This is a cross-sectional study, i.e. a snapshot at one point in time. The study authors used the data from the Austrian Health Interview Survey, a self-reported set of data. There are major biases associated with self-reported data, from self-selection to recall bias.
The authors took the single point-in-time snapshot and then ignored confounders. For example, people with cancer, allergies, and (possibly) mental disorders are more likely to try different diets, including vegetarian, in an attempt to manage their health. Bottom line: a cross-sectional study gives you a look at the prevalence of disease in a selected population at a single point in time. It can't say anything about causation. But the authors suggest that a vegetarian diet is involved in the causation of the various diseases.
BAD SCIENCE!
PLoS One is an open-access online journal that accepts 70% of what is sent to it. It depends on readers to decide what's important; apparently scientific integrity isn't on its list of priorities. That's not to say that more strictly-reviewed journals are perfect, but this one has (reasonably) been called the Public Library of Sloppiness.
I'm going to print this one out for my next Journal Club as an example of a poorly executed study. Sorry, Bruce. But I know you threw this out there to see who'd bite.
