Optimizing care for women experiencing pregnancy loss: Insights from a systematic review and meta-analysis

Protocol and registration

This systematic review and meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines. The study protocol was registered a priori with PROSPERO (International Prospective Register of Systematic Reviews) Prieyadharshini Jayaprakasam, Jeyaram Nadarajan Saraswathy. Psychological Impact of Pregnancy Loss and the Effectiveness of Interventions: A Systematic Review and Meta-Analysis. PROSPERO 2024 Available from https://www.crd.york.ac.uk/PROSPERO/view/CRD42025635112 prior to data extraction to enhance transparency and minimize selective reporting bias. The review synthesized evidence from 18 studies published between 2020 and 2024, with the objective of identifying, appraising, and synthesizing interventions for optimizing care across the pregnancy loss trajectory, encompassing prevention of recurrent pregnancy loss, management of acute pregnancy loss, and treatment of psychological sequelae.

Information sources and search strategy

A comprehensive literature search was conducted in December 2024 across three major electronic databases: PubMed (MEDLINE), Embase, and the Cochrane Central Register of Controlled Trials (CENTRAL). The search was designed to identify all relevant publications from January 1, 2020, through December 31, 2024, ensuring focus on contemporary evidence regarding pregnancy loss interventions. This five-year window was selected to capture recent advances while maintaining manageable review scope. The search strategy combined controlled vocabulary (Medical Subject Headings for PubMed and EMTREE terms for Embase) with free-text keywords to maximize sensitivity.

The PubMed search strategy identified 600 records using the following terms: (“pregnancy loss” OR “miscarriage” OR “spontaneous abortion” OR “recurrent pregnancy loss” OR “stillbirth” OR “ectopic pregnancy”) AND (“management” OR “treatment” OR “intervention” OR “care” OR “support”). The Embase search yielded 400 records using comparable EMTREE terms with free-text variations. The Cochrane Library search identified 200 records of controlled trials and systematic reviews. Supplementary search methods included manual review of reference lists from included studies and relevant review articles, which identified 40 additional records.

No language restrictions were applied during the search process, though practical limitations restricted extraction to English-language publications. The searches were conducted by two independent reviewers to minimize selection bias and ensure comprehensive database coverage. Database-specific search strategies were documented and retained for transparency and reproducibility.

Study selection criteria

Inclusion criteria

Studies were included if they met the following criteria: (1) Population: women of reproductive age (≥18 years) experiencing or at risk for pregnancy loss, including those with recurrent pregnancy loss, recent pregnancy loss, or prior loss history; (2) Intervention: any intervention aimed at preventing recurrent pregnancy loss, managing acute pregnancy loss, supporting psychological recovery following loss, or optimizing reproductive outcomes for women with prior loss; (3) Comparator: studies with comparative data including standard care, placebo, no intervention, or alternative interventions; (4) Outcomes: studies reporting reproductive outcomes (live birth rate, pregnancy loss rate) or psychological outcomes (depression, anxiety, PTSD, quality of life); (5) Study design: randomized controlled trials (RCTs), observational studies with comparative groups (cohort and case-control designs), systematic reviews and meta-analyses synthesizing primary research, and economic evaluations reporting clinical outcomes.

Exclusion criteria

Studies were excluded if they: (1) employed non-comparative designs (case reports, case series); (2) included sample sizes <10 participants; (3) were non-English publications; (4) were duplicate publications from the same cohort (most complete or recent publication retained); (5) had missing outcome data with >50% attrition without intention-to-treat analysis; or (6) reported only surrogate outcomes without clinical endpoints.

Study selection process

The study selection process employed a systematic two-stage approach with dual independent reviewers. Stage 1 involved title and abstract screening of all 1,050 unique records (following removal of 190 duplicates) using standardized screening forms and Covidence systematic review software. Studies were included at this stage if they potentially met inclusion criteria based on title and abstract content. Stage 2 involved full-text assessment of 100 articles against complete inclusion and exclusion criteria. At full-text assessment, 82 articles were excluded with documented reasons: 50 for irrelevant populations, 20 for unsuitable study designs, and 12 for insufficient outcome data. Reviewer disagreement at full-text stage was resolved through discussion and consensus, with 96% inter-rater agreement achieved. The complete selection process is documented in the PRISMA 2020 flow diagram.

Data extraction

Data extraction was performed by two independent reviewers using a standardized data extraction form developed and pilot-tested on a subset of five studies.²⁷ The form captured the following information: (1) Bibliographic information: first author, publication year, country of origin, and funding source; (2) Study design characteristics: study design type, sample size, randomization method (for RCTs), allocation concealment, blinding status, and duration of follow-up; (3) Population characteristics: participant age, baseline pregnancy loss history, inclusion and exclusion criteria, and recruitment settings; (4) Intervention characteristics: detailed intervention description, dose/intensity, delivery method, and duration; (5) Outcome measures: specific outcomes assessed, measurement instruments used, and timing of outcome assessment; (6) Results: numerical outcome data including means/medians with standard deviations, event counts with group totals, p-values, and 95% confidence intervals; and (7) Risk of bias indicators relevant to the quality assessment tools applied.

Extracted data were entered independently by each reviewer, compared for discrepancies, and verified against original manuscripts. Data extraction was managed in Excel spreadsheets with quality assurance checks performed to ensure accuracy and completeness.

Quality assessment and risk of bias evaluation

Quality assessment and risk of bias evaluation were performed using methodology-specific tools appropriate to study design. The Cochrane Risk of Bias Tool 2 (RoB 2) was applied to all randomized controlled trials, assessing risk of bias across five domains: (1) bias arising from the randomization process, (2) bias due to deviations from intended interventions, (3) bias due to missing outcome data, (4) bias in outcome measurement, and (5) bias in selection of reported results. Each domain was rated as “Low risk,” “Some concerns,” or “High risk,” with an overall study-level judgment synthesized across domains.

The Newcastle-Ottawa Scale (NOS) was applied to observational studies (cohort and case-control designs), evaluating selection of study groups, comparability of groups, and assessment of outcomes/exposures. Studies scoring 7-9 on the NOS were classified as high quality, 5-6 as moderate quality, and <5 as low quality. For systematic reviews and meta-analyses, the AMSTAR 2 tool evaluated methodological quality. Economic evaluations were assessed using the Quality of Economics Studies checklist.

Quality assessment was performed by two independent reviewers using Covidence software, with disagreements resolved through discussion and consensus. Risk of bias information was synthesized graphically using traffic light plots (green = low risk, yellow = moderate/some concerns, red = high risk) displaying all studies and assessment domains. The assessment resulted in classification of 6 studies (33.3%) as low risk, 11 studies (61.1%) as moderate risk, and 1 study (5.6%) as high risk of bias.

Data synthesis and statistical analysis

Categorization of studies for analysis

The 18 included studies were categorized into two groups: (1) Quantitative analysis: 11 studies with measurable effect sizes including 6 RCTs, 4 observational studies, and 1 meta-analysis were included in meta-analysis; and (2) Qualitative synthesis: all 18 studies were included in narrative synthesis organized by intervention type and outcome domain.

Effect size extraction and calculation

For binary outcomes (live birth rate, pregnancy loss rate), relative risk (RR) was extracted or calculated as: RR = (a/n1)/(c/n2), where a = events in intervention group, n1 = total intervention group, c = events in control group, n2 = total control group. For continuous outcomes (psychological symptom scores), standardized mean difference (SMD, Cohen’s d) was calculated using: SMD = (M1 - M2)/Spooled, where M1 and M2 are group means and Spooled is the pooled standard deviation. All effect sizes were accompanied by 95% confidence intervals.

Meta-analysis methodology

A random-effects meta-analysis model (DerSimonian-Laird estimator) was employed to combine effect sizes across studies to account for expected heterogeneity in populations, interventions, and outcome assessment approaches. Effect sizes were pooled separately by outcome domain (reproductive outcomes and psychological outcomes). Inverse-variance weighting was used to weight individual study estimates.

Heterogeneity assessment

Between-study heterogeneity was quantified using the I² statistic, representing the percentage of total variation due to heterogeneity rather than random chance, and the Cochrane Q test evaluating statistical significance. I² was interpreted as: <25% = low heterogeneity, 25-50% = moderate, 50-75% = substantial, >75% = considerable heterogeneity. Q-statistic p-values <0.05 indicated statistically significant heterogeneity. The meta-analysis revealed I² = 78.7% in the original analysis, indicating high heterogeneity that was further explored through subgroup and sensitivity analyses.

Subgroup analyses

A priori-specified subgroup analyses were conducted to evaluate potential sources of heterogeneity: (1) Pregnancy loss type (miscarriage vs. stillbirth vs. ectopic pregnancy); (2) Intervention type (pharmacological, psychological, assisted reproductive technology, diagnostic); (3) Study quality (low vs. moderate/high risk of bias); (4) Geographic region (USA/Scandinavia vs. China vs. other countries); and (5) Delivery method for psychological interventions (in-person vs. telemedicine). Subgroup analyses used the same random-effects pooling approach as primary analyses, with statistical interaction tests evaluating whether effect size differences across subgroups were significant (p < 0.05).

Sensitivity analyses

Multiple sensitivity analyses were conducted to evaluate robustness of primary findings: (1) excluding the single high-risk study (S4); (2) restricting to RCTs only; (3) restricting to studies with sample size >100; (4) restricting to studies published 2021 or later; and (5) using alternative effect size metrics (odds ratios vs. relative risk). For each analysis, I² heterogeneity was recalculated to evaluate whether conclusions remained stable across analytical variations.

Publication bias assessment

Publication bias was assessed using funnel plot visualization examining the relationship between study effect size (x-axis) and standard error (y-axis). Symmetric distribution around the mean effect line indicated absence of publication bias, while asymmetry suggested potential bias. Egger’s linear regression test quantitatively evaluated funnel plot asymmetry using the formula: SE (ES) = α + β × ES. Non-significant intercept (p > 0.05) indicated no statistically significant publication bias. Trim-and-fill analysis estimated the number of potentially missing unpublished studies by iteratively removing and replacing extreme effects until funnel symmetry was achieved.

Certainty of evidence assessment

The Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework was applied to evaluate certainty of evidence for major outcomes. Evidence certainty was assessed across five dimensions: study design, risk of bias, consistency, directness, and publication bias. Each outcome was assigned one of four certainty levels: High (multiple well-designed RCTs with consistent findings and low bias), Moderate (RCTs with some limitations or good-quality observational studies), Low (observational studies or RCTs with substantial limitations), or Very Low (serious limitations with inconsistent findings).

Quality assurance and reproducibility

All screening decisions, data extraction, and quality assessments were performed by two independent reviewers with systematic documentation of all decisions and discrepancies. Analyses were conducted using R statistical software version 4.3 with the metafor package for meta-analysis. Systematic review management and screening were conducted using Covidence software. Study characteristics, risk of bias assessments, and outcome data are provided in supplementary tables for full transparency and reproducibility.

Software and tools

Data analyses were performed using RevMan (Review Manager (RevMan) [Computer program]. Version 5.4, The Cochrane Collaboration, 2020.)¹⁷ for meta-analysis, R software (R Core Team. (2024). R: A language and environment for statistical computing (Version 4.4.2), Python (Python Software Foundation. (n.d.). Python Software Foundatio) for advanced statistical analyses, including meta-regression and visualization. For data extraction and management, the free and open-source software LibreOffice Calc (The Document Foundation. (n.d.). LibreOffice Calc (Version 7.5)) was used as an alternative to proprietary spreadsheet tools. This methodologically robust approach, underpinned by the PRISMA guidelines and PICO framework, ensured comprehensive and transparent analysis of the included studies.

Study characteristics and selection process

Study selection outcomes and characteristics

The comprehensive literature search identified 1,200 initial records from three major electronic databases: PubMed (n = 600), Embase (n = 400), and Cochrane Library (n = 200), supplemented by manual reference review (n = 40 additional records). Following removal of 190 duplicate records, 1,050 unique records underwent title and abstract screening. Subsequent full-text assessment of 100 articles resulted in 82 exclusions: 50 articles with irrelevant populations, 20 with unsuitable study designs, and 12 with insufficient data. A total of 18 studies met inclusion criteria and were included in the qualitative and quantitative synthesis. The geographic distribution of included studies demonstrated representation from multiple regions, with the United States contributing 7 studies, China 4 studies, and Scandinavia (Denmark and Norway) 4 studies, along with smaller representations from international collaborations and single studies from South Africa and Afghanistan. This geographic diversity strengthens the evidence base’s applicability across diverse healthcare contexts and economic settings.

Figure 1. PRISMA 2020 flow diagram depicting the study selection process.

The diagram illustrates the flow of records through the identification, screening, eligibility, and inclusion phases, resulting in 18 studies included in the final review. PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

Table 1 presents comprehensive characteristics of all 18 included studies, including study design, participant populations, interventions, primary and secondary outcomes, geographic locations, and risk of bias assessment. The studies employed diverse methodological approaches: 11 randomized controlled trials (RCTs) representing the gold standard for intervention efficacy assessment, 4 observational/cohort studies providing real-world implementation data, 2 systematic reviews and meta-analyses synthesizing broader evidence, and 1 economic evaluation assessing cost-effectiveness. Study designs ranged from small single-center studies to large multicenter trials, with sample sizes varying from not reported in some studies to over 1,000 participants in large efficacy trials.

The interventions evaluated across these 18 studies represented five major categories. Pharmacological interventions (5 studies, 27.8%) included mifepristone combined with misoprostol, low-dose aspirin, tranexamic acid, home-based misoprostol, and intravenous immunoglobulin. Psychological interventions (3 studies, 16.7%) encompassed mindfulness-based stress reduction, art therapy, and cognitive-behavioral therapy delivered through various models including task-sharing approaches. Assisted reproductive technologies (4 studies, 22.2%) focused on frozen embryo transfer protocols, progesterone monitoring, FET timing optimization, and early newborn care integration. Diagnostic and monitoring approaches (2 studies, 11.1%) examined serum lipid profiling and diagnostic protocols in rural settings. Observational and comparative studies (4 studies, 22.2%) evaluated physical activity impacts and comprehensive reviews of interventions.

Figure 2. Timeline of studies - Annual publication distribution (2020-2024).

The Figure 2 temporal distribution of included studies shows research activity concentrated in 2020 and 2024, each with 5 studies, with an intermediate publication in 2022 (1 study) and 2023 (3 studies). This pattern demonstrates sustained research interest in pregnancy loss management interventions, with recent publications suggesting continued emphasis on optimizing care approaches for affected women. The spike in 2024 publications indicates current research momentum in this critical clinical area.

Figure 3. Distribution of studies by intervention type category.

The Figure 3 shows 18 included studies distributed across five major intervention categories shows balanced representation, with no single category dominating the evidence base. Pharmacological interventions (5 studies, 27.8%) and observational/comparative studies (4 studies, 22.2%) represent the largest categories, followed by ART interventions (4 studies, 22.2%), psychological interventions (3 studies, 16.7%), and diagnostic/monitoring approaches (2 studies, 11.1%). This balanced distribution ensures comprehensive assessment of multiple intervention strategies for pregnancy loss management.

Quantitative results and effect sizes

Detailed intervention outcomes with effect sizes and confidence intervals

The meta-analysis compiled quantitative data from 11 major interventions with measurable effect sizes, representing the core of the evidence base for pregnancy loss management. These interventions span pharmacological, psychological, and reproductive technology domains, each with distinct mechanisms of action and targeted outcomes. The quantitative results reveal substantial efficacy for multiple intervention approaches, with confidence intervals demonstrating precision and statistical significance supporting clinical implementation across diverse populations and healthcare settings.

The quantitative results demonstrate exceptional efficacy for mifepristone and misoprostol combination therapy, the cornerstone pharmacological approach for early pregnancy loss management. Two studies (S3, S11) evaluating this combination reported nearly identical effect sizes of OR = 3.5 (95% CI: 2.8-4.3) and OR = 3.8 (95% CI: 3.2-4.4), respectively, indicating a 3.5- to 3.8-fold improvement in tissue expulsion success compared to comparator treatments. This extraordinary consistency across independent studies strengthens confidence in the intervention’s efficacy. The clinical significance extends beyond tissue expulsion rates; the combination regimen reduces the need for surgical intervention by 63%, substantially decreasing procedural morbidity and psychological trauma associated with uterine aspiration procedures. Patient satisfaction with medical management substantially exceeds that of surgical approaches, supporting patient autonomy and reducing iatrogenic distress during already-traumatic circumstances.

Low-dose aspirin therapy emerged as another highly promising pharmacological intervention, particularly valuable for women with recurrent pregnancy loss history. The meta-analysis finding (S9) demonstrated a 22% relative improvement in live birth rates (RR = 1.22, 95% CI: 1.10-1.35, p < 0.05), representing a clinically meaningful benefit with exceptional cost-effectiveness. The mechanism of aspirin’s benefit remains multifactorial, including enhancement of placental function, reduction of thrombotic complications, and modulation of inflammatory responses that may contribute to pregnancy loss. Notably, this intervention demonstrates remarkable safety, with minimal adverse effects reported even in large trials, making it universally applicable across high-resource and resource-limited settings alike.

Assisted reproductive technology interventions showed variable but meaningful benefits, with immediate frozen embryo transfer (FET) protocols demonstrating superior outcomes. The multicenter RCT (S5) reported a 15% improvement in clinical pregnancy rates (RR = 1.15, 95% CI: 1.07-1.23, p < 0.01), while the comprehensive meta-analysis (S18) of FET timing studies demonstrated even more impressive results, with a 1.85-fold improvement in live birth rates (RR = 1.85, 95% CI: 1.65-2.10, p < 0.001). The superiority of immediate FET protocols carries dual significance: enhanced pregnancy achievement combined with reduced psychological burden from eliminated waiting periods between treatment cycles. For women who have already experienced pregnancy loss and face uncertainty regarding future pregnancy success, the elimination of unnecessary delays provides both clinical benefit and psychological relief.

Psychological interventions showed variable but meaningful clinical significance. Mindfulness-based stress reduction (S1) achieved a standardized mean difference of -0.48 (95% CI: -0.65 to -0.30, p < 0.05), indicating nearly half a standard deviation reduction in perceived stress. While this effect size appears modest compared to pharmacological interventions, it represents clinically meaningful psychological benefit for women experiencing acute grief. Art therapy (S2) demonstrated substantial quality of life improvements across multiple domains (physical, psychological, social, environmental), with effect sizes reaching levels comparable to or exceeding those of some pharmacological interventions for specific psychological outcomes. Tranexamic acid (S15), an adjunctive pharmacological approach for postpartum hemorrhage, showed significant effectiveness (OR = 1.8, 95% CI: 1.5-2.1, p < 0.05) with particularly important implications for low-resource settings where blood product availability remains limited.

Figure 4. Forest plot - Comprehensive effect sizes and confidence intervals for major interventions.

Figure 4 forest plot visualization displays all 11 major interventions with their point estimates (diamond markers) and 95% confidence intervals (horizontal lines). The horizontal scale ranges from -1 (protective for stress outcomes) to 5 (large beneficial effect for binary outcomes). Color-coding by intervention type highlights patterns: pharmacological interventions (shown in red) consistently demonstrate the largest effect sizes, ranging from OR = 1.8 to 3.8; psychological interventions (shown in blue) show moderate effects from SMD = -0.48 to 2.1; assisted reproductive technology interventions (shown in green) demonstrate variable effects from RR = 1.0 to 1.85. The width of confidence intervals indicates precision, with larger studies generally showing narrower intervals indicating more stable estimates. All confidence intervals exclude the null value (1.0 for RR/OR; 0 for SMD), confirming statistical significance across all major interventions.

Psychological outcomes: prevalence, severity, and burden

Pooled prevalence of mental health conditions following pregnancy loss

The meta-analysis calculated pooled prevalence rates for three major psychological conditions following pregnancy loss using random-effects models that accounted for between-study variability. Depression emerged as the most prevalent mental health condition, affecting 39% of women (95% CI: 36%-42%) following pregnancy loss. This represents nearly 4 in 10 women experiencing clinically significant depressive symptoms, a staggering prevalence that underscores the profound psychological impact of pregnancy loss. The narrow confidence interval indicates precise estimation across included studies, with upper and lower bounds remaining tightly centered around the point estimate, suggesting consistency of depressive symptomatology across diverse populations and geographic settings.

Anxiety disorders affected approximately 30% of women (95% CI: 27%-33%), representing moderate-to-high prevalence and often co-occurring with depressive symptoms. The overlap in confidence intervals between anxiety and depression (overlapping estimates around 30-39%) suggests significant comorbidity, with many women experiencing both conditions concurrently. Post-traumatic stress disorder was identified in 17% of women (95% CI: 15%-19%), indicating less common but clinically significant psychological sequelae. Notably, the lower confidence bound of PTSD (15%) indicates that approximately 1 in 6 women experiences this serious condition, a prevalence sufficient to warrant systematic screening and intervention protocols in all pregnancy loss care settings.

Approximately 60% of women experienced at least one of these three psychological conditions, indicating that the substantial majority of women experiencing pregnancy loss suffer meaningful mental health burden. This finding has profound implications for clinical care, suggesting that psychological screening and intervention should be considered standard of care rather than optional services for women following pregnancy loss.

The psychological conditions following pregnancy loss carry substantial functional impairment. Depression involves “significant impairment in daily functioning, relationships, work capacity,” with effects often persisting “>6 months if untreated.” Anxiety manifests as “persistent worry, difficulty concentrating, physical symptoms” with “variable: weeks to months” duration. Post-traumatic stress disorder produces “flashbacks, hypervigilance, avoidance behaviors, intrusive thoughts,” and is “often chronic if untreated.” Comorbidity patterns reveal that approximately 50% of women with depression experience concurrent anxiety, 45% of anxious women concurrently experience depression, and 60% of those with PTSD experience concurrent depression/anxiety diagnoses. These high comorbidity rates indicate that women experiencing pregnancy loss often suffer complex, overlapping mental health challenges requiring comprehensive, integrated treatment approaches.

Figure 5. Pooled prevalence mental health conditions with confidence intervals.

Figure 5 visualization displays the three psychological conditions with point estimates (bar heights) and 95% confidence intervals (error bars extending above and below each bar). Depression reaches the highest bar at 39%, with confidence interval extending from 36% to 42%, reflecting the substantial and precise estimate of depression prevalence. Anxiety shows a notably shorter bar at 30% with confidence interval from 27% to 33%, clearly distinguishing it as less prevalent than depression but still affecting approximately 3 in 10 women. PTSD demonstrates the shortest bar at 17% with confidence interval from 15% to 19%, the lowest prevalence but with sufficient magnitude to require clinical attention. The narrow confidence intervals across all three conditions indicate robust estimation with good precision across the included studies.

Subgroup analyses

Psychological outcomes by type of pregnancy loss

Subgroup analysis revealed meaningful variation in psychological burden depending on the specific type of pregnancy loss experienced. This differentiation carries important clinical implications, as the nature and intensity of psychological support should logically align with the magnitude of psychological burden associated with different loss types.

Stillbirth (fetal death ≥20 weeks gestation) was consistently associated with the highest psychological burden across all three mental health conditions. Depression prevalence in stillbirth reached 42%, substantially exceeding that observed in miscarriage (38%) or ectopic pregnancy (37%). Similarly, anxiety affected 33% of women experiencing stillbirth compared to 31% with miscarriage and 30% with ectopic pregnancy. Post-traumatic stress disorder was also elevated in stillbirth (18%) relative to other loss types (miscarriage 17%, ectopic 16%). This elevated burden reflects the greater maternal-fetal attachment occurring in later pregnancy, where women have already developed substantial mental representations of the fetus, anticipated bonding, and future planning. By the time of stillbirth, women have experienced quickening, ultrasound visualization of fetal movement, weight gain, and often shared anticipation with family members and social networks, creating a profound grief response upon loss.

Early miscarriage (pregnancy loss <20 weeks gestation) demonstrated moderate psychological impact, with depression prevalence of 38%, anxiety of 31%, and PTSD of 17%. These rates, while substantial, remain somewhat lower than observed in stillbirth, reflecting the typically lower maternal-fetal attachment in early pregnancy when fewer physical and psychological changes have occurred.

Ectopic pregnancy showed the lowest psychological burden overall, with depression prevalence of 37%, anxiety of 30%, and PTSD of 16%. The reduced burden likely reflects the acute medical nature of ectopic pregnancy loss, where the focus on medical management and potential life-threatening complications may occupy psychological resources that might otherwise be devoted to grief processing in less acute loss circumstances.

These findings have critical clinical implications, suggesting that women experiencing stillbirth require more intensive and specialized psychological support compared to those experiencing early loss. Standard support protocols may be adequate for miscarriage and ectopic pregnancy, while stillbirth management should incorporate intensive, specialized interventions including extended grief counseling, trauma-informed care approaches, and specialized perinatal bereavement support services.

Figure 6. Psychological outcomes by pregnancy loss type.

Figure 6 compares psychological outcome prevalence across three loss types. For each loss type (miscarriage, stillbirth, ectopic), three bars represent depression, anxiety, and PTSD prevalence. The chart clearly illustrates that stillbirth (represented by the tallest set of bars) demonstrates consistently higher prevalence across all three mental health conditions. The depression bars show the most marked difference, with stillbirth at 42% distinctly taller than miscarriage (38%) and ectopic (37%). Anxiety and PTSD show similar patterns, though with smaller absolute differences, maintaining the consistent ranking: stillbirth > miscarriage > ectopic. This visual representation emphasizes the graduated psychological burden by loss type, justifying differential support intensity recommendations.

Psychological outcomes by geographic region and healthcare context

Regional subgroup analysis examined whether geographic location and associated healthcare contexts influenced psychological outcome prevalence, recognizing that reported prevalence may reflect not only true population variation but also differences in detection, assessment, and reporting practices.

Regional variation in psychological outcomes was observed, with Europe reporting the highest prevalence across all three conditions (depression 41%, anxiety 32%, PTSD 18%), followed by Africa (depression 40%, anxiety 31%, PTSD 17%), and Asia showing the lowest reported prevalence (depression 37%, anxiety 30%, PTSD 16%). However, these geographic differences are not necessarily indicative of true population variation in psychological vulnerability. Rather, they likely reflect multiple confounding factors: (1) healthcare infrastructure and mental health service availability, with Europe having “widely available” services, Africa “severely limited,” and Asia “limited in many areas”; (2) diagnostic assessment capabilities and standardized tools used in different settings; (3) cultural frameworks for discussing and reporting mental health symptoms, with Europe showing “high awareness” and Africa “low in many settings”; (4) mental health awareness and literacy levels; and (5) study design and population selection characteristics affecting which women were assessed and how comprehensively.

The interpretation of regional variations requires careful contextualization. The higher rates in Europe may reflect superior mental health detection capabilities in developed healthcare systems with widespread psychological screening and standardized assessment instruments. Healthcare infrastructure classified as “advanced/comprehensive” in Europe enables systematic identification of mood and anxiety disorders through routine screening protocols. In contrast, Africa and Asia, with “basic/limited” and “mixed/variable” healthcare infrastructure respectively, may demonstrate under-reporting due to limited mental health services, lack of screening infrastructure, cultural stigma around mental health conditions, and insufficient trained mental health personnel to identify and diagnose psychological conditions. This important distinction emphasizes that women in resource-limited settings experience not only higher baseline risks of pregnancy loss but also reduced access to mental health services—a critical equity consideration requiring urgent systemic attention.

The diagnostic accuracy assessment reinforces this interpretation. Europe demonstrates “high accuracy” in psychological condition identification, Africa is “limited by resources,” and Asia shows “variable by setting.” These differences in diagnostic capacity directly influence reported prevalence rates, with lower diagnostic accuracy in resource-limited settings likely producing substantial under-reporting of actual psychological burden. Stigma levels also vary considerably, with Europe showing “low-moderate” stigma, Africa “high” stigma, and Asia “moderate-high” stigma, affecting whether women with symptoms seek and receive mental health assessment.

Figure 7. Psychological outcomes by geographic region.

Figure 7 displays regional comparison with three major geographic regions (Asia, Africa, Europe) each represented by three bars showing depression, anxiety, and PTSD prevalence. The chart reveals a clear geographic gradient, with Europe showing the tallest set of bars (depression 41%, anxiety 32%, PTSD 18%), Africa intermediate (depression 40%, anxiety 31%, PTSD 17%), and Asia lowest (depression 37%, anxiety 30%, PTSD 16%). This visual representation demonstrates the systematic variation in reported prevalence across regions. However, the interpretation box emphasizes that these differences likely reflect differential diagnostic capacity and healthcare infrastructure rather than true population differences in vulnerability. The chart is accompanied by annotations indicating healthcare context differences: Europe “advanced/comprehensive,” Africa “basic/limited,” Asia “mixed/variable,” making explicit the structural factors contributing to prevalence variation.

Intervention delivery methods

Effectiveness of in-person versus telemedicine interventions

Subgroup analysis compared the pooled effectiveness of psychological and supportive interventions delivered via in-person versus telemedicine modalities. This analysis carries critical implications for healthcare system planning, particularly in contexts where geographic barriers, resource limitations, and workforce shortages create challenges for in-person service delivery.

In-person interventions demonstrated substantially greater effectiveness, with pooled effect size of 0.85 (95% CI: 0.72-0.98), compared to telemedicine approaches with pooled effect size of 0.59 (95% CI: 0.48-0.70). This difference represents a 44.1% relative improvement in efficacy, a clinically substantial advantage favoring in-person delivery. The relative efficacy comparison shows in-person interventions at 100% effectiveness (baseline), while telemedicine achieves 69.4% of in-person effectiveness, indicating that telemedicine is roughly two-thirds as effective as in-person approaches.

Multiple factors likely contribute to the superior effectiveness of in-person interventions. Therapeutic alliance and personal connection form the foundation of psychological intervention effectiveness, and face-to-face contact enables establishment of stronger therapeutic relationships. In-person practitioners can provide real-time emotional support, observe non-verbal communication and emotional responses to adjust interventions dynamically, and create physical space for grief expression. Session completion rates average 85% for in-person interventions versus 65% for telemedicine, indicating higher engagement and retention. Participants report stronger therapeutic alliance ratings with in-person interventions than telemedicine modalities, an important predictor of treatment success.

However, telemedicine offers important complementary advantages for accessibility. The lower cost per session ($50-150 versus $100-250 for in-person) reduces financial barriers for underserved populations. Location independence enables access for geographically remote populations without transportation barriers, individuals with mobility limitations, and those in healthcare deserts with insufficient mental health professionals. Telemedicine’s higher accessibility score acknowledges its potential to reach populations excluded from in-person services.

The clinical implication is that optimal care should employ a two-tier approach: prioritizing in-person interventions when feasible—particularly for women in acute grief and psychological distress—while utilizing telemedicine to reach underserved populations who would otherwise receive no mental health services. This combination approach maximizes both intervention effectiveness and population access, addressing the effectiveness-accessibility trade-off through strategic deployment of each modality where it offers greatest value.

Figure 8. Comparison bar chart - In-person versus telemedicine intervention effectiveness.

The comparative Figure 8 displays side-by-side bars showing pooled effect sizes for in-person (0.85) and telemedicine (0.59) interventions. The in-person bar extends to 0.85 on the effect size scale, while the telemedicine bar reaches 0.59, creating a visually striking difference emphasizing the 44% relative superiority of in-person methods. Error bars extending above and below each bar represent 95% confidence intervals, with in-person showing CI: 0.72-0.98 and telemedicine CI: 0.48-0.70. The non-overlapping confidence intervals indicate statistical distinctness of the two delivery methods, confirming that the difference is not merely chance variation but reflects genuine superiority of in-person approaches. Supplementary data in adjacent text boxes provide practical considerations: session completion rates (85% vs 65%), cost ($100-250 vs $50-150), and therapeutic alliance ratings (strong vs moderate).

Quality assessment and risk of bias

Risk of bias distribution across all included studies

Comprehensive quality assessment was conducted using the Cochrane Risk of Bias Tool for RCTs and Newcastle-Ottawa Scale for observational studies. Assessment across all six Cochrane bias domains (selection bias, performance bias, detection bias, attrition bias, reporting bias, and overall risk) provides nuanced evaluation of methodological quality.

Summary of risk of bias distribution

Six studies (33.3%) demonstrated low risk of bias across all Cochrane domains: Jensen (S1), Song (S5), Naimi (S9), Grantz (S12), Abbas (S14), and Diop (S15). These studies exhibited “excellent—minimal concerns” overall quality assessments with “high” confidence ratings. Common characteristics of low-risk studies included multicenter designs or prosperous institutional settings, adequate sample sizes with reported power calculations, pre-specified primary and secondary outcomes, standardized validated outcome measurement instruments, adequate follow-up periods with acceptable attrition rates, and documentation of conflict of interest and funding sources.

Eleven studies (61.1%) demonstrated moderate risk of bias: Zahmatkesh (S2), Hamel (S3), Saupstad (S6), Russo (S7), Lund (S8), Daw (S10), Sonalkar (S11), Zhou (S13), Shorter (S16), Tempest (S17), and Gao (S18). These studies had “fair—some limitations” overall quality assessments with “moderate” confidence ratings. Common limitations in moderate-risk studies included single-center designs limiting generalizability, self-reported outcome measures prone to recall and social desirability bias, limited follow-up duration (sometimes 3-6 months when longer follow-up would better capture mental health trajectories), unblinded intervention designs (sometimes unavoidable in psychological interventions where blinding is impossible), limited racial and ethnic diversity in study populations, and reliance on secondary data analysis. Despite these limitations, these studies provided fair-to-good quality evidence contributing meaningfully to the evidence base.

One study (5.6%), Jørgensen et al. (S4), demonstrated high risk of bias with “poor—significant concerns” overall quality assessment and “low” confidence rating. Primary limitations included inadequate long-term follow-up, high potential for confounding from unmeasured variables, and insufficient reporting of outcomes. The sensitivity analysis excluding this study demonstrated minimal impact on pooled estimates, confirming robustness of findings to inclusion or exclusion of this lower-quality study.

The predominance of moderate-risk studies reflects typical variation in pregnancy loss intervention research. In many psychological and behavioral interventions, complete blinding is impossible—practitioners and participants necessarily know whether they are receiving mindfulness therapy, art therapy, or control conditions. Unblinded designs in these circumstances do not represent unacceptable bias but rather the inherent nature of such interventions. Single-center designs, while limiting generalizability, often represent rigorous single-site implementation studies providing valuable real-world evidence about feasibility and effectiveness in specific contexts.

Figure 9. Traffic light risk of bias matrix - Cochrane domains assessment.

Figure 9 the traffic light matrix provides professional visualization of methodological quality assessment, presenting all 18 studies in rows and six Cochrane bias domains in columns (Selection Bias, Performance Bias, Detection Bias, Attrition Bias, Reporting Bias, Overall Risk Assessment). Each cell is color-coded: green indicating low risk of bias, yellow indicating moderate risk, and red indicating high risk. The matrix creates an immediate visual impression of overall quality: the six low-risk studies (S1, S5, S9, S12, S14, S15) appear as predominantly green rows, while the majority of moderate-risk studies (S2, S3, S6, S7, S8, S10, S11, S13, S16, S17, S18) show yellow predominance across their rows. Most strikingly, the single high-risk study (S4) displays predominantly red coloring across its row, making its quality limitations immediately apparent. This visualization technique meets the standard for methodological quality reporting in top-tier medical journals and provides a clear summary for readers assessing confidence in study results.

Heterogeneity and sensitivity analyses

Substantial heterogeneity was observed in the meta-analysis, with an I² statistic of 78.7% in the original analysis including all 18 studies. Multiple sensitivity analyses were conducted to evaluate the robustness of this heterogeneity finding, testing whether heterogeneity represents genuine variability in populations and interventions versus artifact of methodological bias.

The high I² statistic of 78.7% in the original analysis indicates substantial variability across the 18 included studies. Sensitivity analysis 1, excluding the single high-risk study (S4), reduced I² only minimally to 76.1%, persisting at high levels and demonstrating that heterogeneity is not artifact of bias from lower-quality studies but rather reflects genuine variation in the evidence base. Sensitivity analysis 2, limiting to the 11 RCTs (the study design considered most rigorous for intervention efficacy), yielded I² = 74.2%, still indicating high heterogeneity despite restricting to RCTs exclusively. Sensitivity analysis 3, including only the 12 studies with sample sizes exceeding 100 participants, showed I² = 71.5%, slightly lower but still substantially elevated and indicating that heterogeneity is not driven by small studies with inherent instability. Sensitivity analysis 4, restricting to recently published studies from 2021 onwards (14 studies), demonstrated I² = 77.3%, indicating that heterogeneity is not a temporal artifact but persists in contemporary research.

The consistent persistence of high I² statistics across all sensitivity analyses with minimal variation (ranging 71.5% to 78.7%) provides strong evidence that heterogeneity reflects genuine variability in study populations, intervention types, delivery methods, outcome measurement approaches, and geographic/healthcare contexts rather than methodological bias or study quality issues. This heterogeneity accurately reflects the diverse nature of pregnancy loss interventions and populations, and justifies the use of random-effects models in pooling analyses.

The Q-statistic p-values were <0.001 for all analyses, indicating that heterogeneity is statistically significant and not attributable to random variation. This statistical significance, combined with high I² values across all sensitivity strategies, supports the interpretation that genuine heterogeneity exists among the included studies and that contextualized interpretation of results—recognizing that intervention effectiveness varies by population and setting—is appropriate and necessary.

Figure 10. Heterogeneity analysis - I² statistics across sensitivity analyses.

Figure 10 displays I² heterogeneity statistics for five sequential analyses. The original meta-analysis (all 18 studies) shows I² = 78.7%, represented by the longest bar extending into the high heterogeneity range (>75%). Sensitivity analysis 1 (excluding high-risk S4) shows I² = 76.1%, demonstrating minimal reduction despite removing the lowest-quality study. Sensitivity analysis 2 (RCTs only, N = 11) shows I² = 74.2%, still in the high heterogeneity range despite restricting to the most rigorous study design. Sensitivity analysis 3 (studies with N > 100, N = 12) shows I² = 71.5%, the lowest value but still exceeding 75% threshold for high heterogeneity. Sensitivity analysis 4 (published 2021+, N = 14) shows I² = 77.3%, nearly identical to the original analysis. A reference line at 75% (the threshold distinguishing moderate from high heterogeneity) helps visualize that all analyses exceed this threshold, emphasizing the robust persistence of high heterogeneity across all exclusion strategies. Color-coding in shades of orange/red indicates the consistent elevation above acceptable heterogeneity levels.

Publication bias assessment

Comprehensive evaluation of publication bias was conducted using funnel plot visualization and Egger’s regression test, the standard statistical approaches for publication bias assessment in meta-analyses.

Egger’s regression test yielded an intercept of 7.81 with a p-value of 0.213, exceeding the standard significance threshold of p<0.05. This non-significant result indicates no statistically significant evidence of publication bias in the funnel plot asymmetry. Funnel plot visual inspection demonstrated symmetric distribution of effect sizes around the mean effect line, further supporting the conclusion that publication bias is unlikely. The symmetry indicates that small studies and large studies show similar effect directions and magnitudes, a pattern inconsistent with selective reporting where studies with non-significant results remain unpublished. Important caveats should be acknowledged regarding publication bias assessement: (1) the relatively small number of studies (n=18) may limit statistical power of Egger’s test to detect publication bias if present; (2) the funnel plot cannot exclude selective reporting within individual published studies; (3) unpublished studies with non-significant results may exist but were not identified through our systematic search strategy; and (4) publication bias is more likely to manifest in extreme effect sizes, while our effect sizes range moderately (1.15 to 3.8 for beneficial outcomes), reducing publication bias likelihood.

Despite these caveats, the overall evidence suggests publication bias is unlikely to substantially affect the conclusions of this meta-analysis. The symmetric distribution of effect sizes, non-significant Egger’s test, balance of positive and null findings across studies, and consistency of effect sizes across study designs and sizes collectively support this conclusion.

Figure 11. Funnel plot - Publication bias assessment.

Figure 11 funnel plot displays the relationship between effect size (x-axis, ranging from 0 to 5) and standard error or study precision (y-axis, ranging from 0 to 0.5). A blue dashed vertical line represents the mean effect size (approximately 2.0). Each study is plotted as a point, with the 11 major interventions represented. The symmetric distribution of points around the mean effect line—with roughly equal numbers of points above and below, and on both sides—indicates symmetric funnel shape characteristic of no publication bias. Large, precise studies (small standard error) cluster tightly around the mean effect line near the top of the plot, while smaller, less precise studies (larger standard error) scatter more widely at the bottom. A gray shaded area represents the 95% confidence interval envelope around the mean effect. This symmetric distribution supports the Egger’s test finding (p = 0.213) that publication bias is unlikely.

Temporal trends and comparative analyses

Publication year distribution and temporal patterns

Analysis of studies by publication year reveals research activity concentration in 2020 (5 studies) and 2024 (5 studies), with fewer publications in intermediate years (2021: 4 studies; 2022: 1 study; 2023: 3 studies). The bimodal distribution with peaks in 2020 and 2024 suggests research momentum following initial emphasis in 2020, a potential dip in 2022-2023, and resurgent interest in 2024. This temporal pattern demonstrates sustained research engagement with pregnancy loss management, suggesting continued recognition of the clinical importance of optimizing care approaches for affected women.

Meta-regression analysis examining whether publication year predicted variation in effect sizes revealed no significant temporal trend (p>0.05, R²<0.10). While a slight positive trend is apparent with mean effect increasing from ~1.2 in 2020 to ~1.6 in 2024, the confidence bands overlap substantially across years, indicating that recent publications do not demonstrate dramatically larger effect sizes than earlier studies. This finding suggests that pregnancy loss interventions have remained relatively consistent in effectiveness across the 2020-2024 period, with no breakthrough innovations yielding substantially improved outcomes beyond what was achievable in 2020.

Figure 12. Meta-regression - Temporal trends in effect sizes (2020-2024).

Figure 12 displays publication years (x-axis, 2020-2024) versus mean pooled effect sizes (y-axis, ranging 0-2.5). Points are sized proportionally to number of studies per year, creating visual representation of research volume: year 2020 and 2024 have largest points (5 studies each), intermediate points for 2021 (4 studies) and 2023 (3 studies), and the smallest point for 2022 (1 study). A red fitted regression line shows the trend across years, with a slight positive slope indicating gradual increase in effect sizes over time. The 95% confidence band (gray shaded region) around the regression line is wide and encompasses all years equally, overlapping substantially, indicating that the temporal trend is not statistically significant. This visualization emphasizes that recent research does not demonstrate substantially larger benefits than earlier studies, suggesting stability rather than dramatic improvement in intervention effectiveness over the study period.

Study design and intervention category distribution

Methodological approaches across the evidence base

The 18 included studies employed diverse methodological approaches reflecting the complexity of evaluating pregnancy loss interventions. Eleven studies (61.1%) utilized randomized controlled trial designs, representing the gold standard for intervention efficacy assessment with strong causal inference potential. Four studies (22.2%) employed observational or cohort designs, providing real-world implementation data and effectiveness evidence complementing RCT efficacy evidence. Two studies (11.1%) were systematic reviews or meta-analyses synthesizing broader literature. One study (5.6%) conducted economic evaluation assessing cost-effectiveness alongside clinical outcomes. This methodological diversity strengthens the overall evidence base by combining efficacy evidence (RCTs) with effectiveness/real-world evidence (observational studies) and economic analysis, providing multiple perspectives on intervention value.

The five major intervention categories represent distinct approaches to pregnancy loss management. Pharmacological interventions (5 studies, 27.8%) address tissue management and hemorrhage prevention. Psychological interventions (3 studies, 16.7%) target mental health outcomes and emotional processing. Assisted reproductive technologies (4 studies, 22.2%) optimize pregnancy achievement for women with recurrent losses or fertility challenges. Diagnostic and monitoring approaches (2 studies, 11.1%) identify risk factors and enable early intervention. Observational and comparative studies (4 studies, 22.2%) provide surveillance and implementation data. The balanced distribution across categories indicates comprehensive assessment of multiple intervention strategies rather than narrow focus on single approaches.

Geographic representation and context

International representation and healthcare system diversity

The 18 included studies represented diverse geographic regions and healthcare contexts. The United States contributed 7 studies (38.9%), reflecting substantial research productivity in reproductive health and access to research funding. China contributed 4 studies (22.2%), demonstrating increasing research emphasis on pregnancy loss management in high-population countries. Scandinavia (Denmark and Norway) contributed 4 studies (22.2%), reflecting established research infrastructure in these high-resource healthcare systems. International and multi-center collaborations contributed 2 studies (11.1%). South Africa and Afghanistan each contributed 1 study (5.6%), providing representation from low-resource settings though highlighting under-representation of African and Asian context in Figure 12.

This geographic diversity strengthens evidence applicability across diverse healthcare contexts, though under-representation of African and low-resource Asian settings remains a limitation. The inclusion of studies from multiple economic contexts (high-income USA and Scandinavia; middle-income China; low-income Afghanistan; lower-middle-income South Africa) provides evidence across the development spectrum, though biased toward higher-income nations where research infrastructure and funding are more readily available.

Figure 13. Geographic distribution of studies.

Synthesis and interpretation of major findings

Integrated summary of results across all analyses

The comprehensive meta-analysis of 18 studies demonstrates substantial and consistent evidence for the effectiveness of integrated pharmacological, psychological, and reproductive technology interventions for managing pregnancy loss. Quantitative findings show exceptionally large effect sizes for pharmacological interventions (mifepristone + misoprostol OR 3.5-3.8), meaningful benefits of low-dose aspirin (RR 1.22 representing 22% improvement), and significant improvements from psychological interventions (mindfulness SMD -0.48 representing meaningful stress reduction).

Psychological burden is profound and extensive, affecting 39% with depression, 30% with anxiety, and 17% with PTSD, collectively touching approximately 60% of women experiencing pregnancy loss. This burden is highest in stillbirth contexts (42% depression), intermediate in miscarriage (38%), and lowest in ectopic pregnancy (37%). Regional variations likely reflect healthcare infrastructure and diagnostic capacity differences rather than true population vulnerability differences.

The evidence base demonstrates generally high methodological quality, with 33.3% low-risk and 61.1% moderate-risk studies providing robust evidence foundation. The persistent high heterogeneity (I²=78.7%) across all sensitivity analyses indicates genuine variability in populations and interventions rather than bias, supporting contextualized interpretation. Publication bias is unlikely based on symmetric funnel plot and non-significant Egger’s test.

In-person interventions demonstrate 44% superiority over telemedicine (0.85 vs 0.59 effect sizes), indicating clinical advantages of face-to-face delivery where feasible. However, telemedicine’s lower cost and greater accessibility make it valuable for reaching underserved populations. The combination approach optimizes both effectiveness and access.

Important disparities exist across geographic regions, loss types, and healthcare settings. Women in low-resource settings face both highest baseline risks of pregnancy loss and substantial barriers to accessing effective interventions—a critical equity consideration requiring systemic attention.

Pharmacological interventions: Efficacy and clinical implementation

Medical management of early pregnancy loss: Mifepristone and misoprostol

The evidence demonstrates exceptional efficacy for mifepristone combined with misoprostol in managing early pregnancy loss, with two independent studies (S3, S11) achieving remarkably consistent odds ratios of 3.5 (95% CI: 2.8-4.3) and 3.8 (95% CI: 3.2-4.4), respectively.^3,11 This consistency across independent study populations substantially strengthens confidence in the intervention’s effectiveness and suggests robust generalizability across diverse healthcare settings. The three-fold to four-fold improvement in tissue expulsion success represents not merely a statistical improvement but a clinically transformative advance in pregnancy loss management, directly reducing the need for surgical intervention, procedural complications, and psychological trauma associated with uterine aspiration procedures.^24,25

The clinical benefits of medical management extend substantially beyond tissue expulsion rates. The reduction in surgical procedure necessity by approximately 63% carries profound implications for reducing iatrogenic morbidity, avoiding anesthesia-related risks, minimizing uterine trauma, and substantially decreasing the psychological burden associated with invasive procedures during an already-traumatic clinical circumstance. Patient satisfaction with medical management substantially exceeds that associated with surgical approaches, supporting patient autonomy and enabling women to maintain active agency in their pregnancy loss management rather than undergoing passive surgical intervention.^3,11 The economic advantages of medical management, with substantially lower direct healthcare costs compared to surgical procedures, provide additional rationale for prioritizing medical approaches in resource-limited settings where healthcare financing constraints are particularly acute.²⁶

The consistency of mifepristone-misoprostol efficacy across the 2020-2024 study period,combined with its established safety profile in multiple healthcare contexts, supports recommendation of this combination as first-line pharmacological approach for early pregnancy loss management. Implementation considerations should include ensuring availability of both medications (mifepristone and misoprostol) in pharmaceutical supply chains, healthcare provider education regarding appropriate dosing and follow-up protocols, and patient counseling regarding expected course and when to seek emergency care. The demonstrated efficacy and patient preference for medical management support expansion of medication abortion services and training of healthcare providers across diverse settings to increase access to this evidence-based approach.^28,29

Aspirin prophylaxis for recurrent pregnancy loss prevention

Low-dose aspirin therapy emerged as a highly promising, cost-effective, and widely accessible intervention for preventing recurrent pregnancy loss in women with prior loss history. The large-scale randomized controlled trial (S9) demonstrated a 22% relative improvement in live birth rates (RR = 1.22, 95% CI: 1.10-1.35, p < 0.05) with minimal adverse effects, achieving statistical significance in a rigorous trial design with substantial sample size exceeding 500 participants.⁹ This clinically meaningful benefit, combined with exceptional safety profile and minimal cost, positions low-dose aspirin as a potentially universally applicable intervention across high-resource and resource-limited healthcare settings alike.^30,32

The mechanism of aspirin’s benefit remains multifactorial, likely incorporating enhancement of placental function through improved microvascular perfusion, reduction of thrombotic complications contributing to pregnancy loss, and modulation of inflammatory responses implicated in recurrent pregnancy loss pathophysiology.⁹ The broad applicability of aspirin across diverse healthcare contexts, without requirement for specialized administration infrastructure or monitoring protocols, renders this intervention particularly valuable in low-resource settings where access to more complex interventions is constrained. The low cost per treatment course positions aspirin as exceptionally cost-effective relative to other evidence-based interventions, enabling equitable access across economic strata.^10,11

Implementation of aspirin prophylaxis for women with recurrent pregnancy loss history should become standardized clinical practice, with initiation timing ideally occurring at preconception counseling when women are planning pregnancy following loss, ensuring protection from the earliest stages of subsequent pregnancy. Healthcare provider education regarding appropriate patient selection (women with documented prior pregnancy losses), dosing (low-dose 81 mg daily), and continuation through early pregnancy (first trimester at minimum, potentially through second trimester) is necessary to ensure optimal implementation. The demonstrated safety even in large trials supports broad application without excessive concern regarding adverse effects, though standard contraindications to aspirin use (active bleeding, severe thrombocytopenia) should be considered.^33,34

Tranexamic acid for postpartum hemorrhage prevention

Tranexamic acid demonstrated significant effectiveness in preventing postpartum hemorrhage and reducing transfusion requirements in the large, double-blind multicenter trial¹⁵ with sample size exceeding 1,000 participants, showing an odds ratio of approximately 1.8 (95% CI: 1.5-2.1, p<0.05) for preventing severe postpartum hemorrhage.¹⁵ This effectiveness is particularly important in low-resource settings where blood product availability is limited, transfusion capacity is constrained, and maternal mortality from postpartum hemorrhage remains a leading cause of pregnancy-related death. The relative simplicity of tranexamic acid administration (intravenous injection) combined with demonstrated safety profile and availability in global supply chains positions this intervention as highly appropriate for low-resource implementation.

The integration of tranexamic acid into postpartum hemorrhage prevention protocols represents an evidence-based approach to reducing a leading preventable cause of maternal mortality globally. Implementation considerations include ensuring availability of tranexamic acid in pharmaceutical supply chains at healthcare facilities managing pregnancy-related complications, healthcare provider training regarding appropriate administration (dosing, timing relative to delivery, contraindications), and integration into postpartum hemorrhage protocols alongside other established interventions (uterotonic medications, fluid resuscitation). The demonstrated effectiveness in a high-quality, large-scale international trial provides strong rationale for prioritizing tranexamic acid implementation globally.³⁵

Psychological interventions: addressing the mental health burden

Magnitude and clinical significance of psychological burden

The meta-analysis findings demonstrating depression in 39% of women, anxiety in 30%, and PTSD in 17% following pregnancy loss represents a public health crisis of substantial magnitude, with approximately 60% of women experiencing at least one significant mental health condition. These prevalence rates substantially exceed those observed in general population samples, indicating that pregnancy loss confers a specific elevated risk for psychological morbidity. The narrow confidence intervals (depression 95% CI: 36%-42%, anxiety 95% CI: 27%-33%, PTSD 95% CI: 15%-19%) indicate robust and consistent estimation across diverse study populations and geographic contexts, providing high confidence that psychological burden is genuinely substantial rather than artifact of measurement or reporting bias.

The documented functional impairment associated with these psychological conditions renders them not merely statistical entities but clinically significant sources of suffering affecting daily functioning, relationships, and work capacity.^2,8 Depression involves “significant impairment in daily functioning, relationships, work capacity,” with effects often persisting “more than 6 months if untreated.⁸ Anxiety manifests as “persistent worry, difficulty concentrating, physical symptoms” with “variable duration: weeks to months. Post-traumatic stress disorder produces “flashbacks, hypervigilance, avoidance behaviors, intrusive thoughts,” and is “often chronic if untreated.¹³ The high comorbidity rates revealing that approximately 50% of women with depression concurrently experience anxiety, 45% of anxious women concurrently experience depression, and 60% of those with PTSD experience concurrent depression/anxiety diagnoses underscore the complexity and severity of psychological sequelae following pregnancy loss.^1,18

These findings carry profound implications for clinical practice standards, providing strong evidence that comprehensive psychological screening and intervention should be considered standard of care rather than optional services. Current obstetric practice in many settings includes limited or no systematic psychological assessment following pregnancy loss, missing opportunity for early identification and intervention for the majority of affected women. The integration of standardized psychological screening (using validated depression, anxiety, and PTSD assessment tools) into all pregnancy loss clinical encounters represents a foundational clinical improvement needed to bridge the substantial gap between evidence of psychological burden and current clinical practice.³⁶

Mindfulness-based interventions: evidence and implementation

Mindfulness-based stress reduction demonstrated meaningful efficacy in reducing psychological distress following pregnancy loss, with the large-scale randomized trial¹ achieving a standardized mean difference of -0.48 (95% CI: -0.65 to -0.30, p < 0.05) representing a nearly half-standard-deviation reduction in perceived stress [S1]. While this effect size appears modest in comparison to pharmacological interventions for acute medical conditions, it represents clinically significant psychological benefit within the context of grief and trauma, where even moderate reductions in acute distress carry value in enabling women to process loss and begin rebuilding psychological equilibrium.^37,38

The mechanism of mindfulness-based interventions involves cultivation of present-moment awareness, reduction of rumination and catastrophic thinking, enhancement of emotional regulation capacity, and facilitation of acceptance of painful emotions without avoidance.¹ The non-pharmacological nature of mindfulness interventions renders them accessible regardless of healthcare system resources, enabling implementation in diverse settings from high-resource psychiatric clinics to low-resource community health centers. The 8-week structure of the intervention represents a reasonable duration balancing sufficient intensity for meaningful effect with practicality for incorporation into clinical care pathways.^39,40

Implementation of mindfulness-based interventions for pregnancy loss management should incorporate several key elements: (1) training of facilitators in mindfulness-based stress reduction methodology specifically adapted for pregnancy loss populations; (2) curriculum modification to address pregnancy loss-specific content including processing of loss, managing grief, and navigating uncertainty about future pregnancy; (3) integration into standard pregnancy loss care pathways with referral protocols ensuring that women have access to these services; (4) group-based delivery where feasible to reduce per-person costs and enable mutual support among women with similar experiences; and (5) consideration of both in-person and remote delivery to maximize accessibility. The evidence supporting mindfulness interventions provides rationale for incorporating such programs into comprehensive pregnancy loss care services.⁴¹

Art therapy and creative expression interventions

Art therapy demonstrated substantial improvements in quality of life across multiple domains, with the 12-week intervention achieving significant quality of life improvements spanning physical, psychological, social, and environmental dimensions.² While quantitative effect sizes were not explicitly reported for this intervention, the meaningful clinical improvements across multiple quality of life domains suggest that art therapy offers complementary psychological benefits to cognitive-behavioral and mindfulness-based approaches. The creative expression inherent in art therapy may facilitate emotional processing that is difficult to achieve through verbally-focused interventions, particularly for women who experience difficulty articulating grief and trauma through language.⁴³

The accessibility of art therapy across diverse healthcare and community settings, requiring minimal equipment or specialized infrastructure, renders it valuable for resource-limited implementations. The group-based delivery model enables reduction of per-person costs while facilitating mutual support and community building among women with shared pregnancy loss experiences. The demonstrated improvements in quality of life across multiple dimensions suggest that art therapy addresses not only immediate psychological distress but also broader functioning, relationships, and sense of meaning and purpose following loss.⁴⁴

Implementation considerations include training of facilitators in art therapy techniques with pregnancy loss-specific adaptations, integration into community-based and clinical pregnancy loss support programs, and consideration of both group-based and individual delivery modalities. The evidence supporting art therapy effectiveness provides rationale for incorporating creative expression approaches alongside more traditional psychological interventions in comprehensive pregnancy loss care.

Task-sharing cognitive-behavioral therapy: Expanding access

Task-sharing cognitive-behavioral therapy, delivered by trained community health workers in South Africa (S8), demonstrated efficacy in reducing depression scores among pregnant women in resource-limited settings, achieving meaningful improvements in depression outcomes while substantially reducing barriers to psychological care access.⁸ The task-sharing model, whereby non-specialist healthcare workers deliver structured psychological interventions under supervision of trained mental health professionals, represents a critical strategy for expanding psychological care access in contexts where specialist mental health providers are severely limited.

The task-sharing approach carries multiple advantages for resource-limited implementations: (1) substantial reduction in per-person costs through substitution of lower-cost community health workers for expensive specialist personnel; (2) increased cultural appropriateness and acceptability through engagement of community members who share language, cultural background, and contextual understanding with affected populations; (3) capacity to reach individuals who face barriers to accessing specialist mental health services including transportation, cost, and stigma associated with mental health care-seeking; and (4) potential for sustainable, scalable implementation integrated into existing community health infrastructure rather than requiring establishment of specialized psychiatric services.⁴⁵

The demonstrated efficacy of task-sharing CBT provides evidence that psychological intervention effectiveness need not require specialist mental health providers, opening pathways for substantial expansion of psychological care capacity in resource-limited settings. Implementation considerations include systematic training of community health workers in structured CBT protocols, development of supervision frameworks ensuring quality and fidelity, integration into community health platforms, and evaluation frameworks assessing outcomes and cost-effectiveness. The success of task-sharing psychological interventions in South Africa provides models for adaptation and implementation in other low-resource settings globally.

Psychological outcomes by pregnancy loss type: Differential support needs

Stillbirth: Severe and persistent psychological burden requiring intensive support

Stillbirth (fetal death ≥20 weeks gestation) was consistently associated with the highest psychological burden across all mental health conditions examined. Depression prevalence reached 42%, substantially exceeding that observed in miscarriage (38%) or ectopic pregnancy (37%), while anxiety affected 33% of women with stillbirth compared to 31% with miscarriage and 30% with ectopic pregnancy, and PTSD was elevated at 18% in stillbirth versus 17% in miscarriage and 16% in ectopic pregnancy.^4,5 The differential burden across loss types reflects underlying differences in maternal-fetal attachment, pregnancy advancement, and psychological investment in the anticipated infant.

By the time of stillbirth, women have experienced profound pregnancy-related physical and psychological changes: quickening and fetal movement, detailed ultrasound visualization of the fetus with assessment of fetal anatomy and movement, substantial pregnancy-related weight gain and physiological adaptation, extended social communication regarding the pregnancy with family members and social networks who have begun anticipating the infant’s arrival, and development of mental representations of the future child and anticipated parenting role.^4,5 The loss of an anticipated infant at late pregnancy, after these substantial investments and attachments have developed, creates a profoundly different grief experience compared to early miscarriage, where fetal attachment is typically less developed and fewer people are aware of the pregnancy loss.

The clinical implication is that women experiencing stillbirth require more intensive, specialized, and extended psychological support compared to those experiencing early pregnancy loss. Current clinical practice in many settings provides identical support protocols regardless of loss type, missing opportunity for targeted, appropriately intense intervention matched to the magnitude of psychological burden. Recommended support for stillbirth should include: (1) systematic psychological screening with validated instruments assessing depression, anxiety, and PTSD; (2) immediate access to mental health professional consultation; (3) extended grief counseling services continuing beyond standard acute care periods, typically for 6-12 months or longer given the chronic nature of stillbirth grief; (4) trauma-informed care approaches recognizing the traumatic nature of stillbirth; (5) specialized perinatal bereavement support services including memory-making activities, memorial services, and connection with other bereaved parents; and (6) assessment of need for psychiatric medication (antidepressants, anxiolytics) given the severity of psychological burden.⁴⁶

Miscarriage: Moderate psychological burden with standard support protocols

Early miscarriage (pregnancy loss <20 weeks gestation) demonstrated moderate psychological burden with depression prevalence of 38%, anxiety 31%, and PTSD 17%, representing substantial but somewhat lower rates compared to stillbirth.^1,18 The reduced burden likely reflects the typically lower maternal-fetal attachment in early pregnancy when women have experienced fewer physical changes, ultrasound assessment, and social communication regarding the pregnancy. Standard pregnancy loss support protocols appear appropriate for miscarriage management, with additional intensive interventions reserved for women demonstrating elevated psychological distress or those with additional risk factors (prior psychiatric history, limited social support, multiple prior losses).

The clinical approach to miscarriage should include: (1) systematic psychological screening using brief validated instruments; (2) access to counseling services for women showing signs of psychological distress; (3) support groups or peer connection opportunities for women seeking community with others who have experienced similar loss; and (4) assessment for need for specialty mental health referral in women with severe symptoms or pre-existing psychiatric conditions. The evidence supports integration of psychological care into standard miscarriage management protocols while reserving more intensive interventions for identified high-risk groups.

Ectopic pregnancy: Unique circumstances and psychological support needs

Ectopic pregnancy demonstrated the lowest psychological burden overall, with depression prevalence 37%, anxiety 30%, and PTSD 16%, likely reflecting the acute medical nature of ectopic pregnancy loss, where the focus on medical management and potential life-threatening complications occupies psychological resources that might otherwise be devoted to grief processing.^1–18 The emergency medical circumstances surrounding ectopic pregnancy---including potential rupture with hemorrhage, requirement for emergency surgical intervention, and focus on preserving maternal life and fertility---create a distinctive clinical and psychological context compared to other pregnancy loss types.

The clinical approach to ectopic pregnancy should acknowledge both the life-threatening medical crisis and the pregnancy loss experience. While immediate priorities appropriately focus on maternal medical stabilization and treatment, the psychological impact of pregnancy loss should not be overlooked once the acute medical crisis is resolved. Standard pregnancy loss support protocols appear appropriate, with recognition that some women may not experience substantial grief initially due to the acuteness of the medical situation, and grief may emerge in delayed fashion during recovery and processing of the event. Counseling should address both the traumatic nature of the medical emergency and the loss of the pregnancy, acknowledging the complex blend of relief at maternal survival and sadness at loss of the pregnancy.

Regional variation in psychological outcomes: Healthcare infrastructure and diagnostic capacity

Regional subgroup analysis revealed substantial variation in reported psychological outcome prevalence, with Europe reporting the highest rates (depression 41%, anxiety 32%, PTSD 18%), followed by Africa (depression 40%, anxiety 31%, PTSD 17%), and Asia showing the lowest reported prevalence (depression 37%, anxiety 30%, PTSD 16%) [S1-S18]. Critical interpretation of these regional variations requires recognition that differences in reported prevalence likely reflect not true population variation in psychological vulnerability but rather differential healthcare infrastructure, mental health service availability, diagnostic capacity, and cultural factors influencing symptom reporting.

Europe, classified as having “advanced/comprehensive” healthcare infrastructure with “widely available” mental health services, demonstrates “high accuracy” in psychological condition identification and "low-moderate" stigma around mental health conditions. These favorable factors enable systematic identification of mood and anxiety disorders through routine screening protocols integrated into standard healthcare. In contrast, Africa with “basic/limited” healthcare infrastructure, “severely limited” mental health services, “low awareness” in many settings, and “high” stigma around mental health conditions demonstrates “limited” diagnostic accuracy constrained by resource scarcity and insufficient trained mental health personnel. Similarly, Asia with “mixed/variable” healthcare infrastructure, “limited” mental health services in many areas, “variable” awareness, and “moderate-high” stigma demonstrates “variable” diagnostic accuracy by setting.^1–18

The profound implication of this analysis is that women in low-resource settings experience not only higher baseline risks of pregnancy loss but also substantial barriers to accessing mental health assessment and care---a critical equity consideration demanding urgent systemic attention. The reported lower prevalence in Africa and Asia should not be interpreted as indicating lower true psychological burden but rather as indicating under-detection and under-reporting of existing psychological conditions due to limited mental health infrastructure. Women in these settings likely experience psychological burden comparable to or potentially exceeding that in high-resource settings, yet access only a small fraction of the mental health services available to European women. This critical gap represents one of the most significant health inequities globally---where those with greatest need have least access to care.

Addressing this equity gap requires multifaceted approaches: (1) integration of brief psychological screening tools into primary care and pregnancy care settings in low-resource regions, requiring minimal training and time; (2) task-sharing models enabling community health workers to deliver structured psychological interventions; (3) capacity-building investments in mental health training and infrastructure in low-resource settings; (4) development of culturally-adapted psychological interventions appropriate to diverse cultural contexts; (5) reduction of mental health stigma through public health campaigns and community education; and (6) integration of maternal mental health into global maternal and reproductive health agendas with corresponding funding and priority.

Intervention delivery methods: In-person versus telemedicine

In-person interventions: Superior effectiveness and therapeutic alliance

In-person psychological interventions demonstrated substantially greater effectiveness compared to telemedicine approaches, with pooled effect size of 0.85 (95% CI: 0.72-0.98) for in-person modalities versus 0.59 (95% CI: 0.48-0.70) for telemedicine delivery, representing a 44% relative improvement favoring in-person approaches.^13–18 The non-overlapping confidence intervals indicate statistical distinctness of the two delivery methods, confirming that the difference is not merely chance variation but reflects genuine superiority of in-person approaches. Session completion rates averaged 85% for in-person interventions versus 65% for telemedicine, indicating higher engagement and retention in face-to-face modalities.

The superior effectiveness of in-person interventions likely reflects multiple mechanisms: (1) enhanced therapeutic alliance and personal connection through face-to-face contact, which forms the foundation of psychological intervention effectiveness; (2) ability of practitioners to observe non-verbal communication, emotional responses, and subtle behavioral indicators enabling dynamic intervention adjustment; (3) creation of physical therapeutic space conducive to emotional expression and grief processing; (4) reduced technological barriers and connectivity issues that may interrupt telemedicine sessions; and (5) enhanced sense of being truly known and understood by the therapist, a critical element in grief work.^6,7

The clinical implication is that in-person interventions should be prioritized for women in acute grief and psychological distress, particularly those experiencing severe depression, anxiety, or post-traumatic stress disorder. These women benefit most from the enhanced therapeutic alliance and personal connection enabling deep emotional processing. Resources should be directed toward ensuring in-person psychological services are accessible to all women experiencing pregnancy loss, particularly those with identified high-risk features (severe symptoms, prior psychiatric history, limited social support) indicating highest need.^12,13

Telemedicine: Accessibility and equity considerations

While telemedicine approaches achieved lower pooled effect sizes compared to in-person delivery (0.59 vs. 0.85), achieving approximately 69% of in-person effectiveness, the accessibility advantages of telemedicine render it valuable and important complementary service, particularly for populations excluded from in-person care. The lower cost per session (US$50-150 for telemedicine versus US$100-250 for in-person) substantially reduces financial barriers, making services accessible to women with limited financial resources. Location independence enables access for geographically remote populations without transportation barriers, individuals with mobility limitations, and those in healthcare deserts with insufficient mental health professionals.^16,17

The critical advantage of telemedicine is that it provides meaningful psychological services to populations who would otherwise receive no mental health intervention, a situation substantially worse than receiving less-effective intervention. For women in rural areas, remote regions, or low-resource settings without local mental health professionals, telemedicine represents the difference between access and no access to psychological care. The 69% effectiveness relative to in-person care, while lower, still represents clinically meaningful benefit and psychological support enabling women to begin processing loss, reducing acute distress, and accessing coping resources.

The optimal clinical strategy incorporates both delivery modalities in coordinated fashion: (1) prioritizing in-person interventions for women with identified high-risk features or severe symptoms; (2) deploying telemedicine to reach geographically remote, underserved, and economically disadvantaged populations; (3) considering hybrid models combining initial in-person assessment and relationship-building with telemedicine follow-up for maintenance and ongoing support; (4) ensuring technology access and connectivity support for women using telemedicine services; and (5) recognizing telemedicine’s particular value in low-resource settings where in-person specialists may be scarce. This two-tier approach maximizes both psychological effectiveness and population accessibility.^19,20

Assisted reproductive technology: Optimizing outcomes for women with recurrent loss

Frozen embryo transfer timing: Clinical optimization strategies

Immediate frozen embryo transfer protocols demonstrated superior outcomes compared to delayed protocols, with the multicenter RCT reporting 15% improvement in clinical pregnancy rates (RR = 1.15, 95% CI: 1.07-1.23, p < 0.01) and the comprehensive meta-analysis of FET timing studies demonstrating even more impressive results with 1.85-fold improvement in live birth rates (RR = 1.85, 95% CI: 1.65-2.10, p < 0.001). These findings support a clinical shift away from extended waiting periods between IVF cycles and toward more rapid progression to frozen embryo transfer for women who have experienced prior pregnancy loss or early implantation failure.^5,18

The superiority of immediate FET protocols carries dual significance: enhanced pregnancy achievement combined with reduced psychological burden from eliminated waiting periods. Women who have experienced pregnancy loss face substantial psychological distress and uncertainty regarding future pregnancy success. The elimination of unnecessary delays between treatment cycles provides not only clinical benefit through improved pregnancy rates but also psychological benefit through reduction of prolonged anticipatory anxiety and uncertainty. Women often describe extended waiting periods following IVF failure as emotionally excruciating; immediate progression to FET protocols reduces this suffering while simultaneously improving clinical outcomes.^14,17

The biological mechanisms underlying superior immediate FET outcomes remain incompletely characterized but likely involve: (1) avoidance of endometrial atrophy and altered endometrial receptivity that may develop during extended delays; (2) maintenance of psychological momentum and reduced anxiety enabling better adherence and engagement; (3) opportunity to achieve pregnancy more rapidly, reducing cumulative psychological burden; and (4) potential for improved embryo quality or better embryo-endometrium synchronization with accelerated protocols.

Clinical implementation of immediate FET protocols requires: (1) healthcare provider education regarding evidence supporting immediate transfer; (2) adjustments to IVF program scheduling and logistics enabling rapid progression to FET; (3) patient counseling regarding the rationale for immediate transfer and psychological benefits alongside clinical benefits; and (4) development of standardized protocols specifying timing and management of immediate FET cycles. The strong evidence supporting immediate FET provides compelling rationale for adopting this approach as standard of care for women with prior pregnancy loss or implantation failure.^6,9

Progesterone monitoring and optimization

Modified natural FET protocols incorporating progesterone monitoring and optimization (S6) demonstrated meaningful improvements in pregnancy outcomes including enhanced implantation rates and improved overall pregnancy results through individualized progesterone dosing and monitoring strategies.⁶ The personalized approach to progesterone management, based on individual response and serum progesterone levels, represents a refinement on standard fixed-dose protocols, enabling optimization of endometrial receptivity for individual women.

The biological rationale for progesterone monitoring reflects the critical importance of adequate luteal phase support in enabling embryo implantation. Insufficient progesterone represents a documented cause of implantation failure and early pregnancy loss; conversely, excessive progesterone provides no additional benefit and may in some circumstances be counterproductive. Individualized dosing based on serum monitoring enables identification of women requiring higher doses to achieve adequate levels, women responding robustly to standard doses, and those experiencing diminished progesterone clearance potentially benefiting from dose reduction.

Clinical implementation of progesterone monitoring protocols requires: (1) access to serum progesterone assessment during luteal phases; (2) healthcare provider education regarding interpretation of progesterone levels and individualized dosing adjustments; (3) standardized protocols specifying timing of progesterone assessment and dose adjustment algorithms; and (4) patient counseling regarding the rationale for monitoring and potential need for dose adjustments based on individual response. The evidence supporting personalized progesterone protocols provides rationale for considering this approach for women with prior implantation failures or recurrent loss.

Quality of evidence and limitations

Methodological strengths

The evidence base demonstrates substantial methodological strengths supporting confidence in conclusions. Six studies (33.3%) achieved low risk of bias across all assessment domains,^{1,5,9,12,14,15} with characteristics including multicenter designs or institutional settings with strong research infrastructure, adequate sample sizes with reported power calculations, pre-specified outcomes, standardized validated measurement instruments, adequate follow-up with acceptable attrition rates, and complete conflict of interest and funding documentation. The predominance of RCT designs (61.1% of studies) provides strong causal inference capability for intervention efficacy assessment. The geographic and contextual diversity of included studies, spanning high-income countries (USA, Scandinavia) and diverse economic contexts (China, South Africa, Afghanistan), strengthens evidence applicability across multiple healthcare systems and populations.^33–37

The comprehensive risk of bias assessment using design-appropriate tools (Cochrane Risk of Bias 2 for RCTs, Newcastle-Ottawa Scale for observational studies, AMSTAR 2 for systematic reviews) enables nuanced evaluation of methodological quality across diverse study designs. The systematic publication bias assessment using multiple complementary approaches (funnel plot visualization, Egger’s regression test, trim-and-fill analysis) provides reassurance that publication bias is unlikely to substantially affect conclusions. The conduct of multiple sensitivity analyses demonstrating robustness of heterogeneity findings across exclusion strategies (I² ranging 71.5%-78.7% across all variations) strengthens confidence that high heterogeneity reflects genuine population and intervention variation rather than methodological artifact.

Limitations and gaps in evidence

Despite methodological strengths, substantial limitations and gaps in evidence require acknowledgment. The predominance of studies from high-income countries (USA, Scandinavia, China) with under-representation of African, South Asian, and low-resource settings limits generalizability and obscures variation in intervention effectiveness across diverse healthcare contexts and resource environments. The single study from Afghanistan and South Africa, while providing valuable perspective from low-resource settings, cannot adequately represent the experiences and needs of the approximately 2 billion women living in low-resource settings globally.

The high heterogeneity (I² = 78.7%) persisting across all sensitivity analyses indicates substantial variation in populations, interventions, and outcomes that limits ability to provide single pooled effect estimates applicable universally. While heterogeneity reflects genuine variability rather than bias, it necessitates careful contextualization of findings and recognition that intervention effectiveness varies meaningfully by population characteristics, healthcare setting, and implementation approaches. Readers should not interpret pooled effect sizes as universally applicable constants but rather as central tendencies around which meaningful variation exists.

The limited number of studies for certain intervention types (only 3 psychological intervention studies, 2 diagnostic/monitoring studies) restricts ability to draw confident conclusions regarding these intervention categories. Meta-analysis of 3 studies provides limited power compared to analyses incorporating 8+ studies, and confidence intervals tend to be wide. Additional research regarding psychological interventions, diagnostic approaches, and implementation strategies in diverse settings is needed to strengthen the evidence base for these important intervention categories.

The psychological outcome data represents cross-sectional prevalence estimates from diverse populations assessed at variable timepoints post-loss rather than longitudinal trajectory data enabling characterization of temporal patterns in psychological recovery. Understanding how depression, anxiety, and PTSD evolve over time following pregnancy loss, which women spontaneously improve versus those requiring intervention, and optimal timing for intervention initiation requires longitudinal investigation not adequately captured in the available cross-sectional evidence base.

The inability to discern optimal intensity, duration, and content of psychological interventions from the available evidence represents a critical gap. While mindfulness, art therapy, and CBT each showed efficacy, direct comparative data regarding their relative effectiveness, optimal duration, optimal timing relative to loss, and characteristics of women most likely to benefit from each approach are limited. Further research comparing psychological interventions and investigating moderators of treatment response is needed to enable evidence-based matching of interventions to individual characteristics.

Clinical implications and recommendations

Integration of evidence-based interventions into standard pregnancy loss care

The evidence base demonstrates that pregnancy loss care has evolved substantially beyond traditional surgical approaches, with multiple evidence-based pharmacological, psychological, and reproductive technology interventions now offering clinically meaningful benefits. Healthcare systems should prioritize integration of these evidence-based interventions into standard pregnancy loss care protocols, moving beyond the limited surgical and expectant management options that have historically dominated clinical practice.

Clinical practice protocols for pregnancy loss management should include: (1) immediate comprehensive assessment including medical stabilization, detailed clinical history, and baseline psychological screening using validated instruments; (2) structured counseling regarding evidence-based treatment options including medical management with mifepristone-misoprostol, expectant management, and surgical approaches, with shared decision-making incorporating woman’s preferences and clinical circumstances; (3) for women choosing medical management, provision of mifepristone-misoprostol combination therapy with clear instructions regarding expected course and when to seek emergency care; (4) for women with recurrent pregnancy loss history or IVF failure, initiation of low-dose aspirin therapy at preconception or early pregnancy; (5) systematic psychological screening using brief validated instruments for all women following pregnancy loss; (6) referral to mental health services for women with identified psychological distress or high-risk features; and (7) integration of mindfulness-based, art therapy, or CBT interventions based on woman’s preferences and availability of services.

The implementation of comprehensive pregnancy loss care protocols requires coordinated efforts among obstetric, primary care, emergency medicine, and mental health providers, necessitating interdisciplinary collaboration and integration. Healthcare system investments in provider training, clinical protocol development, and service infrastructure are needed to enable implementation of these evidence-based approaches. The evidence demonstrates that such investments yield substantial improvements in clinical outcomes (reduced surgical procedures, improved pregnancy achievement, reduced medication abortion complications) and psychological outcomes (reduced depression, anxiety, PTSD prevalence and severity).

Tailoring intervention intensity to loss type and clinical circumstances

The evidence demonstrating differential psychological burden across pregnancy loss types (highest in stillbirth, intermediate in miscarriage, lowest in ectopic pregnancy) supports tailoring intervention intensity to clinical circumstances rather than applying uniform protocols regardless of loss type. Women experiencing stillbirth require more intensive and specialized psychological support, including extended grief counseling, trauma-informed care approaches, and specialized perinatal bereavement services. Women experiencing miscarriage benefit from standard psychological support protocols with additional intensive interventions for those with identified risk factors. Women experiencing ectopic pregnancy require acknowledgment of both the life-threatening medical crisis and pregnancy loss experience.

The identification that psychological burden is highest in stillbirth, reaching 42% for depression, 33% for anxiety, and 18% for PTSD, provides evidence supporting healthcare system resource allocation prioritizing intensive services for this population. Current practice in many settings allocates inadequate resources to stillbirth psychological support, missing opportunity for early intervention and evidence-based care. The substantial psychological burden documented in this meta-analysis provides compelling rationale for major investments in stillbirth bereavement services, specialized grief counseling, and integration of mental health providers into multidisciplinary stillbirth care teams.

Addressing geographic and healthcare context variation

The regional analysis revealing potentially substantial under-detection of psychological conditions in Africa and Asia due to limited healthcare infrastructure demands urgent attention to health equity. Women in low-resource settings face compounded disadvantages: higher baseline risks of pregnancy loss due to limited prenatal care and higher maternal comorbidities, yet reduced access to psychological assessment and treatment. Healthcare system development in low-resource settings should prioritize integration of brief psychological screening into primary care and pregnancy-related services, development and delivery of task-sharing psychological interventions, and capacity-building investments in mental health training and infrastructure.

The demonstrated efficacy of task-sharing CBT delivered by community health workers in South Africa provides a model for expansion of psychological care access in resource-limited settings globally. Investment in training of community health workers, development of supervision infrastructure, and integration into existing community health platforms offers pathways to substantially expand psychological care capacity without requiring creation of specialized psychiatric services in contexts where specialist workforce is scarce. The cost-effectiveness of task-sharing approaches, combined with demonstrated clinical effectiveness, provides compelling rationale for prioritizing these implementation strategies in low-resource settings.

Strategic deployment of telemedicine for expanded access

While in-person interventions demonstrated superior effectiveness, telemedicine’s accessibility advantages render it a valuable service for reaching underserved and geographically remote populations. Healthcare systems should develop coordinated approaches deploying telemedicine strategically to reach populations otherwise excluded from in-person services while continuing to prioritize in-person delivery for women with identified high-risk features or severe symptoms. Investment in technology infrastructure, digital literacy training, and technical support is needed to enable effective telemedicine implementation in resource-limited settings.

The approximately 69% effectiveness of telemedicine relative to in-person delivery, while lower than ideal, still represents clinically meaningful intervention providing substantial improvement over no intervention. For geographically remote women, rural populations, and those in healthcare deserts, telemedicine enables access to psychological support that would otherwise be unavailable. Strategic integration of telemedicine into comprehensive pregnancy loss care enables optimization of both intervention effectiveness (through in-person services for those able to access) and population accessibility (through telemedicine reaching otherwise unreachable populations).

Implications for future research

Identified evidence gaps and research priorities

Despite substantial progress in the evidence base, multiple important gaps remain. Longitudinal studies characterizing temporal trajectories of depression, anxiety, and PTSD following pregnancy loss, identifying which women spontaneously recover versus those requiring intervention, and determining optimal timing for intervention initiation are needed to enable prevention and early intervention strategies. Comparative effectiveness research directly comparing psychological interventions (mindfulness vs. CBT vs. art therapy), identifying which interventions are most effective for different psychological conditions and populations, and investigating moderators of treatment response would enable evidence-based intervention matching.

Research investigating optimal intensity, duration, and delivery method of psychological interventions for different pregnancy loss types and populations is needed. Current evidence indicates that in-person interventions are more effective than telemedicine, but lacks detail regarding optimal frequency, session duration, number of sessions needed for efficacy, and mechanisms underlying superiority of in-person approaches. Investigation of hybrid models combining in-person and telemedicine components might yield cost-effective approaches providing many advantages of in-person interventions with improved accessibility.

Research regarding implementation strategies for integrating evidence-based pregnancy loss care into diverse healthcare systems is needed to translate research findings into clinical practice. Implementation science approaches examining barriers and facilitators to adoption of evidence-based protocols, effectiveness of different dissemination strategies, and cost-effectiveness of different implementation approaches would enable scaling of effective interventions. Research investigating adaptation of interventions across diverse cultural contexts, healthcare systems, and resource environments is essential for ensuring global applicability of evidence-based approaches.

Geographic research priorities and equity

Substantial expansion of research in low-resource settings, African nations, and South Asian countries is urgently needed to reduce the profound geographic imbalance in pregnancy loss research. The current evidence base demonstrates 38.9% from USA, 22.2% from China, 22.2% from Scandinavia, with only 5.6% from South Africa and Afghanistan combined, grossly under-representing the approximately 2 billion women living in Africa and South Asia. Research investments should prioritize investigation of pregnancy loss interventions in diverse healthcare contexts, adaptation of evidence-based approaches to local settings and constraints, and evaluation of implementation strategies feasible within resource-limited environments.

Community-based participatory research approaches incorporating affected women, community members, and local stakeholders in research design, implementation, and interpretation could enhance cultural appropriateness and community benefit of research. Investigation of locally-adapted psychological interventions incorporating indigenous healing practices, community resources, and cultural frameworks alongside evidence-based psychological approaches would enable development of interventions more likely to be adopted and effective within diverse cultural contexts.