registry data [10].
Using the capture-recapture method for the period of 2008–
2012 and for all sites combined, the cases notified by clinicians and
pathologists formed the majority at 69.2%, followed by those
notified by pathologists only (17.1%) and clinicians only (9.4%). The
direction of dependency between each pair of sources is
determined by comparing the true and estimated number of
cases. If there is positive dependence between two sources, the
capture-recapture method underestimates the true number of
cases, whereas negative dependence overestimates the true
number of cases [19,35]. While pathological sources and death
certificates were negatively dependent, there was positive
dependency between clinical sources and pathological sources.
The strength of dependency between each pair of sources was
determined by evaluating the magnitude of under- or over-
estimation of the true number of cases. The lower the magnitude,
the weaker the dependency [35]. The dependency between
pathological sources and death certificates was stronger than
other pairs as the over-estimated number of clinical cases was the
largest (data not shown). Hence, these cases were combined and
compared with those from clinical sources in a two-way capture-
recapture method to estimate completeness for cases diagnosed in
2008–2012.
The overall completeness for the period 2008–2012 was
estimated to be 97.5% (Table 3). The lowest completeness was
estimated for liver (83.5%), while the highest completeness was
estimated for lip, vermilion (100%). Liver cancer is highly fatal.
Hence, due to strong negative dependency between pathological
sources and death certificates for liver cancer, its completeness was
underestimated based on a two-way capture-recapture method.
The flow method estimates a completeness of registration of
94.3% at one year and 98.1% at five years after cancer diagnosis for
all sites combined, for cancer cases diagnosed from 2008 till 2012
(Fig. 6).
4.3. Validity
Table 4 presents the age-standardised incidence rates (includ-
ing their standard errors) and the ratio of the observed value to the
expected value (O/E). The MV(%), DCO(%) and M:I(%) are also
shown as indicators of validity and completeness. For the period
2003–2007, the proportion of MV cases for Singapore was 88.6% for
males and 93.0% for females for all sites but non-melanoma skin,
and the proportion of DCO cases was 1.5% for males and 1.3% for
females (Table 4) The MV and DCO proportions observed in
Singapore was not significantly different from the region.
Additionally, the O&U proportions were 2.3% for Singaporean
males and 2.0% for Singaporean females for the period 2003–2007
based on data from CI5-X [9].
4.4. Timeliness
The study shows that the timeliness of SCR data had improved
significantly. The median time between diagnosis and registration
was halved from 160 days in 2005 to 76 days in 2013, and this trend
was similar for both male and females (Fig. 7).
The incidence data reported in annual reports [8,36–40] appear
to be stable over time, with the number of cases diagnosed each
year being consistently reported in the year of publication as well
as subsequent years (Table 5). For example, the cancer incidence
for 2012 (12123 cases) was first reported in the 2008–2012 report
which was published in 2014 [39]. This incidence for 2012 was
82 J.W.M. Fung et al. / Cancer Epidemiology 43 (2016) 76–86

Table 2
Total number of cases, number of notifications per case, percentage of morphologically verified cases (MV%) and percentage of cases obtained from death certificate only (DCO
%), Singapore, 2008–2012.

ICD-9 ICD-O-31 Site Total No. of Cases No. of Notifications per case MV% DCO%
140 208 C00-C96 All sites 56697 2.3 91.9 1.1
140 C00 Lip 8 2.3 100.0 0.0
141 C01 C02 Tongue 294 2.4 100.0 0.0
142 C07-C08 Salivary gland 199 2.0 99.5 0.5
143 145 C03-C06 Mouth including gum, palate and floor of mouth 206 2.4 99.5 0.5
146 C09-10 Tonsil and Oropharynx 179 2.2 98.3 0.0
147 C11 Nasopharynx 1510 2.4 99.5 0.2
148 C12-C13 Hypopharynx 134 2.3 95.5 1.5
150 C15 Oesophagus 513 2.4 95.9 1.4
151 C16 Stomach 2502 2.3 97.0 0.8
152 C17 Small intestines 235 2.0 95.7 0.4
153 C18 Colon 5482 2.4 95.6 1.0
154 C19-C21 Rectosigmoid, Rectum and Anus 3309 2.5 97.7 0.3
155 C22 Liver and Intrahepatic ducts 2801 1.8 40.4 6.2
156 C23-C24 Gall bladder and Biliary ducts 561 2.1 81.6 1.1
157 C25 Pancreas 1417 2.2 76.4 3.2
158 C48 Retroperitoneum & Peritoneum 165 2.4 96.4 0.6
160 C30-C31 Nasal cavity, middle ear and accessory sinuses 166 2.2 99.4 0.6
161 C32 Larynx 375 2.3 97.6 0.3
162 C33-C34 Trachea, Bronchus and Lung 6503 2.4 86.0 2.3
163 C38.4 Pleura 82 2.4 95.1 0.0
164 C37 C38 excluding C38.4 Thymus, heart and mediastinum 249 2.0 98.0 0.8
170 C40-C41 Bones and articular cartilage 174 2.2 98.3 0.0
171 C47, C49 Connective and soft tissues 467 2.2 96.8 0.2
172 C43 Melanoma of skin 140 2.5 99.3 0.0
173 C44 Other neoplasms of Skin 2662 1.8 99.4 0.2
174 C50 (female) Female Breast 8507 2.4 99.4 0.2
175 C50 (male) Male Breast 29 1.9 96.6 0.0
180 C53 Cervix uteri 920 3.1 98.7 0.3
181 C58 Placenta 23 1.9 91.3 0.0
182 C54 Corpus uteri 1784 2.7 99.3 0.2
183 C56-C57 Ovary and other parts of female genital organs 1604 2.5 95.4 0.9
184 C51 C52 Vulva and Vagina 128 2.7 97.7 0.8
185 C61 Prostate 3321 2.5 96.5 0.2
186 C62 Testis 169 2.2 99.4 0.0
187 C60 & C63 Penis and other parts of male genital organs 107 2.6 100.0 0.0
188 C67 Bladder 926 2.3 98.1 0.4
189 C64-C66, C68 Kidney and other parts of Urinary organs 1497 2.3 87.5 1.7
190 C69 Eye 37 2.2 91.9 0.0
191 192 C70-C72 Brain, Spinal cord and parts of Central Nervous System 689 2.1 82.9 2.9
193 C73 Thyroid 1299 2.0 99.0 0.1
194 C74-C75 Adrenal gland, other endocrine glands and related structures 143 2.3 88.8 0.0
199 C80 Unspecified2 749 1.9 64.6 5.3
200, 202 C82-C85, C96 Non-Hodgkin's lymphoma 2002 2.2 100.0 0.0
201 C81 Hodgkin's disease 210 2.4 100.0 0.0
203 C88, C90 Multiple myeloma and immunoproliferative neoplasms 454 1.9 93.0 0.0
204 C91 Lymphoid leukaemia 455 2.3 95.6 0.4
205 C92 Myeloid leukaemia 1111 2.0 93.7 0.0
206 C93 Monocytic leukaemia 47 2.6 100.0 0.0
207 C94 Other leukaemia 114 1.4 69.3 0.0
208 C95 Leukaemia, unspecified 39 1.7 94.9 2.6
1
ICD-10 codes were used for C43, C82-95.
2
“Unspecified” include unknown primary sites.

updated in the 2009–2013 report which was published in 2015
(12185 cases) [40]. The percentage difference of 0.51% therefore
shows a slight under-reporting for the 2012 cancer incidence.
5. Discussion
This evaluation of the data quality of the SCR suggests a high
degree of comparability, completeness, validity and timeliness. The
quality of data was also recognized by the IARC in the GLOBOCAN
2012 project where SCR’s data was conferred a data quality index of
A1, indicating high quality national incidence and mortality data
[31]. It meets the prerequisites of high quality data with its
mandatory reporting of neoplasms, unique personal identification
numbers and extensive access to sources of notification.
In terms of completeness, the modified flow method gave
higher estimates than the capture-recapture method. The modified
flow method is more appropriate for the SCR as it is able to adjust
for delayed registration [20] since the SCR collects data retrospec-
tively, but the results from both methods affirm the high level of
completeness of registration of around 97–98%.
A limitation of the capture-recapture method is its assumption
that all notification sources are independent of each other and all
patients have the same probability of being captured in the
registry, which are likely to be violated in cancer registration
[18,41]. However, taking reference from Brenner [19] and Larsen
et al. [14], it is possible to correct for the dependency between the
sources by pooling the two groups into one broad group and
compare it with the remainder group by a two-way capture-
 J.W.M. Fung et al. / Cancer Epidemiology 43 (2016) 76–86 83

Table 3
Capture/recapture estimation on combination of death certificates and pathological sources, Singapore, 2008–2012.

ICD-9 ICD-O-31 Site True Dependency between

Number
D: C versus C: D versus P: C versus Completeness (%)
P3 P4 D5
140 208 C00-C96 All sites 56697 + 97.5
140 C00 Lip 8 NA NA NA 100.0
141 C01 C02 Tongue 294 NA NA NA 99.1
142 C07-C08 Salivary gland 199 NA NA NA 97.7
143 145 C03-C06 Mouth including gum, palate and floor of mouth 206 NA NA NA 99.3
146 C09-10 Tonsil and Oropharynx 179 + + 97.8
147 C11 Nasopharynx 1510 NA NA NA 98.9
148 C12-C13 Hypopharynx 134 + + 98.1
150 C15 Oesophagus 513 + + 98.7
151 C16 Stomach 2502 + 98.4
152 C17 Small intestines 235 + + 93.0
153 C18 Colon 5482 + 98.7
154 C19-C21 Rectosigmoid, Rectum and Anus 3309 + 99.1
155 C22 Liver and Intrahepatic ducts 2801 + 83.5
156 C23-C24 Gall bladder and Biliary ducts 561 95.8
157 C25 Pancreas 1417 + 95.0
158 C48 Retroperitoneum & Peritoneum 165 + + + 97.6
160 C30-C31 Nasal cavity, middle ear and accessory sinuses 166 + + + 97.4
161 C32 Larynx 375 + 98.9
162 C33-C34 Trachea, Bronchus and Lung 6503 + 97.6
163 C38.4 Pleura 82 + + 98.2
164 C37 C38 excluding Thymus, heart and mediastinum 249 NA NA NA 94.0
C38.4
170 C40-C41 Bones and articular cartilage 174 NA NA NA 94.0
171 C49 Connective and soft tissues 467 NA NA NA 95.2
172 C43 Melanoma of skin 140 NA NA NA 98.3
173 C44 Other neoplasms of Skin 2662 + + 99.2
174 C50 (female) Female Breast 8507 + + 99.6
175 C50 (male) Male Breast 29 NA NA NA 95.9
180 C53 Cervix uteri 920 + 99.4
181 C58 Placenta 23 NA NA NA 84.1
182 C54 Corpus uteri 1784 + + 99.3
183 C56-C57 Ovary and other parts of female genital organs 1604 + 97.0
184 C51 C52 Vulva and Vagina 128 NA NA NA 99.0
185 C61 Prostate 3321 + 98.8
186 C62 Testis 169 + 97.2
187 C60, C63 Penis and other parts of male genital organs 107 NA NA NA 98.9
188 C67 Bladder 926 + + 98.3
189 C64-C66, C68 Kidney and other parts of Urinary organs 1497 + 96.9
190 C69 Eye 37 NA NA NA 97.5
191 192 C70-C72 Brain, Spinal cord and parts of Central Nervous System 689 + 91.5
193 C73 Thyroid 1299 + + 96.6
194 C74-C75 Adrenal gland, other endocrine glands and related 143 NA NA NA 94.5
structures
199 C80 Unspecified2 749 + + 89.0
200, 202 C82-C85, C96 Non-Hodgkin's lymphoma 2002 + 97.9
201 C81 Hodgkin's disease 210 NA NA NA 99.5
203 C88, C90 Immunoproliferative neoplasms and multiple myeloma 454 NA NA NA 91.2
204 C91 Lymphoid leukaemia 455 NA NA NA 94.2
205 C92 Myeloid leukaemia 1111 + + 92.2
206 C93 Monocytic leukaemia 47 99.4
207 C94 Other leukaemia 114 NA NA NA 53.8
208 C95 Leukaemia, unspecified 39 NA NA NA 67.5

Notes:
a. Cases from death certificate and pathology sources were combined and compared with those from clinical sources in a two-source capture-recapture analysis to estimate
completeness for all sites.
b. The dependency between death certificate and pathology sources was stronger than other pairs as the over-estimated number of clinical cases was the largest. There is also
a high proportion of sites in the same direction of negative dependency.
c. “NA” refers to “Not Applicable”. The dependency for these sites cannot be calculated as there were no cases from C + P + D.
1
ICD-10 codes were used for C43, C82-95.
2
“Unspecified” include unknown primary sites.
3
D: C versus P: between clinical and pathology sources for all death certificates.
4
C: D versus P: between death certificate and pathology sources for all clinical records.
5
P: C versus D: between clinical and death certificate sources for all pathology records.
84 J.W.M. Fung et al. / Cancer Epidemiology 43 (2016) 76–86

recapture method. Nevertheless, overestimation can still occur due

Fig. 6. Completeness of ascertainment according to time since diagnosis for all
cancer sites combined, 2008–2012.
to positive dependency between two sources and vice versa [19].
The flow method analysis was modelled after Montanaro’s
modified method [20] which was able to take into account delays
between the diagnosis date and registration, since the SCR does not
collect real-time data. A limitation is that the reliability of the
analysis depends on the quality of death data available.
The semi-quantitative evaluation of completeness displayed
relatively stable trends of annual incidence ASR over time and
typical shapes of the age-specific incidence curves. The childhood
cancer incidence rates were within the reference ranges except for
values for boys and girls aged 0–4 years where the incidence rates
were above the upper limits of the reference interval. This is likely
to be an incidental finding as data cleaning was performed to
ensure that there were no duplicate registrations or coding errors.
The M:I ratio was higher than (1-survival) for prostate cancer
and breast, liver, oesophagus and melanoma cancer for females
which may suggest incompleteness in incidence, whereas lym-
phoma for males, uterine and ovary cancer had a M:I ratio that was

Table 4
Singapore (2003–2007), regional comparison of incidence rates (ASRs), MV%, DCO% and M/I ratios.

MALE

SITE Cases ASR (se) O/E MV(%) DCO(%) M/I(%) ICD-10

Lip, oral cavity and pharynx 1770 16.6 (0.41) 1.23 98.8 0.4 46.0 C00 14
Oesophagus 335 3.6 (0.2) 1.08 96.4 0.9 97.9> C15
Stomach 1325 14 (0.39) 1.10 97.4 0.6 70.8 C16
Colon, rectum and anus 3890 40.3 (0.66) 1.12 97.1 0.7 44.3 C18 21
Liver 1663 17.4 (0.44) 1.11 39.7 4.6 91.9 C22
Pancreas 569 5.9 (0.25) 1.12 70.5 4.6 95.3 C25
Larynx 353 3.7 (0.2) 1.11 96.9 0.3 40.8 C32
Lung (incl. trachea) 3852 41.1 (0.67) 1.09 87.0 2.0 88.4 C33 34
Melanoma of skin 50 0.5 (0.07) 1.20 50.0< 0.0 52.0 C43
Prostate 2190 24.2 (0.52) 1.05 96.3 0.2 20.0 C61
Testis 119 1.3 (0.13) 1.08 98.3 0.0 5.0 C62
Kidney etc. 632 6.3 (0.26) 1.16 87.5 1.3 35.4 C64 66, C68
Bladder 637 7.1 (0.28) 1.11 98.1 0.3 26.5 C67
Brain, central nervous system 230 2.4 (0.17) 1.13 83.5 4.8 70.0 C70 72
Thyroid 215 2.1 (0.15) 1.19 98.6 0.0 12.1 C73
Lymphoma 1157 11.7 (0.36) 1.10 99.7 0.3 34.1 C81 88,C90
Leukaemia 537 6.4 (0.29) 0.97 99.3 0.7 51.6 C91 95
All sites but non melanoma skin 20963 219.9 (1.56) 1.11 88.6 1.5 60.2 C00 96bC44

Female

Site Cases ASR (se) O/E MV(%) DCO(%) M/I(%) ICD-10

Lip, oral cavity and pharynx 648 5.5 (0.22) 1.35 99.8 0.2 36.7 C00 14
Oesophagus 83 0.6 (0.07) 1.50 89.2 2.4 95.2> C15
Stomach 850 7.1 (0.25) 1.37 96.1 1.1 74.5 C16
Colon, rectum and anus 3363 28.8 (0.51) 1.34 95.3 1.1 44.6 C18 21
Liver 534 4.6 (0.21) 1.33 39.8 5.1 98.7 C22
Pancreas 466 4.0 (0.19) 1.33 57.9 8.2 98.1 C25
Larynx 33 0.3 (0.05) 1.33 90.9 3.0 42.4 C32
Lung (incl. trachea) 1898 16.3 (0.38) 1.33 83.3 3.2 83.2 C33 34
Melanoma of skin 57 0.5 (0.07) 1.40 100.0 0.0 47.4 C43
Breast 6717 57.8 (0.72) 1.33 99.0 0.3 23.6 C50
Cervix uteri 1013 8.9 (0.29) 1.30 99.5 0.2 38.6 C53
O&U part of uterus 1337 11.8 (0.33) 1.30 99.0 0.1 15.4 C54-55
Ovary 1003 9.1 (0.29) 1.29 94.1 1.4 47.5 C56
Kidney etc. 332 3.1 (0.18) 1.23 82.5 1.2 37.3 C64 66, C68
Bladder 206 1.9 (0.13) 1.37 98.1 0.0 34.5 C67
Brain, central nervous system 191 2.1 (0.16) 1.05 84.3 3.1 52.9 C70 72
Thyroid 650 5.9 (0.24) 1.25 99.4 0.2 11.4 C73
Lymphoma 841 7.8 (0.28) 1.21 99.3 0.7 35.9 C81 88,C90
Leukaemia 443 5.1 (0.26) 1.00 99.3 0.7 47.9 C91 95
All sites but non melanoma skin 21915 192 (1.34) 1.31 93.0 1.3 46.2 C00 96bC44

Notes:
a. Significantly lower (<) or higher (>) values are shown in bold.
b. SCR data were compared with Malaysia (Penang), Thailand (Bangkok), Philippines (Manila), China (Hong Kong), Japan (Osaka), Republic of Korea and India (New Delhi)
based on geographical and ethnicity comparability.
c. Data Source: CI5 X, International Agency for Research on Cancer (IARC), 2014.
d. Statistical tests performed were modelled after the detailed methods described in CI5-VIII, Comparability and quality of data chapter (page 72).
 J.W.M. Fung et al. / Cancer Epidemiology 43 (2016) 76–86 85

Table 5
Number of registered cancer cases in Singapore from 2008 to 2012, comparing the incidence first published in the annual report to the incidence reported in the next annual
report.

Year of annual report publication

2010 2011 2012 2013 2014 2015 (09-13 report) % Difference

(04- (05- (06- (07- (08-12
08 report) 09 report) 10 report) 11 report) report)
Number of registered cancer cases by year 2008 10316 10316 0.00

2009 10568 10579 0.10

2010 11160 11144 0.14
2011 11525 11529 0.03
2012 12123 12185 0.51

between data timeliness and data completeness as both improved

over the years.
180.0
6. Conclusion
160.0

140.0 This review indicated that the SCR adheres to international

120.0
guidelines, has a high degree of completeness, accuracy and
timeliness. It should however be kept in mind that an efficient
Median days

100.0 national registry system is not the end goal; the eventual
80.0 Male objectives are to step up actions to prevent, control and improve
Female interventions for cancer care. This review allows for greater
60.0 confidence in the use of the quality data in the SCR for treatment
40.0 and research, as well as the evaluation of national programmes
such as the National Breast Cancer Screening Programme [43] and
20.0
the National Cervical Screening Programme [44], which in turn
0.0 impacts decision-making by the Singapore health authorities.
2005 2006 2007 2008 2009 2010 2011 2012 2013
Year Conflict of interest

Fig. 7. Median time between diagnosis and registration, Males and Females,
None declared.
Singapore 2005–2013.

Authorship contribution

Study concept and design: JWMF, SBLL. Data acquisition:

lower than (1-survival). Based on regional comparisons shown in
Table 4, the M:I ratio for oesophageal cancer was higher than the
region for both males and females. However, the results need to be
interpreted with caution as under-reporting or over-reporting or a
lack of stability of incidence and mortality trends may distort the
results [10]. Additionally, this method may not be appropriate for
assessing sites that are frequently metastasized due to the
tendency to misclassify them as primary sites [15].
In terms of validity, the proportion of morphologically verified
cases was high at 91.9% and the proportion of cases identified
through death certificates only was low at 1.09%. This shows that
an effective system is in place to trace for other sources of
notifications. Liver cancer had the lowest MV% of 40.4%. The IARC
technical report No. 43 [42] suggested that it was not easy to
perform pathological investigations for liver cancer. Additionally,
its rapidly fatal nature and the high level of accuracy of clinical and
diagnostic tests (such as CT scans, tumour markers) reduce the
need for additional microscopic verification.
Although late notifications continue to add to the cases after in
the year after the first publication, the difference is low at below
0.51%. There is a calculated over-reporting of 0.14% for cancer
incidence for 2010 reported in year 2012 when compared to year
2013. This was attributed to the NRD Act enforced in 2009 for
mandatory cancer notifications, resulting in an influx of retrospec-
tive cases. This incidence count reported for 2010 was corrected in
subsequent years. There does not appear to be significant trade-off
WYTH, BGL. Quality control of data: SBLL. Data analysis and
interpretation: JWMF. Statistical analysis: HZ. Manuscript prepa-
ration: JWMF. Critical revision of the manuscript for important
intellectual content: JWMF, SBLL, KYC, HPL.
References
[1] Trends in Cancer Incidence in Singapore 1968–2007. Singapore Cancer
Registry, Report No. 7, in Singapore Cancer Registry, 2008.
[2] Trends in Cancer Incidence in Singapore 1968–2002. Singapore Cancer
Registry, Report No. 6, in Singapore Cancer Registry, 2004.
[3] Population in Brief 2013, A joint production by National Population and Talent
Division, Prime Minister’s Office, Singapore Department of Statistics, Ministry
of Home Affairs, Immigration & Checkpoints Authority, 2013. Available from:
http://www.nptd.gov.sg/portals/0/news/population-in-brief-2013.pdf
(accessed 19.5.16).
[4] M.C.C. Teo, K.C. Soo, Cancer trends and incidences in Singapore, Jpn. J. Clin.
Oncol. 43 (3) (2013) 219–224.
[5] K. Tomic, et al., Evaluation of data quality in the National Prostate Cancer
Register of Sweden, Eur. J. Cancer 51 (1) (2015) 101–111.
[6] D.M. Parkin, The role of cancer registries in cancer control, Int. J. Clin. Oncol. 13
(2) (2008) 102–111.
[7] National Registry of Diseases Act, Chapter 201B, Singapore, 2007. Available
from http://statutes.agc.gov.sg/ (accessed 20.7.15).
[8] Trends in Cancer Incidence in Singapore 2005–2009. Singapore Cancer
Registry, Interim Annual Report, 2011.
[9] Cancer Incidence in Five Continents Volume X, IARC Scientific Publication No.
164, in: D. Forman, F. Bray, D.H. Brewster, C. Gombe Mbalawa, B. Kohler, M.
Piñeros, E. Steliarova-Foucher, R. Swaminathan, J. Ferlay (Eds.), International
Agency for Research on Cancer, Lyon, 2013.
[10] D.M. Parkin, F. Bray, Evaluation of data quality in the cancer registry: principles
and methods. Part II. completeness, Eur. J. Cancer 45 (5) (2009) 756–764.
86 J.W.M. Fung et al. / Cancer Epidemiology 43 (2016) 76–86
[11] F. Bray, D.M. Parkin, Evaluation of data quality in the cancer registry: principles
and methods. Part I: comparability, validity and timeliness, Eur. J. Cancer 45 (5)
(2009) 747–755.
[12] Trends in Cancer Incidence in Singapore 1968–1977. Singapore Cancer
Registry, Report No. 1, 1983.
[13] Trends in Cancer Incidence in Singapore 2010–2014. Singapore Cancer
Registry, Interim Annual Report, 2015.
[14] I.K. Larsen, et al., Data quality at the Cancer Registry of Norway: an overview of
comparability, completeness, validity and timeliness, Eur. J. Cancer 45 (7)
(2009) 1218–1231.
[15] N. Dimitrova, D.M. Parkin, Data quality at the Bulgarian National Cancer
Registry: an overview of comparability, completeness, validity and timeliness,
Cancer Epidemiol 39 (3) (2015) 405–413.
[16] L. Barlow, et al., The completeness of the Swedish Cancer Register—a sample
survey for year 1998, Acta Oncol. 48 (1) (2009) 27–33.
[17] R. Zanetti, et al., Completeness and timeliness: cancer registries could/should
improve their performance, Eur. J. Cancer 51 (9) (2015) 1091–1098.
[18] K. Tilling, Capture-recapture methods—useful or misleading? Int. J. Epidemiol.
30 (1) (2001) 12–14.
[19] H. Brenner, C. Stegmaier, H. Ziegler, Estimating completeness of cancer
registration in Saarland/Germany with capture-recapture methods, Eur. J.
Cancer 30 (11) (1994) 1659–1663.
[20] F. Montanaro, et al., A modification to the flow method to estimate
completeness in cancer registries with delayed registration, J. Public Health
(Oxf.) 28 (3) (2006) 274–277.
[22] D.M. Parkin, S.L. Whelan, J. Ferlay, L. Teppo, D.B. Thomas, Cancer Incidence in
Five Continents Volume VIII, IARC Scientific Publications, No. 155,
International Agency for Research on Cancer, Lyon, 2002.
[23] A. Fritz, C. Percy, A. Jack, K. Shanmugarathan, L. Sobin, D.M. Parkin, S. Whelan,
International Classification of Diseases for Oncology (ICD-O-3), third ed.,
World Health Organization, 2000.
[24] S.H. Swerdlow, E. Campo, N.L. Harris, E.S. Jaffe, S.A. Pileri, H. Stein, J. Thiele, J.W.
Vardiman, WHO Classification of Tumours of Haematopoietic and Lymphoid
Tissues, fourth ed., IARC Press, Lyon, France, 2008.
[25] O.M. Jensen, D.M. Parkin, C.S. Muir, Cancer Registration: principles and
methods. IARC Scientific Publication No. 95, IARC, Lyon, 1991.
[26] International Rules for Multiple Primary Cancers (ICD-O Third Edition).
Internal Report No. 2004/02, IARC, Lyon, 2004.
[27] C.H. Johnson, S. Peace, P. Adamo, A. Fritz, A. Percy-Laurry, B.K. Edwards, The
2007 Multiple Primary and Histology Coding Rules, National Cancer Institute,
Surveillance, Epidemiology and End Results Program, Bethesda, MD, 2007.
[28] F.L. Greene, D.L. Page, I.D. Fleming, A. Fritz, C.M. Balch, D.G. Haller, M. Morrow,
AJCC Cancer Staging Manual, sixth ed., Springer-Verlag, New York, 2002.
[29] S. Edge, D.R. Byrd, C.C. Compton, A.G. Fritz, F.L. Greene, A. Trotti, AJCC Cancer
Staging Manual, seventh ed., Springer-Verlag, New York, 2010.
[30] National Cancer Registrars Association, Certified Tumor Registrar Certification,
(2016) . Available from: http://www.ncra-usa.org/ (accessed 13.01.16).
[31] J. Ferlay, I. Soerjomataram, M. Ervik, R. Dikshit, S. Eser, C. Mathers, M. Rebelo, D.
M. Parkin, D. Forman, F. Bray, GLOBOCAN 2012 v1.0, Cancer Incidence and
Mortality Worldwide: IARC Cancer Base No. 11 [Internet], International Agency
for Research on Cancer, Lyon, France, 2013 Available from: http://globocan.
iarc.fr (accessed 13.01.16).
[32] L.G. Sigurdardottir, et al., Data quality at the Icelandic Cancer Registry:
comparability, validity, timeliness and completeness, Acta Oncol. 51 (7) (2012)
880–889.
[33] K. Suwanrungruang, et al., Appropriateness of the standard mortality/
incidence ratio in evaluation of completeness of population-based cancer
registry data, Asian Pac. J. Cancer Prev. 12 (12) (2011) 3283–3288.
[34] R. Cendales, et al., Data quality at population-based cancer registries in
Colombia, Biomedica 32 (4) (2012) 536–544.
[35] E. Crocetti, et al., An application of the two-source capture-recapture method
to estimate the completeness of the Tuscany Cancer Registry, Italy, Eur. J.
Cancer Prev. 10 (5) (2001) 417–423.
[36] Trends in Cancer Incidence in Singapore 2004–2008. Singapore Cancer
Registry, Interim Annual Report, 2010.
[37] Trends in Cancer Incidence in Singapore 2006–2010. Singapore Cancer
Registry, Interim Annual Report, 2012.
[38] Trends in Cancer Incidence in Singapore 2007–2011. Singapore Cancer
Registry, Interim Annual Report, 2013.
[39] Trends in Cancer Incidence in Singapore 2008–2012. Singapore Cancer
Registry, Interim Annual Report, 2014.
[40] Trends in Cancer Incidence in Singapore 2009–2013. Singapore Cancer
Registry, Interim Annual Report, 2015.
[41] M.E. Kroll, et al., Childhood cancer registration in Britain: capture-recapture
estimates of completeness of ascertainment, Br. J. Cancer 104 (7) (2011)
1227–1233.
[42] F. Bray, A. Znaor, P. Cueva, A. Korir, R. Swaminathan, A. Ullrich, S.A. Wang, D.M.
Parkin, Planning and developing population-based cancer registration in low-
and middle-income settings, in IARC Technical Publication No. 43,
International Agency for Research on Cancer, 2014.
[43] E.Y. Loy, et al., National Breast Cancer Screening Programme, Singapore:
evaluation of participation and performance indicators, J. Med. Screen. 22 (4)
(2015) 194–200.
[44] A.Z. Jin, et al., Evaluation of the National Cervical Cancer Screening Programme
in Singapore, Singapore Med. J. 54 (2) (2013) 96–101.