Benefits of Epidemiological Studies

Q1, Marking guidelines, screening evaluation (8 marks)

Survival figures and bias (2 marks). 1 mark for identifying that the jump from 52% to 81% five-year survival is not evidence of lives saved. 1 mark for naming lead-time bias and explaining it correctly: screening moves the diagnosis date earlier, so survival measured from diagnosis lengthens even if the person dies on exactly the same day they would have anyway.

Mortality data (2 marks). 1 mark for identifying that mortality is the measure that cannot be inflated by earlier diagnosis, because death is counted per head of population rather than from the diagnosis date. 1 mark for reading it correctly: 44 versus 47 deaths per 100,000 is a difference of only 3 per 100,000, roughly a 6% relative reduction, which is far smaller than the survival figures imply and may not be statistically or practically significant.

Excess detected cases (1 mark). 1 mark for recognising that detection more than doubles (420 versus 190 per 100,000) while deaths barely move, which is the signature of overdiagnosis: the extra cases largely comprise disease that would never have caused symptoms in that person's lifetime, yet those people are counted as cured and flatter the survival statistic.

Harms (1 mark). 1 mark for weighing the recorded harms: 6,300 false positives per 100,000 means about 6.3% of participants are told they may have cancer when they do not, carrying anxiety, further investigation, procedural risk and cost, and the overdiagnosed cases carry treatment side effects they never needed. This indicates low specificity.

Judgement with equity (2 marks). 1 mark for an explicit, data-referenced judgement. 1 mark for addressing equity using the participation split. A Band 6 response argues that the program's benefit is much weaker than the headline survival figure suggests: the only unbiased measure shows a small mortality reduction, purchased with substantial overdiagnosis and a high false-positive burden, so the program cannot be called clearly beneficial on this evidence. It further notes that participation of 71% in the highest income quintile against 29% in the lowest means whatever real benefit exists accrues mainly to the group already advantaged, so the program risks widening the health gap. Accept a judgement that the program should continue subject to raising specificity and targeting participation, provided it is argued from the data.

Band guidance. Band 6 names lead-time bias and overdiagnosis by their proper terms, distinguishes survival from mortality explicitly, and still reaches a judgement. Band 4 describes the table or notes that "screening has limitations" without naming the biases or without judging.

Q2, Marking guidelines, source critique (8 marks)

Award 1 mark for each error correctly identified and 1 mark for each accurate correction, to a maximum of 6, plus up to 2 marks for the judgement. Expected errors include:

• "Because the relative risk is above 1, causation is established." An association is a starting point, not proof. Evidence worth acting on must be strong, consistent across multiple studies and populations, show the exposure clearly preceding the disease, and be biologically plausible. A single relative risk of 1.4 satisfies none of these on its own, and confounding has not been excluded, since coffee drinkers may differ in sleep, stress or workload.
• The study design and size. A survey of 210 adults in a single town is a small cross-sectional sample. It measures exposure and outcome at the same moment, so it cannot establish that coffee drinking preceded the headaches, and the result may not generalise beyond that town.
• "Launch a national campaign immediately." Weak, single-study evidence rarely justifies national action. The defensible response is to treat the finding as a hypothesis for stronger studies. Acting prematurely spends public money and credibility on an association that may not survive replication.
• "A study is beneficial the moment it is published." Publication produces awareness, not benefit. A study delivers benefit only when the whole chain holds: finding, a decision by someone with power to act, implementation the population can reach, and an outcome check showing exposure or disease actually changed.
• Magnitude ignored. The article reports only relative risk. Without the baseline absolute risk of headaches, a 1.4-fold increase cannot be judged for importance, and no attributable-risk estimate is offered to indicate how many cases removing coffee would actually prevent.
• Source quality. No study link, sample details or funding statement is provided, so the figure cannot be traced back to its source.

Judgement (2 marks). A defensible response is that the finding warrants no public-health action at present. It should be recorded as a hypothesis and tested with a larger, preferably prospective study in multiple populations that controls for likely confounders such as sleep, stress and analgesic use, and that reports absolute as well as relative risk. Only if the association proves strong, consistent, temporally correct and biologically plausible would any campaign be justified, and even then attributable risk should be compared against competing prevention priorities before funding is committed. Full marks require the student to reject the article's conclusion without dismissing the underlying question as unworthy of study.