Cancer Diagnosis and Therapy
Population-based survival analysis of colorectal cancer patients in Singapore, 1968–1992
Article first published online: 19 MAR 2002
Copyright © 2002 Wiley-Liss, Inc.
International Journal of Cancer
Volume 99, Issue 3, pages 460–465, 20 May 2002
How to Cite
Du, W.-B., Chia, K.-S., Sankaranarayanan, R., Sankila, R., Seow, A. and Lee, H.-P. (2002), Population-based survival analysis of colorectal cancer patients in Singapore, 1968–1992. Int. J. Cancer, 99: 460–465. doi: 10.1002/ijc.10333
- Issue published online: 25 APR 2002
- Article first published online: 19 MAR 2002
- Manuscript Accepted: 11 JAN 2002
- Manuscript Revised: 2 JAN 2002
- Manuscript Received: 30 APR 2001
- Bill and Melinda Gates Foundation through the Alliance for Cervical Cancer Prevention (ACCP)
- National Medical Research Council, Singapore. Grant Number: NMRC/0145/1996
- International Union Against Cancer (UICC), Geneva, Switzerland
- relative survival;
- observed survival;
- prognostic factors;
- developing country;
- cancer registry;
Since the 1980s, colorectal cancer incidence in Singapore has ranked second to lung in males and females. We describe a population-based analysis of survival of colorectal cancer patients diagnosed from 1968 to 1992 in Singapore. Data of colorectal cancer patients diagnosed during 1968–1992 were retrieved from the Singapore Cancer Registry. Patients were passively followed up for death to the end of 1997. The final dataset consisted of 10,114 subjects. Observed and relative survival rates were calculated by stage (localized, regional metastases and distant metastases), age, ethnicity and calendar period for both genders. Over the study period, a significant progress in survival of colorectal cancer patients was observed. For localized cancer of the colon, the 5-year age-standardized relative survival (ASRS) increased from 36% in 1968–1972 to 66% in 1988–1992 for males and from 32 to 71% for females. For localized rectal cancer, the 5-year ASRS improved from 25 to 66% for males and from 23 to 66% in females. Similarly, improvement was observed in colorectal cancer patients with regional metastases, but not in those with distant metastases. Calendar year period and clinical stage of disease were identified as major significant prognostic factors of survival for colorectal cancer. The substantially improved colorectal cancer survival rates reflected the interplay of cancer control activities in various areas, such as health promotion, early diagnosis and treatment. Our study shows a unique changing pattern of survival experience for colorectal patients from a country undergoing rapid economic development. © 2002 Wiley-Liss, Inc.
In Singapore, the incidence of colorectal cancer has ranked second to that of lung cancer in both males and females since the 1980s. The age-standardized incidence rate (using world standard population as the standard population) for colorectal cancer in males has increased from 19.9/100,000 in 1968–1972 to 45.0/100,000 in 1993–1997 in Singapore. For females, it has increased from 15.6/100,000 to 32.1/100,000 over the same period.1, 2 The incidence of colorectal cancer even surpassed that of lung cancer for both genders combined in 1993–1997. The incidence rates in Singapore Chinese are now slightly higher than those of Chinese in Hawaii and Los Angeles and comparable to those in Europe, Australia and Hong Kong.2
Singapore is a city-state with a 1990 resident population of 2.69 million, of which 78% are Chinese, 14% Malays and 7% Indians.3 In the past 50 years, Singapore has experienced dramatic improvements in socioeconomic development.4 The lifestyle, including dietary pattern, has also changed in tandem with the changing cancer pattern.1 The health services have also developed and expanded considerably into a well-developed and modern health care infrastructure with equitable access to the general population.5
Clinical studies on the survival of patients with colorectal cancer have often been based on surgical or pathologic series in hospitals. Population-based survival estimates are based on unselected patients and could provide a more realistic reflection of cancer care and control in a given area. They are useful for making comparisons between populations and over time. Comparison of survival estimates can help to identify priorities and suggest measures to improve patients' survival. Our study describes the population-based survival of colorectal cancer patients diagnosed from 1968 to 1992 in Singapore.
MATERIAL AND METHODS
In January, 1968, the Singapore Cancer Registry began to obtain basic epidemiologic and clinical data on all cases of cancer diagnosed in Singapore. Notification was voluntary and from all medical professionals in Singapore. Pathology records (biopsies and necropsies), hospital discharge records and death certificates were also routinely checked by Cancer Registry staff and requests were sent to the doctors-in-charge for additional information on cases that were not previously reported. Cases were also registered by Cancer Registry staff if they satisfied one of the following conditions: (i) a pathologic diagnosis of cancer; or (ii) a clinical diagnosis of cancer supported by surgical, radiologic or laboratory findings; or (iii) a record of cancer in the death certificate (i.e., death certificate only [DCO]).
A total of 11,110 invasive colorectal cancers patients (6,522 colon and 4,588 rectum) were registered from 1968 to 1992. Of these, 9,820 (91.5%) were histologically confirmed at diagnosis. There were 367 (3.3%) registered based on DCO information. Another 73 (0.7%) were diagnosed at necropsy. DCO and necropsy cases were excluded from further analysis. Passive follow-up was done by matching with the national death register from January 1, 1968 to December 31, 1997, as well as the electoral register of 1997. The remaining 556 (5.0%) patients who could not be found in these registers were deemed lost to follow-up and excluded from the analysis. The final dataset consisted of 10,114 cases (5,878 colon and 4,236 rectum), of which 91.4% were histologically confirmed.
Death from any cause was regarded as an event and the subjects that were still alive at the end of follow-up were censored. Length of survival was defined as the interval between date of diagnosis and either date of death or the censoring date for follow-up (December 31, 1997). Cumulative observed survival and relative survival rates were calculated with Hakulinen's method, using software developed by the Finnish Cancer Registry (SURV2).6, 7 Relative survival is a good indicator of survival that might occur if the risks of death other than cancer were removed and is defined as the ratio of observed survival to expected survival. The expected survival Si at individual age i was estimated as: Si = 1 − qi where qi was the probability of death of the Singapore general population and is represented by the proportion of individuals alive at the beginning of a specific age interval who would be expected to die by the end of it. Probability of death (qi) is not exactly equal to the corresponding mortality rate (λi). Therefore probability of death was approximated from λi by using the following formula: qi = 2λi/(2 + λi).8, 9 The expected survivals were calculated using the published age-, gender- and calendar year-specific mortality rates of the Singapore general population. When ethnic-specific survivals were estimated, the age-, gender-, calendar year and ethnic-specific mortality rates were used.10, 11
The equality of the relative survival distribution between levels of a variable was examined using Hakulinen's method.12 A proportional hazards regression model on relative survival was employed subsequently to adjust for the relevant factors and identify the independent significant predictors of survival:
where h(t; x) is the hazard function for colorectal cancer patients with covariate vector x; h0(t; x) is the baseline hazard function for comparable individuals from the general population; t is the follow-up time; x is the covariate vector; and β is a vector of regression coefficients measuring the intensity of the effect of the vector x.13 The appropriateness of this proportional hazards regression model relies on the satisfaction of the proportional hazards assumption for each of the individual covariate with the others held constant. This assumption was examined by introducing interaction terms between covariates and the categorical follow-up year. Interactions between follow-up year and gender, ethnicity, age groups, period and stage were tested in this proportional hazards regression model.13
For comparisons over time and between countries, the 5-year relative survival rates in Singapore or other populations of corresponding period were all directly standardized using the world standard cancer patient population as the standard population.14
Over the study period, survival of colorectal cancer patients improved greatly, especially for patients with localized cancer and regional metastases (Figs. 1,2). For localized cancer of colon, the 5-year age-standardized relative survival (ASRS) increased from 36% in 1968–1972 to 66% in 1988–1992 for males and from 32 to 71% for females. For localized rectal cancer, the 5-year ASRS improved from 25 to 66% for males and from 23 to 66% in females. A similar improvement was observed for patients with regional metastases, i.e., from 24 to 54% in males and from 33 to 49% in females for colon cancer and from 10 to 37% in males and from 25 to 37% in females for rectal cancer.
Calendar year period of diagnosis and clinical stage of disease were significant predictors of survival for both cancers. Older patients had worse prognosis than the young except for males with rectal cancer. Chinese and Indians enjoyed better relative survival rates than Malays (Tables I, II).
|No.||OS (%)||RS (%)||p value1||No.||OS (%)||RS (%)||p value1|
|Age group (yr)|
|Calendar year period|
|Clinical stage of disease|
|No.||OS (%)||RS (%)||p value1||No.||OS (%)||RS (%)||p value1|
|Age group (yr)|
|Calendar year period|
|Clinical stage of disease|
In the multiple regression analysis, only interaction between age groups and follow-up year was significant, which implied the nonproportionality of excess hazards of death across age groups over the follow-up years. For cancers of both colon and rectum, the excess hazards of death in patients aged 70 and above were the highest for the first year after diagnosis compared with other age groups and this difference diminished to unity after 5 years of follow-up. After the adjustment for the interaction between age groups and follow-up, calendar year period and clinical stage of disease were identified as the major factors influencing relative survival in the proportional hazard regression analysis (Table III). For colon cancer, patients diagnosed during 1988–1992 had a 53% lower excess hazard of death compared with those diagnosed during 1968–1972. A 50% higher excess hazard of death was found in Malays than in Chinese. For rectal cancer, patients diagnosed during 1988–1992 enjoyed a 60% lower excess hazard of death than those diagnosed in 1968–1972. Malay patients were found to have 66% higher excess hazard of death compared with Chinese.
|Characteristics||Cancer of colon||Cancer of rectum|
|HR||95% CI||HR||95% CI|
|Age groups (yr)|
|Calendar year period|
|Clinical stage of disease|
Singapore had the highest 5-year age-standardized relative survivals (ASRSs) for cancer of the colon among the developing countries (Table IV). They were higher than in some of the developed countries, such as Japan (Osaka) and England and comparable to those in Finland and Switzerland. However, they were behind those of the USA. Similar results were found for cancer of the rectum except that the 5-year ASRS in Singapore were lower than in Osaka (Japan).
|Country or population||Period||Cancer of colon||Cancer of rectum|
|Other developing countries|
|Chiang Mai, Thailand1||1983–1992||35.6||22.6|
|Khon Kaen, Thailand1||1985–1992||29.1||37.4|
|European pooled average3||1985–1989||47.3||42.5|
The Singapore Cancer Registry receives notifications from multiple sources: medical practitioners, pathology records (biopsies and necropsies), hospital discharge records and death certificates. These multiple sources of registration ensure the accuracy of data and improve the coverage of registration. Over the period of 1968–1992, the proportion of patients registered based on DCO was around 5%.
Singapore uses a unique personal National Registration Identity Card (NRIC) number for each citizen and permanent resident. This NRIC number is used for record linkage among the Cancer Registry, Death Register and the electoral register. The death register is complete and well organized in this small city-state. Furthermore, parliamentary voting is compulsory in Singapore for all citizens. We are therefore able to confirm the vital status of most of the cohort (95%). A very small group of patients may have emigrated or died overseas. These were considered as lost to follow-up and hence were excluded from the analysis. Both genders had similar proportions with loss to follow-up. There was a higher proportion of Indian patients, patients who were less than 40 years old and patients with localized cancers in the lost-to-follow-up group. The proportion of those lost to follow-up also decreased over the study periods. The proportion of DCOs and necropsy cases has also been decreasing over time. Most of the DCOs and necropsy cases do not include information on stage of cancer. The highest proportion of DCOs and necropsy cases was observed in patients older than 70 years old and in Malays. The exclusion of DCOs and necropsy cases may lead to an overestimation of the survival rates. However, DCO, necropsy and lost-to-follow-up cases only consist of 9.0% of all colorectal cancers registered and their exclusion is unlikely to bias the survival estimates greatly.
Clinical extent of disease has long been recognized as a predictor of colorectal cancer.18 In the Singapore Cancer Registry, stage of cancer is coded as localized cancers (cancers restricted to primary organ or tissue of origin or local extension beyond primary organ or tissue or origin), cancers with regional metastases (cancers with lymph node involvement) and distant metastases. In our study, information on clinical extent was obtained in 71% of colon and in 80% of rectal cancers. Despite the high proportion of incomplete information on stage, it remains a strong predictor of survival.
In the present study, higher 5- and 10-year survival rates of colon cancer patients were observed than in patients with cancer of the rectum. Similar results were also found in Cuba14 and Italy.19 This could be due to lower recurrence rate20, 21, 22 and more radical surgery23, 24 for cancer of colon than cancer of rectum.
Older patients were found to have worse prognoses than the young. This result concurs with many population-based reports.15, 16, 25 However, some other studies have shown a decreased survival rate in younger patients compared with the elderly. It is believed that the less than favorable outcome of younger patients with colorectal cancer is because of more aggressive tumor biology, different tumour distribution and delay in diagnosis.26 Isbister27 and Paksoy et al.28 found that more malignant mucinous adenocarcinoma occurred with higher frequency in young patients. However, these studies were based on clinical series and could be biased by specific patient selection. In our study, mucinous adenocarcinoma only accounted for 4.1% of all colorectal cancers. Among these, 1.9% occurred in patients younger than 60 years and 2.1% in those over 60 years.
The influence of ethnicity on the survival of colorectal cancer patients has been well studied. In the USA, studies on ethnic disparities among cancer survivors generally show a higher cancer survival rate for whites than for blacks.29, 30 The ethnic variation in survival has several possible explanations, including differences in attitudes toward invasive procedures, extent of disease and utilization of health care. In Singapore, the survival of Malays is significantly lower than that of Chinese and Indian. Their stage distributions were similar. The proportion of localized cancer was 35.9, 35.0 and 39.4% in Chinese, Malays and Indians, respectively. Correspondingly, the proportion of cancer with distant metastases was 15.2, 14.9 and 15.6%. Hence, the most likely explanation for lower survival rates in Malays may be the social and cultural differences in attitudes toward catastrophic illnesses. Anecdotal accounts from practicing oncologists do support the impressions that Malays are less willing to pursue aggressive therapies.
In Singapore, significant progress in survival of colorectal cancer patients was observed over the study period.This is in parallel with the dramatic national development in socioeconomic and health care services in Singapore. A comprehensive national cancer center has been established and standard therapeutic protocols are adopted through tertiary care centers. Surgical services were widely available through the secondary and tertiary health care facilities. The proportion of surgically resected patients correlated positively with the relative survival probability in the EUROCARE study.31 It is believed that with the easier availability of surgical services and increasing education and socioeconomic status of cancer patients, the proportion of colorectal cancer patients receiving surgery may have increased significantly. In an ongoing study in Singapore, 86.1% of colorectal cancer patients in the period 1993–1995 received surgery (W.B. Du, unpublished data, 2000). Advances of adjuvant therapy could also have contributed to the significant progress in survival. In Singapore, chemotherapy and radiotherapy have been well developed at the tertiary care level. Currently, preoperative and postoperative chemotherapy have become common treatment modalities for advanced colorectal cancer. Newer and more acceptable agents have been introduced and hence a greater proportion of patients receive chemotherapy. The management of side effects of chemotherapy has also improved dramatically, increasing the compliance rate.32
In Singapore, age-standardized mortality for colorectal cancer has been fairly stable (16.7/100,000 and 19.2/100,000 for 1978–82 and 1988–92 in males; 14.6/100,000 and 14.4/100,000 in females). However, the colorectal cancer incidence rate has increased from 31.0/100,000 to 42.2/100,000 in males and 26.7/100,000 to 29.4/100,000 in females over the same period.1, 10, 11 Welch et al.33 demonstrated that increasing 5-year survival may not be due to improvement in cancer treatment. Other factors like changes in diagnostic capabilities, lead time bias, stage migration and change in stage distribution may also contribute to the improved survival.34 For example, patients who had micrometastases could be previously diagnosed as localized cancer, but with modern diagnostic techniques, they may be categorized as patients with metastases.
Differences in methods of data capture may account for some interregistry differences in survival. However, the higher survival of developed countries might largely be attributed to the rapid advances in early diagnosis and treatment techniques.19 Different distributions by stage and anatomic subsite could also explain the differences. The proportion of localized colorectal cancer differed greatly in different countries (USA 1986–1991: 38%; Singapore 1983–1992: 33%; Chiang Mai, Thailand 1983–1992: 5%).14 It has been reported that right-sided tumors of the colon have a poorer prognosis compared with left-sided cancers, which are more common in developing countries.14 However, a recent study in Singapore showed no difference in the survival rates between patients with right-sided and left-sided colorectal cancer.34
Over the 25-year period, the survival rates of colorectal cancer patients improved remarkably, for localized cancers and regional metastases. Those at early stages benefited much more from treatment compared with patients with advanced stages. This further enforces the importance of early detection of colorectal cancers. However, endoscopic screening for colorectal cancer is currently not widely practised.35, 36 Noninvasive screening methods on a national scale have not proved to be cost-effective.37, 38 Screening of high-risk populations remains the most pragmatic approach for colorectal cancer control at this stage of knowledge and practice. The 25-year study period is in parallel with the rapid national development of Singapore. Our study displays a unique changing pattern of survival experience of colorectal cancer patients from a developing country changing toward a developed country.
An International Cancer Transfer (ICRETT) fellowship was awarded to W.B. Du by the International Union Against Cancer (UICC), Geneva, Switzerland and allowed analysis of the data to be carried out at the International Agency for Research on Cancer (IARC), Lyon, France.
- 1Trends in cancer incidence in Singapore 1968–1992, Singapore Cancer Registry Report No. 4. Singapore: Singapore Cancer Registry. 1996. 1–217., , , .
- 2Cancer incidence in Singapore, 1993–1997, Singapore Cancer Registry Report No. 5. Singapore: Singapore Cancer Registry, 2000. 1–161., , , .
- 3Singapore Department of Statistics. Singapore census of population 1990, statistical release 1: demographic characteristics. Singapore: Singapore Department of Statistics, 1991. 1–159.
- 61998. 1–105. Available from URL: http://www.cancerregistry.fi/surv2/ [accessed Dec 1, 1999]., , . Surv2: relative survival analysis program, software manual, program version 2.02β,
- 8Population estimation and projection: method for marketing, demographic and planning personnel. New York: Greenwood Publishing Group, 1992. 43–60..
- 9The life table and its applications. Malabar: Robert E. Krieger Publishing, 1984. 1–316..
- 10Department of Statistics. Yearbook of statistics, Republic of Singapore, 1968 to 1997. Singapore: Singapore Department of Statistics, 1968–1997.
- 11Registrar-General of Births and Deaths. Report on registration of births and deaths, Republic of Singapore, 1968 to 1997. Singapore: Singapore Immigration and Registration, 1968–1997.
- 14Cancer survival in developing countries. IARC Scientific Publication No. 145. Lyon: International Agency for Research on Cancer, 1998. 1–173., , .
- 15SEER Cancer statistics review, 1973–1994. National Cancer Institute. NIH Pub. No. 97-2789. Bethesda, MD: National Cancer Institute, 1997. 1–300., , , , .
- 16Survival of cancer patients in Europe: the EUROCARE-2 Study. IARC Scientific Publication No. 151. Lyon: International Agency for Research on Cancer, 1999. 1–572., , , , , , , .
- 17Survival of cancer patients in Osaka, 1975–89. Osaka: Osaka Cancer Registry, 1998. 1–205., Japan.
- 24Surgery for cancer of the colon and rectum. Ann Oncol 1994;5(Suppl 3): 91–95..
- 25Survival of cancer patients in Europe: the EUROCARE study. IARC Scientific Publication No. 132. Lyon: International Agency of Research on Cancer, 1995. 1–444., , , , , .
- 35Cancer: causes, occurrence and control. IARC Scientific Ppublication No. 100. Lyon: IARC, 1990. 1–352., , , , , , , .