TDMA by this reasoning has an obvious hazard, because of the structural patterns at extreme low frequencies: the signal to some degreee “comes and goes” at a regular repetition rate that is within bio-frequencies.
Pure CDMA, or code-division multiple access has no such structural coming and going, or volumetric repetition pattern. The structure is in the signal coding: like three people talking at you at the same time in different dialects, you can easily listen to the one in your own dialect. So there may be patterns in a dialect, but it’s not evident in the power level, so how it would be discerned is harder to see.
Of course, whether they take it in turns (TDMA) or in different dialects at once (CDMA), if they’re shouting loud enough it could damage your ear-drums. But this is probably where the pure energy / volume argument of safety margins breaks down anyway. And then there is the issue of whether the inherent signal markers in CDMA, that are more masked when the channels are active than at night when inactive, are what causes sleep problems around 3G base stations …
So it’s not a question easily answered, but certainly there is no intuitive case for anticipating increasing dangers along the 1G, 2G, 3G, 4G line.