Demographic Pattern, Tumor Size and Stage of Breast Cancer in Africa: A Meta-analysis

According to 2018 global cancer statistics [1], breast cancer (BC) is one of the two most common adult cancers, accounting for nearly 25% of cancers in women worldwide. Africa has disproportionately high age-standardized mortality due to BC [2]. The World Health Organization (WHO) and other experts in the field [3] recommend early diagnosis combined with timely and effective treatment as cost-effective measures for improving BC outcomes in Africa. Understanding the epidemiology of BC in Africa, as well as regional Abstract


Quality assessment
Five quality assessment variables were designed using domains in the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist [6]. The quality assessment classified the report into one of 5 levels, A-E, depending on the quality score, with A being the highest score of five and E being a score of one or zero. The quality score was not used in meta-analytical weighing.

Statistical analysis
The primary outcome was the summary estimate of each outcome variable defined in our PICOTS criteria. The meta-analytical procedure was conducted in MetaXL (www.epigear.com) add-in for Microsoft Excel. A random-effect model was implemented to obtain summary estimates using the double arcsine transformation to avoid overweighting studies with values close to 0 or 100%. I-squared [I2] values above 75% indicated high heterogeneity. Subgroup analysis was conducted based on the United Nations regional classification of African nations as Central Africa (CA), East Africa (EA), Northern Africa (NA), West Africa (WA), and Southern Africa (SNA). By-country analysis was also conducted to compare variables and explain the potential source of heterogeneity. We analyzed summary estimates of all variables where there were at least two observations for the continent, the region, or the country. The variables were analyzed as proportions of the total in each publication (n/N).
Tumor characteristics according to the clinical or pathologic AJCC, were analyzed separately. Results were presented in percentages with 95% confidence intervals (95% CI). The parent forest plots for all analyses are available in the supplementary file. Funding: No funding source.

Results
Full electronic search returned 5661 articles; 80 articles were eligible after the article selection process ( Figure 1, Table 2). Each article contributed data for one country except Sengal et al. [7], which provided data for Sudan and Eritrea in one article. Two articles from South Africa [8,9], Sudan [7,10], and Central African Republic [11,12] shared the same population of subjects but provided different data points (Table 2). Attempts to clarify incomplete or obscured data via email communications with authors yielded variable results as detailed in Table 2 footnote.
Twenty-three countries from all five regions of Africa contributed articles. WA contributed the largest number of articles (35 articles), followed by EA-19, NA-13, SNA-7, and CA 6. Nigeria contributed 23 articles, the largest from a single country ( Figure 2).
There were 33,199 subjects in total. The minimum number of subjects in a study was 42 [13], and the maximum was 3044 [14]. The maximum number of subjects from one country was 5425, contributed by Nigeria ( Figure 2). The majority of studies (n=39) were in the B quality assessment category; the study rationale was well stated in 84%, the design was adequate in 85%, and the participants were adequately stated in all studies. The study outcomes were adequately described in 77%, but the ease of data extraction was present in only 36% (Supplementary File). There was marked heterogeneity (>75%) in the overall summary estimates of the continentwide analysis. The heterogeneity was significantly reduced or eliminated in most by-region and by-country analyses (Supplementary File).
Temporal analysis revealed declining age of BC patients over time. Sixty per-cent of patients from 2010-2019 were less than 50 years of age, compared to 55% in 2000-2010 period. Eight per-cent of patients were less than 30 years of age from 2010-2019 compared to 4% in the 2000-2010 period ( Figure 2, Supplementary file).

Educational and marital status
Eleven publications (3747 subjects) contributed to the marital status analysis. Overall, 61% were married and 39% were unmarried, including 20% single. One study from CA [12] reported the highest proportion (74%) of unmarried singles; sensitivity analysis excluding this study saw the proportion of singles overall drop to 15%. The regional distribution of single patients was 11% EA, 14% WA, and 24% SNA. The largest unmarried population was in South Africa (57%) while the smallest was in Nigeria (27%). The temporal trend showed a slight increase in unmarried women diagnosed with BC in the last decade (41%) compared to the period between 2000-2010 (35%) CA and SNA. By-country analysis showed a similar distribution of 2-3% in Nigeria, Tanzania, Eritrea, and Ghana. Single studies reported 2% male prevalence in Cameroon [14] and South Africa [15]. A single study from Ethiopia recorded male BC prevalence of 18% [16]. Subgroup analysis showed a rising trend in male BC incidence in the last ten years compared to the decade before. (Supplementary file).

Age distribution
Thirty-three articles (14,545 subjects) contributed to age distribution analysis. Overall, more than half of patients (58%) were diagnosed before the age of 50. Twenty-eight per-cent of patients (95% CI 24-31) were diagnosed before the age of 40, and 6.0% (95% CI 5.0-8.0, I2=90%) were diagnosed at 30 years or younger. The youngest patients were in EA, where 8% (95%CI 6.0-11, I2=82%) were diagnosed under the age of 30, 38% (95% CI 31-45, I2=85%) were diagnosed under the age of 40, and 64% were diagnosed before the age of 50. Conversely, in SNA, over 60% were diagnosed at the age of 50 or above, and 37% (95% CI 35-39, I2=0%) were diagnosed at the age of 60 or above. The age distribution analysis was possible for NA only in the 50-year cutoff, showing that the majority were also younger than 50 years (58%, (95% CI 44-72, I2=94%). (Figure 3, Supplementary file) By-country analysis showed Ethiopia had the youngest patients with 10% (95% CI 8.0-13, I2=63%) being Participants/Population We included freely available publications of studies conducted in Africa and reporting on the total female breast cancer patients or both sexes or a representative sample. We excluded articles reporting on breast cancer patients' subpopulations, such as early presentation alone, young women, older women, or treatment subgroups.

Not applicable
Control Not applicable

Outcomes
The outcomes were: patient demographics (including age, sex, marital status, educational status, and menopausal status), and locoregional characteristic (including the primary tumor size, lymph node status, combined tumor staging, proportion of invasive and in-situ tumors, and tumor laterality).
The sex distribution was extracted in studies where the proportion of both sexes were reported. Age distribution was extracted in the range <40, 40-49, 50-59, and ≥60 years and in the binary distribution ≤30 years/ >30 years, and <50 years /≥50 years. Marital status was extracted into three categories: married, unmarried (separated, divorced, or widowed), and single (never married). Education was extracted into three categories: none/primary, secondary, and tertiary. Tumor laterality was extracted from articles that reported both unilateral and bilateral disease.
The proportion of invasive disease and carcinoma in-situ were extracted using articles that reported the two. The primary tumor size was based on the American Joint Committee on Cancer (AJCC) classification for the articles reporting in T1-4 fashion, and staging was based on articles where all four stages could be distinctly identified.
Nodal status was extracted as the presence or absence of nodal metastasis using the clinical or pathologic description according to the American Joint Committee on Cancer (AJCC) clinical staging criteria version 6 or 7.

Time
Articles published between January 2010 and December 2019. Articles including data earlier than January 2000 were excluded.
Study design Study design was not a strict exclusion criterion because demographic characteristics are expected to be fundamental elements in the reporting of any study. Language was also not an exclusion criterion. We included any original article with a sample size of at least 30 subjects providing at least one data point or observation according to the outcomes list above. We excluded review articles. Original articles involving more than one country were included if the observation (s) could be extracted separately for each country.  (Table 3). Twelve articles (3,103 subjects) contributed to the educational status analysis. Thirty-eight per-cent of patients had none or primary education, 36% completed secondary education, and 26% completed tertiary education overall. The proportion of patients with secondary or tertiary level education was highest in SNA (75%) than EA (62%) and WA (62%). The proportion of patients with secondary or tertiary level education in the last decade (64%) was slightly higher than the overall analysis (62%) ( Table 3). Subgroup analysis for 2000-2010 was not feasible. However, a single study from Nigeria in the 2000-2010 period recorded 52% secondary and tertiary education [17], another study from Nigeria with data between 2010 and 2012 recorded 66%, [18], and a separate study in Uganda including data between 2010 and 2013 [19] recorded 62%. (Supplementary file).

Stage distribution
Fifteen articles (9,185) contributed to carcinoma-insitu analysis. Prevalence of carcinoma in-situ was generally low in all regions (CA-4%, EA-2%, and NA-1%, SNA, and WA-1%). The highest proportion of carcinoma-in-situ in individual publications was 6% reported in Central African Republic and 5% in Malawi.
Overall, 98% of BCs were invasive based on analysis of 30 articles (10,352 subjects). Advanced BC (AJCC stage III or IV) accounted for 67%. Overall, 7% (95% CI 4.0-9.0, I2= 98%) of patients were diagnosed stage I disease, ranging from 2-10% in each region. Twenty-six per-cent of disease was stage II, ranging from 21-35% in each region, 50% of disease was stage III, ranging from 39-74% in each region, and 17% of disease was stage IV, ranging from 3-21% in each region. The earliest tumors were in NA; 74%, and 81% were Stage II or III in SNA and NA, respectively, while 70% or above were Stage III or IV in other regions ( Figure 4 and Supplementary File). Trend analysis showed a decreasing prevalence of stage I (from 8% to 4%) and stage IV (from 24% to 12%) disease, with an increasing prevalence of stage II and III disease in the last decade ( Figure 4).

Discussion
The African continent has a total population of approximately 1.  Together WA and EA account for more than 60% of Africa's population, and South Africa (SA) alone accounts for 88% of the population of SNA. We aggregated data from 80 articles published within the last decade, from 23 countries representing Africa's regions. Our findings corroborated previous evidence that African BC patients are younger than those from Europe and the US. In this study, 6% of patients were <30 years of age compared to 0.43% in the UK, 28% were <40 years compared to 6.6% in the US [96,97], and 58% were <50 years compared to 20% in Europe [98].
The age distribution of BC in South Africa showed a reverse pattern, mirroring the Caucasian age distribution seen in Europe. One explanation might be the proportion of Caucasian inhabitants in SA. Nonetheless, in four of the seven studies included from SA where the race was reported, 90% were Black patients (Table 1). However, it was not reported whether these patients might have been mixed-race Black patients. Even then, previous report suggests that black BC patients in SA are older than other races with BC in SA [99], though this may be partiallyattributable to under-reporting.
The declining age of BC found in this study in the setting of the increasing age of Africa's population overall [100] contradicts the view that the earlier age of BC onset can be entirely attributed to the younger population in Africa. Additionally, the elevated proportion of male breast cancer, 3% overall and 4% in the last decade, compared to approximately 1% reported globally [101,102], and previously reported increased rates of triple negative disease raise questions regarding potential genetic predisposition and merit further investigation.
The early age of BC onset in Africa brings numerous challenges regarding screening, early diagnosis, and treatment compliance [103]. Young women may be less likely to complete the diagnostic process or treatment for BC due to social reasons, such as fertility issues and sociocultural isolation. A report in Nigeria found that 31% of  young women outrightly declined the diagnostic biopsy procedure, 60% of those who did not decline failed to return for the result of the biopsy, and only 45% of those offered mastectomy accepted treatment [103]. Future intervention should be directed toward improving early diagnosis and compliance with treatment in this patient population.
Even in high-income countries where screening is ubiquitous, it is recommended to begin after 40 years (or 50 according to some guidelines). A third of BC patients in Africa were <40 and would be missed by applying the same screening age guidelines as in the US. While population-based mammographic screening programs are not feasible in most African countries due to resource constraints, education of the general population, paired with clinical breast examination (CBE) has the ability to downstage clinically apparent disease, and age range recommendations should be based on available data.
Thirty-eight per-cent of breast cancer patients in this analysis had none or primary education, and education level varied widely by region. This underscores the importance of tailoring breast cancer education and breast health awareness for both patients and the general population to the local context, taking into account educational and cultural background.
Although unmarried BC patients' population appears to be increasing in Africa, perhaps due to delaying marriage for education, the current predominantly married BC patient population still provides opportunities to include men as potential intervention targets. Reports suggest men are willing to support women in BC control [104,105].
In this analysis, nearly 90% of tumors were greater than 2cm, half were greater than 5cm, and one-quarter had skin or chest wall involvement. This preponderance of clinically-detectable disease means that the vast majority of patients have the potential for earlier detection by CBE. Marked regional variations in tumor size and stage across regions, might be explained in part by differences in health systems. Coordinating and centralizing local resources provided affordable, comprehensive health financing, and helped to downstage BC in NA [106]. Increasing awareness and reducing the distance and bottlenecks between BC patients and specialists aided downstaging in SA [9]. Countries in SSA could benefit from the experience in NA and SA to attain the goals of downstaging invasive BC. The smaller pathologic T staging compared to the clinical T staging in our analysis might be linked to the widespread use of neoadjuvant chemotherapy or errors of clinical or pathologic measurement and requires further investigation.
The decreasing rate of stage IV disease over time (24% to 12% in the first decade vs. the second decade of this analysis) shows promise for a slow, but positive trend toward earlier diagnosis, which is one of the most important factors in improving outcomes. The generally low prevalence of carcinoma-in-situ, 2-4% in all regions, can be explained by the common lack of populationbased screening. Population-based screening should not be considered until a health system has the resources and ability to effectively diagnose and treat clinically apparent disease. Nonetheless, it is important to note that the rise in the prevalence of small tumors (<1cm ) and carcinoma-insitu (from 2% to >20% ) in developed countries was linked to screening [107]. In light of the emerging evidence supporting the use of ultrasound in detecting early BC in young women [96,108,109], when a health system is ready to consider targeted screening, ultrasound may be considered along with CBE.
The heterogeneity, narrow spread, and paucity of articles capturing some of the variables analyzed limit our findings. Notably, only SA contributed to the findings for SNA. Similarly, in a previous review [110], only SA contributed to the meta-analytical review of the stage at presentation in SNA because the literature on BC is scarce from other SNA countries. Also, the reporting of demographic variables was influenced by the region as articles from the same region reported similar data points in similar formats. The countries of WA and EA reported more information on patient demographics than in the other regions. Availability of data limited our ability to draw conclusions regarding certain critical variables of interest, such as immunohistochemistry or treatment modality by stage.
In conclusion, Africa has a common goal of downsizing and downstaging BC, which is achievable through the early diagnosis of clinically detectable disease. There is marked regional variation in the clinical pattern and patient demographics of BC in Africa, and the interventions developed should be tailored to the local context in each area, while allowing for countries and regions to benefit from shared knowledge and experiences.

Declaration Funding
The authors received no funding for this research

Conflict of interest/Competing interests
The authors declare no conflict of interest.

Availability of data and Materials:
All data used in this article are freely available

Authors Contribution
Agodirin Contributed to all aspect of the research, all authors contributed to approval. Aremu contributed to conception, data acquisition, extraction , and review. Rahman contributed to conception, data interpretation, drafting, review. Olatoke contributed to conception, data interpretation and review. Olaogun contributed to draft, review and data interpretation, Akande contributed to data interpretation, drafting and review. Romanoff contributed to data interpretation, drafting and review.