The Research Group for Population-based Cancer Registration in Japan has been conducting a cooperative study to estimate cancer incidence in Japan since 1975. Estimated incidence data calculated annually were accumulated in 1996 for 18 years. The Group has separately provided another re-calculated incidence estimate series which were prepared for 1975–79, 1980–84 and 1985–89. The former two results, each of five years were previously published elsewhere. These recalculated estimate series made more reliable observations of time trends in incidence feasible by using the same eligible registries' data throughout each 5-year period. This report presents results of the latter 5 years between 1985 and 1989; age-specific, crude and age-standardized incidence rates, as well as the number of incidence according to site and gender, under the cooperation of eight eligible population-based cancer registries in Japan: Miyagi, Yamagata, Kanagawa, Fukui, Osaka, Tottori, Hiroshima City and Nagasaki City. Incidence in Japan was estimated at 187 200 and 150 700 for all cancer sites among males and among females, respectively, in 1985 and 216 700 and 166 900, respectively, in 1989. The leading site was the stomach among both males and females from 1985 to 1989. Among males the second leading site was the lung, followed by the liver, colon and rectum. Among females, it was the breast, followed by the colon, uterus and lung in 1989. The proportion of the cases registered by a death certificate only for all sites was 14.0–15.7 and 13.7–15.3% and the ratio of incidence to mortality was 1.69–1.73 and 1.95–2.01 among males and females, respectively, during the period.
Monitoring cancer incidence and cancer trends for the entire country is essential in order to provide and assess programs of cancer prevention and control in Japan.
The Research Group for Population-based Cancer Registration in Japan has been providing and publishing national estimates of cancer incidence and incidence rates since the Research Group was first organized in 1975 (1,2). A series of annual estimates were conducted according to site, gender and age-group based on qualified data from eligible registries whose reporting rate satisfied a set of standards. The results have been reported in this journal (3–6).
Generally, completeness of reporting was constant or improved in most registries. However, from time to time, some registries failed to maintain satisfactory reporting rates. On the other hand, other registries successfully improved their reporting rates to satisfactory standards. This variance caused a change in the member of eligible registries, resulting in inconsistencies being introduced into the annual trends of incidence estimates. To avoid this, it was suggested that estimates be calculated for at least 5 years, based on the data from the same eligible registries.
Apart from this, a considerable number of delayed cancer reports were subsequently submitted to all relevant registries from their reporting hospitals after each annual estimate was conducted. It was concluded that if a recalculation of incidence using the total cancer reports including those which were delayed was to be conducted in registries after a few years, estimates obtained using these data would more closely reflect true incidence.
For these two reasons, re-estimations were conducted for 1975–79 and 198–84 which provided another estimate series. The results have been presented elsewhere (2).
All member registries of the Research Group agreed to recalculate their annual cancer incidence from 1984 to 1990 by including all their delayed cancer reports in order to conduct a re-estimation of incidence in Japan for each year between 1985 and 1989.
Materials and Methods
Moving Average Process
According to the standard methods defined in the ‘Guidelines for Population-based Cancer Registries in Japan, Third Revision’ (7) published by the Research Group, all member registries recalculated their annual cancer incidence in 1993 according to site, gender and 5-year age-group between 1984 and 1990 and sent their data to the Secretariat of the Research Group, the Osaka Cancer Registry (OCR). Annual average incidence rates were computed at the OCR for each three years 1984–86, 1985–87, 1986–88, 1987–89 and 1988–90 by a moving average process according to site, gender and 5-year age-group for each registry region. The Research Group has used this moving average data to obtain a set of stable age-specific incidence rates for each cancer site, even from small population registries.
To select qualified data, a set of standards was set up by the Research Group: (1) the proportion of cases registered with death certificate only (DCO) ≤30% and (2) the ratio of cancer incidence to cancer deaths in the vital statistics ≥1.4, for all cancer sites in males and females combined (8).
The data from the following eight eligible registries, whose every 3-year moving average data satisfied the standards through the period, were used for the re-estimation of annual incidence in Japan between 1985 and 1989: Miyagi, Yamagata, Kanagawa, Fukui, Osaka, Tottori, Hiroshima City and Nagasaki City. The Kanagawa Cancer Registry alone presented data from a representative area with a population of one million in order to provide a satisfactory reporting rate.
The characteristics of the eight eligible registries are shown in Table 1. The population covered by the eight registries was 16241 600, corresponding to 13% of the total population of Japan in 1989. The proportion of DCO and the ratio of incidence to deaths which present completeness of reporting are shown together with the proportions of the histologically verified cases to incidence as well as those to reported cases for all cancer sites. The highest completeness of reporting was provided by the Fukui Registry and the highest accuracy of diagnosis was provided by the Hiroshima City Registry.
In comparing these registries with the eligible registries of the 198–84 re-estimation, the Chiba Cancer Registry was excluded because of a decrease in completeness of reporting. The Saga Cancer Registry failed to complete a calculation for incidence in 1990 before the closing date and was excluded. The registries of Kanagawa, Fukui and Tottori were newly included as they satisfied the expected reporting standards.
Arithmetic Mean of Incidence Rates from Eight Registries
In estimating cancer incidence in Japan, in order to avoid a bias caused by the size of the registry population, an arithmetic mean of incidence rates of all the eligible registries (by site, gender and 5-year age-group) was used, instead of dividing the total incidence by the total population of the eight registries. These rates were regarded as uncorrected estimates of incidence rates in Japan.
Correction Coefficients Based on Cancer Mortality in Japan
On the assumption that survival rates for cancer patients from all eligible registries do not differ much from those for cancer patients in Japan according to gender and site, the following correction was applied during the final stage of the estimate to minimize any possible bias caused by the selection of too few eligible registries.
Cancer mortality in Japan was estimated by using the exact methods employed for the estimation of incidence, by taking mortality data from the vital statistics (8) of the same eight eligible registries. Correction coefficients were then computed according to site and gender by using the observed cancer deaths in Japan in the vital statistics divided by the number of estimated deaths in Japan.
The estimated uncorrected cancer incidence according to site, gender and 5-year age-group were multiplied by the correction coefficients prepare d according to site and gender to obtain corrected estimates. These corrected figures were then multiplied by the corresponding Japanese population to obtain the incidence figures for Japan. Age-standardized rates were calculated for two standard populations: the world population (9) and the Japanese model population in 1985 (10).
Registration Reliability and Correction Coefficients of Re-estimated Incidence
Indices presenting registration reliability and correction coefficients for the re-estimated incidence are shown in Table 2 for all sites combined for each year in the 1985–89 period.
The proportion of DCO for all sites combined improved yearly from 15.7 to 14.0 among males and from 15.3 to 13.7 among females from 1985 to 1988 and then rose slightly in 1989. This decline in reporting rate in 1989 was caused by the last year data for the average of the three years 1988–90 in which not all of the delayed cancer reports could be included owing to a shortage of time until re-calculation. The proportion by site and gender is shown in Appendix 1 for 1985, 1987 and 1989. Although the figures for 1986 and 1988 are omitted from the Appendix owing to space limitations, those for each site were almost constant during the period. It was consequently concluded that no consideration was necessary regarding changes in the proportion of DCO in the observation of incidence trends.
The ratio of incidence to deaths is shown for all sites combined in Table 2 and according to site and gender in Appendix 2. The ratio for all sites ranged from 1.69 to 1.73 among males and from 1.90 to 1.96 among females. This ratio is regarded here as another index which can be used to indicate the completeness of reporting. However, it is also an index which indicates the survival rates of patients. The index was generally larger for sites with a more favorable survival rate and smaller for sites with an unfavorable survival rate.
The proportion of histologically verified cases to incidence indicating the diagnostic accuracy of registry data was 66–68% in males and 69–71% in females, for all sites combined, as shown in Table 2. The proportion is usually lower than that with respect to the reported cases, reflecting the size of the DCO proportion. The proportions by site and gender are shown in Appendix 3.
During the 1985–89 observation period, DCO (%) had been increasing and ratio of I/D and HV/I (%) had been decreasing in the female ‘lip, oral cavity and pharynx’ and female multiple myeloma, as observed in Appendices 1–3. This was caused by an increasing DCO (%) in the former site in three eligible registries and in four registries for the latter. The reason why the completeness of these sites had been decreasing in these registries is unclear. As both were among the most minor sites of the 25 sites appearing in the Appendices, late reporting might have accidentally had a great influence on these sites. The cancer reporting collection process for these two sites should be examined for these registries.
The correction coefficients for all sites combined are shown in the last two rows of Table 2 and were 0.96–0.97 for males and 0.97–0.98 for females during the 1985–89 period. No differences were observed between genders in these years. Correction coefficients according to site and gender are shown in Appendix 4 for 1985, 1987 and 1989. These coefficients were between 0.85 and 1.15 in most cancer sites. One situated outside the range 0.80–1.20 has been seen only in a few incidence sites for some years, e.g. thyroid in males and larynx, skin and corpus in females.
Cancer Incidence in Japan
The incidence and age-standardized incidence rates in 1985 and 1989 for the five leading sites are shown according to gender in Table 3. Incident cases for all sites combined increased from 187 200 in 1985 to 216 700 in 1989 (1.16-fold) in males and from 150 700 to 166900 (1.11-fold) in females. The leading site was the stomach both among males and females from 1985 to 1989. Among males the second leading site in the same period was the lung, followed by the liver, colon and rectum. Among females, the second leading site was the breast, followed by the uterus, colon and lung from 1985 to 1988; however, in 1989, the number of incidence in the colon and uterus was reversed.
Age-standardized rates (Table 3) for all sites combined rose from 254.4 in 1985 to 261.0 in 1989 (1.03-fold) among males and declined slightly from 167.0 to 165.5 (0.99-fold) among females. Among males, between 1985 and 1989 the rate decreased only for the stomach and increased considerably for the colon, rectum, liver, lung, etc. Among females, the rate decreased for the stomach and uterus. Increases were smaller among females than males even for sites where the incidence rate had increased.
Proportional site distributions based on age-standardized rates (world population) in 1989 are shown in pie chart form according to gender in Fig. 1.
The incidence, crude rates (per 100 000 population), age-standardized rates to the world population and to the Japanese model population in 1985 are shown according to gender and site in Appendices 5–8 for 1985, 1987 and 1989.
Time trends in age-standardized rates (world population) in all sites combined and the leading seven sites are shown in Fig. 2 for each year between 1975 and 1989, obtained from the first to third re-estimation series (2) on semi-logarithmic graph paper. Incidence rates for all sites combined increased until 1987 and then leveled off among males, whereas the rates rose slowly until 1987 and then showed a declining trend among females. The rates for stomach cancer showed a decreasing trend in both genders. The uterine cancer rate also continued to decline in females. Incidence rates for colon, rectum, liver, gallbladder, pancreas, lung and breast cancer rose between 1980 and 1985. However, in recent years the rate increase has slowed for cancers of the liver, gallbladder, pancreas and lung.
Appendices 9–11 show 5-year age-specific incidence rates according to site and gender for 1985, 1987 and 1989. When compared with age-specific incidence rates in 1985, the rate for each age-group younger than 30–34 in 1989 showed decreases of 17 and 8% on average among males and females, respectively, for all sites combined. Conversely, the age-specific rate for the age-group over 85 showed increases of 18 and 12% among males and females, respectively, in the same period. For age-groups between 35 and 84 there was only a 3% increase and 2% decrease among males and females, respectively, although the trends showed a considerable variation between age-groups.
Validity of the Correction in the Estimation Process
Our correction regarding cancer mortality in Japan conducted in the final estimation process, was based on the assumption that survival rates were almost equal for the cancer patients from all the eligible registries and those for the entire country. Regrettably, this assumption cannot be proven at present, because survival rates have been published for only three registries (11–13), and not for the other eligible registries or the country as a whole.
Re-estimation Evaluation on Completeness of Reporting
Incidence and age-standardized rates obtained by the annual estimates are shown in the top row of Table 4 and those obtained by re-estimation are shown in the bottom row for 1985, 1987 and 1989. Indices of registration reliability for each estimate are also shown in the same Table.
The reliability of the data was improved in both quality and quantity in the re-estimation when compared with the annual estimation. In the re-estimation, the proportion of DCO decreased from 16–17 to 13–15%, the ratio of incidence to mortality increased from 1.6–1.8 to 1.7–1.9 and the proportion of verified histologically cases rose 2–5%. The annual number of incident cases increased 8000–12 000 and age-standardized rates rose 9–16 (per 100 000 population) per year in both males and females in the re-estimation.
The proportion of DCO in 1985 was lower in annual estimates than in re-estimates (Table 4), which did not conform to usual trends. Since most of the five registries selected for an annual estimate in 1985 achieved high completeness, the average DCO was 12.7% for both genders, as shown in Table 5 and the completeness improved at the time of the re-estimate even for these registries. However, two additonal registries. whose completeness had not satisfied the standards at the time of annual estimate but barely satisfied it at the time of re-estimation, were included as eligible registries for re-estimation. This resulted in reducing the total completeness in re-estimates to 15.5% (Table 5). The HV/I (%) in 1985 in Table 4 was higher for annual estimate than for re-estimates. This was due to the same reason.
Discontinuity at the Juncture of 1980–84 and 1985–89
Discontinuities were observed in several cancer sites at the temporal juncture of the two re-estimation series: one for 1980–84 and the other for 1985–89. Possible causes are as follows: 1) eligible registries differed between two re-estimations 2) there was a longer time-lag until re-estimation for the first year of each 5-year period and a shorter time-lag for the last year. This could have caused incidence data to be characterized by a higher reporting rate including enough delayed cancer reports for the first year and incidence data with a relatively lower reporting rate for the last year of each 5-year period. 3) The Fukui Cancer Registry was included in the eligible registries in the re-estimation for 1985–89. This Registry started in 1984. Information obtained from medical institutes at the beginning of registration inevitably suffers from an over-reporting of prevalent cases, making it difficult to distinguish true incident cases (14). The above three factors may have caused a higher rate of incidence in the first year of the re-estimation series for 1985–1989.
The names of the eligible registries and the re-estimates reliability for the first and the second re-estimation series are shown in the left and the right sides of Table 6, respectively. Reliability data are shown for all sites and for both genders according to year of incidence.
Reliable 15-Year Re-estimates
Although the incidence data for 1985–89 have some problems as mentioned above, the results of the I–IIIrd re-estimation series provided final estimates of cancer incidence in Japan for 15 years from 1975 to 1989 which virtually achieved internationally acceptable completeness (9). These incidence estimates could be utilized both in cancer epidemiology and in programming and assessing cancer control programs. The Research Group has published future cancer incidence predictions for Japan until 2010 by extrapolations based on these 15 years of incidence estimates; they will be utilized to examine cancer problems in Japan from various aspects in the future (15).
Observation of Recent Changes in Incidence Rates
When time trends in cancer incidence were observed for the 15 years from 1975 to 1989, an increasing rate for sites whose incidence had been increasing seemed to change towards the end of the period. These changes should be confirmed to determine whether or not these are true trends or just a delay in increase for some reason which will be followed by an increasing trend in incidence. Time trends in mortality rates should be observed carefully in relation to this incidence observation.
- DCO/I (%) or DCO (%)
proportion of cases registered with information on the death certificate only to incident cases
ratio of incidence to deaths
- HV/I (%)
proportion of histologically verified cases to incident cases
- HV/R (%)
proportion of histologically verified cases to reported cases
age-standardized incidence rate to world population
The authors acknowledge participation in the study of all member registries of the Research Group for Population-based Cancer Registration in Japan from 1986 to 1993: Miyagi, Chiba, Kanagawa, Fukui, Aichi, Hyogo, Tottori, Kochi, Saga, Nagasaki, Hiroshima City and Nagasaki City. The authors also thank Mrs T. Kitagawa, Mr H. Matsumoto, Mr S. Kosono and MrK. Kawai for their assistance in facilitating the computer work required for this report.
This study was supported by a Grant-in Aid for cancer research from the Ministry of Health and Welfare (62-5 and 5-3).