Advertisement

Week-by-week alcohol consumption in early pregnancy and spontaneous abortion risk: a prospective cohort study

      Background

      Half of women use alcohol in the first weeks of gestation, but most stop once pregnancy is detected. The relationship between timing of alcohol use cessation in early pregnancy and spontaneous abortion risk has not been determined.

      Objective

      This study aimed to evaluate the association between week-by-week alcohol consumption in early pregnancy and spontaneous abortion.

      Study Design

      Participants in Right from the Start, a community-based prospective pregnancy cohort, were recruited from 8 metropolitan areas in the United States (2000–2012). In the first trimester, participants provided information about alcohol consumed in the prior 4 months, including whether they altered alcohol use; date of change in use; and frequency, amount, and type of alcohol consumed before and after change. We assessed the association between spontaneous abortion and week of alcohol use, cumulative weeks exposed, number of drinks per week, beverage type, and binge drinking.

      Results

      Among 5353 participants, 49.7% reported using alcohol during early pregnancy and 12.0% miscarried. Median gestational age at change in alcohol use was 29 days (interquartile range, 15–35 days). Alcohol use during weeks 5 through 10 from last menstrual period was associated with increased spontaneous abortion risk, with risk peaking for use in week 9. Each successive week of alcohol use was associated with an 8% increase in spontaneous abortion relative to those who did not drink (adjusted hazard ratio, 1.08; 95% confidence interval, 1.04–1.12). This risk is cumulative. In addition, risk was not related to number of drinks per week, beverage type, or binge drinking.

      Conclusion

      Each additional week of alcohol exposure during the first trimester increases risk of spontaneous abortion, even at low levels of consumption and when excluding binge drinking.

      Key words

      Introduction

      The line between how alcohol is used before and during pregnancy is blurred in the first weeks of gestation. Although 10% of women continue to use alcohol through pregnancy, as many as half of pregnancies are exposed around conception.
      • Green P.P.
      • McKnight-Eily L.R.
      • Tan C.H.
      • Mejia R.
      • Denny C.H.
      Vital signs: alcohol-exposed pregnancies—United States, 2011-2013.
      • McCormack C.
      • Hutchinson D.
      • Burns L.
      • et al.
      Prenatal alcohol consumption between conception and recognition of pregnancy.
      • O’Keeffe L.M.
      • Kearney P.M.
      • McCarthy F.P.
      • et al.
      Prevalence and predictors of alcohol use during pregnancy: findings from international multicentre cohort studies.
      • Popova S.
      • Lange S.
      • Probst C.
      • Gmel G.
      • Rehm J.
      Estimation of national, regional, and global prevalence of alcohol use during pregnancy and fetal alcohol syndrome: a systematic review and meta-analysis.
      • Tan C.H.
      • Denny C.H.
      • Cheal N.E.
      • Sniezek J.E.
      • Kanny D.
      Alcohol use and binge drinking among women of childbearing age - United States, 2011-2013.
      The tendency to use alcohol until pregnancy detection is consistent among both women with intended and unintended pregnancies, which suggests that preemptive change in alcohol use when planning a pregnancy is not typical.
      • Pryor J.
      • Patrick S.W.
      • Sundermann A.C.
      • Wu P.
      • Hartmann K.E.
      Pregnancy intention and maternal alcohol consumption.
      Previous studies on alcohol neglect or cannot capture information about the timing of exposure in early pregnancy, which may obscure or underestimate the risk of outcomes such as spontaneous abortion.
      • Sundermann A.C.
      • Zhao S.
      • Young C.L.
      • et al.
      Alcohol use in pregnancy and miscarriage: a systematic review and meta-analysis.
      This limitation may fuel the misconception that adverse pregnancy outcomes are only associated with heavy consumption and that modest, occasional use is harmless.
      • Meurk C.S.
      • Broom A.
      • Adams J.
      • Hall W.
      • Lucke J.
      Factors influencing women’s decisions to drink alcohol during pregnancy: findings of a qualitative study with implications for health communication.
      ,
      • Holland K.
      • McCallum K.
      • Walton A.
      ‘I’m not clear on what the risk is’: women’s reflexive negotiations of uncertainty about alcohol during pregnancy.

      Why was this study conducted?

      Alcohol use is common in the first weeks of gestation before pregnancy detection. Alcohol is routinely treated as an unchanging exposure, making information about how timing and duration of use relates to spontaneous abortion risk scarce.

      Key findings

      Each additional week of alcohol exposure during the first trimester increases the risk of spontaneous abortion, even at low levels of consumption and when excluding binge drinking. Alcohol use in weeks 5 through 10 of pregnancy is associated with an increased risk of spontaneous abortion.

      What does this add to what is known?

      The timing of alcohol use matters, and each additional week of even modest consumption is associated with an increased risk of spontaneous abortion. An emphasis on early detection of pregnancy and cessation of alcohol use could curtail spontaneous abortions linked with exposure.
      Spontaneous abortion occurs in an estimated 1 in 6 recognized pregnancies
      • Ammon Avalos L.
      • Galindo C.
      • Li D.K.
      A systematic review to calculate background miscarriage rates using life table analysis.
      ,
      • Wilcox A.J.
      • Weinberg C.R.
      • O’Connor J.F.
      • et al.
      Incidence of early loss of pregnancy.
      and can come at a great emotional cost.
      • Farren J.
      • Mitchell-Jones N.
      • Verbakel J.Y.
      • Timmerman D.
      • Jalmbrant M.
      • Bourne T.
      The psychological impact of early pregnancy loss.
      Alcohol use may increase the risk of spontaneous abortion through several potential mechanisms: oxidative stress secondary to alcohol consumption may disrupt biochemical pathways involved in embryogenesis; exposure can hinder retinoic acid synthesis, thereby affecting epigenetic programming and cell lineage determination; and alcohol use can alter maternal hormone levels, affecting uterine receptivity.
      • Kalisch-Smith J.I.
      • Moritz K.M.
      Detrimental effects of alcohol exposure around conception: putative mechanisms.
      Studies of alcohol use and spontaneous abortion are often hindered by methodologic shortcomings such as recall bias and imprecision in determining gestational age at pregnancy loss.
      • Sundermann A.C.
      • Mukherjee S.
      • Wu P.
      • Velez Edwards D.R.
      • Hartmann K.E.
      Gestational age at arrest of development: an alternative approach for assigning time at risk in studies of time-varying exposures and miscarriage.
      Many recruit participants during prenatal care, meaning enrollment takes place later in gestation than many spontaneous abortions occur. Others are vulnerable to selection bias because of recruitment methods that differ by pregnancy outcome. Prior studies routinely treat alcohol use as an unchanging exposure, which does not reflect the pattern of use for most women.
      • McCormack C.
      • Hutchinson D.
      • Burns L.
      • et al.
      Prenatal alcohol consumption between conception and recognition of pregnancy.
      ,
      • Tan C.H.
      • Denny C.H.
      • Cheal N.E.
      • Sniezek J.E.
      • Kanny D.
      Alcohol use and binge drinking among women of childbearing age - United States, 2011-2013.
      ,
      • Pryor J.
      • Patrick S.W.
      • Sundermann A.C.
      • Wu P.
      • Hartmann K.E.
      Pregnancy intention and maternal alcohol consumption.
      These factors leave women and care providers with limited access to data about how the timing of alcohol use in pregnancy relates to spontaneous abortion.
      In this prospective cohort, we had the opportunity to recruit participants representative of the general obstetrical population. They were enrolled while planning a pregnancy or in early pregnancy and reported alcohol use both before and after any change in drinking. Our primary objective was to incorporate information about week-by-week alcohol use in measures of spontaneous abortion risk.

      Materials and Methods

      Study population

      With institutional review board approval, we recruited women early in pregnancy or planning a pregnancy into Right from the Start (RFTS), a prospective cohort. Women were enrolled between 2000 and 2012 from 8 metropolitan areas in North Carolina, Tennessee, and Texas.
      • Promislow J.H.
      • Makarushka C.M.
      • Gorman J.R.
      • Howards P.P.
      • Savitz D.A.
      • Hartmann K.E.
      Recruitment for a community-based study of early pregnancy: the Right From The Start study.
      Recruitment materials were distributed through businesses, community groups, public advertising, direct mail, and prenatal care providers. Eligibility required participants to be aged 18 years or older, English- or Spanish-speaking, no use of fertility treatments, and intention to carry pregnancy to term. Participants were enrolled before 12 completed weeks of gestation (median gestational age at enrollment, 47 days; interquartile range [IQR], 38–58; n=5353) and gave informed consent. Women intending to become pregnant were provided free pregnancy tests for up to 6 months and enrolled at first positive pregnancy test.
      Participants completed an intake interview at baseline and a computer-assisted telephone interview during the first trimester. Interviews collected information about demographics, medical history, reproductive history, lifestyle, and health behaviors. Participants had a transvaginal research ultrasound in the first trimester. The median gestational age of ultrasound was 58 days (IQR, 49–69 days).

      Exposure

      During the first trimester interview, participants provided detailed information about alcohol consumed in the past 4 months (Appendix). This window was selected to capture alcohol exposure immediately before pregnancy and through the first trimester. Participants reported whether they altered alcohol use during this period; date of change in use; frequency, amount, and type of alcohol consumed before and after change; and number of binge episodes, defined as consumption of more than 4 drinks in an episode. Gestational age at change was determined using self-reported last menstrual period (LMP), which is a validated and reliable dating method in this cohort.
      • Hoffman C.S.
      • Messer L.C.
      • Mendola P.
      • Savitz D.A.
      • Herring A.H.
      • Hartmann K.E.
      Comparison of gestational age at birth based on last menstrual period and ultrasound during the first trimester.
      We used self-reported LMP for gestational dating for all study participants because ultrasound-based dating often underestimates gestational age in pregnancies that go on to end in loss. Number of drinks per week was calculated for before and after change and was evaluated as both a continuous and categorical measure (unexposed, ≤1 drink/week, 1.01–2 drinks/week, 2.01–4 drinks/week, and >4 drinks/week). Beverage type was categorized as wine, beer, or liquor (consumed alone or in mixed drinks).

      Outcome

      Participants provided pregnancy status at 20 to 25 weeks from LMP. Self-reported pregnancy outcome was corroborated by information abstracted by trained study personnel from vital records, birth certificates, and medical records. Spontaneous abortion was defined as loss of pregnancy before 20 completed weeks of gestation. Pregnancies ending in spontaneous abortion were compared with those surviving past 20 weeks of gestation (live births and stillbirths), and participants with an unknown pregnancy outcome were censored at the date of last study contact. We defined timing of pregnancy outcome among women with spontaneous abortion using 2 approaches: gestational age at spontaneous abortion and gestational age at arrest of development estimated using features observed on research ultrasound before loss.
      • Sundermann A.C.
      • Mukherjee S.
      • Wu P.
      • Velez Edwards D.R.
      • Hartmann K.E.
      Gestational age at arrest of development: an alternative approach for assigning time at risk in studies of time-varying exposures and miscarriage.
      Gestational age at arrest of development was estimated using features observed on research ultrasound before loss, which was not available for 28.6% of participants who had a pregnancy that ended in spontaneous abortion (185/645). Because the distribution of gestational age at arrest differs by gestational age at spontaneous abortion, we assigned gestational age at arrest of development for women without a research ultrasound through random sampling of observed values of gestational age at arrest among women who had a spontaneous abortion in the same gestational week.

      Statistical analysis

      We used 2 main modeling approaches to quantify risk associated with alcohol use in pregnancy because the timing of alcohol exposure may influence risk in multiple ways. First, we considered alcohol exposure by gestational week of exposure. Second, we evaluated how duration of alcohol exposure relates to risk.

      Gestational age–specific exposure

      Timing of exposure during pregnancy maps to embryologic development and thus informs risk, so we examined gestational week–specific effects of alcohol use. We performed separate logistic regressions to estimate adjusted odds ratios (aORs) for spontaneous abortion and alcohol exposure (yes/no) in each gestational week of the first trimester. Participants who did not use alcohol during pregnancy were counted as unexposed for all weeks, and participants who did not change consumption or who only altered amount were considered exposed for all weeks. Participants who stopped using alcohol during the first trimester were classified as exposed in weeks before reported change and unexposed thereafter. Participants were included in week-specific models if they had not yet had a loss or been censored by the beginning of the week.
      To evaluate the role of the amount of alcohol consumed, we quantified the association between spontaneous abortion risk and the number of drinks per week in 4 developmental windows in which teratogens are expected to confer risk through different mechanisms: periimplantation (gestational weeks 1–4), early embryonic (gestational weeks 5–7), late embryonic (gestational weeks 8–10), and fetal (gestational weeks 11 and 12).
      • Moore K.L.
      • Persaud T.V.N.
      • Torchia M.G.
      The developing human: clinically oriented embryology.
      We performed separate logistic regressions for amount of alcohol consumed and spontaneous abortion risk for each window. Logistic regression model fit was assessed using Pearson goodness-of-fit test and Hosmer-Lemeshow test.

      Duration of exposure

      We also considered that duration of alcohol use during pregnancy may drive risk. We used extended Cox survival models to measure the association between spontaneous abortion and duration of alcohol use, operationalized as the number of days between LMP and time t or gestational age at cessation of alcohol use, whichever came first. If a participant reported continuing alcohol use, duration of exposure accumulated until the first trimester interview. We present adjusted hazard ratios (aHRs) associated with each additional week of use. Participants contributed time in the model from day of enrollment through 20 weeks’ gestation, arrest of development, or loss to follow-up, whichever came first. Left truncation based on gestational age at enrollment allowed us to more precisely estimate spontaneous abortion risk by taking into account whether a subject had an ongoing pregnancy at cohort entry.
      • Dupont W.D.
      Statistical modeling for biomedical researchers: a simple introduction to the analysis of complex data.
      Information about the duration of exposure was updated in the model for each gestational day. Once a participant entered the cohort, the cumulative number of days exposed during pregnancy was reflected in the model, thus incorporating information about exposure that occurred before cohort entry while protecting for immortal time bias by not counting events that could not be observed. Given the varying amount of congeners,
      • Greizerstein H.B.
      Congener contents of alcoholic beverages.
      such as acetaldehyde, in different alcohols, we also evaluated the risk associated with each additional week of exposure by beverage type in a secondary analysis.

      Commonalities between approaches

      Adjusted models included covariates selected a priori based on a directed acyclic graph of known or suspected relationships with alcohol consumption and spontaneous abortion risk
      • Greenland S.
      • Pearl J.
      • Robins J.M.
      Causal diagrams for epidemiologic research.
      : maternal age (years, spline), race/ethnicity (non-Hispanic white, non-Hispanic black, or other), education (high school or less, some college, and college or more), cigarette use (never smoker or distant quit [more than 4 months before first trimester interview], recent quit, or current smoker), pregnancy intention (intended or unintended),
      • Finer L.B.
      • Zolna M.R.
      Unintended pregnancy in the United States: incidence and disparities, 2006.
      and parity (nulliparous, 1 prior birth, and 2 or more prior births).
      We enrolled 6105 women; data from 5353 women were eligible for analysis (Figure 1). Participants who were excluded because they were missing data for 1 or more variables in the covariate set were younger and more likely to be black and have an unintended pregnancy than participants with complete data (n=71; Supplemental Table 1).
      Figure thumbnail gr1
      Figure 1Flow diagram for study population derivation
      Sundermann et al. Alcohol use and spontaneous abortion. Am J Obstet Gynecol 2021.
      We performed a series of sensitivity analyses to determine robustness of results; analyses were repeated with pregnancy endpoint for losses defined as gestational age at spontaneous abortion as opposed to gestational age at arrest of development, with participants without a research ultrasound excluded, and with women who reported binge drinking excluded. We tested for effect modification by maternal body mass index (continuous) and smoking status using the likelihood ratio test for the inclusion of interaction terms. In a secondary analysis, we used Cox proportional hazard models to measure the association between the number of binge episodes (none, 1–3, and ≥4) and spontaneous abortion.
      We used 2-sided tests with a significance level of .05. Threshold for significance was Bonferroni-corrected by a factor equal to the number of tests performed in the hypothesis. Analyses were performed in Stata (Version 14.2, StataCorp, College Station, TX).

      Results

      Among 5353 women, 14.1% reported never using alcohol, 36.2% quit before LMP, 44.3% quit after LMP, and 5.4% continued use. Among 2926 women who reported a change in alcohol exposure within the month before conception or during the first trimester, 91.5% quit using alcohol, 8.0% decreased use, and 0.5% increased use. Median gestational age at change was 29 days (IQR, 15–35 days) and 41.0% of participants who reported a change altered use within 3 days of a positive pregnancy test (1214/2962). Higher maternal age, household income, level of education, prenatal vitamin use, and illicit drug use were associated with alcohol exposure during pregnancy (Table 1). Non-Hispanic white women, nulliparous women, and smokers were more likely to be exposed to alcohol during pregnancy than their counterparts. Participants who were exposed to alcohol consumed a median of 2 drinks per week at the onset of pregnancy (IQR, 1–4 drinks per week). At least 1 binge episode during the periconception period or first trimester was reported by 11.0% of women (591/5349). The median number of binge episodes was 2 (IQR, 1–4), and 10.3% of participants who binged reported 10 or more episodes (61/591).
      Table 1Participant characteristics by alcohol use during pregnancy
      CharacteristicNo alcohol use (n=2691)Alcohol use
      Alcohol use defined as exposure past last menstrual period
      (n=2662)
      Unadjusted OR95% CI
      Maternal age, n (%), y
       <25623 (23.2)418 (15.7)1.00Referent
       25–29962 (35.7)880 (33.1)1.361.17–1.59
       30–34784 (29.1)936 (35.2)1.781.52–2.08
       ≥35322 (12.0)428 (16.1)1.981.64–2.40
      Race/ethnicity, n (%)
       White, non-Hispanic1723 (64.0)2052 (77.1)1.00Referent
       Black, non-Hispanic634 (23.6)353 (13.3)0.470.40–0.54
       Other334 (12.4)257 (9.7)0.650.54–0.77
      Education, n (%)
       High school or less586 (21.8)340 (12.8)1.00Referent
       Some college520 (19.3)442 (16.6)1.471.22–1.76
       College or more1585 (58.9)1880 (70.6)2.041.76–2.37
      Household income, n (%)
       ≤$40,000967 (35.9)638 (24.0)1.00Referent
       $40,001–$80,000972 (36.1)967 (36.3)1.511.32–1.72
       >$80,000647 (24.0)986 (37.0)2.312.01–2.66
       Missing105 (3.9)71 (2.7)
      Marital status, n (%)
       Married or cohabitating2395 (88.4)2401 (90.2)1.00Referent
       Other296 (11.6)261 (9.8)0.880.74–1.05
      Parity, n (%)
       Nulliparous1149 (42.7)1414 (53.1)1.00Referent
       1 prior delivery984 (36.6)869 (32.6)0.720.64–0.81
       2+ prior deliveries558 (20.7)379 (14.2)0.550.47–0.64
      Prior spontaneous abortion, n (%)
       02020 (75.1)2115 (79.5)1.00Referent
       1518 (19.2)443 (16.6)0.820.71–0.94
       ≥2153 (5.7)104 (3.9)0.650.50–0.84
      BMI, n (%), kg/m2
       <18.567 (2.5)66 (2.5)0.870.62–1.24
       18.5–24.91334 (49.6)1505 (56.5)1.00Referent
       25–29.9645 (24.0)610 (22.9)0.840.73–0.96
       ≥30607 (22.6)460 (17.3)0.670.87–0.77
      Missing38 (1.4)21 (0.8)
      Smoking status,
      Quitting within the 4 months before the end of first trimester interview is considered a recent quit, and quitting before that time is considered a distant quit
      n (%)
       Never or distant quit2454 (91.2)2266 (85.1)1.00Referent
       Current or recent quit237 (8.8)396 (14.9)1.811.53–2.15
      Pregnancy intention, n (%)
       Intended1983 (73.7)1940 (72.9)1.00Referent
       Not intended708 (26.3)722 (27.1)1.040.92–1.18
      Prenatal vitamin use,
      Any history during the 4 months before the first trimester interview.
      n (%)
       No109 (4.1)64 (2.4)1.00Referent
       Yes2582 (95.9)2598 (97.6)1.711.25–2.34
      Illicit drug use,
      Any history during the 4 months before the first trimester interview.
      n (%)
       No2599 (96.6)2400 (90.2)1.00Referent
       Yes92 (3.4)262 (9.8)3.082.42–3.94
      Intimate partner violence,
      Any history during the 4 months before the first trimester interview.
      n (%)
       No2624 (97.5)2579 (96.9)1.00Referent
       Yes62 (2.3)79 (3.0)1.300.93–1.82
       Missing5 (0.2)4 (0.15)
      BMI, body mass index; CI, confidence interval; OR, odds ratio.
      Sundermann et al. Alcohol use and spontaneous abortion. Am J Obstet Gynecol 2021.
      a Alcohol use defined as exposure past last menstrual period
      b Quitting within the 4 months before the end of first trimester interview is considered a recent quit, and quitting before that time is considered a distant quit
      c Any history during the 4 months before the first trimester interview.
      Furthermore, 12% of pregnancies ended in spontaneous abortion (645/5353). When considering week-specific exposure, alcohol use in gestational weeks 5 through 10 was associated with spontaneous abortion after adjusting for multiple comparisons (aOR range, 1.42–4.85; Figure 2; Supplemental Table 2). Risk peaked for exposure in week 9 of gestation (aOR, 4.85; 95% confidence interval [CI], 3.30–7.13). These results were consistent between the 2 approaches used for defining timing of outcome (Supplemental Figure) and when excluding participants who reported binge drinking. A dose-response trend was not detected in any developmental window (Supplemental Table 3).
      Figure thumbnail gr2
      Figure 2Risk of spontaneous abortion by gestational week with alcohol exposure (n=5353)
      Pregnancy endpoint defined as the gestational age at arrest of development. Estimates adjusted for maternal age, race/ethnicity, education, parity, smoking status, and pregnancy intention. Weeks 5 through 10 are significant after adjusting for multiple comparisons (Bonferroni-corrected with a factor of 12).
      Sundermann et al. Alcohol use and spontaneous abortion. Am J Obstet Gynecol 2021.
      Each additional week of alcohol exposure during pregnancy was associated with an 8% relative increase in the risk of spontaneous abortion compared with the risk among women who were unexposed (aHR, 1.08; 95% CI, 1.04–1.12; Table 2). Participants who were exposed to alcohol up until 29 days of gestation (the median gestational age of alcohol use cessation in the cohort) had a 37% greater risk of spontaneous abortion relative to participants who were unexposed (aHR, 1.37; 95% CI, 1.18–1.60). Alcohol use in the lowest exposure category (≤1 drink per week) was associated with elevated risk in a way that was not different than estimates for higher levels of exposure (Table 2). Estimates did not vary by alcohol type (P=.99, Wald test) or when excluding participants who reported binge drinking. In addition, estimates did not differ when excluding pregnancies ending in a spontaneous abortion without a research ultrasound (Supplemental Table 4) or when defining pregnancy endpoint as the gestational age at spontaneous abortion (Supplemental Table 5).
      Table 2Risk of spontaneous abortion associated with each additional week of alcohol use during pregnancy
      Alcohol use
      Alcohol modeled as a time-varying exposure for duration of use, left truncation based on gestational age at enrollment
      characteristic
      Births (n=4708)

      n (%)
      Spontaneous abortions (n=645)

      n (%)
      Crude HR95% CIAdjusted HR
      Adjusted for maternal age (continuous, spline), race/ethnicity, education, parity, smoking status, and pregnancy intention
      95% CI
      Per additional weekPer additional week
      Any use
       No2367 (50.3)324 (50.2)1.00Referent1.00Referent
       Yes
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      2341 (49.7)321 (49.8)1.091.05–1.131.081.04–1.12
      Amount at LMP
      Categories reflect level of alcohol consumption before change in use, duration defined as prechange use
       Unexposed2367 (50.3)324 (50.2)1.00Referent1.00Referent
       ≤1 drink/wk
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      931 (19.8)120 (18.6)1.091.05–1.141.081.04–1.13
       1.01–2 drinks/wk449 (9.5)67 (10.4)1.061.00–1.121.061.00–1.12
       2.01–4 drinks/wk440 (9.3)60 (9.3)1.051.00–1.101.051.00–1.10
       >4 drinks/wk521 (11.1)74 (11.5)1.020.97–1.071.000.96–1.05
      Alcohol type
      Alcohol type categories do not total 100% because they are not mutually exclusive. Women who reported alcohol exposure in pregnancy but did not provide alcohol type are excluded from this analysis (n=30). Referent group is women unexposed to alcohol.
       Wine
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      1545 (32.8)201 (31.2)1.071.03–1.111.071.02–1.11
       Beer
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      1089 (23.1)138 (21.4)1.071.02–1.121.071.02–1.12
       Liquor858 (18.2)106 (16.4)1.030.97–1.091.040.98–1.10
      CI, confidence interval; HR, hazard ratio; LMP, last menstrual period.
      Sundermann et al. Alcohol use and spontaneous abortion. Am J Obstet Gynecol 2021.
      a Alcohol modeled as a time-varying exposure for duration of use, left truncation based on gestational age at enrollment
      b Adjusted for maternal age (continuous, spline), race/ethnicity, education, parity, smoking status, and pregnancy intention
      c Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      d Categories reflect level of alcohol consumption before change in use, duration defined as prechange use
      e Alcohol type categories do not total 100% because they are not mutually exclusive. Women who reported alcohol exposure in pregnancy but did not provide alcohol type are excluded from this analysis (n=30). Referent group is women unexposed to alcohol.
      We did not observe modification of the association between alcohol use and spontaneous abortion by maternal body mass index or smoking status. The number of binge episodes was not associated with spontaneous abortion risk (0 episodes [referent]; 1–3 episodes: aHR, 0.75; 95% CI, 0.48–1.15; ≥4 episodes: aHR, 0.88; 95% CI, 0.43–1.80), and inclusion of binge drinking in the main models did not alter findings. Any illicit drug use within 4 months leading up to the first trimester was reported by 6.6% of participants. Of the participants who reported illicit drug use, 84.2% reported marijuana as the only exposure (298/354). Intimate partner violence within the 4 months before the first trimester interview was reported by 2.6% of participants, and physical harm was reported by 1.3% of participants (n=9 did not respond). Including illicit drug use and intimate partner violence as covariates in the adjusted estimates did not alter results.

      Comment

      Principal findings

      In this prospective, community-recruited cohort, the timing of alcohol use is a key determinant of spontaneous abortion. Alcohol exposure occurred in half of the pregnancies, with many participants not changing use until a positive pregnancy test. Each additional week of alcohol use in the first trimester was associated with a cumulative increase in the risk of spontaneous abortion, and risk was most strongly related to exposure in weeks 5 through 10 of pregnancy. This window aligns with the embryonic stage of development, when organogenesis is occurring and pregnancy is most vulnerable to insults.
      • Yelin R.
      • Ben-Haroush Schyr R.
      • Kot H.
      • et al.
      Ethanol exposure affects gene expression in the embryonic organizer and reduces retinoic acid levels.
      ,
      • Polifka J.E.
      • Friedman J.M.
      Medical genetics: 1. Clinical teratology in the age of genomics.
      These findings persisted when excluding women who reported binge drinking.
      Women who were older than 35 years, white, college-educated, and from high-income households were most likely to use alcohol. Although this is not the population generally flagged for high-risk behaviors, these demographics are consistently linked with alcohol use during pregnancy.
      • O’Keeffe L.M.
      • Kearney P.M.
      • McCarthy F.P.
      • et al.
      Prevalence and predictors of alcohol use during pregnancy: findings from international multicentre cohort studies.
      ,
      • Tan C.H.
      • Denny C.H.
      • Cheal N.E.
      • Sniezek J.E.
      • Kanny D.
      Alcohol use and binge drinking among women of childbearing age - United States, 2011-2013.
      ,
      • Muggli E.
      • O’Leary C.
      • Donath S.
      • et al.
      “Did you ever drink more?” A detailed description of pregnant women’s drinking patterns.
      ,
      • Floyd R.L.
      • Decouflé P.
      • Hungerford D.W.
      Alcohol use prior to pregnancy recognition.
      Clinical biases may result in these women being overlooked for risk counseling even though this group is at the greatest risk for modest, continued alcohol use.

      Results in context

      Prior studies of alcohol exposure and spontaneous abortion risk are limited by the methods used for ascertaining and modeling exposure.
      • Sundermann A.C.
      • Zhao S.
      • Young C.L.
      • et al.
      Alcohol use in pregnancy and miscarriage: a systematic review and meta-analysis.
      Many define exposure as alcohol use after pregnancy recognition. In RFTS, this definition misclassifies 44.3% of participants as unexposed. Others calculate an across-pregnancy average dose or describe prepregnancy alcohol use and its associated risk separately. An across-pregnancy average dose neglects that exposure is likely concentrated in early pregnancy. Evaluating prepregnancy exposure separately without considering how long use persists disregards that risk may be tied to gestational age at exposure. Alcohol use typically occurs before pregnancy detection and rapidly tapers thereafter. Therefore, most exposure co-occurs with the first stages of embryo development. Our results suggest that the timing of exposure is critical in understanding spontaneous abortion risk.

      Strengths and limitations

      Before considering the implications of these findings, let us audit the level of confidence we should have in the results. We relied on self-report to determine alcohol use because no sufficiently sensitive and specific biomarker for alcohol exposure exists.
      • Howlett H.
      • Abernethy S.
      • Brown N.W.
      • Rankin J.
      • Gray W.K.
      How strong is the evidence for using blood biomarkers alone to screen for alcohol consumption during pregnancy? A systematic review.
      Social desirability bias, or responding in a way deemed favorable by others, may lead women to underreport alcohol use during pregnancy.
      • Bailey B.A.
      • Sokol R.J.
      Prenatal alcohol exposure and miscarriage, stillbirth, preterm delivery, and sudden infant death syndrome.
      ,
      • Ernhart C.B.
      • Morrow-Tlucak M.
      • Sokol R.J.
      • Martier S.
      Underreporting of alcohol use in pregnancy.
      We attempted to minimize this bias by conducting telephone interviews in a nonclinical and confidential setting using questionnaires with nonjudgmental wording and unknown interviewers. Prevalence of alcohol use at the onset of pregnancy in this cohort aligns with national data about exposure among nonpregnant, reproductive-aged women,
      • Tan C.H.
      • Denny C.H.
      • Cheal N.E.
      • Sniezek J.E.
      • Kanny D.
      Alcohol use and binge drinking among women of childbearing age - United States, 2011-2013.
      ,
      • Alshaarawy O.
      • Breslau N.
      • Anthony J.C.
      Monthly estimates of alcohol drinking during pregnancy: United States, 2002-2011.
      which provides reassurance that social desirability bias did not unduly suppress reporting about the presence of alcohol exposure.
      Assessment of alcohol exposure followed loss for 67.2% of spontaneous abortions (436/649), allowing potential for recall bias.
      • Rockenbauer M.
      • Olsen J.
      • Czeizel A.E.
      • Pedersen L.
      • Sørensen H.T.
      EuroMAP Group
      Recall bias in a case-control surveillance system on the use of medicine during pregnancy.
      However, the proportion of women with losses who reported alcohol exposure during pregnancy did not differ by interview timing before or after loss (chi-square P value=.78), and gestational age at change in alcohol consumption was similar between the groups (median 31 days vs 32 days; Wilcoxon rank-sum P value=.36).
      We did not observe a dose-response relationship between alcohol exposure and risk. Although many biological relationships operate on a dose-dependent gradient, the timing of alcohol use may drive spontaneous abortion risk with a threshold effect observed at low levels of exposure. In fetal alcohol spectrum disorder, a dose-response relationship is not always the rule.
      • Sokol R.J.
      • Delaney-Black V.
      • Nordstrom B.
      Fetal alcohol spectrum disorder.
      ,
      • May P.A.
      • Gossage J.P.
      Maternal risk factors for fetal alcohol spectrum disorders: not as simple as it might seem.
      Facial abnormalities characteristic of fetal alcohol spectrum disorder can be observed for low levels of alcohol use if exposure occurs when neural crest cells are migrating to form facial structures,
      • Muggli E.
      • Matthews H.
      • Penington A.
      • et al.
      Association between prenatal alcohol exposure and craniofacial shape of children at 12 months of age.
      and changes in neonatal brain activity are observed with low levels of prenatal alcohol exposure.
      • Shuffrey L.C.
      • Myers M.M.
      • Isler J.R.
      • et al.
      Association between prenatal exposure to alcohol and tobacco and neonatal brain activity: results from the Safe Passage Study.
      Alternatively, imprecision or bias in reporting the amount of alcohol consumed may obscure a dose-dependent effect. Because alcohol use during pregnancy is stigmatized, information about the amount consumed may be more vulnerable to reporting biases than responses about mere presence or absence of exposure. In addition, misconceptions about size and alcohol content of a standard drink may lead to error in earnest reporting.
      RFTS prioritized early recruitment of pregnancies to capture as many spontaneous abortion events as possible; 25.8% of participants entered the study before conceiving and 71.6% enrolled before 7 weeks’ gestation. The proportion of participants who were exposed to alcohol in early pregnancy and timing of change in alcohol use was similar when comparing women who enrolled before conception with those who enrolled in the first trimester, and results are unchanged when excluding participants who were enrolled before conception. Recruitment before conception or initiation of prenatal care enabled earlier enrollment than clinic-based studies. Although this is an improvement over many studies of spontaneous abortion, losses occurring very early in gestation are inevitably underrepresented in this sample. We truncated time before enrollment in survival analyses to account for a participant having an ongoing pregnancy at study entry. Risk associated with alcohol use in the first weeks of pregnancy may be higher than estimated if unobserved early losses were highly associated with alcohol exposure.
      Because this cohort required early enrollment, this study also has a higher proportion of planned pregnancies than the general population (73% vs 51%).
      • Finer L.B.
      • Zolna M.R.
      Unintended pregnancy in the United States: incidence and disparities, 2006.
      The proportion of participants exposed to alcohol at pregnancy onset and timing of change in drinking was similar for participants with intended and unintended pregnancies, indicating planned pregnancies do not necessarily involve preparatory changes in alcohol use. Of women who were exposed, 40% reported quitting alcohol use within 3 days of a positive pregnancy test.
      Less than 1% of participants reported a history of type 1 or type 2 diabetes, and results were unchanged when diabetes was included in the adjusted model. A priori, we determined there would be insufficient power to address other medical conditions associated with early pregnancy loss, such as infection or maternal antiphospholipid syndrome. These conditions are rare and unlikely to influence findings of a large cohort. When limiting the analysis to women who had at least 1 prior live birth, and therefore proven capacity for a successful pregnancy, results were unchanged.

      Conclusions and implications

      Studies not accounting for alcohol exposure in early pregnancy obscure the time-dependent effect of alcohol use and underestimate risk. In this prospective cohort, we find that risk of spontaneous abortion accumulates with each successive week of alcohol use, even at low levels of consumption and when excluding binge drinking. These findings underscore the warning of no known safe amount of alcohol in pregnancy.
      Centers for Disease Control and Prevention
      Fetal alcohol spectrum disorders (FASDs): alcohol use in pregnancy.
      Optimally, exposure would be completely prevented; still, half of pregnancies in the United States are unintended and abstaining from alcohol when planning a pregnancy is not typical. Because home pregnancy testing reliably detects pregnancy as early as 4 weeks’ gestation and alcohol use in weeks 5 through 10 is most concerning for risk, there is a window of opportunity. Therefore, efforts to promote early pregnancy recognition and cessation of alcohol use are warranted to curtail risk of spontaneous abortion.

      Supplementary Data

      Appendix

      Questionnaire items about alcohol consumption in first trimester interview.
      • Have you ever had alcoholic beverages, like beer, wine, or liquor including gin, whiskey, rum, or mixed drinks?
      • At this time, do you drink any alcoholic beverages, like beer, wine, or liquor including gin, whiskey, rum, or mixed drinks?
      • How often do you drink an alcoholic beverage, by that I mean at least one beer, one glass of wine, one mixed drink, or one shot of liquor? (# times per day, week, or month, or less than once per month).
      • On those occasions that you drink alcoholic beverages, how many drinks do you usually have?
      • At this time, what type(s) of alcohol do you usually drink? To make it easier for you to respond, I’m going to read you a list of options: beer, wine, mixed drinks, shot of liquor, other alcohol _______.
      • Did you stop drinking alcoholic beverages in the past 4 months or more than 4 months ago?
      • In the past 4 months, have you changed how often and/or how many alcoholic beverages you drink?
      • When did this change occur?
      • Do you remember what week in [month] that was, the first, second, third, fourth, or fifth?
      • Before this change, how often did you drink? (# times per day, week, or month, or less than once per month).
      • Before this change, on those occasions when you drank alcoholic beverages, how many drinks did you usually have on each occasion?
      • What type(s) of alcohol did you usually drink? Did you drink beer, wine, mixed drinks, shot of liquor, other alcohol ______.
      • In the past four months, have you had more than four drinks on any one occasion?
      • How many times in the past four months have you had more than four drinks on any occasion?
      • On those occasions when you had more than four drinks, what type(s) of alcohol did you usually drink? Did you drink beer, wine, mixed drinks, shot of liquor, other alcohol ______.
      Sundermann et al. Alcohol use and spontaneous abortion. Am J Obstet Gynecol 2021.
      Figure thumbnail fx1
      Supplemental FigureRisk of spontaneous abortion by gestational week with alcohol exposure; pregnancy endpoint defined as gestational age at spontaneous abortion
      Estimates are adjusted for maternal age, race/ethnicity, education, parity, smoking status, and pregnancy intention. Participants with complete data for adjusted analysis are included (n=5353). Weeks 5 through 12 are significant after adjusting for multiple comparisons (Bonferroni-corrected with a factor of 12). Black dot represents adjusted odds ratio, black line spans 95% confidence interval, and dotted line indicates an odds ratio of 1.
      Sundermann et al. Alcohol use and spontaneous abortion. Am J Obstet Gynecol 2021.
      Supplemental Table 1Comparison of participants with complete covariate data with those missing data for 1 or more covariates
      CharacteristicComplete case (n=5353)Missing covariates (n=71)P value
      P value calculated using Wilcoxon rank-sum test for continuous variables and Fisher exact test for categorical variables (missing not included)
      Maternal age, median (IQR), y29 (26–32)25 (21–30)<.01
      Race/ethnicity, n (%)<.01
       White, non-Hispanic3775 (70.5)16 (22.5)
       Black, non-Hispanic987 (18.4)39 (54.9)
       Other591 (11.0)13 (18.3)
       Refused0 (0.0)3 (4.2)
      Education, n (%)<.01
       High school or less926 (17.3)42 (59.2)
       Some college962 (18.0)14 (19.7)
       College or more3465 (64.7)14 (19.7)
       Missing0 (0.0)1 (1.4)
      Marital status, n (%)<.01
       Married or cohabitating4796 (89.6)43 (60.6)
       Other557 (10.4)28 (39.4)
      Parity, n (%).01
       Nulliparous2563 (47.9)6 (8.5)
       1 prior delivery1853 (34.6)2 (2.8)
       2+ prior deliveries937 (17.5)5 (7.0)
       Missing0 (0.0)58 (81.7)
      Smoking status, n (%)
      Quitting within the 4 months before the end of first trimester interview is considered a recent quit and quitting before that time is considered a distant quit
      <.01
       Never or distant quit4720 (88.2)53 (74.6)
       Current or recent quit633 (11.8)18 (25.4)
      Pregnancy intention, n (%)<.01
       Intended3923 (73.3)23 (32.4)
       Not intended1430 (26.7)37 (52.1)
       Missing0 (0.0)11 (15.5)
      Alcohol use, n (%)
      Alcohol use is defined as use past LMP.
      .23
       Yes2662 (49.7)30 (42.3)
       No2691 (50.3)41 (57.7)
      Gestational age at change, median (IQR), d29 (15–35)22 (6–35).10
      Consumption at LMP, median (IQR), drinks/wk2.0 (1.0–4.0)2.0 (0.3–3.0).27
      Pregnancy outcome, n (%).14
       Spontaneous abortion645 (12.0)4 (5.6)
       No spontaneous abortion4708 (88.0)67 (93.4)
      IQR, interquartile range; LMP, last menstrual period.
      Sundermann et al. Alcohol use and spontaneous abortion. Am J Obstet Gynecol 2021.
      a P value calculated using Wilcoxon rank-sum test for continuous variables and Fisher exact test for categorical variables (missing not included)
      b Quitting within the 4 months before the end of first trimester interview is considered a recent quit and quitting before that time is considered a distant quit
      c Alcohol use is defined as use past LMP.
      Supplemental Table 2Risk of spontaneous abortion by gestational week of alcohol exposure
      Week of alcohol exposureBirths
      Counts reflect participants who contributed to each week-specific model. Participants were only included if they had complete data for adjusted analysis and had not had a spontaneous abortion or been censored by the beginning of the week
      Spontaneous abortion
      Counts reflect participants who contributed to each week-specific model. Participants were only included if they had complete data for adjusted analysis and had not had a spontaneous abortion or been censored by the beginning of the week
      Adjusted OR
      Adjusted for maternal age (continuous, spline), race/ethnicity, education, parity, smoking status, and pregnancy intention
      95% CI
      Gestational wk 147086450.970.81–1.16
      Gestational wk 247086451.050.89–1.25
      Gestational wk 347086451.100.92–1.31
      Gestational wk 447086451.231.03–1.46
      Gestational wk 5
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 12).
      47086421.421.18–1.69
      Gestational wk 6
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 12).
      47085741.951.59–2.40
      Gestational wk 7
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 12).
      47084462.942.29–3.77
      Gestational wk 8
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 12).
      47082793.602.58–5.02
      Gestational wk 9
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 12).
      47052154.853.30–7.13
      Gestational wk 10
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 12).
      47051362.881.61–5.16
      Gestational wk 114702832.230.93–5.38
      Gestational wk 124700522.210.65–7.47
      CI, confidence interval; OR, odds ratio.
      Sundermann et al. Alcohol use and spontaneous abortion. Am J Obstet Gynecol 2021.
      a Counts reflect participants who contributed to each week-specific model. Participants were only included if they had complete data for adjusted analysis and had not had a spontaneous abortion or been censored by the beginning of the week
      b Adjusted for maternal age (continuous, spline), race/ethnicity, education, parity, smoking status, and pregnancy intention
      c Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 12).
      Supplemental Table 3Risk of spontaneous abortion associated with the amount of alcohol consumed in 3 developmental windows
      Alcohol use characteristicBirthsSpontaneous abortionsCrude OR95% CIAdjusted OR
      Adjusted for maternal age (continuous, spline), race/ethnicity, education, parity, smoking status, and pregnancy intention
      95% CI
      n
      We were not able to estimate dose-specific effects for exposure in the fetal window because alcohol exposure was rare late in the first trimester
      %n
      Counts reflect participants that contributed to analysis for each developmental window. Participants were included if they had complete data for adjusted analysis and had not had a spontaneous abortion or been censored by the beginning of the week
      %
      Weeks 1–44708645
       Unexposed236750.332450.21.00Referent1.00Referent
       ≤1 drink/wk93119.812018.60.940.75–1.180.910.73–1.14
       1.01–2 drinks/wk4499.56710.41.090.82–1.441.110.83–1.48
       2.01–4 drinks/wk4409.3609.31.000.74–1.340.980.72–1.33
       >4 drinks/wk52111.17411.51.040.79–1.360.970.73–1.29
      Weeks 5–74708642
       Unexposed324068.839361.21.00Referent1.00Referent
       ≤1 drink/wk
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 12).
      53711.49615.01.471.16–1.881.441.12–1.84
       1.01–2 drinks/wk2795.9497.61.451.05–1.991.521.09–2.11
       2.01–4 drinks/wk2826.0457.01.320.94–1.831.380.98–1.94
       >4 drinks/wk3707.9599.21.310.98–1.761.320.97–1.80
      Weeks 8–104708379
       Unexposed443494.222379.91.00Referent1.00Referent
       ≤1 drink/wk
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 12).
      1813.84114.74.503.13–6.493.972.71–5.83
       1.01–2 drinks/wk370.862.23.221.35–7.722.981.21–7.37
       2.01–4 drinks/wk320.741.42.490.87–7.092.290.79–6.64
       >4 drinks/wk240.551.84.141.57–10.963.421.25–9.35
      CI, confidence interval; OR, odds ratio.
      Sundermann et al. Alcohol use and spontaneous abortion. Am J Obstet Gynecol 2021.
      a We were not able to estimate dose-specific effects for exposure in the fetal window because alcohol exposure was rare late in the first trimester
      b Counts reflect participants that contributed to analysis for each developmental window. Participants were included if they had complete data for adjusted analysis and had not had a spontaneous abortion or been censored by the beginning of the week
      c Adjusted for maternal age (continuous, spline), race/ethnicity, education, parity, smoking status, and pregnancy intention
      d Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 12).
      Supplemental Table 4Risk of spontaneous abortion associated with each additional week of alcohol use during pregnancy, cases restricted to those with ultrasound data
      Alcohol use
      Alcohol modeled as a time-varying exposure for duration of use, left truncation based on gestational age at enrollment
      characteristic
      Births (n=4708)
      Participants with research ultrasound for estimation of gestational age at arrest of development are included in this table (n=5170)
      Spontaneous abortions (n=462)
      Participants with research ultrasound for estimation of gestational age at arrest of development are included in this table (n=5170)
      Crude HR95% CIAdjusted HR
      Adjusted for maternal age (continuous, spline), race/ethnicity, education, parity, smoking status, and pregnancy intention
      95% CI
      n%n%Per additional weekPer additional week
      Any use
       No236750.323951.71.00Referent1.00Referent
       Yes
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      234149.722348.31.081.04–1.131.071.03–1.12
      Amount at LMP
      Categories reflect level of alcohol consumption before change in use, duration defined as prechange use
       Unexposed236750.323951.71.00Referent1.00Referent
       ≤1 drink/wk
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      93119.87416.01.071.02–1.131.061.01–1.12
       1.01–2 drinks/wk4499.54910.61.071.00–1.131.071.00–1.13
       2.01–4 drinks/wk4409.3459.71.050.99–1.111.050.99–1.11
       >4 drinks/wk52111.15511.91.030.98–1.091.010.96–1.07
      Alcohol type
      Alcohol type categories do not total 100% because they are not mutually exclusive. Women who reported alcohol exposure in pregnancy but did not provide alcohol type are excluded from this analysis (n=30). Referent group is women unexposed to alcohol.
       Wine
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      154532.813829.91.071.02–1.121.061.01–1.11
       Beer
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      108923.19620.81.071.02–1.131.071.01–1.12
       Liquor85818.27917.11.030.96–1.101.030.96–1.10
      CI, confidence interval; HR, hazard ratio; LMP, last menstrual period.
      Sundermann et al. Alcohol use and spontaneous abortion. Am J Obstet Gynecol 2021.
      a Alcohol modeled as a time-varying exposure for duration of use, left truncation based on gestational age at enrollment
      b Participants with research ultrasound for estimation of gestational age at arrest of development are included in this table (n=5170)
      c Adjusted for maternal age (continuous, spline), race/ethnicity, education, parity, smoking status, and pregnancy intention
      d Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      e Categories reflect level of alcohol consumption before change in use, duration defined as prechange use
      f Alcohol type categories do not total 100% because they are not mutually exclusive. Women who reported alcohol exposure in pregnancy but did not provide alcohol type are excluded from this analysis (n=30). Referent group is women unexposed to alcohol.
      Supplemental Table 5Risk of spontaneous abortion associated with each additional week of alcohol use during pregnancy; pregnancy endpoint defined as gestational age at spontaneous abortion
      Alcohol use
      Alcohol modeled as a time-varying exposure for duration of use, left truncation based on gestational age at enrollment
      characteristic
      Births (n=4708)Spontaneous abortions (n=645)Crude HR95% CIAdjusted HR
      Adjusted for maternal age (continuous, spline), race/ethnicity, education, parity, smoking status, and pregnancy intention
      95% CI
      n%n%Per additional weekPer additional week
      Any use
       No236750.332450.21.00Referent1.00Referent
       Yes
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      234149.732149.81.101.07–1.131.091.07–1.12
      Amount at LMP
      Categories reflect level of alcohol consumption before change in use, duration defined as prechange use
       Unexposed236750.332450.21.00Referent1.00Referent
       ≤1 drink/wk
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      93119.812018.61.091.06–1.121.071.04–1.11
       1.01–2 drinks/wk
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      4499.56710.41.081.05–1.121.081.04–1.12
       2.01–4 drinks/wk
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      4409.3609.31.061.02–1.101.061.02–1.10
       >4 drinks/wk52111.17411.51.041.01–1.081.031.00–1.07
      Alcohol type
      Alcohol type categories do not total 100% because they are not mutually exclusive. Women who reported alcohol exposure in pregnancy but did not provide alcohol type are excluded from this analysis (n=30). Referent group is women unexposed to alcohol.
       Wine
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      154532.820131.21.081.06–1.111.081.05–1.11
       Beer
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      108923.113821.41.091.05–1.121.091.15–1.12
       Liquor
      Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      85818.210616.41.071.03–1.111.081.04–1.12
      CI, confidence interval; HR, hazard ratio; LMP, last menstrual period.
      Sundermann et al. Alcohol use and spontaneous abortion. Am J Obstet Gynecol 2021.
      a Alcohol modeled as a time-varying exposure for duration of use, left truncation based on gestational age at enrollment
      b Adjusted for maternal age (continuous, spline), race/ethnicity, education, parity, smoking status, and pregnancy intention
      c Significant after adjustment for multiple comparisons (Bonferroni-corrected with a factor of 4 for amount consumed and 3 for alcohol type)
      d Categories reflect level of alcohol consumption before change in use, duration defined as prechange use
      e Alcohol type categories do not total 100% because they are not mutually exclusive. Women who reported alcohol exposure in pregnancy but did not provide alcohol type are excluded from this analysis (n=30). Referent group is women unexposed to alcohol.

      References

        • Green P.P.
        • McKnight-Eily L.R.
        • Tan C.H.
        • Mejia R.
        • Denny C.H.
        Vital signs: alcohol-exposed pregnancies—United States, 2011-2013.
        MMWR Morb Mortal Wkly Rep. 2016; 65: 91-97
        • McCormack C.
        • Hutchinson D.
        • Burns L.
        • et al.
        Prenatal alcohol consumption between conception and recognition of pregnancy.
        Alcohol Clin Exp Res. 2017; 41: 369-378
        • O’Keeffe L.M.
        • Kearney P.M.
        • McCarthy F.P.
        • et al.
        Prevalence and predictors of alcohol use during pregnancy: findings from international multicentre cohort studies.
        BMJ Open. 2015; 5: e006323
        • Popova S.
        • Lange S.
        • Probst C.
        • Gmel G.
        • Rehm J.
        Estimation of national, regional, and global prevalence of alcohol use during pregnancy and fetal alcohol syndrome: a systematic review and meta-analysis.
        Lancet Glob Health. 2017; 5 ([published correction appears in Lancet Glob Health 2017;5:e276]): e290-e299
        • Tan C.H.
        • Denny C.H.
        • Cheal N.E.
        • Sniezek J.E.
        • Kanny D.
        Alcohol use and binge drinking among women of childbearing age - United States, 2011-2013.
        MMWR Morb Mortal Wkly Rep. 2015; 64: 1042-1046
        • Pryor J.
        • Patrick S.W.
        • Sundermann A.C.
        • Wu P.
        • Hartmann K.E.
        Pregnancy intention and maternal alcohol consumption.
        Obstet Gynecol. 2017; 129: 727-733
        • Sundermann A.C.
        • Zhao S.
        • Young C.L.
        • et al.
        Alcohol use in pregnancy and miscarriage: a systematic review and meta-analysis.
        Alcohol Clin Exp Res. 2019; 43: 1606-1616
        • Meurk C.S.
        • Broom A.
        • Adams J.
        • Hall W.
        • Lucke J.
        Factors influencing women’s decisions to drink alcohol during pregnancy: findings of a qualitative study with implications for health communication.
        BMC Pregnancy Childbirth. 2014; 14: 246
        • Holland K.
        • McCallum K.
        • Walton A.
        ‘I’m not clear on what the risk is’: women’s reflexive negotiations of uncertainty about alcohol during pregnancy.
        Health Risk Soc. 2016; 18: 38-58
        • Ammon Avalos L.
        • Galindo C.
        • Li D.K.
        A systematic review to calculate background miscarriage rates using life table analysis.
        Birth Defects Res A Clin Mol Teratol. 2012; 94: 417-423
        • Wilcox A.J.
        • Weinberg C.R.
        • O’Connor J.F.
        • et al.
        Incidence of early loss of pregnancy.
        N Engl J Med. 1988; 319: 189-194
        • Farren J.
        • Mitchell-Jones N.
        • Verbakel J.Y.
        • Timmerman D.
        • Jalmbrant M.
        • Bourne T.
        The psychological impact of early pregnancy loss.
        Hum Reprod Update. 2018; 24: 731-749
        • Kalisch-Smith J.I.
        • Moritz K.M.
        Detrimental effects of alcohol exposure around conception: putative mechanisms.
        Biochem Cell Biol. 2018; 96: 107-116
        • Sundermann A.C.
        • Mukherjee S.
        • Wu P.
        • Velez Edwards D.R.
        • Hartmann K.E.
        Gestational age at arrest of development: an alternative approach for assigning time at risk in studies of time-varying exposures and miscarriage.
        Am J Epidemiol. 2019; 188: 570-578
        • Promislow J.H.
        • Makarushka C.M.
        • Gorman J.R.
        • Howards P.P.
        • Savitz D.A.
        • Hartmann K.E.
        Recruitment for a community-based study of early pregnancy: the Right From The Start study.
        Paediatr Perinat Epidemiol. 2004; 18: 143-152
        • Hoffman C.S.
        • Messer L.C.
        • Mendola P.
        • Savitz D.A.
        • Herring A.H.
        • Hartmann K.E.
        Comparison of gestational age at birth based on last menstrual period and ultrasound during the first trimester.
        Paediatr Perinat Epidemiol. 2008; 22: 587-596
        • Moore K.L.
        • Persaud T.V.N.
        • Torchia M.G.
        The developing human: clinically oriented embryology.
        10th ed. Elsevier, Philadelphia, PA2015
        • Dupont W.D.
        Statistical modeling for biomedical researchers: a simple introduction to the analysis of complex data.
        2nd ed. Cambridge University Press, Cambridge, UK2009
        • Greizerstein H.B.
        Congener contents of alcoholic beverages.
        J Stud Alcohol. 1981; 42: 1030-1037
        • Greenland S.
        • Pearl J.
        • Robins J.M.
        Causal diagrams for epidemiologic research.
        Epidemiology. 1999; 10: 37-48
        • Finer L.B.
        • Zolna M.R.
        Unintended pregnancy in the United States: incidence and disparities, 2006.
        Contraception. 2011; 84: 478-485
        • Yelin R.
        • Ben-Haroush Schyr R.
        • Kot H.
        • et al.
        Ethanol exposure affects gene expression in the embryonic organizer and reduces retinoic acid levels.
        Dev Biol. 2005; 279: 193-204
        • Polifka J.E.
        • Friedman J.M.
        Medical genetics: 1. Clinical teratology in the age of genomics.
        CMAJ. 2002; 167: 265-273
        • Muggli E.
        • O’Leary C.
        • Donath S.
        • et al.
        “Did you ever drink more?” A detailed description of pregnant women’s drinking patterns.
        BMC Public Health. 2016; 16: 683
        • Floyd R.L.
        • Decouflé P.
        • Hungerford D.W.
        Alcohol use prior to pregnancy recognition.
        Am J Prev Med. 1999; 17: 101-107
        • Howlett H.
        • Abernethy S.
        • Brown N.W.
        • Rankin J.
        • Gray W.K.
        How strong is the evidence for using blood biomarkers alone to screen for alcohol consumption during pregnancy? A systematic review.
        Eur J Obstet Gynecol Reprod Biol. 2017; 213: 45-52
        • Bailey B.A.
        • Sokol R.J.
        Prenatal alcohol exposure and miscarriage, stillbirth, preterm delivery, and sudden infant death syndrome.
        Alcohol Res Health. 2011; 34: 86-91
        • Ernhart C.B.
        • Morrow-Tlucak M.
        • Sokol R.J.
        • Martier S.
        Underreporting of alcohol use in pregnancy.
        Alcohol Clin Exp Res. 1988; 12: 506-511
        • Alshaarawy O.
        • Breslau N.
        • Anthony J.C.
        Monthly estimates of alcohol drinking during pregnancy: United States, 2002-2011.
        J Stud Alcohol Drugs. 2016; 77: 272-276
        • Rockenbauer M.
        • Olsen J.
        • Czeizel A.E.
        • Pedersen L.
        • Sørensen H.T.
        • EuroMAP Group
        Recall bias in a case-control surveillance system on the use of medicine during pregnancy.
        Epidemiology. 2001; 12: 461-466
        • Sokol R.J.
        • Delaney-Black V.
        • Nordstrom B.
        Fetal alcohol spectrum disorder.
        JAMA. 2003; 290: 2996-2999
        • May P.A.
        • Gossage J.P.
        Maternal risk factors for fetal alcohol spectrum disorders: not as simple as it might seem.
        Alcohol Res Health. 2011; 34: 15-26
        • Muggli E.
        • Matthews H.
        • Penington A.
        • et al.
        Association between prenatal alcohol exposure and craniofacial shape of children at 12 months of age.
        JAMA Pediatr. 2017; 171: 771-780
        • Shuffrey L.C.
        • Myers M.M.
        • Isler J.R.
        • et al.
        Association between prenatal exposure to alcohol and tobacco and neonatal brain activity: results from the Safe Passage Study.
        JAMA Netw Open. 2020; 3e204714
        • Centers for Disease Control and Prevention
        Fetal alcohol spectrum disorders (FASDs): alcohol use in pregnancy.
        (Available at:)