You have previously been introduced to three methods which you can use to control for confounding at the design stage of a study: randomisation, restriction, and matching. Going back to the example I use to explain the criteria for confounding. I will now discuss these three methods to examine the association between alcohol consumption and oral cancer. The first message is randomisation. Randomization is the gold standard method for controlling for confounding at the design phase, and is often used in trials. In theory, you would need to randomise participants to drink alcohol. However, as is often the case, this is not possible due to ethical reasons. As a result, you need to consider other methods to control for confounding. Another method is restriction. Smoking has previously been identified as a confounder within the literature. As a result, you could restrict the eligibility criteria so that you would only study non-smokers. This would be a simple, efficient, and inexpensive way to control for the effect of smoking. However, the literature also suggests that an association exists between smoking and alcohol consumption. Remember, in this study, 90 percent of smokers consumed alcohol compared to 27 percent of non-smokers. As a result, this would restrict the pool of potential participants. Restriction also reduces the generalisability of study findings. As you can imagine, we would be reluctant to generalise findings from a sample of non-smokers to a population of smokers. The final message used to control for confounding at the design phase is matching. Matching is particularly relevant to case-control studies. So, in our example, this would mean matching each person with oral cancer, with one or more persons without oral cancer based on smoking status. This would address the issue of confounding by smoking. But as you know, this may not be ideal. Firstly, because it would then be impossible to study the variable we used for matching, smoking, but also because there are usually multiple confounding factors, and matching for all of them makes recruitment complex and difficult. Matching also cannot be undone. As a result, our decision to match should be given a lot of thought. Now that you've seen an example of applying the three methods of controlling for confounding of the design stage, I'm sure you can now see how important it is to consider confounding ahead of commencing any study.