Epidemiology: Incidence, Prevalence, Mortality and Study Design

Q1 — Sample Band 6 response (8 marks), annotated

Definitions. Incidence is the rate of new cases of a disease arising in a population per 100,000 people per year over a defined time period — it measures how fast disease is developing. Mortality rate is the number of deaths attributable to a specific disease per 100,000 people per year — it measures the lethality of the disease at the population level. These are distinct: incidence counts new cases regardless of outcome; mortality counts deaths. [1 — clear distinction with definitions]

Trends. Age-standardised incidence fell from 68 to 53 per 100,000 per year (a 22% decrease). Age-standardised mortality fell more steeply, from 34 to 18 per 100,000 per year (a 47% decrease). Five-year survival improved from 50% to 72% — meaning that by 2016–2020, almost three-quarters of people diagnosed with bowel cancer were still alive five years later, compared to only half in the early 1990s. [1 — trend description with data for all three measures]

Divergence: incidence and mortality moving in different directions. Incidence and mortality can move independently because they measure different things. Falling incidence reflects either genuine disease prevention (e.g. increased dietary fibre, reduced processed meat consumption) or earlier detection through screening programs such as the National Bowel Cancer Screening Program (introduced in Australia in 2006), which detects cancers earlier and thereby removes them from the pool of cases that progress to advanced stages. Falling mortality reflects treatment advances — improved surgical techniques, chemotherapy regimens, targeted therapies (e.g. cetuximab for KRAS wild-type tumours) — that mean a greater proportion of diagnosed cases survive. Both phenomena can be occurring simultaneously. [2 — screening reason + treatment reason, mechanistically explained]

5-year survival as a measure. Rising 5-year survival (50% → 72%) provides strong evidence that treatment effectiveness has improved substantially across three decades — earlier detection gives patients more treatment options and time for intervention. Limitation: 5-year survival is susceptible to lead-time bias — if screening detects cancers earlier (at Stage I rather than Stage III), the clock starts earlier, making patients appear to survive longer with their diagnosis even if their actual date of death is unchanged. This can make survival statistics appear to improve even if treatment effectiveness is constant. [1 — correctly identifies lead-time bias as a limitation]

Confounding variable. Dietary change: Australians have reduced red and processed meat consumption and increased dietary fibre intake over the 1990–2020 period. This dietary shift independently reduces bowel cancer risk and could explain part of the falling age-standardised incidence over time — the trend in incidence may reflect the confounding effect of dietary improvement rather than solely reflecting a fall in underlying disease susceptibility. Researchers would need to control for dietary habits in analyses to isolate the contribution of each factor. [1 — valid confounder with mechanism explaining how it inflates or deflates the trend]

Overall conclusion. The data show a genuine and substantial improvement in Australia's bowel cancer burden across three dimensions: fewer new cases are arising per 100,000 per year, fewer deaths are occurring per 100,000 per year, and people who are diagnosed are surviving longer. However, what cannot be concluded from these data alone is the relative contribution of prevention, screening, and treatment to these trends — and lead-time bias cautions against over-interpreting survival improvements at face value. [1 — appropriately qualified evaluative conclusion]

Marking criteria (8 marks):

• 1 mark — Correctly defines incidence AND mortality rate and explicitly distinguishes them (new cases vs deaths per 100,000 per year).
• 1 mark — Describes the trend in all three measures (incidence, mortality, 5-year survival) with at least two specific data values cited.
• 1 mark — Explains why incidence can fall (screening detects earlier, or genuine prevention) referencing the National Bowel Cancer Screening Program or equivalent named example.
• 1 mark — Explains why mortality can fall independently (treatment advances, better surgery, targeted therapies — named example accepted) — without requiring incidence to fall at the same rate.
• 1 mark — Evaluates the 5-year survival trend as evidence of improved treatment effectiveness.
• 1 mark — Identifies lead-time bias (or equivalent valid limitation of 5-year survival as an epidemiological measure) and explains the mechanism.
• 1 mark — Identifies a valid confounding variable (e.g. dietary change, physical activity trends, alcohol consumption) and explains how it is associated with both the time period AND the outcome (bowel cancer incidence).
• 1 mark — Reaches an appropriately qualified evaluative conclusion that distinguishes what the data show from what cannot be concluded from surveillance data alone.

Q2 — Sample Band 6 response (7 marks), annotated

The journalist has committed two distinct errors: conflating association with causation, and drawing a policy recommendation from a single observational study. [1 — overall evaluative judgement identifying both flaws]

What the cohort study has demonstrated. The study has established a statistical association between high processed meat consumption (5+ servings/week) and a 38% higher rate of bowel cancer development over 15 years compared to low consumers. Because participants were followed forward in time, the cohort study establishes temporal sequence — processed meat consumption was measured before bowel cancer developed, ruling out reverse causation. However, temporal sequence alone is not sufficient to establish causation in an observational study. The word "proves" is scientifically unjustified for any observational study: without randomisation, confounding variables cannot be fully eliminated, and the word "proves" implies a certainty that epidemiological evidence cannot deliver. [1 — correctly distinguishes association from causation; explains temporal sequence does not equal causation]

Confounder 1 — overall dietary quality. People who eat processed meat five or more times per week are also more likely to have generally poorer diets — lower vegetable and fruit intake, higher saturated fat consumption, lower dietary fibre. Lower dietary fibre is independently associated with increased bowel cancer risk. Dietary quality is therefore associated with both the exposure (processed meat eating is a marker of poor diet broadly) and the outcome (poor diet increases bowel cancer risk independently of processed meat). Unless the researchers statistically controlled for overall dietary quality, this confounder could explain part of the 38% higher risk. [1 — valid confounder, both directions of association explained]

Confounder 2 — physical activity and body mass index. High processed meat consumption is also correlated with lower physical activity and higher BMI — both of which independently increase bowel cancer risk through different biological pathways (chronic inflammation, hormonal changes). Without controlling for physical activity level and BMI, the apparent association between processed meat and bowel cancer may be partly attributable to these co-occurring lifestyle factors rather than to processed meat per se. [1 — second valid confounder, both directions of association explained]

Additional evidence required for causal inference. A single cohort study, however large, cannot establish causation. To move toward a reasonable causal inference, the Bradford Hill criteria require: consistency (the association should be replicated across multiple cohort studies, case-control studies, and different populations and countries); dose-response (higher consumption should produce proportionally higher bowel cancer rates); biological plausibility (a known mechanism — e.g. nitrosamines and haem iron in processed meat causing DNA damage to colonic epithelium); and ideally, experimental evidence (an RCT of processed meat reduction would be ethically possible, though practically challenging over the timescales needed for cancer development). Mechanistic laboratory evidence linking processed meat components to colorectal carcinogenesis would also strengthen the case. [1 — identifies at least two further evidence types needed with Bradford Hill framing]

Policy assessment. The recommendation to "immediately ban processed meat advertising" is not scientifically justified by this study alone. Policy recommendations require: (1) evidence of causation rather than merely association; (2) evidence that the association is specific to advertising rather than to consumption driven by other factors; and (3) a cost-benefit analysis of policy intervention. One large cohort study — even a well-designed one — falls short of the convergent evidence needed to support a regulatory ban. [1 — correctly assesses policy recommendation as unjustified by single study, with reasoning]

Reformulated claim. "A large Australian cohort study found that high processed meat consumption (5+ servings/week) was associated with a 38% higher bowel cancer rate over 15 years compared to low consumption, consistent with findings from multiple international studies and biologically plausible mechanisms. This adds to a body of evidence suggesting that reducing processed meat intake may lower bowel cancer risk; however, causation has not been proven from this study alone, and further research is needed before strong regulatory recommendations can be made." [1 — scientifically defensible reformulation that preserves the finding without overclaiming]

Marking criteria (7 marks):

• 1 mark — States an overall evaluative judgement identifying that the journalist has conflated association with causation and/or over-extended the evidence to policy.
• 1 mark — Explains what a cohort study actually demonstrates (association with temporal sequence) and why this is not sufficient to "prove" causation — temporal sequence is necessary but not sufficient.
• 1 mark — Identifies confounder 1 (overall dietary quality / fibre intake or equivalent) and explains both directions of association with exposure AND outcome.
• 1 mark — Identifies confounder 2 (physical activity / BMI / socioeconomic status or equivalent, distinct from confounder 1) and explains both directions of association with exposure AND outcome.
• 1 mark — Identifies at least two types of additional evidence needed (consistency across studies, dose-response, biological plausibility, experimental evidence) — framed with reference to Bradford Hill criteria or equivalent causal framework.
• 1 mark — Assesses the policy recommendation as unjustified by a single observational study, with specific reasoning (association ≠ causation; single study insufficient; additional steps needed).
• 1 mark — Provides a scientifically defensible reformulation of the claim that preserves the epidemiological finding without overstating causal certainty.