There is little information on breast cancer (BC) survival in Ethiopia and other parts of sub-Saharan Africa. Our study estimated cumulative probabilities of distant metastasis-free survival (MFS) in patients at Addis Ababa (AA) University Radiotherapy Center, the only public oncologic institution in Ethiopia. We analyzed 1,070 females with BC stage 1–3 seen in 2005–2010. Patients underwent regular follow-up; estrogen receptor-positive and -unknown patients received free endocrine treatment (an independent project funded by AstraZeneca Ltd. and facilitated by the Axios Foundation). The primary endpoint was distant metastasis. Sensitivity analysis (worst-case scenario) assumed that patients with incomplete follow-up had events 3 months after the last appointment. The median age was 43.0 (20–88) years. The median tumor size was 4.96 cm [standard deviation (SD) 2.81 cm; n = 709 information available]. Stages 1, 2 and 3 represented 4, 25 and 71%, respectively (n = 644). Ductal carcinoma predominated (79.2%, n = 1,070) as well as grade 2 tumors (57%, n = 509). Median follow-up was 23.1 (0–65.6) months, during which 285 women developed metastases. MFS after 2 years was 74% (69–79%), declining to 59% (53–64%) in the worst-case scenario. Patients with early stage (1–2) showed better MFS than patients with stage 3 (85 and 66%, respectively). The 5-year MFS was 72% for stages 1 and 2 and 33% for stage 3. We present a first overview on MFS in a large cohort of female BC patients (1,070 patients) from sub-Saharan Africa. Young age and advanced stage were associated with poor outcome.
Ethiopia is the second most populated country in sub-Saharan Africa. In 2012, Ethiopia reported a population of 86.6 million people (estimated in 2013); nearly half (47%) was under the age of 15 years and life expectancy was 58 years. Ethiopia is one of the most rural countries in the world, with only 17% of the population living in urban areas. Ethiopia is a very diverse country, covering 80 different ethnic groups. The gross national income per capita was $870 in 2008; the per capita total expenditure on health was $30 (10% of total government expenditures) in 2007. The centrally located capital of Ethiopia, Addis Ababa (AA), is a fast growing city with a radical change in lifestyle: the total fertility rate is only 1.5 as opposed to 5.5 children per women in rural Ethiopia. There are around 122 public and about 50 private hospitals in Ethiopia; this is 1:671,402 population. One striking problem compared with other countries of sub-Saharan Africa is the scarcity of medical personnel, e.g., only 0.4 physicians per 10,000 inhabitants compared with five times more physicians on average in the African region. The prevalence of HIV is still low around 2% (2.1%). Communicable diseases are still the major burden of disease as a percentage of total disability-adjusted life years (DALYs) (73%). Similar to other developing countries, non-communicable diseases are emerging. Breast cancer (BC) has now become the most commonly diagnosed cancer in women in several sub-Saharan African countries. This is a shift from previous decades when cervical cancer was the most commonly diagnosed cancer. Furthermore, the burden of BC is likely to increase in the coming decades. This probably reflects increases in the prevalence of known risk factors associated with urbanization.[6-8] In Ethiopia, an estimated age-standardized incidence rate of 19.5 per 100,000 and an estimated age-standardized death rate of 11.8 per 100,000 females are reported. The World Health Organization (WHO) declared the fight against cancer a priority for all governments in 2005. The Breast Health Global Initiative (BHGI) states that before any targeted national program in a specific country can be set up, data on the incidence and prognosis should be obtained.[11, 12] A review of the literature revealed that only few data on BC survival from sub-Saharan Africa are available.
Unlike the Western world, where women present early and have a good chance of survival, women in Ethiopia usually present late and are expected to have a very limited life span. Public oncologic treatment including radiotherapy in Ethiopia is limited to the Radiotherapy Center at AA University Hospital, staffed with four oncologists. Limited oncologic service (chemotherapy without radiotherapy) is offered by these oncologists in four private clinics. Breast surgery is offered at nearly all hospitals by general surgeons throughout the country. A system to supply patients without financial resources is in place for government hospitals (“poor-papers” to ensure free treatments are issued by local authorities). Since 2005, evidence-based standards of BC care have been adapted according to the BHGI. Women with BC account for 19% of the total cancer patients. From 2006 through 2010, an independent project funded by AstraZeneca Ltd. (Cambridge, UK) and facilitated by the Axios Foundation (Paris, France) provided free endocrine treatment (tamoxifen and anastrozole) for estrogen receptor-positive and -unknown patients. Information about the program was widely distributed amongst institutions, cancer organizations and mass media in AA. Because of this unique situation, we considered BC patients at the AA University Radiotherapy Center from 2006 to 2010 as a representative cohort of patients, presuming that the majority of patients found their way into the program and came for regular follow-up to receive their endocrine medication.
The primary aim of our study was to estimate distant metastasis-free survival (MFS) of women with BC diagnosed between June 1, 2005 and May 31, 2010 at the AA University Radiotherapy Center. Furthermore, we estimated the effect of potential determinants including age, stage of the disease and histologic type on MFS.
Material and Methods
Women with a histologically verified primary diagnosis of invasive carcinoma of the breast [International Classification of Disease-Oncology (ICD-O-3) codes C50.0–9] without evidence of distant metastasis consulting the Radiotherapy Department of AA University between June 1, 2005 and May 31, 2010 were included in our study (n = 1,070). All patient and tumor characteristics and information concerning therapy and outcome were documented from patients' files.
According to international coding standards for cancer registries, the date of incidence was defined as the first consultation at a hospital for the cancer in question. Information on tumor size (T) and nodal status (N) was used to derive stage by the American Joint Committee on Cancer staging system AJCC (seventh edition): stage 1 (TxN0); stage 2 (T0N1, T1N1, T2N0, T2N1 or T3N0) and stage 3 (TxN2, T3N1, T4Nx or TxN3). All staging information mentioned within the first 3 months after primary diagnosis was used. T-stage was assessed according to the 2009 TNM classification. For N-stage, the information on involved lymph nodes (LNs) from the pathologist was used, given that an adequate number of LNs was examined (≥double the number of involved nodes for N1 and any number for N2 or N3). This information was used to derive the stage (UICC) as a potential prognostic factor for estimation of MFS-probabilities. Additionally, in patients without operation (n = 140) or neoadjuvant chemotherapy (n = 42), we decided to use clinical TNM to derive the stage (UICC) as a potential prognostic factor for estimation of MFS-probabilities. As these patients had T3 and T4 tumors, we assumed that clinical TNM would not differ from theoretical pathologic TNM. For patient characteristics, any positive LNs described in the pathology report and clinically detected suspect LNs were coded N+. Contralateral regional LNs clinically noticeable without evidence of a local contralateral disease were considered to be distant metastasis (M). M-stage: in low-resource settings, standardized staging by imaging is often not done. In this cohort, a minimum of chest X-ray and abdominal ultrasound were usually performed. A lack of clinical symptoms (n = 149) or radiologically confirmed absence of distant metastasis (n = 921 investigated with chest X-ray and abdominal ultrasound) at diagnosis was considered to be free of distant metastasis. Bone scans were not available. Tumor histology was classified according to written notes from pathology reports as ductal not otherwise specified (NOS) or lobular, and all others were summarized as other/unspecified.
The primary endpoint of our study was MFS because mortality data could not be obtained (patients usually die at home, so no death registration is available). Person time equaled the time from the date of diagnosis to the date of distant metastasis or to the date of last contact. Women without event were right-censored at the last visit to the clinic. Information was collected between October 14, 2010 and the closing date March 23, 2011. All patients who had not come for their appointment since more than 6 months ago were considered to have incomplete follow-up. Different patterns in right-censoring for incomplete follow-up over time were investigated between established prognostic groups of interest (in our case, information on age, stage and histology was available). No differences were seen using a Cox model to determine hazard ratios (HRs) for incomplete follow-up among the prognostic groups. During follow-up, patients were investigated for clinical signs of metastasis. M was confirmed by positive findings on imaging or pathologic diagnosis [two cases (0.7%) out of 285 were diagnosed by strong clinical suspicion only]. The time of occurrence of distant metastasis was defined as the time point of a positive finding on imaging or of a strong suspicion of distant metastasis by clinical signs whichever came first. Local recurrence during follow-up time was not analyzed, because operation notes and pathology reports did not sufficiently report about remaining tumor and surgical margins. Therefore, primary progression and local recurrence could not be precisely distinguished. Of 372 patients who showed local tumor after surgery, 163 also developed distant metastasis. Of 16 patients who were known to be certified dead, 15 died of BC. Only one of 16 patients died of a cause unrelated to BC; this patient was right-censored.
Analyses were done with SAS® (SAS, Cary, NC), Version 9.3. Median follow-up time was 23.1 (0–65.6) months. We estimated MFS by means of Kaplan–Meier survival analysis. We used causal graphs [directed acyclic graphs (DAGs)] to identify confounding factors. Minimally sufficient adjustment sets were calculated (Supporting Information 1) using a standard software program. Adjustment for age was done for stage and histology. We used crude and multivariable Cox proportional hazards regression to estimate unadjusted and adjusted HRs and corresponding 95% confidence intervals (95% CIs) with respect to prognostic factors. We checked the assumption of proportional hazards by the use of Schoenfeld residual plots. Worst-case analysis considered all patients with incomplete follow-up as having developed distant metastasis 3 months after the last visit.
Ethical approval for our study was obtained from Institutional Review Boards of AA University Medical Faculty and Martin Luther University Halle. The study was conducted without individual informed consent as the study relied on retrospective data collected as part of routine patient care.
Of 2,031 registered patients, 1,507 files could be retrieved. Overall, 1,070 women fulfilled the inclusion criteria (Supporting Information 2). Nearly all patients (n = 930; 87%) received an operation. The majority had modified radical mastectomy (n = 880; 95%). Of these, an estimated 20% were operated at the Department of Surgery at AA University Hospital. In 628 surgery reports describing the margins, 69.9% reported surgical margins free of disease and 30.1% had margins involvement [for 302 patients (32%), no information on margins was available]. The majority of patients (n = 893; 83%) also received chemotherapy, mainly anthracycline-containing chemotherapy (n = 782). Of these, about 70% were administered at the AA University Radiotherapy Center. The preferred anthracycline-containing regimen was FAC (5-fluorouracil 500 mg/m2, adriamycin 50 mg/m2 and cyclophosphamide 500 mg/m2) for 666 patients, AC (cyclophosphamide 600 mg/m2 and adriamycin 60 mg/m2) for 116 patients and for 76 patients cyclophosphamide, methotrexate and 5-fluorouracil (CMF) was used. Taxanes were not available in general. Of all chemotherapy patients, 753 patients (83.7%) received a full six cycles of chemotherapy, mainly FAC (n = 577). There were 42 patients who received neoadjuvant chemotherapy. All endocrine therapy for positive or unknown receptor status (n = 864) was administered at the AA University Radiotherapy Center. Patients' everyday compliance to endocrine treatment could not be assessed. Adjuvant treatment (if applied) was given rather well-timed; 79% of patients with chemotherapy started within 6 months and 77% of patients with endocrine therapy started treatment within 12 months after surgery.
In 285 women (26.6%), distant metastasis occurred during follow-up. The majority of women came for regular follow-up visits ranging from 8.1 to 65.6 months after primary diagnosis (median 23.1 months). Altogether, 101 women (9.4%) did not have any follow-up visit (after completion of therapy: maximum 8 months after the date of diagnosis) and 130 women (12.1%) had incomplete follow-up later on. At the end of the study, 78% of women had complete follow-up.
The age of the women ranged from 20 to 88 years (median age 43.0 years), with women aged 30–39 being the largest group of the study population (37.9%). Almost half of the women were premenopausal (49.7%; n = 889 information available). Among the women whose origin had been specifically inquired when their history was taken, half were classified as AA and half as non-AA. The majority of women presented with stage 3 disease. Nodal status was positive in 81% of cases. The most frequent histological type of BC was ductal carcinoma NOS. About half of the women had a pathology report showing grade 2. The extent of incomplete follow-up and differences in distribution by prognostic factors are shown in Table 1.
Table 1. Clinical and pathological characteristics of the patients
At the end of the study—Proportions within different prognostic groups
Complete follow-up (metastasis)
Complete follow-up (without disease)
Place of origin
Ductal not otherwise specified
Surgery within 6 months
Chemotherapy within 6 months
Endocrine therapy within 12 months
Tumor size (mean, standard deviation)
4.96 ± 2.81 cm
MFS of patients after 2 and 5 years was 74 and 46%, respectively. In our worst-case analysis, MFS declined to 59 and 27%, respectively (Fig. 1). The difference between documented results and worst-case analysis was less than 20% for 2- and 5-year MFS (Table 2).
Table 2. Distant metastasis-free survival (MFS) probabilities (in %)
MFS (95% CI)
After 2 years
After 5 years
After 2 years
After 5 years
Abbreviation: CI: confidence interval.
Place of origin
Ductal not otherwise specified
To find out the factors influencing MFS, HRs were calculated for the different patient characteristics (Fig. 2). Differences between unadjusted and adjusted values can be seen in Supporting Information 3. MFS was highest in women aged 50–59 years and was lower in younger age groups (Fig. 3). Women aged 60 years and above also tended to have worse prognosis compared with those aged 50–59 years. The HR for distant metastasis of patients <30 years of age was higher (HR = 3.20, 95% CI 1.99–5.14) compared with that of women aged 50–59 years (Fig. 2).
Women with stage 3 disease had a considerably worse MFS than patients with stage 1/2 disease, showing an HR of 2.62 (Fig. 2). Women with unknown stage had MFS between those with stages 1/2 and stage 3 (Fig. 4).
Tumor histology was grouped as ductal NOS, lobular and other/unspecified (e.g., four phyllodes tumors, two sarcomas and medullary carcinoma). Patients with ductal histology had the best MFS compared with the other entities. The lobular group also included specified lobular cancers; outcome was not as beneficial as expected for pure lobular cancers (Supporting Information 4).
This is the first detailed and largest study on BC survival in Ethiopia and other parts of sub-Saharan Africa. Overall, MFS probability in this cohort of Ethiopian women with BC was 74% after 2 years and 46% after 5 years [median follow-up 23.1 (0–65.6) months]. Patient characteristics in our cohort tended to be unfavorable compared with Western cohorts; more than 50% were premenopausal, aged <40 years and/or with stage 3 disease. Tumor biology was more favorable; the majority was ductal and grade 2. Often cancer survival data in developing countries have the endpoint of overall survival. We only found follow-up data with an endpoint on MFS from Nigeria, without any information on treatment. Out of the described 308 cases from Nigeria, 106 cases with incomplete follow-up were omitted and only 202 were entered into Kaplan–Meier plots. The plot shows only those patients who developed metastasis within 5 years; therefore, a comparison is not possible. Looking at worldwide 5-year overall survival probabilities, the figures ranged from 89.2% in highly developed countries down to 38.8% in Algeria and 12% in The Gambia.[22, 23] A report from Uganda showed an overall survival probability for stage 1–2 cancers of 74% and stage 3–4 cancers of 39%. Our stage 1 and 2 patients showed 72% and stage 3 patients showed 32% MFS probability, and overall survival is expected to be higher. This possibly points to a more favorable outcome in our cohort from Ethiopia.
The median age of 43 years (20–88) in our cohort was slightly lower compared with studies from other countries. A previous study conducted from 1995 through 1999 on 125 consecutive BC cases from AA University Hospital also revealed a median age of only 40 years. Reports show a median age of 46 years in Mali (n = 118), 46 years in Tanzania (n = 328), 49 years in Ghana (n = 330) and 48 years in Nigeria (n = 192).[26-29] The differences in MFS in our age groups point to a possibly favorable prognosis in women between 50 and 59 years of age in terms of MFS. This is similar to Western studies showing that younger women have worse outcome. It is most likely that general health status is responsible for the suboptimal therapy and tendency for poorer MFS in our group of women aged 60 years or more.
As in other developing countries, a large proportion of women presented with a late stage of the disease (70% stage 3). This finding is in line with previous studies from Africa. For example, the study from Tanzania reported that 71% of patients with nonmetastatic BC were stage 3. The study from Uganda reported that 69% of the women had stage 3 BC. In Egypt, 60% of women presented with late-stage disease in a prospective study. In contrast, the North American Surveillance, Epidemiology, and End Results Program database reveals that 60% of BCs are diagnosed at stages 1 and 2. In our study, stage 3 BC patients showed an HR of more than 2.5 compared with the MFS in stage 1/2 patients. This highlights the fact that earlier detection would indeed improve outcome in BC patients in Ethiopia.
Often, tumors in the African setting are described as aggressive with high grading. Ly et al. reported a high incidence of grade 3 tumors (78%) from Mali. In Ghana, 54% were grade 3. In Tanzania, 56% grade 3 cases were seen. In Nigeria, the findings are contradictory: Huo et al. found 44% grade 3 tumors, whereas Adebamowo et al. found only 22% grade 3 tumors. Such high-grade tumors occurred less frequently in our cohort (35%). Because of the subjective nature of the assessment, a degree of uncertainty must be taken into account. Altogether, these findings may reflect a less aggressive nature of tumors from Ethiopia.
We found that the majority of patients received adjuvant treatment according to the BHGI guidelines, which recommend anthracyclines (782 of 893 patients with chemotherapy received anthracyclines) and tamoxifen (for positive and unknown hormone receptor status) for this setting with limited resources. Adjuvant therapy was done in a standardized manner at the AA University Radiotherapy Center for the majority of patients. Surgery was mainly modified radical mastectomy (880 of 930 operated patients). The fact that nearly 80% of surgery was done in peripheral hospitals might account for a suboptimal standard. There are published data about the adequacy of surgery in peripheral hospitals from India. Thorat et al. reported that out of 424 referred cases, 153 were reoperated owing to clinically suspected inadequate removal of tumor. Of those, 64 patients were diagnosed with additional involved LNs; no tumor was found in the remaining breast tissue. Therefore, in our cohort from a similar setting, there might be a considerable proportion of patients with involved LNs in situ. This could explain the high number of local recurrences seen. Additionally, there were 140 patients who did not receive any surgery owing to a lack of plastic surgery options. These patients did receive chemotherapy. Altogether, our cohort consisted of about 700 patients who received optimal care with mastectomy, anthracyclines and endocrine treatment. As we wanted to focus on a consecutive cohort in this setting with limited resources, we decided to describe the total cohort including the group of patients with inadequate treatment. The reasons for different BC-specific survival of ethnic groups are unknown. Among US citizens, Caucasian women have the best, African-American women have poorest and other immigrants have intermediate BC-specific survival. Possible factors that contribute to these differences are unfavorable stage at diagnosis, limited access to care, comorbidities, socioeconomic status, obesity and physical activity. Besides genetic factors leading to differences in tumor biology, patient and lifestyle factors might also contribute to the differences in findings. Women of African origin have a much more diverse genetic background than Caucasian women. In addition, the environment and lifestyle differ considerably throughout the African continent. We hypothesize that East African women have better survival outcome than West African women (Gambia, Nigeria) because of more favorable biological characteristics such as tumors with a lower grade. In addition, in Ethiopia the predominantly rural environment (e.g., less processed food) and non-Westernized lifestyle (e.g., underweight, high physical activity, high number of children, long period of breastfeeding, less hormonal contraceptives or replacement therapy) may reflect differences from other African countries. Also, the lack of awareness, low economic status, lack of knowledge, etc. may result in less treatment uptake and therefore patient selection.
There are some factors that limit our results. First, the non-AA population (97% of Ethiopia) was most likely underrepresented at the AA University Radiotherapy Center. We found that non-AA women tended to be younger, to have a higher stage of disease, less frequent endocrine treatment (differences of up to 6% between groups) and similar MFS. These might be general features of the disease in the rural setting or might be the result of patient selection. Because of the rather high costs of living in the capital city AA, we have to assume that patients with low income, probably mostly from rural areas, may not reach the AA University Radiotherapy Center. Therefore, this cohort is probably exclusive of very poor patients, especially those living further away. It is possible that this group that hardly accesses the health care system would present late, have less compliance, not remain in care and therefore add to the unfavorable higher stage group with reduced probabilities. Second, information on stage was missing in about 40% of the cases. As the MFS of patients with missing stage was between that of patients with stage 1/2 and stage 3 disease, these patients most likely reflected a mixture of patients from the other two groups. Third, about 25% of the registered files were not found. The record system at the AA University Radiotherapy Center is based on hard copy patient files that are stored according to names and patient numbers. Files can easily be misplaced. Illiteracy frequently results in various spellings of names; files are frequently not found. We are not aware of any other reason for missing files and therefore do not suspect a selection bias through this fact. Fourth, our cohort showed a considerable proportion of women with incomplete follow-up (21%). Therefore, the assumption of right-censoring being independent of time to distant metastasis may not be valid. By analyzing the time pattern of right-censoring in incomplete follow-up patients within prognostic groups, we found no differences between prognostic groups. However, this does not prove the above-mentioned assumption of independence for the prognostic factors. In a setting without valid endpoints like vital registration and with a retrospective design, these doubts cannot be ruled out. For sensitivity analysis, all patients with incomplete follow-up were considered to develop distant metastasis 3 months after their last appointment. The resulting differences in MFS did not differ among the subgroups of the prognostic factors of interest. Therefore, we believe that potential bias was not substantial, although we are aware that these observations cannot exclude that right-censoring was related to the risk of distant metastasis.
We reported data from 1,070 women with BC stage 1–3 followed for a median time of 23.1 (0–65.6) months. These women showed a rather favorable 5-year outcome of 45% MFS compared with previously described smaller cohorts from sub-Sahara Africa in the literature. We found a lower median age and a higher proportion of grade 2 tumors in our cohort compared with those previously described in other parts of sub-Saharan Africa, mainly in West-Africa (older patients, more grade 3). In our cohort, outcome was negatively influenced by age below 49 years and age above 60 years, higher stage at presentation and a histology other than ductal carcinoma NOS. These findings are in line with published data from Caucasian patients. The majority of women who presented with stage 3 disease at diagnosis had worse outcome than stage 2 patients. This shows the potential benefit and urgent need of downstaging programs, which have been proven successful in similar settings elsewhere.
The authors acknowledge the help of all staff from the AA University Radiotherapy Center. There are no financial disclosures for any author.