According to John Briffa this is a devastating critique of the Honda et al study that showed there was no relationship between MMR rates and autism in Japan. Basically it claims that even though MMR rates don't correlate with autism rates this is because single measles jabs also cause autism! The data used is not the local single jabs rates for the area studied by Honda et al but the national Japanese rates. I don't know why Briffa thinks this argument supports the MMR-autism link, it looks like a simple post hoc rationalisation to me.
But let's look at this data (plotted top right), vaccination rates (national MMR + single measles, and local MMR) shifted by one year (the Japanese measles vaccines are given at 1-year according to Nakatani et al and Honda et al) plotted against autism rates (cumulative up to seven years) by year of birth - overall autism spectrum and autism with regression are plotted*. Here we can see that there is minimal relationship between overall rates of autism spectrum and combined vaccination rates, and essentially no relationship with autism with regression.
This can be better seen in the figure on the right where the regressions of total measles vaccination rate against autism incidence are plotted - with neither even approaching statistical significance (p>.3), and correlation coefficients .3 and -.2 for autism spectrum and autism with regression respectively. Note that the correlation between MMR vaccination and overall autism spectrum is -.7, so MMR protects against autism!**
*For those interested in these things, autism rates are from Honda et al, and measles vaccine rates are estimated from the figure in Nakatani et al.
**No, obviously I don't think this really.
Here's a bit of fun, let's try another national vaccination rates vs. local autism rate graph. California is apparently a popular target, so how does autism rate (as determined by autism case load - dodgy, but like I say, just a bit of fun) compare with US national vaccination rate (1-year time shift, WHO data)?
Well here it is top right, not very convincing (although r=.7, p<.01). It nicely demonstrates that correlations are fairly common in time series (because all you need for a correlation is a trend for both things to increase with time) but also shows that the relationship is far from a 'dose-response' (sic) since autism keeps climbing while vaccination rates plateau. It accords fairly well with the figure on the left from a study of MMR vaccination rates and autism in California from JAMA which also found little relationship, with a plateau in vaccination rates and a continuing rise in autism.
This is also an interesting figure (below) from a study in the BMJ based on the UK general practice research database. It found that there was no relationship between UK MMR coverage (which remained very high for cohorts from 1988 to 1993) while autism rates continued to climb and nicely reflects the US vaccination plateau.