Think back into the mists of MMR: in 2002, John O'Leary's group in Dublin reported finding measles virus in the intestine of children with autism and bowel problems. The anti-MMR movement were almost delirious with excitement, and so were the media. Andrew Wakefield, working with Kawashima et al in Japan, had already reported finding measles virus in blood cells in similar children.
What if they were mistaken? How would you know? Well, a major paper published in the leading academic journal Pediatrics this month strongly suggests that these earlier results were in error, false positives. This study has been ignored by the media: it has been covered, by my reckoning, in one Reuters piece, and in one post on the lead researcher's boyfriend's blog.
This new MMR study, by D'Souza et al, replicates the earlier experiments pretty closely - and in some respects more carefully - in 54 children with autism (80% also had gastrointestinal symptoms), and 34 controls. All but six had received the MMR vaccine.
All these studies, old and new, used PCR, the same process used in genetic fingerprinting. PCR works by using enzymes to replicate RNA, so you start with a small amount in your sample, but then it is "amplified up" - copied over and over again - until you have enough that you can measure and work with it.
Beginning with a single molecule of genetic material, PCR can generate 100bn similar molecules in an afternoon. Because of this, the PCR process is exquisitely sensitive to contamination so you have to be very careful and clean up as you go. One substance used to prevent this contamination is called "UNG". The new study used 50 times more UNG than the original O'Leary paper, to prevent contamination. The researchers were also careful to use the very same primer sequences for the measles virus genes as their predecessors.
The results were striking. Firstly, using the Kawashima primer pairs, they simply got negative results, where Kawashima et al reported positive results. The replication of the O'Leary work was more interesting. Looking only at the results of the PCR, at first it looked like the O'Leary primers did indeed produce RNA strands that matched measles virus. However, when they looked more carefully at the size of the strands, and the "melting curves", and then sequenced the genetic material, they discovered that what had been amplified was not from measles virus at all: they were false positives. The original O'Leary paper did not pursue these extra "double checking" steps.
The authors are quite clear: there is good reason to suspect that the earlier studies produced false positive results, because of suboptimal contamination control, and because the O'Leary primers can accidentally amplify bits of normal human RNA. This is not about criticising individual researchers. Techniques move on, results are sometimes not replicated, not all double-checking is practical.
· Please send your comments to bad.science@theguardian.com