If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Correspondence to: Sarah C.M. Roberts, DrPH, University of California, San Francisco, 1330 Broadway, Suite 1100, Oakland, CA 94612. Phone: 1-510-986-8962; fax: +1-510-986-8960.
Department of Obstetrics, Gynecology & Reproductive Sciences, Bixby Center for Global Reproductive Health, Advancing New Standards in Reproductive Health (ANSIRH), University of California, San Francisco, Oakland, California
Department of Obstetrics, Gynecology & Reproductive Sciences, Bixby Center for Global Reproductive Health, University of California, San Francisco, Oakland, California
Researchers at the U.S. Centers for Disease Control and Prevention (CDC) recently estimated the number of women at risk for alcohol-exposed pregnancies (AEPs) as 3.3 million per month. The number of women at risk was widely interpreted as the actual number of AEPs. The actual number of AEPs may be easier to interpret and may be more a more relevant public health metric for evaluating efforts to reduce AEPs. We estimated the expected actual number of AEPs among U.S. women 15–44 years of age and the expected actual number of alcohol-exposed births (AEBs).
Methods
Like the CDC researchers, we used data about women aged 15–44 years who were neither pregnant nor sterile from the 2011–2013 National Survey of Family Growth. We identified women who had had sex without contraception in the last 4 weeks and reported binge drinking or drinking on more than 7 of the last 30 days. We then estimated the expected actual number of AEPs and AEBs, accounting for the chances of becoming pregnant and for pregnancy outcomes (birth, miscarriage, and abortion). We also conducted sensitivity analyses with varying assumptions.
Results
Estimated prevalences of AEPs and AEBs were 1.2% (95% confidence interval, 0.9–1.7) and 0.8% (95% confidence interval, 0.5–1.2), respectively. During a 1-month period, we estimate 731,000 U.S. women had AEPs and 481,000 resulted in AEBs. Sensitivity analyses indicate expected actual AEP estimates ranging from 104,000 to 1,242,000 and AEBs from 79,000 to 816,000.
Conclusions
Under our assumptions, the estimated expected actual number of AEPs is 2.5 million less than the CDC estimate of the number at risk of an AEP. By using evidence-informed assumptions for the chances of becoming pregnant and common pregnancy outcomes, our estimate of the expected actual number of AEPs is only 22% as large as the CDC's estimate of number at risk, and our estimate of expected actual number of AEBs only 15% as large. The evidence-informed assumptions used here should inform future efforts to estimate expected actual numbers of AEPs and AEBs.
In February 2016, the Centers for Disease Control and Prevention (CDC) released a new report, infographic, and press release that, in effect, advised sexually active women to not drink unless they were using contraception (
). Some commentators argued that the report and infographic caused backlash that diverted attention away from actual harms associated with alcohol use during pregnancy (
Despite the reaction, little media attention was directed at the analysis that formed the basis for the other CDC materials, that is, its estimate that 3.3 million women each month are at risk for alcohol-exposed pregnancies (AEPs) in the United States (
). To arrive at this number, the CDC researchers used data from the 2011–2013 National Survey on Family Growth and considered a woman at risk for an AEP in the past month if she had sex with a male, drank any alcohol, did not use contraception in the past month, and was not sterile or had a partner(s) not known to be sterile. Although the report clearly described their estimate as the number of women at risk for an AEP, reporting about the new estimate as well as some of the CDC's own publicity materials seemed to interpret the 3.3 million as the actual number of AEPs (
). The actual number of AEPs may be easier to interpret and also may serve as a more relevant metric for evaluating efforts to reduce AEPs. Thus, we build on the CDC's approach to provide estimates of what we argue is the more relevant public health metric, the expected actual number of AEPs. We also estimate the expected actual number of AEPs that result in alcohol-exposed births (AEBs), another relevant public health metric. To arrive at these estimates, we made four key modifications to the CDC researchers' approach.
•
The chances of becoming pregnant. The first step in going from the number of women at risk of an AEP to an estimate of expected actual AEPs is to account for the chances of becoming pregnant from unprotected sex. The best available evidence indicates that, among samples of women trying to become pregnant, the peak probability of becoming pregnant from unprotected sex during the fertile window ranges from 0.097 with one act of intercourse (
Alcohol exposure once a pregnancy is established. The second step is including only those who are likely to have consumed alcohol after a pregnancy was established. The CDC estimate included women who had as few as one drink in the previous 30 days. For women with typical menstrual cycles, pregnancy begins with implantation of a fertilized egg around day 22 (
). To expose a pregnancy to alcohol, a woman would have to drink after implantation, meaning in the last 6 days of a cycle. Women who drank infrequently during the same month they became pregnant may not have drunk on days after a pregnancy was established.
•
The level of alcohol consumption considered harmful. The third step includes focusing on levels of alcohol use with clearly documented harm during pregnancy, which are typically addressed in programs the CDC recommends to reduce AEPs (
). The CDC estimate included women who had as few as one drink per month, which is in line with clinical and public health guidance that recommend abstinence from alcohol during pregnancy (
). Although research is unlikely to ever definitely establish whether there is a safe level of alcohol use during pregnancy, there is no conclusive evidence that drinking as few as one drink per month during pregnancy causes harm (
Pregnancy outcomes include birth, miscarriage, and abortion. The fourth step includes incorporating common pregnancy outcomes. Although the CDC researchers discussed AEP estimates as births, the CDC's estimate actually included women with common pregnancy outcomes other than birth, namely, miscarriage and abortion. Based on published estimates of the number of pregnancies, births, and abortions in 2011, estimates for U.S. pregnancy outcomes are 18.3% ending in miscarriage, 17.3% in abortion, and 64.4% in birth (
The main purpose of our study was to establish an estimate of the expected actual number of AEPs and an estimate of the expected actual number of AEBs in the United States, which we argue are the more relevant public health metrics than the number at risk. A secondary purpose was to assess the impact on estimates of different assumptions about the chances of becoming pregnant, the timing and level of drinking, and miscarriage and abortion rates. A third purpose was to provide a methodological approach to estimate of AEPs and AEBs that could be used by other researchers in the future.
Methods
We used data on women of reproductive age (15–44 years) who were neither pregnant nor sterile from the 2011–2013 National Survey on Family Growth to estimate the expected actual number of AEPs and AEBs in the United States. The National Survey on Family Growth sample design has been described elsewhere (
). To account for the complex sample design, we applied post-stratification weights to reflect the female household population of the United States from 2011 to 2013.
We used the same variables as the CDC to determine current pregnancy, sterility, sex with a male in the last 4 weeks, and contraception use at last sex (
; Figure 1). For our main analyses, we used the following variables and approach. For drinking, we created a dichotomous variable of whether a woman consumed alcohol on more than 7 days in the past 30 days or binge drank in the last 30 days. We focused on binge drinking because the program recommended by the CDC to reduce AEPs (Project CHOICES;
) focuses on this level of drinking. We considered focusing on drinking more than 14 days, because it would correspond with drinking on at least one-half of the days in the past month. However, only 5% of the sample who were not pregnant, not sterile, had had sex with a male in the last 4 weeks, and did not use contraception at last sex reported drinking at that level. We thus focus on drinking on more than 7 days in the past month, which corresponds with approximately 10% of those who met the other criteria. Although we recognize that it is an imperfect estimate of the timing of drinking relative to implantation (i.e., after a pregnancy was established), it seemed plausible that someone who drank on more than 7 days in the past 30 days would have had at least one drink after implantation. We categorized women as at risk of an AEP if they binge drank or drank on more than 7 of the last 30 days, reported vaginal sex with a male at least once in the last 4 weeks, and did not use contraception at last sex. We then used a random number generator to randomly assign women in our at-risk sample to become pregnant or not using two probabilities of pregnancy. The first probability is the maximum established in the literature with three or more acts of intercourse during the fertile window (38%; estimate 1;
). Among these randomly assigned pregnancies, we then used a random number generator to assign pregnancy outcomes (miscarriage, abortion, or birth). We categorized women as having an AEB if they met the criteria for an AEP and were randomly allocated to give birth using a national estimate for this pregnancy outcome (64%;
). To develop national estimates, we then multiplied the weighted proportion of at-risk women in each outcome category by the size of the household population the sample is designed to represent. In cases where outcome category numbers were too small, that is, where only one or two women were in that group, we considered population estimates to be unstable. We rounded our population estimates to the nearest 1,000 and report national estimates for AEPs and AEBs with 95% confidence intervals (CIs).
Figure 1Study flow diagram. Numbers represent actual unweighted n in the dataset.
We also conducted a series of sensitivity analyses with varying assumptions. Estimate 3 used a lower chance of becoming pregnant, based on one random act of unprotected sex (4.5%;
). Estimate 4 again used the maximum chance of becoming pregnant (38%), but assumed that women who drank more than seven drinks or binge drank in the past month and became pregnant would be more likely than average to have a miscarriage (
). Estimate 5 used the maximum chance of becoming pregnant (38%), but relaxed the assumption about level of drinking of concern, including women who reported any drinking in the past month, as did the CDC researchers in their original estimates. Estimate 6 was similar to estimate 5, but included only women who reported drinking on more than 3 days in the past month. Finally, we attempted to estimate the expected actual number of AEPs and AEBs among women who became pregnant while using a method of contraception, based on the monthly probability of contraceptive failure among hormonal contraception, intrauterine device, and condom users (
). However, the number of women in the sample was too small to provide stable estimates, given the low probability of becoming pregnant while using contraception.
Results
For estimate 1, among U.S. women aged 15–44 years who were neither pregnant nor sterile, the weighted AEP prevalence was 1.2% (95% CI, 0.9–1.7), or 731,000 women in the past month (Table 1). The weighted prevalences of pregnancy outcomes for women estimated to have an AEP in the past month were: miscarriage 0.2% (95% CI, 0.1–0.5) or 134,000 women; abortion 0.2% (95% CI, 0.1–0.4) or 116,000 women; and birth 0.8% (95% CI, 0.5–1.2) or 481,000 women. For estimate 2 (using a lower probability of becoming pregnant), among U.S. women aged 15 to 44 who were neither pregnant nor sterile, the weighted AEP prevalence was 0.3% (95% CI, 0.2–0.6) or 183,000 women in the past month (Table 1). The weighted prevalence of AEBs was 0.2% (95% CI, 0.1–0.5) or 128,000 births in the past month. The population estimates for miscarriage and abortion for estimate 2 were unstable.
Table 1Estimated Weighted Prevalence and Population Estimates of Past Month AEPs and Pregnancy Outcomes from AEPs among U.S. Women
Estimate 1. Probability of Becoming Pregnant with ≥3 Acts of Intercourse during Fertile Window (0.38)
Estimate 2. Probability of Becoming Pregnant with 1 Act of Intercourse during Fertile Window (0.097)
Estimated Prevalence (%)
95% CI
Population estimate
95% CI
Estimated Prevalence (%)
95% CI
Population estimate
95% CI
AEP
1.2
0.9–1.7
731,000
548,000–1,035,000
0.3
0.2–0.6
183,000
122,000–365,000
AEP ending in miscarriage
0.2
0.1–0.5
134,000
61,000–304,000
0.1
0.0–0.3
UE
UE
AEP ending in abortion
0.2
0.1–0.4
116,000
61,000–244,000
0.0
0.0–0.1
UE
UE
AEB
0.8
0.5–1.2
481,000
304,000–731,000
0.2
0.1–0.5
128,000
61,000–304,000
Abbreviations: AEB, AEP ending in birth; AEP, alcohol-exposed pregnancy; CI, confidence interval; UE, unstable estimate owing to small n in model.
Estimates based on 2011–2013 National Survey on Family Growth data. AEP estimate 1 is women of reproductive age who were not pregnant, not sterile, had sex with a man in the past 4 weeks, did not use contraception at last sex, and drank >7 days or binge drank in the past month. Survey weights were used to get prevalence estimates, which were then multiplied by the population size to get population estimates of the percent of all women of reproductive age having an AEP in the past month. Pregnancy outcomes are allocated based on 2011 estimates of the different pregnancy outcomes.
Sensitivity analyses showed that estimates of AEPs ranged from 103,500 for estimate 3—with the lowest chance of becoming pregnant—to 1,242,000 for estimate 5, which assumed the maximum chance of becoming pregnant and considered any drinking in past month a risk (Table 2). Estimates of AEBs ranged from 79,000 for estimate 3 to 816,000 for estimate 5.
Table 2Sensitivity Analyses Assessing the Impact of Model Assumptions on AEP and AEB Estimates
As a way to offer what we argue are more relevant public health metrics regarding AEPs, this study estimated the expected actual number of AEPs and AEBs and compared these estimates with the CDC's estimate of the number of U.S. women at risk of an AEP. Under our assumptions, the estimated expected actual number of AEPs in the United States is 2.5 million fewer and AEBs 2.8 million fewer women per month than the CDC researchers' estimate of the number of women at risk of an AEP. For our estimates, we made three assumptions notably different from the CDC analysis. First, we only included women likely to have drunk alcohol after pregnancy was established and women binge drinking regardless of frequency. Second, we only included women likely to have become pregnant during a month when both drinking and unprotected sex occurred. Third, we accounted for three common pregnancy outcomes: miscarriage, abortion, and birth. Under our evidence-informed assumptions, we arrived at an expected actual AEP estimate that is only 22% as large and an expected actual AEB estimate that is only 15% as large as the CDC researchers' estimate of women at risk. Our findings strongly suggest that it is not appropriate to interpret estimates from the CDC's at-risk approach as the actual number of AEPs.
We also conducted sensitivity analyses to assess the impact of different assumptions on expected actual AEP and AEB estimates. We found that the estimates are not robust to these assumptions. Although our lowest estimate of AEPs (104,000) is 3% of the CDC's estimate of women at risk for an AEP, even our maximum estimate of 1.2 million expected actual AEPs is approximately one-third of the CDC's estimate of women at risk. More research can better inform the accuracy of assumptions underlying estimation of AEPs and AEBs. Until then, estimates should be reported as a range, and the assumptions underlying estimates should be explicit. Our main estimates of expected actual AEPs and AEBs (estimate 1) are solidly in the middle of the range of our sensitivity analyses.
Our desire to estimate the expected actual numbers of AEPs and AEBs does not question the importance of alcohol use during pregnancy as a public health problem or deny that public health efforts should be expended on this problem. We are concerned that the CDC estimates have been incorrectly interpreted as actual AEBs, which overestimates the scope of the problem. Overestimating the scope of the problem can contribute to moral panic about alcohol use during pregnancy (
), which can in turn lead to more stigmatizing and punitive policy approaches to alcohol use during pregnancy. Research shows that stigmatizing and punitive policy approaches lead to increases in adverse birth outcomes and less prenatal care use (
Associations between state-level policies regarding alcohol use among pregnant women, adverse birth outcomes, and prenatal care utilization: Results from 1972-2013 Vital Statistics.
Alcoholism, Clinical and Experimental Research.2018; 42: 1511-1517
). We are also concerned that the at-risk estimates are not specific to AEPs, but could rather describe the size of the population at risk for any type of risk related to unprotected sex (e.g., a sexually transmitted infection). Finally, we are concerned that the CDC's decision to not account for the chances of becoming pregnant from unprotected sex, when a pregnancy is established, or the proportion of pregnancies that result in a birth reinforces inaccurate information about pregnancy and abortion.
Furthermore, ignoring the role of abortion in AEP outcomes undermines the ability of public health programs to effectively address alcohol consumption during pregnancy. Ensuring that pregnant women who drink alcohol are neither coerced to end their pregnancies nor continue them is of the utmost ethical importance. Yet it is inappropriate to ignore the fact that some pregnant women who have an AEP will have an abortion, because it overestimates the number of AEBs. Ignoring the role of abortion may also indirectly contribute to policies that restrict abortion access for pregnant women who use alcohol, especially in problematic patterns. This is of concern because evidence shows that restricting pregnant women's ability to obtain abortions does not lead them to stop drinking in problematic patterns (
In terms of a way forward, this approach to studying AEPs and AEBs could be improved on in the future. Estimates could be more precise if we had large enough samples to account for age, race/ethnicity, or education-specific chances of becoming pregnant and pregnancy outcomes. Research could provide more precise estimates of the chances of becoming pregnant among women who drink more heavily and the chances of giving birth among women who have an AEP. However, greater precision for these models may be moot if we do not address the larger questions about what timing and quantity of alcohol use during pregnancy should be the focus of public health intervention and the role of abortion in relation to alcohol use during pregnancy.
There are limitations to our analyses. First, some factors likely affect the precision of our estimates. We used overall rather than age-specific chances of becoming pregnant from unprotected sex and proportions of pregnancy outcomes, despite chances of becoming pregnant varying by age (
). Although more precise estimates would account for age-specific chances of becoming pregnant and giving birth, the estimates we currently report are not robust to model assumptions. Thus, adding age-specific chances of pregnancy would not necessarily make estimates more accurate. For similar reasons, we did not account for race/ethnicity- or education-specific chances of becoming pregnant, having a miscarriage, or having an abortion, which could also contribute to a lack of precision in our estimates. Second, our main estimates of expected actual AEPs and AEBs assumed that all acts of unprotected sex occurred during the fertile window. The impact of assuming all acts occurred during the fertile window can be seen in sensitivity analyses (estimates 2 and 3), with AEPs 25% and 14% lower than our main estimate. Third, we, like the CDC, did not account for women who became pregnant while using contraception (
). If results from this research show that alcohol use before ovulation, fertilization, or implantation causes harms, we would need to revisit our assumptions about the timing of drinking that affects a pregnancy and, thus, our estimates.
Implications for Policy and/or Practice
Although the updated estimates of expected actual AEPs and AEBs are lower than the estimates of women at risk for an AEP that the CDC published and publicized, they are still substantial and do not indicate that there needs to be a change in the amount of public attention to reducing harms owing to alcohol use during pregnancy. Rather, they point to the importance and relevance of basing estimates of the scope of public health problems on the best available scientific evidence, even when there may be political controversy about the topics—such as abortion—covered in the scientific evidence.
Conclusions
By more realistically accounting for conception probabilities and common pregnancy outcomes, our estimates of the expected actual numbers of AEPs and AEBs are lower than CDC estimates of women at risk, although still substantial. These estimates provide an important alternative to the CDC estimates and the approach we used should inform future efforts to estimate AEPs and AEBs, which may be more relevant public health metrics.
Acknowledgments
For editing help, we thank Nicole Nguyen, MPH (UCSF ANSIRH), and Erin Wingo, MSPH (UCSF ANSIRH). For assistance with probabilities of conception, we thank Diana Greene Foster, PhD (UCSF ANSIRH). For assistance with information about research about effects of alcohol exposure prior to conception, we thank Lyndsay Ammon Avalos, PhD, MPH (Kaiser Permanente Northern California Division of Research).
Associations between state-level policies regarding alcohol use among pregnant women, adverse birth outcomes, and prenatal care utilization: Results from 1972-2013 Vital Statistics.
Alcoholism, Clinical and Experimental Research.2018; 42: 1511-1517
Sarah C.M. Roberts, DrPH, is Associate Professor at University of California, San Francisco. Dr. Roberts studies how policies and our health care system punish rather than support vulnerable pregnant women, including women seeking abortion and women who use alcohol and/or drugs.
Kirsten Thompson, MPH, is Program & Communications Director at the UCSF Bixby Center for Global Reproductive Health. Ms. Thompson works to improve sexual and reproductive health and rights through creative solutions and new technologies, including research, communications, advocacy, and design.
Article info
Publication history
Published online: January 13, 2019
Accepted:
November 5,
2018
Received in revised form:
October 25,
2018
Received:
April 20,
2018
Footnotes
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
30242-1&id=gr1.jpg){kind=link}