The usual approach is to divide the FEM slab in zones of different (constant) thickness. It is in the philosophy of FEM that the results will approach provided the math doesn't mess the thing. If ETABs has no 8 nodes' brick or tetrahedral elements that can in other way get your model more to your taste or needs, you may use one of the other programs that have; everything is for something.
For this purpose i suggest you to use Robot Structural Analysis Profesional, by Autodesk (French Software for structural engineers)
This offer more posibylites to defining slab variyng thickness,
1.Slab thickness with two dimensions,
2.Slab thickness with three dimensions,
........
........
Etabs is very good software,but there are some difficulties to model analysis and design some elements,
SAP lacks some features that make ETABS useful in building design. Also, obtaining meaningful output for design using solid elements is more difficult. It would be nice if ETABS incorporated SAP's ability to assign offsets to the slab thickness to vary the centerline while maintaining connectivity using internal constraints.. kind of like what ETABS does now with the insertion point assignment for frame elements. But ETABS should be able to model most variable depth slab design applications for you now.
In ETABS, slabs with different slab sections/thicknesses are typically modeled sharing the same centerline for each floor in order to maintain connectivity. If you want to model a slab with variable depth, you can define different slab section thicknesses in ETABS, but by default the different slab sections share the same centerline on each floor, an assumption which might be problematic in some situations, but in most situations that assumption would be ok.. drop panels, for example, can be reasonably assumed to have the same centerline as the slab for most design applications.
In those design situations where the change in centerline location is a concern, like with a step level, you could add a reference plane in order to model the change in slab centerline locations in different areas of the same floor, in which you would model upper and lower portions of the slab disconnected at first, then connect them together using wall and/or column elements. No need to hassle with other programs like SAP or Robot in most situations involving variable slab depth.
If you were designing an 8 foot thick/tall machine foundation with soil properties varying significantly by depth, then you might be better served using solid finite elements, but for slab designs, even with variable thickness, I haven't seen many situations where solid elements would be useful.
For the past 4 years I have been mostly calculating structures, something quite stringent of what I am by now released. In these years there have been in the tens at least cases where having accurate 3D solid models of one or two floors, mainly to be treated model with a single hypothesis, even if some hypotheses used the same basic model, has been useful.
This was most in order to comparison, accuracy of the geometry or simply to have some reliable model.
In this cases I modeled the solids in autocad and passed the .sat file to visualNastran, that used tetrahedral elements. This way you can model any geometry Autocad can deal with, changes in slab depth one of them. If more complicated in geometry, you may model the solid in Inventor, to be used the same way.
The more recent Autodesk Inventor 2009 has within analysis capabilities and very likely would have performed for me perfectly -without resource to visualNastran, I mean, were it available- for rarely I was investigating something nonlinear.
Furthermore it seems it exports to Ansys (that I don't use) and for these once-a-time items surely can go so as deep as required.
