Epidemiology Notes

Hierarchy of Epidemiological Evidence

Case Control Study (Reverse of Cohort)

 Definition – study design in which a group of people with disease/condition are compared with
suitable controls from the same population, and prevalence of past exposures is collected
 Design
1. Identify cases (case definition, inclusion + exclusion criteria, identify population from which
cases come)
2. Recruit control subjects (without disease) from same population = control definition,
exclusion of disease, matched or unmatched, identify how controls will be identified
3. Measure previous exposure = which methods of collection e.g. questionnaire, records,
biomarkers
4. Compare odds of exposure between cases and controls i.e. odds ratio (cannot determine
relative risk as YOU picked the participants, so ratio of probability in exposed to unexposed
is entirely determined by YOU)

Cohort Study
 Definition – study design where exposure is assessed and samples are followed prospectively
and evaluated with respect to disease/outcome to determine associated risk factors. Can be
prospective or retrospective.
 Advantages
- Subjects in cohort can be matched which limits influence of confounding variables
- Standardisation of criteria/outcome
 - Easier and cheaper than an RCT
 Disadvantages
- Difficult to find cohorts due to confounding variables (stratification eliminates)
- No randomisation – so imbalances in patient characteristics
- Blinding/masking is difficult
- Subjects lost to follow up

Design Pitfalls
 Cohorts must be chosen from separate but similar populations
 Exposure must precede outcome
 Note that study proves only correlation and not causation
Practicalities

1. Define the population and select a sample free of

o Inclusion and exclusion criteria
1. Multivariable regression > stratification of risk factors
o Controls
1. Population – ideal definition = ‘adult resident without this’ but
uptake might be low and biased
2. Hospital controls – may have similar investigations previously,
similar behaviours BUT not random sample (more likely to do other
RFs, so generalisability of study is weakened)
3. Matching – age, gender, anything else?

Are there any BIASES introduced by this grouping

2. Identify exposed and non-exposed (if present at start of trial)

o Methods of collection (questionnaire, records, biomarkers etc)
3. Follow-up the sample over time
o Frequency & method of follow up
o Information to be collected
1. Some risk factors hard to measure / liable to induce recall bias or
social desirability bias – solution = ACASI i.e. computerised self-
administered interview technique which improves accuracy of
behaviour reporting where interviewees reluctant to disclose
behaviour to real interviewers
2. Objective measures = ideal
3.
4. Measure the occurrence of outcome(s) over time
o Case definition
5. Compare the incidence of outcomes between exposure group (risk or rate ratio)

Statistics
- Odds ratio =
 - Relative risk =

Real-life Examples

Hoepner, L., Whyatt, R., Widen, E., Hassoun, A., Oberfield, S., Mueller, N., ... Rundle, A.
(2016). Bisphenol A and Adiposity in an Inner-City Birth Cohort. Environmental Health Perspectives,
124(10), 1644-1650. https://doi.org/10.1289/EHP205

This longitudinal cohort study looked at whether exposure to bisphenol A (BPA) early in life affects
obesity levels in children later in life. Positive associations were found between prenatal BPA
concentrations in urine and increased fat mass index, percent body fat, and waist circumference at
age seven.
Lao, X., Liu, X., Deng, H., Chan, T., Ho, K., Wang, F., ... Yeoh, E. (2018). Sleep Quality, Sleep Duration,
and the Risk of Coronary Heart Disease: A Prospective Cohort Study With 60,586 Adults. Journal Of
Clinical Sleep Medicine, 14(1), 109-117. https://doi.org/10.5664/jcsm.6894

This prospective cohort study explored "the joint effects of sleep quality and sleep duration on the
development of coronary heart disease." The study included 60,586 participants and an association
was shown between increased risk of coronary heart disease and individuals who experienced short
sleep duration and poor sleep quality. Long sleep duration did not demonstrate a significant
association.