Stats Homework

Original article Alcohol consumption amongﬁrst-time mothers and the risk of preterm birth: a cohort study Maria T.G. Dale PhD a,*, Leiv S. Bakketeig PhD a, Per Magnus PhD a,b aDepartment of Genes and Environment, Norwegian Institute of Public Health, Oslo, NorwaybDepartment of Medicine, University of Oslo, Oslo, Norway article info Article history:

Received 10 April 2015 Accepted 31 August 2015 Available online 30 September 2015 Keywords:

Cohort study Alcohol consumption Preterm delivery Pregnancy Healthy drinker effect abstract Purpose:Our aim was to explore the association between alcohol consumption, before and during pregnancy, and the risk of preterm birth among 46,252 primiparous mothers.

Methods:We obtained information on alcohol consumption from questionnaire responses at pregnancy week 15 from the prospective, observational Norwegian Mother and Child Cohort Study. Data on preterm birth, categorized as delivery before gestation week 37, were retrieved from the Medical Birth Registry of Norway.

Results:Among the participants, 91% consumed alcohol before pregnancy and fewer than 20% reported consuming alcohol during pregnancy. The adjusted odds ratio (aOR) for preterm birth associated with prepregnancy alcohol consumption was 0.81 (95% conﬁdence interval [CI], 0.70e0.95). We did notﬁnd a risk reduction for overall drinking during pregnancy, aOR¼1.03 (95% CI, 0.90e1.19). However, dose- response analyses showed tendencies toward adverse effects when drinking 1e3 times per month during theﬁrst 15 weeks of pregnancy, aOR¼1.51 (95% CI, 1.14e2.00).

Conclusions:We did notﬁnd any effects of alcohol consumption during pregnancy, whereas pre- pregnancy drinking was associated with reduced risk of preterm birth. Residual confounding may have inﬂuenced the risk estimates, especially before pregnancy, as nondrinkers have lower socioeco- nomic status and well-being than drinkers.

2016 Elsevier Inc. All rights reserved. Introduction Most public health and medical authorities advise total absti- nence of alcohol both during pregnancy and when contemplating pregnancy[1,2]. This advice is an invocation of the precautionary principle since existing evidence is inconclusive as to whether there is a threshold dose below which alcohol intake is safe. Overall, there are few studies that indicate adverse effects of low levels of alcohol intake[3,4], whereas there is broad consensus in the literature that abusive and heavy drinking has a teratogenic effect on the fetus, such as the risk of fetal alcohol syndrome, growth restriction, birth defects, and neurodevelopmental problems[5e7].

A concern has been that alcohol intake may increase the risk of preterm birth. Preterm birth is among the leading causes of child mortality, to which 35% of neonatal deaths on a global basis areestimated to be attributable[8,9]. In 2013, about 5.8% of all births in Norway were preterm, as deﬁned by the World Health Organization as delivery before 37 weeks (259 days) of gestation. This rate is similar to the rest of Europe and other developed countries, where 5%e9% of all births were preterm[10,11]. From the 1990s, the chances of survival have increased because of better medical intensive care with advances such as antenatal corticosteroids, assisted ventilation, and administration of surfactant to prevent lungs from collapsing[12]. Almost in parallel with these improve- ments, preterm birth rates appear to increase[13]. Overall, from 1990 to 2010, there has been a 19.4% increase in the preterm birth rate in high-income countries (from 7.2% to 8.6%). The United States has a particularly high rate (12%)[8]. Better knowledge of causes and mechanisms is needed to prevent preterm birth.

Several studies on the subject of alcohol and preterm birth exist, but theﬁndings are inconsistent. Apparently, some studies suggest that the relation between alcohol and preterm birth is J-shaped or U-shaped. A study using a Danish pregnancy cohort suggested that an intake of more than seven drinks of alcohol per week increased the risk of preterm birth[14]. Another study showed that the risk * Corresponding author. Department of Genes and Environment, Division of Epidemiology, Norwegian Institute of Public Health, Box 4404, Nydalen, Oslo 0403, Norway. Tel.:þ47-21078106; fax:þ47-21078252.

E-mail address:[email protected](M.T.G. Dale). Contents lists available atScienceDirect Annals of Epidemiology journal homepage: www.annalsofepidemiology.org http://dx.doi.org/10.1016/j.annepidem.2015.08.013 1047-2797/ 2016 Elsevier Inc. All rights reserved. Annals of Epidemiology 26 (2016) 275e282 increased when more than 10 drinks were consumed per week[15].

However, both studies reported that lower levels of consumption might provide a small advantage as the risk of preterm birth was signiﬁcantly decreased compared to nondrinkers (relative risks of 0.7e0.9) with an intake of alcohol below four drinks per week.

Decreased risk at low to moderate alcohol intake was also found in a cross-sectional study by Wright et al. (1998)[16], who reported a relative risk of 0.4 (95% conﬁdence interval [CI], 0.2e0.7) for pre- term birth associated with alcohol intake in the third trimester.

Looking only at induced preterm birth, Meis et al. (1997)[17]found that alcohol intake reported in gestational week 24 was associated with reduced risk (relative risk 0.34, 95% CI, 0.15e0.76). A very recent prospective cohort study by Lundsberg et al. (2015)[18] support these earlierﬁndings, reporting reduced risk of preterm birth with low to moderate alcohol intake both during early and late pregnancy (relative risk of 0.79 and 0.60).

Other studies suggest that daily drinking has an adverse effect on preterm delivery. Jaddoe et al. (2007)[19]found that an average consumption of one drink per day either during early or late pregnancy had a relative risk of 2.5 (95% CI, 0.9e6.8) for preterm birth. Another prospective study showed a dose-response risk for low and moderate intake ( 0.10 oz or>0.10e0.25 oz of absolute alcohol per day) during late pregnancy and risk of preterm birth (relative risks of 2.88 and 2.96)[20]. Both studies indicate that daily alcohol intake has a risk-increasing effect and that greater the exposure the greater the risk, but no studies have found adverse effects of occasional drinking. A recent systematic review and meta- analysis by Patra et al. (2011)[4]sum up the aforementioned ﬁndings: only heavy consumption, 1.5 drinks per day and more, increased the risk of preterm birth, whereas lower levels were associated with reduced risk or no risk.

These inconsistentﬁndings of risk at relatively high consump- tion levels and no risk and/or reduced risk at moderate and low levels may be due to the heterogeneity across studies, large diver- gence in deﬁning and measuring alcohol consumption, and that someﬁndings are not adjusted for important lifestyle and related socioeconomic factors[4,19e21]. Another source of confounding could be undetected differences in health-related behavior due to previous reproductive experience[21]or due to the presence of publication bias[3].

Most pregnant women in Norway either abstain from alcohol or drink occasionally. The proportion of women who are occasional drinkers may be relatively large, a recent study in Norway indicates that one in 10 women maintain light alcohol consumption during pregnancy[22]. A more solid scientiﬁc basis for advice to these women is needed and should be provided by epidemiologic cohort studies rather than studies focusing only on women with high alcohol consumption[23]. Our objective is to examine critically, in a large prospective pregnancy cohort, the effect of alcohol before and during pregnancy on preterm delivery. Studies suggesting that low levels of alcohol intake have a protective effect have been the subject of controversydis there a biological explanation for this association or is the outcome attributed to maternal lifestyle and background variables? A major problem in this area of research is confounding by background factors such as socioeconomic status and lifestyle. We had the opportunity to control for prepregnancy drinking, as well as the drinking pattern of the spouse. Together with educational level, these two variables reveal a major part of the family’s lifestyle that might have confounding effects, which allows us to identify better the alcohol-speciﬁc effects on preg- nancy. This present study provides dose-response information in the drinkers versus nondrinkers analyses, making it possible to explore if low amounts of alcohol relate to preterm risk. As alcohol intake was assessed in theﬁrst trimester, prospective to the birth outcome, we avoid recall bias and under-reporting among womenwith adverse pregnancy outcomes. By including only singletons among primiparous births, we isolated alcohol exposure from other exposures that potentially increase the risk of preterm birth, and we captured uncontrolled confounding by previous reproductive experiences or fetal complications.

Materials and methods Study population and data collection This study is a subproject of the Norwegian Mother and Child Cohort Study (MoBa) conducted by the Norwegian Institute of Public Health[24]. MoBa is a prospective pregnancy cohort comprising all pregnant women in Norway attending routine ultrasound examinations at gestational weeks 17e18. Recruitment started in 1999 and ended in 2008, and participants were recruited to the study through a postal invitation after they have signed up for the routine ultrasound examination in their local hospital. Of the total, 40% of the invited women consented to participate, and the cohort now includes 112,768 pregnancies[25].

Questionnaires used in MoBa can be found online athttp:// www.fhi.no/moba-en. The record of the pregnancy and delivery in the Medical Birth Registry of Norway (MBRN) is included in the MoBa database. The information in MBRN is based on antenatal forms and data recorded at the maternity departments at delivery and during the hospital stay[26]. The present study was based on version 7 of the quality-assured dataﬁles made available in 2013.

Written informed consent was obtained from all participating women, and the study has been approved by the Regional Com- mittee for Ethics in Medical Research and the Data Inspectorate.

Variables We deﬁned preterm delivery as delivery before 37 weeks (259 days) of gestation. Predictions from ultrasound measures determined the outcome variable gestational age, or if this was missing, from the date of the last menstrual period. Information on maternal age at child birth was categorized as“less than 20 years,” “20e24 years,”“25e29 years,”“30e34 years,”and“35 years or more.”The variables gestational age, maternal age, and child’ssex were drawn from the MBRN.

Questionnaire Q1, with assessment point at gestation week 15, asks about alcohol use during the last 3 months before pregnancy and during pregnancy. The woman was asked how often she drinks alcohol with response categories“6e 7 times,”“4e5 times,” “2e3 times,”“1 time per week,”“2e3 times per month,”“less than once per month,”or“never.”There was also a question about how many units (for instance a glass of beer or a glass of wine) the mothers usually consumed on each occasion.Table 1gives the distribution of these variables. If the woman reported to be a nondrinker before pregnancy and did not respond to the preg- nancy intake question, we assumed that she was also a nondrinker during pregnancy.

The response categories on maternal smoking were “nonsmoker,”“occasional smoker,”and“daily smoker.”Maternal educational attainment was categorized as“less than 12 years,” “12 years,”“13e16 years,”and“17 years or more.”The smoking variable and the information on maternal education were taken from Q1 together with variables assessing prepregnancy maternal height and weight and maternal anxiety. Prepregnancy height and body weight were used to calculate body mass index (kg/m 2) which was categorized as“less than 20,”“20e24,”“25e29,”and“30 or more.”The mothers anxiety scores range from 2 to 8 and was a sum of responses“not bothered¼1,”“a little bothered¼2,”“quite bothered¼3,”or“very bothered¼4”on two questions on whether M.T.G. Dale et al. / Annals of Epidemiology 26 (2016) 275e282 276 the woman during the past two weeks have been (1) frightened or anxious or (2) nervous and in turmoil. All questionnaire variables also include a missing category.

For the background variables (Tables 2 and 3), we included additional variables from Q1. First, social support was measured by the item“Do you have anyone other than your husband/partner that you can ask for support in a difﬁcult situation?”and was dichotomized as“no”and“yes.”Second, maternal distress was measured by the SCL-5, a shortened version of the Hopkins Symptom Checklist shown to correlate strongly with the SCL-25 index[27]. SCL-5 hasﬁve items, with four response categories from“1¼not bothered”to“4¼very bothered.”A typical item was “Worrying too much about things.”We scored the index as the mean of the item scores, where an average score less than 2 is considered within the normal range on the SCL-5. Third, well- being comprisesﬁve items rated on a 7-point Likert scale ranging from“1¼strongly disagree”to“7¼strongly agree.”The items were phrased as follows:“in most ways my life is close to my ideal,”“the conditions of my life are excellent,”“I am satisﬁed with my life,”“so far I have gotten the important things I want in life,”and“if I could live my life over, I would change almost nothing.”Fourth, the mother was asked if she did exercise“once a week or more”before pregnancy, and we dichotomized the re- sponses as“no”and“yes”. Fifth, we asked about folate intake and dichotomized the responses as“no”and“yes.”Sixth, if the woman was previously treated for infertility, we dichotomized the re- sponses as“no”and“yes.”Finally, paternal alcohol use during thelast 6 months before pregnancy and after pregnancy was asked for, and the categories were:“6e7 times,”“4e5 times,”“2e3 times,” “1 time per week,”“2e3 times per month,”“less than once per month,”or“never.”We dichotomized the responses as“Non- drinkers”and“Drinkers.” Analytic strategy Alcohol intake in pregnancy reported in Q1 was relatively infrequent, and the number of units consumed on each occasion was small (seeTable 1). Considering the low number of heavy drinkers in pregnancy, we chose to create a simple alcohol variable (drinker or nondrinker) for two of the time periods given inTable 1 for the further analyses. In the analyses, we excluded records with missing values for the main exposure (alcohol intake in pregnancy) but included a missing category for the potentially confounding variables.

Multivariable logistic regression analyses were used to approx- imate crude and adjusted relative risks of preterm birth according to alcohol intake.

According to World Health Organization’sInternational Classiﬁ- cation of Diseases, Tenth Revision, the perinatal period commences after gestation week 22. As a result, earlier births were excluded in this study[28]. The number of pregnancies registered in the MBRN that had gestational length above 22 weeks and shorter than 44 weeks was 107,978. Furthermore, we wanted to isolate alcohol exposure from other exposures reported to increase the risk of Table 1 Proportions of women who drank alcohol during the 3 months before pregnancy and during pregnancy, as reported in week 15 (Q1) Alcohol consumption Frequency Alcohol consumption Amount (units) per occasion * N¼44,300N¼44,300N¼40,515N¼4,121 Before pregnancy,n(%) Week 0e15,n(%) Before pregnancy,n(%) Week 0e15,n(%) Never 3040 (7.7) 35851 (80.9) <1 per mo 10163 (22.9) 3280 (7.4)<1 1492 (3.7) 3111 (75.5) 1e3 per mo 16573 (37.4) 674 (1.5) 1e2 12637 (31.2) 624 (15.1) 1 per wk 8917 (20.1) 131 (0.3) 3e4 13292 (32.8) 51 (1.2) 2e3 per wk 4293 (9.7) 24 (0.1) 5e6 8444 (20.8) 34 (0.8) 4e5 per wk 463 (1.0) 4 (0.0) 7e9 3206 (7.9) 9 (0.2) 6e7 per wk 106 (0.2) 8 (0.0) 10þ782 (1.9) 6 (0.1) No response 381 (0.9) 4328 (9.8) No response 662 (1.6) 286 (6.9) *Subjects who responded“never”and nonresponders to the frequency question are not included. Table 2 Characteristics of women according to prepregnant intake of alcohol Background variables Nondrinker *(n¼3404) Drinker *(n¼40,515) c2/FP y Mother Gestational week (mean SD) 39.3 2.1 39.5 2.0 6.9 .001 Age when child is born (mean SD) 27.5 4.6 28.4 4.4 69.3<.0001 Education>12 y (%) 60.8 71.7 248.3<.0001 Smoking during pregnancy, yes (%) 5.0 8.4 47.7<.0001 Prepregnancy BMI (mean SD) 23.8 4.6 23.7 4.3 0.8 .469 Folate, yes (%)72.4 80.2 116.4<.0001 Infertility, yes (%) 12.2 8.0 87.0<.0001 Social support, yes (%) 91.4 97.8 476.3<.0001 Exercise before pregnancy once a week (%) 75.2 83.7 124.6<.0001 Psychological distress gestation week 18, raw score from 1 to 4 (mean SD) 1.3 0.4 1.3 0.4 6.7 .001 Well-being gestation week 18 raw score from 1e7 (mean SD) 5.6 1.2 5.7 1.1 4.3<.014 Child Boys (%)51.4 51.3 0.9 .472 Birth weight, g (mean SD) 3597.4 597.3 3592.1 580.6 0.8 .378 Father Drinking before pregnancy, yes (%) 99.7 99.4 3.1 .085 ANOVA¼analysis of variance.

*Nondrinkers are women who reported being a nondrinker. Drinkers are women who responded to the frequency categories varying from“less than 1 per month”up to “6e7 times per week”and/or reported drinking“less than 1”up to“10 or more”units of alcohol on each occasion.

yPvalues were calculated using ANOVA for continuous variables and c2test for categorical variables. M.T.G. Dale et al. / Annals of Epidemiology 26 (2016) 275e282277 preterm birth. Therefore, only cases deﬁned as singleton births were eligible for this study (n¼104,225). To capture uncontrolled confounding by previous reproductive experiences, fetal compli- cations and/or change of behavior after adverse pregnancy out- comes or opposite, we excluded multiparous women (57,973). The ﬁnal sample consisted of a total of 46,252 primiparous mothers with singleton births (seeFig. 1).

Sensitivity analysis To be sure that exclusion of earlier preterm births (<22 week) did not induce selection bias, we included them in a sensitivity analysis to investigate if these women were drinkers either before pregnancy (T1) or during pregnancy (T2), and if that could poten- tially inﬂuence our risk estimates in the primary analysis.

Furthermore, to investigate the potential impact of the nonre- sponse of mothers on alcohol intake during pregnancy, we per- formed another sensitivity analysis. We repeated the primary analysis before (T1) and during pregnancy (T2) by recoding the nonresponse of mothers asﬁrst being drinkers and later asnondrinkers. By doing this, we could separately estimate the po- tential largest deviations: the maximum risk effect and the maximum protective effect of alcohol consumption if the missing responses were either possible drinkers or possible nondrinkers.

Covariates Factors that could potentially inﬂuence maternal alcohol use and risk of preterm delivery were identiﬁed as confounders. The following covariates were included: Maternal age at delivery, years of education, pre-pregnancy body mass index, smoking, exercise before pregnancy, and paternal drinking before pregnancy.

The analyses were computed with PASW software, version 22.0, (SPSS Inc., Chicago, IL).

Results Characteristics of the study population Alcohol consumption before and during pregnancy was measured by a general questionnaire sent out in gestational week 15, Q1, with a response rate of 94.9%. In this version, the total sample sizes were 101,769 for Q1 and 108,327 for the MBRN, but after excluding multiparous women, plural births, and births before gestation week 22, and after gestation week 43, it was reduced to 46,252 (seeﬂow chart concerning excluded cases for details).

Nondrinkers and drinkers differed on several background vari- ables, dependent on the two measurement times. Women drinking before pregnancy were more likely to be older, smokers, have higher education, take folic acid during pregnancy, experienced more social support, exercised more before pregnancy, and they reported less treatment due to infertility when compared with nondrinkers before pregnancy (seeTable 2).

Women drinking during pregnancy shared the same background variables as nondrinking women during pregnancy when it comes to age, smoking, and education. On the other hand, it was less likely that they reported taking folic acid supplements, and they experi- enced lower well-being, and they reported more treatment due to infertility when compared to nondrinkers during pregnancy. They also experienced more psychological distress had lower body mass index, and their partners were more likely to drink when compared to nondrinkers during pregnancy (seeTa bl e 3).

Table 3 Characteristics of women according to intake of alcohol during pregnancy Background variables Nondrinker *(n¼35,851) Drinker *(n¼4121) c2/FP y Mother Gestational week (mean SD) 39.5 2.0 39.5 2.0 0.6 .558 Age when child is born (mean SD) 28.2 4.4 30.1 4.4 367.1<.0001 Education>12 y (%) 71.1 77.6 226.9<.0001 Drinking before pregnancy, yes (%) Smoking during pregnancy, yes (%) 7.5 12.0 100.7<.0001 Prepregnancy BMI (mean SD) 23.7 4.4 23.4 3.8 17.3<.0001 Infertility, yes (%) 8.5 6.6 43.5<.0001 Folate, yes (%)80.1 76.7 25.9<.0001 Social support, yes (%) 97.4 97.9 4.1 .049 Exercise before pregnancy once a week (%) 83.1 83.3 0.1 .714 Psychological distress gestation week 18, raw score from 1 to 4 (mean SD) 1.2 0.4 1.3 0.4 51.7<.0001 Well-being gestation week 18 raw score from 1e7 (mean SD) 5.7 1.1 5.6 1.1 50.0<.0001 Child Boys (%)51.2 51.2 0.4 .263 Birth weight, g (mean SD.) 3592.9 580.8 3598.1 57464 0.8 .376 Father Drinking before pregnancy, yes (%) 99.5 98.7 25.8<.0001 Drinking during pregnancy, yes (%) 95.6 99.4 96.1<.0001 *Nondrinkers are women who reported being a nondrinker. Drinkers are women who responded to the frequency categories varying from“less than 1 per month”up to“6- 7 times per week”and/or reported drinking“less than 1”up to“10 or more”units of alcohol on each occasion.

yPvalues were calculated using ANOVA for continuous variables and c2test for categorical variables. Fig. 1.Flow diagram displaying exclusion criteria for theﬁnal study population.

*GW¼Gestational weeks. **Variables from MBRN.M.T.G. Dale et al. / Annals of Epidemiology 26 (2016) 275e282 278 Effect of alcohol and risk of preterm birth In this dataset, the prevalence of preterm birth was 5.9% (2729 of 46252). The median length of gestation was 40 weeks for both drinkers during pregnancy and nondrinkers, whereas the mean gestational length was 39.47 weeks (SD¼2.03) for both groups. The mean pregnancy length for prepregnancy drinkers was 39.48 (SD¼ 2.02), whereas it was 39.34 (SD¼2.12) for nondrinkers (F¼13.4, 1 degrees of freedom,P<.001). The mean pregnancy length for drinkers during pregnancy was 39.50 (SD¼2.02) and 39.47 (SD¼ 2.03) for nondrinkers (F¼1.2, 1 degrees of freedom,P¼.281).

Overall, alcohol consumption was not associated with preterm birth risk.Table 4lists a statistically signiﬁcant tendency toward lower risk of preterm birth of prepregnancy drinking (adjusted odds ratio [aOR]¼0.81, 95% CI¼0.70e0.95). When we analyzed the dose-response relationship, we found that the risk decreased when the average frequency in consumption increased. A similar association was not found among women who reported drinking during early pregnancy aOR¼1.03 (95% CI¼0.90e1.19).

Table 5shows the same tendency: when women who reported being drinking either before pregnancy or during theﬁrst 15 gestation weeks were analyzed together with subjects who re- ported drinking in both periods, the risk of preterm birth was signiﬁcantly reduced for women who reported drinking beforepregnancy. But not after pregnancy, with nondrinkers as the reference (aOR¼0.80, 95% CI¼0.68e0.94).

Potential effect of very early preterm birth and nonresponse on alcohol intake In this data set (after exclusion of multiple births and parity), 93 children were born before gestation week 22 and eligible to be included in the sensitivity analysis. These extra cases increased the total number of preterm births from 2474 to 2567 at T1 and from 2346 to 2439 at T2. When we explored the potential impact of nonresponse mothers, the total number of drinkers increased from 40,515 to 40,896 and from 3404 to 3785 for nondrinkers at T1. At T2, the total number of drinkers increased from 4121 to 8449 and from 35,851 to 40,179 for nondrinkers. AsTable 6shows, ourﬁndings in the main analysis remained the same after exploring the potential impact of very early preterm birth and the nonresponse of mothers.

When we included very early preterm births in the analysis, the estimates were in line with our previous results both before preg- nancy (aOR¼0.84, 95% CI¼0.73e0.96) and during pregnancy (aOR¼1.03, 95% CI¼0.90e1.18). When we included mothers with nonresponse in the analysis and recoded them as being drinkers, the odds for preterm birth were unchanged with no risk for prepregnancy drinking (aOR¼0.82, 95% Table 4 Number and prevalence (percent) of preterm births (<37 weeks), and crude odds ratios and aORs according to self-reported alcohol intake before pregnancy (T1) and theﬁrst 15e17 weeks of pregnancy (T2) for 46,252 pregnancies in the Norwegian Mother and Child Cohort Study First-time mothers and preterm birth (T1 alcohol before pregnancy) Exposure N of births N of cases Percent cOR aOR * Nondrinker T1 y 3404 236 6.93 1.0 1.0 Drinker T1 y 40,515 2338 5.77 0.82 (0.72e0.94) z 0.81 (0.70e0.95) z <1 per mo 10,163 629 6.19 0.89 (0.76e1.03) 0.85 (0.71e1.01) 1e3 per mo 16,573 989 5.97 0.85 (0.74e0.99) z 0.85 (0.72e1.00) x 1 per wk 8917 486 5.45 0.77 (0.66e0.91) z 0.76 (0.64e0.92) z 2e3 per wk 4293 210 4.89 0.69 (0.57e0.84) z 0.66 (0.53e0.83) z 4e5 per wk 463 19 4.10 0.57 (0.36e0.93) z 0.58 (0.34e0.99) z 6e7 per wk 106 5 4.71 0.67 (0.27e1.65) 0.61 (0.22e1.71) No response 381 24 6.23 0.90 (0.52e1.39) 1.45 (0.80e2.63) First-time mothers and preterm birth (T2 alcohol duringﬁrst 15 wk) Exposure N of births N of cases Percent cOR aOR k Nondrinker T2 y 35,851 2104 5.87 1.0 1.0 Drinker T2 y 4121 242 5.87 1.00 (0.87e1.15) 1.03 (0.90e1.19) <1 per mo 3280 180 5.49 0.93 (0.80e1.09) 0.97 (0.82e1.13) 1e3 per mo 674 55 8.16 1.43 (1.08e1.89) z 1.51 (1.14e2.00) z 1 per wk 167 7 3.82 0.70 (0.33e1.50) 0.71 (0.33e1.52) No response 4328 252 6.18 0.99 (0.87e1.14) 0.99 (0.86e1.14) *Adjusted for maternal age, educational attainment, smoking before pregnancy, prepregnancy BMI, exercise before pregnancy and paternal drinking before pregnancy.yNondrinkers are women who reported being a nondrinker. Drinkers are women who responded to the frequency categories varying from“less than 1 per month”up to “6e7 times per week”and/or reported drinking“less than 1”up to“10 or more”units of alcohol on each occasion.

zP<.05.xBorderline statistically signiﬁcant.kAdjusted for drinking before pregnancy, maternal age, educational attainment, smoking before pregnancy, pre-pregnancy BMI and paternal drinkingbefore pregnancy. Table 5 Prevalence of preterm birth and crude (cOR) and adjusted (aOR) odds ratios (with 95% conﬁdence interval (CI)) according to alcohol intake before pregnancy and duringﬁrst 15 weeks of gestation in 46, 252 pregnancies from the Norwegian Mother and Child Cohort Study First-time mothers and preterm birth (T1 and T2) Drinking before pregnancy * Drinking 0e15 wk * No of births No of cases Prevalence cOR aOR y No No 3160 235 4.18 1.0 1.0 Yes No 30,523 1862 4.14 0.82 (0.71e0.94) z 0.80 (0.68e0.94) z Yes Yes 3857 239 3.63 0.83 (0.69e1.00) x 0.83 (0.66e1.03) *Women who report being drinking either before pregnancy or during theﬁrst 15 gestation weeks were analyzed together with subjects who report drinking in both periods.

yAdjusted for maternal age, educational attainment, smoking before pregnancy, prepregnancy BMI, and paternal drinking before pregnancy.zP<.05.xBorderline statistically signiﬁcant. M.T.G. Dale et al. / Annals of Epidemiology 26 (2016) 275e282279 CI¼0.70e0.97) and drinking during pregnancy (aOR¼0.95, 95% CI¼0.85e1.07). The estimates remained stable showing no risk of preterm birth associated with alcohol consumption when we reco- ded nonresponse as nondrinkers at T1 (aOR¼0.80, 95% CI¼ 0.69e0.94) and T2 (aOR¼1.01, 95% CI¼0.86e1.17).

Discussion The message from this study is somewhat complex. Ourﬁnding that alcohol consumption during the last 15 weeks before preg- nancy was associated with reduced risk of preterm birth, whereas drinking during pregnancy was not is intuitively difﬁcult to make sense of.

Notably, we did not observe an overall effect on preterm birth among alcohol drinkers during pregnancy. Instead, the dose- response analyses showed tendencies toward adverse effects when drinking 1e3 times per month during theﬁrst trimester. This ﬁnding is not in accordance with the two Danish prospective studies of alcohol exposure during pregnancy and preterm birth that report a J-shaped relationship between the dose of alcohol and the risk of preterm birth, indicating that moderate drinking during pregnancy is protective[11,12]. An important contrast to these aforementioned studies is that the participants of MoBa report very low alcohol consumption. We had no possibility of studying the effects of alcohol intake above four drinks per week.

Another essential difference between this study and the two Danish studies is that the risk reduction was observed for drinking before pregnancy. Our results suggest a protective dose-response effect of prepregnancy alcohol consumption. Women who consumed alcohol 4e5 times per week had a lower risk (aOR¼ 0.58) than women drinking less than once per month (aOR¼0.85).

This reduced risk with prepregnancy drinking has previously been reported in a review by Patra et al. (2011)[4]. On the other hand, their review showed a dose-response relationship betweenpreterm birth and prenatal drinking, with no effect at an average of 1.5 drinks per day and monotonically increased risk with higher consumptions.

Cloner and Rezkalla (2007)[29]pose the question“To drink or not to drink”in their review of the beneﬁts of low to moderate consumption alcohol. One might suggest that a more relevant question is about the labeling“To be a drinker or a nondrinker?” rather than the alcohol consumption by itself. We assume that a possible and more intuitive explanation can be that differences in socioeconomic status, health-related behavior, and physical and psychological well-being between drinkers and nondrinkers inﬂu- ence the risk estimates.

Interestingly, alcohol consumption among women in northern Europe and the United States has increased during the last decades as alcohol has become more integrated into everyday life. The alcohol consumption in Norway has increased by 40 percent over the last two decades, with the highest growth among well- educated and high-income women [30]. In most western cultures, the use of alcohol is the norm, and abstinence is associated with social deviance[31,32]. A relevant Norwegian population-based study shows that people with low alcohol consumption are at higher risk of depression and anxiety compared with moderate drinkers, and the abstainers have the highest risk of experiencing these negative mental health consequences[32].

This tendency has previously been described as the“healthy drinker effect.”Healthy women tend to drink more than women with chronic diseases[15,18]. However, perhaps a broader term such as the“higher well-being effect”is more appropriate, as intake of alcohol appears to be associated with healthful lifestyle, higher education, and better social network.

In this study, we identiﬁed several signiﬁcant differences in background variables between drinkers and nondrinkers. Drinkers overall appear to have higher education, be more social and have a higher score on well-being. Not surprisingly, the differences are most Table 6 Sensitivity analyses with number and prevalence (percent) of preterm births (<37 weeks), and crude (cOR) and adjusted (aOR) odds ratios according to self-reported alcohol intake before pregnancy (T1) and theﬁrst 15e17 weeks of pregnancy (T2) including very preterm births (>22) and the nonresponse of mothers on alcohol intake in the main analysis First-time mothers and preterm birth (T1 alcohol before pregnancy) Exposure N of births N of cases Percent cOR aOR * Including very early preterm birth (<22,n¼88) Nondrinker T1 y 3411 243 7.12 1.0 1.0 Drinker T1 y 40,596 2419 5.96 0.83 (0.72e0.95) z 0.84 (0.73e0.96) z Including nonresponse as drinkers (n¼381) Nondrinker T1 3404 236 6.93 1.0 1.0 Drinker T1 40,896 2362 5.78 0.82 (0.72e0.95) z 0.82 (0.70e0.97) z Including nonresponse as nondrinkers (n¼381) Nondrinker T1 3785 260 6.87 1.0 1.0 Drinker T1 40,515 2338 5.77 0.83 (0.73e0.95) z 0.80 (0.69e0.94) z First-time mothers and preterm birth (T2 alcohol duringﬁrst 15 wk) ExposureN of births N of cases Percent cOR aOR x Including very early preterm birth (<22,n¼81) Nondrinker T2 35,919 2172 6.05 1.0 1.0 Drinker T2 4134 255 6.17 1.02 (0.89e1.17) 1.03 (0.90e1.18) Including nonresponse as drinkers (n¼4328) Nondrinker T1 35,851 2104 5.12 1.0 1.0 Drinker T1 8449 494 5.85 0.97 (0.90e1.10) 0.95 (0.85e1.07) z Including nonresponse as nondrinker (n¼4328) Nondrinker T1 40,179 2356 5.86 1.0 1.0 Drinker T1 4121 242 5.87 1.00 (0.87e1.15) 1.01 (0.86e1.17) z * Adjusted for maternal age, educational attainment, smoking before pregnancy, prepregnancy BMI, exercise before pregnancy and paternal drinking before pregnancy.yNondrinkers are women who reported being a nondrinker. Drinkers are women who responded to the frequency categories varying from“less than 1 per month”up to “6e7 times per week”and/or reported drinking“less than 1”up to“10 or more”units of alcohol on each occasion.

zP<.05.xAdjusted for drinking before pregnancy, maternal age, educational attainment, smoking before pregnancy, prepregnancy BMI and paternal drinking before pregnancy. M.T.G. Dale et al. / Annals of Epidemiology 26 (2016) 275e282 280 pronounced before pregnancy, since during pregnancy total alcohol abstinence is highly recommended, and therefore, the socioeco- nomic difference between drinkers and nondrinkers decreases at that measurement time. Consequently, it is very likely that studies like our study, measuring the association between the mothers’self- reported alcohol intake and preterm risk, also indirectly measures the impact of maternal social resources and well-being by assessing several factors that accompany both alcohol and the risk of preterm birth. The fact that adverse social circumstances are associated with the risk of preterm birth underscores the complexity of factors and mechanisms involved in understanding the association between alcohol consumption and potential adverse birth effects[9,33].In this sense, if alcohol consumption was the only signiﬁcant predictor and explanation for our results, the exposure should display the same risk estimates during early pregnancy, and not only be asso- ciated with reduced preterm risk when measured before pregnancy.

If there is a biological explanation for this association, the mechanisms are uncertain. One possible explanation of the results is similar causal pathways as for the association between alcohol and cardiovascular disease. Studies have found protective effects of alcohol on the vascular endothelium, resulting in increased angio- genesis, and reduced risk of inﬂammation and infections[33].

However, as already mentioned, this protective effect does not pertain in our study when women report drinking during pregnancy.

Strengths of this study include a large population sample ofﬁrst- time mothers with only singleton births, with relevant data on a number of covariates and background variables that permitted adjustment for important confounders. Alcohol exposure was assessed in theﬁrst trimester prospective to the birth outcome, avoiding potential systematic errors due to under-reporting and recalls bias. Our relatively large cohort of mothers should enable identiﬁcation of variations in maternal attributes, and a study of recruitment bias in MoBa by Nilsen et al. (2009)[34]concludes that exposure-outcome associations are not biased.

However, this cohort study suffers from several limitations that might have affected the validity of ourﬁndings. As with other questionnaire studies, some caution is required when interpreting self-reported measures. Mothers’responses may be affected by social desirability and other response biases. Women who believe that they should not drink any alcohol during pregnancy may under-report alcohol intake, and cause a potential source of error by misclassiﬁcation toward inclusion of drinkers in the nondrinker category. Such misclassiﬁcation may lead to a bias toward the null value in the estimate of the effects of alcohol consumption during pregnancy on preterm birth.

Moreover, drinking before pregnancy were assessed retrospec- tively and is potentially subject to recall bias and under-reporting especially if mothers discover any complications during early pregnancy. For this purpose, we performed a sensitivity analysis on the potential impact of the nonresponding mothers and mothers with births before gestation week 22, to see if these groups could induce selection bias. However, the effect estimates in the cohort analyses remained unchanged. If under-reporting were present in the self-reported alcohol consumption, the effect estimates should have been underestimated.

A relevant substudy of the MoBa cohort explored the risk of underestimates with variables linked to social desirability. They found that self-reported smoking status during pregnancy had a sensitivity of 82% and a speciﬁcity of 99% against plasma cotinine concentrations, signifying that self-reported smoking is a valid measure in the MoBa cohort[35]. This could suggest that reporting bias is not the explanation for our results, especially before preg- nancy, as there is no established stigma attached to prepregnancy drinking in Norway and most of mothers in the cohort (91%) report to drink alcohol.Another possible source of error is that the timing of conception and alcohol consumption may overlap. Other studies have shown that when women are asked about their alcohol consumption before pregnancy, they might report consumption before preg- nancy recognition and therefore include early pregnancy[36,37]. An important limitation is that even if potentially confounding variables are included in the data set, residual confounding cannot be ruled out: The list of potential confounders is extensive, and it is impossible to control for all. Finally, two factors in our study raise an important question of the representativeness of the sample. First, the missing rate is high, and the women who did not participate might have a different exposure-outcome relation than those who did participate. Second, given that we only included primiparous women with singleton births, ourﬁndings are limited to represent women at low risk. Hence, our conclusions might not provide any guidance to other pregnant women. Consequently, we suggest that further studies should explore the effects of alcohol consumption among different groups of women.

Conclusions We cannot conclude from our study that alcohol intake is of beneﬁt to the fetus. More research is needed to understand the causal pathways. The health of the fetus may be affected in many complex ways by alcohol, and any beneﬁcial effect on reduced risk of preterm birth may as well be outweighed by other risks. This question will be investigated further in MoBa by looking at neu- rodevelopmental outcomes. Furthermore, the protective effect of prepregnancy alcohol consumption, also seen in other studies, can reﬂect a biological mechanism where geneeenvironment in- teractions have a part to play and that may lead to better prevention of preterm birth after more detailed clinical and experimental studies have explored the possible underlying mechanisms.

As this effect was not present during pregnancy, we further suggest that any direct positive effects of prepregnancy drinking can be explained by differences in socioeconomic background variables that we have not been able to control for. More impor- tantly, our data suggest that women drinking during pregnancy are not at lower risk of having a preterm birth. Instead, there might be a slight inverse risk that warrants the existing recommendations against any maternal alcohol intake during pregnancy.

Acknowledgments All authors have contributed to the conception and design of the study, revising the article critically for important intellectual con- tent and given theﬁnal approval of this version to be submitted.

The study is supported by The Norwegian Ministry of Health: NIH/ NIEHS (grant no. N01-ESe85433), NIH/NINDS (grant no. 1 UO1 NS 047537) and the Norwegian Research Council/FUGE (grant no.

151918/S10). Maria T.G. Dale has received funding from an unre- stricted grant from Oak Foundation, Geneva, Switzerland. The study sponsors were not involved in the study design; in the collection, analysis, or interpretation of data; in the writing of the report; or in the decision to submit the article for publication.

References [1] Report No. IS-1284 Alcohol and pregnancy, report from an expert group. Oslo, Norway: Norwegian Health Directorate; 2005https://helsedirektoratet.

no/Lists/Publikasjoner/Attachments/24/Alkohol-og-graviditet-hva-er-farlig-for- fosteret-hvordan-forebygge-og-behandle-IS-1284.pdf[accessed 09.04.2015].

[2] International guidelines on drinking during pregnancy. Washington, D.C., USA: Washington, D.C., USA: International Center for Alcohol Policies; 2009 http://www.icap.org/Table/InternationalGuidelinesOnDrinkingAndPregnancy [accessed 09.04.2015]. M.T.G. Dale et al. / Annals of Epidemiology 26 (2016) 275e282281 [3]Henderson J, Gray R, Brocklehurst P. Systematic review of effects of low- moderate prenatal alcohol exposure on pregnancy outcome. BJOG 2007;114(3):243e52.

[4]Patra J, Bakker R, Irving H, Jaddoe VW, Malini S, Rehm J. Doseeresponse relationship between alcohol consumption before and during pregnancy and the risks of low birthweight, preterm birth and small for gestational age (SGA)da systematic review and meta-analyses. BJOG 2011;118(12):

1411e21.

[5]Jones K, Smith D. Recognition of the fetal alcohol syndrome in early infancy.

Lancet 1973;302(7836):999e1001.

[6]Little RE, Wendt JK. The effects of maternal drinking in the reproductive period: An epidemiologic review. J Subst Abuse 1991;3(2):187e204.

[7]Ouellette EM, Rosett HL, Rosman NP, Weiner L. Adverse effects on offspring of maternal alcohol abuse during pregnancy. N Engl J Med 1977;297(10):528e30.

[8]Blencowe H, Cousens S, Oestergaard MZ, Chou D, Moller A-B, Narwal R, et al. National, regional, and worldwide estimates of preterm birth rates in the year 2010 with time trends since 1990 for selected countries: a systematic analysis and implications. Lancet 2012;379(9832):

2162e72.

[9]Hoffman HJ, Bakketeig LS. Risk factors associated with the occurrence of preterm birth. Clin Obstet Gynecol 1984;27(3):539e52.

[10]The prevention of perinatal mortality and morbidity. Report of a WHO Expert Committee. World Health Organ Tech Rep Ser 1970;457:1e60.

[11] Medical birth registry of Norway annual report. Oslo, Norway: Norwegian Institute of Public Health; 2014http://www.fhi.no/dokumenter/ba886ffc80.

pdf[accessed 09.04.2015].

[12]Saigal S, Doyle LW. An overview of mortality and sequelae of preterm birth from infancy to adulthood. Lancet 2008;371(9608):261e9.

[13]Goldenberg RL, Culhane JF, Iams JD, Romero R. Epidemiology and causes of preterm birth. Lancet 2008;371(9606):75e84.

[14]Albertsen K, Andersen A-MN, Olsen J, Grønbæk M. Alcohol consumption during pregnancy and the risk of preterm delivery. Am J Epidemiol 2004;159(2):155e61.

[15]Kesmodel U, Olsen SF, Secher NJ. Does alcohol increase the risk of preterm delivery? Epidemiology 2000;11(5):512e8.

[16]Wright SP, Mitchell EA, Thompson JM, Clements MS, Ford RP, Stewart AW.

Risk factors for preterm birth: a New Zealand study. N Z Med J 1998;111(1058):14e6.

[17]Meis PJ, Goldenberg RL, Mercer BM, Iams JD, Moawad AH, Miodovnik M, et al.

The preterm prediction study: risk factors for indicated preterm births. Am J Obstet Gynecol 1998;178(3):562e7.

[18]Lundsberg LS, Illuzzi JL, Belanger K, Triche EW, Bracken MB. Low-to-moderate prenatal alcohol consumption and the risk of selected birth outcomes:

a prospective cohort study. Ann Epidemiol 2015;25(1):46e54.

[19]Jaddoe VWV, Bakker R, Hofman A, Mackenbach JP, Moll HA, Steegers EAP, et al.

Moderate alcohol consumption during pregnancy and the risk of low birth weight and preterm birth. The Generation R Study. Ann Epidemiol 2007;17(10):834e40.

[20]Lundsberg LS, Bracken MB, Saftlas AF. Low-to-moderate gestational alcohol use and intrauterine growth retardation, low birthweight, and preterm de- livery. Ann Epidemiol 1997;7(7):498e508.[21]Andersen A-MN, Andersen PK, Olsen J, Grønbæk M, Strandberg-Larsen K.

Moderate alcohol intake during pregnancy and risk of fetal death. Int J Epi- demiol 2012;41(2):405e13.

[22]Stene-Larsen K, Torgersen L, Strandberg-Larsen K, Normann PT, Vollrath ME. Impact of maternal negative affectivity on light alcohol use and binge drinking during pregnancy. Acta Obstet Gynecol Scand 2013;92(12):1388e94.

[23]Caetano R, Ramisetty-Mikler S, Floyd LR, McGrath C. The epidemiology of drinking among women of child-bearing age. Alcohol Clin Exp Res 2006;30(6):1023e30.

[24]Magnus P, Irgens LM, Haug K, Nystad W, Skjærven R, Stoltenberg C, et al.

Cohort proﬁle: The Norwegian Mother and Child Cohort Study (MoBa). Int J Epidemiol 2006;35(5):1146e50.

[25]Schreuder P, Alsaker ER. The Norwegian Mother and Child Cohort Study (MoBa)eMoBa recruitment and logistics. Nor Epidemiol 2014;24(1-2):23e7.

[26]Irgens LM. The Medical Birth Registry of Norway. Epidemiological research and surveillance throughout 30 years. Acta Obstet Gynecol Scand 2000;79(6):435e9.

[27]Strand BH, Dalgard OS, Tambs K, Rognerud M. Measuring the mental health status of the Norwegian population: a comparison of the in- struments SCL-25, SCL-10, SCL-5 and MHI-5 (SF-36). Nord J Psychiatry 2003;57(2):113e8.

[28]Langhoff-Roos J, Kesmodel U, Jacobsson B, Rasmussen S, Vogel I. Spontaneous preterm delivery in primiparous women at low risk in Denmark: population based study. BMJ 2006;332(7547):937e9.

[29]Kloner RA, Rezkalla SH. To drink or not to drink? That is the question. Cir- culation 2007;116(11):1306e17.

[30] Annual report. Oslo, Norway: Norwegian Institute of Public Health; 2014 http://www.fhi.no/dokumenter/4313e1cf39.pdf[accessed 09.04.2015].

[31]Peele S, Brodsky A. Exploring psychological beneﬁts associated with moderate alcohol use: a necessary corrective to assessments of drinking outcomes?

Drug Alcohol Depend 2000;60(3):221e47.

[32]Skogen JC, Harvey SB, Henderson M, Stordal E, Mykletun A. Anxiety and depression among abstainers and low-level alcohol consumers. The Nord- Trondelag Health Study. Addiction 2009;104(9):1519e29.

[33]Peacock JL, Bland JM, Anderson HR. Preterm delivery: effects of socioeconomic factors, psychological stress, smoking, alcohol, and caffeine. BMJ 1995;311(7004):531e5.

[34]Nilsen RM, Vollset SE, Gjessing HK, Skjærven R, Melve KK, Schreuder P, et al.

Self-selection and bias in a large prospective pregnancy cohort in Norway.

Paediatr Perinat Epidemiol 2009;23(6):597e608.

[35]Kvalvik LG, Nilsen RM, Skjaerven R, Vollset SE, Midttun O, Ueland PM, et al.

Self-reported smoking status and plasma cotinine concentrations among pregnant women in the Norwegian Mother and Child Cohort Study. Pediatr Res 2012;72(1):101e7.

[36]Day NL, Cottereau CM, Ricardson GA. The epidemiology of alcohol, marijuana, and cocaine use among women of childbearing age and pregnant women. Clin Obstet Gynecol 1993;36(2):232e45.

[37]Whitehead N, Lipscomb L. Patterns of alcohol use before and during preg- nancy and the risk of small-for-gestational-age birth. Am J Epidemiol 2003;158(7):654e62. M.T.G. Dale et al. / Annals of Epidemiology 26 (2016) 275e282 282