Trends in breast cancer incidence in greater Bombay: an epidemiological assessment.

Trends in breast cancer incidence in Greater Bombay: an epidemiological assessment B.B. Yeole,1 K. Jayant,2 & D.J. Jussawalla3 Reliable data from the Bombay CancerRegistryshowan increase in the age-adjusted incidence rate ofbreast cancer in women from 17.9 to 24.9per 100000population between 1965and 1985. Byfittinga linear regression model based on the logarithm of the breast cancer incidence rates, it was found that the average percentage increases in crude, age-adjusted and truncated rates over this 20-year period were highly significant. Evaluation of these trends in the light ofproven etiological factors suggests that the increase in breastcancer incidence is related to a gradual decrease in the proportion of women having a first child before 20 years of age and to an increase in the proportion of "never married" women. These findings were also applicable to the subgroup of Hindus (70% of the population) who show a significant increase in breast cancer over this period as well as a clear cohort effect, the younger birth cohorts in general having higher rates than the 5-year older cohorts. However, the Muslim and Christian subgroups were found to have stable rates. An earlier study of cervical cancer incidence over the same period showed stable rates among Muslims and Christians but a declining trend among Hindus. Breast cancer is now the leading cancer in women in Bombay, while cancer of the cervix uteri predominates in the rest of the country. Well-designed epidemiological studies are urgently needed to explain the phenomenon and to help control the increase in breast cancer before it assumes the magnitude observed in the developed countries. Introduction Reliable data on cancer incidence in Greater Bombay, which have been available since 1964, show significant changes, the incidence at several sites having sharply declined while others (lung and breast) have increased. The decrease in cancers of the tongue, pharynx and larynx are largely due to changes in the type oftobacco usage over the years (1); the decreasing trends in both stomach and cervical cancer arb consistent with those found in many parts of the world (2). The decline in cervical cancer has been shown to be a cohort effect (3). In recent years, breast cancer has become the leading cancer in women in Bombay; in the rest of the country it is cancer of the cervix uteri. The trends in breast cancer incidence in Greater Bombay have therefore been studied in detail and an epidemiological assessment of these trends in the light of proven etiological factors is presented here. Materials and methods Between 1964 and 1985 the Bombay Cancer Registry recorded 42 749 cancer cases in females, 7985 of which I Co-ordinating Officer, Bombay Cancer Registry, Indian Cancer Society, 74 Jerbai Wadia Road, Parel, Bombay-400012, India. Requests for reprints should be sent to this author. 2 Scientific Officer, Cancer Research Institute, Parel, Bombay, India. 3Director, Bombay Cancer Registry, Indian Cancer Society, Parel, Bombay, India. Reprint No. 5069 were breast cancers. A linear regression model based on the logarithm of the incidence rates, the method often used for studying time trends (4), was applied to the entire data. Trends in the crude, age-adjusted and age-specific incidence rates for various religious groups were examined. For calculating these rates the female population, by religion and age, was estimated from the 1961, 1971 and 1981 census reports of Greater Bombay. Since our definition of a resident population differs from the criteria used in the population census, we corrected our population estimates by eliminating all migrants who had been resident for less than one year in Bombay. Age-specific incidence curves for 5-yearly cohorts born between 1914 and 1939, were drawn by consider- ing the incidence rates for the 5-year periods 1964-68, 1969-73, 1974-78 and 1979-83. For example, for a cohort born in 1919 the incidence rates for the age groups of 45-49, 50-54, 55-59 and 60-64 years are given by the corresponding age-specific incidence rates reported for 1964-68, 1969-73, 1974-78 and 1979-83, respectively. Results The number ofnew cases as well as the crude (CR), the age-adjusted (AAR) and the truncated (TR) incidence rates per 100 000 population for breast cancer are shown in Table 1. The CR, AAR and TR were found to have increased from 12.4,22.9 and 47.7 in 1964 to 15.2, 24.9 and 56.3, respectively, in 1985. Bulletin of the World Health Organization, 66 (2): 245-249 (1990) © World Health Organization 1990 245 B.B. Yeole et al. Table 1: Number of new cases and the crude (CR), age-adjusted (AAR), and truncated (TR) Incidence rates for breast cancer, Greater Bombay, 1964-85 Year No. of cases CR AAR TR 1964 224 12.4 22.9 47.7 1965 180 9.6 17.9 38.1 1966 216 11.0 18.3 43.3 1967 210 10.3 18.4 40.4 1968 248 11.7 20.9 45.2 1969 243 11.0 11.5 43.9 1970 259 11.2 19.8 40.4 1971 259 10.8 18.3 37.2 1972 291 11.6 19.7 43.9 1973 291 11.2 18.7 48.3 1974 344 12.6 21.6 48.4 1975 352 12.4 21.3 46.7 1976 379 12.8 21.5 46.2 1977 403 13.1 22.4 48.1 1978 389 12.1 20.7 48.3 1979 441 13.2 22.8 49.4 1980 469 13.4 22.7 50.5 1981 512 14.1 23.3 51.2 1982 500 13.2 21.6 48.1 1983 524 13.2 21.3 49.2 1984 595 14.4 22.8 52.7 1985 656 15.2 24.9 56.3 The regression model utilizing data reported for 1964-68 was fitted and the rates predicted for 1975, by extrapolating directly, were CR = 12.2, AAR = 20.9, and TR= 46.5. It was found that there was no signifi- cant difference between the observed and estimated rates. Having thus established the goodness of fit, the regression model was fitted for the entire data (1964- 85). Trends in various incidence rates are shown in Fig. 1. It will be observed that all three rates (CR, AAR and TR) present an increasing trend. Estimates for the Fig. 1. Trends In incidence rates per 100 000 population for breast cancer In Greater Bombay, 1964-85.' 660r .*. D 30 0 c 20 .7;i0C 10 1.K. _. 0 * 0 . _ . .@ . .A.A.R. C.R. * - - ~ ~~~ I Table 2: Estimates from regression analysis of rates by age for 1964 and 1985, and the average percentage change for breast cancer In Greater Bombay Estimated rate Average percentage 1964 1985 change t-test CR' 10.4 14.3 +1.52 7.38b AAR" 18.8 23.2 +1.00 4.34b TR" 40.5 52.6 +1.25 5.49b Age group (years): 25-34 5.1 7.3 + 1.78 2.85C 35-44 24.6 25.0 +0.07 0.16 45-54 45.0 63.6 + 1.67 4.10b 55-64 57.5 78.4 +1.49 3.95b > 65 76.5 82.5 0.36 0.68 CR=crude rate; AAR=age-adjusted rate; TR=truncated rate. b Significant at P < 0.001 level. c Significant at P < 0.01 level. average percentage change (APC) in CR, AAR and TR obtained from regression analyses are shown in Table 2. The crude rates seem to have increased maximally, i.e., on an average by 1.52% over the study period. The truncated rates, which are supposed to be more reliable, increased by 1.25%. However, the AAR increased by only 1.00%. The average percentage increase seen in each of the estimated rates is highly significant (P <0.001). When the trends were examined by 10-year age groups (25-34, 35-44, 45-54, 55-64, and > 65 years) the increase was observed in almost all groups (Fig. 2). Fig. 2. Trends In age-specific Incidence rates for breast cancer In Greater Bombay, 1964-85. 100 80 60 40 co 00. 02 C> 0) 8 4 000 00o 6 + 0*o 0 0 v 0 55 -64 5 - 5 4 054 0 x x x o x x x Ko 35_44 0x * -25-34 * x * 0 *= 0 0 .0 64 68 72 76 80 84 86 Year of diagnosis WHO Bulletin OMS. Vol 68 1990. 64 68 72 76 80 84 86 Year of diagnosis ' T.R. = truncated rates; A.A.R. = age-adjusted rates; C.R. = crude rates. 4^. _ . . . 50 40 2" Trends In breast cancer Incidence In Greater Bombay The percentage increase was 1.67% for the age group 45-54 years and 1.49% for the age group 55-64 years, which were highly significant (P <0.001), while the maximum increase of 1.78% noted for the 25-34-year age group was significant at P <0.01. For the age groups 35-44 and >65 years, however, the increase was not found to be significant. The age-specific curves for breast cancer, by religious group, are similar except among Parsis who are more westernized and form only 1% of the population (Fig. 3). However, there is a wide variation in breast cancer incidence rates, Christians showing the highest age-adjusted rate (29.62) and Hindus the lowest (15.36 per 100 000 population). Estimates of the annual percentage changes by regression analysis of crude and age-adjusted rates for breast cancer among Hindus, Muslims and Christians are presented in Table 3. For Hindus and Muslims the trends in breast cancer incidence were found to be increasing, while for Christians they were decreasing. The average percent- age change in CR and AAR for Hindus was found to be statistically significant (P <0.001), but the annual percentage increase for Muslims and the annual Fig. 3. Age-specific incidence rates for breast cancer, by religious groups, in Greater Bombay, 1964-85. 300r 2001- c .9 Q 100a o0 0 0 O 00. 04)30 00) ;20 8?10 U)Q 0 0) 5 3 I --- HINDU .. *MUSLIM ----CHRISTIAN _ *PARSI Table 3: Estimates from regression analysis for crude rates (CR) and age-adjusted (AAR) Incidence rates, by religious groups, for 1964 and 1985 and the average percentage change for breast cancer In Greater Bombay Estimated rate Average percentage Group Rate 1964 1985 change t-test Hindu CR 8.78 14.26 2.34 7.80' AAR 15.36 25.78 2.53 8.51' Muslim CR 11.89 12.34 0.35 0.82 AAR 22.10 27.81 1.10 1.98 Christian CR 20.40 17.04 -0.71 1.43 AAR 29.62 28.83 -0.14 0.25 a Significant at P <0.001 level. percentage decrease among Christians are not signifi- cant. Table 4 presents the trends by age group (except for the Parsis where the total number in each 10-year age group was not statistically sufficient). An increas- ing trend was noted in all the 10-year age groups for Hindus and the annual percentage change was found to be statistically significant in all the age groups. However, a significant percentage increase was found for Muslims in the younger (35-44 years) but not the older age groups, and for Christians in the 55-64-year age group only. Table 4: Estimates from regression analysis for age- specific rates for breast cancer by religion, Greater Bombay, 1964-85 Age group Estimated rate Average and percentage religion 1964 1985 change t-test 35-44 years: Hindu 21.1 31.1 +1.80 3.04' Muslim 26.2 38.7 +1.88 2.65b Christian 40.2 27.0 -1.87 1.37 45-54 years: Hindu 36.2 70.2 + 3.20 4.54C Muslim 51.3 74.1 +1.76 1.89 Christian 87.2 69.2 -1.10 1.00 55-64 years: Hindu 51.4 83.2 +2.34 3.46' Muslim 69.9 67.6 -0.14 0.15 Christian 61.1 111.2 +2.96 2.18b >65 years: Hindu 51.8 92.8 2.85 4.51 c Muslim 76.2 119.9 2.21 1.58 Christian 87.3 116.2 1.37 0.69 ' Significant at P <0.01 level. bSignificant at P < 0.05 level. c Significant at P < 0.001 level. WHO Bulletin OMS. Vol 68 1990. Z5 30 40 Bo 6U) 10 80 Age (years) s. . . . . . . . . . . . 247 B.B. Yeole et al. Discussion The results of the regression analysis clearly show that the incidence of breast cancer in females in Bombay has been increasing over the last two decades. As data from the Bombay Cancer Registry are known to be reliable and complete (5) the increase is not an artefact but probably due to a change (over a period ofyears) in exposure to etiological factors. It was ofinterest to find out whether the increase was due to a cohort effect. While the age-specific curves for cohorts of females showed a discernible though not very striking effect, the average percentage change was not significant in some age groups. The reasons for this are not clear. However, the diverse sociocultural and religious back- grounds of the population in Bombay appear to be reflected in the differential rates of cancer in the various groups. Estimation of the average percentage change (APC) in incidence for each of these groups showed a significant APC for the Hindus in all age groups as well as a clear cohort effect (Fig. 4). The risk Fig. 4. Age-specific Incidence rates for breast cancer in various birth cohorts In Hindus. c .Q 1oo0 a so a o 6 uv 500 00. 2 50 .2 X 0 Cu U) 10 OC .1924 I 32-5 375 42-5 475 52 5 Age (years) 57 5 62-5 in each 5-year birth cohort (in Hindus), which was estimated by computing the cumulative incidence over comparable age ranges in successive birth cohorts by the method reported by Jayant & Yeole (1), showed that the percentage increase in risk in a younger cohort compared to the 5-year older cohort varied between 13.4% and 32.9% for cohorts born between 1914 and 1929. However, for cohorts born after 1929 the adja- cent 5-year birth cohorts had similar cumulative incidence rates (Table 5). Thus, the increase in breast cancer observed in Bombay could well be explained as a cohort effect, the more recent birth cohorts having a greater risk of breast cancer among the Hindus who form 70% of the population. An attempt was made to relate the observed trends in breast cancer incidence to changes in expo- sures to some known etiological factors. A first full-term pregnancy before the age of20 years is known to be a protective factor and non-childbearing single women have been shown to be at higher risk than married women. In India, the mean age at marriage for all females in the census synthetic cohorts has progres- sively increased from 13.5 years in 1911-21 to 17.2 years in 1961-71 (6). Hence, it can be inferred that the proportion of women having a first child before 20 years is gradually decreasing. Decennial census data on marital status have also shown that the proportion of "never married" in females has increased from 358 per 1000 females in 1921 to 452 in 1971 (6). Although the corresponding figures for Bombay are not avail- able, similar changes (through social changes in younger cohorts) are probably equally (if not more) applicable to Bombay, which is the premier city in India. This is supported by the results of a survey in Bombay in the 1960s, which showed the mean age at marriage as 17 years in all women (7), and by the finding reported in the decennial census that 431 out of 1000 females in 1961 (8) and 486 in 1981 (9) were never married in Bombay. Thus, the increasing trend in breast cancer incidence is most likely due to the progressive increase in age at marriage leading to delayed first births and an increase in the proportion of "never married" women. Table 5: Cumulative rate percentage over common age ranges for succesive birth cohorts for Hindus Year of birth Percentage increase Common in younger age range 1914 1919 1924 1929 1934 1939 cohort (years) 30-44 - - - - 0.32 0.33 +3.1 35-49 - - - 0.59 0.56 - -5.1 40-54 - - 0.65 0.81 - - + 24.6 45-59 - 0.73 0.97 - - - +32.9 50-64 0.82 0.93 - - - - +13.4 WHO Bulletin OMS. Vol 68 1990248 Trends In breast cancer Incidence In Greater Bombay The above explanation applies to the Hindus, who form the majority in Bombay and show a signifi- cant increase in breast cancer. However, it is not clear why the Muslims whose age at marriage over the years has been similar to the Hindus do not show a signifi- cant increase. On the other hand, among the Christians (whose average age at marriage was 17 years even in 1911 (6), it is possible that during the past two decades there has been no substantial change in the proportion of first births before the age of 20 years. This situation might have contributed to the observed stable rates of breast cancer in them. It may be noted that the incidence of cervical cancer over the same period has been stable among Muslims and Christians in contrast to the declining trend observed in Hindus (10). Thus it appears that, whatever factors increased the risk of breast cancer and decreased the risk of cervical cancer in Hindus, they remained stable over the years for both Muslims and Christians. The above-mentioned survey (7) also reported some differences in the average number of children delivered by the age of 20 years (0.44 for Hindus, 0.67 for Muslims, and 0.08 for Christians), which may have had a bearing on the differential rates of breast cancer observed in these communities. Information on the role of possible etiological factors in these cancers in Bombay is limited and we have not been able to consider the influence of age at menarche and at menopause or the intake of dietary fat in the observed increase in breast cancer. There is an urgent need to undertake well-designed epidemio- logical studies to explain the phenomenon and help control breast cancer before it assumes the magnitude now observed in the developed countries. Resume Tendances de l'incldence du cancer du sein en zone metropolitaine de Bombay: une evalua- tion epidemiologique Des donnees fiables issues du registre du cancer de Bombay montrent une augmentation de l'incidence corrigee pour I'fge du cancer du sein chez la femme, passant de 17,9 a 24,9 pour 100000 entre 1965 et 1985. En se servant d'un modele de regres- sion lineaire base sur le logarithme des taux d'inci- dence, on a trouve que les augmentations moyen- nes, en pourcentage, des taux bruts, des taux corriges de 1'age et des taux par Age sur cette periode de vingt ans sont hautement significatives. L'evaluation de cette tendance a la lumiere de facteurs etiologiques averes laisse a penser que I'augmentation de l'incidence du cancer du sein est li6e A la diminution progressive de la proportion de femmes ayant leur premier enfant avant I'Age de 20 ans et a I'augmentation de la proportion des femmes "jamais mariees". Ces observations sont egalement applicables au sous-groupe des hindoues, qui constituent 70% de la population, et chez lesquelles on observe une augmentation significative du cancer du sein sur cette periode ainsi qu'un effet de cohorte marque, les cohortes les plus jeunes ayant en general des taux plus eleves que les cohortes de cinq ans plus agees. En revanche, les sous-groupes constitues par les femmes musulmanes et chretiennes presen- tent des taux stables. Lors d'une 6tude anterieure sur l'incidence du cancer du col uterin pendant la meme periode, on avait observe des taux stables chez les musulmanes et les chretiennes et une tendance A la baisse chez des hindoues. Le cancer du sein est devenu le principal can- cer chez la femme a Bombay, alors que le cancer du col uterin predomine dans le reste du pays. 11 est urgent de proceder a des etudes epidemiologiques bien conques afin d'expliquer ce phenomene et d'aider a endiguer cette augmentation avant que le cancer du sein n'atteigne ici la meAme ampleur que dans les pays developpes. References 1. Jayant, K. & Yeole, B.B. Cancers of the upper ali- mentary and respiratory tracts in Bombay, India: a study of incidence over two decades. British journal of cancer, 56: 847-852 (1987). 2. Stukonis, M.K. Cancer incidence cumulative rates international comparison. Lyon, International Agency for Research on Cancer, 1978 (Internal Technical Report No. 78/002). 3. Jayant, K. Cancers of the cervix uteri and breast: changes in incidence rates in Bombay over the last two decades. Bulletin of the World Health Organiza- tion, 64: 431-435 (1986). 4. Trends in cancer incidence in Norway, 1955-78. Oslo, Cancer Registry of Norway, 1982. 5. Yeole, B.B. & Jussawalla, D.J. An assessment of reliability and completeness of Bombay Cancer Regis- try data (1983-1985). Indian journal of cancer, .25: 177-190 (1988). 6. Agarwala, S.N. India's population problems. New Delhi, Tata McGraw Hill, 1977, p. 92. 7. Rele, J.R. & Kanitkar, T. Fertilityand family planning in Greater Bombay. Bombay, Popular Prakashan, 1980, pp. 38-87. 8. Census of India 1961, Vol. X, Maharashtra Part-11 (912). Social and cultural tables. Superintendent of Census Operations, Bombay, 1965. 9. Census of India, 1981, Series 12, Maharashtra Part IV-A. Social and cultural tables. Director of Census Operations, Bombay, 1986. 10. Yeole, B.B. et al. Declining trend in cervical cancer incidence in Bombay, India (1984-85). Journal of surgical oncology, 42: 267-271 (1989). WHO Bulletin OMS Vol 68 1990. 249