Flexible maximum likelihood methods for assessing joint effects in case-control studies with complex sampling