Introduction

Despite advances in maternal and fetal medicine in the United States, rates of preterm (PTB) and low birth weight (LBW) births continue to rise. PTB and LBW have been associated with acute and chronic infant morbidity, as well as infant death [1–4]. The increased use of fertility interventions, which subsequently increase the risk for multiple gestations, have contributed to some of the observed increase in rates of PTB and LBW. However, 96% of US births are singleton births and PTB and LBW rates are increasing, or have remained high, in singletons for the past decade [5]. Minnesota reported increased rates in both PTB and LBW among singletons from 1990 to 2006 [6]. Between 1990 and 2006, rates of LBW among MN singletons increased from 4.1 to 4.9% and those of PTB from 6.7 to 8.5%.

The literature reports inconsistent associations between prenatal mental health, often measured by symptoms of depression, anxiety or stress, and adverse birth outcomes. Dayan et al. [7] did not find an association between anxiety and preterm birth, but Dole et al. [8] found a weak association between pregnancy-related anxiety and PTB. Glynn et al. found increased perception of prenatal stress and anxiety were associated with a two- to three-fold increase in PTB in a prospective study of 415 women at a prenatal care clinic [9, 10].

Women may be at higher risk for depressive symptoms during pregnancy; they also report depressive symptoms during their reproductive years at about twice the rate as men [11]. Researchers hypothesize that hormonal fluctuations during the onsets of menarche, pregnancy, postpartum, perimenopause and menopause may make women vulnerable to depressive symptoms throughout the life cycle [12]. A review article of studies examining prenatal depressive symptoms estimated a period prevalence for major and minor depressive symptoms during pregnancy at 18.4% [13]. Most studies have focused on identifying women who meet the clinical diagnosis for depression, yet symptoms of depression may also be important to women and their infants.

Research on substance use and mental health has suggested that depressed mood and use of substances, such as alcohol and tobacco, as well as other drugs, may co-occur [14–16]. Substance use is potentially harmful to the user and there may be teratogenic effects [17–21]. Tobacco and alcohol are the most commonly reported prenatal substances [22, 23] and both are associated with increased risk for birth defects [22, 23] and poor fetal growth [24]. A Swiss study of 6,284 singleton births found an association between tobacco use and LBW delivery (OR = 2.7, 95% CI: 2.1–3.5) [22]. Suzuki et al. found an association between smoking and LBW delivery that indicated tobacco’s main influence was on size at gestational age [25]. Infants of women who smoked were smaller for gestational age resulting in LBW, as opposed to LBW resulting from preterm birth [25].

The combined effect of substance use and poor maternal mental health may be greater than the independent effects of either factor alone. Singer et al. examined postpartum psychological distress, prenatal substance use and infant birth outcomes among 415 women and their infants [26]. Postpartum maternal psychological stress was assessed as soon as possible after the delivery as a surrogate for prenatal stress and was measured by the Global Severity Index, a summary score of nine psychiatric symptom patterns including depression and anxiety. They found that psychological distress, as well as alcohol, tobacco, and cocaine use, was independently associated with measures of fetal growth (i.e. head circumference, birthweight, microcephaly and small for gestational age) [26]. Singer et al. determined that cocaine-users had more severe symptoms of psychological distress that were more likely to be clinically significant compared to non-users. No studies were found that examined the potential interactive associations between maternal mental health, prenatal substance use and birth outcomes.

We examined the independent and interactive associations between maternal mental health, measured by maternal mood, and third trimester self-reported tobacco and alcohol use, and the risks for LBW and PTB among women who participated in Minnesota’s population-based survey, the Pregnancy Risk Assessment System (PRAMS). To our knowledge, such analysis has not been published.

Methods

Study Design

The Minnesota (MN) PRAMS, a cross-sectional written and telephone survey of postpartum women, was used to examine women’s experiences and behaviors during the pre-conception and prenatal periods. MN PRAMS is a component of the national state- and population-based PRAMS, administered by the Centers for Disease Control and Prevention (CDC). The Minnesota Department of Health (MDH) administers MN PRAMS [27]. Data were available for mothers and their infants who were born in 2002–2006 (partial year data for 2002). This study has Institutional Review Board approvals from the University of Minnesota and the MDH.

Sample

Approximately 200 postpartum women were randomly selected for MN PRAMS participation each month from their infant’s birth certificates. Women with infants who died during the early postpartum period were eligible and remained in the sample. The sampling file retains one birth certificate for women who had multiple births. PRAMS sampled 11,891 women during the study years 2002–2006, with 7,655 respondents overall and an annual response rate of 76%. African American and American Indian populations were over-sampled, however their stratum-specific response rates were too low to allow for subgroup analyses by race. For this analysis, 7,475 women had complete data (97.6% of the respondents).

Data Collection

MDH staff conducted the sampling and data collection, following CDC protocol for PRAMS administration [27]. A series of mailings began 2–4 months after delivery and was typically complete by 3–6 months postpartum. Because the questionnaire contained some items about the early postpartum period, mailings were timed to ensure that all women could respond for this period. If there was no response to repeated mailings, women were contacted and completed the questionnaire by telephone. The data collection cycle lasted approximately 60–95 days. Approximately 10% of total respondents provided their survey responses by phone.

Instrument

PRAMS consists of a standard set of core questions included by each state. In addition, states may include optional questions that CDC developed and standardized. The maternal mood question used for these analyses is not part of the core survey. PRAMS surveys vary in length by state and year, because of optional questions, but typically comprise 75–85 questions. Maternal survey responses are linked back to their infant’s birth certificate to create a dataset that includes infant information, including birth outcomes.

Measures

The main outcome measures, PTB (defined as less than 37 weeks gestation and coded as 1), and LBW (defined as less than 2,500 g and coded as 1), were obtained from the infant birth data. We created the main measure of maternal mood by combining responses to the question: “How would you describe the time during your pregnancy?” [27]. Women who responded that it was a “moderately hard time,” “a very hard” or “one of the worst times of my life” were classified as having a low mood level. Women with responses of “one of the happiest times of my life” or “a happy time with few problems” were classified as having a high mood level.

Prenatal substance use was defined by self-reported tobacco and alcohol use during pregnancy. We measured third trimester smoking status with the question: “How many cigarettes did you smoke on an average day?” The response categories were: “41 cigarettes or more; 21–40; 11–20; 6–10; 1–5; less than 1 cigarette; or none (0 cigarettes).” Women who reported smoking one or more cigarettes were classified as smokers. We measured third trimester alcohol use with the question: “How many alcoholic drinks did you have in an average week?” The response categories were: “14 drinks or more; 7–13; 4–6; 1–3; less than 1; or didn’t drink then.” We classified women as using alcohol if they reported any alcohol use.

We created interaction terms for low maternal mood and substance use to assess the interactive associations of prenatal tobacco use and maternal mood, and prenatal alcohol use and maternal mood with risks of having a LBW or PTB outcome. The interactive terms for tobacco and mood were: high mood/no tobacco use (referent), high mood/tobacco use, low mood/no tobacco use and low mood/tobacco use. The interactive terms for alcohol and mood were: high mood/no alcohol use (referent), high mood/alcohol use, low mood/no alcohol use, and low mood/alcohol use.

We examined several covariates as potential confounders. Maternal stress level was assessed with a 13-item set of potential stressors, a subset of the Modified Life Events Inventory [28]. Stressors included moving, being homeless, financial problems, relationship problems and an illness or death in the family. Women received an overall stress score by summing the yes responses to the 13 items. We classified women as having no life stressors, 1 or 2 stressors and 3 or more stressors. We also examined maternal age at delivery (aged 18–24, 25–34, 35+ years), marital status (yes/no), education level (some high school, high school, some college, college or more), multiparous (yes/no), prenatal care as early as wanted (yes/no), prenatal care in the first trimester (yes/no), pregnancy intention at the time of conception (intended/satisfied with timing, unintended/mistimed, unwanted), history of pregnancy loss (yes/no) and history of partner abuse (yes/no).

Analysis

Because so little is known about prenatal mental health, we generated descriptive statistics to describe the independent associations of the independent variable, maternal mood, with covariates. We examined the independent and interactive associations of maternal mood and substance use with preterm birth and LBW using multivariate logistic regression analyses. Multivariate models included interaction terms, variables that were independently associated with the outcome variables in chi-square analysis, and maternal age, parity and education as these variables are potential confounders of the associations studied and are considered standard adjustment factors for PTB and LBW analyses. CDC provides each state with a final data file with analysis weights. Weighting was consistent across years and allowed for the combination of multiple data years for analysis. We used SAS 9.1.3 PROC SURVEYFREQ, PROC SURVEYMEANS and PROCSURVEY LOGISTIC to analyze the weighted data.

Results

Non-respondents were more likely than respondents to be unmarried, younger, multiparous and have lower education levels; they also had higher rates of LBW births (9.6 vs. 6.4%) and PTB (11.1 vs. 8.6%) than respondents (data not shown).

Among respondents, measures of maternal age, marital status, education level, income level, pregnancy-related characteristics, maternal mood and stressful life events were consistent over the study years. Women reported rates of low mood during pregnancy that ranged from 22.0–24.7% over the survey years (data not shown). Women who reported low mood levels were more likely than those with high mood levels to be unmarried, have a history of pregnancy loss, have a history of partner abuse, and have had prior pregnancies; they were less likely to have obtained prenatal care as early as they would have liked. Substance use appeared to co-occur with low maternal mood: women who reported low mood levels were more likely than those with high mood to use tobacco during the third trimester (26.2 vs. 11.7%) and to drink alcohol during the third trimester (7.6 vs. 5.8%) (Table 1).

Table 1 Weighted frequencies: chi-square associations of maternal mood with substance use and selected demographic variables, Minnesota PRAMS, 2002–2006 (n = 7,475)
Full size table

The prevalence of tobacco use appeared stable over the study years. Third trimester alcohol use seemed to increase slightly over the study time frame, from 5.6% in 2002 to 7.2% in 2006 (data not shown). For all years combined, among the 6.2% of women who reported drinking in the third trimester, 59.6% reported drinking 7 or more drinks/week, 15.4% had 1–6 drinks per week and 25% had less than one drink per week. Less than 1% of women who reported alcohol use during the third trimester reported drinking more than 14 drinks/week.

Correlates of Low Birth Weight

The average weight of the study infants was 3,398 g and 5.7% percent of the study infants were LBW. Low maternal mood and prenatal tobacco use were independently associated with LBW, after adjustment for maternal age, parity and education. Women with low mood were almost twice as likely to have a LBW delivery compared to high mood women (OR = 1.93, 95% CI: 1.47, 2.55, P < 0.0001) (data not shown). Women who smoked during the third trimester were also more likely to have a LBW delivery as compared to non-smokers (OR = 1.55, 95% CI: 1.12, 2.15, P = 0.009) (data not shown). Maternal education and age were the only covariates significantly associated with LBW in adjusted models.

Low maternal mood and prenatal tobacco use were interactively associated with having a LBW infant (Wald chi-square for interaction: 4.93, 1 df, P = 0.03). Women with low mood levels who used tobacco during the third trimester were twice as likely to have a LBW infant compared to women with high mood levels and no third trimester tobacco use (Table 2). LBW risk was higher among women who reported low mood levels and no tobacco use (OR = 2.16, 95% CI: 1.60, 2.92, P < 0.0001) and high mood levels and tobacco use (OR = 1.92, 95% CI: 1.30, 2.84, P = 0.001), compared with women who reported no tobacco use and high mood levels. Maternal education, age and parity were also significantly associated with LBW in the multivariate model. Third trimester alcohol use was not independently associated with delivery of a LBW infant.

Table 2 Multivariate associations of low birthweight with maternal mood, tobacco use and selected covariates, Minnesota PRAMS (n = 7,475)
Full size table

Correlates of Preterm Delivery

The average gestational age at delivery for study infants was 38.8 weeks and 7.8% of the study infants were preterm. Low mood was independently associated with PTB, after adjustment for maternal age, parity and education (OR = 1.78, 95% CI: 1.42, 2.25, P < 0.0001) (data not shown). Third trimester tobacco use and alcohol use were not independently associated with delivery of a preterm infant (data not shown). We found a weak interactive effect for maternal mood, alcohol use and PTB (OR = 0.68, Wald chi-square for interaction: 2.27, 1 df, P = 0.13), after adjustment for maternal age, parity and education (Table 3). Women with low mood levels during pregnancy who abstained from alcohol were almost twice as likely to have a PTB (OR = 1.95, 95% CI: 1.54, 2.45, P < 0.0001), compared to women who reported high mood levels and did not drink. Women in the highest age category were 1.7 times more likely to have a PTB as compared to the 18–24 year group. More educated women were less likely to have PTB, as compared to women with less than a high school education.

Table 3 Multivariate associations of preterm birth with maternal mood, alcohol use and selected covariates, Minnesota PRAMS, 2002–2006 (n = 7,475)
Full size table

Discussion

Women who reported low mood levels during pregnancy were twice as likely to have a LBW infant, regardless of their smoking status, compared to women reporting higher mood levels. Women who had low mood and used tobacco had similar risks for LBW infants as women with low mood but no tobacco use, while the risk for LBW delivery for women reporting tobacco use and high mood was slightly lower than those for the other categories. Our findings suggest the relationship between tobacco use and maternal mood may be subtle and more complex than previous studies of these risk factors alone could demonstrate. Tobacco use may buffer the effects of lowered maternal mood. It is also possible that there are biological and behavioral responses to low maternal mood that influence the growth of the infant, regardless of tobacco use.

Our finding of an association between substance use and LBW is consistent with previous studies that have examined the independent effects of alcohol and tobacco on birth outcomes [7, 22, 25, 29]. A retrospective study nested in a cohort of 18,297 children reported a finding for smoking and LBW delivery similar to ours, after adjustment for maternal age, body mass index, parity, alcohol use, maternal education, ethnicity, income and gestational diabetes [29]. The magnitude of association for tobacco use and LBW in other studies was similar to the odds ratios generated from our study [22, 25, 29].

Women with low mood during pregnancy who abstained from alcohol during the third trimester were almost twice as likely as their high mood counterparts to deliver a preterm infant. Our borderline statistically significant interactive association between maternal mood, prenatal alcohol use and PTB suggests that mood may be more important than third trimester alcohol use in predicting PTB. Our results concur with Dayan et al. who found higher rates of spontaneous PTB among women with higher depression scores, as measured by the Edinburgh Postnatal Depression Scale with a cut-off of 14 for prenatal depression [7].

Studies of alcohol use and PTB are equivocal. Dew et al. [30] found no association between alcohol use and PTB, however there was an association for women who used alcohol and tobacco. Lundsberg et al. [31] reported protective effects of mild drinking on intrauterine growth restriction, but higher rates of preterm birth among women who drank during the seventh month. Dole et al. found higher rates of PTB among women reporting more than five drinks per week compared to a reference group of women who consumed five or less drinks per week. This fits with Goldenberg’s conclusion that while heavy drinking could be associated with the risk of PTB, mild or moderate alcohol use is typically not considered a strong risk factor [32]. Our study was unable to detect an association between intensity of alcohol use and PTB, however the majority of our population fit into low use categories.

Our measure of maternal mood is both a limitation and strength of this study. The study question asked women to reflect on pregnancy, without specifying a period of time during pregnancy. Prenatal depressive symptoms may be variable [33] and we do not know if women who experience mood variability during pregnancy retrospectively report the most positive or the most negative, or an average, of feelings when asked to recall a period that may be as long as nine months. Our measure also does not fit the criteria for diagnosed clinical depression or even those of depressive symptoms, limiting comparison to other studies. While this measure does not fit clinical criteria for depression, our prevalence of low maternal mood in this study is only slightly higher than the estimates of major and minor depressive symptoms during pregnancy from other studies. In addition, our prevalence remained fairly stable over the study years. The maternal mood component of the PRAMS survey may serve as a useful proxy for clinical depression. Additionally, it serves as a useful surveillance tool since it is only one question and may reduce respondent burden. Finally, while our measure of mood was simple, it allowed us to identify subsets of women who were at twice the risk for PTB and LBW infants.

The generalizability of our findings is affected by the response rates. PRAMS was designed to be representative of resident births in Minnesota. The fairly high overall response rate provides confidence in the study results, however respondents and non-respondents differed on the outcomes of interest, as well as demographically. Our non-respondents were more likely to have adverse birth outcomes, thus we were missing some of the events of interest. A further limitation relates to biases associated with the design and retrospective nature of PRAMS. Women with adverse birth outcomes may be more likely than those with healthy birth outcomes to recall or to exaggerate prenatal risk factors, including poor mood and substance use. The effect of such biases could exaggerate the associations between such exposures and birth outcomes. Conversely, misclassification bias could occur with respect to substance use and mental health. Underreporting of prenatal substance use has been documented [34]. The social expectation that pregnancy should be one of the happiest times of a woman’s life may also influence women to answer the maternal mood question more positively. Such misclassification of predictor variables, if systematically related to the outcomes of interest, could bias our associations to the null. The PRAMS design focuses on women with a stable address and telephone number, thus there is the potential for selection bias, which can affect estimates of exposure-outcome associations. However, a recent study by Nilsen et al. found that selection bias influenced prevalence estimates of exposures and outcomes in a large pregnancy cohort as compared to all live births in Norway, but not estimates of exposure-outcome associations [35]. The fairly high response rate also minimizes the potential effects of selection bias.

Our study raises the question of how substance use and maternal mood may co-occur and influence birth outcomes. As maternal mental health continues to garner public awareness and the attention of health professionals, it is important to devise strategies for assessing measures of maternal mood in populations. Ongoing surveillance tools are important for identifying potential trends and further identification of factors associated with low maternal mood. The co-occurrence of substance use and low maternal mood in our population indicates an opportunity for addressing both issues when developing interventions and screening strategies, as targeting either issue in isolation may not be as effective. The interactive relationship between tobacco use and maternal mood with the risk for LBW needs further exploration. Studies are also needed to determine how well non-clinical measures of depressive mood during pregnancy reflect true impairment. The PRAMS measure of maternal mood could influence treatment and interventions by alerting professionals to women who may not meet a clinical diagnosis of depression during pregnancy, but may need mental health intervention and be at risk for adverse birth outcomes.