Part I · Foundations and the global landscape · Chapter 2

Epidemiology and the changing burden of disease

Incidence and mortality curves are not read off biology, and separating the two is a skill.

1 · Global incidence, mortality, and prevalence

In the 2022 estimates, cancer of the female breast accounted for 11.6% of all new cancer cases and 6.9% of all cancer deaths worldwide¹. The gap between those two proportions is the chapter in one line. Incidence is concentrated where survival is best.

Estimates from the GLOBOCAN database put the burden at over 2.3 million new cases and 685000 deaths in women in 2020². Demographic projection alone, with no change in age-specific rates, takes that to over 3 million cases and 1 million deaths a year by 2040.

Prevalence follows from the same arithmetic and is often left out. High incidence combined with good survival produces a large population living with a diagnosis, sometimes for decades. Those patients consume follow-up, survivorship care and surveillance imaging, and they are invisible in any figure that reports only cases and deaths.

The single most consequential number in this chapter is the spread in survival. World Health Organization estimates of five-year net survival for women diagnosed between 2017 and 2021 give a regional median of 39.1% in the African Region against 88.5% in the Region of the Americas³. That gap is not a gap in tumour biology. It is the argument that Global breast cancer care exists to answer.

2 · Age, birth cohort, and geographic patterns

Incidence rises steeply with age and then flattens, and the flattening is informative. A single exponential age curve would suggest one accumulating process. The observed shape is more consistent with two overlapping disease groups separated around the menopause, which is developed in PI3K, AKT, mTOR, and related networks and Young adults, fertility, and pregnancy.

Geography separates incidence from mortality cleanly. Age-standardized incidence ranges from under 40 per 100000 women in some Asian and African countries to over 80 per 100000 in Australia and New Zealand, North America and parts of Europe². Variation in mortality is much narrower. Countries in epidemiologic transition therefore carry a share of deaths out of proportion to their share of cases.

Cohort and period effects are different things and are routinely conflated. A period effect shifts rates in every age group at the same calendar time, which is what the introduction of screening looks like. A cohort effect follows a birth cohort through life, which is what a change in reproductive pattern or childhood exposure looks like. Formal age-period-cohort analysis of United States incidence in women aged 20 to 49 years found evidence of both⁴.

Reading a rising national curve as a change in exposure requires ruling out the period effect first. That is the subject of the next section.

3 · Reading incidence and mortality trends without confusing screening for biology

Four artefacts sit between a screening programme and an incidence curve, and all four push in the same direction. Lead time moves the diagnosis earlier without moving the death. Length bias enriches the screen-detected group for slow-growing tumours. Stage migration reassigns patients to better stages without changing anyone's outcome. Overdiagnosis adds cancers that would never have surfaced.

The effect on incidence is unambiguous. In the United States, incidence rose by 1% a year between 2012 and 2021, and the rise was largely confined to localized-stage and hormone receptor positive disease⁵. The national cancer statistics for 2024 place breast among the cancers whose incidence is still rising, by 0.6% to 1% a year during 2015 to 2019⁶. The 2022 report had put the rise at 0.5% a year during 2014 to 2018, with the decline in mortality slowing⁷. An increase concentrated in the stage and subtype that screening finds most easily is a detection signal until proven otherwise.

Mortality is the harder question, because both screening and treatment improved together. Modelling work has tried to separate them twice, a generation apart. Seven independent models covering 1975 to 2000 attributed between 28% and 65% of the mortality reduction to screening, with a median of 46%, and the remainder to adjuvant treatment⁸. Four models covering 1975 to 2019 estimated a 58% reduction in breast cancer mortality overall⁹. Of that reduction, 25% was associated with screening, 47% with treatment of stage I to III disease and 29% with treatment of metastatic disease.

A tumour-size analysis makes the same point from observed data. After screening was introduced, the rate of large tumours fell by 30 cases per 100000 women while the rate of small tumours rose by 162 per 100000¹⁰. If the underlying disease burden was stable, most of those additional small tumours were never going to become large ones.

4 · Reproductive, hormonal, metabolic, and environmental determinants

The established determinants act through cumulative exposure of breast epithelium to oestrogen, and the pattern of associations fits that mechanism closely. Circulating sex hormone concentrations in postmenopausal women are associated with subsequent breast cancer risk¹¹.

Childbirth is the instructive case, because its effect reverses with time. In a pooled analysis of 15 prospective cohorts, risk in parous women peaked about five years after delivery, with a hazard ratio of 1.80 against nulliparous women¹². It then declined, and crossed below 1.0 at about 24 years after birth. The pattern was driven by oestrogen receptor positive disease. Parity is protective in the long run and transiently the opposite, and a study with short follow-up will find whichever half it happened to observe.

Exogenous hormones act in the same direction. In an individual-participant meta-analysis, current users in years 5 to 14 of oestrogen with progestagen had a relative risk of 2.08 against never users¹³. For current users of oestrogen alone over the same interval the relative risk was 1.33. Some excess risk persisted for more than ten years after stopping.

Metabolic and behavioural exposures are weaker per person and large in aggregate. Body mass index is associated with the incidence of several cancers including breast cancer¹⁴. A randomized dietary modification intervention was followed by reduced breast cancer mortality on long-term follow-up¹⁵.

Not every hypothesis survives. Three prospective studies with a meta-analysis of published work found no substantial association between night shift work and breast cancer incidence¹⁶. That null result is worth as much space as a positive one and usually gets less.

5 · The rise of young-onset disease

Incidence below age 50 is rising, and the rise is not obviously a detection artefact. In the United States, incidence increased by 1.4% a year in women under 50 and by 0.7% a year in women aged 50 and over between 2012 and 2021⁵. That steeper rise in the younger group was significant only among White women, which limits how far the argument below can be carried. Most organized screening does not begin before 40, so a faster rise in the unscreened age group cannot be attributed to screening in the ordinary way.

A registry analysis of 217815 women aged 20 to 49 years diagnosed between 2000 and 2019 gives the detail⁴. Overall incidence rose with an average annual percent change of 0.79. The increase was confined to receptor positive disease, with an average annual percent change of 2.72 for tumours positive for both oestrogen and progesterone receptors. Rates of tumours negative for both receptors did not increase.

Three explanations compete and they are not mutually exclusive. Exposure may have changed, through later first birth, lower parity, rising body mass index and other cohort-linked shifts. Detection may have changed, through more imaging in symptomatic young women. Biology may have changed in ways nobody has identified.

The receptor-specific pattern favours the exposure account, because the hormonally driven subtype is the one that is rising. It does not settle it. Management questions specific to this group are taken up in Young adults, fertility, and pregnancy.

6 · Subtype-specific epidemiology across populations

Subtype distribution differs between populations, and that difference is frequently offered as the explanation for an outcome gap. It explains part of one and not most of it, and the distinction is worth being precise about.

The United States data are the clearest available. Black women have 38% higher breast cancer mortality than White women despite 5% lower incidence⁵. Triple-negative disease is more common in Black women, which contributes. It does not account for the gap, because Black women have the lowest survival within every subtype and every stage except localized disease. They are also less likely to be diagnosed at localized stage, 58% against 68%.

Age adds a further pattern. In women aged 20 to 29 years and 30 to 39 years, age-standardized incidence was higher in non-Hispanic Black women than in non-Hispanic White women, with incidence rate ratios of 1.53 and 1.15⁴. In women aged 40 to 49 years the relationship reversed, with a rate ratio of 0.96.

A crossover of that kind is not a property of tumours. It is what a difference in age-specific exposure or in detection looks like. Race, ethnicity and genetic ancestry are three different variables, and they are separated in Disparities across the breast cancer continuum.

7 · Overdiagnosis as an epidemiologic concept

Overdiagnosis is the detection at screening of a cancer that would not have become clinically apparent during that person's remaining lifetime. It is not misdiagnosis, and it is not a false positive. The pathology is correct in every overdiagnosed case.

Two distinct mechanisms produce it, and they call for different responses. The first is non-progressive or extremely indolent disease. The second is progressive disease in a person who dies of something else first. The second mechanism depends on age and comorbidity rather than on tumour biology.

The magnitude depends entirely on the denominator chosen, which is why published estimates appear irreconcilable. An independent United Kingdom review estimated excess incidence at 11% of cancers diagnosed in the invited group over the long term, and at 19% of cancers diagnosed during the active screening period¹⁷. Expressed as absolute numbers, for every 10000 UK women aged 50 invited to screening over the following 20 years, the panel estimated 43 breast cancer deaths prevented and 129 cancers overdiagnosed.

A Bayesian natural history analysis of a United States screening cohort separated the two mechanisms directly¹⁸. Among 35986 women with 82677 mammograms and 718 diagnoses, 4.5% of preclinical cancers were estimated to be non-progressive. Under biennial screening from age 50 to 74, 15.4% of screen-detected cancers were estimated to be overdiagnosed, 6.1% through indolent disease and 9.3% through competing mortality.

8 · Data quality, registry coverage, and what is invisible in global statistics

Every global figure in this chapter rests on registries, and registries are unevenly distributed. The third cycle of the CONCORD programme analysed individual records for 37.5 million patients from 322 population-based registries in 71 countries, of which 47 provided data with complete population coverage¹⁹. Countries without a functioning registry do not appear as low values. They appear as estimates derived from neighbours.

That distinction matters when reading any global comparison. The WHO survival estimates covering all 194 Member States were produced precisely because measured survival data remain limited in low-income and middle-income countries³. A survival figure for a country with no registry is a model output. A survival figure for a country with complete registration is a measurement. The two are printed in the same typeface.

Even complete registration leaves things out. Stage at diagnosis is frequently missing or inconsistently coded. Receptor and HER2 status are often absent, which makes subtype-specific international comparison largely impossible. Treatment actually received is rarely captured. Cause-of-death coding varies by jurisdiction, which is why net survival, estimated relative to background mortality, is preferred to cause-specific survival in cross-national work.

The practical consequence is a reading rule. Where the survival gap in Global incidence, mortality, and prevalence is widest, the data supporting it are weakest, and the direction of that bias is not known. The infrastructure question is taken up in Global breast cancer care, and the equity question in Disparities across the breast cancer continuum.

References

All 19 references

1. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024 74:229-263. PMID 38572751
2. 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
3. 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
4. Xu S, Murtagh S, Han Y, Wan F, Toriola AT. Breast cancer incidence among US women aged 20 to 49 years by race, stage, and hormone receptor status. JAMA Netw Open 2024 7:e2353331. PMID 38277147
5. Giaquinto AN, Sung H, Newman LA, et al. Breast cancer statistics 2024. CA Cancer J Clin 2024 74:477-495. PMID 39352042
6. Siegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024. CA Cancer J Clin 2024 74:12-49. PMID 38230766
7. Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer statistics, 2022. CA Cancer J Clin 2022 72:7-33. PMID 35020204
8. Berry DA, Cronin KA, Plevritis SK, et al. Effect of screening and adjuvant therapy on mortality from breast cancer. N Engl J Med 2005 353:1784-1792. PMID 16251534
9. Caswell-Jin JL, Sun LP, Munoz D, et al. Analysis of breast cancer mortality in the US, 1975 to 2019. JAMA 2024 331:233-241. PMID 38227031
10. Welch HG, Prorok PC, O'Malley AJ, Kramer BS. Breast-cancer tumor size, overdiagnosis, and mammography screening effectiveness. N Engl J Med 2016 375:1438-1447. PMID 27732805
11. Key T, Appleby P, Barnes I, Reeves G. Endogenous sex hormones and breast cancer in postmenopausal women: reanalysis of nine prospective studies. J Natl Cancer Inst 2002 94:606-616. PMID 11959894
12. Nichols HB, Schoemaker MJ, Cai J, et al. Breast cancer risk after recent childbirth: a pooled analysis of 15 prospective studies. Ann Intern Med 2019 170:22-30. PMID 30534999
13. Collaborative Group on Hormonal Factors in Breast Cancer. Type and timing of menopausal hormone therapy and breast cancer risk: individual participant meta-analysis of the worldwide epidemiological evidence. Lancet 2019 394:1159-1168. PMID 31474332
14. Bhaskaran K, Douglas I, Forbes H, et al. Body-mass index and risk of 22 specific cancers: a population-based cohort study of 5.24 million UK adults. Lancet 2014 384:755-765. PMID 25129328
15. Chlebowski RT, Aragaki AK, Anderson GL, et al. Dietary modification and breast cancer mortality: long-term follow-up of the Women's Health Initiative randomized trial. J Clin Oncol 2020 38:1419-1428. PMID 32031879
16. Travis RC, Balkwill A, Fensom GK, et al. Night shift work and breast cancer incidence: three prospective studies and meta-analysis of published studies. J Natl Cancer Inst 2016 108:djw169. PMID 27758828
17. Independent UK Panel on Breast Cancer Screening. The benefits and harms of breast cancer screening: an independent review. Lancet 2012 380:1778-1786. PMID 23117178
18. Ryser MD, Lange J, Inoue LYT, et al. Estimation of breast cancer overdiagnosis in a U.S. breast screening cohort. Ann Intern Med 2022 175:471-478. PMID 35226520
19. 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