Delayed presentation and diagnosis of breast cancer in African women: a systematic review

1Women in Africa currently have one of the lowest incidence rates of breast cancer worldwide. However, the burden from this cancer is expected to increase markedly in the next decades. A growing aging population alone, that is, assuming incidence rates will remain constant, will lead to estimated 119,918 new cases in 2030, a near doubling in the number of incident cases over 20 years. The increase will be even more marked as incidence rates are likely to rise due to the adoption by African women of more westernized lifestyle profiles, particularly reproductive patterns characterized by late age at first full-term pregnancy, lower parity, reduced lifetime, breastfeeding duration, and increases in postmenopausal weight.

2Despite breast cancer incidence rates being still relatively low in Africa, mortality rates from this disease are as high, or higher, than in high-incidence countries due to poor survival. Furthermore, the proportion of breast cancer cases and deaths at premenopausal ages is higher in Africa than in high-income countries (HICs), where disease incidence is highest, reflecting the younger age structure of the continent's population and possibly also distinctive risk factors and/or tumor characteristics. Consequently, breast cancer in Africa disproportionately affects women in the prime of their lives, and hence, it has particularly marked familial, societal, and economic consequences.

3A recent systematic review shows that a high proportion of breast cancer patients in sub-Saharan Africa (SSA) are diagnosed with late-stage disease leading to poor survival. Studies from HICs have shown that delays between onset of symptoms and start of treatment are main determinants of late-stage presentation and poor survival. Previous studies have attempted to examine delays in breast cancer presentation, diagnosis, and treatment in Africa,, but, to our knowledge, these have not been comprehensively investigated across the continent. Knowledge of the length of time intervals between symptom recognition, presentation, diagnosis, and start of treatment—and of the factors that may influence them—is the key to the development of strategies to shorten them. Therefore, we conducted a systematic review to investigate delays in presentation and diagnosis of breast cancer in Africa and their determinants.

4Figure 1 depicts a patient's trajectory from the moment she first notices symptom(s) to the time when treatment starts and the factors that may affect her journey. In HICs with free universal access to health care, the delay from a woman first noticing potential symptoms of breast cancer to her presentation to a health care provider is labeled as “patient delay” as it is essentially driven by patient-mediated factors. In contrast, the time from first medical consultation to the beginning of definitive treatment is labeled as “provider delay” as it is driven predominantly by health system–mediated factors. However, in many African settings, the picture is likely to be far more complex as delays in both presentation and diagnosis are likely to result from a complex interplay between patient-mediated and health system–mediated factors. For instance, a woman may delay presentation not only because of her lack of breast cancer awareness but also because of the unavailability of health care providers in her area of residence. Similarly, a woman who first presents with a suspicious cancer may delay diagnosis due to fear of its consequences (e.g., mastectomy, death). In this review, we will consider presentation delays as the time interval from symptom recognition to presentation to the first health care provider, diagnostic delays as the time interval between presentation and breast cancer diagnosis, and treatment delays as the time interval between diagnosis and start of cancer treatment. These terms do not carry any judgment on whether these delays are primarily induced by patient-mediated or provider-mediated factors.

5The Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement guidelines were followed to select relevant publications on delays in breast cancer presentation and diagnosis in Africa. Articles were eligible for inclusion in the systematic review if they reported findings from primary research studies conducted in Africa; reported on delays in presentation and/or diagnosis of female breast cancer patients; and were published between January 1, 2000 and May 31, 2016. No language restrictions were imposed. Relevant publications were searched in the electronic databases MEDLINE, Embase, and Global Health. A search strategy using synonyms (including truncations) and subject headings of the search concepts “breast cancer,” “late diagnosis,” “Africa,” and “determinants,” and the Boolean operators “AND” and “OR” was used (Appendix A). All titles and abstracts were screened to identify potentially eligible articles and the full text for these retrieved and critically reviewed articles to assess eligibility and, if eligible, to extract relevant data.

6The data extraction from each eligible article was carried out independently by two reviewers (C.E. and I.d.-S.-S.) using a specifically developed standardized data extraction form. The following information was extracted: the type of catchment population (e.g., country, urban, rural, or mixed); the study design (quantitative, qualitative, mixed); the type of recruitment source (primary, secondary, or tertiary hospital/clinic) and approach (eligibility criteria, recruitment period, type of sample: consecutive or convenience, i.e., opportunistic, sample size); patient (e.g., age) and tumor characteristics (e.g., stage, size, histology, symptoms); source (e.g., patient, medical records) and timing of collection (e.g., before or after diagnosis) of data on delays and their reasons; reported times between symptom recognition, presentation, diagnosis, and start of treatment; and patient-mediated and health system–mediated factors that might have influenced them. Disagreements between the two reviewers were discussed, and a consensus was reached.

7The quality of the articles included in the review was assessed independently by the same two reviewers. A standardized quality assessment form was developed, which included parameters to assess the potential for selection and information bias as well as the appropriateness of the analytical methods used, including those for dealing with potential confounders (Appendix B). The overall quality score of an article was expressed as the sum of its parameter-specific scores, which could range from 0 (lowest) to 30 (highest). The higher the score, the higher the methodological quality of the article; the lower the score, the more likely its findings might have been affected by biases.

8Data were synthesized in narrative, tabular, and graphical forms. Study-specific mean (SD) or median (range), presentation, diagnosis, and treatment delays are presented; if only categorical data were reported in the original publication, we used them to estimate the median, or a weighted mean, whenever possible. Studies differ greatly in the way they obtained information on potential reasons for delays and in the way such data were presented (Appendix C). Most studies simply presented data in a descriptive way (e.g., percentages), but a few used logistic regression methods to estimate crude and/or adjusted odds ratios for delayed presentation, diagnosis, or treatment for each variable examined, with studies using different cutoff points to define such delays (e.g., from ≥2.2 to >6 months for delay in presentation and from >2 weeks to ≥6 months for delays in diagnosis; Appendix C). One study in North Africa reported on delays but only examined factors associated with late stage (III/IV) versus early stage at diagnosis; late stage was taken here as a proxy for delays between symptom recognition and diagnosis. Findings are shown separately for studies conducted in North Africa (i.e., in Algeria, Egypt, Libya, Morocco, Sudan, Tunisia, and Western Sahara) and SSA (i.e., countries in East, Middle, South and West Africa) as defined by the United Nations.

9A total of 315 articles (after removal of duplicates) were identified through electronic searches and their titles and abstracts were screened for potential eligibility (Fig. 2). In all, 35 articles were retrieved for full-text review. Of these, only 21 were eligible for inclusion in the review: 16 quantitative studies, three qualitative studies, and two mixed studies (quantitative and qualitative).

10Table 1 summarizes the main characteristics of each participating study. Of the 18 quantitative and mixed-design studies, 8 (44%) were conducted in North Africa and 10 (56%) in SSA, with their sample sizes ranging from 44 to 350. In contrast, all three qualitative studies were conducted in SSA, with sample sizes ranging from 9 to 31. All studies were hospital-based cross-sectional surveys that relied on consecutive samples of patients, except for two small qualitative studies, that relied on convenience samples. Eligibility was restricted to women with advanced breast cancer in one study in North Africa and in four studies (three quantitative,, and one qualitative ) in SSA. The large majority of studies recruited breast cancer patients diagnosed predominantly in the years 2000–2010, but two studies in North Africa, and two in SSA, included patients diagnosed after 2010, whereas one study in SSA recruited patients diagnosed before 2000 (Table 1). The average (mean/median) age at breast cancer diagnosis was in the 40s in the large majority of studies. Most studies involved collection of data through structured or semistructured questionnaires, usually administered by the researchers or medical staff around the time of diagnosis, but four studies were conducted retrospectively using medical records,,,. Information on ethnicity was provided in only one study, which stated that its subjects were all black. Information on tumor stage at diagnosis was available for seven (88%) studies in North Africa and nine (69%) in SSA. Among studies with stage information and whose subject eligibility was not dependent on it, the proportion of patients with late stage (III/IV) was very high (range: 46%–61% in North Africa; 76%–91% in SSA; Table 1).

11Quality scores were low for most quantitative studies (Table 1) albeit slightly higher for those from North Africa (median = 18.5; range: 14–25) than for those from SSA (median = 17.5; range: 7–25). Similarly, the quality of the qualitative and mixed-design studies varied substantially, with three studies presenting more in-depth qualitative results,,.

12The time interval between symptom recognition by the woman to presentation, that is, to first visit to a health care provider, varied substantially across studies but, overall, it was shorter in North Africa than in SSA (Table 2; Fig. 3A). Of the five North African studies that reported on presentation delays, most yielded median estimates of less than 2.5 months; the only exception was a study in Libya with a median presentation time of 4 months. Of the five studies in SSA that provided estimates of time from symptom recognition to presentation, only one reported a median time of less than 2.5 months, with the remaining reporting average times ranging from 3.4 months in Mali to greater than 6 months in South Africa.

13Fewer studies in North Africa,, and in SSA,, gave estimates of the time between presentation and diagnosis, or between diagnosis and start of treatment. Nevertheless, the length of these intervals tended to be shorter than the length of the corresponding intervals between symptom recognition and presentation in North Africa (all <1 month) but not in SSA (Fig. 3B).

14Five North African studies provided median estimates of the total delay from symptom recognition to date of breast cancer diagnosis or start of treatment (Fig. 3C). Two of these studies recruited only advanced breast cancer cases with average total delays of 8 and 12 months. Median estimates of the total delay from symptom recognition to diagnosis for the remaining three studies ranged from 4 to 8.5 months. Five SSA studies provided average times from presentation to diagnosis or start of treatment (Fig. 3C), with their estimates ranging from 7.9 months in Ghana to 15 months in Rwanda; median delays were known to be greater than 6 months for two studies,, but their exact values could not be estimated. In addition, a small qualitative study (n = 11) in Botswana reported a median time from first symptom(s) to presentation at the hospital where the diagnosis was finally made of 3 years.

15The number of health care providers visited before the one where the diagnosis was made was reported by only one study in North Africa and four in SSA,,,, with estimates ranging from a median of 1.5 in Egypt to greater than 5 in Rwanda; however, these estimates are not entirely comparable because traditional and religious healers were included in two of these studies,.

16A few studies examined whether delays were associated with late stage (III/IV) at diagnosis. The study by Benbakhta et al. in Morocco reported a 6.81-fold (95% confidence interval [CI], 3.65–12.7) increase in the odds of late stage among patients who delayed presentation by greater than 64 days relative to those who presented less than or equal to 64 days of symptom recognition. Similarly, the odds of late stage among patients who experienced a diagnostic delay of greater than or equal to 50 days were 1.84 (95% CI, 1.05–3.23) times higher than among those diagnosed less than 49 days of their first presentation to a health care provider. The study by Mousa et al. in Egypt also reported an association between late stage and delays in presentation greater than 3 months (crude odds ratio: 1.99; 95% CI, 1.01–1.99) but not with delays in diagnosis greater than 2 weeks. In Rwanda, late stage was positively associated with both presentation (median [range] in months: 2 [1–12] for stages I/II, 5 [1–13] for stage III, and 9 [3–18] for stage IV; P =.09) and diagnostic delays (4 [2–13] months for stage I/II, 4 [2–10] for stage III, and 11 [5–28] for stage IV; P =.005).

17Appendix C summarizes the reasons most commonly reported by the quantitative studies in the review for late presentation to the first health care provider. They fell into the following categories: (1) socioeconomic factors such as low educational level; (2) lack of breast cancer awareness and poor knowledge of early-detection methods (e.g., breast self-examination); (3) type of initial symptoms: painless, not taken seriously, or hoping they would resolve soon; (4) fear of the disease, its treatment (e.g., mastectomy) or death, or of being a burden to the family; (5) belief in traditional medicine or spiritual cures; (6) financial constraints; and (7) poor access to health care (e.g., living too far away from a health care provider; lack of transportation). Benbakhta et al. found in mutually adjusted analysis that a delay in presentation of greater than or equal to 2.2 months in Morocco was positively associated with low socioeconomic conditions (e.g., living in a rural area, being illiterate, being a housewife [vs. being employed], and having low socioeconomic level) and lack of breast cancer awareness (e.g., negative family history of cancer, no knowledge of breast self-examination) (Appendix C). In contrast, Mousa et al. found no association between delay in presentation greater than 3 months in Egypt and a woman's socioeconomic characteristics or type of symptoms before or after adjustment for potential confounders. In South Africa, Marcus et al. found in mutually adjusted analysis the positive associations with late presentation (>6 vs. 3–6 months) with increasing age and a previous cancer diagnosis but not with educational level, marital status, or being employed/unemployed. A mutually adjusted analysis of data from a study in Rwanda revealed a four-fold to five-fold increase in the odds of late presentation (≥6 months) for patients with low or no education and for those who visited a traditional healer first but no independent associations with other socioeconomic factors, breast cancer awareness, symptom, or health services–related variables (Appendix C). Overall, the findings from the qualitative studies supported the evidence from the quantitative studies,,, (Appendix C).

18The reasons given by the patients for delays between presentation and diagnosis, or start of treatment, included patient-mediated factors (e.g., socioeconomic factors, type of symptoms, having tried traditional treatments first, financial problems, fear of the disease and/or its treatment, and denial) and health care provider–mediated factors (e.g., travel time to health care provider, the number and type of health care providers contacted before diagnosis, delayed referrals or nonreferrals, misdiagnosis, wrong advice or false reassurances, delays in obtaining diagnostic confirmation, and in starting treatment) (Appendix C). The study in Morocco by Benbakhta et al. found in mutually adjusted analyses that a delay greater than 1.7 months between presentation and start of treatment was associated with older age, illiteracy, low socioeconomic level, distance to health care provider greater than or equal to 100 km, and greater than or equal to 3 consultations before the diagnostic one. Mousa et al. in Egypt showed that after adjustment for potential confounders, the odds of a delay greater than 2 weeks from the first medical consultation to arrival at the diagnostic center were not associated with the patient's age, socioeconomic conditions, or type of symptoms but were strongly associated with the type of the first health care provider visited and the navigation pathway followed by the patient (Appendix C). In Rwanda, Pace et al. found in mutually adjusted analyses a 2.69 (95% CI, 1.24, 5.84) higher odds of a delay greater than or equal to 6 months for patients who visited five or more health care facilities before diagnosis but no associations with the patient's socioeconomic conditions, reproductive history, or type of symptoms. In the qualitative studies (Appendix C), some women reported poor clinical practices (e.g., inadequate diagnosis by general doctors, hospital strikes, or having sought alternative care after receiving the diagnosis).

19To our knowledge, this is the first systematic review of studies that reported on delays in a woman's breast cancer journey in Africa. Its findings highlighted three main issues. First, there is a paucity of published data on delays in the presentation and diagnosis of the most common female cancer in Africa. The systematic review identified only 21 published studies over the 16-year period (January 2000 to May 2016), comprising only 2788 breast cancer patients from across the continent (1382 from North Africa; 1406 from SSA). Second, the findings revealed marked delays in presentation and diagnosis of breast cancer patients in both North Africa and SSA. Third, the reported reasons for such delays were complex and included both patient-mediated and health system–mediated factors; however, the relative importance of these two types of factors varied from setting to setting.

20There is strong evidence that a delay from symptom recognition to diagnosis of more than 3 months is associated with later stage at presentation and poorer survival. This review revealed substantially longer delays in both North Africa and SSA, with reported average times from symptoms recognition to diagnosis between 4 and 15 months. These estimates are in line with those observed in other low- and middle-income countries (LMICs) (e.g., 7.6 months in Brazil, 5.5 months in Malaysia ) but much higher than those observed in HICs (e.g., 34 days in France, 48 days in the United States ). The very long time intervals from symptom recognition to diagnosis in Africa resulted from delays in both presentation and diagnosis. All studies in this review, with the exception of two,, reported average presentation intervals between 2.2 months and greater than 6 months, much longer than those observed in HICs (e.g., 9 days in the United Kingdom; 16 days in Germany ). Similarly, reported diagnostic intervals in Africa were much longer than those found in HICs (e.g., from 10 to 42 days in France, Germany, and the United States ) but similar to what has been described for other LMICs (e.g., median of 5 months in Brazil, Colombia, and Mexico ).

21As we had hypothesized in our conceptual model, delays in presentation in Africa were found to be associated not only with patient-mediated factors (e.g., low educational level, poor breast cancer awareness, use of alternative care medicine) but also with health service–mediated factors (e.g., distance to the nearest health care center). These results are similar to those from previous studies—for example, being unaware of the warning signs or tests for breast cancer, patients only seeking conventional care when traditional treatment has failed, or inability to afford the costs of treatment. Similarly, delays in diagnosis in Africa were influenced by both patient-mediated factors (e.g., low educational level, financial problems) and health system–mediated factors (e.g., type of first health care provider visited, number of providers visited before diagnosis, type of navigation pathway followed before reaching the diagnostic center). A high number of referrals make the patient's journey through the health system longer resulting in a more advanced tumor stage at diagnosis; however, it is also conceivable that a low number of referrals might reflect a more aggressive tumor, or a longer time interval before presentation to the first health care provider, and thus a more advanced tumor that was easily identified by the physician. Of note, however, is the fact that none of the articles directly examined health system factors, for example, through interviews with health care providers, relying instead on patients' reports.

22Major strengths of this review include the systematic search strategy used to identify eligible English and non-English publications and the use of standardized methods for data extraction and synthesis. The review also has weaknesses. Its representativeness may have been compromised by several factors. First, publication bias cannot be excluded as gray literature was not included in this review. Second, the review included studies from only 4 of 7 North African countries and 11 of 51 SSA countries, albeit the latter comprised studies from all four SSA regions (i.e., from Eastern, Western, Southern, and Middle Africa). Third, none of the studies in the review were population based; they were all hospital based, predominantly from tertiary hospitals as these are the only ones in most African countries to have appropriate cancer diagnostic and treatment facilities. However, such studies excluded, by design, the large number of patients who never reach tertiary hospitals, some of whom are never diagnosed. Hence, the included patients who reached tertiary facilities are unlikely to be a representative sample of all breast cancer patients in Africa.

23The methodological quality of most articles was low. In particular, measurement errors may have affected the validity of the review's findings as although most of the studies recruited women prospectively, patients were asked to remember the time from first symptom(s) to presentation, and this might have introduced recall errors and even biases. Little detail was provided in the original articles on the specific instruments used to collect information and the methods used to estimate times to presentation, diagnosis, and treatment, including on the way questions to patients on time intervals were formulated and on how relevant time-related events (e.g., dates of contact with a first health care provider, breast cancer diagnosis, and start of treatment) were defined. Between-study differences in these methodological issues may have affected their comparability. When questioned about the reasons for delays, patients might have been reluctant to admit less orthodox behaviors such as the use of traditional medicine. Reassuringly, however, the studies that examined associations between self-reported delays and late stage at diagnosis showed, as expected, strong positive associations. Many studies had relatively small sample sizes, and thus, their ability to precisely quantify delays and their power to detect associations were limited. There were large variations across studies in the way data were analyzed (e.g., only a few quantitative studies attempted to control for confounders; none of the qualitative studies conducted theoretical analyses) and summary findings presented, hampering between-study comparisons, and precluding the conduct of meta-analyses.

24Several studies in Africa have shown that early-stage breast cancer is associated with better survival than late-stage disease,, consistent with early diagnosis and treatment being associated with reductions in mortality from this disease in the region. The long presentation and diagnostic delays identified by this review indicate that there is considerable potential to introduce interventions aimed at shrinking the time intervals between symptom recognition and diagnosis. Mammography screening is often advocated as the best intervention to improving early diagnosis of breast cancer, but the findings from this review strongly argue against adopting such an approach in African settings. Screening can only reduce breast cancer mortality if women with suspicious screen-detected lesions have access to appropriate diagnosis and treatment. Despite the limitations of the existing data, and the high heterogeneity across African settings, the long diagnostic delays highlighted by the review indicate that the addition of women with asymptomatic screen-detected tumors would place significant additional burden on most, already overstretched, health care systems in the region. Instead, downward-stage migration of symptomatic breast cancer should be the priority in most settings as recommended by the Breast Health Global Initiative and the Breast Cancer Initiative 2.5. Achieving this would require increased breast cancer awareness of the population, enhanced ability of primary and secondary health care professionals to diagnose breast cancer, as well as clear patient navigation pathways to facilitate timely referral and admission of patients to tertiary care services for early care. The introduction of such an approach in other LMICs has demonstrated that downward-stage migration of breast cancer is achievable in the absence of screening.