Large scale national surveys with clustering (and perhaps stratification)? SurveyLogistic wins hands down.

SAS's survey-something procedures were tailor-made for such large scale survey observational data, like nationally-representative studies. Surveylogistic allows for model fit testing, using the methods that are already available with logistic regression, without loss of generality. SurveyReg, SurveyLogistic, SurveyFreq procs etc have already been used extensively, for example in the analysis of NHANES (do a google search) study data or see http://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/statug_surveylogistic_sect013.htm for an illustration.