To the Editor.
—The article by Li et al1 presents an objective assessment of smoking behavior in relationship to birth weight. Randomization of patients into a control and an experimental group clearly indicates the value of the interventions provided to the experimental group.The protocol did not, however, randomize study subjects into smokers and nonsmokers. Therefore, self-selection bias was not precluded. One might reasonably wonder whether quitters and reducers might have practiced other positive behaviors as well. Parity, length of time between pregnancies, history of a previous preterm or low-birth-weight infant, maternal illness, and drug use during pregnancy were not taken into account. Since there was no reason to assume low birth weight or preterm birth for those infants for whom those outcomes were not known, leaving them out of the analysis might have been more appropriate.