CL-2022-000676 - [2025] EWHC 2486 (Comm)

From the DHSC’s perspective the answer here was straightforward. 103 out of 140 gowns tested failed the sterility tests. It says that the testing done was sufficient to establish on a balance of probabilities that the gowns did not meet the required SAL of 10^-6. The specified SAL of 10^-6 permits no more than 25 out of 25 million gowns to fail a sterility test. Professor Hutton points out that the minimum failure rates demonstrated by the testing are already well above the failure rate tolerated by the required SAL. For example, of the 3.75m small gowns, the SAL of 10^-6 would require that no more than 4 fail, yet the observed number of failed gowns is hugely in excess of that figure. The number of failures is said to be so high that it does away with any challenge by Medpro on the selection of the samples in that even if DHSC had selected only those gowns that were not sterile in the entire population and sent those to Swann Morton for testing, the required SAL of 10^-6 has not been met.

This gave rise to a very interesting debate about the nature of a sample for statistical purposes and whether any inference could be drawn from a non-representative and non-random sample. This was an area in which it emerged that expert input might well have been useful in the formulation of the questions for the experts, as well as the answers, with DHSC’s eminent (and on her own subject extremely impressive) expert witness Professor Jane Hutton, stating that the question should have been “given the results that we have got, what are the possible explanations of evaluative opinion?” – or to posit that in the terms of this case: what statistically valid conclusions can be drawn from the results of the Swann-Morton testing?

In fact the evidence on sampling seemed only to underscore the reasons why this was not an appropriate way of testing for sterility. Unless one were prepared to descend into the byways of statistical evidence there would be no way of robustly testing unless one could establish a “representative” sample; and in a population of gowns of this nature unless one had some way of knowing what strata were likely to be of interest—and there are many possible candidates—no stratified sampling method could easily be devised. In addition, the evidence as to the nature of the sterility definition meant that extrapolating from actual samples was conceptually entirely at odds with the nature of the exercise.

As to what one might nonetheless gather from the evidence and the statistics, the statistical arguments cannot determine the matter. While, regardless of the debate on the optimum method of sampling (on which it seemed clear that the method used was not remotely the optimum method: in that the Swann Morton testing which formed the original basis for the case involved only 2 out of about 540 containers over 14 storage sites, and related to the work of only one of the sterilisation facilities) I might have been inclined to see force in the submission that as a matter of probability 55 non-sterile gowns in a population of 25 million means that there was probably a breach of the sterility requirement, that “analysis” proceeds on the basis that the samples as tested were representative of the samples as delivered. That creates a real problem given that there was no evidence of this – a topic to which I return further below.