Part XVI · Disparities, access, and health systems · Chapter 91

Global breast cancer care

Five-year survival for the same disease runs from under 40% to near 90%. Almost none of that spread is biology.

1 · The epidemiologic transition and the shifting global burden

Five-year survival for the same disease runs from under 40% to near 90% depending on where it is treated. Almost none of that spread is biology.

Incidence and mortality do not vary together, and the discrepancy is the whole argument of this chapter. In 2020 there were over 2.3 million new cases of breast cancer worldwide and 685 000 deaths. By 2040, on population growth and ageing alone, the projection is over 3 million new cases and 1 million deaths each year¹.

Age-standardised incidence ranges from under 40 per 100 000 women in some Asian and African countries to over 80 per 100 000 in Australia, New Zealand, Northern America and parts of Europe. Geographic variation in mortality is much smaller than variation in incidence. Countries with lower incidence carry a disproportionate share of deaths¹. A country can have half the incidence of another and a similar death rate.

The epidemiologic transition explains the incidence trend. As countries move through demographic and lifestyle change, later first birth, fewer births, shorter breastfeeding, higher body mass index and longer life expectancy all push breast cancer incidence up. Incidence in transitioning countries is rising fastest, which means the burden is growing where survival is lowest.

The measurement caution is substantial and rarely stated. Incidence estimates for many countries are modelled rather than registered. Only a minority of countries have national population-based cancer registries with mortality linkage, and where registration is partial the incidence estimate depends on assumptions rather than counts. This is why registry strengthening appears as an intervention in national cancer control plans rather than only as a research tool. The burden estimates in Epidemiology and the changing burden of disease carry the same caveat.

2 · Late-stage presentation and the determinants of delay

The survival spread is largely a stage spread. The World Health Organization estimated population-based age-standardised five-year net survival for women diagnosed with breast cancer between 2017 and 2021 across all 194 Member States. Median five-year net survival ran across the six WHO regions as follows: 39.1% in the African Region, with an uncertainty interval of 34.1% to 44.7%, 61.0% in the Eastern Mediterranean Region, 66.3% in South-East Asia, 81.1% in the Western Pacific, 84.0% in the European Region and 88.5% in the Region of the Americas². The earlier registry-based estimates showed the same shape, with five-year survival of 90.2% in the United States and 66.1% in India for women diagnosed between 2010 and 2014³.

Concrete stage data show where the spread comes from. At a large tertiary hospital in Ghana, of 214 patients with breast cancer who underwent staging in 2021, 10.3% were stage I or II, 50% were stage III and 39.3% were stage IV⁴.

Delay is conventionally split into three intervals, and they have different causes and different remedies.

The patient interval runs from noticing a symptom to first contact with a health worker. It is driven by symptom recognition, by the belief that cancer is untreatable, by cost of the first consultation and by the availability of alternative providers.

The diagnostic interval runs from first contact to confirmed diagnosis. It is driven by referral pathways and by the availability of imaging and pathology, which is why Pathology and diagnostic infrastructure as the rate-limiting step treats pathology as rate-limiting.

The treatment interval runs from diagnosis to treatment start, and beyond it sits treatment abandonment, which has no counterpart in high-income datasets. In the Ghanaian series, all patients were diagnosed within 31 days of presentation, and only 45 of 243 patients with confirmed breast cancer, which is 18.5%, completed recommended treatment without abandonment⁴.

That pairing is the finding. The diagnostic interval met the international benchmark. Completion did not come close. A programme that had optimised only diagnosis would have reported success.

3 · Resource-stratified guidelines and tiered recommendations

A guideline written for a maximal-resource setting is not merely unachievable elsewhere. It is actively unhelpful, because it gives no ordering for what to acquire first.

Resource stratification solves a different problem from guideline adaptation. Adaptation asks what of this guideline can we do. Stratification asks, given a defined level of resource, what sequence of additions produces the greatest gain in outcome per unit of resource. The output is an ordered list rather than a filtered one.

The method that produces such a list has a specific shape. Interventions are ranked by the contribution each makes to clinical outcome rather than by their position in a conventional care pathway⁵. That ranking frequently inverts the order familiar from high-income practice. Reliable oestrogen receptor testing and tamoxifen supply outrank a wider chemotherapy formulary. Diagnostic imaging to support tissue sampling outranks population screening. Preventing treatment abandonment outranks adding a new agent.

Stratified guidelines are also implementation documents rather than only clinical ones. They specify a phased sequence, so that a ministry of health can cost a step and then cost the next one⁶. A guideline that lists everything at once cannot be budgeted against.

The failure mode of this approach is worth naming. A tier is a planning category and not a description of a country. Within any country, tertiary centres and district hospitals sit at different tiers, and a national designation conceals that. The tiers themselves are set out in Basic, limited, enhanced, and maximal resource tiers.

4 · The WHO Global Breast Cancer Initiative and its three pillars

The Global Breast Cancer Initiative was launched by the World Health Organization to coordinate efforts to reduce breast cancer mortality, particularly in low-income and middle-income countries⁷. Its framework organises the whole of breast cancer control into three pillars, each with one key performance indicator and one numerical benchmark.

The benchmarks are, in order: at least 60% of breast cancers diagnosed at stage I or II, all suspected breast cancer diagnosed within 60 days of presentation to the health system, and at least 80% of patients completing recommended treatment⁸. The stated goal is a 2.5% reduction in breast cancer mortality each year⁴.

Three properties of this design deserve attention, because they are what make it different from a guideline.

The indicators are health system measurements rather than clinical ones. None of them names a drug or a procedure. A country can meet all three with a basic-tier formulary and fail all three with a maximal one.

The three are sequential. Early diagnosis is useless without timely workup, and both are useless without treatment completion. A programme is therefore assessed on the weakest link rather than on its best-performing component.

The benchmarks are thresholds rather than targets to be approached. That is a deliberate choice, and it makes the indicators blunt. A country moving from 20% to 45% of cancers diagnosed at stage I or II has made a large improvement and still fails the indicator.

How far current performance sits from these thresholds is not hypothetical. The African Breast Cancer and Disparities in Outcomes cohort covered Nigeria, Uganda, Zambia and two population groups in Namibia. Achievement of the first two indicators ranged from 22% to 49% across those groups. Achievement of the third was below 30% in every group⁸.

5 · Basic, limited, enhanced, and maximal resource tiers

The Breast Health Global Initiative stratification assigns every intervention to one of four levels, defined by the contribution each resource makes to improving clinical outcome⁵.

Basic is the level at which the core services necessary for a minimally functional breast cancer system exist. Limited adds services that produce major improvements in outcome such as increases in survival, and is typically delivered at secondary level. Enhanced adds services that are optional but important, improving outcome and increasing the number and quality of options available. Maximal covers services that may be used in some high-resource settings and that should be considered lower priority than those in the first three levels on the basis of cost-effectiveness.

Two consequences follow from this structure and both are frequently missed.

The tiers are cumulative. An enhanced-tier service delivered without the basic-tier services beneath it does not deliver the outcome attributed to it. A HER2-directed agent supplied to a centre without reliable receptor testing is an enhanced intervention resting on an absent basic one.

The tiers apply to services and not to countries. Any country contains institutions at several tiers, and the planning unit is the facility.

The stratification has aged in a specific and instructive way. It was constructed when the distinction between basic and enhanced was largely about cytotoxic chemotherapy and surgery. It now has to accommodate biomarker-defined therapy, where the limiting resource is an assay rather than a drug. A tumour-agnostic or biomarker-defined indication cannot be delivered at any tier that lacks the test, which is the argument of Pathology and diagnostic infrastructure as the rate-limiting step.

6 · Task shifting and clinical breast examination based programs

Mammographic screening requires equipment, trained radiographers, radiologists to report, and a recall system. Where those do not exist, the relevant question is not how to acquire them. It is whether earlier detection can be achieved without them.

Clinical breast examination delivered by trained non-physician health workers is the principal answer that has been tested in a randomised design. In a cluster-randomised trial in Mumbai, 20 geographically distinct clusters containing 151 538 women aged 35 to 64 were allocated to screening or to control. Women in the screening arm received four rounds of clinical breast examination by trained female primary health workers every two years, with cancer awareness, followed by active surveillance. Control clusters received one round of cancer awareness and active surveillance⁹.

The downstaging result was clear. The proportion of women with stage III or IV disease was 37% in the screening arm against 47% in the control arm. Breast cancer was detected at a younger age in the screening arm⁹.

The mortality result was not. Breast cancer mortality was 20.82 per 100 000 person-years in the screening arm against 24.62 in the control arm, a rate ratio of 0.85 with limits of 0.71 to 1.01. That 15% reduction did not reach significance. A post hoc subset analysis in women aged 50 and over gave a rate ratio of 0.71, with limits of 0.54 to 0.94. In women under 50 the rate ratio was 0.93, with limits of 0.79 to 1.09⁹.

Task shifting is not confined to detection. Where oncologist density is low, chemotherapy administration, symptom management and follow-up can be delivered by trained nurses and clinical officers under protocol. The limiting factor in every case is supervision capacity rather than the procedure itself.

7 · Pathology and diagnostic infrastructure as the rate-limiting step

Every treatment decision in breast cancer is made on a pathology report. Receptor status determines whether endocrine therapy is given. HER2 status determines whether HER2-directed therapy is given. Grade and nodal status determine much of the rest. Without a report, the only available decision is whether to operate.

Pathology and laboratory medicine services in low-income and middle-income countries are of both insufficient scope and inadequate quality for that role, and four barriers have been identified specifically. Insufficient workforce capacity. Inadequate education and training. Inadequate infrastructure. Insufficient quality standards and accreditation¹⁰.

The ordering matters when a programme is designed. A donated immunohistochemistry platform without trained staff, without cold chain for antibodies and without external quality assessment produces results. It does not produce correct ones, and an incorrect receptor result is worse than an absent one because it is acted upon.

The analytic validity problem is the same one set out in Receptor assessment and the measurement problem, arriving under harsher conditions. Fixation time, antibody clone, scoring convention and observer training all move the result. A centre without external quality assessment has no way of knowing where its own assay sits.

This is why pathology is described as rate-limiting rather than as one component among several. Drug supply, surgical capacity and radiotherapy capacity each deliver their benefit only to patients correctly classified. Adding a targeted agent to a system that cannot reliably determine the target adds cost without adding outcome.

8 · Essential medicines, biosimilars, and access to targeted therapy

The WHO Essential Medicines List identifies medicines that are priority for public health. Whether those medicines reach patients is a separate question, and it has been measured.

In a cross-sectional survey of 948 oncologists across 82 countries, respondents selected the cancer medicines that would produce the greatest public health benefit in their own country. The most frequently selected were doxorubicin, cisplatin, paclitaxel, pembrolizumab, trastuzumab, carboplatin and fluorouracil. Of the twenty most frequently selected medicines, 19 were already on the Essential Medicines List, 12 were cytotoxic agents and 13 had received United States regulatory approval before 2000¹¹.

That is the first finding, and it is about priorities rather than access. Frontline oncologists worldwide prioritised old, cheap, off-patent drugs.

The second finding is about availability. Universal availability of each of those twenty medicines was reported by 9% to 54% of respondents in low-income and lower-middle-income countries, 13% to 90% in upper-middle-income countries and 68% to 94% in high-income countries. Catastrophic expenditure means spending more than 40% of household consumption net of food. Its risk was reported by 13% to 68% of respondents in lower-middle-income countries, against 0% to 9% in high-income countries¹¹.

Two implications follow.

Access failure is concentrated in medicines that are not expensive to manufacture. That points at procurement, supply chain and regulatory capacity rather than at price alone.

Adding further expensive agents to the Essential Medicines List is of questionable value while the existing list is unavailable to half the respondents in the poorest settings. Biosimilar entry lowers the price of trastuzumab and does not by itself build the cold chain, the infusion capacity or the cardiac monitoring that its use requires.

9 · Radiotherapy and surgical capacity

Radiotherapy capacity is the most unequally distributed major cancer resource, and the disparity in machine availability between high-income countries and low-income and middle-income countries remains a principal constraint. Workforce credentialling is a second constraint of comparable size, since a machine without trained physicists and therapists does not treat patients¹².

One mitigation is available immediately and does not require new equipment. Replacing conventional fractionation with hypofractionated protocols for prostate and breast cancer was projected to allow an additional 2.2 million patients to be treated with existing resources. Of those, 1.4 million would be patients with breast cancer¹².

That projection deserves emphasis because of what it does not require. It requires no capital purchase and no new drug. It requires a change in schedule, and the schedule change is one that high-income practice has already made on efficacy grounds. The evidence base sits in Fractionation, hypofractionation, and acceleration.

Surgical capacity is the other half. The Lancet Commission on Global Surgery estimated that 5 billion people lack access to safe, affordable surgical and anaesthesia care when needed¹³. Breast surgery is technically among the less demanding oncological operations, which cuts both ways. It can be delivered in a district hospital with a general surgeon. It also competes for the same theatre time, anaesthesia and post-operative capacity as every other surgical condition in that hospital.

The interaction between the two modalities is a planning variable rather than a clinical one. Breast-conserving surgery requires adjuvant radiotherapy. Where radiotherapy is unavailable or unreachable, mastectomy is not an inferior choice. It is the choice that matches the available system, and offering conservation without deliverable radiation produces worse outcomes than offering mastectomy.

10 · Workforce capacity and training

Workforce is the constraint that money does not immediately relieve, because training takes years and the trainees are mobile.

Four workforce groups are separately rate-limiting in breast cancer care, and a programme can be complete in three and blocked by the fourth. Pathologists and laboratory scientists, without whom Pathology and diagnostic infrastructure as the rate-limiting step applies. Surgeons with oncological training. Medical oncologists or physicians trained to deliver systemic therapy safely. Radiation oncologists, medical physicists and radiation therapists, whose absence renders installed equipment unusable^12,10.

Nursing sits across all four and is the group most often omitted from workforce counts.

Two structural problems recur.

Training abroad produces clinicians competent in a system they will not return to, and it selects for emigration. Training in-country produces clinicians matched to the available resource and requires local trainers, which is the bottleneck that in-country training exists to relieve.

Retention is governed by factors outside the training programme. Salary, availability of equipment to practise with, professional isolation and the presence of a career structure determine whether a trained specialist stays. A scholarship programme with no retention component transfers workforce from a low-income to a high-income country at the low-income country's expense.

11 · Capacity building, twinning, and sustainable partnership

Twinning pairs an institution in a high-resource setting with one in a low-resource setting for a sustained period. The model has a good record in some settings and a poor one in others, and the difference is structural rather than a matter of goodwill.

The features that distinguish the two are identifiable in advance.

Direction of agenda-setting. A partnership in which the low-resource partner defines the priority addresses a problem that partner has. One in which the high-resource partner defines it addresses a problem that partner finds interesting.

Duration relative to the training cycle. Building pathology capacity takes longer than a grant period. Partnerships funded for three years and dependent on a single individual end when either ends.

Whether the intervention survives withdrawal. Equipment donation without a service contract, consumables budget and local biomedical engineering leaves non-functioning equipment. Training without a local trainer leaves a cohort that cannot be replaced.

Whether it fits the local tier. Introducing an enhanced-tier service into a system that has not secured the basic tier moves resource away from the interventions with the larger outcome gain, as Basic, limited, enhanced, and maximal resource tiers sets out.

The most durable form of capacity building is the least visible, which is that it produces local trainers rather than trained locals. A programme is sustainable at the point where it can replace its own staff without external input.

12 · Data sovereignty, equitable authorship, and avoiding extractive research

Research conducted in low-resource settings by high-resource institutions has a measurable pattern of credit allocation.

Across 7100 articles on health in sub-Saharan Africa published between 2014 and 2016, 68.3% included collaborators from the United States, Canada, Europe or another African country. Overall, 54.0% of 43 429 authors and 52.9% of first authors were from the country the paper was about. Where collaborators came from the United States, Canada or Europe, local representation fell. For papers including collaborators from top United States universities, 41.3% of all authors and 23.0% of first authors were from the country of the paper's focus. Local representation was highest when the collaborators were from another African country. Overall, 13.5% of papers had no local coauthor at all¹⁴.

The figure to hold is the last one. One paper in seven about health in a sub-Saharan African country had no author affiliated with the country the paper was about. Authors from elsewhere in Africa are a separate category in that review, and local representation was highest when the collaborators came from another African country.

Authorship is a proxy for a larger issue. Specimens, sequencing data and registry records collected in one country and analysed, stored and published in another leave the originating country without the capacity to use its own data. The result is that the country supplies the observations and imports the conclusions.

Data sovereignty is the principle that a population retains governance over data derived from it. Operationally it means locally held data, local capacity to analyse it, local approval for secondary use, and return of results in a usable form.

The ancestry genomics literature is where this matters most directly for this book. The argument in Ancestry-associated tumor biology, what replicates and what is confounded is that most claims about ancestry-associated tumour biology rest on cohorts in which race was recorded and ancestry was not. Correcting that requires genomic cohorts recruited in the populations concerned. Building those cohorts extractively would reproduce the authorship pattern above in the one literature where the relationship between a population and its own genomic data is the actual subject.

References

All 14 references

1. Arnold M, Morgan E, Rumgay H, et al. Current and future burden of breast cancer: global statistics for 2020 and 2040. Breast 2022 66:15-23. PMID 36084384
2. Girardi F, Nyangasi M, Callender C, et al. Global breast cancer survival estimates in 2017-2021 to advance the WHO Global Breast Cancer Initiative. Nat Med 2026. PMID 42420542
3. Allemani C, Matsuda T, Di Carlo V, et al. Global surveillance of trends in cancer survival 2000-14 (CONCORD-3): analysis of individual records for 37 513 025 patients diagnosed with one of 18 cancers from 322 population-based registries in 71 countries. Lancet 2018 391:1023-1075. PMID 29395269
4. Mensah S, Kyei I, Agbedinu K, Shiako J, Yorke DA, Gyedu A. Assessing breast cancer diagnosis and management in Ghana per the Global Breast Cancer Initiative key performance indicators. JCO Glob Oncol 2025 11:e2400653. PMID 40324119
5. Anderson BO, Shyyan R, Eniu A, et al. Breast cancer in limited-resource countries: an overview of the Breast Health Global Initiative 2005 guidelines. Breast J 2006 12 Suppl 1:S3-S15. PMID 16430397
6. Anderson BO, Duggan C, Scheel JR. Resource-appropriate evidence-based strategies to improve breast cancer outcomes in low- and middle-income countries guided by the Breast Health Global Initiative and Global Breast Cancer Initiative. J Surg Oncol 2023 128:952-958. PMID 37811558
7. Anderson BO, Ilbawi AM, Fidarova E, et al. The Global Breast Cancer Initiative: a strategic collaboration to strengthen health care for non-communicable diseases. Lancet Oncol 2021 22:578-581. PMID 33691141
8. Boucheron P, Zietsman A, Anele A, et al. Measuring the WHO Global Breast Cancer Initiative Pillars' key performance indicators in Sub-Saharan Africa: experience in the African Breast Cancer-Disparities in Outcomes hospital-based cohort study. EClinicalMedicine 2025 81:103104. PMID 40034567
9. Mittra I, Mishra GA, Dikshit RP, et al. Effect of screening by clinical breast examination on breast cancer incidence and mortality after 20 years: prospective, cluster randomised controlled trial in Mumbai. BMJ 2021 372:n256. PMID 33627312
10. Wilson ML, Fleming KA, Kuti MA, Looi LM, Lago N, Ru K. Access to pathology and laboratory medicine services: a crucial gap. Lancet 2018 391:1927-1938. PMID 29550029
11. Fundytus A, Sengar M, Lombe D, et al. Access to cancer medicines deemed essential by oncologists in 82 countries: an international, cross-sectional survey. Lancet Oncol 2021 22:1367-1377. PMID 34560006
12. Abdel-Wahab M, Giammarile F, Carrara M, et al. Radiotherapy and theranostics: a Lancet Oncology Commission. Lancet Oncol 2024 25:e545-e580. PMID 39362232
13. Ng-Kamstra JS, Greenberg SLM, Abdullah F, et al. Global Surgery 2030: a roadmap for high income country actors. BMJ Glob Health 2016 1:e000011. PMID 28588908
14. Hedt-Gauthier BL, Jeufack HM, Neufeld NH, et al. Stuck in the middle: a systematic review of authorship in collaborative health research in Africa, 2014-2016. BMJ Glob Health 2019 4:e001853. PMID 31750000