The National Cancer Registry of Cuba has achieved three survival studies with cases incident in 1976,. 1982 and 1988±1989. This article deals with the global.
European Journal of Epidemiology 15: 521±528, 1999. Ó 1999 Kluwer Academic Publishers. Printed in the Netherlands.
Trends in survival rates of cancer in Cuba Margarita C. Graupera Boschmonar1, Pedro J. JimeÂnez Chaviano2, Antonio A. MartõÂ n GarcõÂ a1, Yaima H. GalaÂn Alvarez1, Leticia M. FernaÂndez Garrote1 & R. Sankaranarayanan3 1
National Cancer Registry, and 2National Cancer Control Program, National Institute of Oncology and Radiobiology, Havana City, Cuba; 3Unit of Descriptive Epidemiology, International Agency for Research on Cancer, Lyon, France
Accepted in revised form 27 March 1999
Abstract. Reports of population-based survival rates of cancer from developing countries are infrequent. In Latin America, only the Cancer Registry of Puerto Rico has published population-based survival data. The National Cancer Registry of Cuba has achieved three survival studies with cases incident in 1976, 1982 and 1988±1989. This article deals with the global observed and relative survival rates estimated in the latter study. Survival time trends are analysed. In the period 1988±1989, 12,985 primary cancer cases were included from the most common cancer sites, with the exclusion of cancer in situ cases and 8900 cases reported by `death certi®cate only' (DCO) (35.8%). The vital status of cases was checked up to 31 December 1994 using a mixed follow-up system with the exclusion of 2900 cases lost to follow-up (11.2%). DCO proportions are shown for the major sites and compared to those of 1982. Observed survival rates were estimated by Kaplan±Meier method using the
SPSS Statistical Software. The relative rates were estimated by the Hakulinen's Computer Program Package for Cancer Survival Studies (1988) using life tables from Cuban population. Statistical comparisons of survival curves by year of diagnosis were achieved by using the Log±Rank and Pearson statistic tests. Global results are shown by year of follow-up and a comparative analysis is done in time and with internationals values. Survival rates decreased in the period 1982/1988±1989 for colon, prostate and lung cancer. Prostate cancer shows a low ®ve years relative survival rate when compared with the USA, but its observed rate is comparable with Puerto Rico's. Figures for mouth and lung cancer were comparable with the corresponding ®gures of the USA and Europe. Breast and cervix cancers rates are comparable with the European mean and the blacks in USA.
Key words: Population-based cancer registry, Survival rates, Time trends Introduction The prevention of transmissible diseases and the elevation of life expectancy have caused the ageing of the population in Cuba with an increment in the incidence and mortality of chronic nontransmissible diseases [1]. Cancer control activities and their systematic evaluation started in Cuba since the 1960s with the foundation of the National Cancer Registry (NCR) in 1964 and the National Program for Cervix Cancer Screening in 1968. The NCR is a population-based nation-wide cancer registry and its foundation was a pioneer step in Latin America; only two cancer registries of this type were established by the time (Puerto Rico (1950) and Chile (1959)) [2]. Since then, it receives information of all new cancer cases diagnosed in Cuba, its major objective being the description of the cancer pro®le in Cuba by the annual estimation of crude and adjusted incidence rates and the incorporation of data from the National Mortality Statistics for the emission of mortality rates. Both rates are important indexes for cancer control evaluation [3].
Survival analysis reports are scarce for developing countries due to the diculties that the follow-up of the cases implie [4]. Cuba counts with an optimum system for death certi®cates as well as a good citizens registry, both necessary for this activity [5, 6]. The ®rst eorts in population-based survival analysis started in 1988 with the estimation of survival rates of cases that were reported to the NCR in 1976. In 1989, another study was accomplished with cases diagnosed in 1982 [7]. These were some of the ®rst steps towards the attainment of this valuable index for the evaluation of the global eectiveness of cancer care in Cuba, even in Latin America [8, 9] In 1994 a new survival study was initiated in collaboration with the International Agency for Research on Cancer (IARC); it was part of a Multinational Study named Cancer Survival in Developing Countries in which the only participant from Latin America was Cuba. It had the aims of estimating the observed and relative population-based survival rates for cases incident to the NCR in 1988±1989 and to validate a more eective follow-up system to be
522 implemented in the NCR for the regular estimation of these rates [10]. This paper deals with some of the global results obtained in the latter study and the possible explanations of some of the survival variations with time in Cuba as well as those observed when international comparisons are accomplished. Materials and methods The study includes 12,985 primary cancer cases incident to the NCR in the period 1988±1989 in the following sites: tongue, oral cavity, oropharinx, colon, rectum, lung, female breast, leukaemias, cervix and corpus uteri, prostate and Hodgkin's disease. The in situ cases and those reported by `death certi®cate only' (DCO) were excluded from the study (8900; 35.8%), as well as cases lost to follow-up (2900; 11.2%). Although, by the time of the survival study, the tracing back of cases informed by death certi®cate (DCI) was not implemented for its regular use in the NCR, a tracing back of the DCI cases was achieved by searching for diagnostic information in the original clinical reports of the patients. This procedure lowered DCO proportions. Some of its results by site are shown in Table 1. With the aim of establishing the vital status of cases up to the closing date of 31 December 1994, a mixed follow-up system was used consisting of: (a) passive follow-up: incidence ®les were matched with the ®les of the National Mortality Registry from the period 1988±1994, for the identi®cation of dead cases as well as their dates and causes of death. For this purpose a computerised record-linkage procedure (RL) was implemented. In it, the matching procedure is achieved using, in the ®rst place, the variables date of birth (®rst six digits of the identity number) and sex.
The mortality ®les are sorted by both variables, creating subsets of records with the same date of birth and sex. In this way, each record of the cancer registration ®le is searched, not in the whole mortality ®le but only in a few records in which an automated routine to compare individual names will be executed. Pairs of records with the same birthday and sex and `similar' names are stored in a third ®le for a visual review. A procedure to update the date and cause of death of the cases will complete the record-linkage procedure. (b) Active follow-up: an active follow-up method was accomplished that consisted of three steps: 1. Matching with the National Identity Registry ®les: unmatched cases resulting from the R-L were considered `apparently alive'. Their vital status was checked by co-ordination with the National Identity Registry (NIR) by the following procedures: ± Emission of lists by province and alphabetically ordered. ± Manual matching of these lists with the lists of the provincial oces of NIR. The `alive' and `emigrant' status and the date of emigration of the cases were established. All cases dead by any cause were considered as dead cases, the date of last contact being the date of death. 2. Paid-reply cards: questionnaires were sent by mail to cases who were not traced in the ®rst step, with an ethical inquiry about the vital state of the patients. 3. Search in the medical records: vital status of cases not traced by the last step were searched in their hospital of attendance, using the clinical record of the patient as a source. The outcome of the follow-up of cases are shown in Table 1. Before survival analysis was accomplished, internal validation was achieved by checking the following items:
Table 1. Results of the follow-up of cases and the tracing-back procedures Site
Mouth (140±46) Colon Rectum Lung Leukemia Breast Cervix uteri Corpus uteri Lymphomas Prostate
Cases lost to follow-up
DCO cases
No.
%
No.
%
%a
TMSb
231 205 141 513 188 459 332 97 163 476
16.0 8.6 12.8 7.4 12.7 13.1 16.2 17.8 13.3 13.6
210 1094 248 3600 714 664 176 45 385 1615
14.6 45.7 22.6 52.2 48.3 18.9 8.6 8.2 31.4 46.3
19.0 13.0 16.9 12.6 11.2 12.3 20.9 16.2 13.1 8.3
387.0 20.2 200.4 611.8 201.6 651.7 252.8 328.2 31.9 148.1
DCO: cases reported by death certi®cate only. DCI: cases informed by death certi®cate. a %: percent of the initial DCI found. b TMS: mean survival time (DAYS).
Traced back DCI cases
523 ± Validation of the data consistence by the CHECKLIST (IARC) Program [11]. Possible errors were checked against the original reports. ± Age at diagnosis was recalculated from date of diagnosis and date of birth and corrected in the incidence data set. ± Survival time was checked with the recti®cation of cases with zero or negative values. Data were processed to estimate survival. The estimation of observed survival rates was made by the Kaplan±Meier method [12] using the SPSS Statistical Software. The estimation of relative survival rates was made by the Computer Program Package for Cancer Survival Studies developed by Hakulinen et al. (1988) [13]. These rates are the ratio of the observed survival rate to the expected survival rate. The latter was obtained from the Cuban general population life-table for a group of people similar to the patient group with respect to sex, age and calendar period. Survival rates were calculated for the major individual cancer sites speci®ed above. Time trends of the observed and relative survival rates resulting from the three studies were analyzed. In the study of cases reported in 1976 only ®ve major sites were analysed and the survival data set was not available, not permitting an statistical comparison of the survival curves. For the survival study including cases incident to the NCR in 1982, information and survival data set for all sites were available. Comparative analysis of curves for survival data for selected cancer sites for cases diagnosed in 1982 and 1988±1989 was achieved using the SPSS statistic package testing the equality of survival distributions by Kaplan±Meier using the Log±Rank statistics test. Comparison of relative survival curves was achieved by the Hakulinen software using the Pearson statistics test.
Global survival rates for some sites of the 1988± 1989 study were compared with rates published in international reports by other cancer registries. Results Global observed and relative survival rates by one, three and ®ve years of follow-up for the sites in study are show in Table 2. As it is observed, only breast and corpus uteri cancers have ®ve-years relative rates higher than 60% and a rate lower than 20% is only shown for lung cancer. Five-years survival rates estimated in the three population-based survival studies are shown in Table 3. For mouth cancer (ICD-O 140±146) an absolute dierence of 6% is observed between the relative ®ve-years survival rates of 1982 and 1988±1989, but a comparative analysis of its observed survival curves showed statistically non-signi®cant dierences (p > 0.1). This was also noted when comparing relative survival curves (Figure 1). In relation to colon and prostate it can be appreciated, for the ®rst period (1976±1982), a 4% and 8% absolute raise in their respectively observed survival rates. Nevertheless, in the second period (1982±1988) decreases of 5% and 18%, respectively, are noted in relative and observed rates and a comparative analysis of their observed survival curves showed statistically signi®cant dierences (p < 0.05 and p < 0.001, respectively). Moreover, when comparing the relative survival distribution (Figures 2 and 3) a similar result was achieved. When comparing rectum cancer data for the 1982 and 1988±1989 studies, 10% absolute increase in its ®ve-years relative survival rates is observed, but no
Table 2. Overall observed and relative survival rates and frequencies by site for both sexes for cases incident to the National Cancer Registry of Cuba in 1988±1989 Site
Mouth (140±146) Tongue Oral cavity Oropharinx Colon Rectum Lung Leukemias Female breast Cervix uteri Corpus uteri Prostate Hodgkin's disease a
Number of cases
994 232 385 109 1087 707 2775 508 2371 1533 404 1388 279
Survival rates (%) at year of follow-up 1 year
2 year
Observed
Relative
67 48 67 65 46 58 23 40 80 74 77 61 68
74 52 74 70 51 63 25 41 83 76 81 72 70
a
3 year
Observed
Relative
47 23 44 32 35 37 10 24 64 57 59 39 56
55 28 52 38 42 45 12 27 69 59 66 55 59
a
Observed
Relative
42 19 39 27 29 31 8 19 54 52 52 26 50
54 25 50 34 38 44 11 22 61 56 62 45 54
Relative survival rates were estimated by the Computer Program Package for Cancer Survival Studies developed by Hakulinen et al. (1988) [13].
524 Table 3. Time trends of survival rates in Cuba for selected sites. Both sexes (1976/1982/1988±1989) Site
Five-years survival rates (%) by year of diagnosis 1976a
Mouth (140±146) Colon Rectum Lung Breast Cervix uteri Corpus uteri Prostate
1982
1988±1989
Observed
Relativea
Observed
Relative
Observed
Relative
) 30 ) 15 43 43 ) 36
) ) ) ) ) ) ) )
37 34 27 11 54 44 56 44
48 43 34 14 60 47 69 67
42 29* 31 8*** 54 52 52 26***
54 38* 44 11*** 61 56 62 45***
a
In this study relative survival rates were not estimated. * Statistical signi®cant dierence. *** Highly statistical signi®cant dierence.
signi®cant increase (p > 0.1) was noted in the comparative analysis of observed as well as relative survival distribution (Figure 4). Nevertheless, a better prognosis is shown in the last years of follow-up for 1988±1989 cases. Regarding lung cancer, a progressive decrease in the ®ve-years survival rates for both periods is observed in Table 3. Also, a statistically signi®cant decrease (p < 0.001) appears when comparing survival curves for the last period. Results are shown in Figure 5. For breast cancer an important rise is found in the ®rst period (Table 3), that is very slight in the second.
Moreover, when the comparative analysis of survival curves for data from 1982 and 1988±1989 was attained, a statistically non-signi®cant change was observed (p > 0.1). In Figure 6 relative survival curves are shown, with lower survival rates for the latter in the ®rst years of follow-up. For cervix cancer an inverse situation is observed with a slight rise in the ®rst period and an absolute dierence of 9% in the second between the ®ve-years relative survival rates (Table 3). Nevertheless, no statistically signi®cant dierence is observed when comparing survival curves (p > 0.1); relative ones are shown in Figure 7.
Figure 1. Relative survival curves by year of diagnosis. Mouth cancer. Both sexes, 1982/1988±1989.
Figure 2. Relative survival curves by year of diagnosis. Colon cancer. Both sexes, 1982/1988±1989.
525
Figure 3. Relative survival curves by year of diagnosis. Prostate cancer, 1982/1988±1989.
Figure 5. Relative survival curves by year of diagnosis. Lung cancer. Both sexes, 1982/1988±1989.
Figure 4. Relative survival curves by year of diagnosis. Rectum cancer. Both sexes, 1982/1988±1989.
Figure 6. Relative survival curves by year of diagnosis. Female breast cancer, 1982/1988±1989.
Discussion The observed dierences shown in Table 3 for mouth cancer might be interpreted at ®rst sight as a result of the implementation in 1984 of the National Screening Program for the early diagnosis of bucal cancer with a greater surveillance of pre-malignant diseases and
the diagnosis of malignant lesions in lower stages as well as of slow-growing tumours prevalent at the moment. Nevertheless, when a comparative analysis of survival curves was attained, this dierence resulted non-signi®cant (Figure 1). Moreover, in 1988± 1989 there is a slight diminution of the rates in the
526
Figure 7. Relative survival curves by year of diagnosis. Cervix cancer, 1982/1988±1989.
®rst years of follow-up that might indicate problems in the diagnosis. Related to the great decrease in survival rates for prostate cancer (Figure 2) in the last period, it is necessary to recall the great diculties that prostate cancer survival comparisons have in relation to the uncertainty of the de®nition of diagnosis [14, 15], as well as the high prevalence of the slow-growing focal prostatic carcinoma that might be diagnosed by incidental biopsy of a benign hyperplasm surgery or by an autopsy ®nding. Moreover, in Cuba there had been disagreements among physicians in relation to the diagnosis and treatment of this cancer and these may have produced a lack of uniformity. The implementation of rectal digital diagnosis by the Family doctor in all males aged more that 50 years started in the 1980s, although it did not have a good response. Nevertheless, it seems probable that a more active attitude towards diagnosis of prostate cancer may have produced a rise in the use of surgical procedures. The surgical treatment of non-malignant entities could have enabled occasional ®ndings of latent slow-growing tumours favouring a rise in survival rates of 1982 data. It is not until 1989 that a very severe blockade started, and may have limited the possibilities of prolonged hormonotherapy treatments. This decrease in survival is supported by the fact that mortality trends changed from a negative one in the period 1976±1981 to a positive, increasing trend in the following periods (1982±1987 and 1988±1994), with a 1% annual increase in the latter. Also, a comparatively slight increase in incidence is observed in the
period 1977±1994. The latter fact makes us think that the decrease in survival is not merely a consequence of the 10% variation in DCO proportions between the periods (Table 4). In relation with colon cancer, great improvements in diagnostic procedures were introduced since 1979 with the use of the colonoscopy. Consequently, the polipectomy was achieved more frequently and initially a great number of slow-growing tumours, prevalent in the population at the moment, were probably diagnosed with a rise in the survival rates. The frequency of this tumour could have tended to decrease in time, producing a decrease of the survival rates in the cases diagnosed in 1988±1989. Nevertheless, this variation in survival rates with time may be due to the nearly 20% reduction in DCO proportions between the compared periods (Table 4). This is supported by the fact that mortality had a low steady rise (0.1% annually) in the period 1982±1995. The implementation in Cuba of new surgical techniques for rectal cancer treatment in the 1980s could be a possible explanation for the improvement in prognosis in the last years of follow-up for patients diagnosed in 1988±1989 with rectal cancer in comparison to those diagnosed in 1982 (Figure 4). This could probably be due to the decrease in the relapses and/or post-operatory problems. The decrease in survival rates for lung cancer in the whole period may be explained by the fact that this site has presented one of the highest proportions of DCO cases. Since 1976 these proportions decreased slowly, from more than 60% in 1976 (Table 3) to 52.2% in 1988±1989. Also, no signi®cant variations have been observed in mortality trends for lung cancer in the whole period. The behaviour of breast cancer time trends when comparing the ®rst two studies (Table 3) presents particularities that are probably due to the creation of services for breast cancer attendance all along the country at the end of the 1970s. These have probably increased the availability of breast cancer diagnosis to the population, with a favourable in¯uence in the early detection of the disease. Also, advanced treatment procedures were applied in this decade with the introduction in 1980 of the adjuvant and neo-adjuvant Table 4. Proportions of excluded DCO cases in the survival studies of cases incident in 1982 and 1988±1989 Site
Mouth (140±146) Colon Rectum Lung Breast Cervix uteri Prostate
Percent of cases (%) 1982
1988±1989
23 64 28 61 24 12 56
15 46 23 52 19 9 46
527 Table 5. International comparison for selected cancer sites. Both sexes. Cuba/USA/Europe Site
Mouth (140±146) Lung Prostate Colon Breast Cervix
Five-years relative survival rates (%) Cuba (1988±1989)
USA (1980±1985)a White
Black
54 11 45 38 61 56
54 13 73 55 76 67
31 12 63 19 64 59
Europeb (1978±1985) ) 8 55c 42 67 59
a
SEER (Surveillance, Epidemiology and End Results Program) [15]. Mean of European rates estimated in Eurocare International Survival Study [16]. c EUROCARE-2 Study [17]. b
chemotherapy. Both facts may have produced an elevation of survival rates in 1982. In the last period, the ®ve-years survival rates do not dier notably (Table 3), but the analysis of survival curves (Figure 6) showed a considerable decrease in the survival rates in the ®rst years of follow-up for the cases diagnosed in 1988±1989. At the last years of follow-up a stabilization is observed pointing probably to a normalization of breast cancer care in Cuba. Time trends in cervix cancer may be explained by the introduction in Cuba in the 1990s of a better diagnostic method, the colposcopy, permitting a more speci®c diagnosis that favoured the application of more adequate treatments. However, the National Program of Cervix Cancer Screening started since 1967 has not produced the expected rise in survival, due perhaps to de®ciencies in its coverage. The fact that survival curves for the last two studies (Figure 7) did not present signi®cative dierences may probably be due to the non-paralelism of the curves, but the last years of follow-up showed higher values for the 1988±1989 cases pointing to a better achievement in the treatment of this cancer. Moreover, an eort in treatment standardisation was attained in the country in the 1980s with the editing of oncology guide-lines despite the fact that cases are frequently handled by other specialists, having its in¯uence in prognosis. International comparisons showed comparable ®gures for the results of the 1988±1989 study (Table 5) instead of the highly in¯ated survival rates that could be expected due to the exclusion of high proportions of DCO cases [14]. Prostate cancer shows a very low ®gure, compared with the USA. Nevertheless, the ®ve-year observed survival rate (45%) is comparable with the one reported by Puerto Rico (44%) for the period 1980±1985 [9]. In mouth and lung cancer, comparable values with the USA and Europe can be appreciated. Breast and cervix cancers are comparable with the European mean and the blacks in the USA. In fact, some European countries have published lower survival values for cervix cancer with relative ®ve-years survival rates of 41% for
Spain and 54% for Scotland and Poland. These countries have a low availability for the diagnosis and treatment facilities, as reported elsewhere [14]. Fiveyears observed survival rates for cervix cancer (52%) are only slightly lower than Puerto Rico's (54%), but for breast cancer, the observed ®gure (54%) is an absolute 7% lower than Puerto Rico's 61% value. Colon is comparable with the European mean and possibly with that of the whole population in the USA. Maybe, the exclusion of DCO cases has not resulted in a serious deviation, as it appears, because these excluded cases may represent for some sites a sample of all the registered cases. This can be assumed from the high values of mean survival time (see Table 1) for traced-back DCI cases of mouth, breast and lung cancer, which are unlikely to represent the poor prognostic cancer cases. The NCR is actually working on the improvement of its coverage and the regular tracing back of the DCI cases, which will result in a better interpretation of survival and incidence data from Cuba. References 1. Annual statistics 1996. Cuba: National Statistics Direction, Health Ministry, 1997. 2. Parkin DM, Wagner G, Muir C, eds. IARC-The role of the cancer registry in cancer control. IARC Scienti®c Publications No. 66 Lyon: International Agency for Research on Cancer, 1985. 3. Jensen OM, Storm HH. Aims and uses of cancer registries. In: Jensen OM, Parkin DM, Mac Lennan R, Muir CS and Skeet RG (eds), Cancer registration. Principles and methods. IARC Scienti®c Publications No. 95. Lyon: International Agency for Research on Cancer, 1991: 7±23. 4. Navarro SaÂnchez C, LizaÂn GarcõÂ a M, Tormo DõÂ az MJ. Uses of death certi®cates in population-based cancer registries. Gaceta Sanitaria 1988; 2(7): 197±202. 5. Rios Massabot E. Actual state of information in mortality systems in Cuba. Havana, Cuba: Health Development Institute, Health Ministry Report, 1980.
528 6. Rios Massabot E, Tejeiro Fernandez A, Mesa Machado AC. Quality of medical death certi®cation. Havana, Cuba: Direction of National Statistics, Health Ministry Report, 1991. 7. Graupera M, Abascal ME, et al. Survival rates of cancer in Cuba, 1982. Revista Brasileira de Cancerologia 1994; 40(3): 135±140. 8. Central cancer registry of Puerto Rico ± Cancer in Puerto Rico, 1987. Lybrary of Congress Cataloging in Publication Data, 1989. 9. Central cancer registry of Puerto Rico ± Cancer in Puerto Rico, 1989. Lybrary of Congress Cataloging in Publication Data, 1991. 10. FernaÂndez L, Graupera M, GalaÂn Y, Lezcano M, MartõÂ n A, Camacho R. Cancer survival in Cuba. In: Sankaranarayanan R, Black R, Parkin DM (eds), Cancer survival in developing countries. IARC Scienti®c Publications No. 145. Lyon: International Agency for Research on Cancer, 1998: 51±59. 11. Parkin DM, Chen J, Ferlay J, GalceraÂn J, Storm HH, Whelan SL. Comparability and quality control in cancer registries. IARC Technical report No. 19. Lyon: International Agency for Research on Cancer, 1995: 79±83. 12. EsteÂve J, Benhamou E, Raymond L. Statistical method in cancer research. Volume 4. Descriptive Epidemiology. IARC Scienti®c Publications No. 128. Lyon: International Agency for Research on Cancer, 1994.
13. Hakulinen T, Abeywickrama KM. A computer program package for relative analysis. Computer Programs in Biomedicine 1985; 19: 197±207. 14. Berrino F, Sant M, Verdecchia A, Capocaccia R, Hakulinen T and Esteve J, eds. IARC ± European Commission ± Survival of cancer patients in Europe. The Eurocare study. IARC Scienti®c Publications No. 132 Lyon: International Agency for Research on Cancer, 1995. 15. WHO-IARC. Prostate. In: Coleman MP, EsteÂve J, Damiecki P, Arsland A, Renard H (eds), Trends in cancer incidence and mortality. Scienti®c Publications No. 121 Lyon: International Agency for Research on Cancer, 1993. 16. SEER (Surveillance, Epidemiology and End Results Program, National Cancer Institute, Division of Cancer Prevention and Control). 1987 Annual Cancer Statistics Review, Including Cancer trends 1950±95. Bethesda: US Department of Health and Human Services, No. 88-2789, 1988. 17. Post PN, Damhuis RAM, van der Meyden APM and the EUROCARE Working Group. Variation in survival of patients with prostate cancer in Europe since 1978. Eur J Cancer 1998; 34 (14): 2226±2231.
Address for correspondence: M.C. Graupera, National Cancer Registry, National Institute of Oncology and Radiobiology, 29 y F Vedado, Habana 4, Cuba
