Inequities in breast cancer treatment in sub-Saharan Africa: findings from a prospective multi-country observational study

ABC-DO is a prospective multi-centric hospital-based study of survival after breast cancer diagnosis in five SSA countries. A detailed protocol has been published [11]. During September 2014 to early 2016, all women aged 18 and older presenting in the participating hospitals and suspected of having breast cancer (either clinically or through histological confirmation) were invited to participate. Benign cases were excluded from participation after diagnostic workup. Across all sites, 99% of invited women (n = 2265) agreed to participate. In the present analysis, after excluding 7 women with no treatment data, the partaking hospitals included 1335 women from (i) Windhoek Central Hospital’s AB May Cancer Care Centre in Namibia (n = 502, among these n = 398 black women and n = 104 non-black women); (ii) Uganda Cancer Institute and Mulago Hospital, Kampala (n = 430) and (iii) the Federal Medical Centre, Owerri, and Abia State University Teaching Hospital, Aba (n = 313, combined as Nigeria public), and, the only private setting, the Maranatha clinic, Aba (n = 80), Nigeria. Two ABC-DO countries that lacked complete or harmonised treatment information at the time of analysis were not included, i.e. Zambia (n = 207) owing to a later recruitment period and South Africa (n = 716) due to a different data capture system. For the three included countries, the catchment populations, and breast cancer treatment options and their approximate costs are listed in Table 1. The sites include diverse public-sector cancer care providers, notably two regional hospitals in Nigeria, a national referral general hospital in Uganda and a specialised national cancer care centre in Namibia, with varying out-of-pocket care costs to the patient. Due to these differences, treatment access is thought to reflect each site as an example of the cancer treatment setting, rather than as representative of the country as a whole.

Due to the complexity of multimodality breast cancer treatment, for the present analysis, we analysed whether any breast cancer treatment was initiated for each woman within 12 months of diagnosis (yes/no), which we defined as having had or commenced at least one of surgery, radiotherapy or systemic (chemotherapy or endocrine) therapy. Palliative care without curative intent (e.g. opioids for pain management) was also recorded, but was not included here. The binary treatment indicator was created as follows: First, data was extracted from two systematically recorded data sources—medical records and patient self-reporting.

Socio-economic, breast-specific and other health-related determinants were examined in relation to treatment received. These included the following: (i) five “population groups” defined according to country (Namibia, Nigeria, Uganda), ethnicity (black vs. non-black) in Namibia and hospital type (private vs public) in Nigeria; (ii) recorded TNM stage at diagnosis: stages I/II combined (as stage I was rare), stage III, stage IV and unknown stage; (iii) age at diagnosis (< 40 years, 40–49, 50–59, 60–69 and ≥ 70); (iv) country-specific socio-economic position (SEP) categories (low, middle and high) which were constructed based on thirds of each country’s distribution of a SEP score derived from combining the following self-reported possessions and facilities: home ownership, indoor water, flush toilet, electricity, vehicle, refrigerator, landline, gas or electric stove and a bed; (v) belief in traditional medicine (yes/no); (vi) employment status as highly skilled/skilled or unskilled/not applicable (containing, e.g. the informal work sector and housewives); (vii) body mass index (BMI) derived from measured height and weight at recruitment (five categories: < 18.5, 18.5 to < 25, 25 to < 30, ≥ 30 kg/m² and unknown); (viii) residential area, i.e. living in an urban (city/town) or rural (village/rural); (ix) previous breast cancer knowledge (ever/never heard of the disease prior to current disease onset); (ix) belief in traditional medicine (yes/no); (xi) belief in spiritual healing (yes/no) and (xii) self-reported HIV status (yes/no).

At the 6-month follow-up interview, we also ascertained a woman’s, and her family’s, out-of-pocket healthcare costs in the past 3 months in the two countries without free cancer care, i.e. Nigeria and Uganda. Answers were provided in the local currency and subsequently converted into US dollars.

Among the 1335 women included, mean age at breast cancer diagnosis was 50.7 years (SD = 13.6, Additional file 1: Table S1). Overall, 597 (45%) of women were diagnosed at stage III and 204 (15%) at stage IV, with a markedly improved stage distribution in Namibian non-black women. Most women had unskilled jobs (n = 923; 70%), and other than in Uganda, where 74% (n = 320) of women lived in rural areas, the majority resided in urban areas. Nine percent (n = 121) were HIV-positive. A higher percentage of women reported believing in conventional medicine (94%) than in spiritual healing (66%) or traditional medicine (24%).

Treatment information was known for 1325 (99%) of women, of whom 83% (n = 1098) had initiated breast cancer treatment within 1 year of diagnosis and 17% (n = 227) had not (Table 2). These percentages varied greatly between settings, with the highest percentage of untreated women in the two regional hospitals in Nigeria, at 38% and 34% in the public and private settings respectively, 17% at the national referral hospital in Uganda and was near zero in the Namibia Cancer Care Centre (0% in non-black women and 1.3% in black women; Fig. 1). Treatment was mostly in the form of surgery and/or chemotherapy, whilst treatment also included radiotherapy for two thirds of women at the Namibian Cancer Care Centre, 15% of women in Uganda—during a period when the radiotherapy machine was not in operation for 18 months—and < 5% of women at the Nigerian regional hospitals.

Odds ratios of associations with receipt of cancer treatment are provided in Table 2, and those for age, stage and SEP are plotted in Fig. 2.

In Nigeria, percentages of untreated women increased linearly with more advanced cancer stage, with 23% (absolute) more untreated women in stage IV vs stage I/II, whereas in Uganda only stage IV (and not stage III) had a higher (by 3%) percentage of untreated women. For stage III cancer, this association was fully attenuated after adjusting for SEP, BMI and residential area, thus in the fully adjusted model, as expected, only stage IV women had lower treatment odds (stage IV v stage I/II: 0.63, 95% CI 0.37, 1.07, Table 2). In both settings, among the 8% women with missing stage information, untreated percentages were particularly high, at 51%, which was not explained by adjustments made (OR 0.26; 0.14, 0.45).

An inverse U-shaped association between age and treatment was observed in both settings, with percentages receiving treatment being highest among 50–59-year-old women. Women aged 70 and over, comprising 6% of patients, were least likely to be treated, with an absolute difference of 15% in the Nigerian public setting and 10% in Uganda (OR 0.48, 95% CI 0.25, 0.93) (Additional file 3: Figure S1, Additional file 2: Table S2). For young women (< 40 years), who represented 26% of patients, treated percentages were 4% less in Uganda and 8% less in Nigeria (stage and age-adjusted OR 0.70; 95% CI 0.44, 1.10). Further adjustment for BMI, SEP and area of residence partially attenuated the associations, but the slight inverse U-shape remained. These associations held when restricted to HIV-negative women and were not restricted to women with low SEP or later-stage disease; instead, they were slightly stronger in early-stage disease and in medium and high SEP groups (data not shown).

A consistent and marked socio-economic gradient in the percentage of untreated women was seen (Fig. 3). Absolute percentages of untreated women in the lowest vs highest SEP tertiles were 21% vs. 11% in the Uganda national referral hospital, 41% vs. 21% and 43% vs. 28% in the two Nigerian regional settings (Additional file 3: Figure S1, Additional file 2: Table S2). This SEP-treatment association was unaltered after further adjustment for BMI and residential area (fully adjusted OR 2.45; 95% CI 1.53, 3.90; Table 2). The association with skilled/unskilled employment was in a similar direction, albeit not statistically significant, whilst there was no clear association between residential area and treatment.

Obese (BMI > 30 kg/m²) and, to a lesser extent and non-significant, underweight (BMI < 18.5 kg/m²) women each had a higher odds of receiving breast cancer treatment compared to those with a normal BMI (18.5–25 kg/m²) (age and stage-adjusted OR 1.76; 95% CI 1.10, 2.81 and 1.58; 0.70, 3.59) respectively; Table 1). Further adjustment for SEP and residential area slightly weakened the obesity association (fully adjusted OR 1.53; 95% CI 0.95, 2.47) but strengthened the underweight association (1.83; 0.79, 4.21). Regarding HIV, after full adjustment, HIV-positive women had one third of the odds of being treated relative to their HIV-negative counterparts (Table 2), reflecting the Ugandan situation, yet almost all Namibian black women, the group with the highest (14%) HIV prevalence (Additional file 1: Table S1), received breast cancer treatment. In Uganda, the total percentage difference between positive and negative women was 16%.

Health beliefs affected treatment received, but the effects were setting-specific (p values for the interaction in Table 2). In Uganda, where 38% of women reported believing in traditional medicine, such women were less likely to receive treatment (27% untreated vs 13% untreated among non-believers (Additional file 2: Table S2); fully adjusted OR 0.39; 95% CI 0.22, 0.66 (population group-specific ORs not shown)). Further, in the 6-month follow-up interview, 40% of these women stated that they were currently attending a traditional healer for treatment. These were eight times more likely not to have received conventional breast cancer treatment (7.84; 3.64, 16.91). In contrast, in the Nigerian private hospital, the odds of being treated for breast cancer was higher among women who believed in traditional medicine (fully adjusted OR 5.1; 95% CI 1.1, 24.7) or in spiritual healing (6.5; 1.2, 34.5) than among their non-believer counterparts, with the wide CIs reflecting the small sample size. Finally, previous breast cancer knowledge was not associated with treatment overall, but had an effect in Uganda, where untreated percentages were lower in women who knew that breast cancer was curable than those who did not (13% v 21%).

Self-reported information on treatment gaps was available for 131 women who did not undergo surgery despite medical recommendation; the reasons for refusing surgery were cost (47%, n = 62) and personal decision (e.g. due to lack of belief in its effectiveness, fear or non-compliance to or rejection of therapy) (31%, n = 40). In addition, for 55 women, a medical reason for not getting surgery was recorded in the clinical notes (e.g. too old and frail, tumour too advanced).

Healthcare expenses were paid out-of-pocket by the patient herself (ranging from 42% in Uganda to 67% in Nigerian public hospitals) or by family and friends (65%, all settings). Median household healthcare expenses in a 3-month snapshot period increased with increasing SEP (Fig. 3B1-B3: range $196 to 270 in Uganda, $56 to 224 in Nigerian public and $102 to 140 in Nigerian Private hospitals) and tended to be associated with increasing percentages of treated women, particularly so for Nigerian public hospitals (Fig. 3B2). Although only asked for a 3-month interval, it is likely that the cumulative costs will exceed the annual gross national income per capita of 600 USD in Uganda and 2100 USD in Nigeria (2017 census; https://​data.​worldbank.​org). Untreated women also incurred healthcare expenses, with median household expenses of 83, 20 and 55 USD, respectively, for women in Uganda, Nigeria public and Nigeria private hospitals (all SEPs combined).

The treatment divide between ABC-DO countries reflects, to a large extent, the marked differences in healthcare systems and out-of-pocket costs to women and their families. In Namibia, where cancer treatment is free, almost all women attending the cancer care hospital were willing to be treated. In contrast, in Uganda and Nigeria, cancer care costs are paid out-of-pocket by the patients, often with the help of relatives and friends. Accordingly, SEP was the most influential determinant of treatment access, in line with, but of larger magnitude than, other studies linking social status to cancer outcomes [14‐17]. Expenditure levels amounted to financial catastrophe, which have also been reported for other LMIC, mainly in the Asian region [8, 9].

We found slight inverse U-shaped relationship of treatment with age, which, to the best of our knowledge, has not been reported so far and warrants investigation in other settings. Elderly women may not receive treatment due to their frail physical condition or patients’ preference [18, 19]. The lower treatment percentage seen in younger women is of concern and possibly arises in these settings due to stigma, fear of disfigurement and abandonment. Preference for traditional medicine also prevented women from initiating cancer treatment in Uganda. Feelings of stigmatisation, hopelessness and fatalism are commonly reported by patients diagnosed with breast cancer in SSA [20, 21]. Such feelings, together with lack of knowledge of the disease, contribute to rejection of allopathic cancer treatments and preference for traditional medical practitioners (TMPs), who are often cancer illiterate and do not take the adequate action [22].

Increases in life expectancy of HIV-positive women combined with rises in the incidence of breast cancer leads to an increasing number of women in SSA with the double HIV-breast cancer burden. In Uganda, HIV-positive women were less likely to get treatment than their HIV-negative counterparts, but the reasons for this were unknown. A personalised treatment management plan is needed to minimise possible drug-drug interactions of chemotherapy with ARVs in the HIV-positive breast cancer patient [23]. Non-medical reasons for the lower treatment proportions rates might include stigmatisation as well as poor treatment literacy of the patients and health care workers.

A limitation of the present study is the difficulty to disentangle whether untreated proportions are rather attributable to a country or the specific hospital setting in this specific country. This is due to the hospital-based nature of ABC-DO, and hence, the treatment proportions reported reflect those among women who reach this level of the cancer care system in a given country. This is particularly relevant for the Namibian specialised Cancer Care Centre, where the high treated percentages are among women who have already made the journey to this national referral centre. Nevertheless, the high treated proportions in this centre also reflect willingness of women to receive treatment when the burden of treatment costs is alleviated from the patient. Further, the national hospital referral network in Namibia might heighten the chance of a patient to reach cancer care at a specialised centre. In every setting, breast cancer patients who do not reach this level of the health system and may be more likely to go untreated, are not included. Such patients may be substantial in number as various patient-level and health system-level barriers prolong the journey to breast cancer diagnosis [12], thus overall untreated proportions may be even higher. We do, however, expect that the profile of women vulnerable to under/no treatment are captured in the associations observed.

The study identified marked inequities in access to breast cancer treatment in SSA settings. The much lower treatment rates in Uganda and Nigeria, when compared to Namibia, underline the importance of providing population-wide affordable cancer care through universal healthcare coverage (UHC) and a national hospital referral network. Even in the absence of UHC, knowledge that lower social groups are vulnerable to treatment gaps should be addressed, and navigational, financial and educational support provided to reduce these. Appropriate treatment facilities across the continuum of cancer diagnosis and care are currently lacking in most SSA countries [15, 24, 25]. In a mathematical cost-effectiveness modelling of cancer treatment in SSA, BC treatment of all stages has been shown cost-effective if also accompanied by early detection programs [26]. In particular at earlier stages, treatment is more cost-effective, as treatment is less expensive with higher chances to cure. Costs of BC treatment in Africa were estimated at $78 per disability-adjusted life year (DALY) for treating early-stage breast cancer with surgery and radiotherapy vs. $4986 per DALY for treating metastatic BC with systemic chemotherapy. In this context [27], the high percentage of patients who were not treated within the first year of diagnosis though diagnosed at earlier stages, especially stage II, needs to be urgently addressed.

The present results also highlight the need to consider how to integrate the informal health care sector into cancer care, particularly in countries where traditional medical practitioners (TMPs) serve a large part of the population. TMPs are a significant source of health care delivery in cancer management in SSA [22]. On average, one TMP serves 400 persons, contrasting with one allopathic practitioner per 12,000 persons and two surgeons per 100,000 persons [28]. In a survey asking TMPs about cancer management, their literacy was still limited: the majority of TMPs recognised cancer, but only at advanced stages [22]. Further, most TMPs believed that they could treat cancer; thus, appropriate referral for the specialised diagnosis and care needed for breast cancer would be delayed. Thus providing training courses on breast cancer awareness and recognition, and the importance of immediate hospital referral and treatment initiation and completion, is a potential promising approach to significantly improve access to breast cancer diagnosis and treatment in SSA. Such educational programs need to include the affected breast cancer patients themselves, but also TMPs and conventional health care providers at the frontline and at key referral nodes to both accelerate the route to diagnosis and increased treatment access.