# paper.pdf ## Metadata - PDFFormatVersion=1.4 - Language=en-US - IsLinearized=false - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - CreationDate=D:20260525193253+02'00' - Creator=Adobe InDesign 18.3 (Windows) - Custom.CrossMarkDomains[1]=springer.com - Custom.CrossMarkDomains[2]=springerlink.com - Custom.CrossmarkDomainExclusive=true - Custom.CrossmarkMajorVersionDate=2010-04-23 - Custom.doi=10.1007/s00737-026-01723-0 - ModDate=D:20260703082618+02'00' - Producer=Adobe PDF Library 17.0; modified using iText® 5.5.13.4 ©2000-2024 iText Group NV (SPRINGER SBM; licensed version) - Trapped=False - crossmark:crossmarkdomains=springer.comspringerlink.com - pdfx:crossmarkdomains=springer.comspringerlink.com ## Contents ### Page 1 Introduction Antenatal depression (AD), also known as prenatal depres-sion, is a mental health condition estimated to affect 28.5%of pregnant people globally (Al-Abri et al.). Whilethere are many evidence-based treatment pathways avail-able, what remains largely unknown is the comparativeefficacy of these interventions, and this meta-analysis syn-thesizes such evidence. The importance of treating depres-sion in pregnancy is underscored by its adverse impact onboth maternal as well as offspring outcomes. These includeadverse birth outcomes, such as preterm delivery (Jardeet al.; Smith et al.; Yedid Sion et al.) andlow birthweight (Field; Jarde et al.) and adversepsychosocial offspring outcomes, like insecure or disorgan-ised attachment (Hayes et al.) and temperamental andsleep-related disturbances (Smith et al.). Indeed, theliterature also portrays AD as among the most prominentrisk factors in the development of postpartum depression(Leigh and Milgrom). Despite these outcomes, only20% of women experiencing AD receive the appropriateintervention (Vigod et al.), with factors such as lackof access, a long waitlist for treatment, and prior negative Received: 28 July 2025 / Accepted: 8 May 2026 © The Author(s) 2026 Extended author information available on the last page of the article # Abstract # Purpose Antenatal depression (AD) is a mental health condition estimated to affect close to 30% of pregnant people glob-ally. The importance of treating depression in pregnancy is underscored by its downstream impact on both maternal andinfant biopsychosocial outcomes, such as preterm birth, low birthweight, and postnatal depression. Despite these outcomes,only 20% of women experiencing AD receive appropriate and timely support and treatment. This meta-analysis is a com-prehensive synthesis of RCTs evaluating psychological and non-psychological interventions to treat antenatal depression,conducted as part of the HappyMums project. # Methods Studies were sourced from PubMed, Embase Medline and MIDIRS (via OVID), and were included if they were(a) randomized controlled trials, (b) with pregnant people with, or at risk for depression, (c) consisted of an intervention, and(d) published in English. To estimate efficacy, Standardized Mean Differences (SMDS) were calculated with 95% confidenceintervals using the post- intervention Ms and SDs of both the intervention and control groups. The I2 statistic was used as anindicator of variation between studies. Analyses were conducted using the software R Studio. # Results In total, 115 studies were included in the analysis. The pooled effect size of all interventions, compared with allpooled control arms, showed a clear therapeutic effect (SMD = 0.65, 95% CI 0.48; 0.83), although the between-study hetero-geneity variance was high (τ2 = 0.78, 95% CI 0.7;1.3). After accounting for outliers, the pooled effect size across 94 treatmentarms was still SMD = 0.5 (95% CI, 0.42, 0.56). No significant subgroup differences were found by intervention type andformat, or measure used. # Conclusion There are numerous interventions that, on average, are moderately effective in reducing depressive symptomol-ogy in pregnancy. Future research should be aimed at addressing how to personalise the choice of treatment and improvetreatment outcomes for the individual pregnant person. Keywords Antenatal depression · Randomised controlled trials · Interventions · Meta-analysis The evidence base of interventions to treat antenatal depression: ameta-analysis of randomized controlled trials Riddhi Laijawala1 · Kristi Priestley1 · Rebecca Bind1 · Katie Hazelgrove1 · Lavinia Rebecchini1 · Madeline Kirkpatrick1 ·Francesca Mancino1 · Samrina Sangha1 · Alessandra Biaggi2 · Malvika Godara3 · Claudia Buss3,4 · Sonja Entringer3,5 ·Anthony Woods1 · Paola Dazzan6 · Annamaria Cattaneo2,7 · Carmine M Pariante1 Archives of Women's Mental Health (2026) 29:103 [ ![http://crossmark.crossref.org/dialog/?doi=10.1007/s00737-026-01723-0&domain=pdf&date_stamp=2026-5-25]() ](http://crossmark.crossref.org/dialog/?doi=10.1007/s00737-026-01723-0&domain=pdf&date_stamp=2026-5-25) ![Springer Logo]() ### Page 2 experience with mental healthcare, posing as barriers toaccessing treatment (Kopelman et al.). Existing systematic reviews and meta-analyses havedemonstrated the efficacy of CBT (Shortis et al.),mindfulness-based interventions (Li et al.), lifestyleand exercise-based interventions (Zhang et al.) in thetreatment of antenatal depression. Indeed, with the advance-ment of digital health tools within the space of perinatalmental health, evidence has shown that digital therapeuticinterventions are also effective tools in managing depres-sive symptomology during pregnancy (Lau et al.).However, to our knowledge, no meta-analyses have syn-thesized the entire evidence base, encompassing psycho-logical, non-psychological, as well as digital interventionsin the treatment of antenatal depression using a commonmethodological approach and effect sizes calculation; there-fore, this was our primary aim. Furthermore, we includedan analysis of biomarkers changes during the interventions,in the smaller set of studies that have this data, to providea mechanistic insight into how these diverse interventionsmight work. While meta-analyses have been undertaken on specificintervention types, such as exercise (Zhang et al.) andpsychotherapy (Li et al.), no meta-analysis at the timeof data extraction had provided an overarching synthesis ofthe evidence base, considering all the available treatmentpathways. In the absence of any randomised controlled tri-als using a pharmacological intervention (Howland),we felt it was important to establish the range of psychologi-cal and non-psychological interventions that were effective,and then we followed-up this global investigation by under-taking subsequent subgroup analyses by intervention formatand delivery type. An increasing amount of literature highlights the roleof specific biomarkers, such as abnormal proinflammatorycytokines (for example, IL-6, IL-8, CRP, TNF alpha, IL-7,IL-10) (Silva-Fernandes et al.; Zuo et al.), anddysregulated HPA axis functioning (Carnevali and Buoli), in the development of antenatal depression. Giventhis evidence, we deemed it appropriate to consider theimpact that interventions have not only on depression, buton biological changes that might be causally related to thedepressive symptoms and/or involved in the mechanisms ofaction of the interventions. Prior syntheses of the literature (Zeng et al.) havefocused on non-pharmacological interventions in the treat-ment of antenatal depression, however, to date no meta-analysis has synthesized the overall evidence base assessingdepression, while also examining a range of biomarkersrelevant to the perinatal period. Numerous biomarkers rel-evant to major depressive disorder have also emerged asprominent in perinatal depression treatment response, suchas TNF alpha, IL-6, IL-10 (Osborne et al.) and alteredmetabolites such as amino acids and the tryptophan-kyn-urenine pathway (Kwok et al.). Therefore, we build onthis knowledge base by exploring the evidence base of inter-vention studies which also include biomarkers. We initiallyaimed to investigate inflammatory biomarkers due to preg-nancy being so clearly associated with the immune system.However, our review of the literature found varied mark-ers (not only inflammatory), such as cholesterol, cortisol,and glucose. Above all, we integrated clinical data alongwith biomarkers so that we could have a more personalizedunderstanding of treatment mechanisms, allowing for moretailored intervention strategies based on individual bio-markers’ profiles (Plant and Barton). Since pregnancyis a time of great biological changes, it is vital to understandhow effective interventions act from a biopsychosocial lens. To mitigate the impact of antenatal depression on mater-nal and offspring outcomes, it is vital to review the evidencebase of interventions. Therefore, as part of the HappyMumsinterdisciplinary research programme (funded by Horizonresearch and innovation) this meta-analysis was conductedto evaluate the effectiveness of all interventions for AD,both psychological and non-psychological (for example,lifestyle or alternative medicine), to gain an understandingof their comparative efficacy and the factors predicting theresponse. Overall, the effects size differences, albeit sta-tistically significant, were small, and indicative of smallchanges in clinical scores which could be of doubtful clini-cal significance, for example approximately 2.3 points onthe EPDS (Lautarescu et al.), 1.6 points on the PHQ-9(Srisurapanont et al.) and 4.4 points on the CES-D(Kliem et al.). We also simultaneously identify, when available, bio-markers that are influenced by the interventions, to providepossible mechanistic insight into the therapeutic action. Methods We followed the Preferred Reporting Items for SystematicReviews and Meta-Analyses (PRISMA 2020) guidelines.Embase, PsycINFO, Medline, and the Maternity and InfantCare Database (MIDIRS) were searched via Ovid and uti-lised to extract relevant studies. Search terms included acombination of the following keywords: (pregnan* or pre-natal or antenatal), AND (treatment or therapy or interven-tion) AND (depress*) to identify publications relevant to theresearch question. At this stage, a preferred study method-ology was not included in eligibility criteria; however, inthe subsequent screening we only selected randomised con-trolled trials. The search was conducted in successive wavesfrom the beginning of HappyMums (1st of November 2022) ### Page 3 until 30th of November 2024; although the review per sewas not registered at inception, we followed the overarch-ing aims of HappyMums as defined in the protocol (Biaggiet al.) as well as the objective of the Project Task3.5 (to conduct meta-analytic pooled estimates assessingthe response to these interventions over depressive symp-tomatology, as wells as possible associations between theclinical improvement and biological factors). All identifiedstudies were exported to Zotero Version 7.0.311, whereduplicates were removed prior to the screening of abstractsand titles. To determine study eligibility, all studies werescreened in a two-stage process: (1) A preliminary abstractand title screening to deem initial eligibility. In this, studieswere categorised into “Eligible”, “Ineligible”, and “Maybe”categories. For studies where the primary reviewer (R.L.)was uncertain about eligibility (i.e. the “Maybe” group),a senior postdoctoral member of the team (K.P.) reviewedthese studies and determined final eligibility. (2) In Stage2, a full text review of “Eligible” studies was completed toidentify the final studies for data extraction. At this stage,only randomized controlled trials (RCTs) were deemedeligible, as they are the ‘gold standard’ for effectivenessresearch (Hariton and Locascio). Studies were included in Stage 1 if they (a) Contained asample of pregnant women (b) Assessed depressive symp-tomology as an outcome measure, either through self-reportor clinician rated measures, (c) Delivered an interventionfor antenatal depression, (d) Had a control condition suchas treatment as usual or other active treatment and (e) Pub-lished in the English language. During Stage 2, RCT studymethodology was introduced as an additional criterion. Exclusion criteria were as follows (a) Studies not pub-lished in English (b) Dissertations, book chapters or confer-ence abstracts (c) Non-randomised controlled trials. Data extraction Data was extracted from papers meeting eligibility cri-teria during Stage 2 of selection. The following data wasextracted and entered in a table on an Excel file: Author,date of publication, country of delivery, gestational age atbaseline, mean age of participants at baseline, interventiondetails, control condition details, study methodology, num-ber of participants in each group, outcome measure used,outcome measure type (clinician or self-rated), Time 1 andTime 2 Means (M) and Standard Deviations (SD). Wherestudies consisted of more than two intervention arms, eachtreatment arm was compared to the arm pre-specified as acontrol group by study authors. In situations where nec-essary data for analysis was unavailable, correspondingauthors were contacted. If no necessary data was available,the study was then excluded at the final stage before statisti-cal analysis. Data extraction was performed by R.L., to limit variabil-ity in the calculation of effect sizes. However, the data werere-examined by doubled entry by another member of theteam (F.M.) before effect sizes were calculated. Moreover,for studies where the primary reviewer (R.L.) was uncertainabout eligibility (i.e. the “Maybe” group), a senior postdoc-toral member of the team (K.P.) reviewed these studies anddetermined final eligibility. Considering the classificationof studies into specific subgroups, R.L. initially completedthe classification, and M.K. reviewed these, helping confirmsubgroup characteristics. Therefore, at each stage, the datawas looked at by two independent researchers. Statistical analysis Effect size calculations To estimate efficacy, Standardized Mean Differences(SMDS), also known as Cohen’s d were calculated with95% confidence intervals using the post-intervention Msand SDs of both the intervention and control groups. Thismethod of estimating effect sizes was chosen based on lit-erature highlighting that within group (or pre-post) SMDscan be potentially contribute to biased outcomes and mightnot provide reliable information about intervention effects(Cuijpers et al.). All calculations were conducted on R, using pre-devel-oped scripts (Harrer et al.). To facilitate interpreta-tion, the following guidelines recommended by Cohen wereused: SMD 0.2 = Small, SMD = 0.5 Medium, SMD = 0.8Large (Cohen; Faraone). Publication bias Publication bias was assessed by examining funnel plots ofthe included studies, and Egger’s tests as well as Duval andTweedie’s trim and fill method (Duval and Tweedie;Harrer et al.)were utilised to examine plot asymme-try. Additionally, a risk of bias assessment was undertakenusing the RoB 2 tool ((RoB 2: A Revised Tool for AssessingRisk of Bias in Randomised Trials | The BMJ, 2019.). R.L.completed the risk of bias screening, with additional supportfrom S.S. Both used the guiding statement from the RoBtool for each domain in determining overall risk of bias ineach study. If disagreements arose, R.L. and S.S. had a dis-cussion to reach consensus and determine final risk of biasassignment. ### Page 4 Heterogeneity The I2 statistic was used as an indicator of variation betweenstudies. 25% heterogeneity was considered low, 50% mod-erate, and 75% was considered high (Melsen et al.). Results Characteristics of included studies After the removal of 950 studies (which were duplicates),2547 titles and abstracts were screened to deem initial eligi-bility. Out of 437 full text articles (deemed eligible at title/abstract screening), 328 met eligibility criteria. Finally, 115trials (125 arms overall) were included in the meta-analysisof interventions. The study selection process is summarisedin Fig.. 78 trials used standard antenatal care as a control group,7 trials utilised a placebo condition, 2 used a sham control,24 had an active control condition, and 4 trials did not spec-ify details about the control group. For the purpose of the analysis, intervention typeswere divided into (a) Psychological (68 studies), (b)Non psychological (43 studies), and (c) Combination (4studies (see below for details)). 76 interventions wereconducted in person, 27 were digitally/online based, 9were conducted through monitored self-administration,and 3 studies utilised a combination of these interventionformats. ### Page 5 Participant and regional characteristics Overall, the trials included a sample of 12,561 participants,6279 from treatment groups and 6282 from control groups.Accounting for missing values, the mean age of participantsacross all trials was 29.7 (SD 2.68), and the average gesta-tional age at baseline was 21.8 weeks (SD 5.7). Supplementary Material Table 1 summarises the geo-graphical and income regions where the various trials wereconducted. For consistency, The World by Income andRegion () of the world by income and geographicalregion were utilised. A similar number of studies were fromhigh income countries (The World by Income and Region,n = 59) and from low to upper-middle income countries(n = 56). More details are in Supplementary Material Table 1. Effectiveness of interventions Outliers were removed in an exploratory manner, onceresults were available on the substantial levels of heteroge-neity and on the presence of publication bias as identified bythe Egger’s test. This was done through the package ‘dem-etar’ on R studio (Harrer et al.), whereby a removalalgorithm searches for outlying studies from the initialmeta-analysis, identifies them (with the option to removeand recalculate results). Once these outliers were identi-fied, we subsequently removed them and then conductedthe Egger’s test again. When 31 outliers were removed, thepooled effect size from 94 treatment arms was still indica-tive of a therapeutic effect (SMD = 0.5, 95% CI 0.42; 0.56,p < 0.001). Between study heterogeneity variance improved,as estimated at τ2 = 0.05 (95% CI 0.02; 0.09), with I2 valueof 46.3% (95% CI 31.5%; 57.9%), and there was no lon-ger any publication bias (see also below). Upon removingoutliers, the 95% prediction interval was 0.06–0.92, thistime clearly indicating that future studies with these inter-ventions would yield therapeutic benefit in the treatment ofantenatal depression. Psychological interventions Interventions were classified as psychological if they pri-marily targeted emotional, cognitive, or interpersonal pro-cesses, and non-psychological if they focused on physical,nutritional, lifestyle, or sensory strategies. Psychological treatments included Mindfulness BasedInterventions (MBIs), Cognitive Behavioural Therapy(CBT), psychoeducation, Interpersonal Psychotherapy(IPT), behavioural management, problem solving therapy,mentalization, counselling, music therapy, psychosocialintervention, cognitive analytic therapy, emotional manage-ment programmes, and behavioural activation. Non psychological interventions The range of non-psychological intervention types waslarge. Trials utilised interventions such as aromatherapy,religious and spiritually based treatments, bright lighttherapy, transcranial direct current stimulation, transcranialmagnetic stimulation, yoga and exercise, Omega 3 supple-mentation, acupuncture, massage therapy, and childbirtheducation. Surprisingly, we could not find any randomisedcontrolled trial of a pharmaceutical agent in AD. Combination treatments Interventions were categorised into the combination groupif they integrated both psychological and non-psychologicalaspects into the interventions. Examples include interven-tions consisting of counselling sessions along with lifestyle-based couple training, and exercise coupled with cognitiveanalytic therapy. Measures used to assess depression Included studies used a range of self-report and clinician-rated measures, specific to perinatal depression or for thegeneral population. The most used measure was the Edin-burgh Postnatal Depression Scale (EPDS) (Cox et al.); while this measure uses the term postnatal, it hasindeed been validated for use during pregnancy (Berginket al.). Other self-report measures included the BeckDepression Inventory (BDI) (Beck et al.), Center forEpidemiological Studies-Depression Scale (CESD) (Rad-loff), and the Patient Health Questionnaire 9 Item(PHQ-9) (Kroenke et al.). Fewer studies reported ondepression assessed using clinician-rated measures, such asthe Hamilton Depression Rating Scale (HAM-D) (Hamilton)and the Montgomery-Åsberg Depression Rating Scale(MADRS) (Montgomery and Åsberg). Intervention format Along with varied intervention types, the format of deliveryalso varied between the trials. 76 trials were conducted in per-son, that is, in a physical setting, where the participants wereface to face with the facilitator. In person interventions wereeither conducted individually, or in a group. 27 digital/onlineinterventions were delivered through a variety of means, suchas smartphone apps, messaging services, Compact Discs(CDs), and via telephone. 9 studies used a self-administratedform of treatment, such as omega 3 supplementation, brightlight therapy, self-help treatment, lavender cream treatment,and aromatherapy. Finally, 3 studies used a combination for-mat, integrating online and in person formats. ### Page 6 Differences in efficacy by intervention type, formatand measure We undertook additional subgroup analyses on the studiespresented in Fig.. No significant differences were observedbetween the three groups, that is, psychological, non-psy-chological and combination treatments (p = 0.63). Furthersubgroup analyses were undertaken to assess numerical dif-ferences in efficacy based on intervention type. Both psychological and non-psychological interventionsshowed moderate efficacy (SMD = 0.46, 95% CI 0.4; 0.55p < 0.005 and SMD 0.53, 95% CI 0.41; 0.65, p < 0.0005).Finally, interventions that utilized a combination of the twoalso showed comparable efficacy (SMD 0.57 95% CI 0.08;1.05, p < 0.05). Another subgroup analysis was undertaken to identifywhether differences in efficacy arose as a result of interven-tion format, measure-; these differences were not significant(p = 0.88). As above, all intervention formats had similarnumerical efficacy. 85 arms utilised a self-report measure to assess depres-sive symptomology, and 9 arms utilised a clinician rated out-come measure. The difference was not significant (p = 0.64).although numerically the efficacy of interventions washigher for clinician-reported outcomes (SMD 0.58; 95% CI0.31; 0.85), as compared to self-reported outcomes (SMD0.48; 95% CI 0.42; 0.56). However, given that only 9 studiesreported on clinician-rated outcome, these results need to beinterpreted with caution. All effect sizes were comparable,and results of subgroup analyses are summarized in Table. Biomarker analysis Within the larger dataset, a small number of studies reportedon biological outcomes. Effect sizes were pooled to under-take a meta-analysis to determine the ability of interven-tions to differentially affect the post-intervention levels ofthe various biomarkers, i.e., we tested for post-interventiondifferences between the intervention and the control group.Biomarkers analysed included oestradiol, high-sensitivityC-reactive protein (hsCRP), interleukin (IL)−6, adiponec-tin, cortisol, and blood pressure. Overall, 7 studies wereincluded, and we acknowledge the physiological heteroge-neity of pooled biomarkers in our analysis. Results of themeta-analysis are displayed in Fig.. For the purpose of simplicity, positive effect sizes indi-cate that the differences in the post-intervention levels of thebiomarker favoured the intervention, i.e., suggested lowerstress levels in the intervention group (see SupplementaryTable 3). Together, the interventions affected the biomarkersin directions consistent with a “stress reduction” effect,suggesting that biological changes accompany the men-tal health improvement (SMD = 0.25, 95% CI 0.01; 0.48).Music therapy had the largest effect size (Cao et al.) onsystolic (SMD 1.38, 95% CI 0.83; 1.93) and diastolic bloodpressure (SMD 0.93, 95% CI 0.4; 1.46, as well as angioten-sin II (SMD 1.81, 95% CI 1.22; 2.4). Some studies howeverdid not show such an effect, for example, in the study by(Wilczynska et al.), the levels of cortisol were higherin the intervention group than in the control group. As these biomarkers were varied, as expected, levels ofheterogeneity were substantially high (I2 = 73%). However,we undertook another meta-analysis after accounting for 3outliers (Cao et al. (a), Cao et al. (b), and Wilczynska et al.),and the overall results remain unchanged (SMD = 0.17, 95%CI 0.01; 0.33). Results of this analysis are summarised inSupplementary Material Fig. 2. Publication bias Publication bias was assessed by visually inspecting funnelplots, administering Egger’s test to identify plot asymme-try, and using Duval and Tweedie’s trim and fill method(Duval and Tweedie) to assess bias. In the main anal-ysis including all studies, publication bias was identifiedusing all three methods. When outliers were removed, nobias was detected. Results of these tests are displayed in theSupplementary Material Figs. 3. As per the RoB V2 tool, the following domains of publi-cation bias were assessed: (a) Bias arising from the random-ization process, (b) Bias due to deviations from intendedinterventions, (c) Bias due to missing outcome data, (d) Biasin measurement of the outcome, (e) Bias in selection of thereported results. Figure summarizes the level of risk ofbias for each domain. Overall, studies had a low risk of bias,however the domain with the highest risk was domain (b). Discussion To our knowledge, this is the first meta-analysis synthesis-ing the entire evidence base of interventions in the treatmentof AD, along with a focus on biomarkers influenced by theinterventions. Results of our analysis found an overall mod-erate treatment effect (SMD = 0.65, p < 0.0001) of interven-tions when compared to control conditions over depressivesymptomology. Given the substantial levels of heterogene-ity (I2 = 88%), our secondary analysis, eliminating outliers,still found a clear treatment effect (SMD = 0.5, p < 0.0001),with lower levels of heterogeneity (I2 = 46.3%). Intervention types were varied, including, but not lim-ited to, psychotherapy, yoga, CBT, music therapy, mindful-ness, Omega-3 supplementation, and bright light therapy. ### Page 7 ### Page 8 Interestingly, we found no differences in efficacy betweenpsychological and non-psychological interventions. Previ-ous meta-analyses across specific intervention types haveshown differences in effect sizes, for example 0.41 for exer-cise (Zhang et al.), and 0.76 for mindfulness (Li etal.)), although these were calculated in different meta-analyses. Contrary to these findings, in our common meta-analysis we did see that psychological, non-psychologicaland combination interventions had comparable efficacy(SMD = 0.46, 0.53, and 0.57 respectively). While part ofthis can be explained by the common methodology for theestimation across all types of studies, we also postulate thatthis might be in part due to the range of risk factors in thedevelopment of antenatal depression, as seen from the Hap-pyMums clinical study (Priestley et al.), which includeprevious mental health history, lack of social support, andhistory of fertility issues. Different intervention types, likepeer support, CBT, and mindfulness might address theserisk factors during treatment, which might explain similarimprovements across intervention types. In terms of inter-vention format, digitally delivered interventions, such asiCBT, use the same principles as interventions deliveredin person, and therefore have the same therapeutic mecha-nisms at play. Our review indicates the potential of non-psychologicalinterventions, such as self-help, or Omega 3 supplementa-tion, to be used as adjunctive therapies, for added therapeu-tic value, or as effective low-intensity interventions whilewaiting to access psychological support. Indeed, non-psy-chological interventions might be offered as “first line” formild depressive symptoms, to offer treatment while reducingburden on healthcare systems, and reserving higher inten-sity or specialized psychological intervention for moder-ate to severe depressive symptomology. This is in line withcurrent National Institute for Health and Care Excellence(NICE) guidelines in the UK, where self-help is offered forsubthreshold depressive symptoms, or mild to moderatedepression, while high intensity psychological interventionis provided to women with moderate or severe depression inpregnancy (Antenatal and Postnatal Mental Health: Clini-cal Management and Service Guidance,). We acknowledge that the range of interventions is verywide, ranging from psychotherapy, to aromatherapy, andbright light therapy, with different therapeutic mechanismsat play. Therefore, we explored through subgroup analyseswhether efficacy across intervention types differed. Indeed,as per the analysis on intervention type, we did not find sig-nificant subgroup differences, as shown in Table. Further-more, we agree with similar work in the field (Dennis andAllen; Wu et al.) on the limited interpretabilityof specific intervention types like aromatherapy for antena-tal depression, due to limited evidence. Therefore, we do | Table 1 Tabular visualisation of subgroup analyses (when outliershave been removed) | | - | | K | SMD | P value | 95%CI | P value(subgroupdifferences) | | Intervention type | | Psychological | 53 | 0.46 | p < 0.0001 | 0.4;0.55 | 0.63 | | Non psychological | 38 | 0.53 | p < 0.0001 | 0.41;0.65 | | Combination | 3 | 0.57 | p < 0.05 | 0.08,1.05 | | Intervention format | | In person | 62 | 0.5 | p < 0.0001 | 0.9 | | Online/digital | 21 | 0.51 | p < 0.0001 | | Monitoredself-administration | 9 | 0.41 | p < 0.05 | | Combination | 2 | 0.56 | p < 0.05 | 0.150.97 | | Measure type | | Clinician | 9 | 0.58 | p < 0.0001 | 0.31;0.85 | 0.5 | | Self-report | 85 | 0.48 | p < 0.0001 | 0.42;0.56 | | Income region | | High income | 44 | 0.36 | p < 0.0001 | 0.3;0.43 | 0.0007 | | Upper-middleincome | 40 | 0.55 | p < 0.0001 | 0.45;0.64 | | Lower-middleincome | 9 | 0.7 | p < 0.05 | 0.34;1.09 | | Low income | 1 | 0.9 | p < 0.0001 | 0.5;1.31 | | Geographical region | | East Asia and Pacific | 26 | 0.47 | p < 0.0001 | 0.34;0.6 | 0.16 | | Europe and CentralAsia | 22 | 0.43 | p < 0.0001 | 0.29;0.58 | | Latin America andCaribbean | 2 | 0.46 | p < 0.22 | −0.27;1.19 | | Middle East andNorth Africa | 18 | 0.58 | p < 0.0001 | 0.45;0.72 | | North America | 17 | 0.41 | p < 0.0001 | 0.32;0.51 | | South Asia | 8 | 0.63 | p < 0.05 | 0.25;1.02 | | Sub Saharan Africa | 1 | 0.9 | p < 0.0001 | 0.49;1.31 | | Control Group Type | | Active control | 16 | 0.42 | p < 0.0001 | 0.26;0.59 | 0.59 | | Placebo | 7 | 0.5 | p < 0.001 | 0.22;0.76 | | Sham control | 2 | 0.55 | p < 0.0001 | 0.45;0.64 | | Unspecified | 2 | 0.94 | p < 0.05 | 0.11;1.7 | | Routine antenatalcare | 67 | 0.5 | p < 0.0001 | 0.420.57 | | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | - | ### Page 9 ### Page 10 feel that the pooled effect sizes need to be interpreted withcaution, due to the heterogeneity levels in our analysis. Nev-ertheless, we believe that the rich set of subgroup analysesacross intervention type and format provide a thorough per-spective on the evidence base. Moreover, the four studies that integrated both psycho-logical and non-psychological aspects also showed compa-rable efficacy. One study combined exercise with CBT (Meiet al.). Given that both CBT (Pettman et al.) andphysical activity (He et al.) show moderate to highefficacy in managing AD, combining the two seems to bea promising approach for treating depression in pregnancy.However, it is important to consider clinician guidancefor pregnant people before undertaking different forms ofphysical activity. In the absence of clear superiority of oneapproach over the other, patient choice and lifestyle andmedical considerations should guide the treatment choice,while future research should investigate predictors that couldpersonalise the best treatment for the individual patient. Aspregnancy is such a biologically sensitive period, identify-ing, for example, biomarkers associated with a response tointerventions, could provide a clinically relevant person-alised approach to the choice of treatment, as is happeningwith inflammatory biomarkers in the context of depressionoutside pregnancy (Miller et al.). In terms of intervention format, we found some sug-gestive evidence that interventions delivered digitally/online were as effective as interventions delivered in per-son. These findings are in line with previous meta-analysesundertaken to review the efficacy of digital interventionsin treating perinatal depression in general and postnataldepression in particular (Ansaari et al.; Ashford etal.; Li et al.). One reason for this could be thatdigital interventions might be easier to access for pregnant(or newly postnatal) women, who might experience physi-cal barriers to accessing interventions delivered in personand instead can attend sessions online. For example, wehave recently shown that an online singing intervention iseffective in reducing depression and anxiety in postnataldepression (Bind et al.). As for combined interventions, combined delivery for-mats also showed comparable efficacy to those employingone intervention format. Integrating the two provides highlevels of flexibility for participants, with digital optionsavailable in between in person sessions, with more fre-quent support from intervention providers. We proposethat having a flexible approach also allows for facilita-tors to engage with participants in-between physicallydelivered sessions and therefore might be associated withhigher levels of participant engagement. It is important to note that our systematic review of theliterature did not yield results of randomised controlledtrials assessing the impact of pharmacological interventionson AD which met our inclusion criteria. This is because, todate, no RCT has been conducted to prospectively assessthe efficacy of antidepressant medication to treat depressedpregnant women and their subsequent impact (Howland). This was already discussed more than 10 years agoby Howland (), who said that research undertakenglobally interprets and emphasises data somewhat differ-ently, and does not provide consistent guidance either forpatients or clinicians. Indeed, there have been no changessince, partly due to the ethical and practical reasons associ-ated with the potentially increased risk of trialing pharmaco-therapy during the perinatal period (Andrade). RCTshave been conducted assessing medications like fluoxetine(Milgrom et al.)and paroxetine (Milis, et al.) totreat postnatal depression, but at the time of data extrac-tion, no RCT had assessed pharmacotherapy for antenataldepression. Our analysis found that most studies used self-reportedoutcome measures. Previous literature on the efficacy ofpsychotherapy on depression has shown that self-reportmeasures are associated with lower effect sizes when com-pared with clinician rated outcomes (Cuijpers et al.).Interestingly, our results were in the same direction, albeitthe difference was not statistically significant. The numberof studies with clinician rated measures (k = 9) is smallerthan those with self-report measures (k = 85), leading to anumerically larger effect size (SMD = 0.58 for clinicianrated measures and 0.48 for self-report measures). There-fore, further research into using both clinician- as well asself-report measures is warranted to truly understand inter-vention efficacy. It is important to emphasize, however, thatclinician rated measures are indeed more objective and lesssusceptible to bias, especially when the clinician or out-come assessor is blinded to intervention allocation, and thismight improve the precision of the measurement and thusthe effect size. Comparisons based on geographical regions did revealsignificant differences, with studies conducted in the MiddleEast, North Africa, and Sub-Saharan Africa having the high-est effect sizes. Given that these regions fall under the UpperMiddle to the Lower Middle-income countries (LMIC), thisindicates that the interventions were most effective at man-aging AD symptomology despite financial and economicresource constraints. In fact, subgroup analyses of efficacyby income region also revealed subgroup differences. Onereason for this difference could be the difference in the stan-dard of care between High-income countries (HIC) andLMICs. Primary care and mental health services in LMICsare in general underfunded, and therefore, those needingaccess to care might not be treated. Therefore, when takingpart in a trial aimed to manage depressive symptomology, ### Page 11 intervention can lead to significant improvements, if com-pared with a very limited or non-existent “control” care asusual. In fact, as also highlighted by Cuijpers et al. (),care as usual in Western or High-income countries mightalready include access to community mental health servicesor other routine treatment pathways, while this might notbe the case in LMICs, where usual care as usual means noaccess to mental health care. Therefore, it is indeed possiblethat the intervention group show larger effect sizes becausethey receive treatment, while the care as usual group doesnot have any access. Given that the prevalence of antena-tal depression is highest in LMICS (Roddy Mitchell et al.), it is important to emphasise that any form of inter-vention could yield larger improvements in depressivesymptomology in these women. In our analyses, high levels of heterogeneity wereexpected due to the wide range of intervention types, for-mat and outcomes, as well as delivery countries and cul-tures. This has been seen in previous meta-analyses ofinterventions in the context of antenatal depression, whichalso indeed reported moderate to high levels of heteroge-neity (Liu et al.; Yasuma et al.) citing similarreasons to ours, such as differences in outcome measures(Yasuma et al.), intervention format, and duration(Liu et al.). Societal factors also may have contrib-uted significantly to the levels of heterogeneity here, as wedid see significant subgroup differences based on incomeregions. Furthermore, our synthesis of the literature foundnumerous types of outcome measures being used, such asthe EPDS, BDI, PHQ-9, CES-D, and clinician rated mea-sures. These outcome measures, though validated for useduring the perinatal period (Sambrook Smith et al.),can lead to between study heterogeneity, due to aspectssuch as cut off scores, method of administration, and con-structs measured. Differences in control conditions mightbe an additional factor which contributes to between-studyheterogeneity. Certain studies utilized an active or placebocontrol, while others had participants on a waitlist or treat-ment as usual control. Therefore, this variation might havecontributed to study heterogeneity. Moreover, “treatment asusual” might differ worldwide: while for some, this meansaccess to community or specialist mental health teams, forothers, it means no treatment at all. That said, we did notfind significant subgroup differences based on control grouptype. For baseline severity, studies, such as required par-ticipants to have a diagnosis of depression, while for others,there was no baseline severity as inclusion. Therefore, whilesome studies have a sample who met criteria for currentantenatal depression, others included low mood, withouthaving to meet threshold to be included in the study. Whileexamining baseline depression scores, some studies, such asChang M.-Y. et al. () had a mean baseline EPDS scoreof approximately 12, while others, like Fontein-Kuiperset al. () had a mean baseline EPDS score of around4.5, highlighting that participants did not meet criteria fordepression. Reassuringly, our risk of bias assessment foundlow risk for. “Bias arising from the randomisation process”. Out of 115 RCTs, 7 reported on biological outcomes.Pregnancy is a time of numerous biological changes, andit is important to note that between-group effect sizes post-interventions are confounded by normal biomarker changesacross this period (Radmark et al.), despite attemptsto mitigate this by focusing on RCTs. Second, biomarkersare examined at a single time point post-intervention andmay be influenced by gestational timing or normal preg-nancy-related changes. Given the varied range of reportedbiomarkers, levels of between-study heterogeneity wereexpectedly high. Despite this, our results highlight the over-arching biopsychosocial impact that interventions have onantenatal depression, affecting both depression and biologi-cal mechanisms linked to stress, and usually in the samedirection, i.e., with a significant overall “stress-reduction”effect. Overall, the biological evidence is heterogeneousand does not point to a single mechanistic pathway. HPA-axis and immune dysfunction are widely implicated inantenatal depression (Osborne et al.), suggesting thatstress-regulation mechanisms may represent one importantintervention target. Some studies showed changes in stress-related biomarkers such as cortisol and blood pressure (Caoet al.; Sanaeinasab et al.) while others indicatedinfluences on reproductive endocrine factors (e.g., oestra-diol) (Nishi et al.) or cardiometabolic markers (lipidsand glucose) (Papandreou et al.), which may interactwith stress-physiology pathways. Although few interven-tion-specific differences were observed in immune mark-ers, inflammation is still thought to contribute to perinataldepression, highlighting the need for more interventionstudies that consistently measure biological biomarkers toclarify underlying mechanisms. We would like to reiter-ate that pooling physiologically different biomarkers hasindeed led to substantial levels of heterogeneity, while atthe same time offering the opportunity to explore a unify-ing conceptual framework based on, broadly speaking, theeffects of intervention on stress-related mechanism. While,based on this meta-analysis, we cannot point to a singlemechanistic pathway, we do add to the growing body ofevidence that HPA axis and immune system dysregula-tion are implicated in antenatal depression, and offer somepromising signal for the role of cardiovascular indicatorssuch as blood pressure and Angiotensin II. Thus, interven-tions targeting stress reduction might provide a mechanismto manage depressive symptomology in pregnancy, butthis interpretation is limited by the largely heterogeneousnature of included studies. ### Page 12 Similar work on this topic was done by Kwok and col-leagues in 2025 (Kwok et al.), wherein they reviewedmajor biomarkers involved in MDD, which also had associ-ations with perinatal depression in the context of treatmentresponse. We now build on this, drawing data from pub-lished work, and identifying specific biomarkers that changeas a result of treatment. This is with the caveat that we didnot do this from the perspective of examining biomarkerspredicting response, but rather seeing the impact that inter-ventions had on biomarkers. Examining intervention effi-cacy from a biological and a psychological lens can helpclinicians with providing a more personalized, precision-based treatment approach in antenatal depression. One study with the highest effect sizes (Cao et al.)delivered a music therapy intervention, and identified sub-stantial reduction in blood pressure levels and in angioten-sin II, a biomarker which has emerged as a critical hormoneaffecting the function of key organs like the kidney, heart,and the brain (Patel and Mehta). This finding high-lights the efficacy of arts in health interventions in treat-ing antenatal depression and their associated biomarkers.Seeing as previous music and singing based interventionshave shown promising results in the treatment of postnataldepression (Bind et al.; Fancourt and Perkins),further research into this intervention type for antenataldepression could prove to be a promising pathway frompregnancy to the first year postpartum. Once again, along the same trend as the clinical outcomesanalysis, we saw significant levels of heterogeneity in thebiomarker effects (I2 = 75%), which reduced (I2 = 45%) oncewe removed outliers. Of course, our search uncovered dif-ferent types of biomarkers, from cholesterol to blood pres-sure, hsCRP, and Angiotensin II, all of which have distinctfeatures, changing during pregnancy. Indeed, there mighthave been numerous ways of extracting samples (blood ver-sus saliva), along with different mechanisms being tested,which contribute significantly to heterogeneity. Our analy-sis was exploratory in nature, and future analyses shouldexamine specific biomarker categories and their associationwith treatment outcomes. In our primary analysis with all studies included, weidentified publication bias through three different methods,along with substantial levels of between-study heterogene-ity. Indeed, as we expected, upon removing outliers, no pub-lication bias was found, and this also led to a reduction inlevels of heterogeneity. Nevertheless, it is possible that stud-ies with very high effect sizes were likely to be influencedby publication bias and their results must be interpreted withcaution. The results of our analysis need to be interpreted in thecontext of certain limitations, the most important being thesubstantial levels of heterogeneity in the primary analysiswith 115 trials, which could have had significant impact oneffect sizes. Moreover, in the context of search methodol-ogy, though we conducted a robust search across differentdatabases, we did not extract papers from the CochraneLibrary and Web of Science. This is because our dataextraction took place on OVID, which allows for searchingacross key databases relevant to perinatal mental health,such as Embase, PsycINFO, Medline, and the Maternityand Infant Care Database (MIDIRS). Reassuringly, whencross checking against other meta- analyses on antenataldepression, we covered all of their included studies. Nev-ertheless, we cannot absolutely exclude the possibility thatusing additional databases might have led to more studiesidentified. While the review was not registered at its incep-tion in 2022, we followed the declared objectives of Hap-pyMums as described in the funded application (Task 3.5)and the protocol paper (Biaggi et al.). Specifically,the protocol paper states that “a literature-based meta-analysis investigating factors associated with responseto interventions for depression in pregnancy will also beconducted”, and that “sociodemographic (…), environ-mental (…) and biological factors (e.g., … inflammation,metabolism-related biomarkers/metabolites, and neuroen-docrine-related biomarkers) will be analysed”. The analyticdecisions were guided by these overarching objectives aswell as the need to dissect the effects of treatment features(e.g., intervention type and format) and socio-demographic/environmental conditions (e.g., income and geographicalregions). Removal of grey literature was intentional, toensure we had sufficient information as normally reportedin published research in order to be consistent across riskof bias assessments. We do understand that exclusion ofgrey literature from meta analyses can have implicationson exaggerated effect sizes (McAuley et al.) howevergiven the wide range of studies currently included in ouranalysis, and the high resulting variability, we believe thisprovides an adequate synthesis of available published lit-erature. Given that unpublished studies in specific popu-lations represent a small proportion of included studies inreviews (Hartling et al.), we felt the above rationaleappropriate to exclude grey literature. In conclusion, this first-of-a-kind meta-analysis demon-strates the efficacy of varied interventions in the manage-ment of depression during pregnancy. Though results seempromising, they must be interpreted with caution owing tothe substantial levels of heterogeneity. Future research willneed to clarify whether combining the most effective inter-ventions can lead to even stronger improvement, or if add-ing biological data to the clinical assessment could facilitatepersonalized, more effective interventions. ### Page 13 Supplementary Information The online version containssupplementary material available at. Author contributions R.Laijawala- Writing - Original Draft, Concep-tualization, Methodology, Formal analysis, K. Priestley- Writing - Re-view & Editing, Conceptualization, Methodology, Formal analysis R.Bind- Writing - Review & Editing K. Hazelgrove- Writing - Review& Editing, L. Rebecchini- Writing - Review & Editing, M. Kirkpat-rick- Writing - Review & Editing, Methodology, Formal analysis F.Mancino- Writing - Review & Editing, Methodology, Formal analysisS. Sangha- Writing - Review & Editing, Conceptualization, Methodol-ogy, Formal analysis A.Biaggi - Writing - Review & Editing, M. Go-dara- Writing - Review & Editing, C. Buß- Writing - Review & EditingS. Etringer- Writing - Review & Editing A. Woods- Writing - Review& Editing, P. Dazzan- Writing - Review & Editing, A. Cattaneo- Writ-ing - Review & Editing, Supervision, Funding acquisition, C.M. Pari-ante- Writing - Original Draft, Conceptualization, Methodology, For-mal analysis, Writing - Review & Editing, Supervision. Funding statement The HappyMums project is supported by the Eu-ropean Union’s Horizon Europe research and innovation programme(Grant Agreement n.101,057,390). Views and opinions expressed arehowever those of the authors only and do not necessarily reflect thoseof the European Union. Neither the European Union nor the grantingauthority can be held responsible for them. The funder played no rolein writing this manuscript and in the decision to submit it for publica-tion. CMP, PD, and KH’s work is supported by the National Institutefor Health and Care Research (NIHR) Maudsley Biomedical ResearchCentre (BRC). The views expressed are those of the author(s) and notnecessarily those of the NIHR or the Department of Health and SocialCare. Declaration of Interest: None. Clinical trial number: not applicable. Data availability Data is provided within the manuscript or supple-mentary information files. Declarations Ethics statement Since data from available (published) literature wasanalysed, and no individual patient data was collected for this paper,formal ethical approval or informed consent was not required. Clinical trial number Not aplicable. Competing interests The authors declare no competing interests. Open Access This article is licensed under a Creative CommonsAttribution 4.0 International License, which permits use, sharing,adaptation, distribution and reproduction in any medium or format,as long as you give appropriate credit to the original author(s) and thesource, provide a link to the Creative Commons licence, and indicateif changes were made. The images or other third party material in thisarticle are included in the article’s Creative Commons licence, unlessindicated otherwise in a credit line to the material. If material is notincluded in the article’s Creative Commons licence and your intendeduse is not permitted by statutory regulation or exceeds the permitteduse, you will need to obtain permission directly from the copyrightholder. To view a copy of this licence, visit. References Al-Abri K, Edge D, Armitage CJ (2023) Prevalence and correlatesof perinatal depression. Soc Psychiatry Psychiatr Epidemiol58(11):1581–1590. Andrade C (2017) Propensity score matching in nonrandomized stud-ies: a concept simply explained using antidepressant treatmentduring pregnancy as an example. J Clin Psychiatry 78(2):e162–e165. Ansaari N, Rajan SK, Kuruveettissery S (2024) Efficacy of in-personversus digital mental health interventions for postpartum depres-sion: meta-analysis of randomized controlled trials. J ReprodInfant Psychol. Antenatal and postnatal mental health: Clinical management andservice guidance. (2020) National Institute for Health and CareExcellence (NICE) Ashford MT, Olander EK, Ayers S (2016) Computer- or web-basedinterventions for perinatal mental health: a systematic review. JAffect Disord 197(h3v, 7906073):134–146. Beck AT, Ward CH, Mendelson M, Mock J, Erbauch J (2011) BeckDepression Inventory [Data set]. Bergink V, Kooistra L, Lambregtse-van den Berg MP, Wijnen H, Bun-evicius R, van Baar A, Pop V (2011) Validation of the Edinburghdepression scale during pregnancy. J Psychosom Res 70(4):385–389. Biaggi A, Zonca V, Anacker C, Begni V, Benedetti F, Bramante A, Bra-niecka A, Brenna V, Bulgheroni M, Buss C, Cavaliere L, CecilCAM, Couch AC, de Barra D, El Marroun H, Entringer S, Grassi-Oliveira R, Jackowska M, Korosi A, Kwant PJC, Lahti J, Leka-dir K, Mansuy I, Manuella F, Marizzoni M, Meyer U, Monk C,Radoš SN, Pariante CM, Pollux BJA, Priestley K, Räikkönen K,Richetto J, Riva MA, Rothmann LM, Simonetti V, Vai B, VernonAC, Žutić M, Cattaneo A (2025) Understanding, predicting, andtreating depression in pregnancy to improve mothers’ and off-spring’s mental health outcomes: the HappyMums study. BrainBehav Immun Health 44:100961. Bind RH, Lawrence AJ, Estevao C, Hazelgrove K, Priestley K, Rebec-chini L, Laijawala R, Miller C, Healey A, Agwuna J, Sevdalis N,Bakolis I, Davis R, Lopez MB, Woods AJ, Crane N, ManoharanM, Burton A, Dye H, Osborn T, Greenwood L, Perkins R, Daz-zan P, Fancourt D, Pariante CM (2025) Clinical effectiveness,implementation effectiveness and cost-effectiveness of a com-munity singing intervention for postnatal depressive symptoms,SHAPER-PND: randomised controlled trial. Br J Psychiatry. Bind RH, Sawyer K, Hazelgrove K, Rebecchini L, Miller C, Ahmed S,Dazzan P, Sevdalis N, Bakolis I, Davis R, Lopez MB, Woods A,Crane N, Manoharan M, Burton A, Dye H, Osborn T, GreenwoodL, Perkins R, Estevao C (2023) Feasibility, clinical efficacy, andwell-being outcomes of an online singing intervention for postna-tal depression in the UK: SHAPER-PNDO, a single-arm clinicaltrial. Pilot Feasibility Stud 9(1):131. Cao S, Sun J, Wang Y, Zhao Y, Sheng Y, Xu A (2016) Music therapyimproves pregnancy-induced hypertension treatment efficacy. IntJ Clin Exp Med 9(5):8833–8838 Carnevali GS, Buoli M (2021) The role of epigenetics in perinataldepression: are there any candidate biomarkers? J Affect Disord280:57–67. Chang M-Y, Chen C-H, Huang K-F (2008) Effects of music therapyon psychological health of women during pregnancy. J Clin Nurs ### Page 14 17(19):2580–2587. Cohen J (2013) Statistical power analysis for the behavioral sciences.Routledge Cox JL, Holden JM, Sagovsky R (1987) Detection of Postnatal Depres-sion: Development of the 10-item Edinburgh Postnatal DepressionScale. Br J Psychiatry 150(6):782–786. Cuijpers P, Karyotaki E, Reijnders M, Purgato M, Barbui C (2018)Psychotherapies for depression in low- and middle‐incomecountries: A meta‐analysis. World Psychiatry 17(1):90–101. Cuijpers P, Li J, Hofmann SG, Andersson G (2010) Self-reported versusclinician-rated symptoms of depression as outcome measures inpsychotherapy research on depression: A meta-analysis, Clin Psy-chol Rev 30(6)768–778 Cuijpers P, Weitz E, Cristea IA, Twisk J (2016) Pre-post effect sizesshould be avoided in meta-analyses. Epidemiol Psychiatric Sci26(4):364–368. Dennis C-L, Allen K (2008) Interventions (other than pharmacologi-cal, psychosocial or psychological) for treating antenatal depres-sion. Cochrane Database Syst Rev Update in: Cochrane DatabaseSyst Rev 20137:CD006795(4):CD006795.. PMID: 23904069 Duval S, Tweedie R (2000) Trim and fill: A simple funnel-plot-basedmethod of testing and adjusting for publication bias in meta-anal-ysis. Biometrics 56(2):455–463. Fancourt D, Perkins R (2018) Effect of singing interventions on symp-toms of postnatal depression: Three-arm randomised controlledtrial. Br J Psychiatry 212(2):119–121.. Cambridge Core Faraone SV (2008) Interpreting estimates of treatment effects: Impli-cations for managed care. P T 33(12):700–711 Field T (2017) Prenatal Depression Risk Factors, DevelopmentalEffects and Interventions: A Review. J Pregnancy Child Health4(1):301. Fontein-Kuipers YJ, Ausems M, de Vries R, Nieuwenhuijze MJ (2016)The effect of Wazzup Mama?! An antenatal intervention to pre-vent or reduce maternal distress in pregnancy. Arch WomensMent Health 19(5):779–788. Hamilton M (1960) A rating scale for depression. J Neurol NeurosurgPsychiatry 23(1):56–62. Hariton E, Locascio JJ (2018) Randomised controlled trials—the goldstandard for effectiveness research: study design: randomisedcontrolled trials. BJOG 125(13):1716. Harrer M, Cuijpers P, Furukawa TA, Ebert DD (2021) Doing Meta-Analysis With R: A Hands-On Guide (1st edn). Chapman & Hall/CRC Press. Hartling L, Featherstone R, Nuspl M, Shave K, Dryden DM, Van-dermeer B (2017) Grey literature in systematic reviews: across-sectional study of the contribution of non-English reports,unpublished studies and dissertations to the results of meta-analy-ses in child-relevant reviews. BMC Med Res Methodol 17(1):64. Hayes LJ, Goodman SH, Carlson E (2013) Maternal antenatal depres-sion and infant disorganized attachment at 12 months. AttachHum Dev 15(2):133–153. He L, Soh KL, Huang F, Khaza’ai H, Geok SK, Vorasiha P, Chen A, MaJ (2023) The impact of physical activity intervention on perinataldepression: a systematic review and meta-analysis. J Affect Dis-ord 321:304–319. Howland RH (2013) Antidepressant medication and pregnancy: time forrandomized controlled trials. J PsychoSoc Nurs Ment Health Serv51(2):11–14. Jarde A, Morais M, Kingston D, Giallo R, MacQueen GM, Giglia L,Beyene J, Wang Y, McDonald SD (2016) Neonatal outcomesin women with untreated antenatal depression compared withwomen without depression: a systematic review and meta-anal-ysis. JAMA Psychiatry 73(8):826–837. Kliem S, Beller J, Tibubos AN, Zenger M, Schmalbach B, Brähler E(2020) A reanalysis of the Center for Epidemiological StudiesDepression Scale (CES-D) using non-parametric item responsetheory. Psychiatry Res 290:113132. Kopelman RC, Moel J, Mertens C, Stuart S, Arndt S, O’Hara MW(2008) Barriers to care for antenatal depression. Psychiatric ServHosp Community Psychiatry 59(4):429–432. Kroenke K, Spitzer RL, Williams JBW (2001) The PHQ-9. J GenIntern Med 16(9):606–613. Kwok W, Wagner-Schuman M, Eisenlohr-Moul T, Hage B (2025) Appli-cation of treatment response biomarkers from major depression toperinatal depression. J Pers Med 15(12):607. Lautarescu A, Victor S, Lau-Zhu A, Counsell SJ, Edwards AD,Craig MC (2022) The factor structure of the Edinburgh Postna-tal Depression Scale among perinatal high-risk and communitysamples in London. Arch Womens Ment Health 25(1):157–169. Lau Y, Cheng J-Y, Wong S-H, Yen K-Y, Cheng L-J (2021) Effective-ness of digital psychotherapeutic intervention among perinatalwomen: a systematic review and meta-analysis of randomizedcontrolled trials. World J Psychiatry 11(4):133–152. Leigh B, Milgrom J (2008) Risk factors for antenatal depression, post-natal depression and parenting stress. BMC Psychiatry, 8(Abidin,R. R. (1986). Parenting Stress Index Manual 2nd edition. Charlot-tesville, VA: Pediatric Psychology Press; 1986.). Li C, Sun X, Li Q, Sun Q, Wu B, Duan D (2020) Role of psycho-therapy on antenatal depression, anxiety, and maternal quality oflife: a meta-analysis. Medicine Baltimore 99(27):e20947. Li L, Yue S-W, Xu J, Qiao J, Redding SR, Ouyang Y-Q (2023) Effec-tiveness of Internet-based psychological interventions for treat-ing perinatal depression: a systematic review and meta-analysis.J Clin Nurs 32(13–14):3087–3101. Liu L, Liu C, Liu X, Yang Y (2023) Summary of the effect of an exerciseintervention on antenatal depression and the optimal program: asystematic review and meta-analysis. BMC Pregnancy Childbirth23(1):293. Li Y, Chen J, Chen B, Wang T, Wu Z, Huang X, Li S (2022) Effect ofmindfulness meditation on depression during pregnancy: A meta-analysis. Front Psychol 13. McAuley L, Pham B, Tugwell P, Moher D (2000) Does the inclusionof grey literature influence estimates of intervention effectivenessreported in meta-analyses? Lancet 356(9237):1228–1231. Mei Q, Chen X, Liu L, Xiao G (2024) An investigation into the cor-relation between early-to-mid pregnancy exercise combined withcognitive behavioral therapy and anxiety levels and quality of ### Page 15 life in patients. Journal of Obstetrics and Gynaecology Research50(3):381–388. Melsen WG, Bootsma MCJ, Rovers MM, Bonten MJM (2014) Theeffects of clinical and statistical heterogeneity on the predic-tive values of results from meta-analyses. Clin Microbiol Infect20(2):123–129. Milgrom J, Gemmill AW, Ericksen J, Burrows G, Buist A, Reece J(2015) Treatment of postnatal depression with cognitive behav-ioural therapy, sertraline and combination therapy: a randomisedcontrolled trial. Australian New Z J Psychiatry 49(3):236–245. Milis SM, Reebye P, Corral M (2004) Lisa. The Use of Paroxetineand Cognitive-Behavioral Therapy in Postpartum Depressionand Anxiety: A Randomized Controlled Trial. Psychiatrist.Com. Miller AH, Berk M, Bloch G, Briquet-Laugier V, Brouillon C, Cuth-bert BN, Dantzer R, Davis MC, De Picker LJ, Drevets WC, EyreHA, Hack LM, Harrison NA, Krystal AD, Lombardo G, MondelliV, Pariante CM, Pulvirenti L, Salvadore G, Leboyer M (2025)Advancing precision psychiatry and targeted treatments: insightsfrom immunopsychiatry. Brain Behav Immun 125:319–329. Montgomery SA, Åsberg M (1979) A new depression scale designed tobe sensitive to change. Br J Psychiatry 134(4):382–389. Nishi D, Su K-P, Usuda K, Chang JP-C, Hamazaki K, Ishima T, SanoY, Ito H, Isaka K, Tachibana Y, Tanigaki S, Suzuki T, HashimotoK, Matsuoka YJ (2020) Plasma estradiol levels and antidepressanteffects of omega-3 fatty acids in pregnant women. Brain BehavImmun 85( ):29–34. Osborne S, Biaggi A, Chua TE, Du Preez A, Hazelgrove K, NikkheslatN, Previti G, Zunszain PA, Conroy S, Pariante CM (2018) Ante-natal depression programs cortisol stress reactivity in offspringthrough increased maternal inflammation and cortisol in preg-nancy: The Psychiatry Research and Motherhood - Depression(PRAM-D) Study. Psychoneuroendocrinology 98:211–221. Papandreou P, Amerikanou C, Vezou C, Gioxari A, Kaliora AC, Skou-roliakou M (2023) Improving adherence to the Mediterranean dietin early pregnancy using a clinical decision support system; a ran-domised controlled clinical trial. Nutrients. Patel BM, Mehta AA (2012) Aldosterone and angiotensin: role in dia-betes and cardiovascular diseases. Eur J Pharmacol 697(1):1–12. Pettman D, O’Mahen H, Blomberg O, Svanberg AS, von Essen L,Woodford J (2023) Effectiveness of cognitive behavioural ther-apy-based interventions for maternal perinatal depression: A sys-tematic review and meta-analysis. BMC Psychiatry 23(1):208. Plant D, Barton A (2020) Adding value to real-world data: The role ofbiomarkers. Rheumatology (Oxford) 59(1):31–38. Priestley K, Laijawala R, Hazelgrove K, Bind R, Rebecchini L, Mar-iani N, Alford S, Kirkpatrick M, Mancino F, Kim S, Pushpak-anthan S, Biaggi A, Cavaliere L, Di Benedetto MG, Matijaš M,Žutić M, Brekalo M, Nakić Radoš S, Żukowska K, Pariante C(2026) HappyMums mobile application study protocol: Use ofa smartphone application to gather data predictive of antenataldepression. BMJ Open 16(2):e106978. Radloff LS (1977) The CES-D Scale: A Self-Report Depression Scalefor Research in the General Population. Appl Psychol Meas1(3):385–401. Radmark L, Osika W, Wallen MB, Nissen E, Lonnberg G, BranstromR, Henje E, Gardner R, Fransson E, Karlsson H, Niemi M (2023)Autonomic function and inflammation in pregnant women par-ticipating in a randomized controlled study of MindfulnessBased Childbirth and Parenting. BMC Pregnancy Childbirth23(1):237 Embase (2022468600). RoB 2: A revised tool for assessing risk of bias in randomised trials |The BMJ. (n.d.). Retrieved 3 March 2025, from Roddy Mitchell A, Gordon H, Lindquist A, Walker SP, Homer CSE,Middleton A, Cluver CA, Tong S, Hastie R (2023) Prevalence ofperinatal depression in low- and middle-income countries: a sys-tematic review and meta-analysis. JAMA Psychiatry 80(5):425–431. Sambrook Smith M, Cairns L, Pullen LSW, Opondo C, Fellmeth G,Alderdice F (2022) Validated tools to identify common mentaldisorders in the perinatal period: a systematic review of system-atic reviews. J Affect Disord 298:634–643. Sanaeinasab H, Saffari M, Sheykh-Oliya Z, Khalaji K, Laluie A, AlZaben F, Koenig HG (2021) A spiritual intervention to reducestress, anxiety and depression in pregnant women: randomizedcontrolled trial. Health Care Women Int 42(12):1340–1357. Shortis E, Warrington D, Whittaker P (2020) The efficacy of cognitivebehavioral therapy for the treatment of antenatal depression: a sys-tematic review. J Affect Disord 272:485–495. Silva-Fernandes A, Conde A, Marques M, Caparros-Gonzalez RA,Fransson E, Mesquita AR, Figueiredo B, Skalkidou A (2024)Inflammatory biomarkers and perinatal depression: a systematicreview. PLoS One 19(5):e0280612. Smith A, Twynstra J, Seabrook JA (2020) Antenatal depression and off-spring health outcomes. Obstet Med 13(2):55–61. Srisurapanont M, Oon-arom A, Suradom C, Luewan S, Kawilapat S(2023) Convergent validity of the Edinburgh postnatal depressionscale and the patient health questionnaire (PHQ-9) in pregnantand postpartum women: their construct correlations with func-tional disability. Healthcare Basel 11(5):699. Vigod SN, Wilson CA, Howard LM (2016) Depression in pregnancy.BMJ i1547. WDI - The World by Income and Region. (n.d.). Retrieved 15 April2024, from Wilczynska D, Walczak-Kozlowska T, Santos-Rocha R, LaskowskiR, Szumilewicz A (2023) Stress is not so bad-cortisol level andpsychological functioning after 8-week HIIT program duringpregnancy: a randomized controlled trial. Front Public Health11:1307998. Wu Y, Wang H, Hui X, Qian Z, Wang M, E F, Xu M, Zhou L, DengX, Li X, Yang K (2025) Efficacy of non-pharmacological inter-ventions for the prevention of antenatal depression: a system-atic review and network meta-analysis. Neurosci Biobehav Rev172:106088. Yasuma N, Narita Z, Sasaki N, Obikane E, Sekiya J, Inagawa T, Naka-jima A, Yamada Y, Yamazaki R, Matsunaga A, Saito T, WatanabeK, Imamura K, Kawakami N, Nishi D (2020) Antenatal psycho-logical intervention for universal prevention of antenatal andpostnatal depression: a systematic review and meta-analysis. JAffect Disord 273(h3v):231–239. ### Page 16 Yedid Sion M, Harlev A, Weintraub AY, Sergienko R, Sheiner E (2016)Is antenatal depression associated with adverse obstetric and peri-natal outcomes? J Matern Fetal Neonatal Med 29(6):863–867. Zeng G, Niu J, Zhu K, Li F, Li L, Gao K, Zhuang Y, Zhang B,Han X, Ye G, Gao Z, Li H (2025) Effects of non-pharma-cological interventions on depressive and anxiety symp-toms in pregnant women: a systematic review and networkmeta-analysis. eClinicalMedicine. Zhang Z, Liu YJ, Sun L, Zhao X-D (2024) The effects of exerciseon antenatal depression: a systematic review and meta-analysis.Front Psychiatry. Zuo H, Chen X, Huang X, Benny C, Fu D, Xiu Q, Cui X, Lyu Y (2025)Using inflammatory biomarkers in early pregnancy to predictsubsequent antenatal depression. J Affect Disord 371:156–163. Publisher’s note Springer Nature remains neutral with regard to juris-dictional claims in published maps and institutional affiliations. Authors and Affiliations Riddhi Laijawala1 · Kristi Priestley1 · Rebecca Bind1 · Katie Hazelgrove1 · Lavinia Rebecchini1 · Madeline Kirkpatrick1 ·Francesca Mancino1 · Samrina Sangha1 · Alessandra Biaggi2 · Malvika Godara3 · Claudia Buss3,4 · Sonja Entringer3,5 ·Anthony Woods1 · Paola Dazzan6 · Annamaria Cattaneo2,7 · Carmine M Pariante1 Riddhi Laijawala Kristi Priestley Rebecca Bind Katie Hazelgrove Lavinia Rebecchini Madeline Kirkpatrick Francesca Mancino Samrina Sangha Alessandra Biaggi Malvika Godara Claudia Buss Sonja Entringer Anthony Woods Paola Dazzan Annamaria Cattaneo Carmine M Pariante 1 Department of Psychological Medicine, Institute ofPsychiatry, Psychology and Neuroscience, Kings CollegeLondon, London, United Kingdom 2 Department of Pharmacological and Biomolecular Sciences,University of Milan, Milan, Italy 3 Institute of Medical Psychology, Charité—Universitätsmedizin Berlin, Berlin, Germany 4 German Center for Child and Adolescent Health, Berlin,Germany 5 German Center for Mental Health, Berlin, Germany 6 Department of Psychosis Studies, Institute of Psychiatry,Psychology and Neuroscience, Kings College London,London, United Kingdom 7 Biological Psychiatry Unit, IRCCS Istituto Centro SanGiovanni di Dio Fatebenefratelli, Brescia, Italy