I'd say: "it depends".
If you have reasonably small loop code (where you can see the whole loop-code without scrolling) its usually ok to use a continue.
However, if the loops body is large (for example due to a big switch), and there is some followup code (say below the switch), you may easily introduce bugs by adding a continue and thus skipping over that code sometimes. I have encountered this in the heart of a bytecode interpreter, where some instrumentation code was sometimes not executed due to a continue in some case-branches.
This might be a somewhat artificially constructed case, but I generally try to avoid continue and use an if (but not nesting too deep as in the Rob's sample code).