New evidence suggests convicted cot death killer Angela Cannings may be the victim of a miscarriage of justice, a BBC Real Story investigation has learnt.
Angela Cannings has always maintained her innocence
When Cannings was convicted in April 2002, the jury had been told that the deaths of three of her children could not have been caused by a genetic defect because there was no evidence of other infant deaths in her close relatives.
But the BBC has learnt there were deaths among young children in Cannings' family.
Cannings, 40, from Salisbury, Wiltshire, was sentenced to life in April 2002 for the murder of seven-week-old Jason in June 1991 and 18-week-old Matthew in November 1999.
The court was told her first child, Gemma, died in 1989 at the age of 13 weeks, although Cannings did not face charges in relation to her death.
The Real Story team discovered Cannings' paternal great-grandmother suffered one infant death and Angela's paternal grandmother two.
Professor Michael Patton, a clinical geneticist at St George's Hospital Medical School in London, said this was compelling new evidence.
Until the Real Story investigation, the police, the judge and the jury all thought Cannings' grandmother had seven children.
But Irish birth records prove she had nine - with two dying in early infancy.
Prof Patton said: "This is an exceptional family. The deaths are more likely to be explained by some form of genetic inheritance where some members of the family are affected and it is passed through other members of the family who don't express the gene.
"We would have seen diseases like measles and whooping cough a couple of generations back. There is no suggestion that these children, from what we know, died from infectious diseases."
The issue of how investigations and prosecutions of unexplained deaths of infants are conducted came under the spotlight with the acquittal on appeal of solicitor Sally Clark.
In that case Professor Roy Meadow told the original trial that two cot deaths in one family were a "one in 73 million chance" - something disputed by statisticians.
After upholding Clark's appeal, the Court of Appeal judges said the medical evidence of a "one in 73 million chance" had been grossly misleading.
Prof Meadow was also a witness in the Trupti Patel case, which saw the pharmacist found not guilty of killing her three children.
In Cannings' case Professor Meadow told the jury her babies could not have died a normal cot death because they appeared healthy immediately before they died.
But Professor Ronald Harper at the University of California said this was perfectly consistent with cot death.
He told Real Story: "Cot death can happen very quickly. We have recordings in which the process appears to occur over one or two minutes.
"We have encountered cases where a child was feeding and the mother had looked down and the child had succumbed.
"There are cases in which the infant is perfectly awake and happy a few minutes before death and then suddenly had died."
The investigation also found that Cannings' half brother, Joshua Connolly, now aged 21 months, suffered three sudden attacks in his first year - all thought to have been caused by an allergic reaction to cow's milk.
Real Story with Fiona Bruce: BBC One, Monday 3 November at 1930 GMT. The programme is also streamed live on the Real Story website.