Skip to content

April 2026: Thirlwall Inquiry final report due after Easter · CCRC still reviewing 31+ independent expert reports · Shoo Lee Panel (Feb 2025): no medical evidence of deliberate harm.

Lucy Letby Facts

Long-form · Legal history

The Sally Clark parallel

Sally Clark was wrongly convicted in 1999 of the murder of her two infant sons. She was acquitted in 2003 after the statistical and pathology evidence that underpinned her conviction was found to be unsafe. The case drove the Royal Statistical Society’s published framework on statistical evidence in criminal trials. This page asks a simple question: was that framework applied in the Letby case?

Last updated
7 min read

What happened to Sally Clark

In November 1999 Sally Clark, a solicitor from Cheshire, was convicted of the murder of her two infant sons. The first had died at 11 weeks; the second at 8 weeks. The prosecution called the eminent paediatrician Sir Roy Meadow, who told the jury that the probability of two cot deaths in the same middle-class family was “1 in 73 million”. He arrived at this number by squaring the probability of one cot death. That calculation was wrong. Squaring a single-event probability treats the two events as independent, and in a real family with shared genetics and shared environment they are not. The Royal Statistical Society publicly corrected the error in October 2001 — two years after Clark was jailed.

Clark’s conviction also rested on pathology evidence from the Home Office forensic pathologist Dr Alan Williams. It later emerged that Dr Williams had failed to disclose microbiology results suggesting that the second child had died of a Staphylococcus aureus infection. When that non-disclosure came to light Clark’s second appeal succeeded. She was acquitted in January 2003. She died in March 2007 of acute alcohol intoxication at the age of 42.

What the Royal Statistical Society said afterwards

The RSS’s response to the Clark case — and to the linked cases of Angela Cannings and Donna Anthony — became the canonical UK reference on statistical evidence in criminal trials. Its published guidance crystallised several principles that are directly relevant here:

  • Selection effects. If a number of events are selected for examination because of a feature that is also used as evidence, that selection must be disclosed and accounted for. Otherwise the “improbability” of the pattern is an artefact of the selection, not a fact about the world.
  • Independence assumptions. Multiplying probabilities only works when the events are independent. Biological events in the same family or the same unit are typically not independent.
  • Base rates. The prosecutor’s fallacy — confusing the probability of the evidence given innocence with the probability of innocence given the evidence — must be avoided. Specifically: courts should not hear bare numerical improbabilities without the denominator.
  • Expert humility. Statistical reasoning should be presented by a statistician, not reverse-engineered by a medical witness.

How the Letby chart looks against that framework

The shift-rota chart shown to the Letby jury fails at least two of those principles on its face:

  1. Selection effect. The 25 “suspicious” events were selected in part because Letby was on shift. Events where she was not on shift were not included. Plotting her shift attendance against a set of events defined by her presence generates a full row whether or not she is doing anything wrong. Prof. Richard Gill (Leiden) has publicly described this as a textbook Texas sharpshooter fallacy. See our statistics deep-dive.
  2. Independence. The collapses occurred on a unit which had, in the same period, a documented superbug outbreak, sewage and plumbing failures, and chronic doctor shortages (see the Guardian investigation summary). Collapses in that environment are not independent trials on a fair coin.

The Royal Statistical Society’s post-Clark guidance was explicitly intended to stop arguments of this shape being put to juries without a qualified statistician explaining their limitations. In the Letby trial, no qualified statistician gave evidence. The chart was presented by counsel and by a causation expert who was not a statistician.

How the medical evidence looks against that framework

The Sally Clark pathology parallel is equally direct. In Clark, undisclosed microbiology results pointing to a natural-causes explanation were not put to the jury. In Letby, the equivalent structural issue is that the Crown’s causation expert Dr Dewi Evans applied a methodology that Dr Shoo Lee and fourteen international colleagues have now publicly said is not a methodology they recognise. The alternative natural-causes explanations for each collapse — sepsis, NEC, resuscitation trauma, thrombosis, gestation-related instability — were not systematically examined in the way the Panel’s additional-10 report and the full Panel review have now done.

Why this is the case, and not the answer

The Sally Clark parallel is a legal-history argument, not a medical-science one. It does not itself prove anything about the Letby evidence. What it does show is that the English criminal justice system has been here before, has told itself what went wrong, and has published guidance on how to avoid repeating the mistake. The specific question the CCRC now has to answer is whether that guidance was, or was not, adequately applied in 2022 and 2023.

Rob Rinder KC, Sir David Davis MP, and the Bar Council letter signatories (all linked on this site) have all made explicit reference to the Clark precedent. The Panel has cited it. Prof. Peter Green — a past president of the Royal Statistical Society — has written publicly about the parallel. The CCRC application filed by Mark McDonald KC in October 2025 rests on it as one of several structural grounds.

Read alongside