Well, medical studies can be confounded by stupidity.

Years ago, eating a good breakfast was considered a great thing to do for your health. I recall Lewis Thomas attacking this notion. He said that people who were very ill and fixing to die generally did not feel good enough to eat breakfast so they didn't. So for that group of very ill people, not eating a good breakfast was a symptom and not a cause.

So you get to fit and fat.

If you look at it, many thin people are unhealthy and should not be included. Those would be people with cancer or anorexics or chronic drug users and so on.

Also, are you healthy because you are fit or are you fit because you are healthy? Obviously, very ill people probably are not going to be fit.

But you can even get fat people who are unfit not because of lack of trying but because of underlying health problems. I recall that Henry the VIII was very fit until he was injured and then could not exercise. His being inactive resulted in his weight gain and corpulance.

Also, much of the cause of ill health in our society is caused by a terrible diet which leads to strokes, heart disease and diabetes. I would think that thin and fit people would avoid most of those circumstances. They are healthy because they work at it and not necessarily because they are fit and thin.

The problem is that being fit is certainly something a person has to work at. People who do it (whether fat or thin) probably avoid the well known risk factors that contribute to ill health like a diet high in saturated fat, chronic heavy smoking, alcoholism, drug abuse and so on.

Another thing that makes problems is what is being fat? Many studies just define fat as Body Mass Index which looks at only two variables: height and weight. Obviously a person can be heavy and short and still have a reasonably low percent body fat. So do lean heavy people get stuck into the group of fit and fat? I bet they do.

I really do not think that the studies are well enough designed to really address the question. If you look at data with confounding facters, of course, your results are going to be problematic.