|
|
Centre for Chronic Disease Prevention and Control
Cancer
The Impact of Stage and Treatment on Short-term Survival for
Lung, Colorectal and Breast Cancer Using Population-based Cancer Registry Data
Jun Zhang, MSc, MD, MPH; WK (Bill) Evans, MD, FRCPC (ORCC);
Kathy Clarke, BA, BEd; Yang Mao, PhD, Cancer Bureau, LCDC, Health Canada
This article first appeared in Oncology Advisor, Volume 6, Number 2, December
1998 and is reproduced by permission of Maclean Hunter Healthcare/Santé.
Abstract
The purpose of this study was to examine the impact of stage and treatment on survival of lung,
colorectal and breast cancer patients. Data for patients diagnosed with primary cancers between 1994 and 1996 were
obtained from the Ottawa Regional Cancer Centre (ORCC) through the initiative of the Canadian Coalition on Cancer
Surveillance (CCOCS). These data included age, sex, TNM stage, initial treatment, and follow-up status. Two-year
relative survival rates (RSRs) were calculated according to age, sex, stage, and treatment. Multivariate relative
survival analysis was applied to examine the relative risk (RR) of dying from cancer and 95% confidence intervals
(95% CI). A statistical package developed at the French Cancer Registry was used for data analyses.
Patients with cancers diagnosed at advanced stages (stage III or IV) had significantly lower
RSRs than those diagnosed at early stages (stage I or II). Two-year RSRs for patients having surgery or surgery
with chemotherapy and/or radiotherapy were higher than for those having chemotherapy and/or radiotherapy, but no
surgery. Patients having no initial treatment had the lowest RSRs for each cancer site.
For lung cancer patients, females had higher survival rates than males, and patients aged 75-86
had lower survival rates than those less than 75 years of age. The lowest 2-year RSR was observed among patients
aged 75-86 diagnosed at stage IV (1.5%). Overall, age at diagnosis and sex had little influence on survival for
colorectal cancer. Age had little impact on survival for breast cancer. The lowest survival rate was seen among
women aged 75-86 diagnosed at stage IV (9.0%).
Stage is the strongest prognostic factor for cancer survival identified in this study. Initial
treatment and age also play important roles. Effect modification between age at diagnosis and stage needs further
study. Systematic collection and analysis of stage and treatment data through the national population-based cancer
surveillance system proposed by the CCOCS will help fill this information gap.
Introduction
Cancer incidence, mortality, and survival are essential components in cancer surveillance systems
that support cancer prevention and control. Population-based cancer survival statistics can be used to investigate
the effects of prognostic factors such as stage and treatment on cancer survival and to compare survival results
worldwide. However, few population-based cancer survival statistics using cancer stage and treatment data are available.1
The Canadian Coalition on Cancer Surveillance (CCOCS) was established in 1996 through the initiative
and leadership of the National Cancer Institute in Canada and in collaboration with many Canadian cancer stakeholder
groups. One of the goals of the CCOCS is to establish a national cancer surveillance system to assemble high-quality
cancer case data, including patient stage information for stage-specific cancer survival analysis. The Ottawa Regional
Cancer Centre (ORCC) is the first institution to participate and provide data for the system. The purpose of this
study is to demonstrate the value of a national cancer surveillance system by calculating relative survival rates
(RSRs) for patients diagnosed with lung, colorectal, or breast cancer in the Eastern Ontario region using data
obtained from ORCC and to estimate relative risk (RR) of dying from these cancers according to stage and treatment.
Materials and Methods
Cancer Patients
A population-based cancer registry at the ORCC covers most of Eastern Ontario and part of the Outaouais
region of Quebec, a population of 1,158,000. Since 1994, centre policy requires that patient staging information
be recorded at the time of cancer diagnosis using TNM classification (tumor, nodes, and metastasis).2
This study included patients with histologically confirmed primary cancers diagnosed between 1994 and 1996 for
which TNM stage information was complete. Patients without information on TNM stage or date of diagnosis and those
aged 87 and over were excluded from the study population. Patients living in the province of Quebec at the time
of diagnosis were also excluded. The study included 75% of the population for whom data were available.
Age at cancer diagnosis was classified into three groups (0 to 59, 60 to 74, and 75 to 86). To
study the impact of stage on survival of colorectal and breast cancer patients, stage I and stage II were grouped
together as the reference category. Initial treatment within four months of cancer diagnosis was classified into
four categories (surgery with chemotherapy and/or radiotherapy, surgery only, chemotherapy and/or radiotherapy,
and no treatment). Treatment data for breast cancer patients were classified into two categories (surgery and no
surgery) due to the small number of breast cancer patients receiving no treatment.
Patient survival status (dead, alive, or lost to follow-up) was followed until December 31, 1996.
ORCC uses a mixed passive and active follow-up of patients: Ontario death files are linked with the ORCC database
annually; death records for patients who die in one of two Ottawa region hospitals (i.e. Civic or General Hospital)
are sent to ORCC weekly; letters/notifications to family on patient's death are received periodically; death announcements
in newspapers are checked daily; and survival status of patients having moved out of the province is followed through
annual contact with their family doctors.
Statistical Analysis
The RSR is the ratio of the observed survival rate for cancer patients to the expected survival
rate of the general population in the same age and sex group. In this study, the 2-year RSRs were calculated using
a method proposed by Ederer.3 Age-, sex-, stage-, and treatment-specific RSRs were computed for each
cancer site. Multivariate relative survival analyses were conducted using a proportional hazard model proposed
by Cox4 and a method described by Esteve et al.5 The data were analyzed using Ontario detailed
life tables for 19916 with a statistical package developed at the French Cancer Registry.7
Results
During the study period 1994 to 1996, 1,001 lung cancer (International Classification of Diseases
[ICD] 9 = 162), 879 colorectal cancer (ICD9 = 153, 154) and 1,542 female breast cancer (ICD9 = 174) patients identified
from the ORCC database met the selection criteria and were included in the survival analysis. The distribution
of the study population is presented in Table 1.
Table 1: Multivariate
survival analysis of patients with lung, colorectal
and breast cancer, 2-year follow-up, ORCC, 1994-1996
1 adjusted for all variables in the table for each cancer site
2 the baseline is TNM Stage I and II
3 due to small numbers in the "no treatment" category for breast cancer patients, the four
categories were redefined as two: surgery and no surgery
Relative Survival Rates
Overall, the 2-year RSR for lung cancer patients was poor: only 22.5% survived two years. Females
were more likely to survive than males: 25.7% of females survived two years in comparison with 20.6% of males.
Patients aged 75 to 86 had a lower RSR (16%) than those less than 75 years of age (24%). Stage at cancer diagnosis
was strongly associated with survival (Figure 1).
The 2-year RSRs were 55.3% for stage I, 43.7% for stage II, 26.8% for stage III, and only 5.7% for stage IV. Patients
having surgery (surgery with chemotherapy and/or radiotherapy, 46.8%; surgery only, 67.7%) were more likely to
survive than those who did not have surgery (i.e. chemotherapy and/or radiotherapy, 17.8%; no treatment, 11.9%).
The lowest RSR was seen among patients aged 75 to 86 diagnosed at stage IV: only 1.5% survived two years.
The overall RSR for colorectal cancer patients was 71.7%. Age and sex had little impact on survival. The 2-year
RSRs for patients aged 75 to 86 (70.5%) were similar to those for patients less than 75 years old (71.6%). The
2-year RSRs were also comparable between male (72.3%) and female (70.5%) patients. Stage-specific RSRs ranged from
92.1% for stage I and II, 81.2% for stage III, to 27.3% for stage IV (Figure
2). Patients having surgery (with chemotherapy and/or radiotherapy, 73.0%; surgery only,
83.1%) were more likely to survive than those who did not have surgery (chemotherapy and/or radiotherapy, 56.8%;
no treatment, 43.1%).
The overall 2-year RSR for breast cancer patients was 94.2%. Women aged 75 to 86 had a slightly
lower RSR (88.5%) than those less than 75 years of age (94.3%). The 2-year RSRs among women diagnosed at stage
I and II (97.6%) were slightly higher than those diagnosed at stage III (85.8%); however, the rate for women diagnosed
at stage IV was much lower (36.7%) (Figure 3).
Breast cancer patients having surgery were more likely to survive (96.0%) than those who did not have surgery,
(86.2%). Age also had a significant impact on stage-specific RSRs: approximately 9% of women aged 75 to 86 diagnosed
at stage IV survived two years, in comparison with 37.6% of women aged less than 59 and 50.6% of women aged 60
to 74.
Relative Risk of Dying
Multivariate relative survival analyses were conducted by adjusting all variables including age,
sex, stage, and treatment in the Cox proportional hazard model (Table 1). Consistent with relative survival rates, female lung cancer patients were less likely to die of the
disease than male patients (RR = 0.8). Elderly patients aged 75 to 86 had a 50% higher risk of dying from lung
cancer within two years than those aged less than 75 years of age. Patients diagnosed with stage II, III, and IV
had a higher relative risk of dying from lung cancer than those diagnosed at stage I. The risk of dying from lung
cancer was not different between patients having surgery only and those having surgery with chemotherapy and/or
radiotherapy. However, the risk of dying was significantly higher for patients not undergoing surgery, especially
for those receiving no treatment (RR = 2.7).
The results of multivariate analyses indicate that age at diagnosis and sex of patients had no
significant impact on survival for colorectal cancer. TNM stage, however, did play an important role. Patients
at stage III were 3.3 times as likely to die of colorectal cancer as those at stage I or II. Patients diagnosed
at stage IV had the highest risk of dying from the disease (RR = 17.7). Patients having chemotherapy and/or radiotherapy
but no surgery were 1.6 times as likely to die of colorectal cancer as those having surgery with other treatments.
Patients having no treatment in the first four months were 2.7 times as likely to die as those in the reference
group (i.e. surgery with chemotherapy and/or radiotherapy). The relative risk of dying among patients having surgery
only was not statistically different from those having surgery with other treatment.
Women with breast cancer aged 75 to 86 were more likely to die of this disease than younger women
(RR = 2.7). The risk of dying among women aged 60 to 74 was non-significantly lower than among women aged less
than 60 years of age (RR = 0.7). Being diagnosed at stage III or IV increased the risk of dying from this disease
within two years by a factor of 5.3 and 33.7, respectively. Women not having surgery were 1.9 times as likely to
die as those having surgery.
Discussion
This study may be subject to selection bias. The referral process in remote areas is biased towards
younger patients and physicians tend to assign more aggressive treatment for younger patients; urban patients may
have more knowledge about and request more treatment; a certain proportion of cancer cases were excluded because
stage information was not recorded in the ORCC database (this latter is unlikely to bias the sample as the characteristics
of the group for which information is available are unlikely to differ substantially from the group for which information
is unavailable). In addition, potential prognostic factors such as histology and grade were not controlled in our
study, due to the small number of cancer patients; and the 2-year follow-up period is relatively short, particularly
for colorectal and breast cancer.
The strengths of this study are that it provides, for the first time, stage- and treatment-specific
RSRs for cancer patients based on records readily available from a large population-based registry in Eastern Ontario.
It also adopts standard methods used worldwide, providing survival statistics suitable for international comparisons.
Two-year RSRs calculated in our study are comparable to those published in the U.S.8
Previous reports on relative survival studies for lung cancer are mostly based on the localized,
regional, and distant (LRD) staging system.9 LRD stage, histology subtypes, age, and sex are reported as important
prognostic factors in lung cancer survival in early studies.10,11 But the effect of initial treatment
on lung cancer survival has not been extensively studied. Our study confirms that TNM stage is the strongest prognostic
factor in lung cancer survival, followed by initial treatment, age, and sex. Due to the small number of lung cancer
patients diagnosed with small cell histology, we did not include histology data in the survival analysis. Significant
effect modification between age and stage was observed in our study. Further study is needed to confirm these results
and to explore possible explanations.
Both LRD stage and TNM stage have been reported as important prognostic factors in colorectal
cancer survival.12,13 The results of our study suggest that age and sex have little impact on colorectal
cancer survival, but TNM stage and initial treatment strongly influence 2-year RSRs. The impact of initial treatment
on colorectal cancer survival requires further study, since few population-based survival analyses have reported
comparable results.
Studies of relative survival analyses for breast cancer consistently report that LRD stage, TNM
stage, histology subtypes, and age significantly influence cancer survival.14-16 As with previous studies,
our study demonstrated that TNM stage as well as age at diagnosis play an important role in breast cancer survival.
The results of our study suggest that TNM stage is the strongest prognostic factor for the survival
of lung, colorectal, and breast cancer patients. Initial treatment and age also play important roles. Certain effect
modifications between age and stage need to be investigated in further studies. Systematic collection and analysis
of high-quality data from population-based cancer registries will help address this gap by providing information
about the effect of stage and treatment on cancer survival. A national population-based cancer surveillance system
currently being built under the guidance of the CCOCS will provide these survival statistics.
Acknowledgments
The authors wish to acknowledge Debbie Read and Judy Morriss of the ORCC and Cathy Mills of the
University of Western Ontario for their contribution to preparing data for analysis and producing descriptive statistics.
References
- McLaughlin JR, Sloan MR, and Janovjak DP. Cancer survival in Ontario. The Ontario Cancer
Treatment and Research Foundation, Toronto (ON), 1995.
- Sobin LH, Wittekind Ch, editors. International Union Against Cancer (UICC). TNM Classification
of Malignant Tumors, fifth edition. John Wiley and Sons, Inc., New York, 1997.
- Ederer F and Heise H. The effect of eliminating deaths from cancer in general population
survival rates. National Cancer Institute, 1959. (Methodological note 11, End results evaluation section)
- Cox DR. Regression models and life tables. Journal of Royal Statistical Society 1972,
34:187-220.
- Esteve J, Benhamou E, Croasdale M, and Raymond L. Relative survival and the estimation of
net survival: elements for further discussion. Stat Med 1990, 9:529-538.
- Statistics Canada. Life tables, Canada and provinces, 1990-1992. Statistics Canada Catalogue
84-537. Ottawa (ON), 1994.
- Hedelin G. RELSURV 2.0: a program for relative survival. Laboratory for Epidemiology
and Public Health, Faculty of Medicine, France, 1997.
- National Cancer Institute. SEER cancer statistics review, 1973-1994. NIH Publication
No. 97-2789. Bethesda (MD), 1995.
- Henson DE, Ries L, and Shambaugh EM. Survival results depend on the staging system. Sem
Surg Oncol 1992, 8:57-61.
- Ries LAG. Influence of extent of disease, histology, and demographic factors on lung cancer
survival in the SEER population-based data. Sem Surg Oncol 1994, 10:21-30.
- Engeland A, Bjorge T, Haldorsen T, and Tretli S. Prognosis of patients with lung cancer diagnosed
in Norway, 1954-93. Cancer Causes Control 1998, 9:57-65.
- Chu KC, Tarone RE, Chow WH, Hankey BF, and Ries LAG. Temporal patterns in colorectal cancer
incidence, survival, and mortality from 1950 through 1990. J Natl Cancer Inst 1994, 86:997-1006.
- Roncucci L, Fante R, Losi L, Di Gregorio C, Micheli A, Benatti P, Madenis N, Ganazzi D, Cassinadri
MT, Lauriola P, and Ponz de Leon M. Survival for colon and rectal cancer in a population-based cancer registry.
Eur J Cancer 1996, 32:295-302.
- Nab HW, Hop WC, Crommelin MA, Kluck HM, Coebergh JWW. Improved prognosis of breast cancer
since 1970 in south-eastern Netherlands. Br J Cancer 1994, 70:285-288.
- Gamel JW, Meyer JS, Feuer E, and Miller BA. The impact of stage and histology on the long-term
clinical course of 163,808 patients with breast carcinoma. Cancer 1996, 77:1459-1464.
- Chu KC, Tarone RE, Kessler LG, Ries LAG, Hankey BF, Miller BA, Edwards BK. Recent trends
in U.S. breast cancer incidence, survival and mortality rates. J Natl Cancer Inst 1996, 88:1571-1579.
|