Reviews
| Date | 01 May 2000 |
| Published date | 01 May 2000 |
| DOI | http://doi.org/10.1111/1468-2230.t01-1-00275 |
REVIEWS
Richard Goldberg,Causation and Risk in the Law of Torts; Scientific Evidence
and Medicinal Product Liability, Oxford: Hart Publishing, 1999, xix + 260 pp, hb
£30.00.
As anyone familiar with the story of thalidomide knows, pharmaceutical products can
cause harm as well as prevent it. In most cases involving injuries alleged to have been
caused by drugs, however, causation is not as easily established as it was with
thalidomide. People who use medicines are usually ill to begin with, and the injuries
they sustain often have a number of possible causes. In the absence of large-scale
epidemiological studies (which involve comparing the rate of injury in a population
exposed to a drug with that in a non-exposed population), it can be difficult to prove
that it is more likely than not that a specific person’s injury was caused by a particular
product. Things are further complicated by the fact that the concept of causation is
subject to competing interpretations, and by the issues of policy involved in balancing
the benefits of pharmaceutical products to society as a whole against the risk they pose
to certain individuals. It is this complex area that Goldberg addresses in Causation and
Risk in the Law of Torts. Unfortunately, he fails to do it justice.
A considerable part of the book is devoted to case studies of various medicines that
have been alleged to cause injury and to exploring how claims against manufacturers
have been dealt with in different jurisdictions. Goldberg has obviously read widelyin
his subject, and he describes the results of a large number of cases as well as the views
of commentators. But while the case studies are thorough, they lack sustained analysis
or synthesis, and this makes it hard for the reader to draw much from them. One of the
difficulties is that Goldberg establishes no clear analytical framework or agreed
terminology. As this is an area where different writers use different terminology, some
of the issues that are introduced as though they are novel seem to me to be no more
than extensions of previously discussed problems. At pp 86–87, for example, the
discussion of ‘indeterminate plaintiffs’ fails to make it clear whether the
indeterminacy differs from the uncertainty present in earlier examples; and at p 144
the reference to general and specific causation appears to repeat a distinction made
earlier in terms of statistical and personal chances. Worse still, Goldberg obviously
misunderstands some of the concepts he discusses. The most glaring example is
statistical significance, which he continually refers to (pp 105, 108, 118, 120, 138–
139, 182, 194) as though it provides a measure of the strength of causal relationship
established by an epidemiological study, when it in fact refers to the probability that a
study would produce the result it did by chance. Given that one of his policy
recommendations is that courts should adopt the level of significance adopted by
epidemiologists, and that ‘[t]he greater the level of statistical significance, the higher
the compensation’ should be (p 246), this is a rather serious mistake. It is also baffling,
given that Goldberg references material (eg Sanders’ and Green’s work on Bendectin)
where the distinction between statistical significance and the strength of association is
clearly explained.
There are similar problems in the theoretical parts of the book. One of Goldberg’s
principal themes is that Bayes’ theorem (BT) should be used to ‘refine’
epidemiological statistics in specific cases, in order to ‘establish the most accurate
probability of a specific causal link between a drug and an injury’ and to ‘promote
more certainty in judicial decision-making’, leading to ‘greater efficiency within the
ßThe Modern Law Review Limited 2000 (MLR 63:3, May). Published by Blackwell Publishers,
108 Cowley Road, Oxford OX4 1JF and 350 Main Street, Malden, MA 02148, USA. 457
. .. civil justice system’ (p 250). Now I have difficulty working out just what Goldberg
means by all this, but it strikes me that the idea of using BT is very much oversold in
this context. One of the reasons for introducing BT is that the statistics produced by
epidemiological studies are said to relate to ‘statistical chances’, but such figures are
‘insufficient’ because they are ‘not personal to the plaintiff’; ‘the statistics do not
relate to a particular case’ (p 37). By using information about the plaintiff, however,
such as the details of her medical history, the timing of the adverse reaction, and her
response to stopping and restarting treatment with the suspect drug, it is claimed that
statistical chances can be converted into ‘personal chances’ (p 40). But the distinction
between the two types of chance is one without very much difference. Suppose the
patient improves when we stop giving her the drug. That fact is meaningless unless we
rely on general information – information that is not specific to the plaintiff – about
how people are likely to react when we stop administering harmful things to them. So
factoring in information personal to the plaintiff does not miraculously convert a
statistical chance into a personal one. All it does is give us more information on which
to base a decision about causation in the specific case.
Nevertheless, BT may still be a useful means of combining the various pieces of
information in a case. But will it make decision-making more accurate or efficient? I
rather doubt it. For one thing, the concept of ‘accurate . . . probabilit[ies]’ (p 40) is
arguably meaningless: all probabilities are in some sense inaccurate because they are
based on incomplete information. What is more, it looks to me as though the process
of developing point-valued probabilities for the case-specific information so that we
can feed them into BT will be more trouble than it is worth. We will be relying on
guesswork, and different doctors will give widely varying values to the probabilities.
The calculations will also be complicated by the fact that we will need to take into
account the dependency relations between the different probabilities. All we will gain
is false precision. But this remains obscured in Goldberg’s discussion, because he
gives no worked-through example of how BT might be used in an actual case. The one
example he gives (pp 53–54) is largely irrelevant. It is also botched because, so far as I
can make out, he puts the wrong probabilities into the equation (he forgets to condition
all of the probabilities on the fact that the patient in his example has taken the drug and
then suffered a headache). He even gets BT wrong (p 39). When he expresses it in
words, he seems to think that the prior probability should be divided by 1; when he
uses notation he decides to use the prior to multiply both numerator and denominator
of the equation (the prior should not be divided by anything and should only multiply
the numerator).
The second limb of Goldberg’s theoretical analysis builds on the first. He analyses
causation in terms of Landes and Posner’s version of the Hand formula (the LPF), and
suggests that, because the LPF contains probabilities, ‘[I]t now appears to be
necessary in complex cases, to invoke statistical constructs [sic] such as the [sic]
Bayes’ Theorem’ (p 199). But, given that all the LPF requires is a determination
whether or not there is an inequality between two probabilities, this is a bit like
arguing that, because most civil cases require a determination whether the plaintiff’s
case is more probable than not, we should regularly employ BT in the courts. Even
ignoring that argument, the problem of false precision remains. Just because we have
managed to get a point-valued probability out of BT, we should be under no illusion
that we will then be in a position to promote efficiency in Landes and Posner’s terms.
It all depends on the information we have in the first place, and in medical causation
cases I cannot see that we will often have good enough information to justify using the
formulae that so beguile Goldberg. Finally, Goldberg misunderstands the LPF. He is
aware that it has been criticised for replacing ex post analyses of causation with ex
The Modern Law Review [Vol. 63
458 ßThe Modern Law Review Limited 2000
To continue reading
Request your trial