Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
International Journal of Epidemiology © International Epidemiological Association 1997 Vol. 26, No. 5 Printed in Great Britain Estimation and Projections of Colorectal Cancer Trends in Italy RICCARDO CAPOCACCIA,* ROBERTA DE ANGELIS,* LUISA FROVA,** GEMMA GATTA,† MILENA SANT,† ANDREA MICHELI,† FRANCO BERRINO,† ETTORE CONTI,‡ LORENZO GAFÀ,§ LUCA RONCUCCI|| AND ARDUINO VERDECCHIA* Capocaccia R (Laboratory of Epidemiology and Biostatistics, Istituto Superiore di Sanità, viale Regina Elena 299, 00161 Rome, Italy), de Angelis R, Frova L, Gatta G, Sant M, Micheli A, Berrino F, Conti E, Gafà L, Roncucci L and Verdecchia A. Estimation and projections of colorectal cancer trends in Italy. International Journal of Epidemiology 1997; 26: 924–932. Background. Occurrence of and prognosis for tumours of the colon and rectum are thought to be changing rapidly due to simultaneous changes in risk factor prevalence, early diagnosis and treatment. In this paper time trends of morbidity, survival and mortality for colorectal cancer during the period 1970–1990 are estimated and analysed. Methods. Mortality trends were obtained from official death certificates. Relative survival rates were computed from population-based cancer registries. Incidence and prevalence rates were estimated from mortality and survival data. Results. Incidence rates were increasing during the period considered, with a lower rate of increase for the youngest birth cohorts. Relative survival rates of both colon and rectum cancers were higher for women, and for younger age groups, and were positively associated with period of diagnosis. No significant survival difference among the cancer registries used was found. A total of about 155 000 prevalent cases, 40% of which had been diagnosed >7 years before, were estimated in the Italian population for the year 1990. Mortality rates were slightly increasing for men and stable for women. Projections of colorectal cancer trends to the year 2000 indicate major expected rises in both incidence and prevalence. Conclusion. Colorectal cancer represents a problem of growing impact for health services in Italy. This conclusion can probably be extended to many developed countries. Keywords: colorectal cancer, occurrence, survival, time trends, Italy the diffusion of effective adjuvant therapies. Recent data show,1 for many countries, diverging time trends between incidence and mortality, with mortality rates remaining almost steady, or even decreasing, despite an increasing trend in incidence rates. A recent large collaborative survival study of data from 30 populationbased cancer registries in 12 European countries,3 showed an increasing proportion of long-term survivors of colon and rectum cancer. Finally, colorectal cancer prevalence is expected to become a major health problem as a result of high and increasing incidence and survival rates. Italy presents all the features of the dynamics described above with, especially in the Northern regions, socioeconomic conditions and cultural attitudes similar to most Western European countries. A marked geographical North-South gradient of colorectal mortality was evident in Italy,4 with two- to three-fold higher mortality in the Northern regions. Reduction of these geographical differences is expected, as a consequence of the progressive homogenization of Italian dietary habits.2 For these reasons Italy provides a good location for the study of colorectal cancer occurrence. From an aetiological point of view, this enables the consistency between colorectal cancer trends and the changing The epidemiology of colorectal cancer is expected to undergo significant changes as a consequence of an evolving pattern of possible risk factors and of continuing improvement in diagnostic and therapeutic activities. These phenomena could substantially modify, directly and through their mutual links, levels and trends of incidence, survival, mortality and prevalence. Colorectal cancer risk, which has been increasing for several generations and is still increasing in populations with low incidence,1 is likely to have been influenced by the dramatic dietary changes over this century and by the progressive diffusion of industrial alimentary products and the subsequent homogenization of dietary habits.2 Moreover, incidence will be significantly modified if the frequency of early diagnosis grows substantially in the near future. Survival is improving as a result of an increase in early diagnosis, advances in surgical technology and * Laboratory of Epidemiology and Biostatistics, Istituto Superiore di Sanità, viale Regina Elena 299, 00161 Rome, Italy. ** National Institute of Statistics, Rome, Italy. † Division of Epidemiology, National Cancer Institute, Milan, Italy. ‡ Latina Cancer Registry, Latina, Italy. § Ragusa Cancer Registry, Ragusa, Italy. || Colorectal Cancer Registry, Modena, Italy. 924 COLORECTAL CANCER TRENDS IN ITALY pattern of its possible determinants to be checked. On the other hand, it gives a reference frame to compare, interpret and evaluate the future trends of the disease. In this paper we present and analyse the observed mortality and survival data and the estimated incidence and prevalence rates for colorectal cancers in Italy during the period 1970–1990. On the basis of past trends, incidence and prevalence projections to the year 2000 are also given. MATERIALS AND METHODS Survival data were collected from all incident cases observed in four Italian provinces covered by cancer registration. The Lombardy cancer registry contributed 2393 incident cases diagnosed in the province of Varese (Northern Italy) during the period 1978–1984. The cancer registries of Modena (period of diagnosis 1985) and Latina (1983–1984), provinces in central Italy, contributed 132 and 189 cases, respectively. Finally, 303 incident cases from the southern province of Ragusa (1981–1984) were included. Twenty-three cases were lost to follow-up. The remaining 2994 cases (1868 colon cancers and 1126 rectum cancers) were followed until December 1991, thus ensuring a minimum potential follow-up of 5 years. Official mortality data for the period 1970–1990 were obtained from the National Institute of Statistics (ISTAT) as individual records. Death codes 153 (colon) and 154 (rectum) were selected for the period 1970– 1979 (ICD 8th revision), while codes 153, 154 and 159.0 (intestine NOS) were considered for the period 1980–1990 (ICD 9th revision). Population size by age, sex and geographical area was estimated from census data and from inter-census births, deaths and migration numbers.5 It is well known1 that separate analysis of mortality data for colon and rectum cancers is made difficult by major misclassification problems in defining and coding the cause of death. As usual in descriptive epidemiological works, the two cancer sites will therefore be analysed jointly. This paper is based on methods for statistical analysis of relative survival rates, for the estimation of incidence and prevalence, and for trend projections, applied in a previous6 work on stomach cancer. Relative survival was computed by the method of Hakulinen et al.7–9 The method of Estève et al.10 was used for multiple regression analysis. In both cases, observed survival rates were corrected for age-, area-, sex- and period-specific general mortality levels. Further details are reported in a previous paper.11 The estimation of incidence and prevalence from mortality and survival data was carried out by the 925 method of Verdecchia et al.12 which has been applied in previous studies.13,14 It assumes incidence to be a continuous function of independent factors such as age, year of diagnosis (period) and year of birth (cohort). The equations relating incidence, survival and mortality rates were used as a link function between incidence and mortality in a non-linear regression model for maximum likelihood estimation of the parameters of the incidence function. These parameters were determined as those giving the best fit of the mortality rates observed during the period 1970–1990, on the basis of the given survival rates. Prevalence, defined as the proportion of all people with previous colorectal cancer diagnosis in the general population, was then derived from the estimated incidence function and the observed survival rates. For method validation purposes, estimates of colorectal cancer incidence in Italian provinces with cancer registration were compared to the corresponding data observed by the local registries. A total of 3008 male cases and 2640 female cases were estimated in the four cancer registries. The corresponding numbers observed by the registries were 3004 and 2605, respectively. Period-specific comparisons were possible for the Lombardy cancer registry, which observed 487 new cases during 1976–1978, 1011 cases during 1978–1982, and 1352 cases during 1983–1987. The corresponding estimates were 481, 993 and 1318 cases, respectively. These results, reported in more detail elsewhere,15 indicated a good agreement between estimated and observed data. Projections of incidence rates for the period 1991– 2000 were calculated on the basis of the incidence function obtained for the estimation period 1970–1990. For this purpose, both age- and cohort-estimated effects were also assumed to hold during the projection period. Furthermore, no incidence changes simultaneously affecting all ages were considered. Regarding the pattern of survival, we adopted a scenario approach by assuming: A) survival continuing to increase at the same rate as observed during the period 1978–1985 or B) survival remaining constant at the rates estimated in 1990. Future prevalence and mortality rates were derived from projected trends of both incidence and survival functions. Birth cohort-specific expected 0–84 years cumulative risks were used to represent colorectal cancer cohort effects. This indicator is calculated1,16 as the sum of agespecific incidence rates estimated (at ages ,1991 minus year of birth) or projected (at ages .1990 minus year of birth) for each birth cohort. It estimates approximately the probability of developing the disease before age 85 for people belonging to a given birth cohort and in the absence of competitive mortality. Estimates of 926 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY FIGURE 1 Estimated cumulative risk of colorectal cancer incidence in male birth cohorts from North, Centre and South of Italy FIGURE 2 Estimated cumulative risk of colorectal cancer incidence in female birth cohorts from North, Centre and South of Italy cumulative risks for birth cohorts born before 1895 and after 1940 should be considered with caution. The former are derived from under 10 years of observed mortality data, while the latter are based on the relatively low mortality rates observed for people under 50. Cross-sectional 0–84 years cumulative rates were also calculated, according to the definition given by standard cancer registries publications, summing up the estimated age-specific incidence rates obtained by the model for a fixed time period. While the estimation method works in reverse, from mortality to incidence and prevalence, the results are better analysed in the opposite sense, i.e. from determinants (incidence and survival) to effects (prevalence and mortality). They will therefore be presented and discussed accordingly. Cross sectional 0–84 cumulative rates, estimated for the whole country in the year 1990, were 9.3 and 6.3 for males and females respectively. Table 1 reports the national age-adjusted estimated and projected incidence rates of colorectal cancer at the national level, according to sex, decade and age group, with the corresponding absolute number of cases for all ages. The increase in incidence is evident in both sexes and in all ages for the period 1970–1990. However, a deceleration of the increase between the first and the second decade is estimated for the 25–49 age group. A decrease of incidence is obtained, for the same age group, during the projection period 1991–2000. Ageadjusted rates are scarcely influenced by the youngest cohorts, and steeply increase during both the estimation and the projection periods. The ratio between male and females age-adjusted rates increases from 1.28 in 1970 to 1.52 in the year 2000. The estimated absolute number of cases rises from 16 000 (49% men, 51% women) for the year 1970, to nearly 50 000 for the year 2000 (53% men, 47% women). We estimated by age standardization that about one half of this growth can be attributed to the higher incidence levels, while the other half was due to the increasing size of the elderly population. RESULTS Incidence Trends of expected 0–84 cumulative risk by birth cohort and geographical area are represented in Figures 1 and 2, for men and women respectively. The probability of having colorectal cancer for people born in 1890 was about 4–5% in the North and Centre and 2% in the South of the country, with a sex ratio very near to one. The cumulative risk increased steeply for subsequent generations. Male cohorts born in 1940 present an expected risk of 18% for the Centre, 16% for the North and 13% for the South. In women, it was about 10% in all areas. A deceleration of the increasing trend is evident however for the most recent cohorts. A plateau is apparently reached for people born around 1950, whose cumulative risks by geographical area vary in men, but are quite homogeneous in women. Survival The excess mortality rate of colorectal cancer patients was analysed according to sex, age group, period of diagnosis, registry and subsite. Figure 3 presents the overall survival curves by subsite. Colon cancer patients had lower survival rates (60% versus 68%) during the first year since diagnosis, afterwards their survival curve quickly tends to level off. Rectum cancer, on the contrary, carried higher mortality from one up to 5 years 927 COLORECTAL CANCER TRENDS IN ITALY TABLE 1 World population age-standardized incidence rates and absolute number of new cases, estimated and projected for colorectal cancer in Italy (rates are per 100 000, no. of cases rounded to 100) Age group 25–49 50–69 70–84 25–84 All ages No. of cases Men Women 1970 1980 1990 2000 1970 1980 1990 2000 7.9 70 220 42 23 7800 9.4 86 275 51 29 11 000 11 122 387 71 40 18 600 10 156 517 90 50 26 000 7.7 54 169 33 18 8100 9.5 66 206 40 22 11 700 11 83 252 49 28 18 300 10 100 314 59 33 23 500 FIGURE 3 Relative survival rates of colon and rectum cancer patients in Italy. 2994 incident cases diagnosed by four Italian cancer registries during the period 1978–1984 since diagnosis. The two curves cross-over at 2 years from diagnosis and are parallel after 5 years. Table 2 reports 5-year observed and relative survival rates, according to site, sex and age group. Colon cancer patients presented higher survival with respect to those with rectum cancer, in both sexes and in all ages. Furthermore, better prognosis was observed for women than for men, and for the youngest versus the oldest age groups. Because of the clear non-proportionality of mortality risks between colon and rectum cancer patients, multiple regression analysis was separately performed for the two subsites. The results are shown in Table 3, which presents the relative risks associated with the factors considered and the corresponding confidence limits. After allowing for the different shapes of the baseline survival curves, colon and rectum cancer present relative risk patterns very similar to each other. No clear geographical pattern was found: relative survival rates for the other registries did not differ significantly from the ones estimated for Varese. However, the factor ‘area’ was not removed from the regression model to avoid possible bias due to different subsite distribution or to different incidence periods covered by the various registries. Survival rates were higher for women than for men, but the significance of sex coefficient was borderline. Survival decreased with age increasing, with relative risks particularly high for the oldest age group (about 2 for the middle age group). Survival increased over time: the relative risk associated with a diagnosis made in 1984–1985 compared to 1978–1979 was 0.76 for colon and 0.72 for rectum. The solid line in Figure 4 shows the overall relative survival curve jointly estimated for colorectal cancer in 1980–1981. The procedure for estimating incidence and prevalence of colorectal cancer required that survival data refer to exactly the same nosological grouping. In fact, survival rates of colorectal cancer at the national level depend on the relative proportion of the two sites, which present slightly different survival levels. The variability of the proportion of colon cancer over all colorectal cancer deaths across Italian regions ranges between 54% and 72%, considering17 the lowest and the highest sex- and area-specific values. The survival curves obtained under these two extreme alternatives, (Figure 4), are very close to each other. We concluded that the sensitivity of survival values to the assumed mixing proportion is only slight. This problem was therefore ignored in the estimation procedure, and the same family of age-, period- and sex-specific survival curves was assumed for colorectal cancer patients in the Italian population. 928 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY TABLE 2 Observed 5-year survival rates and corresponding standard errors for 2994 colon and rectum cancer patients in Italy, by sex and age group Age Colon Rectum No. OBSa SEb RELc SE No. OBS SE REL SE Men 0–49 50–59 60–69 70–79 80+ All 73 172 231 313 107 896 52.1 42.3 32.9 24.3 10.1 31.3 5.8 3.8 3.1 2.4 2.9 1.6 53.6 50.1 38.5 34.9 24.5 40.7 6.0 4.1 3.6 3.5 7.2 2.0 50 123 159 221 70 623 46.0 43.1 27.0 17.2 8.6 26.2 7.1 4.5 3.5 2.6 3.4 1.8 47.3 46.5 31.5 24.8 20.4 33.8 7.2 4.8 4.1 3.7 8.0 2.8 Women 0–49 50–59 60–69 70–79 80+ All 94 133 225 328 192 972 50.0 43.6 41.8 35.5 19.8 36.4 5.2 4.3 3.3 2.7 2.9 1.5 50.0 44.7 44.4 42.8 39.2 44.0 5.2 4.4 3.5 3.2 5.7 1.8 48 79 125 177 74 503 40.0 41.7 35.2 27.6 9.5 30.2 7.1 5.6 4.3 3.4 3.4 2.0 39.9 42.7 37.4 33.2 18.7 35.6 7.1 5.7 4.5 4.0 6.7 2.4 a OBS = observed survival rate. SE = standard error. c REL = relative survival rate. b TABLE 3 Relative risks, (RR) and 95% confidence intervals (CI), of excess hazard rates for colon and rectum cancer patients in Italy Factor Colon Rectum RR 95% CI RR 95% CI Sex Males Females 1 0.85 0.74–0.96 1 0.88 0.75–1.03 Age 0–49 50–59 60–69 70–79 80+ 0.77 0.81 1 1.17 1.96 Period 1978–1979 1980–1981 1982–1983 1984–1985 Registry Varese Modena Latina Ragusa 0.60–1.00 0.66–1.00 0.99–1.38 1.60–2.38 0.72 0.74 1 1.15 1.93 0.52–0.97 0.58–0.94 0.94–1.41 1.48–2.51 1 0.93 0.73 0.76 0.78–1.12 0.61–0.88 0.61–0.94 1 0.93 0.83 0.72 0.74–1.15 0.66–1.04 0.55–0.93 1 1.00 1.27 1.15 0.69–1.45 0.98–1.64 0.93–1.42 1 0.62 0.89 1.10 0.38–1.01 0.61–1.28 0.84–1.44 Prevalence Estimated prevalence is presented by sex and age group in Table 4. In 1990, men and women belonging to the youngest age group present similar levels. Estimated prevalence steeply increases with age, with levels about 20% higher in men than in women, for both the middle and the old age group. Table 4 also reports the absolute number of prevalent cases estimated in 1990 and stratified according to time since diagnosis. The first group considers patients (29% of total cases) diagnosed during the current or preceding year, which are likely to be still in treatment for their disease. About 40% of these patients are expected subsequently to die from colorectal cancer. The second group concerns patients diagnosed 2–6 years before (32% of total cases) who are still at risk of disease recurrence (estimated about 10%), and who present mortality levels higher than the general population. The third group considers cases whose life expectancy is similar to that of the general population,18 and therefore can be considered as cured (40% of total cases). The overall proportion of long-term survivors increases with age: 31% for the age group 25–49 years, 37% for 50–69 and 42% for those aged 70–84 years. Projected levels for the year 2000 were calculated assuming either that survival will continue to increase at the currently estimated rate (Scenario A), or that it 929 COLORECTAL CANCER TRENDS IN ITALY FIGURE 4 Overall relative survival rates of colorectal cancer patients, according to the assumed proportion of colon cases. Solid line: observed proportion (62%); dotted and dashed lines: minimum (54%) and maximum (72%) proportion, respectively will remain constant at 1990 levels (Scenario B). As the behaviour of incidence is considered to be the same, projected prevalence will be higher under the first assumption. During the projection period, the prevalence is expected to remain nearly constant in the young, and to rise in the other age groups. The number of cases almost doubles in 10 years, particularly under Scenario A. Different assumptions on survival trends have a minor influence on 10-year projections of prevalent cases, which are mostly determined by the rise in incidence and by population ageing. Mortality Observed and projected trends in colorectal cancer mortality are presented by sex and age group in Table 5. The general pattern is quite similar between the two sexes. Due to the combined trends of incidence and survival, mortality rates appear slightly increasing during 1970–1980, and slightly decreasing or stable during the subsequent decade. The direction of projected mortality trends for the period 1991–2000 depends on the assumption of future behaviour of survival rates. Increasing survival, as considered in Scenario A, will lead to a future drop in mortality rates in both sexes and in all age groups. On the contrary, survival stable at the late 1980s levels (Scenario B) will be associated with future increases in mortality rates for ages >50 years and in agestandardized rates. A decrease in mortality, because of the estimated declining cohort effects of incidence, will only be observed for the youngest age group. DISCUSSION National incidence and prevalence trends presented above were estimated by assuming that survival rates, derived by sex, age and year of diagnosis from four Italian cancer registries, were valid for the whole population. Although the registries used operate in regions with different socio-economic levels, the Italian public national health service, established in the 1970s, in principle provides fairly uniform availability of health facilities nationwide. However, TABLE 4 World population age-standardized prevalence and absolute number of prevalent cases, estimated and projected for colorectal cancer in Italy (rates are per 100 000, no. of cases rounded to 100). Projection Scenarios: A, survival increasing; B, survival constant Sex/age group 1990 2000 Cases by disease duration Rate Scenario A Scenario B All 0–1 2–6 .6 Rate Cases Rate Cases Men 25–49 50–69 70–84 25–84 All ages 47 473 1462 276 152 4600 30 900 30 300 65 800 71 200 1700 10 000 8800 20 500 21 800 1600 11 000 10 000 22 600 24 300 1300 9900 11 500 22 700 25 100 51 777 2582 449 250 5000 53 000 65 000 123 000 137 000 48 717 2345 412 229 5000 49 000 58 000 112 000 125 000 Women 25–49 50–69 70–84 25–84 All ages 46 407 1163 234 131 4500 30 500 37 500 72 500 84 200 1500 8500 9800 19 800 22 500 1500 9300 10 700 21 500 24 700 1500 12 700 17 000 31 200 37 000 53 612 1841 348 181 6000 46 000 69 000 121 000 138 000 50 575 1705 325 169 5000 43 000 64 000 112 000 128 000 930 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY TABLE 5 World population age-standardized mortality and absolute number of deaths, observed and projected for colorectal cancer in Italy (rates are per 100 000, number of cases are rounded to 100). Projection scenarios: A, survival increasing; B, survival constant Age group Men 1970 1980 Women 2000 1990 A 25–49 50–69 70–84 25–84 All ages No. of cases 4.6 44 165 28 15 5300 4.1 51 190 31 18 7000 4.1 52 200 32 18 9000 1970 2000 1990 B 2.3 3.5 44 64 188 268 28 40 16 24 9500 13 500 endoscopy departments were established later and are less frequent in the South than in the North. The slight survival differences between areas therefore can be related to different distribution of disease stages, even if there was no organized screening during the study period. The definition of stage categories is not uniform in time and between areas, and depends on the thoroughness of diagnostic investigations. This phenomenon is well known as the ‘stage migration effect’.3 Tumour stage, broadly classified as localized, regional and distant spread, was available in our study for about 40% of patients only. The percentage of localized and distant spread cancers decreased, over the study period, from 38% and 31% in 1978–1979 to 25% and 22% in 1984–1985, respectively. Meanwhile, the percentage of regional spread cancers increased from 31% to 53%. This pattern can be due to a mixture, with unknown weights, of both early diagnosis and stage migration effects. For these reasons, stage was not analysed as a prognostic factor in this study. Survival rates reported by Italian cancer registries are similar for colon cancer, and slightly lower for rectum cancer, compared to those estimated for the European population.3 The prognostic role of age and decade of diagnosis is also consistent with findings from the European study,3 as well as the different shapes of survival curves between colon and rectum according to time since diagnosis. This last phenomenon could be explained by earlier symptoms and easier diagnosis for cancer sites which are more accessible, such as the rectum. Consistency with survival data from other European countries strengthen the assumption that age and time trends of the survival rates used in the model give a reliable basis for estimation.3 1980 A 3.7 32 115 20 11 5100 4.1 34 130 22 13 7000 3.7 33 124 21 12 8800 B 2.0 3.0 25 37 103 148 16 24 10 14 7500 10 800 The aetiology of colon and rectum cancers has yet to be well-ascertained. However, strong indications of close relationships with dietary factors come from both analytical and descriptive epidemiological studies.19 Colorectal cancers appear to be positively associated with consumption of saturated fats and animal proteins, and negatively associated with intake of vegetables and fruits. Data from population surveys on food consumption2 revealed major differences among Italian regions, with a typical Mediterranean diet in Southern communities, and a higher risk dietary pattern in Northern communities. These differences, that were striking until the 1950s, appeared to have decreased in recent decades. In the 1980s, Italian dietary habits appeared to be more homogeneous, and to be approaching the high risk pattern of Northern European countries. These observations are consistent with the incidence trends of the disease estimated above, showing an increase of cumulative risk by birth cohort particularly marked for the Southern regions. The different incidence for males and females suggests an aetiologic role for sex-related factors. Some studies found that reproductive factors, such as nulliparity, advanced age at first and last delivery, might be associated with a higher risk of colorectal cancer. The interaction between sex hormone metabolism and nutrition may also play an aetiologic role.20 A fairly constant mortality trend was found in Italy despite increasing incidence rates. This is a direct consequence of increasing survival over time. Coleman et al.1 showed similar patterns of incidence and mortality in other European population-based cancer registries. Projected incidence and mortality rates have been obtained based on the assumption that age and cohort effects estimated for 1970–1990 will continue to be valid COLORECTAL CANCER TRENDS IN ITALY for the projection period 1991–2000. Possible changes in true colorectal cancer risk, associated for instance with changing dietary patterns, are likely to occur gradually over time, and therefore should already be captured by the estimated trends of cohort effects. It is unlikely that risk of colorectal cancer due to other factors will change so suddenly as to affect incidence rates simultaneously and in a substantial way for all age groups during the projection decade. On the other hand, early diagnosis could affect measured incidence rates, including a higher frequency of localized cancers and could also improve survival. Increasing survival could also be expected if adjuvant therapies21 prove to be effective. During the projection period, survival has been assumed either to increase by 4% per year (Scenario A), or to remain constant (Scenario B). The estimates derived under Scenario B are in accordance with previously published projections,22 which are based only on observed mortality trends, and do not take into account possible future increases in survival. Projections of colorectal cancer trends can be better appreciated if compared to the corresponding estimates obtained for stomach cancer, 6 which in Italy still represents the second most common site of cancer death. The absolute number of incident cases of stomach cancer is expected to decrease from about 18 000 to 16 000 during the period 1990–2000, in contrast with the forecasted rise (from 37 000 to 50 000) of cancers of the colon and rectum. The differences in expected trends of prevalence appear even more striking. The number of patients with previous diagnosis of gastric cancer should increase during the present decade from 104 000 to 110 000, under the hypothesis of increasing survival rates. Meanwhile, the number of prevalent cases of colorectal cancer is expected almost to double: from about 150 000 cases in the year 1990 to at least 250 000 cases (assuming constant survival) or to a maximum of 275 000 cases (assuming increasing survival) in the year 2000. Particular intensive care and follow-up protocols should be given to the 30% of patients diagnosed <2 years before, due to the higher risk of local recurrences.21,23 However all patients remain at high risk for a second colorectal cancer.23 The figures presented in this paper may be useful for estimating the need for periodical endoscopic examinations of all prevalent cases throughout the rest of their lives, as routinely recommended. For all these reasons, the potential impact of future trends of colorectal cancer prevalence on the national health service is bound to become a major one. The same problems are expected to arise in European countries presenting demographic and epidemiological trends similar to those reported for Italy in this paper. 931 ACKNOWLEDGEMENTS The authors wish to thank Ms Emily Taussig for editorial revision and preparation of the manuscript. This work was partially supported by the National Research Council (CNR), Progetto Finalizzato ACRO and by the European Union BIOMED Project EUROCARE. REFERENCES 1 Coleman M P, Estève J, Damiecki P, Arslan H, Renard A. Trends in Cancer Incidence and Mortality. Lyon: International Agency for Research on Cancer, 1993; IARC Sci Pub. No 121. 2 Berrino F, Krogh V. I consumi alimentari in Italia in relazione ai tumori dell’apparato digerente. In: Capocaccia R, Verdecchia A, Terracini B (eds). Epidemiologia dei tumori dell’apparato digerente in Italia. Ann Ist Super Sanità 1996; 4: 421–432. 3 Berrino F, Sant M, Verdecchia A, Capocaccia R, Estève J, Hakulinen T (eds). Survival of Cancer Patients in Europe. Lyon: International Agency for Research on Cancer, 1995; IARC Sci Pub No.132. 4 Capocaccia R, Farchi G, Prati S et al. La mortalità in Italia nell’anno 1990. Reports of the Istituto Superiore di Sanità, ISTISAN 93/23, 1993. 5 Capocaccia R, Caselli G. Popolazione residente per età e sesso nelle province italiane. Dipartimento di Scienze Demografiche dell’Università degli Studi di Roma ‘La Sapienza’. Fonti e Strumenti 2, 1990. 6 Capocaccia R, De Angelis R, Frova L et al. Estimation and projections of stomach cancer trends in Italy. Cancer Causes Control 1995; 6: 339–46. 7 Hakulinen T. On long term relative survival rates. J Chron Dis 1977; 30: 431–43. 8 Hakulinen T, Tenkanen L, Abeywickrama K. Testing equality of relative survival patterns based on aggregated data. Biometrics 1987; 43: 313–25. 9 Hakulinen T, Abeywickrama K. A computer program package for relative survival analysis. Comput Prog Biomed 1987; 19: 197–207. 10 Estève J, Benamou E, Croasdale M, Raymond L. Relative survival and the estimation of net survival: elements for further discussion. Stat Med 1990; 9: 529–38. 11 Ponz M, Micheli A, Gatta G et al. La sopravvivenza per tumori del colon e del retto in Italia. In: Capocaccia R, Verdecchia A, Terracini B (eds). Epidemiologia dei tumori dell’apparato digerente in Italia. Ann Ist Super Sanità 1996; 4: 527–536. 12 Verdecchia A, Capocaccia R, Egidi V, Golini A. A method for the estimation of chronic disease morbidity and trends from mortality data. Stat Med 1989; 8: 201–16. 13 Capocaccia R, Verdecchia A, Micheli A, Sant M, Gatta G, Berrino F. Breast cancer incidence and prevalence estimated from survival and mortality. Cancer Causes Control 1990; 1: 23–30. 14 Capocaccia R, Micheli A, Berrino F et al. Time trends of lung and larynx cancers in Italy. Int J Cancer 1994; 57: 1–8. 15 De Angelis R, Frova L, Capocaccia R, Verdecchia A. Incidenza e prevalenza dei tumori dell’apparato digerente in Italia: stime dai dati di mortalità. In: Capocaccia R, Verdecchia A, 932 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Terracini B (eds). Epidemiologia dei tumori dell’apparato digerente in Italia. Ann Ist Super Sanità 1996; 4: 537–550. 16 Capocaccia R, Negri E, La Vecchia C, Decarli A. Time trends of lung cancer in Italy. Eur J Cancer 1993; 29A: 1288–92. 17 Zanetti R, Crosignani P (eds). Cancer in Italy. Incidence data from Cancer Registries, 1983–1987. Torino, Italy: Lega Italiana per la lotta contro i tumori, 1992. 18 Sant M, Capocaccia R, Verdecchia A et al. Comparisons of colon cancer survival among European countries: the EUROCARE study. Int J Cancer 1995; 63: 43–48. 19 Tomatis L. Cancer: Causes, Occurrence and Control. Lyon: International Agency for Research on Cancer, 1990; IARC Sci Pub. No.100. 20 DeCosse J J, Ngoi S-S, Jacobson J S et al. Gender and colorectal cancer. Eur J Cancer Prevention 1993; 2: 105–15. Haller D G. An overview of adjuvant therapy for colorectal cancer. Eur J Cancer 1995; 31A: 1255–63. 22 Negri E, La Vecchia C, Decarli A et al. Projection to the end of the century of mortality from major cancer sites in Italy. Tumori 1990; 76: 420–28. 23 Steele G. Follow-up plans after treatment of primary colon and rectum cancer. World J Surg 1991; 15: 583–88. 21 (Revised version received March 1997)