Updated Trends in Cancer in Japan: Incidence in 1985–2015 and Mortality in 1958–2018—A Sign of Decrease in Cancer Incidence

Background Unlike many North American and European countries, Japan has observed a continuous increase in cancer incidence over the last few decades. We examined the most recent trends in population-based cancer incidence and mortality in Japan. Methods National cancer mortality data between 1958 and 2018 were obtained from published vital statistics. Cancer incidence data between 1985 and 2015 were obtained from high-quality population-based cancer registries maintained by three prefectures (Yamagata, Fukui, and Nagasaki). Trends in age-standardized rates (ASR) were examined using Joinpoint regression analysis. Results For males, all-cancer incidence increased between 1985 and 1996 (annual percent change [APC] +1.1%; 95% confidence interval [CI], 0.7–1.5%), increased again in 2000–2010 (+1.3%; 95% CI, 0.9–1.8%), and then decreased until 2015 (−1.4%; 95% CI, −2.5 to −0.3%). For females, all-cancer incidence increased until 2010 (+0.8%; 95% CI, 0.6–0.9% in 1985–2004 and +2.4%; 95% CI, 1.3–3.4% in 2004–2010), and stabilized thereafter until 2015. The post-2000 increase was mainly attributable to prostate in males and breast in females, which slowed or levelled during the first decade of the 2000s. After a sustained increase, all-cancer mortality for males decreased in 1996–2013 (−1.6%; 95% CI, −1.6 to −1.5%) and accelerated thereafter until 2018 (−2.5%; 95% CI, −2.9 to −2.0%). All-cancer mortality for females decreased intermittently throughout the observation period, with the most recent APC of −1.0% (95% CI, −1.1 to −0.9%) in 2003–2018. The recent decreases in mortality in both sexes, and in incidence in males, were mainly attributable to stomach, liver, and male lung cancers. Conclusion The ASR of all-cancer incidence began decreasing significantly in males and levelled off in females in 2010.


Introduction
Globally, the incidence of major cancers is entering a decreasing phase. For example, a significant decrease in age-standardize rate (ASR) has been observed for colorectal, male lung, female breast, and cervical cancer incidence in North American 1,2 and European countries, 3,4 as well as Asian population. 5 These decreasing trends have been interpreted as having resulted from effective cancer control policies, including tobacco control and screening interventions. [6][7][8][9][10] By contrast, Japan is reported to be experiencing a significant increase in ASR of cancer incidence for all major cancer sites except stomach and liver. 11-13 These reports are relatively outdated, however, with the most recent year of diagnosis being 2012, and subsequent trends in incidence in Japan have yet to be identified. Further, trends in cancers other than major cancer sites have not been sufficiently documented. 11 22 We analyzed site-specific cancers and all cancers combined with reference to the International Classification of Diseases (ICD) version 10 codes (C00-C96, additionally D00-D09 for incidence, and C00-C97 for mortality).
Twenty-five cancer sites were selected according to the list of cancers adopted by the National Cancer Center, Japan, 23 which were the same as those analyzed in our previous analysis. 11, 12 We defined "major cancer sites" as the five leading cancers, and "sub-major cancer sites" as the sixth to tenth leading cancers in the latest cancer statistics (either in males or females and in incidence or mortality). 23 Specifically, major cancer sites were stomach, colon/rectum, liver, pancreas, lung, female breast, uterus, and prostate; A c c e p t e d V e r s i o n 8 sub-major cancer cites were esophagus, gallbladder and bile ducts, ovary, urinary bladder, kidney and other urinary organs (except bladder), thyroid, and malignant lymphoma. For these major and sub-major cancers, we discussed potential factors underlying the observed trends in incidence and mortality. We added the analysis of all-cancer incidence excluding stomach, stomach and liver, prostate and female breast to examine the effect of these influential cancer sites. ASR were standardized to the 1985 model Japanese population for cancer incidence and mortality. A Joinpoint regression model 24 was applied using the Joinpoint Regression Program version 4.7.0.0, developed by the U.S. National Cancer Institute. In the Joinpoint Regression analysis, the number of incidence or death was assumed to follow a Poisson distribution; the maximum number of joinpoints was set at five; the minimum number of observations from a joinpoint to either end of the data was set at two; and the minimum number of observations between two joinpoints was set at three.
To identify cancer sites contributing to the recent decrease in all-cancer mortality rates, we calculated the degree of contribution of each cancer site using the same method as that adopted in our previous study. 12 Briefly, we A c c e p t e d V e r s i o n 9 calculated the average APC (AAPC) during the last 10 years for the trend of all-cancer and site-specific ASR of mortality for each sex, calculated the amount of change in ASR by multiplying the 10th power of (1+AAPC) by the ASR in the first year of the 10-year period, and then calculated the proportion of each cancer site in terms of the amount of change. For cancer incidence, since a joinpoint was observed during the last 10 years for both sexes (more specifically, the trend changed from significant increase to significant decrease or levelling off), the contribution of each cancer site was calculated using the same method as that used for cancer mortality, during each of the last significant increasing segment and the subsequent decreasing segment (if significant). This study was approved by the institutional review board of the National Cancer Center, Japan (2019-202).
Results Trends for sub-major cancers are shown in Figure 1-2. Table 1 shows the results of the Joinpoint regression analysis on the trends in all-cancer incidence in the three selected prefectures. For males and females combined, the ASR of all-cancer incidence intermittently increased from 1985 through 2010 with a significant APC of +1.0% between 1985-1996 and +1.7% between 2000-2010, and then levelled off after 2010.
Similar patterns were seen in the separate analyses for males and females; for males the decrease after 2010 was statistically significant. Overall results were the same when stomach and liver cancers were excluded, when prostate or female breast cancer were excluded (Table 1) [2000][2001][2002][2003][2004]. Major cancers that showed a significant decrease during the most recent segment (the period including the most recent year) were stomach, liver, and lung. Among sub-major cancer sites, esophagus, kidney and other urinary organs except bladder, and malignant lymphoma showed a significant increase in the most recent segment, while gallbladder and bile ducts and urinary bladder showed a significant decrease. For females, significantly increasing major cancers in the most recent segment were colon, rectum, (also colon and rectum combined), pancreas, lung, cervix uteri, and corpus uteri (also uterus as a whole). Notably, a long-term increase in breast cancer (APC: 4.0% in 1985-2010) stopped in 2010. Thyroid cancer significantly increased in 2002-2008 (APC: 6.5%), but levelled off thereafter. Significant decreases were seen in the most recent segment for stomach and liver. Among sub-major cancer sites, esophagus, ovary, kidney and malignant lymphoma showed significant increases in the most recent segment, while gallbladder and urinary bladder showed a significant decrease. Among the other cancer A c c e p t e d V e r s i o n 12 sites, oral cavity and pharynx, skin and multiple myeloma showed a significant increase in both sexes. Figure 2 shows the contribution of cancer sites to the significant increase or decrease in all-cancer incidence. For males, prostate cancer accounted for 64.5% of the most recent significant increase in all-cancer incidence between 2000 and 2010. The contributions of other cancer sites were less than 10% (lung: 9.3%, malignant lymphoma: 5.8%, kidney and other urinary organs except bladder: 5.4%, and oral cavity and pharynx: 3.8%). For females, the largest contribution of cancer site to the significant all-cancer increase in incidence between 2004 and 2010 was breast (51.1%), followed by thyroid (8.8%), lung (8.6%), and colon/rectum (7.2%). For males, all-cancer incidence significantly decreased after 2010, and this decrease was mainly accounted for by stomach (41.1%), lung (26.8%), and liver (24.1%) cancers. For females, there was no significant increase in all-cancer incidence after 2010.   (also colon and rectum combined), liver, lung, and prostate. Among sub-major cancer sites also, all cancer sites showed a significant decrease during the most recent segment, except malignant lymphoma, which significantly decreased in 2001-2005. For females, significantly increasing major cancers in the most recent segment were pancreas, breast, cervix uteri, and corpus uteri, (also uterus as a whole). Similarly to males, all the remaining major cancer sites showed a significant decrease during the most recent segment, except colon, which significantly decreased in 1993-2008 and levelled off thereafter. Among sub-major cancer sites, all cancer sites except kidney and malignant lymphoma showed a significant decrease during the most recent segment. Figure 4 shows the contribution of specific cancer sites to the significant decrease in all-cancer mortality in the most recent 10 years (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018). For males, stomach cancer accounted for 29.8% of the decrease of all-cancer incidence, followed by liver (25.2%) and lung (22.3%). These three sites accounted for 77.3% of the all-cancer decrease. Esophagus, and gallbladder and bile ducts accounted for less than 10% (7.1% and 4.2%, respectively). For females also, stomach, liver, and lung accounted for nearly 75% of the A c c e p t e d V e r s i o n 15 all-cancer decrease (34.4%, 28.7%, and 11.8%, respectively). Unlike the result in males, however, the contribution of gallbladder and bile ducts was slightly larger than that of lung (12.6%), while ovary contributed 3.7%.
The contributions of cancer sites to the significant changes in all-cancer incidence and mortality under 75 years old are shown in eFigures 1 and 2, respectively. The largest contributions of prostate in males and breast in females were the same as in the results for all ages. There was no significant decrease in all-cancer incidence when age was restricted to under 75 years old. The large contributions of stomach, liver, lung, and gallbladder and bile ducts to the recent reduction in all-cancer mortality were also the same as the result of all ages. combined. There was a marked divergence between the trends in incidence and mortality in both males and females. For males, the divergence became wider after the late 1990s due to the decrease in mortality. For females, the gap between incidence and mortality widened constantly.  show the trends in incidence and mortality of major and sub-major cancer sites, respectively. Similar to all cancers combined, a divergence between A c c e p t e d V e r s i o n 16 incidence and mortality was a common feature observed in almost all cancer sites. The results of cancers other than major and sub-major sites are shown in eFigure 3. Table 5 summarizes the description of observed trends in incidence and mortality and potential interpretations for each cancer site. Decreases in exposure to major risk factor such as infectious agents and tobacco smoking were considered to be associated with the decreases in incidence and/or mortality of related cancers (stomach, liver, lung, and urinary bladder).
Effects of the introduction and dissemination of cancer screening were considered to have been reflected in the trends in incidence and/or mortality of several cancers (stomach, colon/rectum, female breast, and prostate), among which the increase in prostate cancer incidence was most remarkable. Improvements in diagnostic measures and treatments were common factors associated with the divergence of incidence and mortality. Potential overdiagnosis was considered to be included in the increase in incidence of prostate and thyroid cancers. This study analyzed the trends in cancer incidence and mortality in Japan with updated representative datasets. A notable finding was that the ASR of all-cancer incidence started to significantly decrease in males and level off in females in 2010, after a long-term intermittent increase. 11,12,25 The leading cancer sites that contributed to the past long-term increase were prostate in males and breast in females, but these slowed down or levelled off during the first decade of 2000s. For males, the main contributing cancer sites to this significant decrease were stomach, liver and lung cancers. The levelling off of ASR of female breast cancer incidence after 2010 A c c e p t e d V e r s i o n 18 observed in the present study is an unprecedented phenomenon. 11-13 11 The mortality of female breast cancer 12 also slowed a slowing down of its increasing trend. Long-term increase in incidence as well as mortality can be interpreted as the effect of changes in reproductive factors in Japanese females (Table 5). 11,29,30 This effect might be converging in breast cancer, but cancers of the corpus uteri and ovary, which share common reproductive risk factors, continued to increase in incidence. The participation rate of female breast cancer screening (mammography) has been increasing in Japan, and early-stage cancer and carcinoma in situ of the breast was reported to have increased in a study using a prefectural cancer registry. 31 Together with the slowing down of the increasing trend in mortality observed in the present study, these changes in trends could have partially reflected the dissemination of breast cancer screening. 11

Discussion
The significant decrease in ASR of all-cancer incidence in males after 2010 was accounted for by stomach, liver and lung cancers. These cancer sites also contributed to the decrease in all-cancer mortality in both males and females (77.3% and 74.9%, respectively; Figure 4). Stomach cancer consistently decreased during the whole observation period with regard to both incidence  (Table 5). 11,32,33 A study using data from a hospital-based registry   (Table 5).
Liver cancer is another site that showed a dramatic decrease in incidence and mortality. As discussed in previous literature, the long-term decrease in liver cancer in Japan is mainly due to the decrease in the prevalence of hepatitis C virus (HCV). 11,13,37 The observed acceleration both in incidence and mortality in 2008 or 2010 (Tables 2 and 4) can be interpreted as a reflection of therapeutic improvements made in the early 2000s such as pegylated interferon in 2004 and a protease inhibitor (Telaprevire) in 2011. 11,38-40 Improvements in differential diagnosis could have also affected the divergence between incidence and mortality since 1980s ( Table 5).
Cancer of the gallbladder and bile ducts had a similar pattern of trends to that of liver cancer. Chronic infections have been proposed as one of the risk factors for gallbladder cancer as well as gallstones and obesity (Table 5). [41][42][43] Control over communicable diseases could have resulted in the reduction of incidence rate of gall bladder cancer. 44  The decrease in lung cancer incidence in males was a phenomenon that had never been observed in previous literature. 11-13 Trends in lung cancer incidence in Osaka by histological type revealed that the ASR of adenocarcinoma continuously increased, whereas those of squamous and small-cell carcinomas decreased from 1990s, which was interpreted to be the result of the spread of diagnostic use of computed tomography and the decreasing trend in smoking prevalence, respectively. 46,47 Another possibility is the shift from nonfilter to filtered cigarettes in the consumption of tobacco products in Japan (Table 5), which may be more influential because the increase in adenocarcinoma was observed even before the introduction of major diagnostic advances. 48 The decrease in overall lung cancer incidence observed in the present study could have reflected the predominant effect of declining smoking prevalence, albeit that analysis stratified by histological type is needed to clarify this possibility. Cancers of the kidney and urinary tracts are also strongly related to tobacco smoking, 49,50 but no similarity was found between the trends in these cancers and smoking prevalence except for bladder cancer incidence in males (Table 5). An important feature of our results is that a decrease in incidence was not observed for colorectal cancer, which can be prevented by organizational screening. The ASR of colorectal cancer has been significantly decreasing in many countries. [1][2][3]5 Using simulation modeling techniques one study revealed that the reduction in colorectal cancer in the U.S. was a combined effect of cancer control measures for prevention and screening. 6 Cervical cancer also showed a sharp contrast; the ASR of this cancer has been consistently decreasing overseas, including the Republic of Korea, 1,2,5 whereas the present study showed a significant increase in incidence and mortality, just as was observed in our previous analysis. [11][12][13] The increase in mortality of cervical cancer (cancer of the corpus uteri as well) in Japan should be interpreted with caution because it could have included the shift from cancer of the 'uterus, not otherwise specified (NOS)'. However, the proportion of NOS had been stable since the late 1990s, and the increase was also observed in incidence. 11,13 Cervical cancer can be prevented by a combination of organizational screening and human papillomavirus (HPV) vaccination. 7,9,51,52 In Japan, the national HPV vaccination program has been substantially halted by the fear of potential adverse effects. 53,54 A simulation Pancreatic cancer was another example that showed a long-term increase both in incidence and mortality. Increase in risk factors such as type 2 diabetes may be related to the increase in incidence 56 and mortality as well.
Improvements in diagnostic measures and biopsy for histologic confirmation have also been proposed as underlying factors of the increasing trend in earlier years. 57 Monitoring cancer incidence trends is useful in examining the possibility of overdiagnosis at a population level. In the U.S., prostate, female breast, skin, kidney, thyroid, and lung cancers have been listed as examples of potential overdiagnosis, characterized as a sharp increase in incidence in the absence of a clear change in mortality. 58 In the present study, this typical pattern seemed to be observed for prostate and thyroid cancers ( Figure 5 and Table 5). A common background factor of these cancers is the availability of Some of the reduction in cancer mortality observed in the present and previous studies 11-13 likely reflects improvements in the prognosis of cancer patients. This effect can be seen in the divergence between trends in incidence and mortality ( Figure 5), which is consistent with the evidence of improved diagnosis, treatment, and disease management cited in Table 5.
Indeed, several studies on hematological cancers showed that the introduction of a new drug or treatment was associated with a reduction in A strength of the present study is the representativeness of the data.
Mortality data were from the national vital statistics and based on a complete mandatory reporting system. Incidence data were from three prefectures, but the representativeness of the data in terms of secular trends has been validated. 12,19 One of the limitations of the present study is that the trends in cancer incidence might have been affected by an improvement in the completeness and data quality of prefectural cancer registries. Indeed, even in the three present prefectures that have long-term high-quality data, there was a slight increase in completeness and quality indices during the first decade of 2000s (eFigure 4). The observed increases in incidence in this period might therefore reflect an improvement in data completeness. A second limitation is that our analysis was only descriptive. Furthermore, the cancers we A c c e p t e d V e r s i o n 26 analyzed were not grouped into clinically relevant subtypes. As stated above, further research is required to clarify factors underlying the observed cancer trends, such as analyses according to clinical stage or histological type and modelling approaches.
b. Standardized to Japanese model population in 1985.      a. Incidence: data from Yamagata, Fukui, and Nagasaki prefectures, Mortality: national data.
b. Standardized to Japanese model population in 1985.

Lung, trachea : Males All ages
Year at diagnosis or death  a. Incidence: data from Yamagata, Fukui, and Nagasaki prefectures, Mortality: national data.
b. Standardized to Japanese model population in 1985.  a. Incidence: data from Yamagata, Fukui, and Nagasaki prefectures, Mortality: national data.
b. Standardized to Japanese model population in 1985.

Urinary bladder : Males All ages
Year at diagnosis or death  b. Standardized to Japanese model population in 1985.

Malignant lymphoma : Males All ages
Year at diagnosis or death