In a viewpoint article published in JAMA today Professors Rory Collins, Zhengming Chen and Jonathan Emberson outline the strategic need to establish some large prospective studies in carefully selected populations around the world, and stress the importance of ensuring that the data generated are widely available to research.
Prospective studies recruit individuals from the general population and then follow their health over many years. By relating information captured on the individuals at recruitment to the diseases they subsequently develop, prospective studies have been used to identify many causes of disease for example, the effect of smoking on cancer, cardiovascular disease and respiratory disease.
However, because only a minority of people in a particular study will ever develop any particular disease – and because most risk factors do not have nearly as large an impact on health as smoking – prospective studies typically need to involve hundreds of thousands of people if they are to provide reliable estimates of the associations of most risk factors with a particular disease (and especially if they are to examine how those associations might vary depending on other characteristics, such as age or sex).
The authors explain how prospective studies do not need to be representative of any given population in order to give reliable and generalisable answers about the causes of disease. Instead, they emphasise the value of studying populations that differ from each other as greatly as possible with regard to risk factor ranges and disease rates.
Using the example of blood cholesterol, they explain how studies in high-income Western populations can, ultimately, only tell us about the effects of cholesterol at levels normally seen in the West. Therefore, to study the effect of cholesterol at lower levels we would need to study a population with blood cholesterol levels much lower than those typically seen in the Western populations (such as rural China). Similarly, the relevance of a risk factor to diseases that occur only rarely in a particular study population may be missed unless other studies are done in populations where those diseases are more common.
To provide more widely generalisable evidence about the relevance of the full range of human exposures (including genetic factors) to many different diseases, the authors highlight the need for well-characterised large prospective cohorts in locations around the world with very different risk factor levels and very different disease incidence rates, and for the data from those studies then to be made widely available for research.