Prenatal alcohol exposure and infant gross motor development: a prospective cohort study

Data were from the Triple B Pregnancy Cohort Study, a prospective Australian study of 1634 pregnant women recruited in 2009–13 at antenatal clinics in NSW (n = 1305) and WA (n = 318) [23]. Ethical approval was granted by university and hospital Human Research Ethics Committees. Eligibility criteria included: pregnancy; aged ≥16 years; no major medical complications (mother/fetus); mother/both parents the primary caregiver/s; mentally able to complete assessments; English literary; and informed consent.

Data were collected in pregnancy at T1a, T1b, T2, T3, and at 8-weeks, and 12-months post-birth. By separating T1 into two periods the effects of early PAE could be examined. Most women reported pregnancy awareness around 5–6 weeks gestation [24]; as such, a six-week split broadly represents pre- and post-pregnancy awareness. Of the participants with GM outcome data, we excluded women with inconsistent drinking behaviour (n = 141) and who had infants with very low birth weight (< 1.6 kg; n = 1). For women who gave birth to twins/triplets (n = 34), one child was selected at random for analysis. The final sample comprised 1324 participants (755 with participating partners).

Study measures are described in Table 1.

Analyses were conducted using STATA 14 and SPSS 20 [30, 31]. Missing data was accounted for using multiple imputation [32, 33]. There were four stages of analysis. First, we described maternal drinking patterns (low, moderate, binge and heavy drinking versus abstinence) at all pregnancy timepoints (T1a, T1b, T2 and T3), and partner drinking patterns at T3. Second, binary logistic regression models estimated the association between mother, infant and partner-related factors and PAE (i.e., any drinking in pregnancy versus abstinence).

In the third stage, logistic regression analyses examined the relationship between PAE (T1a, T1b, T2 and T3) and infant GM outcomes at 12-months, controlling for background socio-demographics, and other potential maternal, infant and partner confounders. At T1a all PAE categories were examined. However, due to the low frequency of moderate, binge and heavy drinking in the sample for T1b, T2 and T3 (see Table 2), only low PAE was examined. The reference category was abstinence. The primary outcome was the BSID-III scaled GM score. The unadjusted results were examined first, followed by increasing levels of adjustment for potential confounders. The adjusted analyses included the maternal socio-demographic background factors, followed by factors found to be associated with maternal drinking at the univariate level (p < .10) [34].

Finally we examined whether the effect of PAE may differ according to an individual’s risk of being exposed to alcohol based on their baseline characteristics [35]. To do this we calculated the propensity of a woman to consume alcohol at low-levels. We then stratified the sample into groups indicating higher and lower risk of low-level alcohol exposure based on their propensity score and compared infant GM outcomes between these groups (see Additional file 2: Section B for a detailed description).

Most women reported alcohol use at some point in pregnancy (61.2%; Table 2). Among T1a drinkers, low-level use was most frequently endorsed (22.0%), followed by binge (15.9%) and heavy drinking (14.0%), respectively. There was a marked change in drinking patterns in T1b. Notably, abstinence increased from 38.8 to 71.2%, and of those who did report drinking alcohol, most did so at low-levels (16.9%). Binge and heavy drinking decreased to 3.1 and 1.8%, respectively. This trend remained consistent through T2 and T3, although some women did return to low-level drinking as their pregnancy progressed (low-level drinkers: T2, 25.6%; T3, 25.5%).

Additional file 1: Table S1 shows the pattern of alcohol use reported by partners. Among those who drank (85.6%), binge drinking was most common (27.9%), followed by low (20.3%), moderate (14.4%) and heavy drinking (13.3%), respectively.

Univariate tests compared whether abstainers and pregnancy drinkers (at any level) differed on background socio-demographics, other substance use, and physical and psychological factors (Table 3). The results show that, relative to abstainers, women who drank alcohol had greater odds of being older (e.g., 30–35 years, 1.97, 95% CI, 1.20–3.24); completing high school (2.61, 95% CI, 1.48–4.61); having moderate (2.29, 95% CI, 1.31–4.02) or high SES (4.42, 95% CI, 2.56–7.64); being born in an English speaking country (1.88, 95% CI, 1.33–2.66); living in a household with multiple parents (single parent: 0.61, 95% CI, 0.39–0.95); and speaking English as their first language (2.34, 95% CI, 1.77–3.09); and lower odds of living in a single parent household (0.61, 95% CI, 0.39-0.95). Other factors associated with pregnancy drinking included: smoking in pregnancy (1.67, 95% CI, 1.18–2.36); reduced anxiety (0.76, 95% CI, 0.57–0.99); and higher estimated IQ (e.g., a score of 100–114, 3.02, 95% CI, 2.01–4.53).

Univariate tests also compared whether infants of abstainers and pregnancy drinkers (at any level) differed on sex and birth indicators (Table 3). Compared to infants of abstainers, infants born to mothers who drank in pregnancy were less likely to be born preterm (< 36 weeks gestation; 0.05, 95% CI, 0.3–0.82). No significant differences were found in sex, birthweight, head circumference or Apgar scores.

Additional file 1: Table S2 shows the results of univariate tests comparing the characteristics of partners of abstainers with partners of pregnancy drinkers. Compared to partners of abstainers, partners of drinkers had three-fold greater odds of being low-level drinkers (2.99, 95% CI, 1.79–5.00); four-fold greater odds of being moderate drinkers (4.27, 95% CI, 2.40–7.63); six-fold greater odds of being binge drinkers (6.08, 95% CI, 3.65–10.12); and almost eight-fold greater odds of being heavy drinkers (7.78, 95% CI, 4.06–14.91). Partners of drinkers were older (e.g., 30–35 years, 2.89 95% CI, 1.52–5.52), less likely to report a non-English speaking background (0.49, 95% CI, 0.34–0.72); and more likely to report English as their first language (2.08, 95% CI, 1.40–3.11), compared to partners of abstainers.

Regression analyses were used to examine the relationship between PAE exposure and infant GM development at 12-months (Table 4).

For T1a, in the unadjusted analyses, there were no significant associations between PAE and GM outcomes. This relationship remained unchanged after adjustment. For T1b, T2 and T3, the results were consistent when PAE exposure was binary (i.e., abstinence versus low-level drinking); namely, low PAE was not significantly associated with GM development in infants at 12-months in the unadjusted, nor in the adjusted analyses. Table 5 shows the marginal means and 95% CIs for PAE and GM outcomes at all levels of adjustment.

As paternal factors have been associated with maternal drinking in pregnancy, a second series of regression analyses were conducted within a sub-sample of the women whose partners participated in the study (Additional file 1: Tables S3 and S4). Again, PAE was not significantly associated with GM development at 12-months (See Additional file 1: Section A for sensitivity analyses).

The sample was then stratified into two levels of risk based on the propensity score matching (see Additional file 2: Section B). Highest risk of alcohol exposure was found to be related to factors including higher SES, higher education, older age, being of English speaking origin, tobacco use, and unplanned pregnancy. In the highest risk subgroup, 122 abstainers were matched to 195 drinkers. In the lowest risk subgroup, there were 186 abstinent women matched to 117 drinkers. Significant differences were only observed between drinkers and abstainers in the highest risk group (t = − 2.92, p = 0.004), with children born to low-level drinkers having higher GM scores (M = 9.20, SD = 2.85) compared to those born to abstainers (M = 8.82, SD = 2.52) (see Additional file 2: Section B Figure S1).

Pregnancy drinking was common in this cohort: 56.1% of pregnant women drank in Trimester one, 0–6 weeks (T1a); 23.6% in Trimester one, 7–12 weeks (T1b); 30.5% in Trimester two (T2); and, 29.6% in Trimester three (T3), respectively. Most women reported drinking at low-levels (average 22.5% across pregnancy). With the exception of T1a, few women reported moderate (2.3%), binge (1.7%) or heavy (1.3%) drinking in pregnancy. Whilst this is consistent with past research, greater specificity in this cohort on PAE timing and dose highlights two findings of public health import [36, 37]. First, rates of drinking prior to pregnancy awareness are close to twice that following awareness; and second, binge and heavy drinking both occur at high rates within the pre-awareness period: 15.9 and 14.0%, respectively. Taken together, the very earliest period of pregnancy may be one of greatest risk of exposure to alcohol.

Consistent with past research, pregnant women who consumed alcohol differed on socio-demographic characteristics compared to abstainers [36, 37]. Specifically, they were more likely to be older, tertiary educated, have moderate to high SEIFA scores (reflective of socio-economic advantage), be born in Australia or another English speaking country, and less likely to live in a multiple parent household. Other factors associated with pregnancy drinking included: smoking in pregnancy; reduced anxiety; higher estimated IQ; and lower odds of obesity. These results suggest pregnancy drinking is common among women from more affluent socio-demographic backgrounds, and among specific at-risk groups such as women who smoke cigarettes. Targeting these populations may result in more effective preventive intervention for pregnancy drinking.

With respect to partners, our results are consistent with the limited extant literature [38, 39]. Specifically, partners of pregnant women who drink were more likely to be from advantaged socio-economic backgrounds (higher SEIFA scores and educational attainment), and to also drink alcohol and smoke tobacco. These characteristics may affect offspring development via their influence within the familial environment (i.e., partner drinking increases the risk for maternal drinking) [40]. Few studies have accounted for these potential influences when examining associations between PAE and offspring development [38].

The third aim was to determine whether GM development was impaired among infants exposed to PAE compared with infant offspring of abstainers. Potential background socio-demographic confounders were included in the adjusted analyses, along with other potential confounders (maternal substance use, physical and psychological factors; infant factors) associated with PAE exposure in the univariate analyses. Finally, to account for the potential role of partner factors in determining infant outcomes, we re-ran the models in a sub-set of the sample for whom partner data were available. We found no evidence to suggest that low PAE was associated with measurable impairments in infant GM development at 12-months. Moreover, at T1a (0–6 weeks), prior to pregnancy awareness for most women, neither moderate, binge nor heavy drinking predicted measurable GM impairment. In all models, and at all levels of adjustment, the pattern of results remained unchanged.

With regard to low-level exposure, the results are consistent with a small number of existing studies showing GM impairment was not linked to low PAE [20, 21]. There are a number of plausible explanations for this finding. First, that low PAE does not have a deleterious effect on early GM development. Alternatively, if harm does occur, the effects are likely to be very small. Our current measurement instruments, even though gold-standard, may not be sufficiently sensitive to detect such small GM deficits. Finally, it is well documented that infant development may fluctuate during infancy [41]. If low PAE does have harmful effects on GM development, these effects may be more readily detected in childhood as GM skills stabilise.

Importantly, when the sample was stratified by propensity score, the relationship between alcohol exposure and GM outcome was different relative to women’s baseline characteristics. In the highest risk subgroup, significant differences were observed between low-level drinkers and abstainers, with children born to low-level drinkers having higher GM scores compared to those born to abstainers. Consistent with other epidemiological findings [35], this result is suggestive of a potential interaction between PAE and other demographic and maternal risk factors in relation to offspring development, Further assessment of this interaction is recommended in samples with greater representation of women from low SES and high-risk backgrounds.

The finding that neither binge, moderate nor heavy PAE were linked to poorer GM development is inconsistent with a number of studies [2]. We did not examine potential harms after T1a due to the low frequency of these drinking patterns. It is possible that harmful drinking patterns later in pregnancy or persistent patterns of harmful drinking across the gestational window may be associated with GM impairment. It has been documented that there are sensitive gestational periods where risk for negative outcomes from teratogens may be heightened [42]. Differential impacts (relative to exposure timing) may explain inconsistency in the literature, and why it is difficult to determine a specific threshold at which PAE is harmful or safe for fetal development. Future work with greater representation of moderate, binge and heavy drinking patterns across the gestational period is needed, either through new cohorts with larger samples, targeted samples of moderate to higher-risk drinkers, and/or through potential data pooling/harmonisation [43].

There are a number of limitations. First, women with low SES backgrounds (and their partners) were underrepresented, and those who were included tended to be abstainers rather than drinkers. As such, this study may not have captured low SES and marginalised families, whose children may be most susceptible to harms relating to PAE [44]. When the sample was stratified by propensity score, the relationship between alcohol exposure and GM outcome was different for women with different patterns of baseline characteristics. Results should thus be interpreted with the caveat that associations between alcohol use and GM outcomes may show a different pattern among lower SES or high-risk populations. Targeted recruitment of these disadvantaged families may result in better representation of the effects of heavier PAE patterns across a range of demographics. Obtaining sufficient representation of heavier PAE post-awareness is difficult due to the reduction that occurs in drinking. Data pooling and/or harmonisation of existing cohorts might be one approach to address this issue [43]. Nevertheless, this study did have good representation of varying PAE within the first six weeks of pregnancy, particularly low PAE from T1a to birth, suggesting that the cohort was well-suited to the assessment of impacts of low PAE. Second, despite the use of the gold-standard, clinically administered BSID-III, GM skills can fluctuate in infancy [41]. Potential PAE effects not identified in this study may emerge as the cohort offspring develop and their capacities stabilise. Thus, follow-up of the cohort to assess development into childhood is important.