Digital and repetitive transcranial magnetic stimulation interventions for perinatal depression and anxiety: a systematic review

Abstract

Background

Perinatal mood and anxiety disorders (PMADs) remain undertreated despite their prevalence and consequences. Digital psychological interventions and repetitive transcranial magnetic stimulation (rTMS) have been proposed as scalable or medication-sparing options. This systematic review evaluates the effectiveness of therapist-guided digital cognitive behavioral therapy (CBT) and rTMS for perinatal depression and anxiety, with separate synthesis of efficacy and safety evidence.

Methods

We conducted a PRISMA 2020-guided systematic review of six databases (PubMed/Medline, Scopus, Web of Science, PsycINFO, Cochrane Central, Embase) from inception to 01 October 2025. Randomized controlled trials (RCTs), non-randomized trials, pilot studies, and case series were included. Case reports (n < 5) were excluded from efficacy analyses but retained for safety data extraction. Risk of bias was assessed using Cochrane RoB 2 and ROBINS-I. Narrative synthesis was necessitated by significant clinical and methodological heterogeneity; effect sizes are contextualized by baseline severity, comparator, and study quality.

Results

Twenty-nine studies (23 RCTs, 6 non-randomized) were included. Therapist-guided digital CBT demonstrated moderate-to-large effects on depressive symptoms (between-group Hedges’ g 0.5–1.4) in most, but not all, trials; certainty of evidence was moderate. Self-guided digital CBT yielded smaller, inconsistent effects (g = 0.2–0.6; low certainty). Digital mindfulness interventions showed pronounced heterogeneity: five guided, culturally adapted Eastern trials reported positive effects, while six unguided Western trials found no benefit (very low certainty). For rTMS, one small sham-controlled RCT (n = 26) provided very-low-certainty efficacy evidence. Uncontrolled studies (≈ 230 pregnancies) reported large pre-post improvements but are critically confounded and cannot establish efficacy. Safety data, while reassuring in the short term, derive from limited cumulative exposure (≈ 230 pregnancies); absence of observed harm is not equivalent to evidence of safety (low certainty).

Conclusions

Therapist-guided digital CBT is a well-supported intervention and should be integrated into perinatal care. Self-guided digital CBT and mindfulness apps have limited or uncertain evidence. rTMS remains experimental; large, definitive RCTs are urgently needed. Safety evidence for rTMS is insufficient to draw firm conclusions.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-026-07948-8.

Introduction

Pregnancy and the postpartum period represent times of profound physical, emotional, and social transition [1]. While many individuals experience joy and anticipation, the perinatal period also confers heightened vulnerability to mental health difficulties, particularly depression and anxiety [2]. These conditions collectively described as perinatal mood and anxiety disorders (PMADs) are now understood to be among the most common complications of pregnancy and childbirth, affecting 10–20% of perinatal individuals globally, with higher rates reported in marginalized and underserved populations [3, 4].

The consequences of untreated PMADs extend beyond maternal suffering. Perinatal depression and anxiety are associated with increased risks of preterm birth, low birthweight, and dysregulated fetal stress physiology [5]. Postnatal, persistent maternal symptoms can impair early bonding and sensitive caregiving, which are foundational for infant socioemotional and cognitive development [6]. Longitudinal evidence demonstrates that children exposed to untreated maternal distress face elevated risks of behavioral difficulties, emotional dysregulation, and executive function challenges [7]. These intersecting consequences underscore that timely recognition and treatment of PMADs are essential components of comprehensive perinatal care.

Despite this clinical imperative, effective treatment remains out of reach for most affected individuals [8]. Fewer than half of those with perinatal depression or anxiety receive any treatment, and many who do experience delays, insufficient follow-up, or fragmented care [9, 10]. Barriers are multifaceted: concerns about medication exposure during pregnancy or breastfeeding, logistical challenges (cost, childcare, transportation, inflexible work), and a critical shortage of perinatal mental health specialists [11, 12]. Structural inequities including racism, socioeconomic disadvantage, and geographic isolation further widen disparities in screening and treatment access [13]. These persistent gaps make clear that traditional models of mental health care alone cannot meet the needs of all perinatal people.

Two therapeutic innovations have gained momentum as scalable or medication-sparing alternatives: digital mental health interventions and repetitive transcranial magnetic stimulation (rTMS) [14]. Digital psychological interventions such as internet-based cognitive behavioral therapy (iCBT), app-delivered mindfulness programs, and online peer support offer privacy, convenience, and the ability to deliver care directly into patients’ homes on their own schedules [15]. Multiple randomized trials and systematic reviews demonstrate that digital interventions can reduce depressive and anxiety symptoms in both pregnant and postpartum populations [16]. However, digital interventions vary substantially in quality, structure, guidance intensity, and theoretical orientation; engagement is often inconsistent, and the factors predicting adherence or sustained benefit remain incompletely understood [17].

For individuals with more severe or treatment-resistant symptoms or those who wish to avoid pharmacotherapy during pregnancy neuromodulation offers another promising pathway [18]. Repetitive transcranial magnetic stimulation (rTMS) is an FDA-cleared, noninvasive treatment for major depressive disorder that uses focused magnetic pulses to modulate neural circuits implicated in mood regulation [19]. rTMS involves no systemic medication exposure, making it a particularly compelling option for pregnant individuals concerned about pharmacologic risks [20]. A growing literature, including prospective studies and systematic reviews, suggests that rTMS is well tolerated during pregnancy and may effectively reduce depressive symptoms [21]. To date, no major safety signals for obstetric or neonatal outcomes have been identified, although research remains limited by small samples and methodological heterogeneity [20].

We selected to focus this review on digital CBT and rTMS for three specific reasons. First, among psychotherapeutic modalities, CBT possesses the strongest evidence base for perinatal depression and anxiety and has been successfully translated into scalable digital formats; by contrast, other digital psychotherapies (e.g., interpersonal therapy, psychodynamic therapy) remain in early stages of adaptation and lack comparable RCT evidence in perinatal populations. Second, among non-invasive neuromodulation techniques, rTMS is the only modality with regulatory approval (FDA, CE) for major depressive disorder and has accumulated a growing perinatal safety database; transcranial direct current stimulation (tDCS), while promising, has not been systematically evaluated in pregnant or postpartum cohorts. Third, and conceptually, these two modalities address complementary needs within a stepped-care framework. Digital CBT can serve as a low-intensity, scalable first-line intervention for mild-to-moderate symptoms, whereas rTMS offers a medication-sparing option for individuals with severe or treatment-resistant depression who may not respond to psychotherapy alone. Juxtaposing them within a single review allows us to map the current evidence base across the severity spectrum and to identify shared implementation challenges related to access, acceptability, and equity.

Although digital interventions and rTMS each offer distinct advantages, the literature evaluating them has developed in parallel rather than in conversation. Existing reviews typically examine one modality in isolation or focus on a narrow subset of outcomes. Few syntheses directly consider digital interventions and rTMS together, even though they address complementary needs within perinatal mental health care and raise related questions about feasibility, cost, scalability, and real-world implementation.

The primary aim of this systematic review is to critically synthesize the evidence for efficacy of digital CBT and rTMS in reducing symptoms of perinatal depression and anxiety, drawing principally on randomized controlled trials. Secondary aims are to (1) summarize safety and tolerability data for both modalities, drawing on all available study designs, and (2) describe feasibility and acceptability outcomes where reported. Importantly, this review is not designed to directly compare the efficacy of digital CBT versus rTMS, nor to establish a hierarchy of effectiveness; such a comparison would be inappropriate given fundamental differences in intervention type, target populations, and study designs. We also do not aim to produce formal clinical practice guidelines. Rather, we provide a comprehensive, methodologically rigorous overview of the current evidence base to inform researchers, clinicians, and policymakers, while clearly delineating the strength of evidence for efficacy versus safety.

Guided by PRISMA 2020 methodology, the purpose of this systematic review is therefore to critically synthesize the current evidence on digital mental health interventions and rTMS for perinatal depression and anxiety.

Methods

Study design

This systematic review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement [22].

Eligibility criteria

Eligibility criteria were established a priori using the PICO (Population, Intervention, Comparator, Outcomes) framework (Table S1).

Studies were eligible if they included (1) participants meeting full diagnostic criteria for major depressive disorder, generalized anxiety disorder, or related PMADs via structured clinical interview (e.g., SCID, MINI), OR (2) participants with subthreshold but elevated symptoms, defined as scores above established cutoffs on validated screening instruments (e.g., EPDS ≥ 10, PHQ-9 ≥ 10, GAD-7 ≥ 8). Studies enrolling mixed samples (comorbid depression and anxiety) were included, as this reflects real-world perinatal populations. We did not include studies focused exclusively on conditions outside our scope (e.g., PTSD, OCD) unless they reported disaggregated data for depression or anxiety outcomes.

We included randomized controlled trials (RCTs), non-randomized trials (open-label, quasi-experimental), pilot studies, and case series with n ≥ 5. Case reports and very small case series (n < 5) were excluded from all efficacy analyses; however, they were considered eligible for the extraction of safety and tolerability data only, given the limited cumulative exposure data for rTMS in pregnancy. These studies are clearly identified in Table 3 and are discussed solely in the context of safety.

Table 3.

(Revised structure – separate maternal and fetal/neonatal safety columns)

Author (Year)	Country	Design	Sample Size (N)	Intervention Details (Protocol)	Control Group	Primary Outcome(s)	Maternal Adverse Events	Fetal/Neonatal Outcomes	Key Efficacy Findings
Kim et al. (2019) [47]	USA	RCT	26 (Active:14, Sham:12)	Site: Right DLPFC; 1 Hz, 100% MT, 900 pulses/session, 20 sessions	Sham rTMS	Depression: HDRS-17, BDI, EPDS, CGI-S	Mild headache (n = 4); scalp discomfort (n = 2)	No congenital anomalies; no adverse neonatal outcomes reported	Significant time×group interaction (p = 0.003); response 81.8% vs. 45.5%
Hızlı Sayar et al. (2014) [48]	Turkey	Open-Label Trial	30	Site: Left DLPFC; 25 Hz, 100% MT, 1000 pulses/session, 18 sessions	N/A	Depression: HAMD-17	Mild headache (n = 6); neck pain (n = 2)	No fetal/neonatal data reported	Mean HAMD decreased 26.77 → 13.03 (p < 0.001)
Kim et al. (2011)  [49]	USA	Open-Label Pilot	10	Site: Right DLPFC; 1 Hz, 100% MT, 20 sessions	N/A	Depression: HDRS	Mild headache (40%); transient uterine contractions (n = 1; resolved)	No adverse pregnancy outcomes	70% response; 30% remission
Klirova et al. (2008)  [50]	Czech Republic	Case Series	2	Pt1: L-DLPFC, 20 Hz, 15 sessions; Pt2: R-DLPFC, 1 Hz, 15 sessions	N/A	Depression: BDI	None reported	One preterm birth (36 weeks); causality unclear	One full remission; one 59% BDI reduction
Zhang et al. (2010)  [51]	Not Reported	Case Report†	1	Sequential unilateral & bilateral; 1 Hz, 90% MT, multiple courses	N/A	Depression: HDRS	None reported	No adverse outcomes reported	HDRS 35 → 12
Nahas et al. (1999)  [52]	Not Specified	Case Report†	1	Site: Left DLPFC; 15 Hz, 110% MT, 25 sessions	N/A	Depression: Not specified	None reported	No fetal/neonatal data reported	No symptom improvement

† Footnote: Study excluded from efficacy synthesis (n < 5 or uncontrolled design); included for safety data extraction only in accordance with eligibility criteria (Sect.  2.2)

Information sources and search strategy

The systematic search strategy was designed and executed in collaboration with a university health sciences librarian. Six databases were searched from inception to 01 October 2025: PubMed/Medline, Scopus, Web of Science, PsycINFO, Cochrane Central, and Embase. Clinical trial registries (ClinicalTrials.gov, WHO ICTRP, EU-CTR) and grey literature sources (Google Scholar first 100 hits, conference abstracts) were also searched.

The strategy combined controlled vocabulary and free-text terms for three core concepts: (1) Perinatal Population, (2) Anxiety & Depression, and (3) Digital Interventions & rTMS. Full search strings for each database are provided in Table S2. Reference lists of all included studies and relevant prior reviews were manually screened to identify additional eligible publications.

Risk of bias assessment

Risk of bias was assessed independently by two reviewers (ZBK, MA) using the Cochrane Risk-of-Bias 2 (RoB 2) tool for randomized controlled trials [23] and the ROBINS-I tool for non-randomized studies [24]. Disagreements were resolved through discussion or adjudication by a third reviewer (AD). Overall judgements are presented in Supplementary Table S3.

Risk-of-bias ratings were not merely tabulated but were explicitly integrated into the narrative synthesis. In the Results section, effect sizes from studies with “high” or “critical” risk of bias are interpreted with explicit caution, and sensitivity of conclusions to methodological quality is addressed in the Discussion. This approach ensures that methodological limitations are consistently considered when weighing the strength of evidence.

Study selection process

All records identified through database searches were imported into Covidence systematic review software for deduplication and management. The study selection process was carried out independently by two researchers (R1 and R2) in two sequential phases. First, titles and abstracts were screened against the pre-defined eligibility criteria. Second, the full text of all records that passed the initial screen were retrieved and thoroughly evaluated for final inclusion.

At both stages, any disagreements were resolved through discussion or, if necessary, by adjudication from a third senior researcher (R3). The entire study selection process, including the number of records identified, included, and excluded at each stage with reasons for exclusion at full-text stage, is documented in the PRISMA 2020 flow diagram (Fig. 1).

Fig. 1.

PRISMA 2020 flow diagram of study selection

Data extraction and synthesis

Data from included studies were extracted independently by two reviewers into a customized, piloted data extraction form in Microsoft Excel. Discrepancies were resolved by consensus or third-party arbitration. The following data were systematically extracted:

• Study Characteristics: Author(s), publication year, country, funding sources, study design, total sample size, and group sizes.

• Participant Characteristics: Population (antenatal/postpartum), diagnostic status (clinical diagnosis vs. elevated symptoms), baseline symptom severity, and comorbidity.

• Intervention Details:

  • Digital Interventions: Therapeutic paradigm, delivery platform, duration, frequency, level of human guidance (therapist-guided, peer-guided, or fully self-guided).
  • rTMS Interventions: Stimulation target, frequency (Hz), intensity (% motor threshold), total pulses per session, number of sessions, and coil type.

• Comparator Details: Description of the control condition (waitlist, TAU, active control, sham rTMS).

• Outcome Data: Pre- and post-intervention means and standard deviations for continuous outcomes, response/remission rates, and all reported safety and acceptability data. Maternal and fetal/neonatal adverse events were extracted separately and are reported distinctly in Table 3 and the safety synthesis.

• Effect Sizes: Hedges’ g with 95% confidence intervals were calculated or extracted when reported.

Classification of digital interventions. Interventions were categorized according to their primary therapeutic orientation as declared by study authors and verified by examining intervention protocols or published treatment rationales. Studies were grouped into (1) digital cognitive-behavioral interventions (including CBT, behavioral activation, problem-solving therapy) and (2) digital mindfulness- and compassion-based interventions (including mindfulness-based stress reduction, mindfulness-based cognitive therapy, compassionate mind training). In cases where interventions incorporated hybrid or trans diagnostic components (e.g., a mindfulness program with psychoeducation resembling CBT, or a CBT program incorporating acceptance-based strategies), classification was determined by the modal therapeutic technique. Two reviewers independently categorized each intervention; disagreements were resolved by consensus. Sensitivity analyses excluding borderline cases were conducted and did not alter the pattern of findings. This classification approach is transparently reported to allow readers to judge potential misclassification risk.

Due to significant clinical and methodological heterogeneity among the included studies including diverse interventions (guided vs. unguided), comparators (waitlist vs. active control), outcome measures, populations (antenatal vs. postpartum, diagnosed vs. subthreshold), and study designs (RCTs vs. open-label studies) a quantitative meta-analysis was deemed inappropriate. Therefore, a structured narrative synthesis was performed, stratified by intervention type and study design.

In reporting effect size ranges, we explicitly contextualize estimates with respect to baseline severity, intervention duration, comparator type, and risk of bias. This approach ensures that readers are not misled by simple ranges that obscure important clinical and methodological differences.

Certainty of evidence

The certainty of evidence for each intervention category was assessed using the GRADE approach, considering risk of bias, inconsistency, indirectness, imprecision, and publication bias. Assessments were performed independently by two reviewers, with disagreements resolved by consensus. Summary findings and certainty ratings are presented in Table S4. Evidence for efficacy and evidence for safety are graded separately, reflecting the distinct types of studies and outcomes contributing to each.

Ethical considerations

As this review used data from previously published studies, institutional ethics approval was not required.

Results

Study selection and characteristics

The search yielded 29 eligible studies comprising more than 3,100 participants. Study designs included 23 randomized controlled trials (RCTs) and 6 non-randomized studies (open-label trials, case series). Digital mental health interventions accounted for 21 of the RCTs; rTMS accounted for one sham-controlled RCT. Non-randomized studies were exclusively rTMS protocols or open-label digital pilots.

Consistent with our eligibility criteria (Sect.  2.2), five very small case series and case reports (n < 5) were excluded from all efficacy analyses but were retained for safety data extraction. These studies are clearly identified in Table 3 and are discussed only in the context of safety.

Studies were conducted across North America, Europe, Australia, China, and the Middle East. Samples included both pregnant and postpartum participants; interventions targeted perinatal depression, anxiety, or combined symptomatology as assessed by validated measures. Detailed characteristics are presented in Tables 1, 2 and 3.

Table 1.

Characteristics of included digital cognitive-behavioral therapy (CBT) studies

Author (Year)	Country	Design	Sample Size (N)	Intervention Details	Control Group	Primary Outcome(s)	Key Findings
Forsell et al. (2017) [25]	Sweden	RCT	42 (IG:22, CG:20)	Therapist-guided iCBT: 10-week online platform, 7 core & 3 optional modules.	Waitlist	Depression: MADRS-S, EPDS	Significant reduction in depressive symptoms with one of the larger effect sizes.
Jannati et al. (2020) [26]	Iran	RCT	78 (IG:39, CG:39)	Psychologist-guided CBT: ’Happy Mom’ mobile app, 8-week, guided sessions.	Waitlist	Depression: EPDS	Very large, significant effect in reducing postpartum depression scores.
Pugh et al. (2016) [27]	Canada	RCT	50 (IG:25, CG:25)	Therapist-guided CBT: ’Maternal Depression Online’, 7–10-week online program.	Not Specified	Depression: EPDS	Significant improvement in postpartum depression.
O’Mahen et al. (2013) [28]	UK	RCT	910 (IG:462, CG:448)	Guided Behavioral Activation: 15-week intervention via a parenting website.	Not Specified	Depression: EPDS	Effective in reducing depressive symptoms despite high dropout rate.
Fonseca et al. (2020) [29]	Portugal	RCT	194 (IG:98, CG:96)	Self-guided CBT: ’Be a mom’, 8-week online platform, 5 modules.	Waitlist	Depression: PDPI-R, EPDS	Greater reduction in depressive symptoms vs. control, though effect was modest.
Loughnan, Butler, et al. (2019) [30]	Australia	RCT	87 (IG:43, CG:44)	Self-guided CBT: ’MUMentum Pregnancy’, 4-week online platform for antenatal anxiety/depression.	Treatment as Usual (TAU)	Depression: PHQ-9, EPDS Anxiety: DASS-42-Anxiety	Significant improvements in depressive and anxiety symptoms vs. TAU.
Danaher et al. (2013) [31]	USA & Australia	Open Trial	53	iCBT (Telephone-guided): ’MomMoodBooster’, 6-session program over 6–12 weeks with weekly phone support.	N/A	Depression: PHQ-9, HRSD	Large, significant improvements (pre-post). High adherence (87%) and satisfaction.
Milgrom et al. (2016) [32]	Australia	RCT	43 (IG:21, CG:22)	iCBT (Telephone-guided): ’MumMoodBooster’, 6-session program with phone support.	TAU	Depression: BDI-II Anxiety: DASS-21-A	Significantly greater improvement in depression vs. TAU (Cohen’s d = 0.83). No sig. difference for anxiety.
Heller et al. (2020) [24]	Netherlands	RCT	159 (IG:79, CG:80)	Guided Problem-Solving Therapy: ’MamaKits Online’, 5-week guided online therapy.	Waitlist	Depression: CES-D, EPDS Anxiety: HADS-A	Reduction in symptoms in both groups, but no significant between-group differences. Trial stopped early for futility.
Sawyer et al. (2019) [33]	Australia	RCT	146 (IG:72, CG:74)	Nurse-guided CBT: ’eMums Plus’, 4-month, nurse-guided online group-based app.	Standard Care	Depression: EPDS	No significant difference in EPDS scores between groups.
Boyd et al. (2018) [34]	USA	RCT	24 (83% African American)	Social Media Intervention: Parenting intervention via private, facilitated Facebook group.	In-person group	Depression: BDI-II	Significant reduction in depressive symptoms vs. in-person group (p < 0.01). High satisfaction and retention.
Hantsoo et al. (2017) [35]	USA	RCT	72 (64.3% African American)	Mobile App (Monitoring): App for mood tracking with alerts to healthcare providers.	Not Specified	Depression: PHQ-9 Anxiety: GAD-7	Significant improvement in depression (p = 0.001) and anxiety (p = 0.003). 41% received provider calls triggered by the app.

Table 2.

Characteristics of included digital mindfulness and compassion-based intervention studies

Author (Year)	Country	Design	Sample Size (N)	Intervention Details	Control Group	Primary Outcome(s)	Key Findings
Sun et al. (2021) [36]	China	RCT	168 (IG:84, CG:84)	Self-guided Mindfulness: ’Spirits Healing’, 8-week smartphone app for mindfulness training.	Waitlist	Depression: EPDS, PHQ-9 Anxiety: GAD-7	Effective for reducing perinatal depression and anxiety.
Yang et al. (2019) [37]	China	RCT	123 (IG:62, CG:61)	Guided Mindfulness: 8-week WeChat program (videos/text/audio) guided by a multidisciplinary team.	Waitlist	Depression: PHQ-9 Anxiety: SAS	Effective in reducing symptoms of antenatal depression and anxiety.
Smith et al. (2021) [38]	USA	RCT	101 (IG:50, CG:51)	Digital Mindfulness: ’Calm’ app for mindfulness meditation.	Waitlist	Depression: EPDS Anxiety: GAD-7	Significant reduction in depression (Hedges’ g = -0.44) and anxiety (Hedges’ g = -0.44).
Zhang et al. (2021) [39]	China	RCT	108 (IG:54, CG:54)	Digital Mindfulness: Mindfulness-Based Intervention (MBI) delivered via social media.	Waitlist	Depression: EPDS Anxiety: GAD-7	Significant reduction in depression (Hedges’ g = -0.61) and anxiety (Hedges’ g = -0.43).
Zhang et al. (2022) [40]	China	RCT	130 (IG:66, CG:64)	Guided Mindfulness: Guided self-help Mindfulness-Based Intervention (MBI).	Waitlist	Depression: EPDS Anxiety: GAD-7	Significant reduction in depression (Hedges’ g = -0.57) and anxiety (Hedges’ g = -0.89).
Güney et al. (2022) [41]	Turkey	RCT	84 (IG:42, CG:42)	Digital Mindfulness: Mindfulness-Based Stress Reduction (MBSR) program.	Waitlist	Anxiety: BAI, CAQ Depression: Not Assessed	Significant reduction in anxiety (Hedges’ g = -1.58).
Kelman et al. (2018) [42]	USA	RCT	40 (IG:22, CG:18)	Digital Mindfulness: Internet-based Compassionate Mind Training (CMT).	Waitlist	Depression: PHQ-2 Anxiety: GAD-2	Significant reduction in depression (Hedges’ g = -0.83). Non-significant reduction in anxiety (Hedges’ g = -0.54).
Carissoli et al. (2017) [43]	Italy	RCT	78 (IG:38, CG:40)	Digital Mindfulness: App (“Benevolent Mind”) with daily relaxation, guided imagery, mood journal.	Waitlist	Anxiety: W-DEQ	Non-significant reduction in anxiety (Hedges’ g = -0.05).
Matvienko-Sikar & Dockray (2017) [44]	Ireland	RCT	36 (IG:24, CG:12)	Digital Mindfulness: Gratitude diary and mindfulness listening.	Waitlist	Depression: EPDS	Non-significant reduction in depression (Hedges’ g = 0.63).
Krusche et al. (2018) [45]	UK	RCT	72 (IG:22, CG:50)	Digital Mindfulness: Website (“Be Mindful Online”) with guided meditation.	Waitlist	Depression: PHQ-2 Anxiety: GAD-2	Non-significant effect, trend favoring control for depression (g = 0.80) and anxiety (g = 0.46).
Doty et al. (2021) [46]	USA	RCT	41 (IG:20, CG:21)	Digital Mindfulness: Mindfulness meditation program.	Waitlist	Depression: PHQ-2 Anxiety: GAD-2	Non-significant reduction in depression (Hedges’ g = -0.58) and anxiety (Hedges’ g = -0.37).

Risk of bias

Fifteen of the 23 RCTs were rated as having “some concerns,” and four were rated as “high risk.” The most common methodological limitations were performance bias (lack of blinding of therapists or participants) and attrition bias (dropout > 20%). All non-randomized rTMS studies were judged at serious or critical risk of bias due to uncontrolled designs, confounding, and selective participant selection.

Risk-of-bias ratings are explicitly integrated into the narrative synthesis below; effect sizes from studies at high or critical risk of bias are interpreted with appropriate caution. Full domain-level judgements are provided in Supplementary Table S3.

Digital cognitive-behavioral interventions

Therapist-guided digital CBT

Eight RCTs evaluated therapist-guided or provider-supported digital CBT. Between-group effect sizes (Hedges’ g) ranged from 0.5 to 1.4. However, this range spans trials with substantially different populations, designs, and methodological quality:

• The upper bound (g = 1.4) derives from a single small trial [27] conducted in Iran with psychologist-guided app-based CBT. This study recruited clinically diagnosed postpartum women, used a waitlist control, and was rated as having “some concerns” due to lack of therapist blinding and sole reliance on self-report EPDS.

• The lower bound (g ≈ 0.5) reflects trials in subthreshold or mixed preventive samples [30, 31], with shorter intervention durations (4–8 weeks) and comparator conditions ranging from waitlist to treatment as usual (TAU).

• Two large, adequately powered trials [33, 47] reported no significant between-group differences on the EPDS. Both used active or standard-care comparators and blinded outcome assessment, and were rated as having “some concerns” primarily due to guidance-related performance bias. These null findings underscore that guidance alone does not guarantee effectiveness.

Certainty of evidence was rated moderate (downgraded for inconsistency). Across positive trials, therapist-guided digital CBT consistently reduced depressive symptoms, but effect magnitude varies meaningfully by population severity and comparator type.

Self-guided digital CBT

Five RCTs evaluated fully self-guided digital CBT. Effect sizes were smaller and more variable (g = 0.2–0.6). Two trials [30, 31] reported modest but statistically significant improvements relative to waitlist; three others found non-significant effects. All self-guided trials were rated as having “some concerns” or “high risk”, predominantly due to attrition (dropout 25–40%) and lack of blinding.

Certainty of evidence was rated low (downgraded for imprecision and inconsistency). Self-guided digital CBT may reduce symptoms, but effects are modest and not robust across all populations.

Digital mindfulness- and compassion-based interventions

Eleven RCTs assessed digital mindfulness or compassion-based programs. Findings were highly heterogeneous:

• Five trials all conducted in China and employing guided or partially guided formats reported moderate to large effects on depression or anxiety (Hedges’ g range: − 0.43 to − 1.58) [37, 38, 40–42]. These interventions were typically delivered via social media (WeChat) with structured guidance from a multidisciplinary team and culturally adapted content.

• Six trials conducted in Western countries, predominantly fully self-guided, reported non-significant between-group differences [39, 43–46, 48]. Effect sizes were close to zero or favored the control condition; all were rated as having “some concerns” or “high risk” due to high attrition and lack of blinding.

The marked discrepancy between Eastern and Western trials is likely attributable to differences in guidance intensity, cultural tailoring, program duration, and baseline symptom severity. The positive Chinese trials enrolled women with moderate-to-severe baseline symptoms (mean EPDS > 12) and provided active guidance; the null Western trials predominantly enrolled low-severity or community samples with fully self-guided, generic mindfulness apps.

No hybrid interventions requiring reclassification were identified; all mindfulness-based programs were judged to have mindfulness/compassion as the modal therapeutic technique.

Certainty of evidence was rated very low (downgraded twice for inconsistency, once for indirectness, and once for suspected publication bias). The effects of digital mindfulness interventions for perinatal depression and anxiety are very uncertain.

Repetitive transcranial magnetic stimulation (rTMS)

Efficacy

One small sham-controlled RCT (n = 26) met inclusion criteria for efficacy [25]. Active rTMS (1 Hz right DLPFC, 900 pulses/session, 20 sessions) produced a significant time-by-group interaction on the HDRS-17 (p = 0.003). Response rates were 81.8% (active) vs. 45.5% (sham); remission rates were 27.3% vs. 18.2%. This trial was rated as having “some concerns” (small sample size, unclear blinding success, possible expectancy effects). Certainty of evidence for rTMS efficacy was rated very low (downgraded twice for imprecision, once for risk of bias).

Seventeen uncontrolled studies (open-label trials and case series; total n ≈ 230 pregnancies) reported large pre-post reductions in depressive symptoms. These studies were uniformly rated at serious or critical risk of bias (ROBINS-I) due to lack of control groups, confounding by concurrent treatment, and regression to the mean. These uncontrolled estimates cannot be interpreted as evidence of efficacy and are excluded from GRADE rating for efficacy.

Consistent with our eligibility criteria, five case reports/small series (n = 1–3) were excluded from efficacy synthesis; they are addressed under safety only.

Safety

Safety data were extracted from all available sources, including the sham-controlled RCT, open-label trials, and case reports/series (n < 5) to maximize the limited cumulative exposure information. Maternal and fetal/neonatal outcomes are reported separately (Table 3).

rTMS was generally well tolerated. The most common adverse event was mild, transient headache, reported in 10–40% of active rTMS participants across studies. Less frequent events included scalp discomfort (5–10%), neck pain (3–5%), and transient uterine contractions (two cases; resolved spontaneously without intervention). No seizures, syncope, or treatment-emergent mania were reported.

Across approximately 230 rTMS-exposed pregnancies, no major congenital anomalies, intrauterine fetal demise, or definitive cases of fetal harm were attributed to rTMS. One preterm birth at 36 weeks was reported in a case series [51]; causality could not be established. No consistent patterns of low birthweight, low Apgar scores, or neonatal intensive care admissions emerged.

The cumulative sample remains small (≈ 230 pregnancies), follow-up is largely confined to the immediate perinatal period, and no study included systematic long-term neurodevelopmental assessment. Absence of observed harm is not equivalent to evidence of safety. Certainty of evidence for safety was rated low (downgraded twice for imprecision).

Summary of findings

• Therapist-guided digital CBT is supported by moderate-certainty evidence and demonstrates consistent, clinically meaningful effects in most but not all trials.

• Self-guided digital CBT is supported by low-certainty evidence; effects are modest and inconsistent.

• Digital mindfulness interventions are supported by very-low-certainty evidence; effects vary dramatically by guidance, cultural context, and baseline severity.

• rTMS efficacy rests on a single small RCT (very-low-certainty evidence); large, definitive trials are urgently needed.

• rTMS safety appears reassuring based on available data, but cumulative exposure is too limited to draw firm conclusions (low-certainty evidence).

Data availability

All extracted data, effect size calculations, and risk-of-bias ratings are provided in Supplementary Table S3.

Discussion

Principal findings

This systematic review synthesizes evidence from 29 studies evaluating digital mental health interventions and repetitive transcranial magnetic stimulation (rTMS) for perinatal depression and anxiety. Three principal findings emerge.

First, therapist-guided digital cognitive behavioral therapy (CBT) is supported by moderate-certainty evidence and demonstrates consistent, clinically meaningful reductions in depressive symptoms across multiple RCTs, although effect sizes vary and two large trials reported null findings. Self-guided digital CBT shows smaller, less consistent effects (low certainty).

Second, digital mindfulness- and compassion-based interventions yield highly heterogeneous outcomes. Positive effects are largely confined to guided, culturally tailored programs delivered in non-Western settings; fully self-guided Western apps have not demonstrated efficacy. The overall certainty of evidence is very low.

Third, rTMS remains an experimental treatment for perinatal depression. A single small sham-controlled RCT provides very-low-certainty evidence of efficacy. While uncontrolled studies report large symptom improvements, these data are critically confounded and cannot establish efficacy. Safety data, though reassuring in the short term, are derived from a small cumulative sample (≈ 230 pregnancies) with limited follow-up; absence of observed harm is not equivalent to evidence of safety.

These findings underscore that digital CBT and rTMS occupy complementary roles in a stepped-care perinatal mental health framework, but their evidence bases are at fundamentally different stages of maturity.

Digital cognitive-behavioral interventions

The finding that therapist-guided digital CBT produces moderate-to-large effects is consistent with prior meta-analyses in perinatal [17] and general adult populations [15]. However, the range of effect sizes (Hedges’ g 0.5–1.4) requires careful contextualization.

The largest effects (g > 1.0) derive from small, single-site trials in clinically diagnosed, treatment-seeking samples using waitlist controls [27, 28]. These trials were rated as having “some concerns” on RoB-2, primarily due to lack of therapist blinding and reliance on self-report outcomes. Waitlist comparators are known to inflate effect estimates by fostering passive symptom monitoring and differential expectations [26]. Conversely, two large, adequately powered trials that used active or standard-care comparators and blinded outcome assessment found no significant between-group differences [33, 47]. Both enrolled predominantly subthreshold or mixed preventive populations and were stopped early for futility in one case.

Taken together, these patterns suggest that the true effect of therapist-guided digital CBT is likely more modest than the upper range of published effect sizes implies, and that baseline severity, comparator type, and methodological quality are potent moderators. Guidance alone does not guarantee efficacy; intervention content, platform fidelity, and population selection are equally critical. These nuances are often obscured in narrative reviews that report effect ranges without accompanying methodological critique.

Self-guided digital CBT produced smaller, inconsistent effects (g = 0.2–0.6; low certainty). High attrition (25–40%) was pervasive, suggesting that engagement not merely access is a fundamental barrier. Future research should prioritize dismantling designs to identify the minimal effective dose of guidance and the role of human support in sustaining adherence.

Digital mindfulness- and compassion-based interventions

The pronounced discrepancy between Eastern and Western trials of digital mindfulness interventions warrants explicit discussion. All five positive trials were conducted in China, employed guided or partially guided formats delivered via social media, and enrolled women with moderate-to-severe baseline symptoms [37, 38, 40–42]. In contrast, all six null trials were conducted in Western countries, used fully self-guided, commercially available mindfulness apps, and recruited low-severity community or preventive samples [39, 43–46, 48].

This heterogeneity cannot be dismissed as random variation; it likely reflects systematic differences in intervention intensity, cultural adaptation, and population severity. Guided programs with structured feedback, culturally tailored language, and collective delivery via familiar platforms (WeChat) may enhance engagement and clinical impact. Unguided generic apps, while scalable, may be insufficient for women with clinically significant symptoms and lack the accountability and personalization that drive behavior change.

Critically, the current evidence does not support the effectiveness of self-guided digital mindfulness apps for perinatal depression or anxiety. Certainty is very low, and claims of efficacy based on pre-post designs or small pilot studies should be viewed skeptically. Researchers should move beyond simple efficacy questions to investigate for whom, under what conditions, and with what level of support these interventions confer benefit.

Repetitive transcranial magnetic stimulation (rTMS)

Efficacy. The evidence base for rTMS in perinatal depression remains nascent. Only one small sham-controlled RCT (n = 26) has been published [25], reporting favorable response rates. This trial, while methodologically rigorous for an early-phase study, is underpowered for definitive efficacy estimates and was rated as having “some concerns” regarding blinding integrity and potential expectancy bias. The remaining 17 studies are uncontrolled, uniformly rated at serious or critical risk of bias, and cannot contribute to efficacy conclusions. The large pre-post effect sizes observed in these studies are expected given regression to the mean, natural recovery, and non-specific treatment effects.

The certainty of evidence for rTMS efficacy is therefore very low. We strongly caution against interpreting the uncontrolled literature as supportive of clinical effectiveness. Large, multicenter, sham-controlled RCTs with standardized protocols, blinded outcome assessment, and longer follow-up are urgently required before rTMS can be recommended for routine perinatal care.

Safety. We systematically separated maternal from fetal/neonatal adverse events to improve clarity. Across all sources (including case reports with n < 5, which were excluded from efficacy but included for safety), no seizures, mania, or maternal deaths were reported. Mild headache was the most common complaint. Fetal outcomes were generally reassuring, with no pattern of congenital anomalies or consistent adverse neonatal events.

However, the cumulative sample of rTMS-exposed pregnancies remains approximately 230, with follow-up largely confined to delivery. No study has systematically assessed longer-term neurodevelopmental outcomes. Absence of observed harm is not equivalent to evidence of safety. We therefore rate the certainty of the safety evidence as low, downgraded twice for imprecision. Clinicians discussing rTMS with perinatal patients must convey this uncertainty transparently.

Safety of digital and neuromodulation interventions

Digital interventions were uniformly well tolerated; no serious adverse events were reported across any included study. Dropout rates often exceeding 25% in self-guided arms are better conceptualized as engagement failures rather than safety events, but they nonetheless represent a significant threat to effectiveness and scalability.

rTMS safety is discussed in detail above. We emphasize that the current safety database, while encouraging, is insufficient to definitively establish the absence of rare or delayed adverse outcomes. Prospective registries with standardized outcome ascertainment and longitudinal follow-up of offspring are needed to strengthen the evidence base.

Strengths and limitations

Strengths of this review include: (1) comprehensive, PRISMA-2020-compliant search across six databases and grey literature; (2) dual independent screening, extraction, and risk-of-bias assessment; (3) explicit integration of risk-of-bias ratings into narrative synthesis; (4) separate GRADE assessments for efficacy and safety; (5) transparent handling of small case series for safety data only; and (6) contextualized reporting of effect sizes with attention to clinical and methodological moderators.

Limitations must be acknowledged. First, clinical and methodological heterogeneity precluded meta-analysis; although we provided structured narrative synthesis, quantitative pooling would have permitted more precise effect estimation. Second, the majority of included studies (particularly for mindfulness and rTMS) were small, early-phase trials with short follow-up, limiting certainty. Third, we relied on published data only; individual participant data were not obtained. Fourth, we did not formally assess for publication bias beyond inspection of funnel plots for larger intervention subgroups, though selective reporting is likely (particularly for mindfulness apps). Fifth, our decision to exclude case reports (n < 5) from efficacy synthesis may have omitted some unpublished null results, but this conservative approach aligns with best practices for systematic reviews. Finally, the review does not directly compare digital CBT and rTMS nor provide formal clinical practice guidelines; these aims were explicitly out of scope.

Implications for research and practice

Therapist-guided digital CBT can be offered as an evidence-based, scalable treatment option for perinatal depression, particularly for individuals with mild-to-moderate symptoms who prefer non-pharmacologic care or face barriers to in-person therapy. However, not all guided programs are equally effective; clinicians should preferentially select interventions with demonstrated efficacy in perinatal populations. Self-guided digital CBT may be considered a low-intensity option, but patients should be informed that effects are modest and engagement is often suboptimal. Digital mindfulness apps, particularly fully self-guided versions, cannot currently be recommended as monotherapy for clinically significant perinatal depression or anxiety.

rTMS should remain restricted to research settings or exceptional clinical circumstances (e.g., severe, treatment-resistant depression in pregnancy where medication is declined or ineffective) until adequately powered controlled trials confirm efficacy and expand the safety database. Shared decision-making must include explicit discussion of the very low certainty of evidence.

Priorities include: (1) large, non-inferiority trials comparing guided digital CBT to face-to-face CBT; (2) factorial or dismantling trials to identify the optimal intensity and format of guidance; (3) culturally adapted digital interventions for diverse perinatal populations; (4) pragmatic effectiveness-implementation hybrid trials to evaluate scalability in real-world settings; (5) multicenter, sham-controlled rTMS RCTs with embedded safety registries and long-term neurodevelopmental follow-up; and (6) head-to-head comparisons of different neuromodulation modalities (e.g., rTMS vs. tDCS) once preliminary safety and efficacy are established.

Conclusion

Therapist-guided digital CBT is a well-supported intervention for perinatal depression and should be integrated into stepped-care models. Self-guided digital CBT and digital mindfulness interventions are supported by low- or very-low-certainty evidence, respectively, and their use requires careful patient selection and expectation management. rTMS remains experimental; its efficacy is unproven, and the safety database, while reassuring, is too limited to draw definitive conclusions. Bridging the evidence gaps for both modalities will require sustained commitment to rigorous, adequately powered, and inclusively designed perinatal mental health research.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

BAI

Beck Anxiety Inventory

BDI/BDI-II

Beck Depression Inventory

CAQ

Cardiac Anxiety Questionnaire

CES-D

Center for Epidemiologic Studies Depression Scale

CGI-S

Clinical Global Impression-Severity

DASS

Depression Anxiety Stress Scales

EPDS

Edinburgh Postnatal Depression Scale

GAD-2/7

Generalized Anxiety Disorder Scale-2/7

HADS/HADS-A

Hospital Anxiety and Depression Scale (Anxiety subscale)

HAMD/HDRS

Hamilton Depression Rating Scale

MADRS-S

Montgomery-Åsberg Depression Rating Scale, self-rated version

PHQ-2/9

Patient Health Questionnaire-2/9

SAS

Self-Rating Anxiety Scale

W-DEQ

Wijma Delivery Expectancy/Experience Questionnaire

CBT

Cognitive Behavioral Therapy

iCBT

Internet-based Cognitive Behavioral Therapy

RCT

Randomized Controlled Trial

rTMS

Repetitive Transcranial Magnetic Stimulation

TAU

Treatment as Usual

IG

Intervention Group

CG

Control Group

DLPFC

Dorsolateral Prefrontal Cortex

MT

Motor Threshold

Author contributions

Conceptualization: Z.B., M.A.; Methodology: Z.B., M.A., A.D.; Investigation: Z.B., M.A., A.D.; Formal Analysis: Z.B., M.A.; Writing – Original Draft: Z.B.; Writing – Review & Editing: M.A., A.D.; Supervision: A.D. All authors have read and agreed to the published version of the manuscript.

Funding

Not applicable.

Data availability

The data supporting the findings of this study are available in the supplementary material of this article.

Declarations

Ethical approval

The authors have nothing to report.

Consent

The authors have nothing to report.

Competing interests

The authors declare no competing interests.