It addresses a question that is as important today as it was in 1988. Countless studies are conducted each year of the effects of what might be termed ‘everyday’ or lifestyle factors that might be thought detrimental to health: tobacco, alcohol, sugar, coffee and so on. How can we distinguish between authentic adverse effects – things that we might do something about – and those that might be caused by chance, coincidence and additional risk factors that have not been considered, or simply bad science?
To put it another way, should we suspend belief when a national newspaper reports that eating chocolate cake will only make you put on weight if you’re worried that it will? The discipline that examines such problems is epidemiology. Feinstein, who died in 2001, summarises his vision of the future of epidemiological research with the words: 'During the next 25 years, the methodological lessons taught by randomized trials can lead to new paradigms, concepts, and approaches that will achieve fundamental scientific standards when randomized trials are not possible. The investigators will have to focus more on the scientific quality of the evidence, and less on the statistical methods of analysis and adjustment.'
The message is that, in comparison with randomized trials, the quality of many epidemiological studies is inadequate. No amount of statistical analysis can salvage a flawed study, one in which the data is incomplete or inadequate. There is a message here for both disciplines. Feinstein’s paper also gives a reminder of how difficult it can be to anticipate the future direction of any discipline.
Among the topics he implies will generate substantial additional research in years to come are Berkson’s bias, an inconvenient truth relating to differential rates of diagnosis that is still often simply ignored by researchers; and ‘epidemiologic necropsy’, a method that uses post-mortems to reduce the rate of disease underdiagnosis. It is fair to say that neither has been at the forefront of epidemiological research during most of the last 25 years.
It remains as difficult as ever to identify the real 'menaces of daily life' today, to 'separate the wheat from the chaff', so to speak. It is more difficult still to communicate them in a way that is clear, objective and balanced. Indeed, the whole subject is so difficult that it has become somewhat fashionable for epidemiologists to sound the death knell of their own discipline. ‘Science’ published another paper in 1995 with the title ‘Epidemiology Faces Its Limits’, while even the editors of one of the most globally prestigious journals in the discipline were moved to ask ‘Epidemiology – is it time to call it a day?’ in 2001.
I will make one prediction for 2038: epidemiology as a subject will have survived, but we'll still be asking the same questions.