Case-control studies with contaminated controls