Measurement of key outcome variables – particularly gestational age – changed over time as well

Analyses of APGAR scores were for the years 1978-2013 because APGAR scores were not reported on birth certificates prior. Cases missing outcome data were typically dropped from analyses. The main exposure variables were time-varying state-level indicators regarding whether states had particular policies in the month and year of conception. These policies were: Mandatory Warning signs, Priority Treatment for Pregnant Women, Priority Treatment for Pregnant Women and Women with Children, Reporting Requirements for Data and Treatment Purposes, Prohibitions on Criminal Prosecution, Civil Commitment, Reporting Requirements for Child Protective Services Purposes, and Child Abuse/Child Neglect. These policies have been detailed elsewhere and are briefly described in Table 1. The first policies, Reporting Requirements for CPS and Child Abuse/Child Neglect, went into effect in Massachusetts in 1974. Next, Washington DC adopted Mandatory Warning Signs in 1985 and Kansas adopted Reporting Requirements for Data and Treatment Purposes in 1986. In 1989, California established Priority Treatment for Pregnant Women, and both Florida and Washington established Priority Treatment for Pregnant Women and Women with Children. Kentucky, Missouri, and Virginia put Prohibitions on Criminal Prosecution into effect in 1992. South Dakota and Wisconsin established Civil Commitment in 1998. All policies were still in effect in at least four states in 2013. Each policy indicator variable is dichotomous, coded as 0 if it was not in effect for that state in the month/year of conception and 1 if it was in effect for the month/year of conception. Linking the policy indicators to the month and year of conception improves the accuracy of exposure timing . Models controlled for both individual-level maternal characteristics and for state level characteristics and policies in effect during the pregnancy.

Individual-level maternal characteristics included maternal age, race, marital status, education, nativity,growing cannabis and parity. If data for individual-level controls were missing, we created a missing category to include all available data. Version of birth certificate was also included as an indicator variable. State-level controls included state- and year- specific poverty, unemployment, per capita cigarette consumption, and per capita total ethanol consumption, as well as indicators for whether government control of wine sales and government control of spirit sales were in effect for that state in that year. Data for state-level controls came from secondary sources, including the U.S. Census, the U.S. Centers for Disease Control and Prevention, APIS, the National Highway Traffic Safety Administration, National Beverage Control Association, and published research . State-level per capita cigarette consumption and per capital alcohol consumption were included because these variables could not be controlled at the individual-level due to lack of data documented on birth certificates in the earlier years and concerns with the quality of these data in the later years . Multi-variable logistic regression was used for all outcomes. Regression models included all policy indicators simultaneously, fixed effects for state and year, state specific cubic time trends, and adjusted for both individual and state-level control variables. Regression models also accounted for clustering of standard errors according to mother’s state of residence. Taking the most conservative approach, analyses included year fixed effects and birth certificate version indicator variables to account for changes in Vital Statistics data gathering over time as well as other relevant events in those states and years. State-specific cubic time trends were added to address possible concerns with endogeneity. All analyses were performed in Stata v14.2. Sensitivity analyses We performed a number of sensitivity analyses post hoc. First, we assessed each policy individually in multi-variable regression models and found no differences compared to models including all policies simultaneously. Second, because information regarding Hispanic ethnicity was not available until 1989, we analyzed data for births for 1989-2013 separately using a combined race/ethnicity variable; results did not change.

Finally, we fit both the preliminary and final models using a 10% sample of the full dataset, and compared these results to those from the full dataset; results did not differ between the 10% sample and full datasets. This is the first study to comprehensively assess whether state-level policies targeting alcohol use in pregnancy are related to adverse birth outcomes, outcomes that indicate measurable harms due to alcohol use during pregnancy. We find that most policies targeting alcohol use during pregnancy – MWS, CACN, CC, PCP, RRDATA, and PTPREG – appear associated with increased adverse birth outcomes, possibly due to some of these policies leading women to avoid prenatal care. In addition, it appears that generally applicable alcohol policies – specifically retail control of wine sales and any other policies that lead to decreased population-level consumption – are associated with improved birth outcomes. Although the magnitudes of effects are generally small, they are still meaningful in such a large population. Overall, these findings do support our hypotheses that policies punishing alcohol use during pregnancy are associated with increased adverse birth outcomes and may lead to avoidance of prenatal care. They do not, however, support our hypothesis that the more supportive policies – including Mandatory Warning Signs – are associated with decreased adverse birth outcomes. They also are inconsistent with our expectation that supportive policies would be unlikely to be associated with prenatal care utilization. With a few exceptions , scholars have consistently distinguished policies targeting substance use during pregnancy as either supportive or punitive; our study findings do not support this distinction. Rather, our findings suggest that state level policies targeting alcohol use during pregnancy at best do not improve birth outcomes and, at worst, lead to increases in adverse birth outcomes and lead women to avoid prenatal care. This pattern of findings is not completely surprising for three key reasons. First, qualitative research has found that information that leads women to worry that their substance use has already irreversibly harmed their fetus leads women to avoid prenatal care .

Similarly, our findings suggest that rather than providing women with information that helps them change their behavior and engage with health care services that may support such behavior change, MWS may operate by scaring women and leading women to avoid such help. Second, this same previous qualitative research has found that policies related to CPS and child removal lead women to avoid prenatal care. Our findings related to CACN policies are consistent with this previous research, and extend prior findings by indicating that this avoidance of prenatal care may be linked to worse birth outcomes. This is crucial,cannabis grow tray as ongoing research on alcohol outcomes has found some associations between states with CACN and less alcohol use during pregnancy . The current analyses show that even though defining alcohol use during pregnancy as child abuse/neglect is associated with decreases in self-reports of binge and heavy alcohol during pregnancy, this does not translate to better birth outcomes. Third, and perhaps most vitally, previous research indicates that policy making related to alcohol use during pregnancy appears more related to policy making in the area of reproductive rights than to policy making that reduces public health harms from alcohol use in the population overall . This means that the problem of alcohol use during pregnancy likely has not benefited from the same public health policy development process used to address public health harms from alcohol use in the general population. The current results show that reduced population-level alcohol consumption and government control over wine retail sales are associated with improved birth outcomes, which is in line with previous studies; therefore policymakers and public health professionals who wish to improve birth outcomes through state-level policies targeting substance use should look to the broader alcohol policy field for lessons and approaches, rather than continuing with the types of policies currently in effect. We do note that some of the patterns of findings are more difficult to understand. For example, the policy that mandates priority treatment for pregnant women was related to lower odds of inadequate PCU, but higher odds of low birthweight, premature birth, and late PCU. These mixed findings could be because the policy indicator does not capture actual treatment availability. States prioritizing treatment for pregnant women might have fewer treatment slots than states without such laws, meaning our finding could be just an indication of lack of treatment availability; future research should examine this. Laws giving pregnant women priority could also prevent women from getting treatment prior to becoming pregnant, especially in states with limited treatment availability. Similarly, laws giving pregnant women priority might prevent other people – including partners of women who become pregnant – from getting treatment, leading to adverse birth outcomes due to harms from others’ drinking. Our findings are inconsistent with the only other published study that examined associations between MWS and adverse birth outcomes across both states and time. In that study, MWS were associated with decreased odds of very low birthweight and very preterm birth .

This discrepancy could be because the previous study 1) only examined MWS without accounting for other policies; 2) used data only for the years 1989-2006; 3) examined different outcomes; 4) only used a subset of states; 5) did not link policy data to individual outcomes based on the month of conception; 6) controlled for state-level policies alcohol and tobacco policies and not actual per capita consumption; and/or 7) controlled for individual-level alcohol use data from birth certificates, which are of poor quality and which could be more likely to be assessed and documented in cases of adverse birth outcomes. Notably, our post hoc sensitivity analyses of race/ethnicity only utilized the years 1989-2013, indicating that the discrepancies between our findings and Cil’s probably are not due to the different time frames. Strengths and Limitations This is the first study to examine all policies related to alcohol use in pregnancy simultaneously across all 50 states using a time frame long enough to capture the period before any laws were enacted . Furthermore, for most of the time frame the data include the entire population of singleton births born in the United States and for the years 1972-1984 include a 50% sample, which makes questions regarding inference and generalizability essentially irrelevant. Another major advantage of these data over, for example, survey data regarding alcohol use during pregnancy, is that biases due to self-report are not present here. Finally, our results were robust across various model specifications, further strengthening our conclusions. The main limitation of this study is that Vital Statistics birth certificate data are not collected for research purposes; therefore, we cannot adjust for maternal-level alcohol or tobacco use. Although maternal alcohol and tobacco use have been recorded on birth certificates since 1989, these data have been shown to be invalid We adjusted for state-level alcohol and tobacco consumption instead. Another limitation is that race has been measured inconsistently on birth certificate data over time. Only in 1989 did states begin to document ethnicity as well as race, although this was phased in over the 1990s. Our primary analyses did not account for ethnicity, e.g. White Hispanic and White Non-Hispanic women are in a single group. Such an approach is reasonable because birth outcomes are similar between White non Hispanic and Hispanic births, both of which differ from Black birth outcomes. We applied approaches developed later to correct for implausible gestational age values to earlier years of Vital Statistics to improve consistency. Also, for these analyses, we focused specifically on policies targeting alcohol use during pregnancy. Preliminary examinations of these policies suggest that many of them may also address drug use. Future research is needed to explore whether the findings generalize to policies targeting drug use during pregnancy.Stress-induced analgesia is mediated by the activa tion of endogenous pain inhibitory systems. Both opioid dependent and opioid-independent forms of SIA have been identified . These mechanisms are differentially activated according to stressor parameters and duration . SIA elicited by intermittent foot shock is blocked by opioid antagonists , whereas SIA elicited by continuous foot shock is blocked by cannabinoid antagonists . We recently demonstrated that this nonopioid form of SIA is mediated by mobilization of two en docannabinoids, 2-arachidonoylglycerol and ananda mide, in the dorsal midbrain . Opioid and nonopioid SIA share similar neuroanatomical substrates. For example, opioid and cannabinoid receptors populate brain regions regulating nociceptive responding, such as the periaqueductal gray and the raphe nuclei of the medulla. Like opioids, cannabinoids modu late distinct circuits within the midbrain PAG and the brainstem rostral ventromedial medulla .