Associations of physical activity during pregnancy with antenatal and postpartum depression: a systematic review and dose-response meta-analysis Authors: Wei Zhang, Ye Yang, Yidi Fu, Siqi Li, Jianfei Chen, Yujia Chen, Jiarun Wang, Enjing Li, Xiaoli Chen, Zhijie Zou Published: 2026 ABSTRACT AbstractObjectiveTo explore the dose-response associations of physical activity (PA) during pregnancy with antenatal depression (AD) and postpartum depression (PPD).MethodThirteen electronic databases were searched from inception to April 8, 2026 and additionally screened references manually. Interventional and longitudinal observational studies in pregnant populations were eligible if they reported dose-response associations for PA during pregnancy in relation to AD or PPD; cross-sectional studies were excluded. Linear and nonlinear dose-response meta-analyses were conducted.ResultsTwenty-three studies from 10 countries involving 122,847 participants were included, comprising 5 randomized clinical trials and 18 cohort studies. In the linear dose–response meta-analysis, each 300 MET-min/wk increase in PA was associated with a 6% lower risk of AD (RR, 0.94; 95% CI, 0.90–0.97), with substantial heterogeneity (I2 = 70.6%, P < 0.001). AD also showed a significant nonlinear dose-response relationship, with the lowest risk observed at approximately 1000–2000 MET-min/wk and no evident protective association above approximately 4000 MET-min/wk. PA during pregnancy was associated with a lower risk of depression in late pregnancy, whereas the overall associations with mid-pregnancy depression and PPD were not statistically significant.ConclusionsPA during pregnancy was associated with a lower risk of AD and showed a significant nonlinear dose-response relationship, with the lowest risk observed at approximately 1000 to 2000 MET-min/wk, whereas no significant overall association was observed with PPD.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12884-026-09426-1.Keywords: Physical activity; Antenatal depression, Postpartum depression; Perinatal depression; Dose-response meta-analysis Introduction The perinatal period is a vulnerable stage for mental health problems because of substantial physiological, psychological, and social changes [1]. Perinatal depression, including antenatal depression (AD) and postpartum depression (PPD), is among the most common mental health disorders in women during pregnancy and after childbirth. In 2017, the WHO identified depression as the leading cause of disability worldwide [2]. Approximately 13% of women are affected by depression during the perinatal period [3, 4]. Untreated perinatal depression is associated with adverse maternal and infant outcomes [5], including prematurity [6], low birth weight, and maternal self-harm or suicide [7, 8]. In the general population, physical activity (PA) is recommended as a low-risk intensive treatment for depression compared to long-term pharmacologic treatment with antidepressants [9, 10]. This treatment may regulate mood by increasing the production of certain chemicals in the brain, such as endorphins, dopamine and serotonin [11]. PA in perinatal women is associated with physical, mental and other positive health benefits. For example, PA during the antenatal and postpartum periods may confer multiple health benefits, including healthier weight management [12, 13], better sleep quality and duration [14], improved psychosocial well-being [15], enhanced cardiovascular fitness [16], and reduced urinary stress incontinence [17]. Previous systematic reviews and meta-analyses suggest that PA during pregnancy may help improve AD outcomes; however, the observed effects appear to vary according to outcome type, activity intensity, dose [18], duration, and timing of exposure [19, 20]. Davenport et al. suggested that exercise during pregnancy may improve AD outcomes and provided early evidence for a possible dose-related association between exercise during pregnancy and AD, although its effect on PPD has not been consistently supported [22]. Subsequent studies further explored whether PA characteristics such as intensity, dose, and timing of exposure might also influence PPD risk, but the findings have remained inconsistent [18, 21]. Nevertheless, several important gaps remain in the current literature. First, most previous studies used categorical comparisons or subgroup analyses [21–24], which may suggest dose or threshold effects but cannot adequately define a continuous dose-response relationship under a unified exposure metric. Second, exposure windows have varied across studies, with some including perinatal or postpartum physical activity, making stage-specific effects difficult to compare. Third, Prior research has also been fragmented, focusing separately on antenatal depressive symptoms, overall intervention effects on perinatal depression, or postpartum depression prevention. Accordingly, updated integrated evidence is still lacking to evaluate both AD and subsequent PPD within the clearly defined exposure window of PA during pregnancy. A systematic reassessment is therefore needed to clarify both the association and its potential dose-response pattern. This study aimed to systematically evaluate the associations between PA during pregnancy with AD and PPD through a systematic review and dose-response meta-analysis, and to further quantify the potential dose-response relationships of activity volume, intensity, and duration with depressive outcomes, so as to inform prevention of perinatal depression and future psychological and clinical guidance on physical activity. Methods Protocol and registration The protocol for this systematic review and meta-analysis was registered in the International Prospective Register of Systematic Reviews (PROSPERO; CRD42023467516; Available from: https://www.crd.york.ac.uk/PROSPERO/#recordDetails). This study was designed, conducted, and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline [25]. Given that this study was based on previously published data and did not involve the collection of individual participant data, institutional review board approval and informed consent were not required. Patients and the public were not involved in the design, conduct, reporting, or dissemination of the research. The PRISMA checklists is provided in Supplementary Ⅱ. Data sources and search strategy Two researchers independently conducted a systematic search using a strategy that combined Medical Subject Headings and free-text terms in Pubmed, Embase, Web of Science, Scopus, EBSCO, CNKI, Wanfang Data, VIP, The Cochrane Library and JBI from database inception to April 8, 2026. The search strategy focused on terms related to physical activity, antenatal depression, and postpartum depression, including “physical activity”, “exercise”, “antenatal depression”, “prenatal depression”, “postpartum depression”. To identify additional eligible studies, we also performed supplementary searches in Google Scholar, Baidu Scholar, and Research Access; screened the reference lists of included studies and relevant review articles; and conducted forward citation tracking. The full search strategy is provided in Supplementary I(S1). Study selection and eligibility criteria Two researchers (WZ and YY) independently screened studies for eligibility, and disagreements were resolved through discussion with a third researcher. Our inclusion criteria were as follows: (1) study population: pregnant women, with no restrictions on age, parity, singleton pregnancy status, or history of anxiety or depression; (2) exposure: PA during pregnancy, including leisure-time, occupational, transportation, and household PA; (3) outcome: antenatal or postpartum depressive symptoms or depression, assessed by clinical diagnosis, database records, or validated instruments; (4) study design: cohort or interventional studies; and (5) available data: extractable or convertible effect estimates, such as odds ratios (ORs), risk ratios (RRs), or hazard ratios (HRs) with 95% CIs, or sufficient raw data to calculate effect estimates. For cohort studies, the comparator was the group with the lowest level of PA, an inactive group, or another reference category as defined in the original study. For interventional studies, the comparator was the control group defined by the original study. The exclusion criteria were as follows: (1) reviews, cross-sectional studies, case reports, conference abstracts, animal studies, and other nonhuman studies; (2) studies in which the exposure was not PA during pregnancy or in which sedentary behavior or related measures were the primary exposure; (3) interventional studies in which the independent effect of PA could not be isolated; (4) studies without full-text availability or sufficient data to estimate effect sizes; (5) studies explores the effects of AD or PPD on PA; (6) duplicated publications or studies based on overlapping data sources, in which case only the most informative report was retained. Cross-sectional studies were excluded because exposure and outcome are typically assessed concurrently, limiting temporal inference and increasing susceptibility to reverse causation. Physical activity exposure definition and harmonization PA was the primary exposure in this study. According to the original reports, PA included recreational, household, occupational, and transportation activities, and its dose was typically determined by duration, frequency, and intensity. To improve comparability across studies, PA dose was harmonized as metabolic equivalent task minutes per week (MET-min/wk). For studies using internationally standardized instruments, such as the International Physical Activity Questionnaire (IPAQ) or the Pregnancy Physical Activity Questionnaire (PPAQ), intensity values were assigned according to the original instrument. For studies not using standardized instruments, intensity values were assigned according to prespecified rules informed by the 2024 Adult Compendium of Physical Activities [26]. Walking or light-intensity PA, moderate-intensity PA, moderate-to-vigorous PA, and vigorous-intensity PA were assigned values of 3.3, 4, 6, and 8 METs, respectively [27]. Data harmonization was performed on the basis of the original study reports, supplementary data provided by authors, and available validation information; detailed harmonization methods are provided in Supplementary Ⅰ (S5) [10]. According to the IPAQ scoring protocol, PA levels were further categorized as low (< 600 MET-min/wk), intermediate (600–1500 MET-min/wk), high (1500–3000 MET-min/wk), and very high (> 3000 MET-min/wk) [28]. Data extraction We used a standardized form for data extraction. Extracted study characteristics included first author, publication year, study design, country, sample size, number of cases, mean gestational age, follow-up time, type of physical activity, exposure assessment tool, outcome type and assessment tool, adjusted confounders, and risk-of-bias assessment results. The main characteristics of the included studies are presented in Table 1. Table 1Characteristics of the Included StudiesAuthors, YearDesignCountry, ConditionsTotal SampleTotal CasesMean ageFollow-up timeType of PAExposure measuring toolOutcomesOutcome measurement toolRisk of bias (JBI)Adjusted confounding factorsFeng YH et al 2018 [49]CohortChina, Developing3579163929.142017.7-2018.7NAIPAQ-SADEPDS10Age, household registration, education level, occupation, annual household income, smoking, alcohol consumption and unplanned pregnancyGan YX et al 2018 [50]CohortChina, Developing258029328.82013–2015NAIPAQ-SPPDEPDS10Age, ethnicity, level of education, BMI, economic status, birth time, health status and sleep quality at 16 weeks gestation, smoking, alcohol consumption, perceived social support, vaginal bleeding, and history of asthmaRuan LL et al 2019 [48]CohortChina, Developing1615 / 1589 / 1570 (early/mid/late pregnancy)585 / 380 / 399 (early/mid/late pregnancy)27.892016.7-2018.7RecreationalSelf-designed questionnaire questionsADSDS10Age, BMI before pregnancy, education, marital status, smoking before pregnancy, drinking before pregnancy, family care in the last 3 months, pregnancy reaction and pregnancy events, pregnancy numberSi JH et al 2020 [51]CohortChina, Developing465412137.152016.1-2019.1NASelf-designed questionnaire questionsPPDEPDS10/Wu SS et al 2020 [47]CohortChina, Developing8132177229.012017.7-2018.11NAIPAQ-SADEPDS10/Wu SS et al 2020 [46]CohortChina, Developing5555139328.962017.7-2018.11NAIPAQ-SADEPDS9/Yue HX et al 2020 [45]CohortChina, Developing245416728.842017.7-2018.11NAIPAQ-SADEPDS10Age, educational level, occupation, annual household income and pre-pregnancy BMIZhong YX et al 2023 [44]CohortChina, Developing5488 / 2610 / 1745 (early/mid/late pregnancy)2042 / 655 / 445 (early/mid/late pregnancy)28.382020.9-2022.6NAPARS-3ADSDS9/Zuo HX et al 2019 [43]CohortChina, Developing4788429.92016.10-2017.12NASelf-designed questionnaire questionsADSDS9/Ana Y et al 2021 [52]CohortIndia, Developing140644924.652016.4—NAPALPPDEPDS11Age, socioeconomic status, education, EPDS score during pregnancy, social support, obesity, hypertension, gestational diabetesDemissie Z et al 2013 [53]CohortUSA, Developed6524330.692001.1-2005.12Occupational, Recreational and FamilySelf-designed questionnaire questionsPPDEPDS10Depression in early pregnancyGjestland K et al 2013 [42]CohortNorway, Developed348292531.32008.11-2010.3RecreationalSelf-designed questionnaire questionsADEPDS11Age, race, years of education, marital status, smoking, pre-pregnancy BMIPadmapriya N et al 2016 [41]CohortChina, Malaysia, India, Developing12478430.72009.6-2010.9NAIPAQ-SADEPDS, STAI11Age, race, education, employment during pregnancy, household income, smoking history, parity, and pre-pregnancy BMIStrøm M et al 2009 [54]CohortDenmark, Developed70,86615730.191996–2002RecreationalVoice recordings of telephone interviewsPPDMedical diagnosis based on ICD-10 and electronic medical records11Age, race, pre-pregnancy BMI, alcohol consumption, smoking, occupation, years of education, home ownership, marital status, social support, and previous history of depressionSzegda K et al 2018 [40]CohortUSA, Developed157925621.62006–2011NAPPAQ, Actigraph and accelerometerMild AD / Severe ADEPDS10Age, education level, living with partner or spouse, gestational order, mild depressionShakeel et al 2018 [55]CohortNorway, Developed64360302008–2011NAActivity recorderPPDEPDS10Race, depression at 28 weeks, pelvic zone syndrome at 28 weeks of pregnancyKolu P et al 2014 [39]CohortFinland, Developed399/29.712007–2009NA/AD/10Age, parity, education, and preconception BMIJennifer G et al 2012 [56]CohortUSA, Developed6026198927.862007–2008RecreationalSelf-designed questionnaire questionsPPDPRIME-MD9Age, race, marital status, years of education, whether you have children, smoking, drinking, first-time mothersSongoygard KM et al 2012 [57]RCTNorway, Developed8551230.582007.4-2009.6Recreational/PPDEPDS11/Daley A et al 2018 [38]RCTUK, Developed7895127.52009.4-2012.11Recreational, TrafficAccelerometer, pedometerADEPDS13Baseline EPDS score, years of education, Age, BMI, and marital statusColl CVN et al 2019 [37]RCTBrazil, Developing29434827.32014.8-2016.3Recreational/ADEPDS11/Navas et al 2021 [58]RCTSpain, Developed3201031.32014.11-2017.3RecreationalIPAQ-SPPDEPDS13/Perales M et al 2016 [36]RCTSpain, Developed1483232.72009.10-2013.1Recreational/ADCES-D13/1. Domain of PA: recreational, family, occupational, transportation, NA: No specific classification assigned, additional information is provided in Annex I) Characteristics of the Included Studies 1. Domain of PA: recreational, family, occupational, transportation, NA: No specific classification assigned, additional information is provided in Annex I) To further describe baseline characteristics and exposure definitions, we also extracted pre-pregnancy and pregnancy BMI, mean monthly household income, mean years of education, type (recreational, family, transportation, occupational) and period (early/middle/late pregnancy) of onset of PA, duration, frequency, intensity, physical activity dose, and the main study conclusions. These data are presented in Supplementary Table S2. For studies included in the dose-response meta-analysis, we additionally extracted exposure category definitions, dose levels, category-specific sample sizes, case numbers, and corresponding effect estimates. These data are presented in Supplementary Table S3. For studies reporting multiple exposure categories or effect estimates, we selected the estimate most comparable across studies for the conventional meta-analysis. For the dose-response meta-analysis, we extracted data from all relevant exposure categories and used the lowest physical activity category or inactive group defined in the original study as the reference category. Risk of bias assessment Risk of bias was assessed using the appropriate JBI critical appraisal tool for each study design. The assessment covered potential sources of bias, including selection, reporting, performance, detection, attrition, and “other” sources of bias. After literature screening, the two researchers (WZ and YY) independently performed the assessments. Disagreements were resolved through discussion and, when necessary, adjudication by the third researcher. The results of the risk-of-bias assessment are provided in Supplementary I(S4). Statistical analysis We first conducted a conventional meta-analysis of the effect estimates reported in the included studies. ORs, RRs, or HRs and their 95% CIs were extracted and pooled on a comparable scale. Because methodological heterogeneity was anticipated across studies with respect to study design, exposure measurement, outcome assessment, and confounder adjustment, a random-effects model was prespecified as the primary analytic approach. We also evaluated dose-response associations between total physical activity during pregnancy and antenatal and postpartum depression. For studies with at least 3 exposure categories and available data on dose, sample size, case numbers, and effect estimates, linear dose-response meta-analyses were conducted using the generalized least squares trend method proposed by Greenland and Longnecker [29, 30]. Study-specific coefficients were then pooled using restricted maximum likelihood [31]. Nonlinear dose-response associations were modeled using restricted cubic splines with three knots place at the 10th, 50th, and 90th percentiles [32]. If the statistical model todid not converge for these knots, the upper knot was incrementally increased until convergence was achieved. Linear and nonlinear associations were assessed using Wald test; a P value for nonlinearity less than 0.05 was considered evidence of a nonlinear dose-response association. For open-ended exposure categories, representative dose values were approximated using the width of the adjacent category. The reference category was defined as the lowest physical activity category or inactive group reported in the original study. As supplementary analyses, we further examined dose-response associations of physical activity duration, frequency, and intensity with antenatal and postpartum depression and conducted gestational period–specific analyses for antenatal depression. These results and figures are provided in Supplementary S8. Heterogeneity between studies was assessed using the Cochran ‘s Q-test and the I2 statistic, with values of 25%-50%, 50%-75%, and > 75% interpreted as low, moderate, and high heterogeneity, respectively [33]. Sensitivity analyses included leave-one-out analyses, excludion of studies at high risk of bias, and, when feasible, exclusion of studies using device-based exposure measurements. Prespecified subgroup analyses and meta-regression were performed to explore potential sources of heterogeneity, including study design, geographic region, exposure assessment tool, and outcome assessment tool. For analyses with asymmetric funnel plots, trim-and-fill analyses were performed when appropriate. Publication bias was assessed, when the number of included studies was sufficient, using the Egger ‘s regression asymmetry test [34] and Begg ‘s [35] rank correlation test. All analyses were performed in RStudio. A 2-sided P value less than 0.05 was considered statistically significant. Results Literature search A PRISMA diagram of the search process, including reasons for exclusion, is shown in Fig. 1. A total of 23 articles from 10 countries were included, comprising 5 randomized clinical trials and 18 cohort studies, with a total sample size of 122,847 participants. The systematic literature search identified 20 studies, and 3 additional studies were included through manual searching. The included studies primarily reported associations of PA during pregnancy with AD (n = 14) and PPD (n = 11). Study characteristics are presented in Table 1, and additional details are provided in Supplementary I(S2). Flow diagram of studies selected for the present study AD Across the included studies on AD, the overall pattern of evidence suggested that higher levels of physical activity during pregnancy were generally associated with a lower risk of AD or fewer depressive symptoms, although the magnitude and statistical significance of the associations varied across studies. This pattern was observed in both randomized trials and prospective cohort studies, but the findings were not entirely uniform. In particular, variability was noted according to the timing of depression assessment during pregnancy, the type and domain of physical activity, and the methods used to assess both exposure and outcome. Evidence from studies focusing on specific gestational periods suggested that the association may differ across pregnancy stages, with findings appearing more consistent for overall AD and late-pregnancy depression than for early- or mid-pregnancy depression. Nonlinear dose-respond meta-analysis Fourteen studies were included in the dose-response meta-analysis of PA during pregnancy and AD [36–49]. Of these, 14 studies reported overall AD, 4 reported early-pregnancy depression, 5 reported middle-pregnancy depression, and 7 reported late-pregnancy depression (Table 1). In the overall nonlinear dose-response meta-analysis, each 300 MET-min/w increment in PA during pregnancy was associated with a pooled RR for AD of 0.94 (95% CI, 0.90–0.97; I² = 70.6%; Cochran Q = 44.22; P < 0.001 for heterogeneity), corresponding to a 6% lower risk of AD (Fig. 2a). Fig2a. Forest plot of PA in pregnancy and AD; Fig2b. Forest plot of PA in pregnancy and PPD; Fig2c. Dose-response analysis of metabolic equivalent levels of PA in pregnancy; Fig2d. Dose-response analysis of metabolic equivalent levels of PA in pregnancy and risk of PPD In pregnancy-period–specific analyses, each 300 MET-min/w increment in PA during pregnancy was associated with pooled RRs of 0.93(95%CI 0.86-1.00, I2 = 73.1%) for depression during early pregnancy, 0.95(95%CI 0.89–1.02, I2 = 78.1%) for depression during middle pregnancy, and 0.91(95%CI 0.84–0.99, I2 = 65.3%) for depression during late pregnancy. These findings suggested an inverse association between PA during pregnancy and depression during late pregnancy, a borderline association for depression during early pregnancy, and no statistically significant association for depression during middle pregnancy (Supplementary Ⅰ, S8). Subgroup analyses and meta-regression Subgroup analyses (Supplementary I, S6) showed significant between-subgroup heterogeneity for study design (P = 0.019), type of PA (P = 0.026), timing of PA assessment during pregnancy (P = 0.005), restricted family history (P = 0.009), restricted single birth (P = 0.041), adjusted for race (P = 0.001) and adjusted for social support (P = 0.009). In contrast, no significant between-subgroup heterogeneity was observed for region, national development status, gestational age, exposure assessment method, outcome assessment method, BMI, years of education, household income, or most other restriction or adjustment factors. In meta-regression analyses (Supplementary I, S7), timing of PA assessment during pregnancy (β = -0.025, 95% CI [-0.048, -0.001]; P = 0.041) and adjustment for number of pregnancies/births (β = -0.070, 95% CI [-0.127, -0.013]; P = 0.020) were negatively correlated with the pooled effect estimate for AD, whereas restricted family history (β = 0.597, 95% CI [0.031, 1.164]; P = 0.040) and adjusted social support (β = 0.597, 95% CI [0.031, 1.164]; P = 0.040) were positively correlated with the pooled effect estimate for AD. No statistically significant associations were observed for the remaining variables. Publication bias and sensitivity analysis Publication bias was assessed using Egger’s regression asymmetry test and Begg’ rank correlation test. Neither test was statistically significant in the funnel plot (both P > 0.05, Cochran’ Q = 44.22), suggesting no evidence of publication bias (Supplementary I, S9). In leave-one-out sensitivity analyses, sequential exclusion of individual studies did not materially alter the pooled effect estimate, and the findings were consistent with the primary analysis (Supplementary I, S10). Dose response analysis The dose-response curve (Fig. 2c) showed a significant nonlinear association between PA during pregnancy and the risk of AD (Wald test: P nonlinear < 0.001). As PA increased during pregnancy, the risk of AD initially decreased and then increased, reaching its lowest level at approximately 1,000 to 2,000 MET-min/w. Thereafter, the risk of AD gradually increased with higher PA levels; when PA exceeded approximately 4,000 MET-min/w, the estimated RR exceeded 1, suggesting that the protective association of PA was no longer apparent. Because of the limited number of available studies, period-specific dose-response analyses were performed only for depression in mid-pregnancy and late pregnancy. Dose-response curve showed a significant nonlinear association between PA during pregnancy and depression in mid-pregnancy (Wald test: P nonlinear < 0.001). The risk of depression in mid-pregnancy was lowest at approximately 1,000 MET-min/w, whereas the point estimate for RR exceeded 1 when PA exceeded approximately 4,000 MET-min/w. Despite this nonlinear dose-response pattern, the overall association between PA during pregnancy and the risk of depression in mid-pregnancy was not statistically significant (P > 0.05). For depression in late pregnancy, no significant evidence of nonlinearity was observed (Wald test: P nonlinear >0.05), suggesting that the dose-response pattern could be interpreted as linear. Overall, higher levels of PA during pregnancy were associated with a lower risk of depression in late pregnancy. In Fig. 2c and d, the solid line represents the pooled relative risk estimate, and the dashed lines represent the 95% confidence intervals. The x-axis indicates total physical activity during pregnancy (MET-min/w), and the y-axis indicates relative risk. PPD For postpartum depression, the findings of the included studies were more heterogeneous than those for AD. Several randomized trials and cohort studies suggested that physical activity during pregnancy may be associated with a lower risk of postpartum depressive symptoms or PPD, whereas other studies reported little or no clear association. Differences in study design, timing and type of physical activity exposure, outcome definition, and postpartum assessment period may have contributed to this inconsistency. Overall, compared with the evidence for AD, the direction and strength of the association between physical activity during pregnancy and subsequent PPD appeared less consistent across studies. Nonlinear dose-respond meta-analysis Eleven studies were included in the dose-response meta-analysis of PA during pregnancy and PPD [37, 38, 50–58] (Table 1). Overall, each 300 MET-min/w increment in PA was associated with a pooled RR for PPD of 0.96 (95% CI, 0.87–1.07; I² = 77.7%; Cochran Q = 44.92; P < 0.001) in the nonlinear analysis (Fig. 2b), indicating no statistically significant association between PA during pregnancy and PPD. Subgroup analyses and meta-regression Subgroup analyses (Supplementary I, S6) showed significant between-subgroup heterogeneity according to study design (P = 0.011), adjustment of effect estimates (P = 0.006), restricted single birth (P = 0.006), restricted complications (P = 0.006) and restricted serious illness (P = 0.026). No significant between-subgroup heterogeneity was observed for type of PA, region, national development status, gestational age, timing of PA during pregnancy, exposure assessment method, outcome assessment method, body mass index, years of education, household income, or most other restriction or adjustment factors. Meta-regression analyses (Supplementary I, S7) showed that none of the study-level characteristics was significantly associated with the pooled effect estimate for PPD. Gestational age was not significantly associated with the pooled effect estimate for PPD (β = -0.087, 95% CI [-0.199, 0.026]; P = 0.115), and no statistically significant associations were observed for the remaining variables. Publication bias and sensitivity analysis Publication bias was assessed using Egger’s regression asymmetry test and Begg ' rank correlation test, and the results showed possible publication bias in the funnel plot (P<0.05, Cochran’ Q = 44.92, Supplementary I, S9). However, trim-and-fill analysis did not materially alter the pooled effect estimate after adjustment. Leave-one-out sensitivity analyses likewise yielded findings consistent with the primary analysis (Supplementary I, S10). Dose response analysis The dose-response curve (Fig. 2d) showed no significant evidence of a nonlinear association between PA during pregnancy and the risk of PPD (Wald test: P nonlinear>0.05), suggesting no clear nonlinear dose-response pattern. However, the estimated curve remained close to an RR of 1.0, with wide confidence intervals, indicating that the overall association between PA during pregnancy and the risk of PPD was not statistically significant (P>0.05). Discussion In this dose-response meta-analysis, PA during pregnancy was associated with a lower risk of AD, with evidence of a significant nonlinear dose-response relationship. Specifically, AD risk decreased as PA increased, reached its lowest level at approximately 1000–2000 MET-min/wk, and then increased at higher activity levels. When PA exceeded approximately 4000 MET-min/wk, the apparent protective association was no longer evident. Period-specific analyses further showed that PA during pregnancy was associated with a lower risk of depression in late pregnancy, whereas the overall correlations with mid-pregnancy depression and PPD were not statistically significant. Most previous meta-analyses based on categorical comparisons have shown that, compared with lower levels of PA during pregnancy, higher levels of PA are associated with a lower risk of depression among perinatal women [59], a finding that is consistent with the overall direction of our results. However, these studies generally relied on relatively crude exposure categories and therefore could compare only overall risk differences across activity levels, without further characterizing the dose-response relationship [9]. In contrast, by modeling PA on a continuous dose scale, our study quantitatively evaluated the association between PA during pregnancy and depression risk and suggests that this association is more likely to be nonlinear, with a potentially optimal range of activity, rather than a simple pattern in which progressively higher activity is associated with progressively lower risk. Although our findings were broadly consistent with previous evidence in direction, the specific association pattern appeared to be more complex. This may in part reflect differences in PA measurement and categorization, PA type, and the timing of outcome assessment. Previous studies have suggested that the association between different types of PA and depression may vary across stages of the perinatal period [19]. For example, sedentary behavior [60–63], , occupational activity [20], and household activity [64] may be associated with a higher risk of depression, whereas leisure-time PA may be protective. Accordingly, failure to adequately distinguish the type, intensity, and context of PA in previous studies may have compromised the comparability of findings across studies. In addition, our findings suggest that at higher levels of PA, the protective association with AD may be attenuated or no longer evident. This pattern may be related to the greater physical burden associated with higher activity levels, including fatigue, heightened stress responses, and sleep problems, which could partly offset the potential mental health benefits of PA [65–67]. Residual confounding factors (environment [68], noise pollution [69], neighborhood deprivation [70], etc.) may also have influenced the observed association. From a clinical and public health perspective, our findings may also be considered in relation to current physical activity guidelines during pregnancy, which generally recommend at least 150 min of moderate-intensity physical activity per week. In our study, the lowest observed risk of antenatal depression was identified at approximately 1000–2000 MET-min/wk, which is higher than the minimum guideline target. However, the observed reduction in risk associated with each 300 MET-min/wk increment suggests that even smaller, more achievable increases in physical activity may still be meaningful. Therefore, our findings do not necessarily imply that pregnant women need to reach the apparent lowest-risk range immediately, but rather support the idea that gradual increases in physical activity during pregnancy may still confer mental health benefits. From a mechanistic perspective, the association between PA during pregnancy and a lower risk of AD may be explained by both biologic and behavioral pathways [71]. Previous studies suggest that exercise may benefit mental health during pregnancy by mitigating stress responses, improving sleep quality, enhancing self-efficacy, and promoting positive affect, which may partly support our finding that PA during pregnancy was associated with a lower risk of AD [72–74]. At the same time, we observed a nonlinear pattern, suggesting that these benefits may not increase indefinitely with greater activity. At higher levels of PA, the additional physical burden, including accumulated fatigue, heightened stress responses, and sleep disturbance, may partly offset its potential mental health benefits [74]. In contrast, we did not observe a significant overall association between PA during pregnancy and PPD, suggesting that the effect of antenatal PA on PPD may be more indirect [37]. One possible explanation is that, although exercise may influence mood through neuroendocrine pathways, including modulation of beta-endorphin levels, these potential effects during the postpartum period may be more readily influenced by postpartum-specific factors, such as sleep disruption, nighttime infant care demands, breastfeeding-related stress, and social support, making a stable overall association more difficult to detect [75–77]. A major strength of this study is that it evaluated the dose-response association between PA during pregnancy and depression risk on a continuous scale. Given that the intensity, frequency, and duration of PA may jointly influence its association with depression, we assessed risk changes more systematically across the exposure continuum and complemented the analysis with both linear and nonlinear dose-response models. In addition, we conducted separate dose-response analyses for AD and PPD and used subgroup analyses and meta-regression to explore potential sources of between-study heterogeneity. Overall, this analytic framework not only allowed a more precise characterization of the association between PA during pregnancy and depression risk but also may help inform future refinement of recommendations regarding the dose, intensity, and duration of PA in pregnancy in the context of existing physical activity guidelines. This study has several limitations. First, most included studies relied on self-reported measures of PA exposure and depressive outcomes, which may have introduced recall and social desirability bias. Only a small number of articles used objective outcome assessment methods, such as clinically diagnosed depression. Second, due to the extremely limited amount of data with very high levels of PA in the original studies, the risk of bias in the data included in the analyses was still unavoidable, although we manipulated the data through harmonisation techniques. Third, different types of PA may have distinct associations with depression, but evidence on household, transportation, and occupational PA during pregnancy remains limited, precluding dose-response analyses by PA domain. Finally, although we initially planned to include anxiety as an outcome, this was not feasible because of the limited number of eligible studies. Further high-quality studies are needed to clarify the dose-response associations, particularly potential nonlinear associations, between different types of PA during pregnancy and perinatal depression and anxiety. Conclusion PA during pregnancy was associated with a lower risk of AD and showed a significant nonlinear dose-response relationship, suggesting the presence of a relatively favorable activity range. Specifically, the lowest risk was observed at approximately 1000 to 2000 MET-min/wk, whereas the apparent protective association was no longer evident above approximately 4000 MET-min/wk. By contrast, the overall association between PA during pregnancy and PPD was not statistically significant. These findings suggest that the association of PA during pregnancy may differ across perinatal depressive outcomes and may help inform further refinement of recommendations regarding PA during pregnancy from a dose-based perspective. Supplementary Information Supplementary Material 1. Supplementary Material 2. Acknowledgements This study is being conducted by the School of Nursing, Wuhan University. The authors are grateful to the Research Centre for Maternal and Child Health, School of Nursing, Wuhan University for their help with our study. and the study was conducted with funding from Professor Xiaoli Chen’s National Social Science Foundation of China (Project No. 21BSH073). Authors’ contributions WZ and YY had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. YDF made substantial contributions to the revised manuscript, including major revision of the manuscript and the point-by-point responses to the reviewers’ comments. WZ, YY and YDF contributed equally and served as co-first authors. Conceptualization and design: WZ, YYData curation: WZ, YYFormal analysis: AllFunding acquisition: XLCMethodology: WZ, YYProject administration: EJL, ZJZSupervision: EJL, XLCValidation: WZVisualization: WZ, SQL, YJCWriting – original draft: WZ, YYWriting – review and editing: AllMajor revision of the manuscript and point-by-point responses to reviewers’ comments: YDF. Funding Our research is supported by the National Social Science Foundation of China (Project No. 21BSH073) to the author Xiaoli Chen. Data availability No participant data was collected for this manuscript. All data will be available within the manuscript and has been collected by the respective authors of papers included in the meta-analysis. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Wei Zhang, Ye Yang and Yidi Fu contributed equally to this work. Contributor Information Enjing Li, Email: leeej@ccnu.edu.cn. Xiaoli Chen, Email: chenxl72@whu.edu.cn. Zhijie Zou, Email: zouzhijie@whu.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Data Availability Statement No participant data was collected for this manuscript. All data will be available within the manuscript and has been collected by the respective authors of papers included in the meta-analysis.