American Journal ol EpWemiotogy Copyright C 1992 by The Johns Hopkins Urwefsrty School of Hygiene and Pubfic Health Al rights reserved
Vol. 136, No 5 Panted in U.S.A.
Maternal Nutrition and Spontaneous Preterm Birth
Michael S. Kramer,1-2 Frances H. McLean,3 Erica L. Eason,3 and Robert H. Usher13
Previous studies suggesting that maternal undemutrition increases the risk of preterm birth have suffered from several methodological shortcomings, including use of total gestational weight gain rather than net rate of gain in maternal tissue, inclusion of induced preterm deliveries, and error-prone gestational age measurements based solely on menstrual dates. The authors have attempted to overcome these shortcomings by investigating the potential etiologic roles of prepregnancy body mass index, net rate of maternal weight gain, height, and a number of other potential biological and sociodemographic determinants of spontaneous (i.e., noninduced) preterm birth in a cohort of 13,102 women with early ultrasound-confirmed gestational age who delivered at the Royal Victoria Hospital in Montreal, Quebec, Canada, between January 1, 1980 and March 31, 1989. Total weight gain, but not body mass index, was highly significantly associated with spontaneous preterm birth, averaging 14.6, 12.5, 9.9, and 9.1 kg, in women delivering at 37 or more, less than 37, less than 34, and less than 32 completed weeks, respectively. Although the relation persisted when weight gain was expressed as an overall rate, it disappeared when the analysis was based on net rate; mean net rates of gain were 0.28, 0.29, 0.27, and 0.27 kg/week, respectively. On the basis of multiple logistic regression analyses, significant determinants of birth at less than 37 weeks included maternal short stature; noncompletion of high school; unmarried status; smoking; diabetes; urinary tract infection within 2 weeks of delivery; prepregnancy hypertension; severe pregnancy-induced hypertension; and previous history of preterm delivery, low birth weight, or neonatal death. Most of these factors retained their significance for birth at less than 34 and less than 32 weeks. In fact, the effect of low maternal education was even stronger at these more severe "levels" of preterm birth. The authors conclude that prepregnancy weight-for-height and gestational weight gain are not important determinants of spontaneous preterm birth and that some previous studies have mistaken an effect of shortened gestation for its cause. Other biologic and social determinants, however, indicate priorities for future research and intervention. Am J Epidemiol 1992; 136:574-83. gestational age; infant, low birth weight; infant, premature; nutrition
Maternal nutrition both prior to and during gestation is of known etiologic importance in the growth of the human fetus. The
evidence is extremely strong that prepregnancy weight-for-height and gestational weight gain each have significant and inde-
Received for publication August 12, 1991, and in final form May 26, 1992. 1 Department of Pediatrics, McGill University, Montreal, Quebec, Canada. 2 Department of Epidemiology and Biostatistics, McGil University Faculty of Medicine, Montreal, Quebec, Canada. 3 Department of Obstetrics and Gynecology, McGill University Faculty of Medicine, Montreal, Quebec, Canada.
Repnnt requests to Dr. Michael S. Kramer, Department of Epidemiology and Biostatistics, McGill University, 1020 Pine Avenue West, Montreal, Quebec, H3A 1A2, Canada. This work was carried out when Dr. Kramer was a senior chercheur-boursier of the Foods de la recherche en sante du Quebec. Presented in part at the annual meeting of the Society for Pediatric Research, Anaheim, California, May 8,1990.
574
Maternal Nutrition and Spontaneous Preterm Birth
pendent impacts on fetal growth and that low prepregnancy weight-for-height and low gestational weight gain are important determinants of intrauterine growth retardation (1-4). Maternal stature is also known to affect fetal growth independently of weightfor-height (1-8). Because a mother's height may be partly determined by her nutritional status during childhood (especially in developing countries) and because it may be reflected by uterine size, placental blood flow, and, hence, the delivery of placental nutrients, maternal stature may also be considered, at least in part, a nutritional determinant. Compared with the extensive literature and convincing evidence regarding fetal growth, relatively few reports have appeared concerning the relation between maternal nutrition and gestational duration, and particularly the relation between poor prepregnancy or gestational nutrition and the risk of preterm birth (9). Although methodologically sound studies have been unanimous in not finding any relation between maternal height and gestational duration (7, 10-14), several have reported that thin women are at significantly elevated risk of preterm birth (7,10, 15-17). The evidence of an important etiologic role for gestational weight gain is suggestive but mixed (11, 12, 14, 17-27). Unfortunately, previous studies relating maternal nutritional factors to preterm birth have suffered from four major methodological shortcomings. First and foremost, many previous studies have based their assessment of gestational weight gain on total weight gain rather than rate of weight gain. Obviously, women who deliver preterm will have had less time to gain weight and will, therefore, have a lower total weight gain; in other words, use of total weight gain does not permit one to distinguish the effect of a shortened gestation from its cause. Second, even among those studies that expressed weight gain as a rate, most based their analyses on overall rate of gain. Since weight gains are generally far smaller in the first trimester than in the second and third (4), women with extremely preterm deliveries will have had a lower overall rate of gain, on
575
average, than those who deliver at term. Moreover, fetal weight increases exponentially, with highest gains in the third trimester, while the overall rate of maternal weight gain is fairly constant after the first trimester (4). Thus, with advancing gestation, the portion of overall gain attributable to maternal tissues and nutritional stores diminishes relative to that of increased fetal size. Associations between overall rate of gain and preterm birth may, therefore, reflect the effect of fetal growth, rather than maternal nutrition per se. A recent analysis based on the 1980 US National Natality Survey (26) is illustrative in this regard; the relation between total gestational weight gain and the risk of preterm low birth weight (gestational age less than 37 weeks and birth weight less than 2,500 g) virtually disappeared when the weight gain measure was based on the net gain ((total weight gain minus infant birth weight)/gestational age) per week. A third methodological problem in previous studies is that most have been based on all preterm births, including induced deliveries. Since one of the major indications for induction in modern obstetrics is a growth-retarded fetus and because there is a known relation between maternal nutrition and intrauterine growth retardation (1-4), inclusion of induced preterm births could also lead to a biased association. Finally, previous studies (7, 10-27) have almost exclusively used definitions of preterm birth based on gestational age measurements estimated from the mother's recollection of her last normal menstrual period. This method has been shown to be prone to significant error, particularly at the extremes of gestational age, thus leading to potential misclassification of some growth-retarded infants as preterm and vice versa (28). Although we are aware of no evidence that such errors are more likely to occur in women with poor nutritional status, differential misclassification could bias the findings in favor of an association between maternal nutrition and preterm birth. We have attempted to overcome these methodological shortcomings in a hospitalbased cohort study of 13,102 women with
576
Kramer et al.
gestational age confirmed by early ultrasound. Moreover, because the implications for morbidity and mortality differ by gestational age even among preterm infants, we have separately investigated the etiologic roles of maternal nutrition and a number of other potential biologic and sociodemographic determinants of delivery prior to 37, 34, and 32 completed weeks. Finally, we have attempted to interpret our findings in the context of previous research in this field by comparing our results based on total weight gain versus rate of weight gain, overall rate of weight gain versus net rate of gain, with and without inclusion of inductions, and with and without ultrasound confirmation of menstrual dates. MATERIALS AND METHODS
The study is based on infants born at Montreal's Royal Victoria Hospital between January 1, 1980 and March 31, 1989. The data concerning these births are contained in the McGill Obstetric and Neonatal Database (29). The study hospital serves an ethnically diverse population. Although the majority of study women are Canadian born and Caucasian, they also comprise sizable numbers of Greek, Portuguese, Italian, and southeast Asian immigrants; relatively few are native or immigrant blacks. We restricted our analysis to live singleton births without evidence of congenital intrauterine infection, chromosomal anomalies, or other major malformations whose gestational age calculated from the last normal menstrual period agreed, within ±7 days, with an estimate based on early second-trimester (usually 16-18 weeks) ultrasound determination of the fetal biparietal diameter. As a further means to ensure accurate measures of gestational age, we excluded infants whose fetal growth ratio (the ratio of the observed birth weight to the mean birth weight for the observed gestational age, in days, based on published standards from this hospital (30)) was less than 0.40 or greater than 1.40. These restrictions and exclusions resulted in a study sample of 13,130 infants. Further restriction of the study sample
was required to exclude preterm infants born after induced labor or Caesarean section for reasons other than threatened preterm delivery. Thus, only those preterm deliveries were included for which there was either 1) documented spontaneous onset of labor (progressive uterine contractions and cervical dilatation and effacement) leading to delivery, or 2) induction for premature rupture of membranes or chorioamnionitis, or 3) Caesarean section for maternal or fetal indications with an increased risk of preterm delivery (e.g., abruptio placentae, placenta previa). After the above restrictions, the number of eligible study subjects was 13,102; the principal study analyses and results are based on this study sample. Preterm birth was defined as delivery prior to 37 completed weeks (259 days). The potential nutritional determinants investigated (also recorded in the McGill Obstetric and Neonatal Database) were maternal height, prepregnancy body mass index (prepregnancy weight (kg)/height (m)2), and gestational weight gain. Gestational weight gain was analyzed in four ways: 1) total weight gain, i.e., last (recalled) weight prior to delivery minus (recalled) prepregnancy weight; 2) net weight gain, defined as total weight gain minus the infant's birth weight; 3) overall rate of gain, i.e., total weight gain/gestational age; and 4) net rate of gain (net gain/gestational age). (Because serial weights during gestation are not recorded in the McGill Obstetric and Neonatal Database, we were not able to assess the influence of weight gain pattern.) Other (nonnutritional) potential determinants investigated include maternal age (less than 20, 20-34, and 35 or more years); years of education completed (10 or less, 11-12, 13-15, and 16 or more years); smoking (0, 1-10, and 11 or more cigarettes per day, on average, during pregnancy); alcohol consumption (less than 1 vs. 1 or more drinks per day, on average, during pregnancy); parity (primiparous vs. multiparous); marital status; prior history (yes vs. no) of preterm birth, low birth weight (less than 2,500 g), or neonatal death; onset of prenatal care during the first trimester (yes vs. no); the
Maternal Nutrition and Spontaneous Preterm Birth
presence or absence of diabetes of any degree and duration, including gestational diabetes (a large majority of study women underwent a glucose tolerance screen early in the third trimester, in which blood glucose was measured 1 hour after a standard oral glucose load); culture-positive urinary tract infection within 2 weeks of delivery (untested women were assumed to be culture negative); prepregnancy hypertension; and severe pregnancy-induced hypertension, based on either specific mention of "severe preeclampsia" in the medical record by at least one obstetrician or documented evidence of frank eclampsia. Maternal education was incompletely recorded in the McGill Obstetric and Neonatal Database during the early 1980s. Since this variable is routinely recorded on the Quebec provincial birth certificate, Montreal metropolitan area birth certificate files were merged with the McGill Obstetric and Neonatal Database for the years 1980-1985. Most statistical analyses were carried out using SAS-PC (31). BMDP (32) was used for all logistic regression analyses, which were performed nonstepwise. Because significant effect modification has not been demonstrated in previous etiologic studies of preterm birth and was not hypothesized by us a priori, logistic models did not include interaction terms. RESULTS
A description of the study cohort, in terms of outcomes and potential determinants under investigation, is contained in table 1. The 13,102 mother-infant pairs included in the study sample were similar to the 15,124 who were otherwise eligible but were excluded (because of missing ultrasound data, discrepant ultrasound and last normal menstrual period estimates of gestational age, or biologically suspect fetal growth ratios) with respect to height (162.3 vs. 162.1 cm), prepregnancy body mass index (22.1 kg/m2 in both), and total gestational weight gain (14.5 vs. 14.3 kg). Mean birth weight and assigned gestational age (based on last normal menstrual period when the ultrasound estimate
577
was concordant or missing and based on ultrasound when last normal menstrual period was unknown) were slightly higher in the study cohort, however (3,395 vs. 3,334 g and 276 vs. 274 days). In bivariate analyses, prepregnancy body mass index was not significantly associated with spontaneous birth prior to 37, 34, or 32 weeks (table 2). Women who delivered prior to 37 weeks were, on average, 1.3 cm shorter {p < 0.0001) than those who delivered at term, with a significantly (p < 0.001) higher proportion with short stature (< 157.5 cm), although there were no significant differences in the heights of women who delivered before versus after 34 or 32 weeks. Total and net gestational weight gains were highly significantly (p < 0.0001) associated with spontaneous preterm birth, with total weight gains of 14.6, 12.5, 9.9, and 9.1 kg in women delivering at 37 or more, less than 37, less than 34, and less than 32 completed weeks, respectively. Women who delivered prior to 34 and 32 weeks also had significantly (p < 0.001) lower overall rates of weight gain. When weight gain was expressed as a net rate, however, the association disappeared; net rates were 0.28, 0.29, 0.27, and 0.27 kg/ week in the four gestational age groups, respectively. The remainder of the bivariate analyses are also summarized in table 2. Based on the results of the bivariate analyses, multiple logistic regression analyses were carried out for each of the three "levels" of preterm birth and the following independent variables: maternal age, maternal education, marital status, cigarette smoking, alcohol consumption, diabetes, urinary tract infection, prepregnancy hypertension, and severe pregnancy-induced hypertension. Low prepregnancy body mass index (
Vol. 136, No 5 Panted in U.S.A.
Maternal Nutrition and Spontaneous Preterm Birth
Michael S. Kramer,1-2 Frances H. McLean,3 Erica L. Eason,3 and Robert H. Usher13
Previous studies suggesting that maternal undemutrition increases the risk of preterm birth have suffered from several methodological shortcomings, including use of total gestational weight gain rather than net rate of gain in maternal tissue, inclusion of induced preterm deliveries, and error-prone gestational age measurements based solely on menstrual dates. The authors have attempted to overcome these shortcomings by investigating the potential etiologic roles of prepregnancy body mass index, net rate of maternal weight gain, height, and a number of other potential biological and sociodemographic determinants of spontaneous (i.e., noninduced) preterm birth in a cohort of 13,102 women with early ultrasound-confirmed gestational age who delivered at the Royal Victoria Hospital in Montreal, Quebec, Canada, between January 1, 1980 and March 31, 1989. Total weight gain, but not body mass index, was highly significantly associated with spontaneous preterm birth, averaging 14.6, 12.5, 9.9, and 9.1 kg, in women delivering at 37 or more, less than 37, less than 34, and less than 32 completed weeks, respectively. Although the relation persisted when weight gain was expressed as an overall rate, it disappeared when the analysis was based on net rate; mean net rates of gain were 0.28, 0.29, 0.27, and 0.27 kg/week, respectively. On the basis of multiple logistic regression analyses, significant determinants of birth at less than 37 weeks included maternal short stature; noncompletion of high school; unmarried status; smoking; diabetes; urinary tract infection within 2 weeks of delivery; prepregnancy hypertension; severe pregnancy-induced hypertension; and previous history of preterm delivery, low birth weight, or neonatal death. Most of these factors retained their significance for birth at less than 34 and less than 32 weeks. In fact, the effect of low maternal education was even stronger at these more severe "levels" of preterm birth. The authors conclude that prepregnancy weight-for-height and gestational weight gain are not important determinants of spontaneous preterm birth and that some previous studies have mistaken an effect of shortened gestation for its cause. Other biologic and social determinants, however, indicate priorities for future research and intervention. Am J Epidemiol 1992; 136:574-83. gestational age; infant, low birth weight; infant, premature; nutrition
Maternal nutrition both prior to and during gestation is of known etiologic importance in the growth of the human fetus. The
evidence is extremely strong that prepregnancy weight-for-height and gestational weight gain each have significant and inde-
Received for publication August 12, 1991, and in final form May 26, 1992. 1 Department of Pediatrics, McGill University, Montreal, Quebec, Canada. 2 Department of Epidemiology and Biostatistics, McGil University Faculty of Medicine, Montreal, Quebec, Canada. 3 Department of Obstetrics and Gynecology, McGill University Faculty of Medicine, Montreal, Quebec, Canada.
Repnnt requests to Dr. Michael S. Kramer, Department of Epidemiology and Biostatistics, McGill University, 1020 Pine Avenue West, Montreal, Quebec, H3A 1A2, Canada. This work was carried out when Dr. Kramer was a senior chercheur-boursier of the Foods de la recherche en sante du Quebec. Presented in part at the annual meeting of the Society for Pediatric Research, Anaheim, California, May 8,1990.
574
Maternal Nutrition and Spontaneous Preterm Birth
pendent impacts on fetal growth and that low prepregnancy weight-for-height and low gestational weight gain are important determinants of intrauterine growth retardation (1-4). Maternal stature is also known to affect fetal growth independently of weightfor-height (1-8). Because a mother's height may be partly determined by her nutritional status during childhood (especially in developing countries) and because it may be reflected by uterine size, placental blood flow, and, hence, the delivery of placental nutrients, maternal stature may also be considered, at least in part, a nutritional determinant. Compared with the extensive literature and convincing evidence regarding fetal growth, relatively few reports have appeared concerning the relation between maternal nutrition and gestational duration, and particularly the relation between poor prepregnancy or gestational nutrition and the risk of preterm birth (9). Although methodologically sound studies have been unanimous in not finding any relation between maternal height and gestational duration (7, 10-14), several have reported that thin women are at significantly elevated risk of preterm birth (7,10, 15-17). The evidence of an important etiologic role for gestational weight gain is suggestive but mixed (11, 12, 14, 17-27). Unfortunately, previous studies relating maternal nutritional factors to preterm birth have suffered from four major methodological shortcomings. First and foremost, many previous studies have based their assessment of gestational weight gain on total weight gain rather than rate of weight gain. Obviously, women who deliver preterm will have had less time to gain weight and will, therefore, have a lower total weight gain; in other words, use of total weight gain does not permit one to distinguish the effect of a shortened gestation from its cause. Second, even among those studies that expressed weight gain as a rate, most based their analyses on overall rate of gain. Since weight gains are generally far smaller in the first trimester than in the second and third (4), women with extremely preterm deliveries will have had a lower overall rate of gain, on
575
average, than those who deliver at term. Moreover, fetal weight increases exponentially, with highest gains in the third trimester, while the overall rate of maternal weight gain is fairly constant after the first trimester (4). Thus, with advancing gestation, the portion of overall gain attributable to maternal tissues and nutritional stores diminishes relative to that of increased fetal size. Associations between overall rate of gain and preterm birth may, therefore, reflect the effect of fetal growth, rather than maternal nutrition per se. A recent analysis based on the 1980 US National Natality Survey (26) is illustrative in this regard; the relation between total gestational weight gain and the risk of preterm low birth weight (gestational age less than 37 weeks and birth weight less than 2,500 g) virtually disappeared when the weight gain measure was based on the net gain ((total weight gain minus infant birth weight)/gestational age) per week. A third methodological problem in previous studies is that most have been based on all preterm births, including induced deliveries. Since one of the major indications for induction in modern obstetrics is a growth-retarded fetus and because there is a known relation between maternal nutrition and intrauterine growth retardation (1-4), inclusion of induced preterm births could also lead to a biased association. Finally, previous studies (7, 10-27) have almost exclusively used definitions of preterm birth based on gestational age measurements estimated from the mother's recollection of her last normal menstrual period. This method has been shown to be prone to significant error, particularly at the extremes of gestational age, thus leading to potential misclassification of some growth-retarded infants as preterm and vice versa (28). Although we are aware of no evidence that such errors are more likely to occur in women with poor nutritional status, differential misclassification could bias the findings in favor of an association between maternal nutrition and preterm birth. We have attempted to overcome these methodological shortcomings in a hospitalbased cohort study of 13,102 women with
576
Kramer et al.
gestational age confirmed by early ultrasound. Moreover, because the implications for morbidity and mortality differ by gestational age even among preterm infants, we have separately investigated the etiologic roles of maternal nutrition and a number of other potential biologic and sociodemographic determinants of delivery prior to 37, 34, and 32 completed weeks. Finally, we have attempted to interpret our findings in the context of previous research in this field by comparing our results based on total weight gain versus rate of weight gain, overall rate of weight gain versus net rate of gain, with and without inclusion of inductions, and with and without ultrasound confirmation of menstrual dates. MATERIALS AND METHODS
The study is based on infants born at Montreal's Royal Victoria Hospital between January 1, 1980 and March 31, 1989. The data concerning these births are contained in the McGill Obstetric and Neonatal Database (29). The study hospital serves an ethnically diverse population. Although the majority of study women are Canadian born and Caucasian, they also comprise sizable numbers of Greek, Portuguese, Italian, and southeast Asian immigrants; relatively few are native or immigrant blacks. We restricted our analysis to live singleton births without evidence of congenital intrauterine infection, chromosomal anomalies, or other major malformations whose gestational age calculated from the last normal menstrual period agreed, within ±7 days, with an estimate based on early second-trimester (usually 16-18 weeks) ultrasound determination of the fetal biparietal diameter. As a further means to ensure accurate measures of gestational age, we excluded infants whose fetal growth ratio (the ratio of the observed birth weight to the mean birth weight for the observed gestational age, in days, based on published standards from this hospital (30)) was less than 0.40 or greater than 1.40. These restrictions and exclusions resulted in a study sample of 13,130 infants. Further restriction of the study sample
was required to exclude preterm infants born after induced labor or Caesarean section for reasons other than threatened preterm delivery. Thus, only those preterm deliveries were included for which there was either 1) documented spontaneous onset of labor (progressive uterine contractions and cervical dilatation and effacement) leading to delivery, or 2) induction for premature rupture of membranes or chorioamnionitis, or 3) Caesarean section for maternal or fetal indications with an increased risk of preterm delivery (e.g., abruptio placentae, placenta previa). After the above restrictions, the number of eligible study subjects was 13,102; the principal study analyses and results are based on this study sample. Preterm birth was defined as delivery prior to 37 completed weeks (259 days). The potential nutritional determinants investigated (also recorded in the McGill Obstetric and Neonatal Database) were maternal height, prepregnancy body mass index (prepregnancy weight (kg)/height (m)2), and gestational weight gain. Gestational weight gain was analyzed in four ways: 1) total weight gain, i.e., last (recalled) weight prior to delivery minus (recalled) prepregnancy weight; 2) net weight gain, defined as total weight gain minus the infant's birth weight; 3) overall rate of gain, i.e., total weight gain/gestational age; and 4) net rate of gain (net gain/gestational age). (Because serial weights during gestation are not recorded in the McGill Obstetric and Neonatal Database, we were not able to assess the influence of weight gain pattern.) Other (nonnutritional) potential determinants investigated include maternal age (less than 20, 20-34, and 35 or more years); years of education completed (10 or less, 11-12, 13-15, and 16 or more years); smoking (0, 1-10, and 11 or more cigarettes per day, on average, during pregnancy); alcohol consumption (less than 1 vs. 1 or more drinks per day, on average, during pregnancy); parity (primiparous vs. multiparous); marital status; prior history (yes vs. no) of preterm birth, low birth weight (less than 2,500 g), or neonatal death; onset of prenatal care during the first trimester (yes vs. no); the
Maternal Nutrition and Spontaneous Preterm Birth
presence or absence of diabetes of any degree and duration, including gestational diabetes (a large majority of study women underwent a glucose tolerance screen early in the third trimester, in which blood glucose was measured 1 hour after a standard oral glucose load); culture-positive urinary tract infection within 2 weeks of delivery (untested women were assumed to be culture negative); prepregnancy hypertension; and severe pregnancy-induced hypertension, based on either specific mention of "severe preeclampsia" in the medical record by at least one obstetrician or documented evidence of frank eclampsia. Maternal education was incompletely recorded in the McGill Obstetric and Neonatal Database during the early 1980s. Since this variable is routinely recorded on the Quebec provincial birth certificate, Montreal metropolitan area birth certificate files were merged with the McGill Obstetric and Neonatal Database for the years 1980-1985. Most statistical analyses were carried out using SAS-PC (31). BMDP (32) was used for all logistic regression analyses, which were performed nonstepwise. Because significant effect modification has not been demonstrated in previous etiologic studies of preterm birth and was not hypothesized by us a priori, logistic models did not include interaction terms. RESULTS
A description of the study cohort, in terms of outcomes and potential determinants under investigation, is contained in table 1. The 13,102 mother-infant pairs included in the study sample were similar to the 15,124 who were otherwise eligible but were excluded (because of missing ultrasound data, discrepant ultrasound and last normal menstrual period estimates of gestational age, or biologically suspect fetal growth ratios) with respect to height (162.3 vs. 162.1 cm), prepregnancy body mass index (22.1 kg/m2 in both), and total gestational weight gain (14.5 vs. 14.3 kg). Mean birth weight and assigned gestational age (based on last normal menstrual period when the ultrasound estimate
577
was concordant or missing and based on ultrasound when last normal menstrual period was unknown) were slightly higher in the study cohort, however (3,395 vs. 3,334 g and 276 vs. 274 days). In bivariate analyses, prepregnancy body mass index was not significantly associated with spontaneous birth prior to 37, 34, or 32 weeks (table 2). Women who delivered prior to 37 weeks were, on average, 1.3 cm shorter {p < 0.0001) than those who delivered at term, with a significantly (p < 0.001) higher proportion with short stature (< 157.5 cm), although there were no significant differences in the heights of women who delivered before versus after 34 or 32 weeks. Total and net gestational weight gains were highly significantly (p < 0.0001) associated with spontaneous preterm birth, with total weight gains of 14.6, 12.5, 9.9, and 9.1 kg in women delivering at 37 or more, less than 37, less than 34, and less than 32 completed weeks, respectively. Women who delivered prior to 34 and 32 weeks also had significantly (p < 0.001) lower overall rates of weight gain. When weight gain was expressed as a net rate, however, the association disappeared; net rates were 0.28, 0.29, 0.27, and 0.27 kg/ week in the four gestational age groups, respectively. The remainder of the bivariate analyses are also summarized in table 2. Based on the results of the bivariate analyses, multiple logistic regression analyses were carried out for each of the three "levels" of preterm birth and the following independent variables: maternal age, maternal education, marital status, cigarette smoking, alcohol consumption, diabetes, urinary tract infection, prepregnancy hypertension, and severe pregnancy-induced hypertension. Low prepregnancy body mass index (
