I recommend preparing a design criteria before embarking on cable sizing. I have supervised too many designers who simply look in the table in the Code without thinking about the factors that go into cable selection and sizing. By preparing a design criteria, you are forced to think of the factors you should consider.
The design criteria would normally include the items jbartos mentions, but some other things to think about:
1. Using the next-larger size cable. Often the cost is minimal, but is repaid with lower copper losses (higher efficiency) and increased equipment life because of reduced operating temperature.
2. Equipment terminal temperature rating. Sizing a cable based on 90 C operating temp does no good if the equipment lugs are rated 75 C.
3. Emergency overload capability. On some feeders you can "cheat" a little on the sizing if there is a definite overload condition anticipated. For example, XLPE insulation is normally rated for 90 C, but can operate at 130 C for a period usually defined as not exceeding 100 hours per year and no more than five overload periods in the cable's lifetime. By using 130 C instead of 90 C for sizing the cable under overload conditions, you can use a smaller cable, a useful cost-saving measure when dealing with MV systems and larger kcmil sizes.
4. System grounding and protective relaying will affect voltage insulation level, especially above 5 kV.