Abstract
Wondering why we no longer give aspirin to people with atrial fibrillation, or how we know that smoking is a major risk factor for lung cancer? Those questions have already been answered by studies, but because treatment effects are not often very dramatic, (and it is often not clear which treatment is best, or whether new ones work/cause harm), we need well-designed studies to tell us answers to new questions. This brief article tells you what you need to know about the different types of study designs, covering the key competencies from the AKT curriculum.
As bad studies lead to misleading evidence, research teams need to:
Choose the right study design to answer their particular question Spend time making sure the study is well-designed (i.e. there are no flaws in the design of the study that would make the final answer unreliable)
There are pros and cons of each study design.
Case-control studies
Case-control studies are observational studies (meaning that they only describe what is there, they do not intervene); they are often used to identify factors that might be associated with a specific medical condition. They do this by taking a group of people with the condition of interest (the ‘cases’), and matching them to patients who are otherwise similar, but who do not have the condition (‘controls’). They then retrospectively look back in time at risk factors to see if they might be statistically associated with the condition of interest. Richard Doll and Bradford Hill used a case-control study design to demonstrate the link between smoking and lung cancer (Doll & Hill, 1950). Case-control studies are relatively cheap to perform, and are also quick (because the cases have already happened). They are less reliable evidence due to their susceptibility to a range of biases.
Cohort studies
A cohort is a group of people sharing a common characteristic or experience within a defined period (this could be year of birth, exposure to a certain drug/environment, etc.). Cohort studies are observational studies that take a cohort of people and follow them over time to see who/how many develop the outcome of interest (usually a specific disease). These are usually prospective, but can also be retrospective. They are good for studying the effects of risk factors on an outcome (e.g. living by a telephone mast and the occurrence of brain tumours). Cohort studies are cheaper to perform than randomised control trials (RCTs), and you can have quite a lot of control over who you choose to put in your study. The main disadvantage is usually length of time to perform the study (particularly if you’re looking at a rare condition or one that takes a long time to develop). The Whitehall study (originally a cohort of British civil servants) and the Framingham Heart Study are two famous examples of cohort studies.
Cross-sectional studies
Cross-sectional studies are also observational studies, but they only look at one specific point in time. Whereas a case-control study looks at individuals with a certain characteristic (e.g. people with lung cancer), cross-sectional studies usually look at whole populations. They are often used to assess the prevalence of conditions in the population. They are relatively cheap and simple, but can only ever show association not cause. They may also be subject to recall bias.
Randomised control trials
RCTs are usually used to test whether treatments work, and how well they work. RCTs are experimental trials in which you allocate participants to either treatment or control groups using a random mechanism. The idea of randomisation is that all known and unknown biases are evenly distributed between groups in the trial. Because (in theory) RCTs reduce bias they are considered a reliable form of scientific evidence. However, it is important to realise that a poorly conducted trial will give unreliable results. RCTs can also be expensive, take a long time to perform and be ethically difficult (for example, randomising groups of people who lack capacity to consent). The All Trials campaign calls for greater transparency in the publication of trial data, so that the scientific community can make more informed judgements about what does and does not work.