Reliability of PCR directly from stool samples: usefulness of an internal standard.