Selection bias in occupational sperm studies.