• Open Access

Trends in cancer incidence and mortality in Osaka, Japan: Evaluation of cancer control activities

Authors


1To whom correspondence should be addressed.
E-mail: itou-yu2@mc.pref.osaka.jp

Abstract

It is essential to analyze trends in cancer incidence and mortality in the evaluation of cancer control activities. Previous studies from Japan, however, described trends in cancer incidence and mortality only qualitatively. There have been few studies that evaluated the trends quantitatively. We calculated age-standardized mortality rates (1968–2006) and incidence rates (1968–2002) for overall cancer sites and for each major site (stomach, colorectal, liver, lung, prostate, breast, and uterus) in Osaka. We applied a joinpoint regression model to the trends in incidence and mortality, in order to identify the joinpoint and estimate annual percentage change. Then, we quantified the contribution of individual cancer sites to the change in overall cancer mortality rate. For the sites that made a major contribution, we estimated the contribution of the incidence reduction to the mortality reduction. In Osaka, the overall cancer mortality started to decrease from 1998. The decrease was largely attributable to the reduction of stomach and liver cancer mortality (73% for men, 53% for women). The reduction of mortality from the two cancer sites could be explained by the decrease in their incidences (more than 80% for stomach, approximately 100% for liver). Female breast cancer incidence and mortality were both increased probably due to lifestyle changes and delayed introduction of an effective screening program among Japanese. In conclusion, the decreased overall cancer mortality in Osaka during the study period was mainly due to natural decreases in the incidence of stomach and liver cancer, which were attributable to the decrease in risk factors. (Cancer Sci 2009: 100: 2390–2395)

It is essential to analyze trends in cancer incidence and mortality in the evaluation of cancer control activities. Almost all of the studies from Japan in the past, however, described trends of cancer incidence and mortality only qualitatively. There have been few studies that evaluated the trends quantitatively.

There has been no national-based cancer registry in Japan. Since 1975, the cancer incidence data as national statistics have been estimated from the data from selected cancer registries that met certain data quality requirements. The number and geographical distribution of registries that provided data for the national estimate were different from period to period. Therefore it is difficult to evaluate the trends in cancer incidence and mortality from the national estimate data.

The Osaka Cancer Registry, established in 1962, is one of the largest population-based cancer registries in the world and its database can be used for the evaluation of long-term trends in cancer incidence. We evaluated the trends in cancer incidence and mortality in Osaka by sex and major cancer sites in order to assess the cancer control program in the prefecture.

Methods

Database.  From the Osaka Cancer Registry database, we calculated age-standardized mortality from 1968 to 2006 and age-standardized incidence rates from 1968 to 2002 for overall cancer sites and for each major site. We used the 1985 Japanese Model Population as the standard population for age standardization. Stomach (ICD [International Classification Code]-10 code C16), colorectal (C18-C21), liver (C22), lung (C33, C34), breast (C50, female only), uterus (C53-C55), and prostate (C61) cancer were selected for analysis, because those sites of cancer were the most common in Japan. In the present study, we could not analyze cervix and corpus uteri separately, because the number of death certificates with ‘uterus, not otherwise specified (NOS)’ was high in the 1970s and 1980s.

Statistical analysis.

Joinpoint regression model.  We applied the piecewise log linear regression model, which is known as the joinpoint regression model, to identify the years when statistically significant changes in incidence or mortality trends occurred, using the Joinpoint 3.3 package (US National Cancer Institute, Bethesda, MD, USA).(1,2) These years are called ‘joinpoints’. The model can estimate the annual percentage change (APC) of each segment between joinpoints and test if it is significantly different from zero (P < 0.05). The logarithmic age-standardized incidence or mortality rate was used as the dependent variable, and the year of diagnosis or death was used as the independent variable in the model. We set the number of joinpoints in each cancer trend to a minimum of 0 and maximum of 3 to find best fit model using permutation test method and assumed constant variance and uncorrelated errors in the calculation. The latest 10-year trends of incidence and mortality were evaluated using the estimated average APC.

Contribution of individual cancer sites to the change in overall cancer mortality rates.  We calculated the contribution of individual cancer sites to the total change in overall cancer mortality rates based on the methods presented in the annual report ‘Cancer Statistics’ from the American Cancer Society since 2007.(3,4) In brief, all cancer sites were divided into two categories, namely ‘cancers with decreasing mortality’ and ‘cancers with increasing mortality’, and the total decrease or increase in each category was calculated, as well as the contribution of individual cancer sites to the total decrease or increase of each category. As a point where the overall cancer mortality starts to decrease, we used the peak year identified by the joinpoint regression model. The contribution of individual sites was calculated using the following equation, where rk(t) represents mortality rate for site k in year t, t0 is the peak year (1998 in this study), and t1 is the latest year available (2006 for mortality):

image

Contribution of the incidence reduction to the mortality reduction.  For the cancer site with decreasing mortality that made a major contribution to the total reduction in mortality described above, we calculated the contribution of the incidence reduction to the mortality reduction. We calculated this contribution using the relative changes for mortality rate and incidence rate, assuming that the contribution of incidence was 100% when the trends in incidence and mortality were parallel.

The relative change for mortality rate (RCmort) was calculated as follows, where M(t) represents the total mortality of the major cancer in the respective year:

image

We assumed that the year of incidence on average preceded the year of mortality for the duration of the median survival time (MST) of each cancer site. The MST was estimated based on the relative survival for cancer patients diagnosed in 1995–1999 in Osaka. Then the relative change for the incidence rate (RCinc) was calculated as follows, where I(t) represents the total incidence of the major cancer in the respective year:

image

Then the contribution of the incidence reduction to the mortality reduction (%) in each cancer site was defined as:

image

When the percentage was more than 100%, we substituted 100% for it.

Results

Trends in overall incidence and mortality.  The results of joinpoint regression analysis are shown in Table 1 for trends of mortality and in Table 2 for trends of incidence. The trends in the overall cancer incidence and mortality by sex are shown in Figure 1. The trends in incidence and mortality by sex for the major cancer sites are shown in Figure 2.

Table 1.   Annual percentage changes (APC) in the trend of age-standardized mortality rate in Osaka, Japan, between 1968 and 2006 for all ages by joinpoint analysis
 Observed age standardized mortality rate (per 100 000)Line segment 1Line segment 2Line segment 3Line segment 4Last 10 years (1997–2006)
19682006YearsAPC95% CIYearsAPC95% CIYearsAPC95% CIYearsAPC95% CIAAPC95% CI
  1. *Statistically significant change compared with zero. AAPC, average annual percentage change; CI, confidence interval.

Men
 All sites222.4217.51968–1974−0.07−0.81, 0.671974–19851.56*1.21, 1.911985–1998−0.02−0.28, 0.24 1998–2006−2.06*−2.53, −1.59−1.80*−2.20, −1.40
 Stomach106.736.01968–1984−2.67*−2.87, −2.471984–1993−3.48*−4.05, −2.901993–1996−0.33−5.64, 5.28 1996–2006−3.10*−3.51, −2.69−3.10*−3.50, −2.70
 Colorectal12.224.41968–19843.49*2.99, 4.001984–19961.48*0.61, 2.351996–2006−0.69−1.67, 0.30    −0.70−1.60, 0.30
 Liver20.732.01968–19865.49*5.07, 5.911986–19970.41−0.55, 1.381997–2006−5.06*−6.13, −3.99   −5.10*−6.10, −4.00
 Lung25.951.11968–19795.24*4.57, 5.921979–19892.81*1.92, 3.711989–2006−0.38*−0.71, −0.05   −0.40*−0.70, −0.10
 Prostate3.47.71968–20062.97*2.58, 3.36         3.00*2.60, 3.30
Women
 All sites142.0106.71968–1999−0.45*−0.54, −0.361999–2006−1.54*−2.41, −0.67      −1.30*−2.00, −0.60
 Stomach54.813.41968–2006−3.75*−3.84, −3.66         −3.80*−3.80, −3.70
 Colorectal10.113.01968–19961.63*1.35, 1.921996–2006−1.49*−2.78, −0.18      −1.50*−2.70, −0.20
 Liver9.110.61996–19735.47*1.75, 9.341973–1979−3.29−6.70, 0.251979–19991.98*1.50, 2.471999–2006−4.32*−6.36, 2.25 −3.00*−4.50, −1.40
 Lung7.914.91968–19883.75*3.28, 4.221988–2006−0.33−0.86, 0.20      −0.30−0.80, 0.20
 Breast5.612.91968–20061.86*1.67, 2.06         1.90*1.70, 2.10
 Uterus22.15.51968–1993−4.96*−5.24, −4.691993–2006−1.01*−1.78, −0.24      −1.00*−1.80, −0.30
Table 2.   Annual percentage changes (APC) in the trend of age-standardized incidence rate in Osaka, Japan, between 1968 and 2002 for all ages by joinpoint analysis
 Observed age standardized incidence rate (per 100 000)Line segment 1Line segment 2Line segment 3Line segment 4Last 10 years (1993–2002)
19682002YearsAPC95% CIYearsAPC95% CIYearsAPC95% CIYearsAPC95% CIAAPC95% CI
  1. *Statistically significant change compared with zero. AAPC, average annual percentage change; CI, confidence interval.

Men
 All sites297.5344.91968–1974−0.38−1.33, 0.571974–19853.11*2.66, 3.571985–19990.07−0.23, 0.371999–2002−4.10*−6.77 , −1.35−1.30*−2.20, −0.40
 Stomach140.467.01968–1974−3.25*−4.46, −2.011974–1984−0.461.13, 0.231984–2000−1.97*−2.28, −1.66 2000–2002−6.31−13.08, 0.97 −3.00*−4.50, −1.40
 Colorectal16.347.51968–19733.11*0.86, 5.411973–19807.38*5.61, 9.181980–19934.48*3.87, 5.091993–2002−1.42*−2.30, −0.53−1.40*−2.30, −0.60
 Liver23.144.31968–19720.25−3.59, 4.241972–19867.82*7.13, 8.521986–19960.19−0.94, 1.32 1996–2002 −5.56*−7.51, −3.57 −3.70*−5.00, −2.40
 Lung29.663.01968–19854.58*4.18, 4.981985–20020.210.17, 0.60      0.20−0.20, 0.60
 Prostate5.117.61968–1971−4.71−16.11, 8.241971–19876.76*5.61, 7.911987–1990−4.84−26.24, 22.77 1990–20025.64*4.06, 7.245.60*4.10, 7.20
Women
 All sites211.4203.21968–1971−4.12*−6.86, −1.291971–19851.28*0.97, 1.591985–2000−0.01−0.29, 0.26 2000–2002−3.80−9.23, 1.96 −0.90−2.10, 0.40
 Stomach71.825.71968–1971−5.52*−9.81, −1.021971–1985−1.71*2.19, −1.231985–2002−3.14*−3.45, −2.82    −3.10*−3.40, −2.80
 Colorectal12.828.01968–1972−0.11−3.59, 3.511972–19796.00*4.01, 8.041979–19933.76*3.15, 4.381993–2002−1.12*−2.13, −0.10 −1.10*−2.10, −0.20
 Liver9.115.01968–1977−0.64−1.75, 0.481977–19885.49*4.48, 6.511988–20000.94*0.11, 1.782000–2002−10.05−20.48, 1.75 −1.60−4.20, 1.10
 Lung9.021.11968–19844.21*3.67, 4.751984–20021.11*0.67, 1.55      1.10*0.70, 1.50
 Breast16.840.61968–19854.13*3.48, 4.791985–20022.27*1.63, 2.91      2.30*1.70, 2.90
 Uterus44.213.01968–1971−7.40*−13.51, −0.871971–1980−1.57*3.03, −0.10 1980–1992−5.03*−5.90, −4.15 1992–2002−1.81*−2.85, −0.76 −1.80*−2.80, −0.80
Figure 1.

 Trends in age-standardized incidence and mortality rate for all sites of cancer in Osaka, Japan, 1968–2006.

Figure 2.

 Trends in age-standardized incidence and mortality rate by site and sex in Osaka, Japan, 1968–2006.

The overall cancer mortality for men increased until 1985, and leveled-off between 1985 and 1998, then decreased by-2.06% per year since 1998. Similar trends were observed for the overall incidence. For women, the trends of incidence increased by 1.28% per year between 1971 and 1985 (P < 0.05), then leveled-off from 1998. The overall mortality rates for women decreased gradually all through the observation period.

Trends of individual cancer sites.

Stomach cancer.  For both sexes, the trends of incidence and mortality rates decreased. Compared with the 1970s and 1980s, the difference between incidence rate and mortality rate widened after the 1990s. But during the latest period, the trends of incidence and mortality rates paralleled.

Colorectal cancer.  The trends of incidence rate for both sexes were similar. Incidence rates for both sexes increased significantly until 1993, and gradually decreased since 1993. The mortality rates have not decreased yet, though.

Liver cancer.  The incidence and mortality rates for both sexes decreased from the middle to end of the 1990s. The differences between the incidence and mortality rate were minimal.

Lung cancer.  Until the middle to end of the 1980s, both incidence and mortality rates increased. Both incidence and mortality rates for men and the mortality rate only for women leveled-off since the middle to end of the 1980s. But the incidence for women was still increasing.

Breast cancer.  Both incidence and mortality rates had increased significantly. The difference between incidence and mortality rates became a little wider between the 1970s and 1980s. The trends in the difference paralleled since 1985.

Uterus cancer.  The latest joinpoint was the year of 1992 for incidence and the year of 1993 for mortality rates. Before the joinpoints for both incidence and mortality rate, those rates were rapidly decreased (APC: -5.03 for incidence, -5.24 for mortality). After the joinpoints degree of the decreases was getting smaller (APC: −1.81 for incidence, −1.01 for mortality).

Prostate cancer.  Both incidence and mortality rates increased. The incidence increased remarkably from 1990.

Contribution of individual cancer sites to the change in overall cancer mortality rates. Table 3 shows the contribution of individual cancer sites to the total changes in overall cancer mortality rates. Overall cancer mortality rates peaked in 1998 for men and did not show a peak for women. Therefore, we calculated the contribution using the change between 1998 and 2006 both for men and women. Overall cancer mortality rates decreased by 14.7% for men and by 10.8% for women between 1998 and 2006. Stomach and liver cancer contributed by more than 25% to the total decrease in mortality; the two sites combined contributed 73% for the total decrease in men, and 53% in women.

Table 3.   Contribution of individual cancer sites to the change in overall cancer mortality rates in Osaka, Japan
SiteAge-standardized mortality (per 100 000)ChangeContribution (%)
19982006AbsoluteRelative
Men
All255.0217.5−37.5−14.7 
Decreasing
 Liver50.032.0−17.9−35.947.0
 Stomach46.036.0−10.1−21.926.3
 Lung55.551.1−4.4−7.911.5
 Lymphoma7.56.4−1.1−14.72.9
 Gallbladder7.86.8−1.0−12.92.6
 Leukemia5.34.6−0.7−12.81.8
 Oral cavity and pharynx5.24.7−0.5−9.81.3
 Esophagus11.511.2−0.3−2.80.8
 Colon and rectum24.524.4−0.1−0.50.3
 Bladder3.93.90.0−0.40.0
 Others18.216.1−2.1−11.45.4
 Total of decreasing sites  −38.2 100.0
Increasing
 Pancreas12.112.50.43.455.5
 Prostate7.37.70.34.444.5
 Total of increasing sites  0.7 100.0
Women
All119.6106.7−12.9−10.8 
Decreasing
 Stomach17.413.4−4.1−23.327.5
 Liver 14.310.6−3.7−26.025.1
 Lung17.014.9−2.1−12.514.4
 Colon and rectum14.513.0−1.5−10.310.1
 Gallbladder6.45.2−1.3−19.78.6
 Uterus6.25.5−0.8−12.25.2
 Leukemia3.02.8−0.2−7.61.5
 Bladder1.00.9−0.1−10.60.7
 Oral cavity and pharynx1.41.3−0.1−7.00.7
 Lymphoma4.14.0−0.08−1.90.5
 Others13.913.1−0.8−6.15.7
 Total of decreasing sites  −14.8 100.0
Increasing
 Breast12.012.90.97.447.6
 Pancreas6.87.50.811.240.4
 Esophagus1.51.70.214.812.0
 Total of increasing sites  1.9 100.0

Contribution of the incidence reduction to the mortality reduction.  The contributions of the incidence reduction in stomach and liver cancer to the reduction in their respective mortality are shown in Table 4. The relative change for mortality was calculated between 1998 and 2006. Median survival time was 5 years for stomach cancer and 1 year for liver cancer. Then the relative change for incidence was calculated between 1993 and 2001 for stomach cancer, and between 1997 and 2004 for liver cancer (2004 is the latest year of incidence at the moment). Nearly 100% of the decrease in liver cancer mortality could be attributed to the decrease in incidence for both sexes. For stomach cancer, more than 80% of the decrease in mortality was attributed to the decrease in incidence.

Table 4.   Contribution of the incidence reduction at major cancer sites to the total reduction in mortality in Osaka, Japan
 Age-standardized mortality rate (per 100 000)Relative change (%)Age-standardized incidence rate (per 100 000)Relative change (%)Explained by incidence reduction (%)
1998200619932001
  1. †The latest incidence data was 2004, thus we could not use the 2005 data here. ‡The figure was substituted as 100%, because the % was more than 100 (109.7 for men and 119.1 for women).

Stomach
 Men46.036.0−21.992.074.9−18.584.7
 Women17.413.4−23.335.829.0−19.081.6
Liver19982006 19972004†  
 Men50.032.0−35.959.936.3−39.4100.0‡
 Women14.310.6−26.017.812.3−31.0100.0‡

Discussion

The decrease in the overall cancer mortality in Osaka since the end of the 1990s is largely attributable to the reduction in the mortality rates of stomach and liver cancer for both sexes. The decrease in mortality from the two cancer sites could be explained mostly by the reduction in each incidence (more than 80% for stomach cancer, 100% for liver cancer). The prevalence of risk factors for stomach cancer, such as high salt intake, low fresh vegetable consumption,(5) and infection of Helicobacter pylori,(6,7) decreased in Japan in the past several decades, which reduced the incidence in part. A possible explanation for the remaining 20% of the decrease could be early diagnosis due to a screening program and improvement of treatments. Previous study about trends of cancer survival in Osaka found that approximately 60% of improved survival could be explained by the effect of early diagnosis of stomach cancer.(8) Hepatitis C virus (HCV) is known as the main risk factor for liver cancer in Japan. A decrease in HCV-related hepatocellular carcinoma incidence has been confirmed in Osaka.(9) Both stomach and liver cancer mortality were reduced by the decrease in the prevalence of risk factors.

Both incidence and mortality of breast cancer were increased in Osaka during the study period. These trends seemed due to changes in lifestyle that were related to risk factors(10–14) and the delayed introduction of an effective screening program for breast cancer among Japanese. In parts of Europe and North America, mortality of breast cancer started decreasing in the 1990s mainly due to the effect of systematic mammography screening.(15–17) In Osaka, some municipalities started the mammography screening from around 2000, but the screening is not so effective. The proportion who were participated in the screening program was 14.9% in Osaka, which was lower than that in the whole of Japan (20.9%)(18) and much lower than that in Finland (87% in 2005, ages 50–69 years),(19) the UK (73.8% in 2006/2007, ages 50–70 years),(20) and the USA (76%).(21)

A slight decrease in lung cancer mortality for men could be explained by the decrease in its incidence among the birth cohort born in the late 1930s, which may be partly explained by limited supply of tobacco after World War II.(22–24) Possibly, this decreasing trend is temporal because the tobacco supply increased as the economy recovered later. Among women, lung cancer incidence in Osaka is slightly increasing as in European countries(25) and the USA.(4,26) In particular, the prevalence of smoking and lung cancer mortality in Osaka has been higher than in other prefectures in Japan.(27) We need to monitor the trends carefully and continue promoting tobacco control.

Trends in colorectal cancer incidence have leveled-off since 1993 as have those for mortality since 1996. There was a slight decrease in incidence for women and in mortality for both sexes, but the APC was small. The difference between incidence and mortality for colorectal cancer was wider from the 1970s to the mid-1990s for both sexes. These trends would be due to improvements both in treatment and early diagnosis. But the magnitude of impact from different contributing factors, such as the change in risk prevalence and improvement in screening and treatments, is still unclear. We need more detailed study of the trends in survival, age, and stage-specific incidence and mortality, as reported from the USA.(28)

The incidence and mortality of uterus cancer consistently decreased during the study period. But the trend in mortality leveled-off from 1993 and the APC of incidence decreased from 1992 and both trends became parallel. This might be due to the effect of the Papanicolau screening test in reducing the risk of invasive cancer incidence and mortality. A previous study reported that the incidence of carcinoma in situ among young people (under the age of 30 years) has increased in recent years and the incidence of invasive cancer decreased in Japan, due to screening of pregnant women.(29) But in Osaka, the coverage of cervical cancer screening was lower than the coverage of Japan as a whole (18.3% in Osaka, 21.3% in Japan).(18) So, we can expect a further reduction in incidence and mortality if the population-based cancer screening is more effectively conducted.

As in other countries in Europe and North America, prostate cancer incidence increased partly due to the overdetection of the cancer from the wide use of the prostate-specific antigen (PSA) test. The mortality of prostate cancer increased too, which could be also explained, at least in part, by wider use of the PSA test for otherwise undiagnosed patients. The effect of population-based PSA screening is still controversial.(30,31)

Limitations

The population of Osaka was large enough to have stable trends of statistics annually. But the characteristics of the cancer statistics in Osaka might be different from those of Japan overall. For example, mortalities from liver and lung cancers in Osaka were higher than those of Japan,(32) and the survival from some cancers was lower in Osaka than in other prefectures.(33) Therefore we may not be able to generalize the results of the present study to Japan as a whole. To evaluate cancer control in Japan, we strongly hope that the Japanese government will establish a cancer surveillance system in the near future.

We should keep in mind the changes in the completeness of cancer registration in Osaka when we evaluate the incidence trends. The percentage of cases registered by death certificate only, which is often regarded as an index for the completeness, was approximately 10–15% and stable in the Osaka Cancer Registry during the most recent two decades.(34,35) Therefore, we considered that the influence of the change in completeness on the trends was small.

In the method used to estimate the contribution of the incidence reduction to the mortality reduction, we assumed the median survival time as the time lag between incidence and death. Under this assumption, the contribution of the incidence reduction to the mortality reduction (%) for liver was more than 100%, which was unrealistic.

We only analyzed the age-standardized rate for all ages. The age of cancer patients in general is rapidly getting higher, but the effectiveness of some cancer control activities, such as primary prevention, are unclear among the older population. Future study will be needed to identify the trends in incidence and mortality by age group, because some cancer control programs, such as screening, are targeted at specific age groups.

Conclusion

The results of our quantitative analyses suggest the possibility that the decreased overall cancer mortality in Osaka during the study period was mainly due to natural decreases in the incidence of stomach and liver cancer, which were attributable to the decrease in risk factors, rather than to the effect of secondary prevention or improved treatment. For the reduction of mortality in other major cancers, we need improved cancer control programs.

Acknowledgments

Yuri Ito 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. This study was supported by a Grant-in-Aid for Cancer Research from the Japanese Ministry of Health, Labour and Welfare (20-2).

Ancillary