Classifying tai-chi-for-health
What are we talking about when we talk about forms-based tai chi? Evaluating the practice as an ignorant outsider, it's essentially a slow dance. There are a number of one-footed balancing postures, deep lunges or otherwise low stances, and opportunity to stretch the limbs as well as flex and extend. In addition to the meditative aspects, there is a degree of social interaction if it is performed in a group or class. These are all swell, but not particularly revolutionary compared to anything else.
As with many group dance and movement programs, tai chi's primary appeal is its compliance factor. People like doing it! A mediocre program that people do over the long term is better in every way than a perfect program that people do for six weeks then quit. Similarly, a mediocre program that people enjoy doing is generally better than a perfect program that people find boring or stressful. This may be where tai chi shines.
In reviewing some scientific studies, we find that there may or may not be awesome healing properties to tai chi. Most likely, it's just that it's a good idea to get up and move, particularly if you don't do so very often.
A Survey of PubMed Science by an Amateur
Context
There are lots of studies on forms-based tai chi as a useful exercise modality. Largely, what we can conclude from this glut of research is:
In particular, I find it preposterous that so few studies use a reasonable control group, such as a group that takes up chess or hopskotch or skeeball or Irish shin-kicking, to see whether the beneficial effects are due to something about Tai Chi, or sports, or mental exercise, or meditation, or another factor. It belies an egregious lack of rigor.
Inconclusive studies
I found a number of negative or inconclusive studies from a quick and dirty PubMed search:
These do not suggest that we should avoid tai chi. They just mean science is hard, and these scientists are good and honest folk. Admitting null results is good for science. It should be applauded.
Something is better than nothing, particularly for geezers
In terms of positive results, that quick-and-dirty search lead me to the entirely unsurprising conclusion that something is better than nothing. Again, it's important to keep the admonitions in the Context section in mind when reviewing these.
At this point I am confident that forms-based tai chi is superior to sitting in a wheelchair while waiting to die.
Tai chi actually beat something
There were a few studies that used control groups more useful than "watching Dynasty reruns".
Conclusion
A study of studies provides a good perspective for evaluating the science, in tai chi and beyond:
Most trials were judged at unclear risk of selection bias, generally reflecting inadequate reporting of the randomisation methods, but at high risk of performance bias relating to lack of participant blinding, which is largely unavoidable for these trials. Most studies only reported outcome up to the end of the exercise programme.
...
There is weak evidence that some types of exercise (gait, balance, co-ordination and functional tasks; strengthening exercise; 3D exercise [such as tai chi] and multiple exercise types) are moderately effective, immediately post intervention, in improving clinical balance outcomes in older people. Such interventions are probably safe.
Overall the science is not convincing in any specific way. I'm sticking with my unscientific "moving > not moving" perspective detailed above. This would point to forms-based tai chi as indeed a productive but not necessarily superior practice for health.