What you need to know
- 1.16Sample selection, sample size, validity and reliability of risk-factor studies.
- 1.15Interpreting associations and conflicting evidence.
By the end, you should be able to…
- Compare cohort, case-control and intervention studies.
- Explain selection bias, confounding and loss to follow-up.
- Suggest specific improvements to a study.
Build your understanding
Follow forwards or look backwards
A prospective cohort study measures exposures in people and follows them for later outcomes. It helps establish that exposure preceded the outcome, but can take years and may lose participants. Exposure can also change during follow-up.
A case-control study starts with people who have a condition and a suitable comparison group without it, then compares past exposures. It can be useful for rare outcomes but is vulnerable to recall bias. Because the researcher chooses the number of cases and controls, their proportions do not directly give populationAll the individuals of one species living in one area at the same time. disease risk.
Interventions and fair comparisons
In a randomised controlled trial, participants are assigned to intervention or comparison groups by chance. This reduces systematic differences in confounders between groups, although it does not guarantee perfect balance in every sample. Blinding, where feasible, reduces behaviour or assessment influenced by knowledge of group allocation.
Random sampling selects people from a population; random allocation assigns enrolled people to study groups. They solve different problems. An ethical intervention can test a treatment, but researchers cannot deliberately assign people to years of harmful smoking simply to test its effects.
Validity and reliability
A representative sample improves how well findings apply to the target population. Recruiting only highly motivated gym members could distort a study intended to describe all adults. Use clearly defined outcomes, standardised measurements and comparable groups to improve validity.
Larger samples reduce random sampling variation and improve precision, but do not remove systematic bias. Repeated, standardised measurements can improve reliability. Record and account for confounders such as age and smoking, and examine whether people lost to follow-up differ from those retained.
Evaluate the actual study
Identify a concrete limitation, explain its likely effect, then suggest an improvement. For example, self-reported smoking may be underreported, misclassifying exposure; confidential standardised questionnaires and an appropriate independent measure can improve assessment.
Conflicting results may reflect different populations, exposure definitions, durations, sample sizes or chance. Compare the methods and uncertainty before deciding whether studies truly contradict one another.
Make your reasoning visible
Avoid generic evaluation
“Use more people” needs a reason, such as reducing sampling variation. It does not solve recruitment bias or a poorly defined outcome.
Separate randomisation and blinding
Random allocation addresses group differences. Blinding addresses expectations and biased assessment. Explain the purpose of each.
Avoid this misconception
The claim: A very large sample guarantees a valid conclusion.
The correction: Large samples can give a precise but biased estimate if sampling or measurement is systematically flawed.
Check your understanding
Try each question before opening the suggested answer.
Practice question
Retrieval and explanation · suggested answer
A dietary study recruits 5000 volunteers from a fitness club. Identify a limitation and explain it.
Show suggested answer
Fitness-club volunteers may differ from the wider population in activity, diet or motivation. The large sample does not remove selection bias, limiting generalisation.
Practice question
Retrieval and explanation · suggested answer
Why might a cohort with 40% loss to follow-up give a biased result?
Show suggested answer
If loss is related to exposure or disease, those remaining no longer represent the original groups. Comparing characteristics and reasons for loss helps assess this problem.
Take the idea further
Apply the same specification knowledge to an unfamiliar problem. Any extra assumptions are supplied in the question.
Stretch question
A trial uses random allocation, but only participants receiving the treatment know their group. Is randomisation sufficient to remove all bias?
Show a hint
Distinguish baseline differences from behaviour after allocation.
Explore the reasoning
No. Randomisation reduces systematic baseline differences, but knowing the treatment can affect behaviour, reporting or adherence. Blinded outcome assessment and objective outcomes can reduce some remaining bias; complete participant blinding is not always possible.
Recall this lesson
Lesson recall questions
Question 1 of 5: Compare correlation with causation using a fictional table in which ice-cream sales and heart-attack admissions both rise in July.
Illness data: correlation and causation
Practice question
Retrieval and explanation · suggested answer
Compare correlation with causation using a fictional table in which ice-cream sales and heart-attack admissions both rise in July.
Lesson: 1C 1: Risk, Correlation and Cause
Try this from memory first.
Put your biology to work
Explore where these ideas appear in the world. Enrichment, not extra required recall.
Following families across generations
1C · Investigating the Causes of CVDsThe Framingham Heart Study follows cardiovascular health over time and across generations. Its long-term design illustrates how researchers investigate factors that precede disease. Following participants over time can…Specification 1.16Read & use the scienceClose article
Why this matters
The Framingham Heart Study follows cardiovascular health over time and across generations. Its long-term design illustrates how researchers investigate factors that precede disease. Following participants over time can establish that a measured exposure preceded an outcome, an advantage over a one-off snapshot. Repeated measurements can also reveal that risk factors change. Yet participants are not randomly assigned to their lives: smoking, diet, income and other influences can overlap. An epidemiologist uses careful comparisons and adjustment, while acknowledging remaining uncertainty. A long-running study is valuable evidence for cardiovascular risk, but its duration alone cannot turn every association into a proven causal mechanism.
Back to the specification
Pearson Edexcel IAL Biology · Issue 2, February 2021 · Unit 1, pp. 15–18. The application extends beyond required recall.
- 1.16 be able to evaluate the design of studies used to determine health risk factors, including sample selection and sample size used to collect data that is both valid and reliable
Use the science
Why could measuring an exposure repeatedly be better than measuring it only at enrolment?
Explore a possible response
Diet, activity or smoking can change. Repeated measurements reduce misclassification and provide a more accurate picture of exposure over time.