The main requirement is that you deal with the sophisticated probability sampling technique employed by NHANES. Basically, some groups are oversampled so they aren't missed. The usual way this is done is by using the SUDAAN software program.
Other than that, I don't think there are other requirements - just cite sources, etc. If you are using the 1999-2000 dataset, the files are in SAS export format which makes it much easier to deal with. NHANES III and previous versions have to be imported into SAS with some code but its pretty simple.
