Learning from the Sally Clark case

In 1999 Sally Clark was convicted in the crown court at Chester of the murder of her infant sons, Christopher and Harry. In the absence of crucial evidence that was later forthcoming, Clark’s first appeal to the Court of Appeal was dismissed on 2 October 2000, based in part upon judicial endorsement of a prosecution witness’s opinion that the chances of two successive natural infant deaths is one in 73 million.

On 2 July 2002, the Criminal Cases Review Commission referred Clark’s case back to the Court of Appeal as unsafe due to newly introduced evidence of natural causes of death. The court confirmed the convictions to be unsafe and set them aside. Mrs Clark was released, but sadly never recovered from her ordeal of three years’ imprisonment, and died in 2006.

The new evidence comprised microbiological tests showing evidence of lethal infections in both infants.

Here, we consider the clinical, epidemiological, and statistical evidence presented during the prosecutions, and the subsequent impact on UK jurisprudence of one of the most serious miscarriages of justice in the English courts.

Circumstantial similarities

At eleven weeks of age, Clark’s first child, Christopher, died suddenly on 13 December 1996. Home Office pathologist Dr Williams concluded that the cause of death was a lower respiratory tract infection, noting minor resuscitation trauma. The case was treated as a case of sudden infant death syndrome (SIDS or ‘cot death’).

On 26 January 1998, Harry, Clark’s second child, died suddenly aged seven and a half weeks. Dr Williams concluded the resuscitation injuries to be consistent with vigorous shaking as the likely cause of death. Williams’s findings led him to revise his opinion of Christopher’s cause of death as suggestive of smothering.

Clark and her husband were arrested on suspicion of murdering both infants.

Absent compelling evidence, and in spite of defence testimony that Clark’s infants were well cared for and loved by their parents, the prosecution elicited weak circumstantial similarities in the detailed history of the death of each child, suggesting that they had died from deliberately induced injury.

Three other expert witnesses testified for the prosecution: Professor Sir Roy Meadow, pediatrician (St James’s University, Leeds); Dr Keeling, paediatric pathologist; and Professor Michael Green, forensic pathology (University of Sheffield).

Expert evidence for the defence was given by: Professor Berry, paediatric pathologist, a recognised expert in SIDS; Dr Rushton, paediatric and peri-natal pathologist; Dr Whitwell, forensic pathologist; Professor David, paediatrician; and Professor Luthert, ophthalmologist.

There were disparate conclusions by doctors at the trial. Dr Williams argued in the case of Christopher’s death there was a ‘possibility that the child was smothered, a possibility in a broad range’. Professor Meadow rejected lethal respiratory infection and SIDS, concluding that Christopher’s death was not natural.

Professor Green, and Drs Keeling, Berry, and Rushton, all considered the cause of Christopher’s death to be ‘unascertained’. Professor David considered that possible causes of death were idiopathic pulmonary haemosiderosis and suffocation.

These expert witnesses disagreed on the post mortem findings in the eyes, lungs, brain, and spine in respect of each death.

Obvious factors which were of relevance to the appeal included:

• The changed opinion of Dr Williams that the respiratory findings could not have led to Christopher’s death;

• The crown’s acceptance of Professor Luthert’s attribution of the intra-retinal haemorrhaging of the eyes to the result of an error in slide preparation;

• Disparate expert opinions regarding the assignment of the deaths as unnatural; and

• The validity of Dr Meadows’s opinion of the statistical unlikelihood of two successive infant deaths by natural causes.

It is helpful to consider two overlapping epistemological frameworks which may be employed by epidemiologists and pharmacologists to evaluate causation of biologically induced injury, namely the Bradford Hill criteria for causation and probability analyses.

The Bradford Hill framework for causation was developed in 1965 by the eminent statistician and epidemiologist Sir Austin Bradford Hill. He described the criteria for persuasive linkage of a given cause (in this case SIDS or murder) and an observed outcome (infant death):

• Contemporality between the outcome and the causative explanation;

• Effect of re-challenge and de-challenge of possible cause(s) on the outcome;

• Plausibility of the putative mechanism of causation; and

• Presence of similar instances leading to similar outcomes.

Although information on each Hill criterion may not be fully available, the approach is commonly used to evaluate drug-induced adverse events and environmental hazards. It is also employed in the assessment of causal linkages with injuries in the civil courts. Statistical analysis of available data can sometimes be employed.

In our view there is a strong case for its judicious use in the criminal courts. However, as Robyn Lucas and Anthony McMichael cautioned in the Bulletin of the World Health Organisation in October 2005, ‘any resulting causal explanations must be viewed as an aid to judgement, not as arbiters of reality’.

Probability analyses

This approach can be integrated into the Hill framework to estimate the likelihoods of competing causal explanations for an observed outcome. Applied here, the approach relies upon explicit or implicit estimates of the probabilities (called ‘prior’ probabilities) of each possible cause (SIDS, murder) accounting for the outcome (sudden death).

When more than one sudden death occurs in a single family, it matters how the ‘prior’ probabilities of SIDS and murder are taken into account. The Sally Clark case is illustrative of incorrect and correct incorporation of probability analyses in assessing the causes of the deaths of Clark’s sons:

• Incorrect (two errors): In changing his first opinion, Meadows retrospectively assumed the ‘prior’ probability of SIDS in Christopher to be 1 in 8,500 and calculated the probability of SIDS accounting for the deaths of both of Clark’s sons (1/8,500 x 1/8,500 = 1 in 73 million) based on the incorrect assumption that the occurrence of SIDS is a random, independent event. Hearing expert testimony of such an extreme improbability of such occurrences in one family may have misled jurors to conclude a greater likelihood of double murders.

• Correct: When the occurrence of a second SIDS death in one family is correctly assumed to be non-randomly higher, based upon knowledge of correlating factors (hereditary traits, epidemiological clustering in families), the probability of two SIDS in one family can be calculated according to Bayes’ rule. This involves incorporating the probability of the first SIDS occurrence (1/1,300) with the higher probability of the second SIDS (1/100) along with the probability of single (1/650,000) and still lower frequencies of double infant murders (1/1.65 million) in one family.

Using this approach, the likelihood of the deaths being due to SIDS is 6 in10.

The court reviewed the statistical testimony of Professor Roy Meadow and concluded that this figure had unduly influenced the jury. Following the trial, the Royal Statistical Society (RSS) issued a public criticism of the expert evidence given by Professor Meadow and considered his approach was statistically invalid.

The criticism noted that no empirical justification for the assumption of randomness was provided in the case, but there were a priori reasons for the assumption of randomness of SIDS to be false, including unknown genetic or environmental factors that predispose families to SIDS, so that a second case within the family becomes much more likely. The statement from the RSS read: ‘The Court of Appeal has recognised these dangers (R v Deen (1993), R v Doheny and Adams (1996)) in connection with probabilities used for DNA profile evidence, and has put in place clear guidelines for the presentation of such evidence. The dangers extend more widely, and there is a real possibility that without proper guidance, flawed frequency estimates and incorrect statistical analyses presented by an “expert” in court mislead the jury in ways that are very prejudicial to defendants.

‘Society does not tolerate doctors making serious clinical errors because it is widely understood that such errors could mean the difference between life and death. The case of R v Sally Clark is one example of a medical expert witness making a serious statistical error, one which may have had a profound effect on the outcome of the case.

‘Although many scientists have some familiarity with statistical methods, statistics remains a specialised area. The Society urges the courts to ensure that statistical evidence is presented only by appropriately qualified statistical experts, as would be the case for any other form of expert evidence.’

The third key issue leading to the referral to the Court of Appeal was the failure by the prosecution to disclose the microbiological results which had been known to Dr Williams.

Throughout the trial there had been significant and ongoing problems in the investigation of the deaths. Standard protocols were not followed and essential steps such as routine dissection and histology were omitted, which prevented verification of alleged autopsy findings. Also, a number of potentially important diagnoses and conclusions were altered over time.

The court concluded that the very detailed review of the evidential and interpretive differences between the prosecution and defence witnesses, particularly as regards the statistical error and non-disclosure of the microbiological reports, meant both Christopher and Harry’s deaths were from natural causes. The verdicts were unsafe and were quashed.

Fundamental flaws

Sally Clark’s case raises some very disturbing issues regarding the admissibility of evidence and the evaluation of causation. It was also worrying that the judges at the original trial and the first referral to the Court of Appeal failed to understand the fundamental technical flaws in the statistical evidence by Sir Roy Meadow. These may be examples of relatively common problems at trial with highly specialised technical medical issues such as the diagnosis of SIDS. We suggest that lay judges in criminal trials should sit with expert technical assessors who would be qualified to opine on the weight, credibility, and technical accuracy of such evidence.

The Law Commission has published a comprehensive report on this issue (Law Com 325). The commission recommended a new statutory admissibility test, which would provide that expert opinion evidence is admissible in criminal proceedings when it satisfies a ’reliability test’. This would provide that if there is any doubt on the matter, expert evidence presented as evidence of fact should be treated as expert opinion evidence.

Trial judges should be provided with a single list of generic factors to help them apply the reliability test and these factors should be set out in the primary legislation containing the test. The trial judge should take into consideration the factors which are relevant to the expert opinion evidence, and any other factors they consider to be relevant.

Also, criminal courts should have a limited
power to disapply the reliability test so that it does not have to be applied routinely and unnecessarily; equally, the power to disapply must not be such that the reliability test becomes only a nominal barrier to the adduction of unreliable expert opinion evidence.

As regards the impact of expert evidence on
jury considerations, the commission proposed
that the reliability hearing should ordinarily take place before the jury is sworn but, exceptionally,
it should be possible to hold a hearing in the absence of the jury.

Unfortunately, these proposals have not
yet been accepted or put on to the legislative agenda. SJ

Peter Feldschreiber, pictured, is a barrister at 4 New Square and Carl Peck is the founder of pharmaceutical advisers NDA Partners