A major advantage of this type of study is that data can be collected quickly. One disadvantage is that if a relationship is detected between two variables, we cannot say whether these two variables are causally linked and if so, which is cause and which is effect.
Some variables may be strongly associated with several others. For example, the type of person who chooses to consume a high fibre diet is also likely to eat a low fat diet, to be a non-smoker, to take more exercise than average and is unlikely to be overweight. Some of these confounding factors can be taken into account in the statistical analyses, but to ascertain whether associations are causal, randomised controlled trials must be conducted.
The population to whom the results are applicable needs to be clearly defined. Then a random sample of that population needs to be selected to ensure that those who take part in the survey are representative of the defined population. A high response rate is important as those who refuse to take part cannot be assumed to be similar to study participants. The volume of data to be collected and the number of tests each subject is asked to undergo needs to be carefully considered. It is better to do a small number of tests that are likely to be acceptable to most subjects rather than undertake very elaborate procedures such that the response rate is poor.
Survey methods need to be standardised and reproducible. In some types of survey, interviewer variation can be a problem - both within interviewer and between interviewer variation. Laboratory measurements can be subject to 'drift' over time. This problem can be eliminated by having all specimens analysed together, if possible. Alternatively, results could be monitored to see if there are tendencies for mean values to go up or down over time.
