According to the second national retrospective mortality survey conducted during 1990–1992, cancer had become the second most common cause of death in China.1, 2 The age-standardized mortality rate (using the Chinese census data of 1982 as standard population) was 94.4 per 100,000 (123.6 for males and 66.3 for females).1, 2 However, the patterns of cancer mortality in urban and rural areas were quite different.1, 2, 3, 4, 5 As cancer becomes increasingly important as a public health problem, the need to evaluate the trends in incidence and mortality for the major cancers becomes necessary, not only to provide clues to etiology but also to help in providing projections of the future burden of disease.
Trends in cancer incidence and mortality in China have been published for limited populations and time periods,6, 7, 8, 9, 10 especially for the city of Shanghai, where cancer registration has been established since the mid 1960s,6, 7 and also for some specific cancers in high-risk areas, such as esophageal cancer in Linxian and Cixian counties and stomach cancer in Changle county.11, 12, 13, 14 However, to date there has been no attempt to evaluate trends of cancer incidence or mortality at the national scale. With respect to incidence, this is virtually impossible since most cancer registries in China were established rather recently, cover only a very small percentage of the national population and are mostly located in urban areas in the eastern (coastal) provinces, and are therefore not representative of the country as a whole.1 Mortality data are, however, available for more extensive populations.
At present, there are 2 ongoing national mortality reporting systems in China. One is operated by the Center for Health Information and Statistics (CHIS) of the Ministry of Health, the results of which are regularly published by World Health Organization (WHO). The other is a project of the Chinese Academy of Preventive Medicine, called the “Disease Surveillance Points” (DSP) sample survey. The nature of the information for both systems are described by Yang et al.3 In a comparison study of the mortality rates from these two sources, Yang et al.4 concluded that although the CHIS reporting system is not entirely satisfactory in representing the national population (in terms of age, sex and urban-rural area), it probably provides a better estimate of cancer mortality rates at the national level than does the DSP system. In addition, the population size covered is quite large, deriving from a 10% sample (non-random) of the whole population and comprising between 100 and 120 million persons.3 Several year's data from the CHIS system are now available.15
Most previous mortality analyses have relied on the visual inspection of plots of mortality rates to describe the changes in the patterns, calculating a constant average annual percent change over time and using age-period-cohort models to evaluate whether the changes can be attributed to a period or a cohort effect. Recently, the joinpoint regression model was developed to identify and describe the occurrence of changes in recent trends in distinct periods of time.16, 17 In this paper, based on the cancer mortality data from the CHIS system during a 13-year period (1987–1999), we investigate time trends in cancer mortality by area (urban/rural), sex and age group in China, using a joinpoint regression model.
MATERIAL AND METHODS
Two data sources are used in this paper: 1) Mortality data from 1987 to 1999, from CHIS as reported to WHO;15 and 2) the national census data for 1990 in China.
The CHIS data set includes mortality rates for 10 cancer sites, and trends for the 9 most common sites (excluding bladder) were examined. They were cancers in the oesophagus, stomach, lung, liver, colon-rectum, female breast, cervix uteri, nasopharynx and leukaemia, as well as all cancers combined. Age-adjusted annual mortality rates over the 13 years were calculated by the direct method, standardized to the national census population in 1990, and trends by site, sex, age group (15–34, 35–44, 45–54, 55–64, 65–74, 75+ (plus, for leukaemia, age groups 0–4 and 5–14) and area (urban and rural) plotted on a logarithmic scale against year.
Changes in the age-standardized mortality rate over the 13-year period were analyzed for each cancer by fitting the joinpoint regression model. For a model with k joinpoints, the ith response is
where ϵ is the error and μ is defined implicitly. The analysis starts with the minimum number of joinpoints and tests whether one or more joinpoints are statistically significant and must be added to the model. The tests of significance use a Monte Carlo permutation method. In the final model, with the best fitting joinpoints where the rate changes significantly, each joinpoint informs of a statistically significant change and an estimated annual percent change (EAPC) is computed along with its 95% confidence intervals. The model expresses the logarithm of the expected number of deaths as a linear function of the factors and follows a Poisson distribution. Significant changes include changes in direction or in the rate of increase or decrease. The annual percent change is tested to determine if it is different from the null hypothesis of no change.16, 17 The analysis was performed using the “Joinpoint” software from the Statistical Research Applications Branch of the National Cancer Institute, U.S.A.18 Age effects were analyzed based on the Poisson regression model for the age-specific mortality rates by cancer, sex and area. The annual percent change for each trend was also evaluated.19, 20, 21
Table I shows the number of deaths and age-standardized mortality rates, by cancer site, sex and area (urban/rural), at the beginning (1987) and the end (1999) of the study period, together with the overall average percentage change in the rate in the 13-year period. The periods between joinpoints, where the trend changes significantly, and the annual percent change for each sub-period by cancer, area and sex are also shown in Table I. There are between 1 and 3 such periods for each set. The estimated annual percent changes by age group (15–34, 35–44, 45–54, 55–64, 65–74 and 75+) are shown in Table II based on the Poisson regression model. Graphs showing these results [the age-standardized and age-specific cancer mortality trends by site, sex and area (urban/rural)] are presented as Figures 1–10 (for leukaemia, certain age groups are shown in Fig. 10 and using a different Y-axis scale than used in other cancers in order to show the younger age-specific mortality trends for leukaemia).
Table I. The Number of Deaths and Age-Standard Mortality Rates (in 1987 and in 1999), Joinpoint Analysis (1987–1999) For Some Common Cancers By Area and Sex
For all cancer sites combined, mortality rates are higher in males than females and in urban areas than in rural areas (Fig. 1). During the study period 1987–1999, mortality rates declined slightly in rural areas for both sexes (−0.1% per year in males and −0.4% in females). In urban areas, rates increased until 1989 (annual increase 4.6%), followed by a significant steady decrease of −1.6% until 1996 and then increased again at 1.3% per year in males; in females, mortality rates dropped significantly (−1.5% per year) until 1996 and then increased 2.2% annually (Table I). For the age-specific mortality rates, there is a significant decline for age groups 0–4 and 55–64 in both sexes and areas. Changes in other age groups/areas were mostly small or nonsignificant (Table II).
The mortality rates for oesophagus cancer are considerably higher in rural areas than in urban areas and among males than females. Mortality rates in rural females are almost the same as among urban males (Fig. 2). Throughout the period, mortality decreased steadily in both areas and sexes and more in urban areas and for females (Table I). These declines are seen in all age groups and populations and for the most part are statistically significant (Table II).
For stomach cancer, the patterns are similar to those of oesophagus cancer: higher rates in rural than urban areas, among males than females and declines in all groups during 1987–1999 (except in urban males, who showed a nonsignificant increase during 1987–1989) (Fig. 3 and Table I). The annual changes by age were also similar to those of oesophagus cancer: all age groups showed a declining trend, except young people (age between 15–34) in rural areas (slightly increasing but not statistically significant for both sexes). The decrease was significant in most age groups, especially in urban areas (Table II).
The mortality rates from lung cancer are much higher in urban than rural areas, and among males than females (Fig. 4). Rates have been rising throughout the study period in all 4 populations, more obviously in rural areas than urban areas (Table I). These increases occurred in all age groups in both sexes, and in both urban and rural areas (except for males aged 55–64, females aged 45–54 and the youngest age group in urban areas) (Table II).
For liver cancer, mortality rates are higher in rural than urban areas, and in males than in females (Fig. 5). Over the 13-year period, there were small increases in mortality in rural areas, in most age groups (Table I and Fig. 5). In urban areas, the mortality rates decreased in young and middle age groups (age between 15–64), especially among females, while there was a small increase in the oldest age group, as in rural areas (Fig. 5 and Table II).
Mortality from colon-rectum cancer is higher in urban than rural areas. Within each area, rates in males are higher than in females, although mortality in urban females exceeds that in rural males (Fig. 6). Time trends were quite varied between the four groups (Table I) but showed some distinct recent increase (Fig. 6). With respect to age-specific rates, the youngest age group (age between 15–34) shows declining trends, especially in urban areas, while the rates in the older age groups (65+) are generally increasing (except in rural females).
Female breast cancer
Female breast cancer mortality increased during the study period in both areas, and this is more obvious in rural areas where the rates, nevertheless, remain lower than in urban females (Fig. 7). The estimated annual percent change in rural areas was +0.9% until 1996, with a dramatic jump (11.6% per year) thereafter, while in urban areas there was a steady, slight, but statistically significant increasing trend (0.9% annually) throughout the 13-year period. The oldest age group (above age 75) showed significantly declines in both areas, but there were different trends in the youngest age group (age 15–34), with a dramatic increase in young rural women (4.8% per year), but a significant decline among young urban women (−2.2% annually).
Cervix uteri cancer
The decreasing trend in cervix cancer mortality is sharper in urban areas (dramatically decreasing at −9.4% per year until 1993 and then at −4.3% per year) than in rural areas (−3.5% annually for the whole period) (Fig. 8 and Table I). The decline is confined to the older age groups (older than age 55); younger women showed increasing trends (for urban women aged 35–44, the increase of 4.1% per year was a statistically significant) (Table II).
Mortality rates for nasopharynx cancer are much lower than for the other cancers presented in this study; the rates are similar in rural and urban areas but higher in males than in females (Fig. 9). The age-standardized mortality declined during 1987 to 1999 in both areas and sexes, especially in urban areas, although there was a nonsignificant increase in urban males during 1987–1992 (Table I).
Mortality rates for leukaemia are similar in urban and rural areas but higher in males than females (Fig. 10). The age-standardized mortality rates decreased in all populations between 1987 and 1999, except for a dramatic jump in urban males after 1996 (Table I). In the age-specific mortality data, a sharp decline is seen in the youngest age groups (younger than age 4) during the 13 years, rates increased a bit in age group 5–14, and then declined again in both areas and sexes during the study period (Table II).
The changes that have occurred in age standardized mortality rates in the 13-year period have resulted in changes in the ranking of the major causes of cancer death, as shown in Fig. 11. Lung cancer has become a relatively more common cause of death in rural populations, while breast cancer is becoming relatively more frequent in females in both areas.
This is the first analysis of time trends in cancer mortality on a national scale in China. The data-set used, from the Center for Health Information and Statistics (CHIS), is considered to provide the most comprehensive and representative information on mortality patterns in the country.3, 4 The information on deaths is based on the death certificates collected and coded by the local vital statistics offices in the participating counties and municipalities. Death certification is considered to be complete in these areas. Cause of death is obtained from medical sources, except for deaths at home, where it may be based on the information provided by relatives. Home deaths are relatively few, however, except in some of the more rural counties. The quality of information has probably varied little over time. Populations at risk data are derived from local population registers maintained by the Security Bureau in the corresponding county or municipality.
Since only 13 years of mortality data are available at present, period and cohort effects were not considered when fitting the models. Joinpoint analysis provides a useful summary of the direction and size of the trend in recent years, and detects when a significant change in trend occurs. It thus provides a much clearer picture of what is happening during a given period than does a single summary statistic for the whole interval.16, 17 Age-standardized and age-specific mortality trends, along with the corresponding estimated annual percent change based on the fitted model, provide practical clues to the prevention and control of cancer, and some epidemiological and etiological hypotheses can be developed and tested.
The focus of this paper is the change in age-specific and age-standardized rates of cancer mortality in China. Changes in these indices may be the result of changing in the risk of disease due to changes in exposures to environmental risk factors, dietary and socioeconomic, or to the effects of early detection and treatment mediated via governmental health care programs, policies and resource management.
Differences in trends in urban and rural areas may provide clues to possible differences in environmental exposures or health care interventions; urban populations have significantly higher mortality rates for lung, colon-rectum and female breast cancers than rural populations, and lower rates for oesophagus, stomach, liver and cervix cancers. However, although the urban and rural areas included in the CHIS survey were geographically constant over the study period, there are ongoing demographic changes within these populations, as a consequence of the massive rural-urban migration that has been taking place and that is projected to last through the next few decades. Rapid economic development has resulted in significant migration from the rural hinterlands of central China to the booming coastal provinces like Beijing, Shanghai, Tianjin, Guangdong, Liaoning and Jiangsu. According to official statistics, the percentage of the population living in urban areas increased from 19.4% in 1980 to 30.4% in 199822 and was estimated to be 36.2% in 2000.23 Three factors are driving this urbanization: the huge “excess population” in agriculture, the income gap between rural and urban employment, and the growing labor demand of urban industries and service sectors.24 Although the statistics cited are the official documentation of China's urbanization, they give an incomplete picture of the reality of population movement. A large number of temporary laborers (so called “floating population”) and a certain amount of “illegal” rural-urban migration have been tolerated. It is estimated that Beijing and Shanghai each have a floating population of 2–3 million people.24 Thus, as well as the huge differences in socioeconomic circumstance, life-style, diet and health service provision between 2 areas, this substantial migration of populations within China is likely to result in significant changes in cancer mortality within rural and urban areas.
Since the economic reforms started 2 decades ago, the national economic and health care situations have improved dramatically. The per capita gross domestic product increased 2.5 times in the last decade (US$339 in 1990 and US$853 in 2000).23 The percentage of GDP devoted to health care also increased from 4.1% in 1991 to 5.3% in 2000 and per capita health care costs increased from 76.6 Chinese Yuan in 1991 to 376.4 Chinese Yuan in 2000.25 Adult literacy increased from 87.0% of males and 68.1% of females in 1990 to 93.8% and 81.2 % respectively, in 2000.23 Within this improving socioeconomic profile, there have been impressive changes in the organization and provision of health care, facilities and human resources in China. This is most obvious with respect to primary health care, health services and education, public hygiene, medical diagnostic methods and treatment techniques and the quality of registration and classification of the causes of death. Professional training has provided thousands of doctors and nurses in the urban areas, and local health workers (sometimes referred to as “barefoot doctors”) with sufficient medical skills to provide basic health care to local residents in rural areas. This improved health infrastructure allows more extensive and effective early detection, diagnosis and treatment of cancer. Programs for the early detection of cervix and breast cancers have been introduced in some urban areas, and for nasopharynx and stomach cancer for certain high-risk populations. These, together with improvements in therapy, may have contributed to better survival, which in turn will be responsible for some of the changes in the observed mortality rates.26 The economic reform programme, as well as leading to an improvement in the average values of economic indicators, has probably increased disparities in income and health care provision, which coupled with the extensive migration referred to above, has probably also influenced mortality patterns over time.
Along with the changing social and economic conditions, there are profound changes occurring with respect to environment and lifestyle. For example, Helicobacter pylori infection, an established risk factor for non-cardia gastric cancer, has been shown to be linked to crowded living conditions, family size, sharing a bedroom, low socioeconomic status, low education level and poor sanitation, and infrequent handwashing before meals.27, 28, 29, 30 The improved socioeconomic status, sanitary condition of housing, education level and other lifestyle improvements may be responsible for the declining risk of gastric cancer, through a reduction in the prevalence of infection with Helicobacter pylori.
Today, the Chinese population has a more diversified diet structure than in the past, eating considerably more fruit, vegetables and all sorts of animal products. The daily calorie supply increased from 2711 per capita in year 1990 to 3044 per capita in the year 2000 and the daily protein supply from 65 to 83 grams per capita.23 According to estimates of the Food and Agriculture Organization of the United Nations (FAO/UN), the average per capita supply of vegetables increased from 47.7 kg in 1980 to 205.4 kg in 2000, and fruit from 5.9 kg to 41.6 kg.31 During these 2 decades, there has been a huge increase in consumption of eggs (from 2.5 to 16.2 kg), fish (from 4.4 to 24.7 kg) and meat (from 13.7 to 49.2 kg) per person per year.31 These dramatic changes might be expected to influence the risk of certain cancers in China. Besides this increased availability and consumption of foodstuffs, substantial improvements in food storage (the widespread use of the refrigerator), and more convenient transport have increased availability of fresh food and reduced the need for pickled, salted or nitrified food preservation, which have been identified as risk factors for upper gastrointestrial tract cancers in China.32, 33, 34, 35, 36 These substantial improvements in living conditions and diet may have contributed to the statistically significant decline in the mortality of cancers such as oesophagus, stomach and nasopharynx, especially in urban areas.3, 7, 30, 32, 33, 34 It has been hypothesized that chronic deficiencies of multiple micro-nutrients, such as vitamins A and C, selenium and riboflavin, are etiologically linked with esophageal and stomach cancer, and supplements of such vitamins and minerals may have helped to lower the risk of these cancers, as suggested by studies in Linxian and Linqu.37, 38 Epidemiological studies have found that chronic infection with hepatitis B virus (HBV) and dietary aflatoxin exposure are the major etiological determinants of the high rates of liver cancer in China, the role of hepatitis C virus (HCV) infection is less important.39, 40, 41 Vaccination against HBV infection in infants has been introduced recently but will not impact on incidence and mortality rates for many years. Exposure to dietary aflatoxin is mainly through the intake of moldy foods such as corn, soy source and peanuts in some regions of the country. Dietary changes, leading to the abandonment of corn as one of major dietary staples, and improvements in food storage and transport may have resulted in declines in aflatoxin exposure and intake in recent years.42, 43 Campbell et al.42 have suggested that increased consumption of foods of plant origin might also reduce the risk of disease. Even the declines in leukaemia may reflect better nutrition during pregnancy, which may be associated with a reduced rate of childhood leukaemia.44
However, increasing intake of a nutritionally richer diet may increase the risk for certain cancers. Campbell noted that as a society begins to acquire additional capital and resources, people quickly begin to consume nutritionally richer diets, particularly diets richer in added fat and sugar and foods of animal origin.42 This phenomenon has obviously also happened in China. The increasing intake of dietary fat, poultry and pork, together with the overall increase in dietary calories and a reduction in physical activity levels, particularly among urban residents (caused by increased industrialization and transformation of life style and leisure activities),45 are resulting in an increase in obesity. According to the China Health and Nutrition Survey (1989–1997), the prevalence of overweight has tripled in men and doubled in women.46 Among pre-school children, the overall prevalence of obesity has increased from 4.2 to 6.4%, largely in urban areas where it increased from 1.5 to 12.6% during that period.46 Such changes may contribute to the increasing mortality from cancers of the colon-rectum and female breast, which are associated with high intake of animal fat, body size and low physical activity.47, 48, 49, 50, 51, 52 The link between breast cancer and increasing intake of dietary fat and high blood cholesterol may be indirect, through increased childhood growth rates, causing earlier menarche and greater body mass, thus increasing the risk for breast cancer later in life.42 The rising trend in female breast cancer may also be related to the delay in age at first childbirth and the decline in the number of births. Between 1990 and 1999, the birth rate dropped from 21.1 to 15.2 per 1,000 population.23, 53 These demographic changes are reflected in the number of children aged 0–14, which accounted for 27.7% of the Chinese population in 1990 and 22.9% in 2000; the rate of population increase dropped from 1.4% to 0.5% correspondingly.23
The large declines in mortality from cervix cancer may be partly due to improvements in survival, as treatment becomes more effective and more generally available. The introduction of screening programs based on Pap smear cytology may also have contributed, as suggested by the more marked declines among urban women. Screening programs have been mainly implemented by local health departments and in occupational settings, and urban women have better access to these services.54 It is also quite likely that there has been a general reduction in the incidence of disease, due to a decreased exposure to risk factors of cervix cancer in older generations of women. Infection with oncogenic human papillomavirus (HPV) is the major determinant of cervical cancer; however other factors such as high parity and poor genital hygiene are independent risk factors,55, 56, 57, 58 and changes in the later 2, both more marked in urban areas, will have contributed to a decline in the mortality rate. The dramatic decline in the incidence of cervix cancer in urban Shanghai between 1972 and 1974 and between 1993 and 1994 has been ascribed to a combination of screening, and to reduction in risk (changing sexual behavior and use of barrier contraception).7 However, it is notable that the mortality declines have ceased, and even reversed, among younger women, and this is more obvious in urban rather than rural areas. In Shangdong Province, Li et al.8 observed that the decreasing mortality for cervix uteri cancer was largely a cohort phenomenon. They suggested that the increased rates among younger women reflect rapid changes in sexual mores, with increasing high-risk sexual behavior, and greater prevalence of infection with HPV and other sexually transmitted agents.59, 60 The increasing trends in other sexually transmitted diseases make this a plausible hypothesis.61
During the last 2 decades, the annual consumption of cigarettes in China increased from 500 billion in 1980 to 1,800 billion in 1996; two thirds of men now become smokers before age 25 and few give up. The average daily cigarette consumption by males was 1 in 1952, 4 in 1972 and 10 in 1992.62 One in 3 of all the cigarettes smoked in the world today are smoked in China. This increasing epidemic of cigarette smoking is the major factor that underlies the increasing trends in lung cancer mortality, but it may also contribute to the rising mortality from other smoking-related cancers, such as stomach, oesophagus, liver and leukaemia.62, 63, 64, 65 A large-scale retrospective study (done in 1989–1991) and one ongoing prospective study of smoking and death have been carried out in China.62, 66, 67 They suggest that tobacco currently causes 12% of adult male deaths (0.6 million deaths in 1990 and 0.8 million expected in 2,000). Among the tobacco-related deaths, 45% were from chronic lung disease, 15% from lung cancer and 5–8% from each of cancers in oesophagus, stomach and liver, stroke, heart disease and tuberculosis. The hazards of tobacco are similar for both sexes. The investigators predict that, if current smoking patterns persist, the country will face a huge epidemic of smoking deaths, for example, one third (about 100 million) of Chinese males now aged 0–29 will be killed by smoking, and this figure would reach 3 million a year by the time the young smokers of today reach middle and old age (around the year 2050).62, 67 Control of the tobacco epidemic is the greatest challenge to public health in China at the beginning of the 21st century. However, the smoking situation among Chinese women is quite different from that among men. Few Chinese women have smoked, and a high prevalence of smoking is only seen among older women in big cities. In 2 large nationwide surveys the prevalence of smoking among women aged 15–24 was 0.5% both in 1984 and in 1996.62 Nevertheless, the mortality rates from lung cancer in Chinese females are relatively high, for example, the estimated age standardized (world standard) mortality in Chinese women in 2000 was 13.5 per 100,000, compared with 1.8 per 100,000 in India.68 This observation has highlighted the impact of other factors, such as environmental tobacco smoke (ETS), coal smoke, cooking fumes, air pollution, exposure to indoor radon and occupational exposures.69, 70
Cancer is accounting for an increasing part of the health burden in China, not only because of the overall increases in risk of the disease, but even more importantly because of huge population growth and aging. When the People's Republic of China was founded in 1949, it had a population of just 550 million. Only 3 decades later its population was nearly 1 billion.71 This rapid population growth was the result of past high levels of fertility and resulted in national policies to reduce population growth. The total fertility rate had fallen to just 1.8 children per woman by 2000.72 Nevertheless, population growth will continue for some time; in its most recent (medium variant) projection,71 the UN Population Division estimated that Chinese population was 1.28 billion in 2000 and would reach to 1.45 billion in 2020, 1.49 billion in 2040, and then slightly decline to 1.46 billion in 2050. During last decade, the crude birth rate declined (from 21.0‰ in 1990 to 15.2‰ in 2000)25 and the expectation of life at birth rose from 68.6 in 1990 to 70.8 in 199725 and is projected to be 76.3 in the years 2025–2030.72 The proportion of elderly (aged 65 or over) in the population has increased steadily from 7.5% in 1950 to 8.6% in 1990 and to 10.1% in 2000, and is projected to be 19.5% in 2025 and 29.9% in 2050.71 Even with currently observed mortality rates, this growth and aging of the population would dramatically increase the number of cancers and cancer deaths in the coming decades of the new century. The projected number of cancer deaths would rise from 1.4 million in 2000, to 1.8 million in 2010, to 2.4 million in 2020 (i.e., by more than 30% in each of the next 2 decades) and then will reach 3.7 million in 2050. Of these cancer deaths, 49% would be people aged 65 or over in 2000, rising to 69% of cancer deaths in 2050, because of population aging.68 Implementing an effective cancer control programme is clearly important, including treatment, early diagnosis and, most importantly, prevention, in which, tobacco control measures are an urgent priority.