Diet studies only compare plant foods against a poor standard diet
Some studies do compare against a poor baseline, but not all of them. Substitution analyses model swapping red meat for plant protein within the same diet, and they find lower risk of death and heart disease. And the case for veganism does not depend on plants winning a health contest, only on a plant-based diet being adequate, which it is.
Most of the evidence that plant foods are healthy comes from comparing people who eat more of them with people on a typical Western diet, which almost everyone agrees is poor. Beating a bad baseline tells you little about whether a plant-based diet beats a carefully built whole-food diet that includes meat. Until those two are compared directly, the honest position is that the literature is mixed and the health case for going fully plant-based is unproven.
The comparison problem is real, but the evidence does not rest on it alone. Many cohort studies compare people who eat more of a food with people who eat less, against a background diet that is often poor. That limits what any single study can show.
Researchers know this, and they have a tool for it. Substitution analyses ask a narrower question: what happens when someone swaps one food for another and keeps everything else the same? These analyses compare protein with protein, not plants with junk food. Replacing unprocessed red meat with legumes or nuts is associated with lower cardiovascular risk across meta-analyses of prospective cohorts. In a large US cohort of over 400,000 adults, replacing just 3% of calories from red meat protein with plant protein was associated with roughly 13 to 15% lower all-cause mortality. These are observational findings and carry the usual caveats. They also point consistently in one direction.
It is true that there is no long-term randomised trial of a whole-food plant diet against a whole-food meat diet. Trials of that length on hard outcomes barely exist for any diet. A missing trial is not evidence that the two diets come out even. It means the question has to be decided on the evidence we do have, and that evidence leans towards plants.
"Mixed literature" also needs care. Some individual studies find weak or null effects for vegetables on particular outcomes. A null result on one endpoint sits alongside clear benefits on others, including all-cause mortality in the study under discussion. That is ordinary variation, not a split verdict.
Finally, the ethical argument does not need plant diets to be healthier than the best possible omnivorous diet. It needs them to be adequate. The major dietetic associations say well-planned vegan diets are. If both diets can be healthy, the health question no longer tips the balance, and the remaining question is whether the harm to animals is necessary.