The essence of our system is the pulse. When people are on medications that are biologically active, then the pulse changes. And since we always and only treat what is in the pulse, as long as we read it correctly and give the herbs that the pulse is asking for, we are always in the safe zone.

For example, if a person is on an effective dose of blood thinners, than possibly that has taken some of the stasis signs out of the pulse and abdomen, and as such we will be less likely to find dry hard blood stasis in the picture requiring blood cracking formulas. I know this is a way too simplified an example, but you get the idea. 

So the first part of the argument is that since we don't take people off of their medication anyway, the medication becomes part of the overall picture. This is why we rarely see side effects due to herb-drug interaction.

Second part of the argument is that there indeed, as mentioned in previous comments, is very little research in this field. Pharmacological knowledge about Chinese herbal medicine is still in its infancy with most studies only being conducted in vitro, or in vivo. Very few human subject studies have been performed. And all the studies that have been done focus heavily on single herbs and not on formulas.

The reason for this is that the analytical models used are designed for observation of as few variables as possible. And given that even a single herb contains multiple active compounds, studying formulas becomes virtually impossible with Western analytical methods. The only studies that do yield some clinically relevant data are the pragmatic trials (outcomes-based studies). But in the eyes of Western science, these are hierarchically not very important.

Which brings us to our point one which is that because Western methods have not been able to properly assess herbal formula efficacy and safety, we can but continue to rely on our traditional methods of pulse and abdomen diagnosis to ensure safety and efficacy.

In closing, this does not eliminate the problem of duplication or redundancy. Sometimes a patient is on medication x, which yields result y. And the pulse asks for herb z, so we administer it. But al of a sudden, the person's results change drastically (e.g. major decline in serum glucose, big change in blood clotting factor, drop in blood pressure, etc). This could be attributed to the fact that why drug x was trying to fix their aforementioned problems, it was only able to do so up to a certain level, but now the addition of herbs also fixes the same problem and picked up where the drugs left off, going where the drugs couldn't even go. So this is like a redundancy or treatment doubling. At this point the patient should either reduce the dosage of their medications, or go off the herbs. The other option would be for the herbs to shift strategy and treat other problems that are also going on, and disregard the problem the meds are trying to fix. But since we rely on the pulse, that is not always a decision that is easy to make.