n this study, we observed substantial improvements in birth
weight (>200 g) among infants of women who reported low
cigarette use who then reduced exposure. However, we
observed no improvement in birth weight among infants of
women with low cotinine concentration who then reduced their
tobacco exposure. One potential limitation of using cotinine as
a measure of tobacco exposure is that the serum cotinine concentration of sporadic smokers (such as those who smoke on
weekends only) may not be at a steady state. Hence, the urine
cotinine concentrations used in our study may not have accurately reflected tobacco exposure, and this could have resulted
in the exposure misclassification of some women. This type of
misclassification could have diminished potential differences
in infant birth weight between women who truly reduced their
exposure and those who did not change.
In previous studies of the effects of smoking reduction on
birth outcomes, researchers have relied on categorizing women
according to changes in exposure (5, 13). This approach was
part of our analytic strategy as well; however, it is inherently
limited. It diminishes power to detect exposure effects because
information contained within categories of exposure is lost. To
overcome this limitation, we also performed regression by
using smoothing, which allowed us to characterize the functional relation between tobacco exposure and birth weight in
detail. We found that third-trimester tobacco exposure is a
more important determinant of birth weight than is early exposure. This is consistent with previous studies of smoking cessation in which women who quit smoking during pregnancy
delivered infants who weighed as much as infants of never
smokers (7, 12). Because we did find a significant association
between urine cotinine concentration early in pregnancy and
birth weight, we cannot rule out the possibility that early
tobacco exposure has some minor effect on fetal growth.
However, any effects of early exposure on birth weight are
likely to be overshadowed by effects of third-trimester exposure. Finally, we found that most of the deleterious effects of
tobacco on infant birth weight appear to occur at low levels of
exposure (i.e., less than eight cigarettes per day or less than 700