Proving a case on causation by reference to statistical probability

Clyde & Co recently successfully defended a clinical negligence claim in which the Judge examined the role of statistics when looking at the issue of causation in clinical negligence cases.

Introduction 

Clyde & Co was instructed by NHS Resolution to represent the Hampshire Hospitals NHS Foundation Trust in this claim brought for alleged injury from admitted negligent medical treatment afforded in 2016. The judgment can be found here.

The Claimant had a medical history of Polycystic Kidney Disease. She was seen by a midwife for a booking appointment on 10 March 2016 when she was about 8 weeks pregnant and was referred for a consultant review. This review did not unfortunately take place until 16 June 2016 by which time she was 23 weeks pregnant; because of her medical history the Claimant was at high risk of pre-eclampsia and was prescribed aspirin at this stage. 

It was admitted by the Defendant that, in accordance with the relevant NICE guidelines, there was a failure to arrange an earlier prescription of Aspirin at a daily dose of 75mg. 

On 1 September 2016, the Claimant was advised to stop taking Aspirin in the light of her continuing complaints of heartburn. On 15 September 2016 she was diagnosed as suffering from HELLP syndrome and on 17 September 2016 the Claimant gave birth to her first child via caesarean section. 

It was the Claimant’s case that had she been prescribed Aspirin from 12 (or alternatively 14 or 16) weeks, she would she have avoided developing HELLP – a severe variant of pre-eclampsia – and various associated consequences which included PTSD and fibromyalgia. 

Damages were agreed, subject to causation being proved, at £430,000. 

The law

His Honour Judge Glen noted that causation was generally a binary issue, and that the Claimant must prove to the usual civil standard that, but for the alleged or admitted breaches of duty, the various conditions and symptoms that she suffers from would not have arisen. He noted that causation was complex in this case as the Claimant sought to prove her case by reference to statistical probability. 

The Judge noted and relied upon the following paragraphs of the Judgment in Gregg v. Scott [2005] AC 176 when considering causation:- 

"27. In cases of medical negligence assessment of a patient's loss may be hampered, to greater or lesser extent, by one crucial fact being unknown and unknowable: how the particular patient would have responded to proper treatment at the right time. The patient's previous or subsequent history may assist. No doubt other indications may be available. But at times, perhaps often, statistical evidence will be the main evidential aid. 

28. Statistical evidence, however, is not strictly a guide to what would have happened in one particular case. Statistics record retrospectively what happened to other patients in more or less comparable situations. They reveal trends of outcome. They are general in nature. The different way other patients responded in a similar position says nothing about how the claimant would have responded. Statistics do not show whether the claimant patient would have conformed to the trend or been an exception from it. They are an imperfect means of assessing outcomes even of groups of patients undergoing treatment, let alone a means of providing an accurate assessment of the position of one individual patient."
…
32.  The value of the statistics will of course depend upon their quality: the methodology used in their compilation, how up to date they are, the number of patients involved in the statistics, the closeness of their position to that of the claimant, the clarity of the trend revealed by the figures, and so on. But to reject all statistical evidence out of hand would not be acceptable. This argument, if accepted, would effectually nullify the use of statistics in all cases of delayed treatment save perhaps where the figures approached 0% or 100%. Despite its imperfection, in practice statistical evidence of a diminution in perceived prospects will often be the nearest one can get to evidence of diminution of actual prospects in a particular case. When there is nothing better courts should be able to use these figures and give them such weight as is appropriate in the circumstances. This conclusion is the more compelling when it is recalled that the reason why the actual outcome for the claimant patient if treated promptly is not known is that the defendant by his negligence prevented that outcome becoming known."

The Judge felt that his task was not to review the epidemiological papers presented in this case and choose between them. Rather, he felt that his task was to examine the evidence of the experts called before him and test that evidence against the medical literature when assessing the validity of their conclusions.

The evidence

This was an extremely complex case, and one cannot do justice to the nature and depth of the issues which were considered. Focusing however on the main literature relied upon by the experts (Professor Siassakos for the Claimant and Mr Tuffnell for the Defendant) in relation to the effect of aspirin generally:

• The Claimant relied upon the paper Roberge et al 2017. This was a systematic review using aggregate data to investigate the effect of different dosages of Aspirin before and after 16 weeks. The overall sample size was 20,909 women but specific reliance was placed upon a subgroup analysis looking at the effect of 75mg Aspirin at ≤ 16 weeks on severe pre-eclampsia. This subgroup analysis constituted two papers totalling 373 women. The relative risk was 0.24. i.e. a reduction of risk of 76%; and
• The Defendant relied upon the paper Cochrane (Duley et al) 2019. This was a systematic review of 40,249 women using Individual Patient Data where available (for 34,514 women). The overall relative risk was found to be 0.82 – a reduction of risk of 18%. The relative risk for high risk women was 0.78 – a reduction of risk of 22% and the relative risk for doses of 75mg or more was 0.78 – a reduction of risk of 22%.

Findings

It was found that there is an obvious danger (identified in the Roberge paper itself) in taking two relatively small trials and over-extrapolating outcomes from the data. This was all the more so when there were question marks over the reliability and relevancy of the data which had been identified by the Defendant. 

Furthermore, it was found that there appeared to be considerable force in the Defendant’s suggestion that studies based on aggregate data observed greater effects than those based on individual patient data. There was also a broad agreement that aggregate data is at risk of publication bias and that individual patient data is more precise and allows for more sophisticated analysis. 

The Judge considered that considerable weight must instead attach to the outcome of a large, “gold standard’” review by Cochrane whilst recognising the impact that other factors may have. 

These findings were made within the context of the Judge’s findings on the expert evidence. He found that Professor Siassakos lacked “flexibility of mind” and was prone to making sweeping statements. The Judge was particularly struck by Professor Siassakos suggestion, made more than once, that to take a different view to his own was verging on being clinically improper. 

By contrast, Mr Tuffnell was found to be “more measured, more willing to make concessions and recognise contrary arguments whilst at the same time providing a coherent basis for his opinion.” 

Overall, taking all the current evidence together, the Judge found that it was simply not possible to find that on a balance of probabilities this Claimant would have avoided developing HELLP had she been prescribed 75mg Aspirin at 12 (or 14) weeks instead of at 23.

Conclusion

The Court’s task in these circumstances is to weigh the expert evidence and to give the statistics such weight as deemed appropriate, noting what is said in Gregg v Scott. i.e. they are an imperfect means of assessing outcomes even of groups of patients undergoing treatment, let alone a means of providing an accurate assessment of the position of one individual patient.