In nineteenth-century Britain, advertisements for patent medicines often carried the testimonials and encomia of duchesses, though there was absolutely no reason to suppose that duchesses were better informed or more intelligent on matters of health than wheelwrights or dustmen. Endorsement of products by sportsmen or film stars is similarly ridiculous. Even those who act upon such endorsements probably know that they are bought and paid for, which suggests that rational considerations play only a minor role in much of humanity’s decision-making.
Something similar seems to be at work when the president of the United States advises pregnant women not to take acetaminophen (also known as paracetamol) during pregnancy because to do so might result in childhood autism.
Doctors are universally of the opinion that pregnant women should take as little medication as possible, especially in the first three months of their pregnancy, and take only that which is really necessary. But this general precautionary principle is a far cry from condemning any individual medication as being dangerous or developmentally harmful to a fetus.
It is probable, of course, that no politician has the faintest idea of what would count as valid evidence to justify such advice. And they are not alone. Few people are aware of the complexity of proving a statement such as that, “Acetaminophen taken in pregnancy causes autism.” A statistical association, even if present, is far from sufficient to do so, because association is not causation.
Unfortunately, it is part of human psychology, even of the sophisticated and the educated, to take association as causation, and to think that such association provides guidance about prevention or treatment of a disease. Researchers who publish findings of statistical associations in medical journals almost always speculate on the reasons for these associations, as if they were causative in nature.
In 1965, the British statistician, Austin Bradford Hill, laid down principles or guidelines for the assessment of whether an association was likely to be causative. These principles or guidelines are not invariable rules but are nevertheless useful and should be kept in mind when statistical evidence of causation is either to be assessed or asserted. Among the factors Hill identified are Strength of Association, Consistency of Findings, Temporal Relationship, a Dosage Relationship, Specificity of Findings, and Biological Plausibility.
Is the association a strong one? The stronger it is, the more likely it is to be causative. A very weak association is most likely to be a statistical artifact. For example, in one of the best studies of the acetaminophen question, a Swedish examination of 2.5 million children born in Sweden between 1995 and 2019, found that the rate of autism was 1.33 percent in those whose mothers had not taken acetaminophen during pregnancy, and 1.42 percent in those who had. This is a very small difference, well within a margin of error from several different angles, and no one should conclude anything from it.
Here it should also be borne in mind that in any such study it is vanishingly unlikely that the figures for the two groups should be exactly the same, and furthermore that if you examined the consumption of, say, a hundred comestibles during pregnancy, you would find that some of them were statistically associated with the development of autism, even though none of them was causatively related in any way. If you look at enough factors, you are bound to find some that are statistically, but not causatively, associated.
Are the findings consistent? The more consistent such studies are, the more likely the association is to be causative. For example, no study fails to show an association between smoking and lung cancer. This is far from the case with acetaminophen and the development of autism.
To suggest without firm evidence that mothers may have helped to cause, however inadvertently, the autism of their infants by having taken acetaminophen during their pregnancy is to add casually to their suffering.
Is there a temporal relationship? Is there a connection between the consumption of acetaminophen by pregnant mothers and the rising prevalence of autism, as might be expected if one caused the other? The answer is “No.” Here, it might also be pointed out that the true nature of the rise in prevalence is not clear. Is it fashion in diagnosis, an increase in recognition of the condition, or a genuine rise in a real phenomenon? Changes in terminology do not necessarily reflect changes in reality.
Is there a relationship between the dose and the response? In this case, is there a relationship between the amount of acetaminophen taken and the likelihood of the development of autism? If there were, it would strengthen the case for a causative relationship, though it must also be borne in mind that the latter might be idiosyncratic, so that any exposure at all to the drug might cause autism. At any rate, no such dose-response relationship has been found.
How likely is it that the outcome is the result of the specific cause being studied? Are the pregnant mothers who take acetaminophen the same in all relevant respects as those who do not? The answer clearly is “No.” Either they take the drug because they are suffering from pain or fever, or because they are neurotically anxious, although, of course, some pregnant mothers may refuse to take acetaminophen for pain or fever (in which case, they are also different from those who do). It is very difficult to control for these factors, but the smaller the association between the drug and the condition, the more important those factors are likely to be. Both fever and diabetes in pregnant mothers, by the way, have been associated with the development of autism in infants—but, once again, association is not causation.
Is it biologically plausible? Is it scientifically possible that acetaminophen should cause autism? The mechanism of action of the drug is not fully understood, nor is the cause of autism fully understood, except that it is likely to be complex. At the very least, it is unlikely that the drug should contribute greatly to that causation; and sweeping pronouncements, such as have been made, that the cause, the magic bullet, will soon be found, raise hopes without any real prospect of fulfilment. In that sense, they are cruel.
Cruel also are warnings against drugs when these scientific guidelines do not indicate a clear danger. It is natural for parents of infants with abnormalities, especially psychological ones, to agonise over what they did, to examine whatever mistakes they made, that resulted in the abnormalities of their children. In the 1960s and ’70s, for example, some psychiatrists suggested that the manner in which families communicated could induce schizophrenia in adolescents and young adults, a view which was popularised in magazines, books, and films, though in fact there was no basis to it. Parents suffered unwarranted guilt about, and even blamed, for the sufferings of their children, which added to their own considerable sufferings. To suggest without firm evidence that mothers may have helped to cause, however inadvertently, the autism of their infants by having taken acetaminophen during their pregnancy is to add casually to their suffering.
There is an interesting sociological question to be asked about the president’s pronouncement: why should anyone take or have taken any notice of it? It is extremely improbable that he studied the question in any depth or even appreciated the methods by which the question could be answered. Yet many people are more likely to take notice of him than of real authorities. Other equally uninformed celebrities, in the wake of this pronouncement, are likely to be reckless in distributing the same advice.
It is, perhaps, wasted breath to protest against people’s propensity to invest the wrong people—presidents, duchesses, or film stars—with authority to pronounce on matters of health, because it seems ineradicable. In these circumstances, however, those with what might be called charismatic authority, rather than with the authority of true expertise, have an inescapable duty to remain silent on subjects that they have not studied but on which their advice might be heeded by many people if given.
No doubt some will say that experts can be mistaken, which is certainly true. Furthermore, the mistakes of experts can be catastrophic. Fallibility is built into our nature, and no one can escape it. Nor are experts always in agreement on every subject.
But this is no reason to doubt the value of expertise, or to resort uncritically to the opinions of those who have not considered a subject for more than a minute. We are utterly dependent on expertise and have good reason not to mistrust it most of the time. Whenever we take an aeroplane, whenever we drive a car, whenever we turn on a light, whenever we drink a glass of tap water, whenever we take a pill, and on myriad other occasions throughout the day, we rely on the expertise, past and present, scientific and practical, of others. The attempt at blanket scepticism of expertise is not only foolish, it is also both impossible and dishonest.
Expert opinion suggests that it can sometimes be medically advisable for pregnant women to take acetaminophen. Celebrities with no expertise in the subject have an inescapable duty to remain silent.