In Japan, cancer has been the leading cause of death since 1981. The total number of cancer deaths is actually increasing annually. The number of deaths due to cancer in 2004 was 320,358, exceeding the 288,680 deaths due to cardiovascular disease, and comprised approximately 31% of total deaths.1 Age-standardized mortality rate (ASR) (using the World Standard Population) was 99.4 per 100,000 (138.0 for men, 69.9 for women). Recently, the trend for ASR of overall cancer deaths has tended to decline for both men and women, but has shown diversification for different anatomical sites. Because cancer is one of the major targets of disease control programs in Japan, an accurate understanding of current trends in overall cancer and common cancers using statistical methods is important.
Recently, Joinpoint analysis has been commonly used to describe the changing trends over distinct periods of time and significant increases or decreases in cancer mortality for foreign countries.2–5 However, similar analyses of trends for cancer have yet to be performed in Japan. We therefore analyzed the long-term trends in overall cancer and major cancers mortalities in Japan between 1958 and 2004 using Joinpoint analysis.
Material and methods
Data regarding the number of deaths due to cancer were obtained according to sex, anatomical sites and age (grouped by 5-year age groups: 0–4, 5–9, 10–14, 15–19, 20–24, 25–29, 30–34, 35–39, 40–44, 45–49, 50–54, 55–59, 60–64, 65–69, 70–74, 75–79, 80–84, 85+) for the years 1958–2004 from the National Vital Statistics in Japan. Corresponding Japanese population data were derived from the Statistics Bureau, Ministry of Internal Affairs and Communications: the Japanese Population Census Report from 1960 to 2000 (conducted every 5 years) and population estimates for the remaining years. During the specific calendar period, cancer deaths in Japan were coded according to the International Classification of Diseases, 7th Revision (ICD-7) for 1958–1967, 8th Revision (ICD-8) for 1968–1978, 9th Revision (ICD-9) for 1979–1994 and 10th Revision (ICD-10) for 1995–2004. Sites for analysis included overall cancer and the 15 most common cancers: esophagus (C15), stomach (C16), colon (C18), rectum (C19-C21), liver and intrahepatic bile duct (liver) (C22), gallbladder and extrahepatic bile duct (gallbladder) (C23-C24), pancreas (C25), lung and bronchus (lung) (C33-C34), breast (C50, D05), uterus (C53-C55), ovary (C56), prostate (C61), urinary bladder (C67), malignant lymphoma (C81-C85, C96) and leukemia (C91-C95). Annual ASRs for the 47-year time span were calculated by the direct method using the World Standard Population.
Long-term trends in ASRs of cancer in Japan were analyzed using Joinpoint regression model.6 Joinpoint regression program is a trend analysis software developed by the US National Cancer Institute for the analysis of data from the Surveillance Epidemiology and End Results Program. This method describes changes in data trends by connecting several different line segments on a log scale at “joinpoints.” Analysis starts with the minimum number of joinpoints (i.e., 0 joinpoint, representing a straight line) and tests for model fit with a maximum of 4 joinpoints. Tests of significance use a Monte Carlo permutation method. In addition, an annual percent change (APC) in ASRs for each line segment and the corresponding 95% confidence interval were estimated. The APC is tested to determine whether a difference exists from the null hypothesis of no change (0%). In the final model, each joinpoint informs a statistically significant change in trends (increase or decrease) and each of those trends is described by an APC.5–7
Tables I and II show the crude death rate and ASR by cancer site at the beginning (1958) and end (2004) of the study period, together with results of Joinpoint analysis for ASRs in men and women for all ages, respectively. The results of Joinpoint analysis on ASRs from overall cancer, stomach, colon, rectum, liver, pancreas, lung, prostate, breast and uterus are presented in Figure 1. Changing trends in ASRs in Joinpoint analysis will be described in 3 patterns as follows in this study. Pattern 1, ASRs reached a plateau or started decreasing in the 1990s after constant increases from 1958. Pattern 2, ASRs decreased continuously from 1958. Pattern 3, ASRs increased continuously from 1958.
Table I. Crude Mortality Rates, ASR1 and Joinpoint Analyses for 1958 Through 2004 in Japan: Men of All Ages
In men, ASRs for overall cancer showed Pattern 1, increasing by 0.32% per year from 1958 to 1989, stabilizing from 1989 to 1993 and 1993 to 1996, and declining by 1.71% per year from 1996 to 2004. ASRs for cancers of the esophagus, colon, rectum, liver, gallbladder, pancreas, lung, prostate and urinary bladder, malignant lymphoma and leukemia also showed Pattern 1. ASRs for cancers of the colon, liver, gallbladder and lung decreased from the 1990s, was similar to the results for overall cancer, whereas ASRs for cancers of the rectum and prostate leveled off from the 1990s. ASRs of malignant lymphoma and pancreatic cancer leveled off from 2001 and 1987, respectively, leukemia decreased from 1987 and cancers of the esophagus and urinary bladder decreased from the 1970s. ASRs for stomach cancer showed Pattern 2, with a clear consecutive decreasing trend since 1958 (Table I and Fig. 1).
In women, ASRs for overall cancer showed Pattern 2, leveling off from 1958 to 1967, decreasing by 0.98% per year from 1967 to 1993, leveling off from 1993 to 1996 and decreasing again by 1.22% per year to 2004. ASRs for cancers of the colon, liver, gallbladder, lung and ovary showed Pattern 1, with ASRs decreasing from the 1990s after an increasing trend. ASRs for rectal cancer and leukemia could also be classified into Pattern 1, although ASRs decreased from the 1970s. ASRs for stomach cancer in women resembled that in men, decreasing continuously by Pattern 2. ASRs for cancers of the uterus and urinary bladder could also be categorized into Pattern 2, although the trend for cancer of the uterus leveled off from the 1990s and cancer of the urinary bladder increased slightly after 1993. ASRs for esophageal cancer showed similar changing trends to uterine cancer, although the gradient of declining ASR was more gradual. ASRs for cancers of the pancreas and breast and malignant lymphoma showed Pattern 3, increasing continuously to 2004 (Table II and Fig. 1).
In general, ASRs for cancer in Japan in the most recent line segment (from the last joinpoint until 2004) have been declining for the majority of cancers, including overall cancer. The few exceptions include increases in mortality from cancers of the pancreas, breast and urinary bladder and malignant lymphoma in women. For other cancers such as cancer of the rectum, pancreas, prostate and urinary bladder and malignant lymphoma in men and cancer of the esophagus and uterus in women, the most recent trend has remained relatively stable (Tables I and II).
During the study period, an increasing trend in crude death rate was observed in overall cancer and for all observed cancer sites in both men and women, except for cancers of the stomach in both sexes and cancer of the uterus in women. Crude death rates for stomach cancer in both sexes showed a decreasing tendency during the observed period, whereas crude death rate for uterine cancer in women decreased until the early 1990s, then increased slightly thereafter.
Compared to those in all ages, truncated (35–64 years) mortality rates were higher in both men and women, with similar patterns in most cancer sites. Different patterns between all ages and truncated age were only observed for overall cancer in men, which decreased continuously during the study period in truncated mortality rates, and uterine cancer in women, for which truncated mortality rates showed an increasing trend in the most recent period (increased slightly from the early 1990s) (data not shown).
To the best of our knowledge, this study offers the first nationwide analysis of trends over time using statistical testing for mortalities from cancer in general and the most common cancers in Japan. The use of Joinpoint analysis has allowed statistical testing of directions and sizes of trends in mortality rates for various cancers, detecting some significant changes. Accordingly, Joinpoint analysis provides a much clearer picture of what is happening during a distinct period in specific terms (identifying the years in which significant changes in trends occurred) than a single summary trend statistic.5, 6
A decline in mortality rates of overall cancer starting in 1996 was identified in both sexes in Japan. The overall decline in recent decades is mostly attributable to reductions in stomach, lung and liver cancers for men and in stomach and gallbladder cancers for women.
The changing trends in cancer mortality rates may be interpreted as resulting from changes in environmental, dietary and socioeconomic factors and potential consequences of early detection, treatment and various public health strategies, which have profoundly changed along with economic growth in Japan in the postwar decades.
The mortality rate from stomach cancer in Japan still ranks as the highest in the world,1 but has declined markedly in the last few decades, in similar fashion to observations from the United States and some European countries.3, 8 The trend in stomach cancer mortality rate has been attributed to the effects of substantial improvements in food storage and preservation through refrigeration,9–11 a more affluent diet with increased consumption of fresh vegetables and fruit9–11 and reductions in the prevalence of Helicobacter pylori infection.9, 12 The recent decline trend in mortality rate could also be explained in part by efforts made in early detection13, 14 and improved treatment, as mortality rates declined almost in parallel with incidence in the 1970s and the gap between these rates widened from the early 1980s to the early 1990s.15
Prostate cancer remains lower in Japan than in Western populations,16, 17 although the mortality rate has increased more rapidly than in those countries.16 The traditional Japanese diet is high in isoflavones which are chiefly derived from soybean products and might be protective against this kind of tumor.18, 19 The leveling off of the mortality rate might have been partly caused by improvements in treatment, as incidence has continued to increase since 1996.15
Consumption of meat increased sharply from the 1940s to the mid-1970s in Japan20 and this could be argued to represent the major cause of the increasing trend in colon cancer until the 1990s.21, 22 Meanwhile, increases in alcohol consumption22 and obesity rate20 along with declines in physical activity22 may also account for some of the increase in colon cancer in men.
The mortality rate of breast cancer is high in developed Western industrialized countries and relatively low in Japan and developing countries.23 Continuous increases in breast cancer mortality rate could be interpreted as indicating increasing lifetime exposure to estrogen, related to early menarche, late menopause, late pregnancies and low parity caused by increases in the ranks of working women.21, 23, 24
The influence of smoking on mortality rate for smoking-related cancers other than lung cancer is less visible, because of the influence of other factors. Declines in mortality rates from lung cancer were not a simple reflection of the decreasing prevalence of smoking over the last few decades in men. According to data from Japan Tobacco Industry, trends in smoking prevalence in Japan over the past 40 years have shown a decline in smoking habits among men, but a leveling off in women. Life-time smoking prevalence among men showed a temporary decrease in the late 1930s birth cohort and increased until the 1950s birth cohort.25, 26 These trends corresponded with changing patterns in lung cancer mortality rates in men, which was lower in the late 1930s birth cohort and increased successively after the 1940s birth cohort.26–28 As for women, we have no clear explanation for the observed declines in mortality rate, as smoking prevalence was not sufficiently high to impact lung cancer mortality trends. Nevertheless, mortality rates among women aged 65–74 years decreased in 2003 and 2004, at least reflecting the lowest life-smoking prevalence observed in the 1930s birth cohort.25 Careful monitoring is needed for recent birth cohorts in women, as life-smoking prevalence increased after the 1940s birth cohort.25 This analysis suggests that aggressive promotion of antismoking measures targeting the birth cohort born after the 1940s is necessary for further reductions in lung cancer mortality in Japan.
The epidemiology of liver cancer is distinctive in Japan, where chronic infection from hepatitis C virus (HCV) rather than hepatitis B virus (HBV) plays the major role in the etiology.29 Prevalence of antibodies against HCV was highest (3.4%) in the 1930s birth cohort, whereas prevalence of hepatitis B surface antigen (HBsAg) was highest (1.5%) in the 1940s birth cohort among the first blood donor candidates in Japan.30 These two generations with high prevalence of HCV antibodies and HBsAg corresponded with those observed in liver cancer mortality rates, which peaked in the 1930s and 1940s birth cohorts for men, and in the 1930s birth cohort for women.31 The social problem of HCV and HBV infection may have contributed to increasing mortality rates for liver cancer from the 1970s. National projects have been implemented with the aim of reducing the threat of infection by HCV and HBV since 1970s,32 because of the serious social problems associated with these infections. The declining mortality rate of liver cancer since 1996 was largely interpreted as due to declines among men of 45–64 years and women of 50–69 years, which in turn were mainly ascribed to the decreasing prevalence of HCV infections after the 1930s birth cohort.
The drastic decline in the mortality rates for uterine cancer has been ascribed to improvements in sanitation, early detection and therapy. However, this decreasing trend appears to have been stabilizing since 1993. Notably, truncated mortality rates increased from 1993. Analysis of mortality rates by age suggests that younger women (≤59 years old) have shown consecutive increases since the 1940s–1950s birth cohorts, whereas older women (≥60 years old) have shown a consecutive decreasing after the 1905s birth cohort (data not shown). This change may be partly due to the greater prevalence of infection from oncogenic human papilloma virus caused by changes in sexual behavior among younger women.33 The increase in mortality rate among younger women is offset by declines in older women, leading to the stabilization observed in mortality rate for uterine cancer since the early 1990s. This trend must be monitored among younger women in future.
Cancer screening has been started nationwide for cancers of the stomach and cervix since 1983, for cancers of the breast and lung since 1987 and for colorectal cancer since 1992 under the Health Services Law for the Aged in Japan. Although Pap smears and mammography have been recommended for screening of cervical and breast cancers, respectively, and no methods were recommended for other cancer site screenings in the UICC report of 1990,34 unique cancer screening policies were conducted in Japan except for cervical cancer at that time. Barium X-ray was adopted for stomach cancer screening, and was accepted only for Japan by the UICC report,34 physical examination was adopted for breast cancer screening at first, before being changed to mammography with physical examination from 2000, X-ray and sputum cytology were adopted for lung cancer screening and fecal occult blood tests were adopted for colorectal cancer screening. However, these screening seems unlikely to have made any major contribution to reductions in cancer mortality rates by the year 2004, with the exception of cancers of the stomach and uterus (mainly for the cervix), as participation rates in screenings of eligible populations are still relatively low for cancers of the lung, breast and colorectum in Japan.35
Potential problems exist in the reliability and validity of death certification for cancers, which may vary across calendar years, when interpreting mortality trends. Deaths certificates are usually considered complete in Japan and cause of death is also considered to have been identified with reasonable accuracy. Mortality rates for cancers in both sexes were unchanged during the coding process related to revisions of the ICD classification system, though the classification and coding on death certificates tended to influence secular trends of mortality rates. We did not distinguish between cancers of the cervix uteri, corpus uteri and unspecified uteri, given acknowledged problems in accurate classification from death certificates.36 Colon cancer and rectal cancer mortality rates were analyzed separately, because those changes in diagnosis and certification are unlikely to explain the recent fall and stability in these two sites and differences in risk factors associated with cancer at each site.21, 22, 32, 37–39
Joinpoint analysis in this study demonstrated favorable declines in overall cancer and the majority of observed cancers in Japan in recent years. However, the aging population and overall increases in risk of cancer have still increased the crude death rate and absolute total number of cancer deaths in Japan. An effective cancer control program including primary prevention (such as antismoking), early detection and treatment should thus be implemented to further reduce cancer mortality, with special emphasis on those anatomical sites of cancer showing higher mortality rates and increasing trends.
D. Qiu received a Research Resident Fellowship from the Foundation for Promotion of Cancer Research (Japan) for the 3rd Term Comprehensive 10-year Strategy for Cancer Control.