L09 Cancer Epidemiology of Cancer

Overview

This lecture introduces cancer epidemiology through three parts: the global and New Zealand burden of cancer, the descriptive-epidemiology framework of person, place and time, and two worked examples (colorectal cancer, stomach cancer) that show how descriptive data generate causal hypotheses.

Global and NZ burden of disease

• Worldwide (2022 estimate): about 20 million new cancer cases per year, with more cases in men than women, and almost 10 million cancer deaths per year. Cases are projected to exceed 35 million per year by 2050, driven by population ageing and growth.
• Age-standardised incidence is highest in North America, most of Europe/Scandinavia, Australia/New Zealand and parts of East Asia (e.g. Japan, South Korea), and lowest across most of sub-Saharan Africa and parts of South/Southeast Asia.
• Incidence rises with country development level (Human Development Index): very high HDI (~285/100,000) > high HDI (~185) > medium HDI (~110) ≈ low HDI (~105), i.e. incidence roughly plateaus between medium and low HDI.
• Mortality does not track incidence directly: lung and liver cancer have mortality rates close to their incidence (poor survival), whereas breast, prostate and thyroid cancer have much lower mortality relative to incidence (better detection/survival). Globally by site, breast has the highest incidence (~46/100,000), followed by prostate (~31) and lung (~23); lung has mortality (~18) nearly as high as its incidence, and liver’s mortality (~9) is almost equal to its incidence (~9).
• Age-standardised DALYs from all neoplasms fell across most high-income comparator countries (including NZ) from 1990 to 2016, converging toward the lower end of the range by 2016 (Netherlands remained highest, Finland lowest).
• NZ, five years to 2019: the ten most common cancers in men are led by prostate (~3,900 cases/yr, ASR 52.4/100,000), colorectal (~1,700/yr, 24.7), and melanoma (~1,500/yr, 23.8). In women, the leading cancers are breast (~3,400/yr, ASR 78.1), melanoma (23), and colorectal (22.5).
• NZ ethnic disparities in cancer burden (2022): incidence per 100,000 is highest in Māori (409) and Pacific (396), then European/Other (348), then Asian (206, lowest). Māori and Pacific populations have higher cancer mortality than others; the Māori cancer death rate is 1.7 times the non-Māori rate; Asian populations have the lowest cancer death rates.
• By cancer site (Māori vs non-Māori, non-Pacific), Māori have markedly higher registration and mortality for lung cancer (the largest disparity of any site, both for incidence and especially mortality), and higher rates for liver and stomach cancer, while melanoma incidence is markedly lower in Māori than non-Māori.
• NZ time trends (2004–2013): the absolute number of cancer registrations and deaths rose (population growth and ageing), while the age-standardised registration rate (men) and mortality rate fell over the same decade; female registration rates stayed roughly flat.
• General age/time conclusions for developed countries: cancer incidence and mortality increase with age (with a small excess in early childhood); as life expectancy rises, cancer becomes a proportionately larger cause of death (both because other causes of death have fallen faster and because of population ageing); for women, cancer becomes the proportionately most important cause of death by around age 50.

Descriptive epidemiology: person, place, time

Descriptive epidemiology examines how disease varies by person, place and time, and these patterns generate hypotheses about causation.

• Person: mortality from lung cancer (men), breast cancer and cervical cancer all rise with age in England & Wales, but with different shapes — lung cancer mortality rises steeply and near-exponentially with age; breast cancer mortality rises but the rate of increase flattens at older ages; cervical cancer mortality rises then plateaus from about age 50–70. Ethnicity within the same country also produces different cancer-site profiles: in 1978–82 Singapore, Chinese people had much higher nasopharyngeal, stomach, colorectal, liver and lung cancer incidence than Malay or Indian people, and Chinese women had the highest breast cancer incidence, while Indian women had the highest cervical cancer incidence — person-based (ethnic) variation in the pattern of cancer sites within one country.
• Place: incidence of specific cancer sites varies hugely by geographic area, e.g. (male) oesophageal cancer: Iran (high) vs Nigeria (low), ratio 300:1; prostate cancer: US (Black men, high) vs Japan (low), ratio 40:1; stomach cancer: Japan (high) vs Uganda (low), 25:1; liver cancer: Mozambique (high) vs England (low), 100:1. In women, uterine cancer varies 30:1 (US high vs Japan low) and cervical cancer 15:1 (Colombia high vs Israel, Jewish, low). Geographic risk factors can be identified this way: the world map of Hepatitis B surface antigen carrier prevalence closely matches the world map of liver cancer risk (both concentrated in sub-Saharan Africa and East/Southeast Asia), implicating HepB as a place-linked cause of liver cancer.
• Time: within NZ, cancer case numbers rose 2004–2013 (population growth/ageing) while the age-standardised registration rate (male) and mortality rate declined over the same period.
• Migration studies combine person and place: for colon cancer, people of NZ origin retain an elevated risk even after moving to England & Wales (relative risk 1.63–1.90 vs 1.0 reference for E&W-origin/E&W-resident), while English & Wales-origin migrants to NZ show only a modest risk increase (RR ~1.1–1.16) — consistent with risk being set by early-life exposure or a strong genetic component, with NZ residence itself also carrying some elevated (environmental) risk for both origin groups.

Worked example 1: colorectal cancer in New Zealand

• Global incidence in women is highest in North America, Western Europe and Australia/New Zealand.
• Age: incidence rises steadily with age for both sexes, converging up to about age 50–55 (~100/100,000), then diverging, with male incidence rising more steeply thereafter (reaching ~385/100,000 at 85+ vs ~330 for women). Overall male:female incidence rate ratio is 1.2.
• Place: NZ has among the highest colon cancer incidence rates internationally in both women (~29/100,000, 1993–97, highest of the countries shown) and men (~31/100,000, second only to US SEER Black men at ~33); India (Bombay) has the lowest rate in both sexes (~3/100,000).
• Ethnicity within NZ: colorectal cancer registration rates are far higher in non-Māori than Māori (non-Māori roughly 1300–1600/100,000 vs Māori roughly 50–120/100,000, 2008–2017), with males higher than females in both groups. [This is a different, ethnicity-specific comparison from the earlier person/place NZ vs England migration data, which used national-origin, not ethnic, groupings.]
• A birth-cohort effect underlies the overall time trend: relative risk of colorectal cancer (both incidence and mortality) rose across cohorts born from ~1906 to a peak for cohorts born in the 1930s, then declined steadily for cohorts born from the early 1940s onward, reaching a trough for cohorts born in the late 1950s/1960s before rising slightly again toward 1972.
• Consequence of the cohort effect: people aged 35–44 experienced large reductions in colon cancer incidence (55% men, 57% women) and an even larger reduction in mortality (82% men, 70% women) between 1975 and 2009.
• Any causal theory must explain: risk is predominantly set before age 25 and persists for life (with a smaller possible later-life influence); the reduction in risk is greater for colon than rectal cancer; the unknown protective factor took effect sometime between the start of WWII and about 1966, probably by affecting adenoma prevalence, and probably reflects a cumulative rather than one-off exposure.
• Hypothesis and test: Cox & Sneyd’s national case-control study (562 cases, 571 controls; cases from the NZ Cancer Registry, controls from the electoral roll) tested whether school milk consumption explained the cohort effect. School milk programme participation was high (~80–92%) and similar between cases and controls for cohorts starting school 1942–47 through 1963–67, then fell sharply for those starting school ≥1968 (to ~17% cases, ~36% controls).
  • Any school milk vs none: adjusted OR 0.70 (95% CI 0.51–0.96).
  • Dose-response by weekly bottles: 1–4 bottles/wk OR 0.96; 5 bottles/wk OR 0.68; 6–9 OR 0.70; ≥10 OR 0.39 (test for trend p = 0.002).
  • Dose-response by total lifetime bottles: 1200–1599 OR 0.62; 1600–1799 OR 0.57; ≥1800 OR 0.62 (all p < 0.05 vs none; test for trend p = 0.002).
  • In the multivariable model, total milk bottles consumed remained protective (OR 0.984 per 100 bottles, p < 0.05); family history of colorectal cancer was a risk factor (OR 1.84); dairy consumption in childhood/adolescence showed a similar but non-significant trend.
  • Conclusion: the a priori hypothesis that school milk reduced adult colorectal cancer risk was not rejected, but as a single observational study, additional research was needed.
• A later systematic review and meta-analysis (Barrubés et al. 2019, 15 cohort + 14 case-control studies, >22,000 cases) found higher total dairy and total milk consumption associated with lower colorectal cancer risk (RR 0.80 and 0.82 respectively), low-fat milk associated with lower risk restricted to colon cancer (RR 0.73), and cheese inversely associated with colorectal and proximal colon cancer risk; most associations were not supported by the case-control subset. A 2025 Nature Communications study of >500,000 women found an extra 300 mg/day of dietary calcium (about one large glass of milk) was linked to a 17% lower risk of bowel cancer, with the protective effect of milk/dairy generally attributed largely to calcium (though other milk components may also contribute).
• Established risk factors for colorectal cancer: age (>90% of cases occur at ≥50 years), inflammatory bowel disease (Crohn’s disease, ulcerative colitis), personal or family history of colorectal cancer or polyps, genetic syndromes (familial adenomatous polyposis, hereditary non-polyposis colorectal cancer), type II diabetes, prior radiation, and race/ethnicity.
• Lifestyle risk factors: physical inactivity, low fruit and vegetable intake, low-fibre/high-fat diet, higher red and processed meat intake, overweight/obesity, alcohol, and tobacco use.
• Protective factors: exercise; a diet high in fruit, vegetables and fibre and low in animal fat; possibly protective medications/supplements including aspirin/NSAIDs, calcium, vitamin D and selenium; and screening.

Worked example 2: stomach cancer in New Zealand

• By age (NZ, 1981–1984): both stomach cancer registration and mortality rates rise steeply with age, from near zero below ~40 to about 150 (registrations) and 130 (mortality) per 100,000 person-years by age 80–85, with registrations consistently slightly above mortality.
• By ethnicity and sex over time (mortality, 1951–1986): Māori men had the highest mortality throughout (declining from ~51–56 to ~22 per 100,000), followed by Māori women (~32 to ~12–15), then non-Māori men (~24 to ~11–12), with non-Māori women lowest throughout (~13 to ~4–5) — all four groups declined over the period.
• By ethnicity and sex over time (registrations, 1949–1984): Māori men and women registration data begins later (from ~1967–69) and runs higher than non-Māori, with fluctuation (Māori men peaking around 45/100,000 in 1973–75); non-Māori men declined gradually (~21 to ~13–15); non-Māori women were lowest and flattest (~6–9) throughout.
• International place comparison (2018): the highest age-standardised stomach cancer incidence rates were in South Korea (39.6/100,000), followed by Mongolia (33.1), Japan (27.5), China (20.7) and Bhutan (19.4) — a pattern epidemiologically linked to diet in these populations (illustrated by a photograph of grilled/preserved fish in a Japanese market context).
• Overall NZ time trend (1948–1986): both registrations and mortality declined steadily from around 20–21 per 100,000 in the late 1940s/early 1950s to around 7–8 per 100,000 by the mid-1980s, tracking closely together throughout (mortality slightly trailing registrations).

Self-test

1. Describe how global cancer incidence and mortality (2022) differ between more-developed and less-developed regions, and explain why some cancer sites (e.g. lung, liver) show mortality close to their incidence while others (e.g. breast, thyroid) do not.
2. Explain the disparity in cancer burden between Māori and non-Māori/other ethnic groups in New Zealand, naming the cancer site with the largest disparity and stating the direction of the melanoma disparity.
3. Define the three components of descriptive epidemiology and give one example from the lecture of how each generates a hypothesis about cancer causation.
4. Describe what the Hepatitis B / liver cancer maps demonstrate about place as a determinant of cancer risk.
5. Explain what the England & Wales vs New Zealand colon cancer migration study shows about the relative contribution of early-life/genetic factors versus place of residence to colon cancer risk.
6. Describe the birth-cohort pattern of colorectal cancer risk in New Zealand, including the approximate years cohort risk peaked and then declined.
7. List the four features that any causal theory for the NZ colorectal cancer cohort effect had to incorporate.
8. Describe the design of Cox and Sneyd’s school milk case-control study, and state its main dose-response finding for weekly and lifetime school milk consumption.
9. Was the school milk hypothesis rejected by this study? Explain the caveat the lecture places on this conclusion.
10. List the established (non-lifestyle) risk factors and the lifestyle risk factors for colorectal cancer, and list the protective factors.
11. Distinguish the age-pattern and place-pattern of stomach cancer incidence/mortality in New Zealand from the corresponding patterns for colorectal cancer described earlier in the lecture.
12. Describe the overall time trend in New Zealand stomach cancer registrations and mortality from the late 1940s to the mid-1980s.

Answers

Reveal answers

1. Age-standardised cancer incidence is highest in more-developed regions (North America, Europe, Australia/NZ, parts of East Asia) and rises with Human Development Index, plateauing between medium and low HDI. Mortality does not simply track incidence: lung and liver cancer have poor survival, so their mortality rates sit close to their incidence rates, whereas breast, prostate and thyroid cancer have much better detection/survival, so their mortality is much lower than their incidence.
2. Māori (and Pacific) populations have higher overall cancer incidence and mortality than European/Other and Asian populations in NZ (Māori cancer death rate 1.7x non-Māori). The largest single disparity is in lung cancer, where Māori registration and especially mortality rates are far higher than non-Māori. Melanoma is the exception: its incidence is markedly lower in Māori than in non-Māori.
3. Person = variation by individual characteristics such as age, sex or ethnicity (e.g. ethnic differences in cancer-site profile among Chinese, Malay and Indian people in Singapore suggest ethnicity-linked exposures or susceptibility). Place = variation by geographic location (e.g. the 300:1 difference in oesophageal cancer between Iran and Nigeria points to a strong environmental or dietary cause). Time = variation within a population over calendar time or birth cohort (e.g. the NZ colorectal cancer birth-cohort effect points to an exposure that changed around WWII to the mid-1960s).
4. The map of Hepatitis B surface antigen carrier prevalence and the map of liver cancer risk show almost the same geographic pattern (both concentrated in sub-Saharan Africa and East/Southeast Asia), which is strong descriptive evidence that Hepatitis B infection is a place-linked cause of liver cancer.
5. People of NZ origin kept an elevated colon cancer risk (relative risk ~1.6–1.9) even after moving to and living in England & Wales, while people of English & Wales origin who moved to NZ showed only a modest risk increase (RR ~1.1) relative to those who stayed in E&W. This pattern points to a strong genetic or early-life exposure component that travels with the person, though NZ residence itself also carried some additional (environmental) risk for both origin groups.
6. Relative risk of colorectal cancer rose across birth cohorts from about 1906 to a peak for those born in the 1930s, then declined steadily for cohorts born from the early 1940s onward, reaching its lowest point for cohorts born in the late 1950s/1960s, before rising slightly again toward 1972.
7. (1) Risk is predominantly determined before age 25 and persists for life, with only a small possible later-life influence; (2) the risk reduction is greater for cancer of the colon than of the rectum; (3) the unknown causal factor took effect sometime between the start of WWII and about 1966; (4) it probably acts by affecting adenoma prevalence and reflects a cumulative exposure rather than a single one-off event.
8. It was a national case-control study (562 cases from the NZ Cancer Registry, 571 controls from the electoral roll) that measured school milk exposure (whether provided, bottles/week, and ages of starting/stopping) by questionnaire. Any school milk consumption vs none reduced odds of colorectal cancer (OR 0.70); risk fell further with more bottles per week (down to OR 0.39 for ≥10 bottles/week) and with more total lifetime bottles (down to OR ~0.57–0.62 for the highest categories), with a significant trend test (p = 0.002) in both dose-response analyses.
9. No — the a priori hypothesis that school milk reduced adult colorectal cancer risk was not rejected by the study. The lecture’s caveat is that this was only a single (observational, case-control) study, so additional research was needed before drawing firm causal conclusions.
10. Established risk factors: age ≥50 (>90% of cases), inflammatory bowel disease (Crohn’s, ulcerative colitis), personal/family history of colorectal cancer or polyps, genetic syndromes (FAP, HNPCC), type II diabetes, prior radiation, and race/ethnicity. Lifestyle risk factors: physical inactivity, low fruit/vegetable intake, low-fibre/high-fat diet, high red/processed meat intake, overweight/obesity, alcohol, and tobacco. Protective factors: exercise, a diet high in fruit/vegetables/fibre and low in animal fat, possibly protective medications/supplements (aspirin/NSAIDs, calcium, vitamin D, selenium), and screening.
11. Both cancers rise steeply with age, but stomach cancer’s age pattern was shown only as a rate-vs-age curve with no reported sex divergence, whereas colorectal cancer’s age curve diverges by sex after age ~50–55 (male incidence overtaking female). For place, colorectal cancer incidence is highest in high-HDI countries including NZ, whereas the highest stomach cancer incidence rates internationally are in South Korea, Mongolia, Japan, China and Bhutan — an East Asian pattern linked to diet, in contrast to colorectal cancer’s Western/high-HDI pattern.
12. Both age-standardised stomach cancer registrations and mortality in New Zealand declined steadily from around 20–21 per 100,000 in the late 1940s/early 1950s to around 7–8 per 100,000 by the mid-1980s, with the two measures tracking closely together throughout (mortality slightly trailing registrations).