The future of epidemiology

It may seem a strange question--"Where will epidemiology go in the 21st century?" Why is that a consideration at all? Hardliners have suggested that epidemiology is no longer useful¹ and have gone as far as to suggest closing epidemiology departments.² Their arguments are as follows: (a) that epidemiology did well finding strong associations between common conditions and disease, but the methods we use are too crude to find anything but the largest of effects; (b) that using epidemiological methods we have found all that we can find; and (c) that the future of understanding causes of disease lies in the laboratory not epidemiology.

Not surprisingly, I disagree with those arguments. Epidemiology is a rapidly changing discipline. It has changed with the times. In the early part of the 20th century, the important public health problems in the United Kingdom (barring poverty) were infectious diseases, such as smallpox and tuberculosis. With the improvement of living conditions and the introduction of antibiotics and immunisations, the importance of infectious diseases decreased, while that of chronic diseases such as cerebrovascular disease and cancer increased. How will epidemiology influence the expected changes such as the ageing population and antibiotic resistance expected during the 21st century?

Changes in epidemiology need to occur not only in response to altering patterns of disease, but also to changes in society. Given how much life has changed over the last century, it would be naïve not to expect quite radical changes over the next 100 years. If epidemiology remains stuck in the 20th century mode of thinking it will not reflect "real life" of the time to any useful extent.

Given that changes are necessary, therefore, what can we expect to see in the future? I will describe four areas in which important changes may occur. Following up the key references given here will lead you to the disparate thoughts of famous epidemiologists on this subject.

Molecular epidemiology
A real frustration in epidemiology is that measurements we make are sometimes imprecise. This is caused by reporting errors, both deliberate and unintentional, which lead to inaccurate results. Imagine if 10% of smokers say that they do not smoke, then the difference in risk of disease between the smokers and the non-smokers will be diluted because the group that we think are non-smokers are actually smokers, with a higher risk of disease. Can molecular methods refine our measurement techniques?

There are a host of biomarkers of exposure, biological factors which reflect someone's exposure levels. Usually these are measured in samples of blood or urine. For example, measuring the amount of cotinine, the major metabolite of nicotine, can tell you whether someone has recently smoked a cigarette. Similar examples exist for various substances and are used in testing for drug misuse among sports people.

There are, however, problems with using biomarkers of exposure. Firstly, they often reflect only recent exposures, whereas many disease processes are influenced by cumulative exposure over a long period. Secondly, they tend to be expensive to measure, which often prevents their use in epidemiological studies when resources are limited. Thirdly, it is important for us to identify modifiable behaviours which can be altered to decrease risk of disease. To know the relationship between cotinine and disease is useful only if we know whether reducing smoking will directly reduce cotinine levels, and by how much. I cannot overemphasise the importance of making results applicable to real people.

Molecular epidemiology can also be used to determine biomarkers of disease. In cohort studies we are often faced with a long wait to see whether individuals will develop a disease. If we know before clinical signs develop that someone is likely to get the disease we can save much research effort, as well as alerting the individual to their high risk, and implementing a suitable screening strategy for that person.

Genetic epidemiology
Medical genetics has expanded beyond the boundaries of identifying single gene abnormalities and the role of genes in chronic diseases is now emerging. The importance of genetic epidemiology will steadily increase in this post-genome mapping era. This discipline will come into its own in the realm of gene-environment interactions. Understanding these interactions--that is, those between risk factors and a person's genetic make up--means that those people who are genetically susceptible to an environmental risk factor can be identified.

It has been suggested that genotype assessment will become a part of all epidemiological studies.³ Not taking into account the genetic make up of study participants may make it harder to find out who is at risk, since the risk factor that we are studying may affect only a proportion of the population.

It may seem odd to you that genetic epidemiology is so important. Surely epidemiology aims to identify modifiable factors, so that we can actually change our disease risk? We can influence gene-environment interactions. Take for example people who are, as a result of a genetic fault, deficient in the enzyme which metabolises phenylalanine. Babies are screened for this defect, and excluding phenylalanine from the diet of those who test positive prevents them from developing learning disabilities. Imagine a similar scenario for cancer. If we know the people in whom excessive alcohol consumption may cause breast cancer we have a targeted population in whom we can hope to reduce disease risk. Genetic epidemiology may allow the Human Genome Project to be translated into public health medicine.

Statistical models
The way that epidemiological studies are analysed has changed radically over the past 20 years with advances in computing power. Past changes are not related only to technological advances, though. Other changes as a result of the development of epidemiological methods and statistical techniques have also taken place. The introduction of the randomised controlled trial (RCT) heralded an important advance in obtaining the evidence for evidence based medicine. Furthermore, combining the results of many RCTs which address the same question uses a technique called meta-analysis. Future developments in statistical methods will allow the analysis of increasingly complex sorts of data, which, if implemented correctly will enhance epidemiological studies.

It is not only developing new methods which will take us forward. Novel uses of existing methods can be equally exciting. For example, most studies on risk factors for cancer have focused on behaviours and events in adulthood, despite the knowledge that cancer may take many decades to manifest itself. Building latency into statistical models to take into account that time lag may add to the interesting findings about how events in preadult life (prenatal, childhood, and adolescence) are related to adult disease. This area of lifecourse epidemiology, in which exposures from all ages and stages are considered as risk factors for adult disease, is an exciting theme which has emerged in the epidemiology of the past 20 years.⁴

Implementation of findings
An interesting discussion, and an area where epidemiology may well change in the future, is the translation of findings from basic research into practice.⁵ It is all very well doing useful research with interesting findings, but it is important not to let these simply grow dusty in a library somewhere. There are a multitude of pressures which have an influence on the determination of policy and practice. These include economic and political considerations as well as society's desire and ability to change. Traditionally, epidemiologists have left the area of implementation to others such as public health policymakers, health promotion professionals, and clinicians. That is fine as long as the message gets across. My hope for the future is that communication between these various disciplines improves and that useful findings are used directly to inform policymakers.

1. Taubes G. Epidemiology faces its limits. Science 1995; 269:164-9.
2. Le Fanu J. The rise and fall of modern medicine. New York: Little Brown, 1999.
3. Khoury M. Genetic epidemiology and the future of disease prevention and public health. Epidemiol Rev 1997;19:175-80.
4. Kuh D, Ben-Shlomo Y. A life course approach to chronic disease epidemiology. Oxford: Oxford University Press, 1997.
5. Colditz G. Epidemiology--future directions. Int J Epidemiol 1997; 26: 693-7.