Components of intervention and control groups
The intervention group received alcohol use screening (T-ACE, AUDIT-10 or TLFB) in all studies. In addition to alcohol use screening, 15 (57.7%) studies also provided Motivational Interviewing (MI) (37–39, 42, 43, 45–48, 50, 51, 54, 55, 57, 58), 3 (11.5%) studies used Motivational Enhancement Therapy (MET) and Cognitive Behavioural Therapy (CBT) combined (MET-CBT) (36, 38, 52), and 2 studies used MET alone (41, 49). One (3.8%) study each used MI + CBT (44). In these studies, counseling focused on the importance of alcohol abstinence in pregnancy (35); harm reduction with drink-size assessment (34), or health communication for healthy lifestyle (40); brief advice to reduce alcohol intake (59); non-stigmatizing counseling advising a reduction in alcohol consumption for women not able to abstain completely (56); or a brief discussion with no specific recommendation on alcohol use(53). Counseling in most of the studies (n = 10 studies each, 38.5%) was provided by health professionals: in 4 (15.4%) studies by clinicians/ psychiatrics (36, 37, 41, 59), three studies by nurses (11.5%) (38, 46, 49), two studies by midwives (11.5%) (56, 58), one study (3.8%) by a nutritionist (44); and in 10 studies by trained field researchers (39, 40, 42, 43, 47, 50–54). In terms of the format of the intervention, three (11.5%) studies had self-administered computer-based counselling (45, 48, 57), and 3 (11.5%) studies had both intervention personnel and computer-based counseling (35, 55, 60). Most studies (n = 14) included one single session varying in length from 5–60 minutes (35, 36, 40, 42–47, 53, 54, 57, 59, 60), (Additional File 3).
Of the 25 studies that had a control group, two studies provided controls with no screening or other treatment component at baseline, who were only screened at follow-up to record their change in alcohol use (54, 56). While 23 studies provided their control group with alcohol use screening at both baseline and at follow-up (35–49, 51–53, 55, 57–60). Among these 23 studies, three provided only screening in their control groups with no other treatment component (35, 42, 46) while the remaining 20 (76.9%) studies had other treatment components in combination with the screening (36–41, 43–45, 47–49, 51–53, 55, 57–60). In these studies, the control group received advice or counseling to abstain from or reduce alcohol use or to minimize the impact of drinking during pregnancy on the fetus by: healthcare staff in seven (26.9%) studies (37, 39, 44, 49, 55, 57, 58), or in the form of educational material in the form of brochure/videotape/manual in eight (30.8%) studies (36, 38, 39, 43, 47, 48, 51, 59), or received information regarding local places to assist them with alcohol management in two (11.5%) studies (38, 60). Two studies mentioned providing usual care to the controls, but no detailed information was provided about the components (40, 53). In fact, two (7.7%) studies received more extensive treatment than control groups in other studies (41, 52), at a level of intensity comparable to that of the intervention condition. In one of these studies, for example, the control group received at least 3 sessions of MET from clinicians that were 60 minutes or more in duration (same as the intervention), with the only difference being the intentional removal of some MET principles (e.g., avoiding confrontation, asking open-ended questions, reflective listening) (41). In the other study, both the intervention and control groups were provided with the same CBT treatment (4 sessions, 7 minutes each), but the control group did not receive two post-intervention monitoring calls (52) (Additional File 3).
Changes in antenatal alcohol use
Of 25 studies reporting change in alcohol use pre-post intervention (Additional File 2), only 6 (24%) demonstrated significant reductions in alcohol use (36, 43, 44, 48, 51, 55). A total of 17 of the 25 studies (68%) found no significant changes in alcohol use between BI and control groups (37–39, 41, 42, 45–47, 49, 53–55, 57–60). One study involving adolescent pregnant women reported a substantial reduction in pre-post alcohol use in BI (22.3–13.1%) and controls (2.4–1.7%), without providing between group differences (40). Another study without a control group found that pregnant women with heavy drinking showed a significant drop in mean drinks/week in the second trimester (8.6, P < 0.001), and third trimester (8.1, P < 0.001) after receiving BI compared to baseline (16.0) (50). Finally, a study in Brazil found that both groups receiving BI with 2 weekly monitoring follow-up components (2 monitoring calls by the researcher in the first and second week post-intervention) versus those receiving BI without the monitoring component show higher reduction in mean-AUDIT, and mean T-ACE scores. No comparison for the change provided for between groups difference (52). However, the percentages of abstinent pregnant women observed post-intervention were (92.3%) in the BI alone group compared to (100%) in the BI with monitoring component group. Regardless of the monitoring component, the study highlighted the importance of early intervention (from the first antenatal visit) in pregnancy to achieve significant reduction in prenatal alcohol use.
Meta-analysis of alcohol abstinence post-intervention (BI v/s control)
Meta-analyses of 12 BI arms versus control groups (38, 39, 42, 43, 45, 55–58, 60) for a combined total of 2,620 pregnant women indicate that the BI group has 56% higher odds of being abstinent during pregnancy at any time-point (OR = 1.56, 95%CI = 1.15–2.13, moderate heterogeneity = 46.75%) (Fig. 2).
Please insert Fig. 2 about here -
Meta-analysis of mean AUDIT post-intervention (BI v/s control)
Only 3 studies reported mean AUDIT scores (46, 53, 54) during pregnancy. The pooled estimates of mean AUDIT scores for a total of 610 pregnant women show a small and statistically insignificant difference between the BI group versus the control group (hedge’s g = 0.10, 95%CI = − 0.06 to 0.26, heterogeneity that can be ignored = 0.0%) (Fig. 3).
Please insert Fig. 3 about here -
Meta-analysis of mean drinks/week
The pooled estimates of 166 participants (55) (two intervention arms, one study) observed small and statistically insignificant difference in the mean drinks/week between BI versus control group (Cohen’s d = − 0.21, 95%CI = − 0.78 to 0.36, substantial heterogeneity = 67.24%) (Fig. 4).
Please insert Fig. 4 about here –
Subgroup analysis
As this review found fewer than the required number of studies in meta-analysis for change in prenatal alcohol use and neonatal outcomes, sub-group/moderator analyses were not conducted.
Publication bias and small study effect
The funnel plot for the percentage of alcohol abstinence (Fig. 5) shows asymmetry indicating publication bias. However, the small study effect obtained from Peters test was not significant (P = 0.255), suggesting that smaller studies with larger effect size did not contribute to the publication bias.
Please insert Fig. 5 about here –
Neonatal outcomes (BI versus control)
In total, seven studies reported neonatal outcomes (39, 46, 47, 55, 58). The neonatal outcome measures reported in these studies were: preterm delivery (35); NICU admission (35); healthy pregnancy (live birth of ≥ 2,500 grams with no admission to NICU) (45); mean difference in birth weight (47, 54); mean difference in head circumference (47, 49); body length; Appearance, Pulse, Grimace, Activity, and Respiration (APGAR 1- or 5-minute) (36, 40, 54); percentage of neonates born preterm (38), and LBW (35, 38).
Only 3 studies showed significant difference in the neonatal outcomes between intervention and control groups, two in favour of the intervention and one in favour of the control (35, 47, 49). Armstrong and colleagues observed that the intervention group had 72% lower odds of LBW compared to the control group (OR = 0.28 (0.10–0.80; 0.02) (35). Similarly, Tzilos and colleagues demonstrated a significant differences in birthweight (in favor of intervention) (F(1, 44) = 0.13), P = 0.03), mean birthweight of intervention group versus control was 3189.6 ± = 328.0 and 2965.3 ± 387.7, respectively; d = 0.62 (47). In contrast, Rubio et al. reported significantly lower mean birth weight in the intervention group compared to the controls (3014 grams vs 3160 grams; P = 0.04) (49).
Meta-analysis of mean birth weight
The pooled estimate of difference in the standardized mean difference in birth weight (grams) (Cohen’s d) (47, 49, 54) is small and statistically insignificant when comparing BI (n = 406) and control group (Cohen’s d = 0.16, 95%CI= -0.36 to 0.68, with considerable – substantial heterogeneity = 81.40%), (Fig. 6).
Please insert Fig. 6 about here –
Meta-analysis of low birthweight
The pooled estimate of 2 studies (35, 38) for odds of LBW in the offspring of 1415 mothers studied does not show a significant difference between the BI (cases = 28) versus control group (cases = 33) (OR = 1.02, 95%CI = 0.44 to 2.40, moderate heterogeneity = 59.03%), (Fig. 7).
Please insert Fig. 7 about here –
Meta-analysis of preterm birth
The meta-analysis for 3 intervention arms (2 studies) versus control groups, in 740 participants (35, 38) observe 33% lower odds of preterm birth among pregnant women in the intervention groups (cases = 47) compared to the control groups (cases = 79) (OR = 0.67, 95%CI = 0.46 to 0.98, small heterogeneity that can be ignored = 0.00%), (Fig. 8).
Please insert Fig. 8 about here –