Brain, Vol. 125, No. 6, 1414-1416,
June 2002
© 2002 Guarantors of Brain
Book Review |
PRINCIPLES OF NEUROEPIDEMIOLOGY
MRC Environmental Epidemiology Unit, University of Southampton, Southampton, UK
PRINCIPLES OF NEUROEPIDEMIOLOGY
Edited by Tracy Batchelor and Merit E. Cudkowicz
2001. Oxford: Butterworth-Heinemann
Price £65. Pp. 400. ISBN 0750670428.
Pots and pans but no pudding
Rinyaszentkirály is a small village in the south west of Hungary with a population of less than 500 souls. Eleven of the 15 babies born there in 1989 and 1990 had serious congenital abnormalities. The first reaction, when news of this apparent epidemic reached official ears, was to put it down to misdiagnosis by a country doctor. But when a team of geneticists and paediatricians was sent to examine the babies, the outbreak proved to be real enough. Epidemiologists know from bitter experience that most clusters of disease are chance events but, even allowing for the likelihood of inbreeding in an isolated community, this explanation wouldn’t wash. The probability of 11 affected births out of 15, when the expected rate of congenital abnormalities is 
6%, is less than one in a billion.
The investigating epidemiologists found no evidence of unusually high rates of congenital abnormality in nearby towns and villages. In Rinyaszentkirály itself, congenital abnormalities, miscarriages and still births before 1989 had been no more frequent than expected. None of the parents had even minor expressions of the abnormalities found in the children. Baffled, the investigators went on to carry out a case control study to compare family histories and environmental exposures of the mothers of the abnormal babies with those of mothers of healthy children born in Gorgeteg, a nearby village, and of mothers of healthy children born in Rinyaszentkirály before 1989. The results ruled out most of the likely candidates for the cause of the cluster. There was no indication that the mothers of the affected babies had been exposed to any known teratogenic factors, rates of consanguinity were no higher than in the control groups and there had been no epidemic of rubella.
But the mothers of the affected children did report that they had all eaten fish from the local fish farm while they were pregnant. When this unpromising lead was pursued, it was discovered that, unknown to the inhabitants of the village, this fish farm had been experimenting with a new method of eradicating parasites known as chemical fast bathing. The process involved fish being taken out of their ponds and briefly dunked in trichlorfon, an organophosphorus insecticide. Following treatment, the fish were returned to their ponds where they often remained, apparently lifeless, near the surface for several hours. Fishing was prohibited during the period following treatment but, as the investigators were later to phrase it in their report: ‘this is a poor village and people continued to eat the fish which were often taken out by hand in a comatose state’.
Although trichlorfon has not been proven to have a teratogenic or germinal mutagenic effect in man, the circumstantial evidence strongly suggests that this organophosphate was the culprit. Trichlorfon is well absorbed orally and rapidly distributed to the tissues of the body. It can cross the placenta. The concentration in treated fish was reckoned at 100 mg/kg so it was very likely that anyone eating them greatly exceeded the acceptable daily trichlorfon intake of 0.01 mg/kg bodyweight. What is more, the epidemic of congenital abnormalities ceased when this chemical treatment was banned. All 10 children born in the village in 1991 and 1992 were healthy.
Unless I have told it exceptionally badly, I shall be surprised if you didn’t find this miniature of applied epidemiology an interesting tale. It has, after all, most of the elements that make the detective story a popular genre: an apparently inexplicable event that eventually yields to tenacious investigation to provide a satisfying dénouement on the last page. And it’s hardly unique. I could have recounted many other fascinating epidemiological detective stories: the epidemic of thyrotoxicosis that turned out to be caused by the incomplete exclusion of bovine thyroid gland from hamburgers; the outbreak of congenital cataract that led to the recognition of the teratogenic influence of rubella infection in pregnancy; the analyses of deaths by occupation that contributed to the understanding of the causes of nasal cancer and mesothelioma. The list is long. Epidemiological investigations directed at uncovering the causes of diseases are clearly not intrinsically dull. So it’s something of a puzzle that Principles of Neuroepidemiology (and, I am sorry to say, most books about epidemiology) is so grindingly boring.
If it isn’t the subject, perhaps it’s the people who do it. Could it be that epidemiologists are by nature dull—the quantity surveyors of the medical world? As an epidemiologist myself, I cannot claim impartiality here but I don’t think this is a credible explanation. My experience is that epidemiologists are amusing people to sit next to at dinner. They drink the claret for its life enhancing properties, not for its cardioprotective effects. Their professional engagement with the estimation of risk and attribution of cause engenders a robust and philosophical view of the vicissitudes of life. Like clinicians, whose daily contact with sick patients provides a yardstick against which their own misfortunes seem trivial, epidemiologists are inclined to make the best of what life offers rather than whinge about its unfairnesses.
So why are epidemiological textbooks so dreary? This is not a facetious question. If epidemiologists believe that their subject is important and interesting—and why else would they devote so much of their waking lives to its pursuit?—their inability to communicate this belief is a serious failure. There are, I suggest, three reasons. One is a preoccupation with methods. Because epidemiological information is derived from observational studies rather than the controlled environment of the laboratory, we are neurotically aware that our data may have been degraded by a triple whammy of bias, confounding and chance. We feel compelled to pre-empt our critics with an explanation of how we deal with these problems, expatiating upon study design, choice of control groups and multivariate modelling. It’s a bad mistake. Nobody invited for a meal wants to feel obliged to admire her host’s batterie de cuisine before tucking in.
A second is that a lot of what claims to be neuroepidemiology is little more than counting cases, dividing the number by the size of the population from which the cases were drawn and calling the quotient an estimate of prevalence. This is nursery school stuff. The exercise rarely leads to a conclusion more thought provoking than: ‘there’s a lot of it about’—which was often the reason for doing the survey in the first place. The prevalence of a chronic disease depends not only on factors that influence the appearance of new cases but on how long cases survive. So an area may have a high prevalence of a disease simply because the complications of that disease are well treated there. Not surprisingly, there are few examples where the results of prevalence surveys have generated useful hypotheses about causation. To be fair to epidemiologists, who are well aware of the limitations (one could almost write futility) of these surveys, I ought to point out that they are more often initiated by clinicians in want of a research project than epidemiologists themselves.
The third reason is that many epidemiological accounts of disease leave the reader high and dry. What starts as a voyage of discovery ends not with a new horizon but stranded on mudbanks of inconsistent findings and publication bias. One case-control study reports a raised odds ratio for this exposure, two others for that. But the control group in one was drawn from hospital patients and from the community in the others. Bias may have been introduced because it isn’t clear that the cases were blind to the hypothesis being investigated. And perhaps there have been studies with negative findings that were never written up? Trying to be even-handed, the writer reviewing the subject ends up dancing a may-be might-be jitterbug. To some extent, this is unavoidable. After all, it is only in retrospect—only after the aetiology of the disease is understood—that one can confidently sort the red herrings from the promising leads. But reviewers can overcome this problem if they are brave enough to synthesize the evidence, however inconsistent and incomplete, to suggest lines of enquiry that might be fruitful and to create specific testable hypotheses.
Unfortunately, Principles of Neuroepidemiology makes all these errors. The first four chapters are devoted to statistical and epidemiological methods, to the design of clinical trials and intervention studies and to methods of measuring health status. There is nothing wrong with them but they do not contain much that is new. A large number of easily accessible texts cover this ground well enough already. The target readership is identified as clinicians with an interest in the epidemiology of neurological diseases but the authors’ shots are often wide of the mark. They aim too high when they deal with statistical principles; most clinicians will find the combination of brevity and algebra numbing. I would defy anyone not already familiar with Cox’s proportional hazards analysis to gain anything useful from the description on pages 23 and 24. And their aim is almost insultingly low when they advise that ‘scatterplots are useful for displaying relationships among quantitative variables’ or that ‘histograms are graphic representations of the distribution of the data’. These chapters would have been more worthwhile if they had focused on problems peculiar to neurology. Many diseases of the nervous system are rare, and approaches such as cohort studies, that epidemiologists find useful in the investigation of common conditions, are often inapplicable. There are solutions to the difficulties of investigating rare diseases in population based studies but they are not explored in any depth here.
Then follow chapters on specific diseases: Alzheimer’s disease, stroke, multiple sclerosis, epilepsy, Parkinson’s disease, headache, backpain, traumatic brain injury, motor neurone disorders, brain tumours, sleep disorders and neurogenetic disorders. Again, there is nothing much wrong with them but there is nothing that is exciting or inspiring either. I longed to find some new evidence unearthed or to read a radically different interpretation, however wacky.
But the main trouble with this book is its lack of ambition. Despite the confident, almost hubristic, title, it is concerned only with adult neurological diseases occurring in the US. Worse, it is almost entirely grounded in a traditional framework of risk factor epidemiology in which individuals are treated as black boxes and investigators infer causal relationships from what they have found out about the input and output of the box. Inputs consist of the sociodemographic characteristics of the study subjects and estimates of their exposure to putatively harmful environmental factors while outputs are measures of the occurrence of the disease under investigation. There is very little about the possibilities opened up by molecular and genetic epidemiology to investigate gene–environment interactions and throw light on the mechanisms of pathogenesis occurring within the black box.
The limitations of risk factor epidemiology have been widely debated recently. One telling criticism of the approach is that it fails to take adequate account of the social, demographic, economic, political and cultural forces that constrain and shape the way people live their lives. Choices made by individuals about what they eat, where they work, what they do in their leisure time, the age at which they start their families and how they look after their children are heavily influenced by the society in which they live. Many forces that affect health and disease operate not at an individual level but on groups. One example is herd immunity which, while a powerful determinant of a person’s risk of infectious disease, is not a property that can be adequately captured by making measurements on individuals. Similarly, the effects on health of large-scale forces such as urbanization, industrialization, population growth, changes in the age structure of the population, racial discrimination, poverty and inequality are likely to be profound even if inaccessible to investigators working within a risk factor paradigm. It is disappointing that the book contains no discussion of how the development of this new area of epidemiology might help us better understand the causes of diseases of the nervous system.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||