IJE Advance Access originally published online on March 15, 2006
International Journal of Epidemiology 2006 35(3):581-583; doi:10.1093/ije/dyl004
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Commentary |
Commentary: Complex disease—responding to the challenge
MRC Epidemiology Resource Centre, University of Southampton, UK
MRC Epidemiology Resource Centre, Southampton General Hospital, Southampton SO16 6YD, UK. E-mail: dnc{at}mrc.soton.ac.uk
The thought-provoking paper by Buchanan et al.1 raises fundamental questions. Has epidemiology passed its zenith? If so, why? And what should be our response?
In its early years, epidemiology generated many important discoveries, but despite more sophisticated methods and much greater investment of resource, output over the past two decades has been less remarkable. This could be because today's epidemiologists are less competent than their predecessors, but while there is no doubting the innovation and application of those who pioneered the discipline, a substantial decline in standards seems unlikely. A more plausible explanation is that nowadays, major discoveries are harder to make.
Buchanan et al. attribute this to the complexity of the diseases that we now address, drawing a contrast with disorders resulting from simple Mendelian inheritance, infection, or environmental toxins, for which it is possible to identify a single causal agent that is both specific and highly predictive. When a cause of disease is both necessary and sufficient, or nearly so, it should be easier to spot and confirm. However, causes that do not satisfy this strict requirement should still be readily demonstrable if they carry a high relative risk. Exposure to aromatic amines is neither necessary nor sufficient as a cause of bladder cancer, but the relative risk is high, and with the appropriate study design, was clearly established.2 Thus, although by Buchanan's definition, bladder cancer is a complex disease, it has, nevertheless, been possible to make useful progress in its prevention, if only for a subset of the population.
The limit to important new discoveries is not the complexity of disease as such, but the number of unrecognized causes that carry high relative risks. Exposures carrying large and statistically robust relative risks for chronic diseases do still come to light from time to time—for example, the risk of malignant melanoma in people with multiple benign moles,3 and the risk of knee osteoarthritis in relation to obesity.4 However, it is unusual for a new risk factor affecting large numbers of people to be confirmed as carrying a relative risk >2 for a major disease.
Epidemiologists have responded to this challenge in various ways. One approach has been to conduct very large studies with detailed assessment of exposure to a wide range of known and suspected risk factors and then to apply advanced statistical techniques in an attempt to distinguish their independent effects. With this method, even relative risks <1.5 can be statistically significant, and by careful control for potential confounding factors the chance of inappropriate causal attribution should be minimized. In a few cases, the findings have been sufficiently consistent across a number of studies that they have achieved wide acceptance and have formed the basis for public health policy. An example is the regulation of particulate air pollutants, which even at relatively low concentrations, have been found to carry a small but consistently increased risk of cardio-respiratory mortality.5
But even when multiple, carefully controlled studies produce similar results, caution is needed if relative risks are small. Observational studies strongly suggested that beta carotene protects against cancer and that hormone replacement therapy protects against coronary heart disease, but neither of these findings was confirmed when randomized intervention studies were carried out.6,7 Presumably, the problem lay in unrecognized residual confounding.
Another ploy has been to exploit technological advances that allow new exposures to be studied. The most salient example is the use of genotyping in genetic studies. Here, there is less potential for bias and confounding than in other areas of epidemiology, and lower relative risks can be interpreted with greater confidence.8 Some findings hold promise of useful therapeutic applications. For example, demonstration that asthma and bronchial hyper-responsiveness are linked with a gene, ADAM33, that codes for a membrane metalloprotease, has opened up the possibility of new pharmacological targets.9 Overall, however, the return from substantial research effort over more than 10 years is disappointing when set against what was achieved by a small number of epidemiologists in a single decade from 1950.
A third strategy has been to focus on diseases that in the past received less attention because they were relatively rare or regarded as less serious. Here, some useful advances in understanding have been achieved, such as the unusually high risk of hip osteoarthritis in farming and other occupations that entail frequent heavy lifting.10 But even for these previously ‘neglected’ diseases, it has not always been easy to identify causes carrying high relative risks.
Overall, therefore, there has been an increase in activity, but in proportion to the outlay of time and money, fewer important discoveries. Clearly, there is a possibility of disenchantment. Funding bodies may become less inclined to invest in epidemiology, and it is notable that currently the largest single epidemiological initiative in the UK, the Biobank project, depended for its support on backing from the powerful genetics lobby. In Britain, this threat to funding is particularly worrying because it coincides with an increasingly restrictive research environment. Primary research in epidemiology is becoming progressively more difficult because of concerns about data protection, ever increasing emphasis by ethics committees on the personal autonomy of potential study participants, disproportionate complexities of research governance, and growing reluctance of the public to participate in research when they are perpetually bombarded with commercial surveys by both mail and phone. A consequence is that studies have become more expensive, the biggest proportionate increase being for smaller and simpler investigations.
How should we respond? First, we must be clear what we are trying to achieve. Is the ultimate aim to understand exactly why one person gets a disease and another does not? If so, as has been argued elsewhere, success is unliklely.11 In many cases, whether or not someone develops a disease may depend on combinations of events at a molecular or cellular level that we can never hope to measure. And even if a causal factor systematically increases the risk of disease by 50%, this will be difficult to demonstrate with confidence by epidemiology. At best we might hope to obtain clues to causation that can be tested in other ways.
The alternative is to view epidemiology not as a means of understanding the natural world but as an applied science, which can help to form practical decisions in the prevention and management of illness. This goes back to its roots. When John Snow carried out his research on cholera, his principal goal was the prevention of disease and not the advancement of scientific knowledge. And there are many areas of public health and clinical medicine that require epidemiological input. How should we respond most effectively when new infectious diseases such as SARS emerge? Does new technology such as mobile telephony pose health risks that require regulatory control? What are the most appropriate restrictions on driving for patients with epilepsy? In responding to questions like these, it does not matter if answers cannot be definitive. Decisions must be made on the best information that is available.
Linked to this, there is a continuing need for effective systems of health surveillance so that new problems such as AIDS can be detected and investigated in timely fashion. Surveillance systems also provide a check on the effectiveness of control measures. For example, the continuing increase in mortality from mesothelioma in Britain is an indication either that earlier limits on asbestos exposure were unsatisfactory or that they were inadequately implemented.12
In picking on aetiological hypotheses to investigate, we need to be more discriminating. The key to many of the most important epidemiological discoveries has not been an advance in technique but asking the right question. A good example is the demonstration that risk of sudden infant death is importantly reduced if babies are placed to sleep on their back.13 Once the hypothesis had been conceived, testing it was not too difficult and did not require complex methods.
Of course, identifying good study questions is easier said than done, but there are helpful pointers. One situation with the potential to be more fruitful occurs when the incidence of a disease is increasing. As Buchanan and colleagues point out, such trends must have environmental causes, and if the increase in incidence is large (as, for example, with testicular cancer14) then those causes are likely to carry high relative risks. In addition, some types of exposure offer more promise than others. One of the most fruitful areas for epidemiological research in recent decades has been the hazards associated with pharmaceuticals. Not only can relative risks be fairly high but also exposure is easier to ascertain than for many other risk factors. Moreover, the associated burden of disease may be large, as for example, with myocardial infarction and rofecoxib.15,16
Finally, we should extend our horizons to end-points other than disease incidence and mortality. For example, one area that is of immediate practical relevance, and much in need of further research, is the distribution and determinants of people's personal exposure to chemical pollutants in the environment. Here, the outcome of interest may be an internal dose measured by a biomarker and the risk factors different aspects of individual lifestyle and environment. Thus, it has been shown that residence near a petrochemical plant had negligible impact on an individual's uptake of benzene in comparison with active smoking, living with a smoker, frequently filling a car with petrol, or driving in traffic.17 This information is extremely useful in responding to leukaemia clusters near to refineries.
In summary, epidemiology still has much to offer, but we cannot expect it to provide a complete understanding of why some people get a disease and others do not. Nor should we expect major discoveries to occur with the same frequency that they did in the 1950s, 1960s, and 1970s. However, by focusing on questions of immediate practical importance, together with judicious selection of more speculative hypotheses for exploration, epidemiologists will continue to make a telling contribution to public health.
 |
References |
|---|
 Top References |
|---|
1 Buchanan A, Weiss K, Fullerton S. Dissecting complex disease: The quest for the philosopher's stone? Int J Epidemiol 2006;35:562–71.
2 Case RA, Hosker ME, McDonald DB, Pearson JT. Tumours of the urinary bladder in workmen engaged in the manufacture and use of certain dyestuff intermediates in the British chemical industry. I The role of aniline, benzidine, alpha-naphthylamine and beta-naphthylamine. Br J Ind Med 1954;11:75–104.[ISI][Medline]
3 Bataille V, Bishop JA, Sasieni P, Swerdlow AJ, Pinney E, Griffiths K et al. Risk of cutaneous melanoma in relation to the numbers, types and sites of naevi: a case-control study. Br J Cancer 1996;73:1605–11.[ISI][Medline]
4 Coggon D, Reading I, Croft P, McLaren M, Barrett D, Cooper C. Knee osteoarthritis and obesity. Int J Obes Relat Metab Disord 2001;25:622–27.[CrossRef][ISI][Medline]
5 Bell ML, Sarnet JM, Dominici F. Time-series studies of particulate matter. Ann Rev Pub Health 2004;25:247–80.[CrossRef][ISI][Medline]
6 Lawlor DA, Davey Smith G, Bruckdorfer KR, Kundu D, Ebrahim S. Those confounded vitamins: what can we learn from the differences between observational versus randomised trial evidence? Lancet 2004;363:1724–27.[CrossRef][ISI][Medline]
7 Lawlor DA, Davey Smith G, Ebrahim S. Commentary: The hormone replacement - coronary heart disease conundrum: is this the death of observational epidemiology? Int J Epidem 2004;33:464–67.
8 Davey Smith G, Ebrahim S. What can mendelian randomisation tell us abut modifiable behavioural and environmental exposures? BMJ 2005;330:1076–79.
9 Van Eerdewegh P, Little RD, Dupuis J, Del Mastro RG, Falls K, Simon J et al. Association of ADAM33 gene with asthma and bronchial hyperresponsiveness. Nature 2002;418:426–30.[CrossRef][Medline]
10 Coggon D, Kellingray S, Inskip H, Croft P, Campbell L, Cooper C. Osteoarthritis of the hip and occupational lifting. Am J Epidemiol 1998;147:523–28.
11 Coggon D, Martyn C. The stochastic nature of disease causation. Lancet 2005;365:1434–37.[CrossRef][Medline]
12 Peto J, Hodgson JT, Matthews FE, Jones JR. Continuing increase in mesothelioma mortality in Britain. Lancet 1995;345:535–39.[CrossRef][ISI][Medline]
13 Fleming PJ, Gilbert R, Azaz Y, Berry PJ, Rudd PT, Stewart A et al. Interaction between bedding and sleeping position in the sudden infant death syndrome: a population based case–control study. BMJ 1990;301:85–89.[ISI][Medline]
14 Joffe M. Are problems with male reproductive health caused by endocrine disruption? Occup Environ Med 2001;58:281–88.
15 Juni P, Nartey L, Reichenbach S, Sterchi R, Dieppe PA, Egger M. Risk of cardiovascular events and rofecoxib: cumulative meta-analysis. Lancet 2004;364:2021–29.[CrossRef][ISI][Medline]
16 Graham DJ, Campen D, Hui R, Spence M, Cheetham C, Levy G et al. Risk of acute myocardial infarction and sudden death in patients treated with cyclo-oxygenase 2 selective and non-selective non-steroidal anti-inflammatory drugs: nested case–control study. Lancet 2005;365:475–511.[ISI][Medline]
17 Wallace LA. Major sources of benzene exposure. Environ Health Perspect 1989;82:165–69.[ISI][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M. J KHOURY and M. GWINN Genomics, epidemiology, and common complex diseases: let's not throw out the baby with the bathwater! Int. J. Epidemiol., October 1, 2006; 35(5): 1363 - 1364. [Full Text] [PDF]
|
|
|
|
|
|
A. V Buchanan, K. M Weiss, and S. M Fullerton On stones, wands, and promises Int. J. Epidemiol., June 1, 2006; 35(3): 593 - 596. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||