Frontiers in Psychiatry 发表产后抑郁运动干预系统综述与网状元分析

Abstract

Objectives:

To synthesize randomized evidence on exercise-based interventions for postpartum depressive symptom scores and to explore comparative signals across exercise types and pragmatic weekly-volume categories using pairwise and network meta-analysis.

Methods:

Eight bibliographic databases were searched from inception and updated through July 23, 2026, with three trial registries searched in the final update, for randomized controlled trials of postpartum exercise interventions reporting Edinburgh Postnatal Depression Scale (EPDS) scores. Pairwise meta-analysis used mean differences (MDs), and network meta-analyses evaluated exercise-type and weekly-volume nodes. The review was registered in PROSPERO (CRD420251275738).

Results:

Twenty-four randomized controlled trials involving 1, 792 unique participants (912 intervention and 880 control) were included. Exercise was associated with lower EPDS scores than usual care/non-exercise control (MD = -2.73, 95% confidence interval (CI) -3.57 to -1.88; I² = 95%), representing an average effect across highly heterogeneous studies. Delivery format showed no difference in an omnibus comparison across three reported categories (p = 0.29), and intervention composition also showed no subgroup difference (p = 0.66). Interventions lasting <12 weeks showed a larger pooled effect than those lasting ≥12 weeks (subgroup p = 0.009), but this was exploratory. In the exercise-type network, yoga/mind-body exercise ranked highest (surface under the cumulative ranking curve (SUCRA) = 94.2%) and was estimated to lower EPDS scores versus aerobic exercise (MD = -2.62, 95% CI -5.16 to -0.08); active-vs-active estimates were entirely indirect and network certainty was very low. The 45–90 min/week category ranked highest in the weekly-volume network (SUCRA = 84.6%), but no between-volume contrast was statistically significant.

Conclusion:

Exercise may reduce postpartum depressive symptom scores, but mixed prevention-oriented and treatment-oriented samples, substantial heterogeneity, co-interventions, and star-shaped networks preclude identifying an optimal exercise type or dose. Exercise may be considered as an adjunct within individualized postpartum mental-health care, not a substitute for clinical assessment or indicated treatment.

Systematic review registration:https://www.crd.york.ac.uk/PROSPERO/view/CRD420251275738, identifier CRD420251275738.

Introduction

Postpartum depressive symptoms are common and clinically consequential during the first year after childbirth, and clinically diagnosed postpartum depression (PPD) is one important presentation within the broader perinatal mental-health spectrum. In this review, prevention-oriented exercise refers to support offered to women with low or subclinical symptom levels, whereas treatment-oriented exercise refers to symptom-reduction programs enrolling women with elevated screening scores or higher symptom-based eligibility thresholds. An elevated Edinburgh Postnatal Depression Scale (EPDS) score indicates symptom burden or risk but does not by itself establish a clinical diagnosis. Symptoms may coincide with physical recovery, infant-care demands, sleep disruption, and changing family roles (). Population estimates indicate that approximately 10-20% of women experience perinatal or postpartum depressive symptoms, with considerable variation by screening method, timing, and population characteristics (). The burden appears higher in many low- and middle-income settings, where social adversity, limited access to care, and weaker health-system infrastructure may compound risk (). Persistent or untreated symptoms can undermine maternal functioning, mother-infant interaction, breastfeeding confidence, and child cognitive and socioemotional development (). Symptoms may also emerge or continue beyond the earliest postnatal screening period, which makes interventions that remain feasible across the first postpartum year clinically important ().

Although evidence-based care for clinically significant postpartum depression includes psychological therapies and pharmacotherapy, treatment access and acceptability remain uneven across postpartum populations. Contemporary clinical guidance emphasizes screening, assessment of severity, suicide-risk evaluation when indicated, and individualized treatment planning rather than a single uniform pathway (). In practice, however, many women do not receive timely care because of stigma, limited specialist availability, cost, competing childcare demands, and uncertainty about whether symptoms warrant formal treatment (). Medication decisions can also be complicated by preferences around breastfeeding, concerns about infant exposure, and the perceived trade-off between maternal symptom relief and medication safety (). These issues do not reduce the importance of medication or psychotherapy for women who need them, but they strengthen the rationale for acceptable, low-cost, and scalable non-pharmacological options that can complement clinical care, support self-management, and provide an early step in stepped-care models (). Reviews of perinatal depression management have similarly emphasized that postpartum care should be flexible, patient-centered, and responsive to the practical constraints faced by women after birth ().

Physical activity and structured exercise are plausible postpartum mental-health interventions because they can be adapted to different functional levels, delivered individually or in groups, and integrated into community or home routines. In the general adult population, exercise has demonstrated antidepressant effects across randomized trials and meta-analytic evidence (). Postpartum-specific evidence suggests that physical activity can reduce depressive and anxiety symptom scores, although effects vary by intervention type, baseline symptom severity, delivery mode, supervision, and adherence (). Earlier reviews of perinatal and postpartum exercise interventions have supported the overall promise of exercise while highlighting heterogeneity in trial designs and intervention prescriptions (, ). This heterogeneity has practical consequences: walking, cycling, aerobic exercise, Pilates, and yoga-informed programs differ in intensity, social context, accessibility, and the relative contribution of movement, breathing, relaxation, and psychological components. Public-health guidance generally encourages postpartum women without contraindications to progress toward regular moderate-intensity activity, but clinical decisions for women with depressive symptoms often require more specific information about feasible modalities and weekly exercise volumes ().

Despite increasing evidence, several questions remain about how comparative exercise evidence should be interpreted in postpartum mental-health care. Conventional pairwise meta-analysis can estimate whether exercise is associated with lower symptom scores than usual care or other non-exercise control conditions, but it is less able to compare several active exercise types when head-to-head trials are sparse. Recent network meta-analyses have begun to address this problem, but uncertainty remains about how Yoga/mind-body exercise, Walking/cycling, and Aerobic exercise compare when newer trials and Chinese-language evidence are considered (, ). An overview of systematic reviews confirmed the relevance of physical-activity interventions while noting that secondary-evidence syntheses are limited in their ability to generate detailed exercise-prescription guidance (). The present review focuses on randomized trials reporting EPDS outcomes, incorporates evidence from English- and Chinese-language databases, and evaluates exercise type and pragmatic weekly-volume categories within a single comparative framework. It synthesizes a mixed prevention-oriented and treatment-oriented evidence base; it does not assume that all participants had clinically diagnosed PPD. Rather than proposing a definitive exercise prescription, the review aims to identify cautious, clinically interpretable signals that may inform individualized postpartum mental-health support and future head-to-head trials.

Methods

This systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement and the PRISMA extension for network meta-analysis where applicable (, ), and was registered in PROSPERO (CRD420251275738). Because this study synthesized published aggregate data and did not involve direct contact with human participants, institutional review board approval and informed consent were not required.

Eligibility criteria

Eligibility criteria were structured according to PICOS. Participants were women aged 18 years or older assessed within 12 months after childbirth. Interventions were exercise-based programs initiated during the postpartum period. Comparators were usual care or non-exercise control conditions, including routine follow-up, health education, social support, psychotherapy, or other non-exercise approaches. The primary outcome was the Edinburgh Postnatal Depression Scale (EPDS), a 10-item self-report measure used to screen, rather than diagnose, postpartum depressive symptoms, measured at baseline and after the intervention (). Requiring EPDS outcomes maintained a common measurement scale across trials and permitted pooling as mean differences without cross-scale standardization. Eligible study designs were randomized controlled trials.

Studies were excluded if exercise began during pregnancy, EPDS data were unavailable, the intervention was not exercise-based, or outcome data could not be extracted. Samples defined by serious physical illness or a major mental disorder, such as bipolar disorder or psychotic depression, were also excluded when these conditions were central characteristics of the recruited population. This restriction improved population comparability but may limit generalizability to women with complex psychiatric or medical histories.

Information sources and search strategy

Two reviewers independently searched PubMed, Embase, Web of Science Core Collection, the Cochrane Library, CINAHL via EBSCOhost, China National Knowledge Infrastructure (CNKI), Wanfang Data, and the VIP Database. The original searches were conducted through December 15, 2025 and were updated and broadened on July 23, 2026; ClinicalTrials.gov, the WHO International Clinical Trials Registry Platform (ICTRP), and the Chinese Clinical Trial Registry (ChiCTR) were also searched in the final update. The updated strategies used three principal concept blocks—postpartum status, depression or depressive symptoms, and exercise or physical activity—and did not require EPDS or randomized-trial terminology as mandatory search blocks. The updated searches were not restricted to abstract fields. No language filter was applied during database searching, although eligible reports were restricted to English- or Chinese-language publications. No publication-status or document-type restrictions were applied at the search stage. Reference lists of relevant reviews and eligible articles were screened manually, yielding two additional records through citation searching. Duplicate citations were removed electronically and then verified manually. Database- and registry-specific strategies, search dates, fields, and record counts are provided in Supplementary Material 1.

Study selection

After deduplication, two reviewers independently screened titles and abstracts and then assessed full texts for eligibility. Disagreements were resolved through discussion, and a third reviewer adjudicated unresolved discrepancies. Screening decisions and exclusion reasons were documented in a PRISMA flow diagram.

Data extraction and risk-of-bias assessment

Two reviewers independently extracted study characteristics, participant information, intervention details, comparator descriptions, and outcome data; disagreements during data extraction and risk-of-bias assessment were resolved through discussion, with unresolved discrepancies adjudicated by a third reviewer (Liuhong Zang). Extracted variables included country, age, EPDS eligibility threshold, group-specific baseline EPDS mean ± SD, completed follow-up sample size at the selected post-intervention EPDS assessment, intervention format, intervention components, postpartum start time, intervention duration, weekly exercise dose, and post-intervention EPDS values. When required outcome data were missing, unclear, or incompletely reported, the corresponding author was contacted up to three times over a three-week period. When a standard error (SE) was reported instead of an SD, it was converted using SD = SE × √n. No other missing EPDS outcome values were statistically derived or imputed. When multiple eligible post-baseline assessments were reported, the assessment corresponding most closely to the end of the active intervention period was selected for the primary analysis; later assessments after cessation of the active intervention were treated as follow-up measurements and were not used as the primary post-intervention endpoint. Baseline EPDS values were used to describe variation in symptom level across trials, not as standardized clinical diagnoses. When both change scores and post-intervention scores were available, post-intervention EPDS scores were used for the primary analysis to maintain consistency across trials. Baseline imbalances were not statistically adjusted in the study-level mean differences; instead, baseline EPDS values were extracted and considered when interpreting heterogeneity, transitivity, and clinical comparability. Because the two active intervention arms in Liu et al. () were assigned to the same yoga/mind-body exercise analytical node, they were combined before analysis using standard formulas for combining groups: means were weighted by sample size, and the combined standard deviation incorporated both within-arm variance and between-arm differences in means. The shared control group was included only once. The combined active group comprised 57 participants (baseline EPDS 8.43 ± 4.44; post-intervention EPDS 6.70 ± 3.87), and the control group comprised 33 participants (baseline EPDS 8.98 ± 3.97; post-intervention EPDS 7.76 ± 4.10). Risk of bias was assessed independently with Cochrane RoB 2 for the effect of assignment to the randomized intervention on the post-intervention EPDS outcome. Domains were D1 randomization process, D2 deviations from intended interventions, D3 missing outcome data, D4 outcome measurement, and D5 selection of the reported result; overall judgments were Low risk, Some concerns, or High risk (). For D5, the availability of prespecified outcome information and the potential for selection among multiple eligible outcome measurements or analyses were considered when judging bias in selection of the reported result. The full trial report was the primary source, supplemented by trial registrations, protocols, or other prespecified outcome information where available.

Intervention grouping

For subgroup analyses, interventions were classified according to three contextual features: delivery format (Individual/home-based exercise, Group-based exercise, or Format unclear/not reported), intervention composition (Exercise-only or Exercise plus co-intervention), and duration (<12 weeks or ≥12 weeks). The Format unclear/not reported category indicates insufficient reporting in the original study and is not a distinct clinically meaningful delivery strategy. The delivery-format subgroup-difference test was specified as an omnibus comparison across all three reported categories. These subgroup comparisons were intended to explore contextual sources of heterogeneity rather than to establish optimal program characteristics.

For the exercise-type network meta-analysis, the node taxonomy was defined pragmatically from the prescribed physical exercise and interventions were classified by their dominant movement component into Walking/cycling, Aerobic exercise, Yoga/mind-body exercise, and the common comparator node Usual care/non-exercise control. Walking/cycling included walking, pram walking or pushing, and cycling/pedal-based programs. Aerobic exercise included structured aerobic, gymnastic, or cardiorespiratory programs and mixed exercise programs when aerobic/cardiorespiratory activity was the predominant prescribed physical component. Yoga/mind-body exercise included yoga, laughter yoga, Pilates, mindfulness-yoga, and related programs when structured postures or movement integrated with breathing, relaxation, or mindfulness formed a core physical component. When programs contained multiple exercise or non-exercise elements, node assignment was based on the principal prescribed physical activity described in the intervention protocol rather than on adjunctive psychological, educational, digital, social-support, or nutrition/hygiene components. This pragmatic classification organized heterogeneous interventions but did not imply equivalence of all active ingredients within a node; accordingly, network estimates should not be interpreted as pure exercise-type effects.

For the weekly-volume network, weekly volume was derived from the prescribed dose information available in each trial. Fixed frequency and duration were multiplied directly; midpoint values were used for reported ranges, minimum values when only a minimum dose was specified, and reported weekly means were retained. If a reported dose range crossed categories, the resulting representative value determined node assignment. After dose extraction, studies were grouped pragmatically into 45-90, 91-150, and 151–360 min/week to organize the observed distribution into analyzable network nodes; these intervals were evidence-grouping categories rather than clinical prescription thresholds. Usual care/non-exercise control constituted the common comparator node. Because classification was based on prescribed rather than achieved exercise volume, it did not incorporate adherence or attendance.

Statistical analysis

Pairwise meta-analyses were performed in Review Manager 5.4 using mean differences (MDs) with 95% confidence intervals (CIs), because all included studies reported EPDS scores on the same scale. Heterogeneity was assessed using the I² statistic and Cochran’s Q test (). Fixed-effect models were used when I² ≤ 50% and the Q-test p value was > 0.10; random-effects models were used when I² > 50% or p ≤ 0.10. Random-effects models were used when heterogeneity criteria were met because a common true effect was not assumed across clinically and methodologically heterogeneous trials. For random-effects analyses, between-study variance (τ²) was estimated using the DerSimonian-Laird method. A 95% prediction interval was calculated for the overall random-effects estimate to describe the expected range of true effects in a new comparable study.

Network meta-analyses were performed in Stata 17.0 using the network suite, with multivariate meta-analytic estimation implemented through mvmeta, under a frequentist random-effects consistency framework and with a common heterogeneity variance (τ²) estimated by restricted maximum likelihood (REML) within each network. Network plots displayed the geometry of the available evidence, with node size proportional to sample size and edge thickness proportional to the number of direct comparisons. Transitivity was assessed descriptively by comparing EPDS eligibility and baseline symptom levels, postpartum timing, intervention duration, supervision, co-interventions, and control-condition type across nodes; no formal between-node tests were performed. Distributions are reported in Supplementary Tables S5, S6. The common comparator node is labeled Usual care/non-exercise control and includes routine follow-up or other eligible non-exercise conditions. Comparator conditions were additionally classified as passive/usual-care (no trial-specific structured non-exercise intervention beyond routine care) or active (a trial-specific structured non-exercise intervention); both were retained within the primary Control node to preserve network connectivity without assuming clinical equivalence. Both evidence networks were star-shaped, centered on this common comparator, and lacked closed loops. Active-vs-active estimates were therefore entirely indirect, and incoherence could not be empirically evaluated. Treatment rankings were summarized using the surface under the cumulative ranking curve (SUCRA) only as exploratory ranking signals, not as evidence of a best intervention or optimal dose. Sensitivity analyses included the existing leave-one-study-out approach, an active-versus-passive comparator subgroup analysis, and pairwise, exercise-type network, and weekly-volume network analyses restricted to passive/usual-care-controlled trials; detailed results are provided in Supplementary Tables S1–S4. Publication bias was assessed with funnel plots and Egger’s regression test.

Certainty of the main network estimates was assessed using a structured CINeMA-informed framework covering within-study bias, reporting bias, indirectness, imprecision, heterogeneity, and incoherence (). For imprecision and heterogeneity judgments, a 4-point EPDS difference was used as an evidence-informed clinically important threshold, defining a symmetric equivalence range of -4 to +4 points. This choice was informed by a 4-point reliable-change criterion derived for the EPDS in a postnatal sample and a 4-point MCID for improvement reported in a pregnancy sample (, ); because these thresholds were developed primarily for within-person change, the 4-point value was treated as an evidence-informed decision threshold rather than a universally established between-group MID. Indirectness judgments were informed by the distributions of potential effect modifiers in Supplementary Tables S5, S6. Because both networks lacked closed loops, incoherence could not be empirically evaluated; consistent with the CINeMA default rule when the design-by-treatment interaction test is not estimable, this was treated as a major concern in the certainty tables. Overall confidence ratings were summarized as high, moderate, low, or very low and are reported in Supplementary Tables S7, S8.

Results

Study selection

The updated search identified 6, 206 records in total: 6, 204 from bibliographic databases and trial registries and two through citation searching. The broadened July 23, 2026 update identified no additional eligible trials. After removal of 2, 787 duplicates, 3, 419 records were screened, of which 2, 759 were excluded. Of 660 reports sought for retrieval, eight were not retrieved; 652 reports were assessed for eligibility and 628 were excluded, leaving 24 randomized controlled trials for quantitative synthesis (Figure 1).

Figure 1

Study characteristics

The 24 included trials contributed 24 independent study-level comparisons involving 1, 792 unique participants (912 intervention and 880 control) from eight countries or regions. For Liu et al. (), the two active arms were combined before analysis and the shared control was included once. Interventions were initiated between the first postpartum month and 12 months after childbirth and lasted 4–24 weeks. Delivery format, supervision, and adjunctive non-exercise components varied across studies, as summarized in Table 1. Baseline EPDS criteria and scores also varied: some trials were prevention-oriented or early-support studies enrolling women with low or subclinical symptom scores, whereas others were treatment-oriented symptom-reduction studies using higher EPDS eligibility thresholds. Because EPDS is a screening scale, these thresholds do not establish a uniform population with clinically diagnosed PPD. Thirteen studies were conducted in middle-income settings (China, Turkey, or India), whereas none were conducted in low-income countries.

Table 1

Characteristics of the included studies.

E, experimental group; C, control group; EPDS, Edinburgh Postnatal Depression Scale; PP, postpartum; SD, standard deviation; NA, not available. Age and EPDS values are as reported; sample size is E/C. “None reported” indicates no adjunctive non-exercise component was reported; “Format unclear/not reported” indicates that the original report did not permit a clear individual/home-based or group-based classification.

Exercise content was organized into the analytical nodes Walking/cycling, Aerobic exercise, Yoga/mind-body exercise, and Usual care/non-exercise control. Active programs were delivered alone or with co-interventions. All included studies assessed depressive symptoms using the EPDS. Because several yoga/mind-body exercise and aerobic exercise programs included psychological, educational, digital, social-support, or nutrition/hygiene components, network estimates should not be interpreted as pure exercise-only effects.

Risk of bias

Overall RoB 2 judgments were Some concerns for 20 studies and High risk for four studies (, , 48, 50); none were Low risk. All 24 trials had Some concerns in D4 because EPDS was participant-reported and participants were generally aware of assignment (Figures 2, 3).

Figure 2

Figure 3

Pairwise meta-analysis

The pooled pairwise meta-analysis indicated that exercise was associated with lower postpartum depressive symptom scores than usual care/non-exercise control (MD = -2.73, 95% CI -3.57 to -1.88; I² = 95%) (Figure 4). The 95% prediction interval was -6.50 to 1.05, indicating that the true effect in a new comparable study could range from a larger reduction in EPDS scores to little or no benefit. This MD is an average effect across highly heterogeneous populations and interventions and should not be interpreted as a common effect expected in all postpartum populations.

Figure 4

In the delivery-format subgroup, pooled estimates favored exercise over usual care/non-exercise control for Individual/home-based exercise (MD = -2.80, 95% CI -4.89 to -0.71; I² = 95%), Group-based exercise (MD = -2.83, 95% CI -4.06 to -1.61; I² = 94%), and Format unclear/not reported (MD = -1.93, 95% CI -2.25 to -1.61; I² = 0%). The omnibus subgroup-difference test across all three categories showed no evidence of a difference (p = 0.29) (Figure 5; it was not a pairwise comparison of Group-based exercise with Individual/home-based exercise).

Figure 5

In the intervention-composition subgroup, pooled estimates favored Exercise plus co-intervention (MD = -2.98, 95% CI -4.52 to -1.45; I² = 97%) and Exercise-only (MD = -2.57, 95% CI -3.66 to -1.48; I² = 93%) over usual care/non-exercise control. There was no statistically significant subgroup difference (p = 0.66) (Figure 6).

Figure 6

Intervention duration was the only contextual factor with a statistically significant subgroup-difference signal (p = 0.009) (Figure 7). Programs lasting <12 weeks had a pooled MD of -3.72 (95% CI -5.18 to -2.26; I² = 97%), whereas programs lasting ≥12 weeks had a pooled MD of -1.57 (95% CI -2.24 to -0.89; I² = 62%). Because heterogeneity remained very high in the <12-week subgroup and the comparison was exploratory, this pattern should not be interpreted as evidence of an optimal duration.

Figure 7

Network meta-analysis

In the exercise-type network, all 24 studies contributed data across Walking/cycling (6 direct comparisons), Aerobic exercise (), Yoga/mind-body exercise (), and Usual care/non-exercise control (Figure 8). SUCRA values were 94.2% for Yoga/mind-body exercise, 62.2% for Walking/cycling, 42.3% for Aerobic exercise, and 1.3% for Usual care/non-exercise control. The common heterogeneity variance was τ² = 6.59. The estimated comparison of Yoga/mind-body exercise with Aerobic exercise favored Yoga/mind-body exercise (MD = -2.62, 95% CI -5.16 to -0.08) (Table 2). This is an exploratory, entirely indirect active-vs-active comparison: the star-shaped network lacked closed loops, incoherence could not be empirically evaluated, and several active programs included co-interventions. The comparison and SUCRA ranking should be interpreted in light of the very-low-certainty evidence in the CINeMA-informed assessment (Supplementary Table S7).

Figure 8

Table 2

League table for the network meta-analysis of exercise type.

Values are mean differences (95% confidence intervals) in post-intervention EPDS scores for Walking/cycling, Aerobic exercise, Yoga/mind-body exercise, and Usual care/non-exercise control. Negative values favor the column intervention over the row intervention. EPDS, Edinburgh Postnatal Depression Scale. *95% confidence interval excludes 0.

In the weekly-volume network, all 24 studies contributed data across 45–90 min/week, 91–150 min/week, 151–360 min/week, and Usual care/non-exercise control (Figure 9). SUCRA values were 84.6%, 71.9%, 36.9%, and 6.5%, respectively. The common heterogeneity variance was τ² = 7.52. None of the contrasts among the three active weekly-volume nodes was statistically significant. The SUCRA ordering is therefore an exploratory ranking signal that should be interpreted in light of very-low-certainty evidence and does not establish an optimal weekly exercise dose (Table 3; Supplementary Table S8).

Figure 9

Table 3

League table for the network meta-analysis of weekly exercise volume.

Values are mean differences (95% confidence intervals) in post-intervention EPDS scores. Negative values favor the column intervention over the row intervention. EPDS, Edinburgh Postnatal Depression Scale. *95% confidence interval excludes 0.

Descriptive assessment showed some imbalance in potential effect modifiers across nodes. Yoga/mind-body trials more often included co-interventions and supervised delivery, while weekly-volume nodes differed more clearly in baseline EPDS, EPDS-based eligibility, postpartum timing, and comparator type; for example, active comparators occurred in 40.0%, 50.0%, and 0.0% of the 45-90, 91-150, and 151–360 min/week nodes, respectively. Transitivity was therefore considered plausible only with caution, particularly for the weekly-volume network (Supplementary Tables S5, S6).

Certainty was rated very low for all six main comparisons in each network (Supplementary Tables S7, S8). Within-study bias raised some concerns because none of the included trials was overall low risk in RoB 2, whereas reporting bias was rated as undetected, with limited power at the comparison level. Indirectness raised some concerns for exercise-type active-vs-active comparisons and major concerns for weekly-volume active-vs-active comparisons because of the effect-modifier imbalances described above. Imprecision and heterogeneity were judged relative to the -4 to +4 EPDS equivalence range using the corresponding 95% CIs and prediction intervals. Because neither network contained a closed loop, incoherence could not be empirically evaluated; under the CINeMA default rule, this non-estimability was treated as a major concern and contributed to the very-low overall confidence.

Sensitivity analysis

Leave-one-out analyses indicated that omission of any single study did not materially change the direction or statistical significance of the overall pooled effect (Table 4). The pooled MD remained close to the main estimate in all iterations; however, heterogeneity remained substantial in every iteration (I² = 92%-95%). The analysis therefore supports robustness of the pooled effect direction and statistical significance to single-study omission, but not consistency of effect magnitudes across studies.

Table 4

Leave-one-out sensitivity analysis for the overall pairwise meta-analysis.

MD, mean difference; CI, confidence interval; I², inconsistency statistic. Negative MD values favor exercise over usual care or other non-exercise control conditions. The overall analysis includes all eligible comparisons, and each subsequent row shows the pooled estimate after excluding the indicated study.

Comparator-type subgroup analysis yielded pooled MDs of -3.04 (95% CI -5.09 to -0.99) for passive/usual-care comparators and -2.21 (95% CI -3.01 to -1.40) for active comparators, with no subgroup difference (p = 0.459). Restricting the pairwise analysis to the 15 passive/usual-care-controlled trials preserved the direction of effect (MD = -3.04, 95% CI -5.09 to -0.99), although heterogeneity remained very high (I² = 95%). In the passive-only exercise-type network, Yoga/mind-body exercise remained highest ranked and favored Aerobic exercise (MD = -4.40, 95% CI -8.23 to -0.57). In the passive-only weekly-volume network, 45–90 and 91–150 min/week were nearly tied (SUCRA = 77.9% and 77.7%, respectively; MD = 0.03, 95% CI -4.38 to 4.44). Full results are provided in Supplementary Tables S1–S4.

Publication bias

The pairwise funnel plot was broadly symmetrical, and Egger’s regression did not indicate significant small-study effects (p = 0.81; exact bias-term p = 0.8052) (Table 5). Comparison-adjusted funnel plots for the exercise-type and weekly-volume networks were interpreted cautiously; Egger’s regression tests did not indicate statistically significant small-study effects (p = 0.93 and p = 0.44, respectively), while visual asymmetry may suggest small-study effects and the limited number of studies per comparison constrained interpretation (Supplementary Figures S1, S2 in Supplementary Material 1).

Table 5

Egger regression test for the overall pairwise meta-analysis.

SE, standard error; CI, confidence interval. The bias term represents the Egger regression intercept used to assess small-study effects. A non-significant bias term indicates no statistical evidence of funnel plot asymmetry.

Consolidated summary of findings

To consolidate the main findings, Table 6 summarizes the overall meta-analytic result, the subgroup analyses, and the key network rankings.

Table 6

Summary of key contextual findings across the pooled and network analyses.

MD, mean difference; CI, confidence interval; EPDS, Edinburgh Postnatal Depression Scale; SUCRA, surface under the cumulative ranking curve. Pooled estimates are average effects across heterogeneous studies. SUCRA values are exploratory ranking signals and should be interpreted together with the corresponding forest plots, league tables, and network plots.

Discussion

Principal findings

This review found that exercise was associated with lower postpartum depressive symptom scores than usual care/non-exercise control (MD = -2.73, 95% CI -3.57 to -1.88), but the pooled evidence was highly heterogeneous (I² = 95%). The estimate therefore represents an average across substantially different populations and interventions, not a common effect expected in every postpartum population. The average between-group difference of 2.73 EPDS points was smaller than the 4-point evidence-informed threshold used for the certainty assessment, while the 95% prediction interval (-6.50 to 1.05) extended from a potentially larger benefit to little or no effect. Because the 4-point threshold is derived mainly from within-person change and is not a universally established between-group minimal important difference, the clinical relevance of the pooled MD depends on baseline symptom severity, population characteristics, and the context in which change is interpreted. Delivery format did not differ in an omnibus comparison across Individual/home-based exercise, Group-based exercise, and Format unclear/not reported (p = 0.29), and intervention composition also showed no subgroup difference (p = 0.66). The duration comparison favored programs lasting <12 weeks (p = 0.009), but substantial residual heterogeneity limits causal interpretation. In the exercise-type network, yoga/mind-body exercise ranked highest and showed an estimated advantage over aerobic exercise, whereas 45–90 min/week ranked highest in the weekly-volume network (SUCRA = 84.6%) without statistically significant between-volume contrasts. These are exploratory, hypothesis-generating signals, and the CINeMA-informed certainty was very low for all main network comparisons (Supplementary Tables S7, S8); neither the league-table contrasts nor SUCRA rankings should therefore be interpreted as establishing superiority or an optimal dose. Comparator type did not significantly modify the pooled effect (p = 0.459), and restricting analyses to passive/usual-care controls preserved the direction of the pairwise effect and the Yoga/mind-body ranking. However, heterogeneity remained high and the 45–90 and 91–150 min/week nodes became nearly indistinguishable, indicating that comparator composition alone did not explain the observed heterogeneity and further weakening any claim of an optimal weekly dose. The transitivity assessment also supports cautious interpretation: exercise-type nodes differed in co-interventions and supervision, and weekly-volume nodes showed clearer imbalance in baseline symptom burden, eligibility, postpartum timing, and comparator composition, so active-vs-active network estimates may partly reflect effect-modifier differences rather than exercise type or volume alone.

The favorable ranking signal for yoga/mind-body exercise should not be interpreted as a specific effect of movement alone. Yoga-informed programs may combine gentle physical postures with breathing regulation, relaxation, and attentional focus, elements that may support stress regulation and mood (51). In the postpartum period, these features may be relevant because women often experience fatigue, sleep disruption, and emotional overload. However, the included yoga/mind-body exercise programs varied in the contribution of active movement, meditation, mindfulness, CBT, counseling, education, VR, social support, digital support, and nutrition/hygiene components. The available evidence therefore cannot separate the contribution of exercise type from the contribution of these adjunctive components.

The subgroup signal for interventions lasting <12 weeks (p = 0.009) should be interpreted as an exploratory study-level association rather than evidence that shorter interventions are intrinsically superior. Residual confounding and ecological bias are plausible because the duration groups also differed across intervention type, baseline symptom severity, co-interventions, comparator conditions, and participant characteristics. Greater feasibility or adherence with shorter programs is one possible explanation, but adherence, attendance, and intervention completion were not systematically evaluated in this review; this explanation therefore remains a hypothesis rather than a finding. Likewise, the higher exploratory ranking of 45–90 min/week does not show that higher volumes are ineffective or that lower volume is optimal. None of the active between-volume contrasts was statistically significant, and the category boundaries were pragmatic evidence-grouping intervals rather than clinical prescription thresholds.

Implications for perinatal mental health practice

These findings have different implications for prevention-oriented and treatment-oriented postpartum care and should not be translated into a uniform exercise prescription. In prevention-oriented or early-support samples with low or subclinical EPDS scores, exercise may be offered as preference-sensitive support for well-being and symptom management. In treatment-oriented samples with elevated screening scores, exercise may be considered as an adjunct to clinical assessment and indicated care; an EPDS score is not itself a diagnosis. Exercise should not replace diagnostic evaluation, suicide-risk assessment when indicated, psychotherapy, pharmacotherapy, or referral. Community health workers, midwives, nurses, and primary-care providers could offer brief activity guidance and referral pathways, particularly for women with persistent or moderate-to-severe symptoms. Yoga-informed programs may be attractive for some women, but they will not be acceptable, culturally appropriate, affordable, or physically comfortable for everyone. Trauma-informed yoga literature has emphasized that survivors of sexual or interpersonal violence may need choice-based, trauma-sensitive delivery and that standard yoga environments can present barriers if safety and autonomy are not prioritized (52, 53). Walking-based and other low-burden home or community programs therefore remain important alternatives, especially for women who face time, transport, childcare, or financial barriers to participation (54).

Strengths and limitations

This review has several strengths. It included only randomized trials, searched both English- and Chinese-language databases, used contextual subgroup analyses, and combined pairwise meta-analysis with network meta-analysis to organize evidence on modalities and weekly-volume categories. Independent screening, extraction, and risk-of-bias assessment further strengthened internal consistency. These strengths support a broad synthesis of the available evidence, while the results remain unsuitable for a definitive exercise prescription because of the diversity of interventions and the limited direct comparative evidence.

Several limitations should temper interpretation. First, heterogeneity was substantial in the overall meta-analysis and remained high in several subgroups, so the pooled MD is an average across heterogeneous studies rather than a uniform expected clinical effect. Second, the evidence mixed prevention-oriented or early-support samples with low or subclinical baseline scores and treatment-oriented samples selected using higher EPDS thresholds. Baseline symptom severity and eligibility thresholds varied, EPDS is a screening instrument rather than a diagnostic interview, and the overall result cannot be interpreted as a single treatment effect for clinically diagnosed PPD. Restricting eligibility to EPDS outcomes improved scale comparability but may have excluded otherwise relevant randomized trials that used other validated depression measures. Third, both networks were star-shaped, centered on usual care/non-exercise control, and lacked closed loops. Active-vs-active estimates were therefore entirely indirect, and incoherence could not be empirically evaluated. Fourth, yoga/mind-body exercise and other nodes often contained multicomponent programs, making it difficult to isolate the effect of exercise type. The weekly-volume nodes were based on prescribed dose and pragmatic range, midpoint, and minimum-value rules; classification uncertainty, adherence differences, and within-node heterogeneity may therefore have influenced the volume comparisons. Fifth, some Chinese-language trials provided limited methodological detail, which may have affected risk-of-bias judgments and confidence in pooled estimates. The CINeMA-informed assessment rated certainty as very low for all main network comparisons, reflecting within-study limitations, heterogeneity and indirectness/transitivity concerns, together with the inability to empirically evaluate incoherence in the star-shaped networks. Sixth, although 13 studies were from middle-income countries, no included study was conducted in a low-income country, limiting generalizability to settings where the burden of postpartum depressive symptoms may be greatest. Finally, because populations defined by serious physical illness or major mental disorder were excluded, the findings may be less applicable to women with complex psychiatric or medical histories.

Conclusion

Across 24 randomized trials involving 1, 792 unique participants, exercise was associated with lower postpartum depressive symptom scores than usual care/non-exercise control (MD = -2.73, 95% CI -3.57 to -1.88; I² = 95%), but this was an average effect across highly heterogeneous studies. Yoga/mind-body exercise ranked highest in the exercise-type network and showed an exploratory estimated advantage over aerobic exercise (MD = -2.62, 95% CI -5.16 to -0.08), while 45–90 min/week ranked highest in the weekly-volume network (SUCRA = 84.6%) without statistically significant differences among active volume categories. The underlying network comparisons were of very low certainty, so these rankings and active-vs-active estimates remain exploratory. Because the evidence mixed prevention-oriented and treatment-oriented samples, included multicomponent interventions, and came from star-shaped networks in which active-vs-active comparisons were entirely indirect and incoherence could not be empirically evaluated, it does not identify an optimal exercise type, duration, or weekly dose. Exercise may be considered as an individualized adjunct within postpartum mental-health care and should not substitute for clinical assessment or indicated treatment.

Statements

Data availability statement

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Author contributions

SL: Formal analysis, Writing – original draft, Software, Resources, Visualization, Methodology, Supervision, Conceptualization, Data curation, Writing – review & editing, Investigation. YT: Methodology, Data curation, Writing – original draft, Investigation, Conceptualization, Writing – review & editing. LZ: Writing – original draft, Project administration, Visualization, Supervision, Writing – review & editing, Funding acquisition.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This work was supported by the Xinjiang Normal University Think Tank Bidding Project Key Project “Research on the Practice of Digital Empowerment for High-Quality Development of Rural Physical Education in Ethnic Minority Areas” (Grant No. ZK2025B13) and 2026 Xinjiang Normal University Graduate Research Innovation Project, ‘Research on the Construction of an Active Health System for Smart Fitness Communities for Older Adults in Xinjiang’ (Grant No. XSY202601056).

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyt.2026.1921342/full#supplementary-material

References

• 1

  O’HaraMWMcCabeJE. Postpartum depression: Current status and future directions. Annu Rev Clin Psychol. (2013) 9:379–407. doi: 10.1146/annurev-clinpsy-050212-185612

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  WoodyCAFerrariAJSiskindDJWhitefordHAHarrisMG. A systematic review and meta-regression of the prevalence and incidence of perinatal depression. J Affect Disord. (2017) 219:86–92. doi: 10.1016/j.jad.2017.05.003

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  Roddy MitchellAGordonHLindquistAWalkerSPHomerCSEMiddletonAet al. Prevalence of perinatal depression in low- and middle-income countries: A systematic review and meta-analysis. JAMA Psychiatry. (2023) 80:425–31. doi: 10.1001/jamapsychiatry.2023.0069

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  SteinAPearsonRMGoodmanSHRapaERahmanAMcCallumMet al. Effects of perinatal mental disorders on the fetus and child. Lancet. (2014) 384:1800–19. doi: 10.1016/S0140-6736(14)61277-0

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  RobbinsCLKoJYD’AngeloDVSharmaAJBarfieldWD. Timing of postpartum depressive symptoms. Prev Chronic Dis. (2023) 20:230107. doi: 10.5888/pcd20.230107

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  American College of Obstetricians and Gynecologists. Screening and diagnosis of mental health conditions during pregnancy and postpartum: ACOG Clinical Practice Guideline No. 4. Obstet Gynecol. (2023) 141:1232–61. doi: 10.1097/AOG.0000000000005200

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  Bodnar-DerenSBennEKTBalbierzAHowellEA. Stigma and postpartum depression treatment acceptability among Black and White women in the first six months postpartum. Matern Child Health J. (2017) 21:1457–68. doi: 10.1007/s10995-017-2263-6

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  BattleCLSalisburyALSchofieldCAOrtiz-HernandezS. Perinatal antidepressant use: Understanding women's preferences and concerns. J Psychiatr Pract. (2013) 19:443–53. doi: 10.1097/01.PRA.0000438183.74359.46

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 9

  StewartDEVigodS. Postpartum depression. N Engl J Med. (2016) 375:2177–86. doi: 10.1056/NEJMcp1607649

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 10

  McDonaldMAlhusenJ. A review of treatments and clinical guidelines for perinatal depression. J Perinat Neonatal Nurs. (2022) 36:233–42. doi: 10.1097/JPN.0000000000000661

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 11

  HeisselAHeinenDBrokmeierLLSkarabisNKangasMVancampfortDet al. Exercise as medicine for depressive symptoms? A systematic review and meta-analysis with meta-regression. Br J Sports Med. (2023) 57:1049–57. doi: 10.1136/bjsports-2022-106282

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 12

  DepratoARuchatSMAliMUDavenportMHMottolaMFSkowRJ. Impact of postpartum physical activity on maternal depression and anxiety: A systematic review and meta-analysis. Br J Sports Med. (2025) 59:550–61. doi: 10.1136/bjsports-2024-108478

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  CarterTBastounisAGuoBMorrellCJ. The effectiveness of exercise-based interventions for preventing or treating postpartum depression: A systematic review and meta-analysis. Arch Womens Ment Health. (2019) 22:37–53. doi: 10.1007/s00737-018-0869-3

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 14

  HeLPSohKLHuangFFZhaoLLiuWFLiJet al. The impact of physical activity intervention on perinatal depression: A systematic review and meta-analysis. J Affect Disord. (2023) 321:304–19. doi: 10.1016/j.jad.2022.10.026

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  BullFCAl-AnsariSSBiddleSBorodulinKBumanMPCardonGet al. World Health Organization 2020 guidelines on physical activity and sedentary behaviour. Br J Sports Med. (2020) 54:1451–62. doi: 10.1136/bjsports-2020-102955

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 16

  XuHLiuRYWangXBYangJH. Effectiveness of aerobic exercise in the prevention and treatment of postpartum depression: Meta-analysis and network meta-analysis. PloS One. (2023) 18:e0287650. doi: 10.1371/journal.pone.0287650

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 17

  WangJCarruCSeddaSFioriPLLiZChenZ. Comparative impact of exercise-based interventions for postpartum depression: A Bayesian network meta-analysis. Int J Gynaecol Obstet. (2024) 165:67–75. doi: 10.1002/ijgo.15091

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 18

  Campos-MarinAGarcia-MunozCMatias-SotoJMartinez-CalderonJ. What interventions focused on physical activity could improve postpartum depression symptoms? An overview of systematic reviews with meta-analysis. Healthcare (Basel). (2025) 13:1419. doi: 10.3390/healthcare13121419

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 19

  PageMJMcKenzieJEBossuytPMBoutronIHoffmannTCMulrowCDet al. The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ. (2021) 372:n71. doi: 10.1136/bmj.n71

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 20

  HuttonBSalantiGCaldwellDMChaimaniASchmidCHCameronCet al. The PRISMA extension statement for reporting of systematic reviews incorporating network meta-analyses of health care interventions: Checklist and explanations. PloS Med. (2015) 12:e1001829. doi: 10.1371/journal.pmed.1001829

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 21

  CoxJLHoldenJMSagovskyR. Detection of postnatal depression: Development of the 10-item Edinburgh Postnatal Depression Scale. Br J Psychiatry. (1987) 150:782–6. doi: 10.1192/bjp.150.6.782

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 22

  LiuNDengJLuFXiaoJ. Virtual reality enhanced mindfulness and yoga intervention for postpartum depression and anxiety in the post-COVID era. Sci Rep. (2025) 15:11766. doi: 10.1038/s41598-025-96165-6

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 23

  SterneJACSavovicJPageMJElbersRGBlencoweNSBoutronIet al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ. (2019) 366:l4898. doi: 10.1136/bmj.l4898

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 24

  HigginsJPTThompsonSGDeeksJJAltmanDG. Measuring inconsistency in meta-analyses. BMJ. (2003) 327:557–60. doi: 10.1136/bmj.327.7414.557

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 25

  NikolakopoulouAHigginsJPTPapakonstantinouTChaimaniADel GiovaneCEggerMet al. CINeMA: An approach for assessing confidence in the results of a network meta-analysis. PloS Med. (2020) 17:e1003082. doi: 10.1371/journal.pmed.1003082

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 26

  MattheyS. Calculating clinically significant change in postnatal depression studies using the Edinburgh Postnatal Depression Scale. J Affect Disord. (2004) 78:269–72. doi: 10.1016/S0165-0327(02)00313-0

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 27

  MaoFSunYWangJHuangYLuYCaoF. Sensitivity to change and minimal clinically important difference of Edinburgh postnatal depression scale. Asian J Psychiatr. (2021) 66:102873. doi: 10.1016/j.ajp.2021.102873

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 28

  ArmstrongKEdwardsH. The effects of exercise and social support on mothers reporting depressive symptoms: A pilot randomized controlled trial. Int J Ment Health Nurs. (2003) 12:130–8. doi: 10.1046/j.1440-0979.2003.00229.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 29

  ArmstrongKEdwardsH. The effectiveness of a pram-walking exercise programme in reducing depressive symptomatology for postnatal women. Int J Nurs Pract. (2004) 10:177–94. doi: 10.1111/j.1440-172X.2004.00478.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 30

  DaleyAJWinterHGrimmettCMcGuinnessMMcManusRMacArthurC. Feasibility of an exercise intervention for women with postnatal depression: A pilot randomised controlled trial. Br J Gen Pract. (2008) 58:178–83. doi: 10.3399/bjgp08X277195

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 31

  KellerCAinsworthBRecordsKToddMBelyeaMVega-LopezSet al. A comparison of a social support physical activity intervention in weight management among post-partum Latinas. BMC Public Health. (2014) 14:971. doi: 10.1186/1471-2458-14-971

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 32

  RenWWuCWXieY. Effects of moderate-intensity aerobic exercise on mild to moderate postpartum depression and self-efficacy. Mod J Integr Tradit Chin West Med. (2019) 28:1500–3. doi: 10.3969/j.issn.1008-8849.2019.14.005

  • CrossRef
  • Google Scholar
  • View reference in article

• 33

  RobichaudK. The effects of an exercise intervention on the psychological well-being of postpartum women. Doctoral Dissertation (2008) Murfreesboro, TN: Middle Tennessee State University.

  • Google Scholar
  • View reference in article

• 34

  ChenF. Application effect of a mobile network platform combined with yoga relaxation training in women with postpartum depression. Womens Health Res. (2025) 2025(9):165–8. doi: 10.3969/j.issn.2096-0417.2025.09.041

  • CrossRef
  • Google Scholar
  • View reference in article

• 35

  GaoXJJiangYWXuKMaRLiWY. The impact of yoga intervention on the physical and mental health of postpartum women. J Henan Norm Univ Nat Sci Ed. (2021) 49:113–20. doi: 10.16366/j.cnki.1000-2367.2021.06.016

  • CrossRef
  • Google Scholar
  • View reference in article

• 36

  KocaHUKucukkelepceDSKucukkelepceOKocyigitA. Effectiveness of laughter yoga in reducing postpartum depression symptoms: A randomized controlled study. Explore (NY). (2025) 21:103195. doi: 10.1016/j.explore.2025.103195

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 37

  LiLGaoL. Effects of yoga combined with psychological counseling on mental health, body shape, and pelvic floor function recovery in primiparas with postpartum depression. China J Health Psychol. (2019) 27:118–23. doi: 10.13342/j.cnki.cjhp.2019.01.035

  • CrossRef
  • Google Scholar
  • View reference in article

• 38

  WangLH. Effects of cognitive-behavioral intervention based on the medical-nursing collaborative model combined with group mindfulness yoga training on depressive symptoms and sleep quality in primiparas with postpartum depression. J Chin Foreign Med Pharm Res. (2024) 3:48–50. doi: 10.3969/j.issn.2096-6229.2024.05.017

  • CrossRef
  • Google Scholar
  • View reference in article

• 39

  Da CostaDLowensteynIAbrahamowiczMIonescu-IttuRDritsaMRippenNet al. A randomized clinical trial of exercise to alleviate postpartum depressed mood. J Psychosom Obstet Gynaecol. (2009) 30:191–200. doi: 10.1080/01674820903212136

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 40

  DaleyAJBlameyRVJollyKRoalfeAKTurnerKMColemanSet al. A pragmatic randomized controlled trial to evaluate the effectiveness of a facilitated exercise intervention as a treatment for postnatal depression: The PAMPeRS trial. Psychol Med. (2015) 45:2413–25. doi: 10.1017/S0033291715000409

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 41

  ForsythJBoathEHenshawCBrownH. Exercise as an adjunct treatment for postpartum depression for women living in an inner city: A pilot study. Health Care Women Int. (2017) 38:635–9. doi: 10.1080/07399332.2017.1295049

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 42

  HarunaMWatanabeEMatsuzakiMOtaEShiraishiMMurayamaRet al. The effects of an exercise program on health-related quality of life in postpartum mothers: A randomized controlled trial. Health. (2013) 5:432–9. doi: 10.4236/health.2013.53058

  • CrossRef
  • Google Scholar

• 43

  MeiDY. The effect of moderate-intensity aerobic exercise on postpartum depression in advanced maternal age. J Shandong Med Coll. (2024) 46:93–5. doi: 10.3969/j.issn.1674-0947.2024.01.043

  • CrossRef
  • Google Scholar

• 44

  NormanESherburnMOsborneRHGaleaMP. An exercise and education program improves well-being of new mothers: A randomized controlled trial. Phys Ther. (2010) 90:348–55. doi: 10.2522/ptj.20090139

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 45

  TeychenneMAbbottGStephensLDCostiganSAJacksonM. Mums on the Move: A pilot randomised controlled trial of a home-based physical activity intervention for mothers at risk of postnatal depression. Midwifery. (2021) 93:102898. doi: 10.1016/j.midw.2020.102898

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 46

  ThiruppathiAPrasanaBMastaniahEVamsidharNHimabinduP. A structured physical activity and health care education beats postpartum depression for primipara mothers: A pilot randomized controlled trial. Int J Physiother. (2014) 1(3):144–51. doi: 10.15621/ijphy/2014/v1i3/53468

  • CrossRef
  • Google Scholar

• 47

  WangMYZhaoY. Analysis of the impact of moderate-intensity aerobic exercise on patients with mild to moderate postpartum depression. J Today Health. (2025) 2025(17):172–4. doi: 10.3969/j.issn.1671-5160.2025.17.062

  • CrossRef
  • Google Scholar

• 48

  YangCLChenCH. Effectiveness of aerobic gymnastic exercise on stress, fatigue, and sleep quality during postpartum: A pilot randomized controlled trial. Int J Nurs Stud. (2018) 77:1–7. doi: 10.1016/j.ijnurstu.2017.09.009

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 49

  ÖzcanSEryilmazG. Can postpartum depression be prevented with care provided to primiparae using Levine's Conservation Model? A single-blind randomized controlled trial. Curr Psychol. (2024) 43:25973–87. doi: 10.1007/s12144-024-06271-3

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 50

  ÖzkanSAKucukkelepceDSKorkmazBYılmazGBozkurtMA. The effectiveness of an exercise intervention in reducing the severity of postpartum depression: A randomized controlled trial. Perspect Psychiatr Care. (2020) 56:844–50. doi: 10.1111/ppc.12500

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 51

  PascoeMCBauerIE. A systematic review of randomised control trials on the effects of yoga on stress measures and mood. J Psychiatr Res. (2015) 68:270–82. doi: 10.1016/j.jpsychires.2015.07.013

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 52

  HongALouisACummingsKFrazierCHwangCRuhlandJ. Scoping review of trauma-informed yoga for survivors of sexual, domestic, and interpersonal violence and development of the WELLNESS mnemonic to characterize outcomes. J Integr Complement Med. (2025) 31:635–49. doi: 10.1089/jicm.2023.0582

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 53

  MazzioAKMendozaNBrownMLThompsonKWalterKH. Yoga as a complementary approach to healing for adult victims and survivors of interpersonal violence. Complement Ther Clin Pract. (2021) 44:101431. doi: 10.1016/j.ctcp.2021.101431

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 54

  BeanCLesserI. Increasing health equity for postpartum women through physical activity. J Phys Act Health. (2024) 21:417–9. doi: 10.1123/jpah.2024-0077

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

Keywords

exercise, network meta-analysis, perinatal mental health, postpartum care, postpartum depressive symptoms

Citation

Liu S, Tan Y and Zang L (2026) Exercise-based interventions for postpartum depressive symptoms: a systematic review and network meta-analysis of exercise modality and weekly volume. Front. Psychiatry 17:1921342. doi: 10.3389/fpsyt.2026.1921342

Received

27 June 2026

Revised

10 August 2026

Accepted

13 August 2026

Published

30 September 2026

Volume

17 - 2026

Updates

Copyright

© 2026 Liu, Tan and Zang.

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Liuhong Zang, zlh751225@126.com

Disclaimer

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.