The undesirable consequences of controlling for birth weight in perinatal epidemiological studies.