Incidence and mortality from stomach cancer in Japan, Slovenia and the USA

Authors


Abstract

The mortality and incidence from stomach cancer were compared in Japan (a country with a high incidence where there was full application of mass screening during this period) and 2 countries with no screening policy: the USA (with a very low incidence) and Slovenia (with an intermediate rate). The registered cases of stomach cancer were from the Osaka Cancer Registry, the Slovenian National Cancer Registry and the Surveillance, Epidemiology, and End Results (SEER) registries in the USA. In the period 1975–95, the age-adjusted incidence rate (/100,000) of stomach cancer declined in the 3 countries, as follows: Japan, from 76.0 to 53.0 in men and 38.4 to 21.3 in women; Slovenia, from 40.2 to 24.1 in men and 16.6 to 10.8 in women; and the USA, from 9.5 to 6.9 in men and 4.3 to 2.9 in women. During the same period, the age-adjusted mortality rate declined, as follows: Japan, from 60.2 to 34.2 in men and 30.5 to 14.1 in women; Slovenia, from 37.7 to 21.2 in men and 13.8 to 9.0 in women; the USA, from 5.6 to 4.7 in men and 2.5 to 2.3 in women. In the period studied, specific trends on incidence and mortality with a cohort effect occurred only in Japan: analysis by the age period-cohort model confirmed that the decline has occurred since the generations born in 1910. The trend therefore corresponds to unplanned prevention through changes in environmental factors occurring since the early 20th century. The study of stage-specific incidence rates confirmed the declining trend for regional cancer, whereas there was an increase in the incidence of localized cancer, associated with a period effect in 1975–95. This is attributed to the policy of early detection of stomach cancer, with the inclusion of intramucosal lesions of favorable prognosis and explains why mortality decreased faster than incidence during the period. © 2001 Wiley-Liss, Inc.

In spite of declining incidence rates worldwide, stomach cancer is the second most important cause of death from cancer, and the actual numbers of cases diagnosed will increase each year, because of the aging world population. In Japan, based on data from 8 population-based cancer registries, stomach cancer accounted for 31% of all new cancer cases in males and 22% in females in 1985–89;1 the stomach remains the leading tumor site.2, 3 Japan is the only country where mass screening for stomach cancer has been a national policy (encouraged since 1966) and has been included in the Health Services Law for the Aged since 1983. It is believed that early detection (either early or silent disease) may have had an impact on mortality rates in Japan.4–10 In the Western world, in the absence of a screening policy, most cases of stomach cancer are detected at an advanced stage, with a correspondingly poor prognosis: indeed, a recent analysis of 57,407 cases in selected series in the USA (National Cancer Data Base11) shows that most cases are still detected at an advanced stage, with a large proportion in stage IV.

The etiology of stomach cancer is linked to environmental factors, including Helicobactor pylori-induced inflammation and atrophy of the gastric mucosa and a diet rich in salt and nitrites and poor in fruit and vegetables. Migrant studies confirm the importance of lifestyle factors in etiology and also suggest the importance of “environment” in early life (childhood) in determining the risk for stomach cancer. Studies of migrants from Japan (high risk) to the USA (low risk) show a decline in incidence post migration. Rates are even lower in Japanese born in the USA (Nisei), but are still higher than in US-born whites;12, 13 similar observations have been made in Hawaii.14 The persistently higher incidence is generally attributed to some retained elements of Japanese lifestyle, notably with respect to diet.

The almost universal decline in risk of stomach cancer has been interpreted in terms of changes with respect to food preservation and storage, and to a reduction in the prevalence of H. pylori infection in childhood.15, 16 The decline in incidence and mortality from stomach cancer has also occurred in Japan, and the contribution of mass screening to the decrease in mortality is an important point of interest. A comparison with countries without a screening policy may help to distinguish the relative contributions of “unplanned prevention” and a mass screening policy. We have therefore examined trends in incidence and mortality for the period 1975–95 in 3 countries: Japan, where there was full application of mass screening during this period, and 2 countries with no screening policy, the USA, with a very low incidence rate, and Slovenia, with an intermediate rate. We have tried to determine the components of the trends, which can be attributed to birth cohort, period and age, as indicating the relative importance of lifestyle change, versus early detection and treatment. Here we present the results for the trends in incidence and mortality rates. Temporal trends in survival will be presented in a later article.

MATERIAL AND METHODS

Sources

Data on registered cases of stomach cancer (by stage of disease) and deaths were obtained from population-based registries in 3 countries: the Osaka Cancer Registry in Japan,17, 18 the Slovenian National Cancer Registry19 and the Surveillance, Epidemiology, and End Results (SEER) registry in the USA.20.

The Osaka Cancer Registry was established in 1962; it covers the population of Osaka City and the surrounding communities. The target population (around 8 millions) amounts to 7% of the Japanese population (121 million in the 1985 census). Incidence and mortality data were available for the period 1975–95.

The Cancer Registry of Slovenia was established in 1950 and covers the population of the entire country (just under 2 million). Incidence and mortality data for the period 1975–95 were included in our study. Mortality data were missing in the files for the year 1981, and the missing data were estimated as the average values for 1980 and 1982.

The SEER Program covers 10% of the US population (therefore over 20 million). Incidence data were available for the period 1975–95 and mortality data only for the period 1978–92.

Cases

Cases were grouped by sex into 5-year age groups at diagnosis, between the limits of 20 and 85 years. Crude and age-specific incidence and mortality rates were calculated using midperiod population denominators for each age group. Age-adjusted incidence and mortality rates were calculated by direct standardization, using the world standard population. Stomach cancers occurring as second primary tumors were excluded from the calculation of incidence.

Stages of stomach cancer

In each data base, the cancer cases were classified in 3 stages: localized, regional and distant, with unstaged tumors being grouped under unknown.

Localized cancer is limited to the gastric wall, and lymph nodes are negative. In Western classifications, invasion of the lamina propria is considered the hallmark of malignancy; earlier lesions are considered to be either dysplasia or at an in situ stage. In Japan, however, intramucosal tumors are classified together with localized cancers, even in the absence of invasion of the lamina propria.21 This divergence produces a large discrepancy in prognosis (survival) and is reflected in the ratio of incident cases to cancer deaths. The recent Vienna consensus classification of digestive tumors may reduce this bias in the future.22

Regional cancer is a tumor limited to the gastric wall with regional lymph node involvement, or a tumor with extension to nearby viscera.

Distant cancer is a tumor with distant extension or metastases, either in the viscera or distant lymph nodes.

As well as stage-specific incidence rates (age-adjusted), the proportions of tumors registered in the localized stage has been determined in each of 4 successive 5-year periods covering the interval from 1976 to 1995 (unstaged tumors having been previously excluded from the analysis).

Time trends in incidence and mortality

An estimate of the linear trend in the age-adjusted stomach cancer incidence and mortality rates and the stage-specific incidence rates was obtained by fitting Poisson regression models to the data observed over the period 1975–95, adjusting for the effects of age.

The periods 1975–79, 1980–84, 1985–89 and 1990–84 were constructed, and a set of 10-year approximate birth cohorts was estimated by subtracting the midpoint of the 5-year age group from the corresponding 5-year period. A formal statistical examination of whether trends were better described by secular changes in risk (period effects) or changes in risk from generation to generation (birth cohort effects) was then made by fitting the standard age period-cohort model.23 The models fitted included terms for age, drift (the rate of change of regular trend, which cannot be attributed exclusively to either period and cohort) and the nonlinear effects of period and of cohort. Hierarchic fitting strategies were employed to determine the best model.23 To examine the effects of period and cohort as measures of relative risk, the groups with midpoints 1882 and 1914 were used as baselines, respectively.

RESULTS

The age-adjusted incidence rates (world population) per 100,000 in men and women are shown in Figure 1. There is a decline in incidence in all 3 countries. Japan (Osaka Registry) has the highest incidence rate; it decreases from 76.0 in 1975 to 53.0 in 1995 for men and 38.4 to 21.3 for women. Slovenia, with a smaller population, shows more fluctuations in incidence, and the rate decreases from 40.2 to 24.1 for men and 16.6 to 10.8 for women. The lowest rates occur in the USA, with a constant decline from 9.5 in 1975 to 6.9 in 1995 for men and from 4.3 to 2.9 for women.

Figure 1.

Time trends in age-adjusted (world population) incidence and mortality rates per 100,000, in 1975–95 in Japan (Osaka Cancer Registry), Slovenia (Slovenia Cancer Registry) and the USA (SEER Registries). Note: in the USA, the mortality rates are for the period 1978–92.

The age-adjusted mortality rates decreased in Japan from 60.2 in 1975 to 34.2 in 1995 for men and from 30.5 to14.1 in women. In Slovenia, the rates decreased from 37.7 in 1975 to 21.2 in 1995 for men and from 13.8 to 9.0 for women. In the USA, the rates decreased from 5.6 in 1978 to 4.7 in 1992 for men and from 2.5 to 2.3 for women.

The average mortality/incidence (M/I) ratio was estimated for a 3-year period as the ratio of age-adjusted mortality rates to incidence rates, expressed as a percentage. The ratio decreased in Japan from 76% in 1975–77 to 60% in 1993–95 for men and from 76% to 63% for women. In Slovenia the average M/I ratio in a 3 year period remained stable and high during the study period: 90% in 1974–76 and 88% in 1993–95 for men; the respective values in women were 85% and 83%. In the USA the average M/I ratio in a 3-year period increased slightly during the study period: 57% in 1978–80 and 64% in 1990–92 in men; the respective values in women were 58% and 66%.

Stage-specific incidence

The age-adjusted incidence rates per 100,000 have been estimated for localized, regional and distant stages, and for the tumors registered as unstaged, in the 3 countries.

In Japan, most of the decline in incidence is accounted for by a declining proportion of unstaged tumors at registration: the age-adjusted incidence rate for unstaged tumors decreased by 57%, from 38.4 in 1975 to 16.3 in 1995 in men and from 20.2 to 6.3 in women. The rates for regional and for distant stomach cancer were stable. On the other hand, there was a significant increase in the rate of localized stomach cancer: from 8.1 in 1975 to 16.8 in 1995 for men and from 4.1 to 6.7 for women. As shown in Figure 2, the increasing incidence of localized cancers stabilizes toward the end of the 1980s in both sexes.

Figure 2.

Time trends in age-adjusted (world population) and stage-specific incidence per 100,000, in 1975–95 in Japan (0saka Cancer Registry), in men and women. Stage categories are localized, regional and distant. The adjusted incidence of unstaged tumors is also represented.

In Slovenia (Fig. 3), the rates decline for all stages (including unstaged), but with fluctuations, due to the small number of cases.

Figure 3.

Time trends in age-adjusted (world population) and stage-specific incidence per 100,000, in 1975–95 in Slovenia (Slovenia Cancer Registry), in men and women. Stage categories are localized, regional and distant. The adjusted incidence of unstaged tumors is also represented.

In the USA (Fig. 4), the incidence of unstaged tumors was stable and low. Most of the decline is attributable to regional and distant cancers, and the decline was more marked for women. Incidence of localized cancer remained stable.

Figure 4.

Time trends in age-adjusted (world population) and stage-specific incidence per 100,000, in 1975–95 in the USA (SEER Registries), in men and women. Stage categories are localized, regional and distant. The adjusted incidence of unstaged tumors is also represented.

Among tumors with stage recorded, the proportion of localized tumors has been determined during the period 1976–95 in 5-year intervals (Table I). In Japan, the proportion increased in each of the 4 successive intervals. In contrast the proportion decreased in Slovenia and increased only slightly in the USA. When comparing the 1st and the 4th 5-year intervals, the proportion almost doubled in Japan, decreased by 30% in Slovenia and increased by 19% in men and 26% in women, in the USA

Table I. Proportion (%) of Localized Stomach Cancer, in Japan, Slovenia and USA1
YearsOsakaSloveniaSEER
MaleFemaleMaleFemaleMaleFemale
  • 1

    At successive 5-year intervals during the period 1976–95, estimated from the age-standardized, stage-specific incidence rates (unstaged tumors excluded).

1976–8024.724.633.832.018.719.6
1981–8530.730.030.026.416.921.4
1986–9039.638.624.425.719.323.4
1991–9544.042.823.124.922.126.7

Regression analyses on the age-adjusted rates

Estimates of the annual linear trend (and standard error) of the age-adjusted incidence and mortality rates are shown in Table II. The negative slopes confirm the downward trends seen in Figures 2, 3 and 4 in all 3 countries. In Japanese males, the regression slope is more marked for mortality (−1.32) than for incidence (−1.12), perhaps an indication of improving prognosis during this period. In Slovenia, the mortality and incidence rates have identical slopes, suggesting stability of survival, whereas in the USA, the decline in incidence is greater than for mortality.

Table II. Annual Mean Change in Age-Standardized Incidence and Mortality1
 OsakaSloveniaSEER
MaleFemaleMaleFemaleMaleFemale
  • 1

    Estimates of slope based on linear regression model.

Incidence/−1.12−0.86−0.86−0.37−0.12−0.07
 100,000, total(±0.07)(±0.03)(±0.07)(±0.03)(±0.01)(±0.005)
Mortality/−1.32−0.81−0.84−0.34−0.03−0.02
 100,000, rate(±0.04)(±0.03)(±0.05)(±0.03)(±0.01)(±0.01)

Estimates of the stage-specific incidence rates are shown in Table III. In Japan, the positive slope of localized cancer (0.53 per year in men and 0.15 per year in women) confirms the increasing incidence rates. This in contrast to the negative slopes for regional cancer and unknown stages, the declining trend being less for distant cancer.

Table III. Stage-Specific Annual Mean Change in Age-Standardized Incidence1
 OsakaSloveniaSEER
MaleFemaleMaleFemaleMaleFemale
  • 1

    Estimates of slope based on linear regression model.

Incidence/100,000+0.53+0.15−0.36−0.12−0.003−0.00
 Localized(±0.06)(±0.03)(±0.02)(±0.02)(±0.005)(±0.002)
 Regional−0.53−0.32−0.07−0.03−0.06−0.03
(±0.05)(±0.03)(±0.03)(±0.01)(±0.005)(±0.003)
 Distant−0.06−0.06−0.27−0.15−0.05−0.03
(±0.05)(±0.05)(±0.04)(±0.02)(±0.01)(±0.003)
Incidence−1.06−0.63−0.20−0.11−0.004−0.007
 unknown(±0.11)(±0.05)(±0.03)(±0.01)(±0.004)(±0.002)

Table IV shows the overall percentage change in incidence in 10-year age groups and for all ages (age standardized) over the 2 decades. The magnitude of the decline is similar in each age group, from 40–49 to 70–79.

Table IV. Percentage Change in the Incidence of Stomach Cancer Over the Period 1975–95 in Japan, Slovenia and USA1
Age (yr)OsakaSloveniaSEER
MaleFemaleMaleFemaleMaleFemale
  • 1

    Age-specific change and age-standardized (all ages) change. +, increased value; −, decreased value.

40–49−32.6−46.6−41.2−34.1−26.6−34.8
50–59−32.9−47.6−38.6−33.0−27.2−34.4
60–69−33.4−49.8−38.6−30.2−27.5−32.3
70–79−31.4−41.8−33.9−36.6−25.8−39.5
80+−2.1−23.8−24.9−33.3−40.6−29.6
All ages−30.2−44.4−40.1−34.9−27.3−33.2

The corresponding results for localized stomach cancer are shown in Table V. In Japan, the increase in localized cancer contrasts with the overall declining incidence of the disease. The increase is more prominent among the younger age groups. In contrast, there are declines in the incidence rates in Slovenia in both sexes, and in all age groups. In the USA, age-specific trends in incidence rates have been fairly stable in men, in the last 20 years, whereas some downward trends have been detected in women.

Table V. Percentage Change in the Incidence of Localized Stomach Cancer Over the Period 1975–95 in Japan, Slovenia and USA1
Age (yr)OsakaSloveniaSEER
MaleFemaleMaleFemaleMaleFemale
  • 1

    Age-specific change and age-standardized (all ages) change. +, increased value; −, decreased value. For Slovenia the number of cases in the age group 80+ is very small, and the percentage of change must be considered with caution.

40–49+110.6+70.6−63.2−49.0−0.8−10.2
50–59+106.5+58.7−59.6−53.1−3.0−13.6
60–69+96.1+40.4−59.7−44.0−1.5−4.2
70–79+84.4+53.4−66.4−55.3−4.5−16.0
80++43.4−20.9−67.1+67.1−8.2+26.4
All ages+109−61.8−64.1+44.8+0.7−2.1

Modeling

The best fitting model for each country, after adding terms representing age, drift, period and birth cohort to the overall stomach cancer mortality and incidence rates, as well as to the incidence rates of localized stomach cancer, are shown in Table VI. Most of the changes in the age-adjusted incidence and mortality rates in Slovenia and the USA can be attributed to nonidentifiable drift or to nonlinear period effects. In Japan, there are significant nonlinear cohort-specific trends in incidence and mortality for all stomach cancers combined, although such an effect was not observed in the localized incidence and mortality rates, for which significant nonlinear period effects are reported. Figure 5 shows the relative risk of death from stomach cancer in Japan in successive cohorts (with reference to the cohort 1914). The earlier cohorts (1895–1905) had a relatively high risk, peaking in 1900 for males (RR = 1.4) and females (RR =1.2). In cohorts born after 1914, the risk decreased in successive generations, reaching 0.2 by 1950 in both sexes. For incidence, the highest relative risk was also for the 1900 cohort in both males and females. The risk declined after 1914 to reach about 0.4 in males and 0.25 in females in cohorts born about 1950. The comparison of successive cohorts by sex indicates that the decline occurred earlier in women than in men and that the decline in mortality rates occurred 5 to 10 years earlier than that of incidence. For example, a relative risk of 0.6 was established in men for cohorts born in 1925 (mortality) and 1935 (incidence); the respective dates in women were approximately 1920 and 1925.

Table VI. The Best Fitting Models to Explain Changes in Mortality Total Incidence and Incidence of Localized Cancer in Japan, Slovenia and USA (SEER)1
 SexOsakaSloveniaSEER
  • 1

    A, age effects; P, period; C, birth cohort; D, drift; df, degrees of freedom; dev, deviance.

MortalityMaleA+C dev: 8.44, df: 26 p < 0.001A+P dev: 15.36, df: 36 p < 0.01A+P dev: 2.37, df: 26 p < 0.001
FemaleA+C dev: 7.56, df: 26 p < 0.001A+P dev: 8.78, df: 36 p < 0.2A+P dev: 1.09, df: 26 p < 0.001
Incidence TotalMaleA+C dev: 13.49, df: 26 p < 0.001A+P dev: 30.61, df: 36 p < 0.01A+D dev: 4.75, df: 41 p < 0.001
FemaleA+C dev: 7.56, df: 26 p < 0.001A+P dev: 27.54, df: 36 p < 0.02A+D dev: 2.27, df: 41 p < 0.001
LocalizedMaleA+P dev: 12.44, df: 39 p < 0.02A+D dev: 22.91, df: 38 p < 0.001A dev: 4.35, df: 42
FemaleA+P dev: 10.49, df: 39 p < 0.02A+P dev: 10.87, df: 36 p < 0.1A dev: 1.92, df: 42
Figure 5.

Cohorts effects for incidence and mortality in Japan (Osaka Cancer Registry). The relative risk for each cohort is expressed in relation to the risk in the cohort 1910.

DISCUSSION

Temporal trends in the 3 countries during the 1975–95 period

One objective of examining these 3 data sets is to investigate the effects of the relative contributions of unplanned prevention and early detection through screening have been on mortality, by studying time trends in both incidence and mortality in Japan (a country with a high incidence and a screening policy) and in 2 other countries with no screening program (the USA, with a low incidence, and Slovenia, with an intermediate incidence). Overall, there is a decline in incidence in all 3 countries, over the whole period in all age groups. The linear drift of the regression models implies that the observed trends are uniform and that the contributions of birth cohort effects and the period-specific risk cannot be distinguished; they also imply that the existing trends may be continued in the future. The decline in incidence is greater in Japan than in the other 2 countries and concerns all age groups, except for those over 80 years. Japan is the only country where a significant cohort effect was observed: the highest risk occurred for men and women born around 1900 and then decreased in successive birth cohorts throughout the 20th century. The birth cohort-specific decline in incidence (which is reflected in mortality) is evident in Japan because, unlike the other countries, there was an increase in risk in cohorts born before 1900. It suggests that the declines in incidence were the result of lifestyle change (probably dietary). On the other hand, the changes in localized disease are clearly due to screening and are discernible as a period effect.

In our study, temporal trends of age-adjusted incidence by stage suggest that, despite the increasing role of endoscopic exploration, no progress in the detection of cases at earlier stages has occurred in Slovenia and the USA: the ratio of the tumors staged as localized versus advanced (regional or distant) was stable during the period. In Japan the incidence of regional cancer has declined while the rate for distant cancer has remained stable, and the proportion of localized tumors almost doubled between 1976–80 and 1991–95 (Table III). These trends suggest that diagnostic methods have changed over the period. Two factors have to be considered in interpreting the results:

  • 1The high proportion of unstaged tumors in the data from Osaka in the early years. In the first 5-year interval of the study (1976–80) the age-adjusted incidence of the unstaged tumors for men was 42.6% of the total rate for men and decreased to 29.6% in the last interval (1991–95). The proportions in women were 42.8% and 28.8%, respectively. The distribution of cases of known stage must be interpreted accordingly.
  • 2In Japan, the localized cancer group included intramucosal tumors and therefore, according to the Japanese classification, cases with no invasion of the lamina propria. Such cases would not be included as gastric cancers in Western countries. This bias may interfere with the proportion of localized cancer, and of course with survival.

However, the extent of the change in stage distribution in Japan is too large to be explained by variations in proportions of unstaged tumors. The increasing incidence of localized cancer occurs in all age groups, up to those aged 70–79 years. The increasing incidence of localized cancer is period-specific, suggesting that it relates to a specific intervention, presumably the Japanese mass screening program.

A decline in mortality is evident in all 3 countries, except for men aged 80 years or more in Japan. Declines in mortality are period-specific in Slovenia and the USA, whereas in Japan (as for incidence) the decline involves successive birth cohorts since 1900. The higher survival from stomach cancer reported in Japan compared with the USA has been a matter of debate. A low M/I ratio is considered an indication of good survival; however, the M/I ratio is in the same range for both countries. Factors such as the completeness of case registration may explain the discrepancy. In Japan, the decline in mortality is greater than for incidence; the persisting trend throughout the period suggests that this is not an artifact and that survival from stomach cancer is improving

Respective roles of lifestyle and screening

The generalized decline in incidence of stomach cancer, evident in most developed countries (“unplanned prevention”), relates to changes in lifestyle, particularly with respect to diet and food storage.15 Such influences have certainly been important in the decline in risk in Japan.24, 25 The role of H. pylori in increasing the risk of gastric cancer through the promotion of chronic atrophic gastritis is also clearly established.26, 27 The prevalence of H. pylori infection in Japan has been declining in successive birth cohorts.28–30 A recent study suggests that the declines in gastric cancer incidence in young subjects in Japan involves the intestinal (rather than the diffuse) type of carcinomas,31 which would be consistent with the greater risk for this subtype conferred by H. pylori infection.26, 32 These changes are often attributed to the generalization of a western style of life in Japan just after World War II in Japan, under American influence. Most studies focus on changing environmental factors during this period. Actually, although the full impact of the declining trend occurred in the post-World War II period, the causes were active at least 30 years earlier. Helicobacter pylori infection is just one factor among others in gastric carcinogenesis, and etiologic factors should be examined in perspective since the early 20th century. Japan entered the industrial arena and behaved as a world power much earlier, under the influence of Emperor Meiji, in the period just after the Sino-Japanese war (1894–95) and the Russo-Japanese war (1904–05), ending with the Port Arthur surrender to the Japanese. Changing environmental factors occurring in the infancy of persons born in the cohort 1910 and after must play a major role.

The period studied (1975–95) coincides with the full-scale introduction of a mass screening program for gastric cancer in Japan, organized either in municipalities or through private companies.4–10, 33–41 The results of the screening examinations are given in the annual reports on mass screening for digestive cancer throughout the country The latest report, for 1997,34 confirms that of 5,151 stomach cancers detected by mass screening, superficial tumors staged T1 (m or m+sm) represent 67.8%, whereas advanced cancer accounts for only 32% of cases. The proportion of intramucosal cancers (m) was 40% (2058 cases). However, even in Japan, the proportion of the population examined in this organized program is not so large; in 1983 the coverage of the screening was 9.6% of the target population.5 According to the report of the Ministry of Health and Welfare of Japan, the proportion covered by the program of the health services mandated by law was 13.3% of the target population in 1998. As a further illustration, it may recalled that the estimated number of incident cases of stomach cancer in Japan in 2000 is 115,294 according to the IARC data base Globocan 2000.42 However, in addition to the organized screening program, there have been an increasing number of opportunistic tests, carried out in various settings. According to the survey of the Osaka Prefecture in 1997, the proportion of those people who had stomach cancer screening tests and/or stomach x-rays during the previous year was 39.4% (A. Oshima, personal communication) Extensive early detection explains why the proportion of localized cancer doubled during the study period in the Osaka Registry, whereas it was stable in the USA. On the other hand, the decline in the proportion of localized cancer in Slovenia during the period is largely attributable to the progression in staging procedures and the higher resectability rate in the country since 1982.43

Among the cases detected in Japan at the stage of intramucosal neoplasia, many would be classified high-grade dysplasia in the Western world21 and not recorded as “cancer.” This is a major cause of the different prognosis of stomach cancer in Japan and in the West. The potential for progression from early to advanced cancer is uncertain, and early gastric cancer has even been called a pseudo-cancer. However, a study conducted in Japan confirms that lesions fulfilling the endoscopic criteria for early gastric cancer in Japan usually progress to advanced cancer, the 5-year risk for progression being 63%.39, 40 Thus, despite staging based on endoscopy alone, the malignant potential of intramucosal neoplasia can be accepted. The higher survival from stomach cancer in Japan has also been attributed to different strategies in surgical treatment, mainly the routine use of the D2 resection, which involves a radical extended regional lymphadenectomy. However, the improvement in results obtained with that operation in Japan have not been confirmed in Western countries.44, 45

The evidence for the efficacy of gastric cancer screening in Japan in preventing advanced disease as well as death from gastric cancer rests on the results of observational studies, so the contribution of selection bias to any favorable effect (low-risk individuals choosing to be screened) is hard to evaluate. Only one randomized trial has been reported,33 and, although no results were available at the time, it is highly unlikely that, given the small difference in screening experience between intervention and control groups, any benefit could emerge later. Three case control studies, of prevention of cancer deaths6, 35 or of advanced disease,36 all suggested a beneficial effect of screening, as did 2 prospective cohort studies comparing mortality rates in individuals participating or not in the mass screening program.37, 38 The study in Osaka37 showed a rather higher than expected incidence of gastric cancer in individuals undergoing screening, implying possible overdiagnosis (detection of nonprogressive “cancer”) in this group.10 On a population level, a previous study of incidence and mortality in Miyagi prefecture between 1960 and 198533 showed that, after the introduction of screening (around 1970), mortality fell more rapidly than incidence, particularly in the age groups targeted for mass screening (50–79 years). Similar data were observed in Osaka.39

CONCLUSIONS

The incidence of stomach cancer has declined in Japan (as shown by data from the Osaka Cancer Registry) rather more rapidly than in Slovenia and the USA during the period 1975–95, with, in addition, a decline in incidence of regional stage cancer and an increase in incidence of localized stage cancer. In addition, in Japan, a decline in mortality that is significantly greater than that of incidence has occurred. Fitting age period-cohort models to the trend in Japan shows that incidence and mortality trends are cohort-specific, whereas the incidence of localized cancer changes is period-specific.

The declining trends of incidence and mortality rates are not recent, and the analysis by the age period-cohort model confirms that they have been present since the generations born in 1910. The trend corresponds to “unplanned prevention” through changes in environmental factors and is responsible for the reduction in the incidence of stomach cancer in Japan and elsewhere.

The increased incidence of localized cancer, associated with a period effect in 1975–95, can be attributed to the general policy of early detection of stomach cancer, with the inclusion of intramucosal neoplastic lesions of favorable prognosis. This explains why mortality has decreased faster than incidence during the period.

Ancillary