The evolving epidemic of breast cancer in sub‐Saharan Africa: Results from the African Cancer Registry Network

Breast cancer (BC) is the leading cause of cancer in sub‐Saharan Africa (SSA) with rapidly increasing incidence rates reported in Uganda and Zimbabwe. However, the magnitude of these rising trends in premenopausal and postmenopausal women is unknown in most African countries. We used data from the African Cancer Registry Network on incident breast cancers in women from 11 population‐based cancer registries in 10 countries representing each of the four SSA regions. We explored incidence changes among women before and after age 50 by calendar period and, where possible, generational effects in this unique sub‐Saharan African cohort. Temporal trends revealed increasing incidence rates in all registries during the study period, except in Nairobi where rates stabilised during 2010 to 2014 after rapidly increasing from 2003 to 2010 (APC = 8.5 95%, CI: 3.0‐14.2). The cumulative risk varied between and within regions, with the highest risks observed in Nairobi‐Kenya, Mauritius and the Seychelles. There were similar or more rapidly increasing incidence rates in women aged 50+ compared to women <50 years in all registries except The Gambia. Birth cohort analyses revealed increases in the incidence rates in successive generations of women aged 45 and over in Harare‐Zimbabwe and Kampala‐Uganda. In conclusion, the incidence of BC is increasing rapidly in many parts of Africa; however, the magnitude of these changes differs. These results highlight the need for urgent actions across the cancer continuum from in‐depth risk factor studies to provision of adequate therapy as well as the necessity of supporting the maintenance of good quality population‐based cancer registration in Africa.

varied between and within regions, with the highest risks observed in Nairobi-Kenya, Mauritius and the Seychelles. There were similar or more rapidly increasing incidence rates in women aged 50+ compared to women <50 years in all registries except The Gambia. Birth cohort analyses revealed increases in the incidence rates in successive generations of women aged 45 and over in Harare-Zimbabwe and Kampala-Uganda. In conclusion, the incidence of BC is increasing rapidly in many parts of Africa; however, the magnitude of these changes differs. These results highlight the need for urgent actions across the cancer continuum from in-depth risk factor studies to provision of adequate therapy as well as the necessity of supporting the maintenance of good quality population-based cancer registration in Africa.

K E Y W O R D S
Africa, breast cancer, incidence, population-based cancer registry, trends 1 | INTRODUCTION Breast cancer incidence rates in Africa are increasing. The average annual percentage change (AAPC) in the breast cancer incidence rate was estimated at 4.5% between 1991 and 2006 in Kampala, Uganda, 1 and at 4.9% between 1991 and 2010 in Harare, Zimbabwe. 2 In a recent study describing cancer trends in people aged 60 and above in sub-Saharan Africa, breast cancer incidence was estimated to be increasing at an annual rate of 5% in Harare and Kampala, while slower rates of increase were observed in women under the age of 60. 3 Conversely, an earlier study in Western Africa (Bamako [Mali] and The Gambia) had described a more rapid breast cancer incidence rate among women aged 55 and below. 4 In contrast, the U.S. Surveillance, Epidemiology and End Results Program (SEER) in the period 1992 to 2011 observed stabilising rates among White non-Hispanic women with an estimated AAPC at −0.4%, while incidence rates among black and Asian/Pacific women were still moderately increasing, at 0.2% and 0.6%, respectively. 5 The rapid rate of increase in breast cancer incidence in sub-Saharan Africa in recent years has been attributed to a "westernisation" of lifestyles, that encompasses the effects of changing reproductive patternsdelayed age at first birth, fewer children and reduced breastfeeding duration-as well as changes in diet, alcohol intake and body weight, among other factors. 6 Other determinants unique to women of African origin have been suggested, such as the use of skin lighteners and increased exposures to hormone modulators in skincare and hair products, as often used by women of African descent. [7][8][9] Although this changing risk is common across populations, the incidence of breast cancer in sub-Saharan Africa varies across countries and regions. GLOBOCAN 2018 10 estimated the age standardised breast cancer incidence at 29.9 per 10 5 person-years in Eastern Africa, 27 6 Thus, little is published on long-term cancer trends from other SSA countries. Clearly more studies are needed to understand breast cancer incidence trends in the countries of the continent, separating out the contributing effects of ageing of the population, the aetiological effects on successive generations, and changes that simultaneously affect all studied age groups in a particular time period, that may represent changes in diagnostic capacity or completeness of cancer registration.
In this article, we describe the temporal trends observed in breast cancer incidence rates, using data generated by member registries of the African Cancer Registry Network (AFCRN). We explore in greater depth incidence changes among women before and after age 50 and examine period and, where possible, cohort effects in 11 populationbased cancer registries from 10 countries representing each of the four sub-Saharan African regions.

| Data sources
Completely anonymised data were obtained on incident cases of breast cancer (International Classification of Diseases, ICD-10 C50) from the database of the AFCRN (http://afcrn.org). This is a network

What's new?
Breast cancer is the leading cause of cancer in sub-Saharan Africa (SSA), and may be on the rise. In this study, the authors examined registries from ten SSA countries, and found that this is indeed the case, especially in older women.
Changing risk-factor profiles may account for these trends.
These results indicate an urgent need for strengthening the healthcare systems of SSA, including improved public health programs such as screening programs for breast cancer, indepth risk-factor analysis, etc., as well as planning for adequate therapy for an increasing number of patients. of sub-Saharan African population-based registries that facilitates homogenisation of registration activities, collaboration, advocacy and research. For a population-based cancer registry to obtain membership into the AFCRN, they must have attained at least 50% coverage of their target population on admission and at least 70% coverage by its third year. 11 There are currently 32 population-based member registries within the network, although for this incidence trend analyses, we included only those with at least 10 years of continuous data on breast cancer incidence. We grouped them according to the United Nations geoscheme for Africa. 12 From the Eastern African region, we included data from Blantyre, Country-specific population censuses were obtained from their respective National Statistics Office reports. To obtain the population-at-risk for each year, intercensal interpolations were made by sex, and within 5-year age groups assuming a constant logarithmic increase. Postcensal projections were made assuming a linear rate of increase at the rates observed between the preceding censuses.

| Statistical analyses
For each registry, the proportion of cases registered based only on a death certificate-death certificate only (DCO)-and of morphologically verified (MV) cases were estimated. These proportions are used as indicators of the data quality for each registry. 14 We estimated age-specific (5-year age groups) and crude incidence rates per 100 000 women by registry for the entire period and by 5-year periods. Age standardisation was performed by the direct method, using the World Standard population. 15 We also estimated the cumulative risk for each 5-year period. The cumulative risk expresses the overall risk of developing breast cancer before age 75, in the absence of competing risks of death. 16 Temporal trends in the age-standardised incidence rate were examined using joinpoint regression, 38 Breast cancer crude incidence rates, age-standardised rates and cumulative risk by registry area and calendar period the economic crisis in those years). 2 We used a modification of the method proposed by Schifflers et al, 18 fitting a fourth degree polynomial rather than cubic splines to interpolate values for the missing years, based on the observations in adjacent periods. This graphical analysis is critical to understanding and interpreting what is observed in the summary agestandardised rates, as sometimes summary rates cannot adequately represent time trends in the presence of influential cohort effects. 19 3 | RESULTS 3.2 | Time trends in age-standardised incidence rates, cumulative risks and Joinpoint analyses    averages. Incidence rates are higher in women aged ≥50 for all registries, although in The Gambia there was a progressive convergence of rates in these two age-groups.  Figure 3 shows the age-specific rates by registry in successive time periods. We observe rapid increases in incidence with advancing age in premenopausal women, followed by a decline in the gradient of the curves after age 45. In Brazzaville and Seychelles, this decline in the gradient of the curves is seen after age 55. There are fluctuations in age-specific rates in the earlier time periods in The Gambia. Overall, these trends point towards a gradual increase in age-specific incidence rates in successive time periods for most registry areas, with the relative difference in age-specific rates over time being greater for women aged 45 and above.

| Time trends in age-specific incidence rates
Trends by birth cohort in Harare-Zimbabwe and Kampala-Uganda are shown in Figure 4. We observe an increase in the breast cancer Age at diagnosis (years) 1991-1994 1996-1999 2000-2004 2005-2009 Ibadan-Nigeria women compared to older women, as has been noted previously. 23 Similarly, the relatively small increase in incidence with age in older women, evident in Figure 3 (compared to that observed in western populations) represents a cohort effect, with progressively higher incidence rates in more recent generations 23 -an effect visible in Figure 4-most clearly for Harare. As a result, at any period of observation, older women will appear to have relatively lower rates, and there may even be an apparent decrease in risk with age (as in some of the curves in Figure 3).
For areas for which we have data as far back as the 1960s, we observe a more than threefold increase in the risk of breast cancer over four decades. In CI5 Volume 1, 24  Despite the increasing rates, observed particularly in postmenopausal women, they are lower than rates observed in more developed countries. In Table S2, we use data from GLOBOCAN 2018 to compare the ratio of the cumulative risk and ASIR of post-menopausal (ages 50 +) to premenopausal (ages <50) using national level estimates for different African regions, Europe and North America. Rates of postmenopausal breast cancer are much higher than premenopausal in all world regions; although the ratios are lower in Africa, where increases in incidence (generation-specific) appear to be continuing. This is partly due to the birth cohort effects seen in Figure 4, whereby rates increase within each age group among successive generations of women. However, the higher differential among post-menopausal women in developed countries in part reflects the effects of population-level screening, with increases in recorded rates in women in screened age groups (eg, ages 50-74) resulting from lead-time and overdiagnosis effects. In the prescreening era in Denmark the cumulative risk of developing breast cancer was estimated at 8%, 27 and in 2018 the cumulative risk was estimated at 9.5%. 28 In our study, the cumulative risk of developing breast cancer before age 75 ranges from less than 1% in The Gambia to 7.3% in Nairobi-Kenya. The lifetime risk of developing breast cancer in the U.S. for all ethnics in 2010 to 2012 was estimated at 12.3% (1 in 8 women). 29 The influence of oestrogens on breast cancer pathogenesis is probably similar across regions of the world, 30  breast cancer (which is more common in older women), and a decrease in the risk of triple-negative breast cancer. 33 In Ghana, Figueroa et al showed that increased parity (≥3 births) was associated with a reduced risk of both ER+ and ER− breast cancer among women aged 50+ but was associated with an increased risk of early onset ER − tumours. Extended breastfeeding was also associated with a protective effect for both ER+ and ER-breast cancer, although the protective effect was stronger for ER+ tumours. 34 In a small Ugandan hospital-based study, no differences in distribution of reproductive risk factors by ER status were found among breast cancer cases. 35 Use of oral contraceptives has been shown to increase breast cancer risk in current and recent users. The use of modern contraceptives, among which are exogenous hormones, varies by sub-Saharan African region with the lowest use in Middle and Western Africa and the highest use in Southern Africa 39 ; there is a greater uptake of contraceptive use among women of higher socioeconomic status. 40 The use of injectables and implants have increased among sexually active women over time in SSA according to data from the DHS and the Performance Monitoring and Accountability Surveys 2020 (PMA 2020). 39 The use of injectable and/or oral contraceptives in South Africa was associated with significantly increased breast cancer risk, and this increased risk persists for about 5 to 10 years after cessation. 41 Reduced physical activity has been associated with increased breast cancer risk in SSA, 42 and physical activity during leisure in many SSA countries is relatively low. 43 There are also ongoing changes in diet with "westernisation" characterised by an increased consumption of diets richer in calories and poorer in fruits and fibre. The increased use over time of animal over plant-based products in sub-Saharan Africa correlates with increased breast cancer incidence rates. 44 Consumption of diets rich in fruits and cruciferous vegetables has been associated with a decreased breast cancer incidence rates in South Africa, while diets rich in calories and which are nutrient poor conferred higher incidence rates. 45 These changing risk factor patterns across Africa surely account for much of the increase in breast cancer incidence rates, the larger increases observed in older (postmenopausal) women, as well as prob- Given the long time series available, and previously reported time trends of breast cancer, 4 we included data from The Gambia. The fluctuations in the incidence rates observed in the Gambia are likely due to a deficit of cases in age groups 55 to 59 and 65 to 69 years as well as the uncertainty concerning age in older women, as many more women are registered aged exactly 60 and 70 ( Figure S1). In addition, the registry is known to suffer incompleteness of case findingparticularly in older subjects 50 -and this appears to be an increasing problem, 26 which probably accounts for the aberrant results from this registry. In addition to potential problems of registry quality, calculation of rates relies upon interpolations of population censuses usually done at 10-year intervals, so that the accuracy of our denominators will depend on the available census data.
Despite these challenges and limitations, these results highlight the need for urgent actions across the cancer continuum, from more in-depth risk factor profile studies across Africa, to the provision of adequate curative treatment and palliative care services at the national level to meet the demands of an increasing number of breast cancer patients.
They also highlight the necessity of supporting the maintenance of good quality population-based cancer registration activities.

ACKNOWLEDGEMENTS
We would like to acknowledge the contribution of the cancer registry staff -including the unpaid interns and volunteers -who collected the data presented here, often under difficult circumstances, and then coded and entered the data into the local databases from which the information for this article was obtained. We express our gratitude to

CONFLICT OF INTEREST
No conflicts of interest.

DATA ACCESSIBILITY
The data that support the findings of our study are available on request. All data requests will be evaluated by the AFCRN research committee. Details of the data application process are outlined on the AFCRN website http://afcrn.org/index.php/research/how-to-apply/ 76-research-collaborations.

ETHICS STATEMENT
Approval for our study was obtained from the AFCRN Research Committee and from participating registries. The study made use of routinely collected population-level anonymised data. The study was performed in accordance with the Declaration of Helsinki.