Misclassification bias. I suspect they began with somewhat equal sample sizes then participants went on meat, so they dumped them in the meat group not to underpower the study. You see trials (if this is a trial, haven't read the article) need to pre-register with a pre-determined consent process. Common sin is dumping treatment groups (vegetarian) into controls (meat and vegie). Technically meat people don't have a meat only diet. So there's a confounder from onset. Vegie only diet peeps tend to have pre-existing disorders like cancer and other nutrition issues if they don't take adequate supplements. This is perhaps more a study on supplement behaviour among vegetarians. But I guess that's not hot like meat vs vegie.