1Istituto Nazionale per lo Studio e la Cura dei Tumori, Milano, Italy; 2National Cancer Intelligence Centre, Office for National Statistics, London, UK; 3London School of Hygiene and Tropical Medicine, London, UK; 4Finnish Cancer Registry, Helsinki, Finland; 5Istituto Superiore di Sanità, Rome, Italy
Received 12 July 2001; revised 26 November 2001; accepted 13 December 2001.
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Information on cancer prevalence is of major importance for health planning and resource allocation. However, systematic information on cancer prevalence is largely unavailable.
Materials and methods
Thirty-eight population-based cancer registries from 17 European countries, participating in EUROPREVAL, provided data on almost 3 million cancer patients diagnosed from 1970 to 1992. Standardised data collection and validation procedures were used and the whole data set was analysed using proven methodology. The prevalence of stomach, colon, rectum, lung, breast, cervix uteri, corpus uteri and prostate cancer, as well as of melanoma of skin, Hodgkins disease, leukaemia and all malignant neoplasms combined, were estimated for the end of 1992.
Results
There were large differences between countries in the prevalence of all cancers combined; estimates ranged from 1170 per 100 000 in the Polish cancer registration areas to 3050 per 100 000 in southern Sweden. For most cancers, the Swedish, Swiss, German and Italian areas had high prevalence, and the Polish, Estonian, Slovakian and Slovenian areas had low prevalence. Of the total prevalent cases, 61% were women and 57% were 65 years of age or older. Cases diagnosed within 2 years of the reference date formed 22% of all prevalent cases. Breast cancer accounted for 34% of all prevalent cancers in females and colorectal cancer for 15% in males. Prevalence tended to be high where cancer incidence was high, but the prevalence was highest in countries where survival was also high. Prevalence was low where general mortality was high (correlation between general mortality and the prevalence of all cancers = 0.64) and high where gross domestic product was high (correlation = +0.79). Thus, the richer areas of Europe had higher prevalence, suggesting that prevalence will increase with economic development.
Conclusions
EUROPREVAL is the largest project on prevalence conducted to date. It has provided complete and accurate estimates of cancer prevalence in Europe, constituting essential information for cancer management. The expected increases in prevalence with economic development will require more resources; allocation to primary prevention should therefore be prioritised.
Key words: cancer, cancer registry, Europe, prevalence
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
This paper presents the main findings of the EUROPREVAL project, a European concerted action for studying cancer prevalence in order to reveal and evaluate differences in requirements for cancer-related health care. EUROPREVAL provides the first large-scale comparative overview of the prevalence of selected major cancers in Europe.
![]() |
Materials and methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
The rules for including cases in the analysis were those used by the EUROCARE study on the survival of European cancer patients [15]. An advantage of this is that the EUROPREVAL results are consistent with those of EUROCARE. The participating registries were asked to provide incidence and follow-up data for cancer cases diagnosed up to 31 December 1992. The Iceland, Saarland and Geneva CRs each had diagnosis periods of 23 years (1970 to 1992)the longest observation period of all the CRs in this studywhile the Tyrol and Warsaw CRs had the shortest included registration periods: 5 years each (1988 to 1992) (Table 2). The information provided on individual patients with cancer consisted of: gender; dates of birth, diagnosis and end of follow-up or death; life status at end of follow-up; tumour site (according to ICD-9 code) [16]; type of diagnosis (histological, cytological or other); and tumour histotype (according to ICD-O code) [17].
To protect confidentiality, only the month and year of dates of birth, diagnosis and end of follow-up or death were included in the data provided by the registries. All these events, except the end of follow-up (31 December 1992), were assumed to have occurred on the 15th day of the month in question.
Malignancies 204208 of the ICD-9 classification were grouped as leukaemias. The category all malignant neoplasms includes all malignancies (ICD-9 140208) except non-melanoma skin cancer (ICD-9 173). The prevalence of all malignant neoplasms may be of little interest from the clinical point of view, but is of great public health interest because it provides an overall indication of the demand for cancer-related health care in a population.
Cases known to CRs by death certificate only (DCO) and those diagnosed at autopsy were not included in the analysis (Table 1 shows the overall percentages of DCO cases by CR). When more than one cancer was diagnosed in a patient, only that diagnosed first was considered. For cases of multiple synchronous tumours, only the most advanced or that causing death was considered. Bilateral synchronous tumours of symmetrical organs were considered as one cancer. The implication of these rules is that we considered the prevalence of persons with cancer and not the prevalence of cancers.
The percentage of patients lost to follow-up in each registry ranged from 0% (many CRs) to 10% (Somme) (Table 2). All case records were checked for errors and inconsistencies (unusual or inconsistent dates or cancer codes, unusual or inconsistent sexsite-morphology combinations) according to the EUROCARE protocol [15, 18]. Defective records were sent back to registries for correction or completion; considerable effort was made to complete and correct individual case records so that as many as possible could be included in the analysis, thereby reducing to a minimum the underestimation of prevalence.
Definition of terms
This study produced point prevalence estimates pertaining to a specific reference day (31 December 1992).
Estimation of prevalence
When the population has been covered by cancer registration for a very long time then the prevalence can be calculated basically by counting directly from the CR data, since we may assume that no cases are surviving that were diagnosed before the CR began registering cases. As shown in Table 2, this was not the case in our study for any of the CRs, because either the registry has been established recently, or the full registry series is not included in the EUROCARE-2 database. Therefore, depending on the total time a CR has been providing cases (the observation period), there must be some additional surviving patients who were diagnosed before the date of available data. These cases must be estimated and their number added, as an adjustment, to the observed prevalence.
Observed prevalence
The observed prevalence was calculated by the PREVAL counting method [13, 19]. PREVAL employs a matrix with three time dimensions where the unit is a year: calendar time; age; and years from diagnosis. Each cancer patient is defined at a given point in time (i.e. a specific day) by age at diagnosis, calendar year of diagnosis and years from diagnosis (which takes the value zero initially). The case is added to other cases with the same values forming a cohort. At each calendar year, the method verifies whether each patient is still alive, and for each age group counts the total number of patients remaining in the cohort. The prevalence on a certain day in a given calendar year is obtained by adding the results from all cohorts. The PREVAL method also incorporates an adjustment to take account of patient loss during follow-up. To implement this adjustment, the following formula is applied to each i, j cell of the matrix along the k axis (time since diagnosis):
where Ak is the number of patients of initial age i and alive at the end of the calendar year j; Ds is the number of patients who died during that year; Lm is the number of patients lost to follow-up at a given m year since diagnosis; and As is the number of patients alive at the end of the year. Thus, the formula multiplies the number of lost to follow-up cases by the time interval survival probability [As/(As + Ds)]. Ek is therefore the expected number of patients diagnosed at age i and alive at the end of year j, taking into account the survival of those lost to follow-up. This adjustment assumes that the lost patients have the same probability (specific for sex and age at diagnosis in a given calendar year) of surviving as those not lost to follow-up.
Total prevalence
The observed prevalence, corrected for lost cases as above, was adjusted by a prevalence completeness index determined by a previously published and validated method for estimating the unknown fractions of the total prevalence [20]. The prevalence completeness index (R) defined by the following formula:
R = NO(m)/NT(m)
is an estimate of the proportion of the total prevalence expressed by the observed prevalence, where NO(m) and NT(m) are model estimates of observed and total prevalence, respectively [20]. These quantities are derived from a mathematical expression relating prevalence to incidence and survival probabilities [20]. The completeness index varies between 1, when all prevalent cases are observed (i.e. the CR has been operating for a very long time), and (theoretically) 0, when no prevalent cases are observed by the CR. R depends on the length of the registration period, cancer-specific incidence rates by age class and cancer-specific survival rates by age class.
Values of R were estimated for each cancer site for each of the four broad European areas defined previously. We had difficulties in estimating past incidence and survival trends by age for cancer of the cervix uteri and for Hodgkins disease [21]; consequently we had problems in estimating R values and hence the total prevalence for these two malignancies. However, in these two cases the method allowed us to produce an adjustment of the observed prevalence that furnished estimates of the 15 year prevalence.
Weighted European mean
In order to take into account that the extent of cancer registration varied between countries, when calculating the crude European mean prevalence we used a weighting factor, w = (100 x n)/c, where n is the average annual number of patients with a given cancer registered by the CR or CRs representing a country, and c is the level of national coverage. In this study c ranged from 1.7 (Germany) to 100 (for those countries completely covered) (Table 1).
Presentation of results
The results for each cancer site or group of cancers are presented in separate figures (Figures 112). All figures have the same layout. The upper part shows a bar chart ranking the crude prevalence in each country as a proportion of the population (for both sexes combined or one sex only, depending on the cancer) and by time from diagnosis (2, 5, 10, 15 years and total prevalence). To facilitate the interpretation of prevalence differences between countries, bar charts show world age-standardised incidence. Age-standardised 5-year relative survival and world age-standardised total prevalence are shown in the lower half of each figure. These world age-standardised figures are directly comparable with those produced in the USA [9]. Note that the x axis scales on these charts vary with cancer site. The data in the bar charts, in the lower part of each figure, are given in rank order of the world age-standardised total prevalence. The incidence data are from Cancer Incidence in Five Continents Volume VII [4] and the survival rates are from EUROCARE-2 [22]. Appendices A and B show all the results used to compile the bar charts in Figures 112.
|
|
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
|
|
|
|
|
|
|
|
|
|
For all cancer sites combined, prevalence in women was higher than in men in all countries. The proportion of prevalent cancer cases that were women ranged from 53% in Spain to 71% in Poland, with a weighted European mean of 61%.
Age
In all countries, the largest proportion of all prevalent cases were aged 65 years of age or over, and in most countries (the exceptions being Poland and Slovenia) such cases formed >50% of the prevalent population. The European weighted mean indicated that 57% of prevalent cases were 65 years of age or over.
Time from diagnosis
About 22% of all prevalent cases consisted of patients who had a cancer diagnosis within 2 years of the index date. This proportion did not vary greatly between countries, ranging from 19% in Slovakia to 25% in France.
Prevalence by site
Female breast cancer had the highest prevalence in all countries, and accounted for about 34% of the total prevalence in women in Europe. Colorectal cancer ranked second in females and first in males, accounting for 10% and 15% of the total female and male prevalence, respectively. In men, prostate cancer accounted for 12% and lung cancer for 10% of the total prevalence.
Incidence, prevalence and survival
Figure 13 shows the age-adjusted total prevalence for all malignancies grouped by country, plotted against the age-adjusted incidence for those registries. Each country group is represented on the plot by discs of diameter proportional to the estimated relative survival. The overall cancer prevalence correlated significantly with incidence when the CRs were grouped by country (R = 0.73, P <0.01). Thus, high prevalence was associated with high incidence, and low prevalence with low incidence. A few countries, however, did not adhere to this general pattern; in particular, Sweden had one of the two highest prevalence figures, but a relatively low incidence, and Poland had lower prevalence than expected from its incidence data. Sweden had the highest level of survival (the largest circle in Figure 13), while Poland had the lowest level of survival (the smallest circle in Figure 13). Countries with high levels of survival (large circles in Figure 13) tended to have high levels of prevalence. The ratio of prevalence to incidence, whose weighted European mean value was 5.2, ranged from 3.5 in Polish CR areas to 7.1 in south Sweden CR, but 11 of the countries had values for this ratio in the range 4.85.8.
|
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The percentage of cases lost to follow-up varied between registry areas, thus potentially affecting the comparability of the prevalence data. This bias was taken into account by assigning the same survival probability to the cases lost in each registry as that of the cases successfully followed. A more serious source of bias is that due to the different times that the registries have been operating. Registries cannot include cases diagnosed before they came into existence and recently established registries therefore have shorter series of incident cases contributing to the prevalence than long-standing registries. We corrected for this bias using completeness indices [20], which make it possible to estimate the unobserved part of the prevalence.
The effects of migration and DCO cases on the prevalence figures were also examined and it was concluded that in no case could these explain more than an insubstantial fraction of the large geographical differences found in prevalence [21]. CRs were asked to check that our prevalence estimates were consistent with their expectations and all their suggestions were considered in order to reduce errors and inconsistencies to a minimum.
Screening and early diagnoses have a variable impact on prevalence. Breast cancer screening anticipates diagnosis: a rise in incidence therefore follows the onset of screening, even in the absence of a time trend, but this lasts only for a few years, after which incidence (except in the youngest women being screened) returns to pre-screening levels, provided no over-diagnoses are introduced. Although breast cancer screening was adopted earlier in some northern European countries and the UK, it was widely implemented in Europe only during the 1990s and mainly after the index date of the present study (31 December 1992). Thus, the expected modest increase in breast cancer prevalence due to screening is not evident in our data.
The widespread use of endoscopy to detect colorectal cancer may be expected, eventually, to lower both the incidence and prevalence of this cancer as it leads to the removal of pre-cancerous lesions. A similar phenomenon was observed following the widespread adoption of cervical screening.
In general, only a small proportion of incident cases is detected by screening, even in areas where screening is well established. This is changing dramatically, however, as the prostate-specific antigen (PSA) assay for the early diagnosis of prostate cancer is being adopted. A notable fraction of new cases detected by PSA are cancers that would never have become clinically symptomatic. This inflates the incidence and consequently the prevalence of prostate cancer. The very large variation in the prevalence of prostate cancer found in this studya 13-fold difference between the highest and the lowestcan be interpreted as due to the differential spread of PSA testing across European countries.
It is important to break down cancer prevalence figures according to time since diagnosis, thereby providing more precise indications of health care needs for specific sections of the population. Cases diagnosed in the 2 years before the reference date are likely to be still undergoing primary treatment for their cancer or suffering from its side effects. The group of prevalent patients diagnosed 25 years prior to the reference date is at high risk for recurrence and should be followed closely. The 510 year prevalence group consists of patients who can be considered cured of their disease (particularly for cancers of colon, rectum and stomach) and in whom the probability of recurrence is low. However, for patients diagnosed 510 years previously, continuing but less intense follow-up is sometimes recommended. Lastly, prevalent patients diagnosed 10 years previously can be considered cured and will make minor cancer-related demands on health care services.
A possible future development would be to classify prevalent cases into four groups: recently diagnosed patients who are receiving primary treatment; those who can be considered cured of their cancer; those in the terminal phase of their illness; and the remainder of intermediate status, also referred to as those in the continuing phase [24]. The definition of these groups requires the availability of population-based information on cancer stage at diagnosis and clinical follow-up. Such groups are much more homogeneous in terms of predictable health needs than those defined solely by time since diagnosis.
We found that high cancer prevalence was associated with low general and infant mortality and with high gross domestic product and high total expenditure on health. These associations suggest that cancer prevalence will rise as the level of economic development rises. In countries with well-developed economies, general mortality is falling, life expectancy is increasing and the age distribution of the population is shifting towards the elderly. Because the incidence of almost all cancers rises steeply with age, the number of cancer cases is increasing, while major investment in early detection and treatment contributes to the longer survival of cancer patients [5]. All these factors result in higher cancer prevalence.
Thus cancer prevalence is an indicator of both the positive and negative aspects of economic development: increasing life expectancy and survival from cancer on the one hand, and increasing cancer incidence on the other. This in turn suggests that, although notable results have been achieved, the campaign against cancer in Europe has not concentrated sufficient energy or resources on primary prevention. Perhaps primary prevention should now take a much more prominent role in the battle against cancer.
![]() |
Acknowledgements |
---|
EUROPREVAL Working Group. Austria: W. Oberaigner, Cancer Registry of Tyrol. Denmark: H. Storm & G. Engholm, Danish Cancer Society, Institute of Cancer Epidemiology. Estonia: T. Aareleid, Estonian Cancer Registry. Finland: T. Hakulinen, Finnish Cancer Registry. France: G. Hédelin, Bas-Rhin Cancer Registry; H. Lefevre, Calvados Digestive Cancer Registry; J. Mace-Lesech, Calvados General Cancer Registry; J. Faivre, Côte dOr Digestive Cancer Registry; G. Chaplain, Côte dOr Gynaecologic Cancer Registry; P.M. Carli, Côte dOr Malignant Haemopathies Registry; P. Arveux, Doubs Cancer Registry; J. Estève, University of Lyon; M. Colonna, Isère Cancer Registry; N. Raverdy & P. Jun, Somme Cancer Registry. Germany: J. Michaelis, German Registry of Childhood Malignancies; H. Ziegler & C. Stegmaier, Saarland Cancer Registry. Iceland: H. Tulinius, Icelandic Cancer Registry. Italy: R. Capocaccia, Project Leader; I. Corazziari, R. De Angelis, S. Francisci, S. Hartley, F. Valente, A. Verdecchia & A. Zappone, National Institute of Health, Rome; F. Berrino, G. Gatta, A. Micheli, E. Mugno & M. Sant, National Institute for the Study and Cure of Tumors, Milan; P. Crosignani, Lombardy Cancer Registry; E. Conti, Latina Cancer Registry; M. Vercelli, C. Casella & A. Puppo, Liguria Cancer Registry, NCI, Genova; M. Federico, Modena Cancer Registry; M. Ponz De Leon, Modena Colorectal Cancer Registry; V. De Lisi, Parma Cancer Registry; R. Zanetti, Piedmont Cancer Registry; C. Magnani, Piedmont Childhood Cancer Registry; L. Gafà, Ragusa Cancer Registry; F. Falcini, Romagna Cancer Registry; E. Paci & E. Crocetti, Tuscany Cancer Registry; S. Guzzinati, Venetian Cancer Registry. Poland: J. Rachtan, Cracow Cancer Registry; M. Bielska-Lasota, Warsaw Cancer Registry. Slovakia: I. Plesko, National Cancer Registry of Slovakia. Slovenia: V. Pompe-Kirn, Cancer Registry of Slovenia. Spain: I. Izarzugaza, Basque Country Cancer Registry; A. Izquierdo, Girona Cancer Registry; C. Martinez-Garcia, Granada Cancer Registry; I. Garau, Mallorca Cancer Registry; E. Ardanaz & C. Moreno, Navarra Cancer Registry; J. Galceran, Tarragona Cancer Registry; V. Moreno, Catalan Institute of Oncologia. Sweden: T. Möller & H. Anderson, Southern Swedish Regional Tumour Registry. Switzerland: J. Torhorst, Basel Cancer Registry; C. Bouchardy, J.M. Lutz & M. Usel, Geneva Cancer Registry; J.E. Dowd, WHO, Geneva. The Netherlands: J.W.W. Coebergh & M. Janssen-Heijnen, Eindhoven Cancer Registry; R.A.M. Damuhis, Rotterdam Cancer Registry. Scotland: R. Black, V. Harris & D. Stockton, Scottish Cancer Intelligence Unit. United Kingdom: T.W. Davies, East Anglian Cancer Registry; M.P. Coleman & S. Harris, London School of Hygiene and Tropical Medicine; E.M.I. Williams, The Merseyside and Cheshire Cancer Registry; D. Forman & R. Iddenden, Northern and Yorkshire Cancer Registry and Information Service and Centre for Cancer Research; M.J. Quinn, Office for National Statistics; M. Roche, Oxford Cancer Intelligence Unit; J. Smith, South and West Cancer Intelligence Unit; H. Moller, Thames Cancer Registry; P. Silcocks, Trent Cancer Registry; G. Lawrence & K. Hemmings, West Midlands Cancer Intelligence Unit.
![]() |
Footnotes |
---|
Members of the EUROPREVAL Working Group are listed after the Acknowledgements.
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
2. Muir CS, Waterhouse J, Mack T et al. (eds). Cancer Incidence in Five Continents, Vol. V. IARC Scientific Publications No. 88. Lyon, France: International Agency for Research on Cancer, 1987.
3. Parkin DM, Muir CS, Whelan SL et al. (eds). Cancer Incidence in Five Continents, Vol VI. IARC Scientific Publications No. 120. Lyon, France: International Agency for Research on Cancer, 1994.
4. Parkin DM, Whelan S, Ferlay J et al. (eds). Cancer Incidence in Five Continents, Vol. VII. IARC Scientific Publications No. 143. Lyon, France: International Agency for Research on Cancer, 1997.
5. Estève J, De Angelis G, Verdecchia A. Trends in cancer survival probability over the period 197889. In Berrino F, Capocaccia R, Estève J et al. (eds): Survival of Cancer Patients in Europe. The EUROCARE-2 Study. IARC Scientific Publications No. 151. Lyon, France: International Agency for Research on Cancer, 1999.
6. Sant M and the EUROCARE working group. Overview of EUROCARE-2 results on survival of cancer patients diagnosed in 198589. In Berrino F, Capocaccia R, Estève J et al. (eds): Survival of Cancer Patients in Europe. The EUROCARE-2 Study. IARC Scientific Publications No. 151. Lyon, France: International Agency for Research on Cancer, 1999.
7. Micheli A, De Angelis G, Giorgi Rossi A et al. General mortality and the survival of cancer patients in Europe. In Berrino F, Capocaccia R, Estève J et al. (eds): Survival of Cancer Patients in Europe. The EUROCARE-2 Study. IARC Scientific Publications No. 151. Lyon, France: International Agency for Research on Cancer, 1999.
8. Zanetti R, Micheli A, Rosso S, Sant M. The prevalence of cancer: a review of the available data. Tumori 1999; 85: 408413.[ISI][Medline]
9. Feldman AR, Kessler L, Myers MH, Naughton MD. The prevalence of cancer. Estimates based on the Connecticut Tumor Registry. N Engl J Med 1986; 315: 13941397.[Abstract]
10. Adami HO, Gunnarson T, Sparén P, Eklund G. The prevalence of cancer in Sweden 1984. Acta Oncol 1989; 28: 463470.[ISI][Medline]
11. Hakama M, Hakulinen T, Teppo L, Saxen E. Incidence, mortality or prevalence as indicators of the cancer problem. Cancer 1975; 36: 22272231.[ISI][Medline]
12. Tulinius H, Storm HH, Pukkala E et al. Cancer in the Nordic countries, 198186. A joint publication of the five Nordic Cancer Registries. APMIS 1992; 31 (Suppl): 146S170S.
13. Micheli A, Francisci S, Krogh V et al. Cancer prevalence in Italian cancer registry areas: the ITAPREVAL study. ITAPREVAL Working Group. Tumori 1999, 85: 309369.[ISI][Medline]
14. Quinn MJ, Babb PJ, Brock A et al. Cancer Trends in England and Wales 19501999. Studies on Medical and Population Subjects No. 66. London, UK: The Stationery Office, 2001.
15. Capocaccia R, Gatta G, Chessa F et al. The EUROCARE-2 study. In Berrino F, Capocaccia R, Estève J et al. (eds): Survival of Cancer Patients in Europe. The EUROCARE-2 Study. IARC Scientific Publications No. 151. Lyon, France: International Agency for Research on Cancer, 1999.
16. World Health Organization. Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death. Geneva, Switzerland: World Health Organization, 1977.
17. World Health Organization. International Classification of Diseases for Oncology (ICD-O). Geneva, Switzerland: World Health Organization, 1976.
18. Sant M, Gatta G. The EUROCARE Database. In Berrino F, Sant M, Verdecchia A et al. (eds): Survival of cancer patients in Europe. The EUROCARE Study. IARC Scientific Publications No. 132. Lyon, France: International Agency for Research on Cancer, 1995.
19. Krogh V, Micheli A. Measure of cancer prevalence with a computerized program: an example on larynx cancer. Tumori 1996; 82: 287290.[ISI][Medline]
20. Capocaccia R, De Angelis R. Estimating the completeness of prevalence based on cancer registry data. Stat Med 1997; 16: 425440.[ISI][Medline]
21.
Capocaccia R, Colonna M, Corazzieri I et al. Measuring cancer prevalence in Europe: the EUROPREVAL project. Ann Oncol 2002; 13: 831839.
22. Berrino F, Capocaccia R, Estève J et al. (eds): Survival of Cancer Patients in Europe. The EUROCARE-2 Study. IARC Scientific Publications No. 151. Lyon, France: International Agency for Research on Cancer, 1999.
23. OECD HEALTH DATA 98. A Comparative Analysis of 29 Countries. CD-ROM. Paris, France: OECD, Health Policy Unit, 1998.
24. Brown ML, Riley GF, Potosky AL, Etzioni RD. Obtaining long-term disease specific costs of care: application to Medicare enrollees diagnosed with colorectal cancer. Med Care 1999; 37: 12491259.[ISI][Medline]