The problem with nearly all 'font readability studies' is that they use a completely arbitrary set of samples and an arbitrary sample size. An example.
As such, what they are typically measuring is "within the confines of these 8 typefaces, typeset in these very specific ways, and based on the results of these 48 people"...which is fine, but ultimately is only testing that specifically.
And then we proclaim universal "truths" like "italics is bad for dyslexia". It may be, but we shouldn't assume that's absolutely true based on simple studies like this.
There are general rules of typography to ensure something is readable, and those should always be followed. But "never use italics" isn't one of them.
Ultimately if someone is dyslexic and they do have a preference for a very specific font, that is great, and thanks to the web, they can set that up on their local machine and use it, regardless of what your web site is using. And that's the ideal solution.