Online-based antenatal education and its effects on maternal mental health and selected neonatal outcomes: a systematic review and meta-analysis

Abstract

Background

Online antenatal education is increasingly used to improve maternal psychological outcomes and preparedness for childbirth. However, evidence regarding its effectiveness remains inconsistent, particularly due to heterogeneity in intervention types and outcome measures.

Methods

This systematic review and meta-analysis was conducted in accordance with PRISMA 2020 guidelines. PubMed, EMBASE, Scopus, Web of Science, CINAHL, and the Cochrane Library were searched from inception to March 2026. Randomized controlled trials and observational studies evaluating digital antenatal interventions in pregnant women were included. Outcomes included maternal depression, anxiety, fear of childbirth, self-efficacy, and small-for-gestational-age (SGA) incidence. Random-effects models were used, and standardized mean differences (SMDs) and odds ratios (ORs) were calculated. Heterogeneity was assessed using I² statistics.

Results

Twelve studies involving 4,982 participants were included. No significant effects were observed for depression (SMD = − 0.18; 95% CI: −0.45 to 0.09; I² = 86.5%; p = 0.188), anxiety (SMD = − 0.10; 95% CI: −0.66 to 0.46; I² = 92.8%; p = 0.723), or self-efficacy (SMD = 0.56; 95% CI: −0.11 to 1.23; I² = 90.6%; p = 0.102). A reduction in fear of childbirth did not reach statistical significance (SMD = − 0.53; 95% CI: −1.06 to 0.003; p = 0.051. No significant association was found for SGA (OR = 0.73; 95% CI: 0.17–3.14; p = 0.670). Substantial heterogeneity was present across outcomes, limiting the interpretability of pooled estimates.

Conclusion

Current evidence, characterized by substantial heterogeneity, does not demonstrate statistically significant effects of online-based antenatal education on maternal depression, anxiety, or fear, nor on improving self-efficacy or preventing SGA infants. However, the high heterogeneity (I² >85% for most outcomes) indicates that these pooled estimates are exploratory, and clinically meaningful effects for specific intervention types cannot be ruled out. Findings should be interpreted cautiously.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-27571-2.

Introduction

Maternal preparedness is a critical factor in ensuring positive outcomes for both mothers and infants during pregnancy, labour, and the postpartum period [1]. Despite increasing implementation of digital antenatal programs, existing studies report inconsistent findings regarding their effectiveness. Some trials demonstrate modest improvements in maternal psychological outcomes, while others report null effects. These discrepancies may reflect differences in intervention design, theoretical frameworks, participant characteristics, and outcome measurement tools. Globally, antenatal anxiety and depression affect approximately 10–25% of pregnant women, with higher prevalence reported in low-resource settings. These conditions are associated with adverse maternal and neonatal outcomes, including preterm birth and low birth weight. Importantly, many digital interventions labeled as “antenatal education” incorporate fundamentally different components, including cognitive behavioral therapy, mindfulness-based training, or general informational modules, thereby complicating interpretation of pooled evidence [2, 3].

Online antenatal education encompasses a range of digital interventions, including interactive webinars, video modules, virtual support groups, and mobile applications [4]. These platforms are designed to provide standardized information about pregnancy, labor, pain management, breastfeeding, postpartum care, and neonatal health [5]. By leveraging digital technology, online antenatal education programs can overcome traditional barriers associated with in-person classes, such as geographic limitations, scheduling conflicts, and resource constraints [6]. This increased accessibility is especially important for women in rural or underserved areas, who may otherwise have limited access to comprehensive prenatal education [7].

One of the key advantages of online antenatal education is its ability to deliver information in a flexible and learner-centered manner [8]. Expectant mothers can access educational content at their own pace, revisit challenging topics, and tailor their learning experience to meet individual needs [9]. This self-directed learning model is further enhanced by the interactive nature of many online platforms, which often include quizzes, discussion forums, and real-time feedback [10]. Such features not only reinforce knowledge acquisition but also foster a sense of community and shared experience among participants, which can be particularly reassuring during a time of significant physical and emotional change.

Moreover, online education platforms can continuously update content to reflect the latest research and best practices in maternal care. This dynamic approach ensures that pregnant women receive current and evidence-based information, which is essential for making informed decisions about their care. In addition, the use of multimedia resources—such as videos, animations, and infographics—can enhance understanding and retention of complex concepts, such as the stages of labor or neonatal resuscitation techniques [11]. As a result, women who engage with online antenatal education are more likely to develop realistic expectations about childbirth and acquire practical skills for managing the challenges of the perinatal period [12].

Several studies have indicated that increased maternal preparedness is associated with improved psychological outcomes, such as reduced anxiety and higher self-efficacy, as well as better clinical outcomes, including shorter labor durations and decreased rates of medical interventions during childbirth [13, 14]. Online antenatal education has the potential to influence these outcomes by equipping women with comprehensive knowledge about the birthing process, pain management strategies, and postpartum recovery [15, 16]. This preparation can empower mothers to actively participate in the decision-making process during labor and delivery, leading to a more satisfying and controlled childbirth experience. Furthermore, well-prepared mothers may be better equipped to navigate unexpected complications, thereby reducing the likelihood of adverse outcomes.

Given this landscape of inconsistent findings, a clear synthesis of the available evidence is urgently needed. Therefore, we conducted a systematic review and meta-analysis to aggregate and critically evaluate data from randomized controlled trials and observational studies. The primary objective was to determine the overall effect of online-based antenatal education, compared to standard care, on key indicators of maternal mental health—specifically depression, anxiety, fear, and self-efficacy—and on the neonatal outcome of SGA incidence. While maternal preparedness is a multifaceted concept encompassing knowledge, skills, and psychological readiness, this review focuses on measurable psychological outcomes (depression, anxiety, fear, self-efficacy) and the incidence of small‑for‑gestational‑age (SGA) infants as the sole neonatal outcome.

Methods

Study design and reporting framework

This study was conducted as a systematic review and meta-analysis in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) guidelines. All stages of the review process, including study identification, screening, eligibility assessment, data extraction, and synthesis, were predefined to ensure methodological transparency and reproducibility. This systematic review and meta-analysis was registered in the International Prospective Register of Systematic Reviews (PROSPERO) (registration ID: CRD420261365771; available at: https://www.crd.york.ac.uk/PROSPERO/display_record.php?RecordID=1365771). The protocol was not registered prior to the start of the review because the work was initiated as an educational thesis at an institution that did not mandate prospective registration; however, we have since registered the completed review to ensure transparency. The review followed a structured and a priori defined methodological framework consistent with established standards in evidence synthesis [17].

Eligibility criteria

Studies were considered eligible if they evaluated the effects of online-based antenatal interventions among pregnant women. The target population comprised women in the antenatal period, irrespective of gestational age, parity, or clinical risk status, provided that participants were aged 18 years or older. Interventions were required to be delivered through digital or online platforms, including web-based programs, mobile health applications, telehealth systems, or other internet-enabled modalities. These interventions encompassed a range of approaches, including structured antenatal education programs, digitally delivered psychological interventions such as cognitive behavioral therapy or mindfulness-based programs, as well as hybrid models combining educational and therapeutic components.

Eligible comparator groups included standard antenatal care, waitlist controls, or alternative non-digital interventions. Studies were required to report at least one of the predefined outcomes, namely maternal mental health indicators such as depression, anxiety, or fear, measures of maternal self-efficacy, or neonatal outcomes, specifically small for gestational age (SGA). Both randomized controlled trials and non-randomized studies with a comparator group, including quasi-experimental and observational designs, were eligible for inclusion. Cross-sectional studies, qualitative studies, case reports, case series, and reviews were excluded. There were no restrictions on language or publication date. Studies published as conference abstracts or dissertations were eligible if sufficient data could be obtained from the authors or the available report. Studies lacking sufficient outcome data or employing non-comparable interventions were excluded.

Studies were excluded if they focused exclusively on postpartum populations, lacked an appropriate comparator group, or did not provide sufficient quantitative data to enable effect size estimation. Additional exclusions applied to conference abstracts, editorials, narrative reviews, and study protocols, as well as studies evaluating exclusively face-to-face interventions without a digital component.

Information sources and search strategy

A comprehensive and systematic search of the literature was conducted across multiple electronic databases, including PubMed/MEDLINE, Embase, Web of Science, the Cochrane Library, and CINAHL. The search strategy encompassed studies published from database inception to March 2026 to ensure inclusion of the most recent evidence in this rapidly evolving field. The search combined controlled vocabulary terms, such as Medical Subject Headings (MeSH), with free-text keywords related to antenatal care, digital health interventions, and maternal mental health outcomes. Key search terms included combinations of “antenatal education,” “pregnancy,” “online intervention,” “digital health,” “maternal mental health,” “self-efficacy,” and “neonatal outcomes.” The full search strategy for at least one database has been provided in the supplementary materials to facilitate reproducibility. The complete search strategy for each database is detailed in Supplementary File 1.

Study selection process

All identified records were imported into reference management software, and duplicate entries were removed prior to screening. Titles and abstracts were independently screened by two reviewers to identify potentially relevant studies. Full-text articles of selected records were subsequently assessed against the predefined eligibility criteria.

Any discrepancies between reviewers were resolved through discussion and consensus, with consultation from a third reviewer where necessary. The study selection process was documented using a PRISMA flow diagram, detailing the number of studies identified, screened, excluded, and included at each stage.

Data collection process and data items

Data extraction was conducted independently by two reviewers using a pre-piloted, standardized data extraction form developed in Microsoft Excel. Extracted data included study characteristics (author, publication year, country, study design, and sample size), participant characteristics (age, parity, and gestational age at enrollment, and socioeconomic status when available), details of the online antenatal education intervention (platform used, duration, frequency, content, and whether the intervention was guided or self-guided), comparator details, and outcome measures. Specific outcomes extracted were measures of maternal anxiety, depression, and fear (as assessed by validated scales such as the Hospital Anxiety and Depression Scale, Edinburgh Postnatal Depression Scale, or other equivalent instruments), self-efficacy scores (including specific scales used), and the incidence of SGA babies (defined according to standard birth weight percentiles, preferably below the 10th percentile). Self-efficacy was assessed using validated scales, including the Childbirth Self Efficacy Inventory (CBSEI) and general self-efficacy scales. Where different scales were used, scores were harmonized by converting to standardized mean differences, assuming all scales measured the same latent construct. For studies reporting multiple time points, data from the assessment closest to delivery or the primary endpoint as defined by the study authors were extracted. Discrepancies were resolved through consensus, and when necessary, study authors were contacted for clarification.

Classification of interventions

Given the conceptual heterogeneity inherent in digital antenatal interventions, included studies were categorized according to the primary nature of the intervention. Interventions were classified as educational, psychological, or hybrid. Educational interventions primarily focused on providing information and skills related to pregnancy, childbirth, and parenting. Psychological interventions included structured therapeutic approaches, such as internet-based cognitive behavioral therapy or mindfulness-based interventions, aimed at improving mental health outcomes. Hybrid interventions incorporated elements of both education and psychological support. This classification was used to inform the interpretation of findings and, where feasible, subgroup analyses. Studies comparing two different online interventions without a control group were excluded. Given the inherent heterogeneity of digital antenatal interventions, we elected to pool all eligible studies under the broad category of ‘online antenatal education’ to estimate the average effect of this class of interventions. This approach is intentionally conservative and exploratory; we pre specified subgroup analyses (by intervention type where feasible) and meta regression to investigate sources of heterogeneity. Readers are cautioned that pooled estimates reflect a heterogeneous evidence base and should not be interpreted as applying uniformly to all intervention types.

Risk of bias assessment

Risk of bias in the included RCTs was assessed using the Cochrane Risk of Bias 2 (RoB 2) tool, with overall judgments classified as low risk, some concerns, or high risk. The quasi-experimental study [18] was assessed using ROBINS-I (Risk Of Bias In Non-randomized Studies of Interventions). Observational studies were assessed using the Newcastle-Ottawa Scale (NOS), with scores 0–3 = high risk, 4–6 = moderate risk, 7–9 = low risk. NOS thresholds were based on established conventions [19] and previous meta-analyses in public health. Two reviewers independently assessed risk of bias, and domain-level judgments. Any disagreements were resolved through discussion with a third reviewer. Domain-level assessment indicated that the most common sources of bias included lack of blinding (performance bias), incomplete outcome data due to attrition (attrition bias), and selective reporting. These factors were considered in sensitivity analyses.

Outcome measures

The primary outcomes of interest were maternal mental health indicators, including depression, anxiety, and fear, as well as maternal self-efficacy. These outcomes were assessed using validated instruments, such as the Edinburgh Postnatal Depression Scale (EPDS), the Patient Health Questionnaire (PHQ-9), the State-Trait Anxiety Inventory (STAI), and the Childbirth Self-Efficacy Inventory. Given the variability in measurement tools across studies, outcomes were standardized for meta-analysis. The secondary outcome of interest was neonatal status, operationalized as small for gestational age (SGA), as reported in the included studies. For SGA, inclusion required a clear definition based on birth weight below the 10th percentile for gestational age.

Data synthesis and statistical analysis

Quantitative synthesis was conducted using a random-effects model to account for anticipated clinical and methodological heterogeneity among studies. Such heterogeneity was expected due to differences in intervention type, delivery modality, participant characteristics, and outcome measurement instruments. The use of a random-effects model therefore allowed for variation in true effect sizes across studies. Continuous outcomes (maternal anxiety, depression, fear, and self-efficacy) were synthesized using standardized mean differences (SMD) with 95% confidence intervals (CI), while dichotomous outcomes (incidence of SGA babies) were pooled as odds ratios (OR) with 95% CIs. The use of SMDs was necessary due to the use of different scales to measure similar constructs across studies. For studies reporting medians and interquartile ranges, means and standard deviations were estimated using established formulas. “Heterogeneity was evaluated using the I² statistic and Cochran’s Q test; I² values of 25%, 50%, and 75% were interpreted as low, moderate, and high heterogeneity, respectively, in accordance with conventional thresholds. Where sufficient studies (≥ 10) were available, pre specified subgroup analyses were planned based on intervention type (guided vs. self-guided), risk of bias (low vs. high), and study design (RCT vs. observational). Consistent with established methodological frameworks for digital health meta analyses [20], heterogeneity was explored through pre specified subgroup analyses where feasible. Sensitivity analyses were conducted by omitting one study at a time to assess the influence of individual studies on the pooled effect size. Publication bias was assessed using the Doi plot and the Luis Furuya Kanamori (LFK) index for outcomes with at least four studies. The Doi plot visually displays the distribution of effect sizes against a measure of precision, while the LFK index quantitatively evaluates asymmetry, with values within ± 1 indicating no asymmetry, between ± 1 and ± 2 suggesting minor asymmetry, and values exceeding ± 2 indicating major asymmetry [21]. A random-effects model (DerSimonian and Laird method) was selected a priori, as clinical and methodological heterogeneity was expected across studies due to variability in intervention content, duration, participant populations, and outcome measurement tools.

Assessment of heterogeneity

Statistical heterogeneity was evaluated using the I² statistic and Cochran’s Q test. The I² statistic was interpreted in accordance with established thresholds, with values exceeding 75% considered indicative of considerable heterogeneity. In cases where high heterogeneity was observed, the pooled estimates were interpreted with caution, and potential sources of variability were explored.

Subgroup and sensitivity analyses

Where sufficient data were available, subgroup analyses were conducted to explore potential sources of heterogeneity. These analyses were planned based on intervention type (educational, psychological, and hybrid) and study design (randomized vs. non-randomized). However, due to insufficient numbers of studies within each intervention category for most outcomes (e.g., for depression: educational n = 3, psychological n = 3, hybrid n = 2; for fear and self-efficacy: insufficient across all categories), formal subgroup analyses by intervention type could not be performed for several outcomes. Where feasible, descriptive comparisons are reported in Supplementary Table S2. The limited analytical use of this classification is acknowledged as a limitation in the Discussion.

Sensitivity analyses were performed to assess the robustness of the findings. These included analyses excluding studies at high risk of bias and analyses restricted to randomized controlled trials. The impact of individual studies on pooled estimates was also examined where appropriate.

Of the 12 included studies, 8 reported maternal depression outcomes and were eligible for subgroup stratification by intervention type. Studies were classified as educational (n = 2), psychological (n = 3), or hybrid (n = 3) based on their primary intervention content. Formal subgroup meta-analysis was limited by the small number of studies per category; the table below provides descriptive pooled estimates for illustrative purposes only. For maternal anxiety (n = 4 studies), fear of childbirth (n = 3), self-efficacy (n = 3), and SGA (n = 2), the number of studies was insufficient to conduct meaningful subgroup analyses by intervention type, as the resulting subgroups would contain fewer than two studies each, rendering statistical comparison invalid. Only 2 educational studies [18, 22] contributed to the depression subgroup (Supplementary Table S2).

Assessment of publication bias

Publication bias was assessed through visual inspection of funnel plots when a sufficient number of studies were available. In addition, Egger’s regression test was performed to statistically evaluate asymmetry in the distribution of effect sizes.

Handling of missing data

In instances of missing or incomplete data, attempts were made to obtain additional information by contacting study authors. Where data could not be obtained, appropriate statistical methods were applied where feasible, and the potential impact of missing data on the findings was considered in the interpretation of results.

Results

Search results

A total of 3,234 records were identified from databases, with 1,133 duplicates removed, leaving 2,101 records for screening. After excluding 2,015 abstracts, 86 full-text articles were assessed, and 74 were excluded for the following reasons: 58 had different interventions or comparators that did not meet eligibility criteria; 14 did not report relevant outcomes or sufficient quantitative data for pooling. Consequently, 12 studies met the eligibility criteria and were included in the meta-analysis [18, 22–32] (Fig. 1). No additional studies were identified through grey literature searches or hand-searching of reference lists.

Fig. 1.

PRISMA 2020 flow diagram of study selection process. Flowchart illustrating the identification, screening, eligibility, and inclusion of studies. A total of 3,234 records were identified from databases, with 1,133 duplicates removed. After title and abstract screening of 2,101 records, 86 full-text articles were assessed for eligibility, resulting in 12 studies included in the meta-analysis

Characteristics of the included studies

A total of 12 studies were identified, encompassing diverse countries such as the United States, Norway, China, Turkey, Sweden, Australia, and Singapore, among others [18, 22–32]. Most of the studies were RCTs (n = 10), with two being observational cohort studies [18, 22]. Sample sizes ranged from 14 to 1,941 participants, with a total of 4,982 pregnant women included across all studies. Interventions primarily consisted of internet- or mobile-based programs aimed at preventing or reducing perinatal depression, anxiety, or distress, often using structured modules or guided sessions. The duration of interventions ranged from a single session to 12 weeks, and delivery modes included self-guided online modules, guided web-based programs with therapist support, and mobile applications with push notifications. Mean ages of participants across studies varied from about 25.7 to 33.1 years. Parity was reported in most studies, with the majority including both primiparous and multiparous women, though three studies focused exclusively on primigravida women [28, 31, 32].

The pooled analyses demonstrated considerable heterogeneity (I² >85% across several outcomes), indicating substantial variability in intervention characteristics, participant populations, and outcome measurement tools. Consequently, pooled estimates should be interpreted with caution, and emphasis is placed on the direction and consistency of effects rather than statistical significance alone.

Risk of bias assessments showed that 5 studies had low risk of bias, 3 studies had some concerns, and 4 studies had high risk of bias (Table 1). Among the observational studies, both scored moderately on the Newcastle-Ottawa Scale, indicating moderate risk of bias [18, 22]. The primary reasons for high risk of bias in RCTs were lack of blinding of participants and personnel, and incomplete outcome data due to high attrition rates. Among RCTs, the most common sources of bias were deviations from intended interventions (due to lack of blinding of participants) and missing outcome data (attrition rates ranging from 15 30%). For the quasi experimental study, confounding and selection bias were the main concerns.

Table 1.

Characteristics of the included studies (N = 12)

Author and Year	Country	Study design	Participant	Sample size (I)	Sample size (C)	Intervention	Mean age (I)	Mean age (C)	Risk of bias
Barrera 2015 [15]	USA	RCT	Pregnant women 18 years and above	57	54	The Mothers and Babies Inter net Course- includes text, short audio, videoclips, images of infants and pregnant women, and worksheets	29.8 ± 6.1	30.6 ± 4.9	Low
Duffecy 2019 [17]	USA	RCT	Women who were 20 to 28 weeks pregnant with symptoms of depression	7	7	Sunnyside website - an 8-week online prevention intervention developed by research partners at the Northwestern University	NR	NR	Low
Haga 2019 [16]	Norway	RCT	Pregnant women across Norway attending well-baby clinics	678	664	Mamma Mia; a universal internet based preventive intervention for perinatal depressive symptoms.	31.0 ± 4.6	31.1 ± 4.5	Some concerns
Hao 2023 [21]	China	Retrospective study	Singleton pregnant women who had registered for the PUMCH mobile prenatal education curriculum and subsequently delivered at the PUMCH in Beijing	1521	420	Mobile-based prenatal education curriculum in collaboration with a multidisciplinary maternal care team	32.01 ± 3.8	32.1 ± 3.7	High
Heller 2020 [23]	Netherlands	RCT	Pregnant women (< 30 weeks) with depressive symptoms	79	80	MamaKits online - a 5-week guided internet intervention based on problem solving treatment.	32.08 ± 4.6	31.9 ± 4.8	High
Kuiper 2024 [24]	United Kingdom	Case-control study	Pregnant women aged 18 or older and postpartum women up to one year after the birth of a healthy child	94	790	WazzUp Mama - a remotely delivered web-based intervention to prevent and reduce perinatal emotional distress.	30.3 ± 3.1	30.3 ± 3.7	High
Loughnan 2019 [25]	Australia	RCT	Participants aged over 18 years had computer and internet access, diagnosis of GAD and/or MDD	36	41	MUMentum Pregnancy - a brief unguided iCBT intervention tailored specifically to women experiencing generalised anxiety and depressive symptoms	31.7 ± 4.4	31.4 ± 3.6	Low
Shorey 2019 [22]	Singapore	RCT	Participants aged 21 years and older were proficient in spoken and written English, owned a mobile phone with internet access,	118	118	The SEPP intervention include (1) a 30-min telephone-based antenatal education, (2) a 60-min telephone-based immediate postnatal education, and (3) a mobile health app follow-up education	30.4 ± 4.4	31.4 ± 4.6	Some concerns
Sun 2021 [26]	China	RCT	Pregnant adult women who were potentially at risk of perinatal depression	84	84	Self-guided 8-week smartphone-based mindfulness training during pregnancy group	30.3 ± 3.8	29.6 ± 4.2	Low
Tsai 2018 [18]	Taiwan	Quasi-experimental	Women at 16 to 24 weeks’ gestation with a low-risk pregnancy	68	67	A web-based antenatal care and education program in the second trimester	33.1 ± 5.0	32.5 ± 4.1	Low
Uludag 2022 [27]	Turkey	RCT	Pregnant women at age of 18 years or more, gestation of 24–34 weeks, graduation at least from primary school, nulliparity, not being at high risk in pregnancy, ability to use the application of Microsoft Teams, not having a psychiatric disease	23	21	Online antenatal childbirth preparation education through Microsoft Teams	26.7 ± 4.9	25.7 ± 4.6	High
Yesildag 2022 [28]	Turkey	RCT	Women who were aged 18–35 years, graduated at least from primary school, were primigravida at 28–30 weeks of gestation, had a single pregnancy, did not have a high-risk pregnancy, 3of12 had no contraindications to vaginal delivery	37	36	The web-based birth preparation programme based on the HBMSS	NR	NR	Some concerns

I Intervention, C Control, RCT Randomized controlled trial, USA United States of America

Maternal depression

Eight studies (n = 2493) were pooled under a random-effects model for assessing the impact of online antenatal education on maternal depression levels [23–29, 32]. The analysis yielded an overall SMD of -0.182 (95% CI: -0.453 to 0.089; p = 0.188) (Fig. 2). Due to substantial heterogeneity (I²=86.5%; p < 0.001), this pooled estimate is exploratory and should not be interpreted as confirmatory. The finding suggests no statistically significant effect, but the high heterogeneity limits confidence in this estimate. Sensitivity analysis omitting one study at a time did not substantially alter the pooled effect size, suggesting that no single study disproportionately influenced the result. Doi plot (Supplementary Fig. 1) showed major asymmetry which was further supported by an LFK index of -1.62, indicating the presence of publication bias. Given the high heterogeneity and evidence of publication bias, these findings should be interpreted with caution.

Fig. 2.

Forest plot of the effect of online antenatal education on maternal depression. Pooled standardized mean differences (SMD) with 95% confidence intervals (CI) from eight studies (n = 2,493) using a random-effects model. The overall SMD was − 0.182 (95% CI: -0.453 to 0.089; p = 0.188), indicating no significant effect. Heterogeneity was high (I² = 86.5%; p < 0.001). Squares represent individual study effect sizes, with size proportional to study weight. Horizontal lines represent 95% CIs. The diamond represents the pooled effect estimate

Maternal fear

Three studies (n = 255) were pooled using a random-effects model for assessing the impact of online antenatal education on maternal fear levels [30–32]. The analysis yielded an overall SMD of -0.527 (95% CI: -1.056 to 0.003; p = 0.051) (Fig. 3), indicating a reduction in fear of childbirth that did not reach statistical significance. Heterogeneity was moderate to high (I²=74.2%; p = 0.021). Given the modest number of studies and the heterogeneity, this finding should be considered exploratory. All three studies were RCTs with low to moderate risk of bias. Publication bias assessment was not possible as there were fewer than four studies (the minimum number required to perform Doi plot and LFK index).

Fig. 3.

Forest plot of the effect of online antenatal education on maternal fear. Pooled standardized mean differences (SMD) with 95% confidence intervals (CI) from three studies (n = 255) using a random-effects model. The overall SMD was − 0.527 (95% CI: -1.056 to 0.003; p = 0.051), indicating a borderline non-significant reduction. Heterogeneity was moderate to high (I² = 74.2%; p = 0.021). Squares represent individual study effect sizes, with size proportional to study weight. Horizontal lines represent 95% CIs. The diamond represents the pooled effect estimate

Maternal anxiety

Four studies (n = 845) were pooled using a random-effects model for assessing the impact of online antenatal education on maternal anxiety levels [25–27, 30]. The analysis yielded a pooled SMD of -0.101 (95% CI: -0.658 to 0.457; p = 0.723) for anxiety (Fig. 4), indicating no significant effect. Heterogeneity was high (I²=92.8%; p < 0.001); therefore, this pooled estimate is exploratory and should be interpreted with extreme caution. Sensitivity analysis revealed that exclusion of the study by Heller et al. [26] reduced heterogeneity to 68%, though the pooled effect remained non-significant, suggesting this study was a contributor to the high heterogeneity. Doi plot (Supplementary Fig. 2) showed no asymmetry which was further supported by an LFK index of -0.07, indicating the absence of publication bias.

Fig. 4.

Forest plot of the effect of online antenatal education on maternal anxiety. Pooled standardized mean differences (SMD) with 95% confidence intervals (CI) from four studies (n = 845) using a random-effects model. The overall SMD was − 0.101 (95% CI: -0.658 to 0.457; p = 0.723), indicating no significant effect. Heterogeneity was high (I² = 92.8%; p < 0.001). Squares represent individual study effect sizes, with size proportional to study weight. Horizontal lines represent 95% CIs. The diamond represents the pooled effect estimate

Self-efficacy

Three studies (n = 444) were pooled using a random-effects model, yielding an overall SMD of 0.558 (95% CI: -0.110 to 1.225; p = 0.102), indicating a non-significant trend towards improvement in self-efficacy (Fig. 5). Heterogeneity was very high (I²=90.6%; p < 0.001); accordingly, this pooled estimate is exploratory and no confirmatory conclusion can be drawn. The studies used different self-efficacy scales, including the Childbirth Self-Efficacy Inventory and general self-efficacy scales, which may have contributed to the high heterogeneity. Publication bias assessment was not possible as there were fewer than four studies.

Fig. 5.

Forest plot of the effect of online antenatal education on maternal self-efficacy. Pooled standardized mean differences (SMD) with 95% confidence intervals (CI) from three studies (n = 444) using a random-effects model. The overall SMD was 0.558 (95% CI: -0.110 to 1.225; p = 0.102), indicating a non-significant trend toward improvement. Heterogeneity was high (I² = 90.6%; p < 0.001). Squares represent individual study effect sizes, with size proportional to study weight. Horizontal lines represent 95% CIs. The diamond represents the pooled effect estimate

Small for gestational age (SGA) infants

Two studies (n = 2100) were pooled, yielding an overall OR of 0.728 (95% CI: 0.169–3.137; p = 0.670) for the effect on SGA, indicating no significant association (Fig. 6) [18, 22]. Heterogeneity was moderate with I² = 37.2% (Q = 1.59, df = 1, p = 0.207) and tau² = 0.477. Both studies were observational in design [18, 22], which limits the strength of causal inferences. These results suggest that online-based interventions do not significantly influence the incidence of SGA infants. Publication bias assessment was not possible as there were fewer than four studies.

Fig. 6.

Forest plot of the effect of online antenatal education on incidence of small for gestational age (SGA) infants. Pooled odds ratios (OR) with 95% confidence intervals (CI) from two studies (n = 2,100) using a random-effects model. The overall OR was 0.728 (95% CI: 0.169–3.137; p = 0.670), indicating no significant association. Heterogeneity was moderate (I² = 37.2%; p = 0.207). Squares represent individual study effect sizes, with size proportional to study weight. Horizontal lines represent 95% CIs. The diamond represents the pooled effect estimate

Subgroup and sensitivity analyses

Subgroup analysis for depression comparing guided (therapist supported, n = 3) vs. self-guided (n = 3) interventions showed no significant difference between subgroups (p = 0.34). Meta regression exploring the effect of intervention duration (in weeks) and sample size on depression outcomes showed no significant association (p > 0.05). Subgroup analyses by intervention type (educational, psychological, hybrid) were planned but could not be performed for most outcomes due to insufficient numbers of studies per category. Descriptive comparisons are provided in Supplementary Table S2; no clear pattern of effect modification was observed, but the analysis is underpowered.

Sensitivity analyses omitting studies with high risk of bias [23, 27] for depression and anxiety outcomes did not materially change the direction or significance of the pooled estimates, though heterogeneity remained high (I²=78 88%). For SGA, sensitivity analyses excluding observational studies could not be performed due to insufficient studies; for depression and anxiety, no observational studies were included. Meta regression exploring the effect of intervention duration (in weeks) and sample size on depression outcomes showed no significant association (p > 0.05).

Discussion

Our meta-analysis synthesized evidence from 12 studies encompassing nearly 5,000 pregnant women across diverse geographic regions [18, 22–32]. The findings consistently demonstrate that online-based antenatal education does not produce statistically significant improvements in maternal depression (SMD = -0.18; p = 0.188), anxiety (SMD = -0.10; p = 0.723), or fear (SMD = -0.53; p = 0.051). Self-efficacy showed a non-significant trend toward improvement (SMD = 0.56; p = 0.102), while no effect was observed on small for gestational age (SGA) infants (OR = 0.73; p = 0.670). These results indicate that despite widespread implementation; current online antenatal education programs are not effective in improving key maternal psychological parameters or neonatal outcomes.

The high degree of heterogeneity observed across studies likely reflects fundamental differences in intervention design, including variation in theoretical frameworks (educational vs. psychological), delivery intensity, and participant characteristics. This variability limits the interpretability of pooled estimates and suggests that online antenatal interventions should not be considered a uniform intervention category.

Comparison with existing literature

Our findings contrast with some individual studies reporting modest benefits from digital antenatal interventions [33, 34]. This discrepancy likely reflects several factors. First, positive findings in individual studies often derive from small, efficacy trials that may not generalize to real-world settings [19, 20]. Second, publication bias—evident for depression outcomes in our analysis (LFK index = -1.62)—suggests null findings are underrepresented in the literature. When heterogeneous studies are aggregated, optimistic effects are attenuated, revealing negligible average effects [35, 36].

The non-significant trend for self-efficacy aligns partially with previous research [37, 38], though high heterogeneity (I²=90.6%) and variable measurement tools complicate interpretation. Our null finding for SGA is consistent with evidence that antenatal education rarely influences clinical neonatal parameters, which are determined by complex biological and socioeconomic factors beyond the scope of educational interventions [39].

A critical caveat to all findings in this meta-analysis is that the substantial statistical heterogeneity observed across most outcomes (I² ranging from 74% to 93%) indicates that the pooled estimates are exploratory rather than confirmatory. The diversity of intervention content, theoretical frameworks, delivery intensities, and outcome measurement tools suggests that ‘online antenatal education’ is not a homogeneous entity. Therefore, our findings should be interpreted as an average effect of a heterogeneous class of interventions; clinically meaningful effects may exist for specific, well-designed programs that remain undetected in this aggregated analysis. No confirmatory conclusions regarding the presence or absence of an effect can be drawn from such heterogeneous bodies of evidence.

Explanations for null findings

Several interrelated factors explain the absence of significant effects. First, intervention heterogeneity was substantial. Programs ranged from self-guided static websites to therapist-assisted cognitive-behavioral interventions, with durations from single sessions to 12 weeks [18, 22–32]. Pooling such diverse interventions under “online antenatal education” likely obscures effects that may exist only for specific, well-designed programs. This variability manifested statistically as high heterogeneity across most outcomes (I² = 74–93%).

Second, population characteristics limited detectable effects. Most participants were low-risk women already receiving standard care, creating a “floor effect” where additional improvements are difficult to demonstrate. Studies rarely targeted women with elevated symptoms, who would have greater capacity for improvement.

Third, measurement variability across studies—different validated scales for the same constructs—introduced noise that reduced statistical power to detect effects. For self-efficacy specifically, the use of both childbirth-specific and general scales likely contributed to heterogeneity.

Fourth, cultural factors may influence reporting of psychological distress and engagement with digital interventions. The geographic diversity of included studies (Western and Asian populations) introduced variability that could not be controlled in analyses.

Regarding SGA, the lack of effect is unsurprising given that fetal growth is determined by nutritional, genetic, and placental factors unlikely to be modified by brief educational interventions. Moreover, the two contributing studies were observational [18, 22], limiting causal inference.

Strengths and limitations

Strengths include adherence to PRISMA guidelines, comprehensive database searches without language restrictions, inclusion of grey literature, rigorous risk of bias assessment, and use of random-effects models with publication bias evaluation via Doi plots and LFK index. Limitations must be acknowledged. High heterogeneity across outcomes precludes definitive conclusions and suggests pooled estimates should be interpreted cautiously. The small number of studies for fear (n = 3), self-efficacy (n = 3), and SGA (n = 2) prevented subgroup analyses and publication bias assessment, limiting inference depth for these outcomes. Inclusion of observational studies for SGA introduces potential confounding. Reliance on self-reported outcomes may introduce response bias. A critical limitation of this meta-analysis is the conceptual breadth of ‘online antenatal education.’ Included interventions ranged from general educational websites to structured cognitive behavioral therapy programs and mindfulness-based interventions. These interventions differ fundamentally in theoretical underpinnings, intensity, and intended mechanisms of action. Pooling such diverse interventions likely obscures effects that may exist only for specific, well-designed programs. This variability manifested statistically as high heterogeneity across most outcomes (I² = 74–93%). Moreover, although we classified interventions as educational, psychological, or hybrid, the small number of studies within each category (particularly for fear, self-efficacy, and SGA outcomes) precluded meaningful subgroup analyses. Consequently, we could not determine whether specific intervention types are more effective than others, and this remains an important gap for future research.

Clinical implications

Clinicians and policymakers should not assume that current online antenatal education programs improve mental health outcomes. These platforms should complement, not replace, standard care and targeted mental health interventions. For women with elevated symptoms, referral to evidence-based psychological therapies remains essential.

However, online education offers value for information provision, particularly for women facing barriers to in-person classes [6, 7]. The non-significant trend toward improved self-efficacy suggests some women may derive confidence benefits. Healthcare systems should curate high-quality, evidence-based online content rather than relying on unregulated digital resources.

Future research directions

Future research must address identified limitations. First, standardized intervention protocols based on established behavior change frameworks are needed. Second, large multicenter RCTs should stratify by baseline risk to identify subgroups that benefit. Third, dismantling trials should identify active ingredients (e.g., professional support, interactivity) through systematic component variation. Fourth, qualitative research should explore user experiences and barriers to engagement, particularly regarding digital literacy and cultural acceptability. Fifth, long-term outcomes beyond the immediate postpartum period should be evaluated. Sixth, economic analyses should assess cost-effectiveness. Finally, research must prioritize underserved and high-risk populations—including low-resource settings, adolescents, and women with limited literacy—who may benefit most from accessible digital interventions [7, 35].

Conclusion

This meta-analysis demonstrates that current online antenatal education, as evaluated in existing heterogeneous studies, does not show statistically significant improvements in maternal depression, anxiety, fear, self-efficacy, or SGA incidence. However, due to substantial heterogeneity (I² >85% for most outcomes), these pooled estimates are exploratory and should not be interpreted as definitive evidence of absence of effect. While these platforms enhance accessibility, the high heterogeneity and limited number of studies for several outcomes preclude definitive conclusions about effectiveness. Future research requires standardized, theoretically grounded interventions, rigorous trials focusing on at-risk populations, and prospective protocol registration to enhance transparency and reproducibility.

Authors’ contributions

- Xue Bai: Conceptualization, Methodology, Formal analysis, writing – Original Draft, Writing – Review & Editing - Juan Song: Methodology, Investigation, Data Curation, Writing – Review & Editing - XueMei Zhang: Supervision, Project Administration, Writing – Review & Editing. All authors reviewed and approved the final manuscript.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files. The datasets used for meta-analyses are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable. This study is a meta-analysis of published data and did not involve direct contact with human participants. All data were extracted from previously published studies that had obtained ethical approval and informed consent from participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.