Cohort studies

Question 1

What are cohort studies?

Answer 1

Cohort studies are observational rather than experimental. They allow us to study at the population level and across populations and lifespans. They can give evidence of effectiveness in real life situations. Most importantly, they allow us to study prevalence, risk factors and intervention for clinical groups where robust research data are not yet available

Question 2

Which kind of bias is the biggest problem in cohort studies?

Answer 2

Selection bias- designing a good cohort study is essential to reduce this

Question 3

Comparing cohort studies to RCT- populations studied

Answer 3

Cohort study- diverse populations of patients observed in a range of settings

RCT- highly selected populations recruited on the basis of detailed criteria

Question 4

Comparing cohort studies to RCT- allocation to the intervention

Answer 4

Cohort- based on decisions made by providers or patients

RCT- Based on change and controlled by investigators

Question 5

Comparing cohort studies to RCT- outcomes

Answer 5

Cohort- can be defined after the intervention and can include rare/unexpected events.

RCT- primary outcomes are determined before patients are entered into the study and are focussed on predicted benefits/risks.

Question 6

Comparing cohort studies to RCT- Follow-up

Answer 6

Cohort- many cohort studies rely on existing experience (retrospective studies) and provide opportunity for long-term follow up. Longitudinal- can track what happens over time

RCT- Prospective studies: often have short follow-up because of costs and pressure to produce timely evidence.

Question 7

Comparing cohort studies to RCT- Analysis

Answer 7

Cohort- sophisticated multivariate techniques may be required to deal with confounding

RCT- Analysis is straightforward.

Question 8

How to design a good cohort study?

Answer 8

Who does or does not receive an intervention is determined by things like practice patterns, personal choice, or policy decisions. It is not random, introducing an element of selection bias.
So there should be
1. Clear choice of comparison groups - Helps maintain selection bias whilst maintaining clinical relevance. This includes a clear definition of the two comparison groups and how they were selected.

2. Restricted characteristics of the group so they become more the same- removes potential biases or confounders, but this is also a limitation (reduced sample size and power; limits reproducibility to larger populations (cf. RCTs).
3. Clearly defined confounders

Confounders are factors that might influence the results of the outcome, because they are in some way related to the outcome or may be unequally distributed across the groups, for example age. They may be known or unknown. For known confounders you should ensure a thorough search of the evidence base to ensure you control for all of these. For example, it is often reported that girls have better language development than boys, so sex might be a confounder in a study of child language. Unknown confounders are more difficult and often a limiting factor in the results of a cohort study. These are often detected in statistical analysis.

4. Clear and robust statistical analysis

Question 9

Potential limitations of cohort studies

Answer 9

Uneven groups (number or characteristics)
Selection bias
Cofounders (is there info about how potential confounders are distributed between comparison groups?)
Poor statistical analysis