Discussion

This NMA of 78 RCTs involving 11 345 participants represents one of the most comprehensive comparative syntheses of psychological treatments for perinatal depression and anxiety to date. CBT demonstrated consistent therapeutic effects for both conditions at post-intervention, supported by moderate certainty of evidence. While other treatments (e.g. IPT, MBI and behavioural therapy) also showed promising effects, confidence in these findings was limited by small evidence bases, high risk of bias and substantial heterogeneity, yielding low or very low certainty. Furthermore, we identified that participant age, country income level and risk of bias were key factors contributing to the observed heterogeneity.

Three treatments (CBT, IPT, MBI) showed efficacy for perinatal depression compared with TAU, with CBT maintaining long-term benefits. For perinatal anxiety, CBT, MBI and behavioural therapy were effective in the short term, yet only behavioural therapy maintained its benefits at follow-up, based on limited evidence. These findings have significant clinical importance, given the persistence of perinatal depression and anxiety and their established association with subsequent maternal mental health disorders. ^{Reference Astbury, Pinnington, Milgrom and Bei40–Reference Meltzer-Brody and Stuebe42} These findings align with, and extend, previous pairwise meta-analyses reporting moderate effects for CBT (d = 0.61–0.65) ^{Reference Sockol17} and IPT (d = 0.33–0.69) ^{Reference Sockol18} for perinatal depression, and large MBI effects (d = 1.44–2.37) ^{Reference Leng, Yin and Ng19} across conditions. Two recent NMAs similarly identified large MBI effect sizes and supported CBT efficacy. ^{Reference Zeng, Niu, Zhu, Li, Li and Gao43,Reference Smith, Dang, Shen, Hung, Lam and Kwok44} Extending prior work that lacked CINeMA evaluations ^{Reference Smith, Dang, Shen, Hung, Lam and Kwok44} or clinically meaningful change thresholds, ^{Reference Zeng, Niu, Zhu, Li, Li and Gao43} we integrated both to comprehensively evaluate evidence certainty. This approach yielded more prudent clinical recommendations: while MBI showed the largest effect size, very low evidence certainty precludes strong recommendations; conversely, CBT demonstrated moderate certainty with robust effects, supporting its role as a first-line therapy. The present study also extends previous NMAs by focusing exclusively on clinically significant populations, providing the first comprehensive NMA for perinatal anxiety and empirically demonstrating differential effects between waitlist and TAU control conditions. Nevertheless, the efficacy evidence for perinatal anxiety remains comparatively limited, with CBT currently the only intervention supported by moderate certainty of evidence. These findings may reflect the larger research base for depression. Perinatal anxiety presentations are more heterogeneous and complex, ^{Reference Zimmermann, Mahanna, Shashkova, Drouhard, Carr and Sheldrick45} requiring flexible and multifaceted clinical responses. These factors highlight the urgent need for further research in this understudied area.

Differences in the short- and long-term efficacy across treatments likely result from variations in the evidence base and treatment-specific mechanisms. The transdiagnostic efficacy of CBT may reflect its targeting of cognitive–behavioural mechanisms potentially shared by perinatal anxiety and depression, providing generalisable skills for managing evolving perinatal stressors such as sleep deprivation, role transitions and infant care demands. ^{Reference Forsell, Bendix, Holländare, Szymanska von Schultz, Nasiell and Blomdahl-Wetterholm46–Reference O’Mahen, Himle, Fedock, Henshaw and Flynn48} IPT’s sustained efficacy for depression may relate to its focus on social relationships and role transitions, ^{Reference Bright, Charrois, Mughal, Wajid, McNeil and Stuart49} which are recognised as key determinants of perinatal depression. ^{Reference Qi, Zhao, Liu, Li and Hu50,Reference Yin, Sun, Jiang, Xu, Gan and Zhang51} However, limited evidence precluded definitive conclusions for anxiety. For other interventions, the lack of sustained effects stems mainly from an insufficient number of long-term trials and underpowered analyses. Another contributing factor may be the nature of the intervention strategies themselves. One possible explanation for the lack of sustained anxiety benefits from MBI is the challenge of maintaining daily mindfulness practice (typically 30–45 min) amid postpartum demands. ^{Reference Pan, Lin, Kuo, Chiu and Ling52,Reference Lönnberg, Jonas, Bränström, Nissen and Niemi53} Similarly, the benefits of PST for depression may not persist because its narrow focus on resolving current problems may leave patients inadequately equipped to handle the ongoing challenges of evolving infant developmental stages. ^{Reference Nezu, D’Zurilla, Zwick and Nezu54} Behavioural therapy showed promise for anxiety, but as an umbrella construct for goal-directed techniques such as behavioural activation and relaxation, ^{Reference Johnson, Edge, Holmes and Carver55,Reference Toussaint, Nguyen, Roettger, Dixon, Offenbächer and Kohls56} the sustainability and transferability of its gains require further evaluation through high-quality trials.

Our study also highlights several methodological limitations in the current literature. In our NMA, a substantial proportion of the included studies (79.5%) raised concerns regarding the selective reporting of outcomes. We advocate for the future use of publicly accessible pre-registration platforms, which could enhance transparency, increase the visibility of null findings and reduce publication bias. ^{Reference Cashin, Richards, DeVito, Mellor and Lee57} Furthermore, 52.6% of the studies were at potential risk of bias due to the measurement of outcomes, ^{Reference Li, Laplante, Paquin, Lafortune, Elgbeili and King58} primarily attributable to a lack of blinding during allocation and outcome assessment. While blinding is inherently challenging in psychosocial intervention trials, this limitation can potentially affect protocol adherence or amplify expectancy effects, potentially resulting in overestimation or underestimation of true treatment effects. ^{Reference Huneke, Fusetto Veronesi, Garner, Baldwin and Cortese59} Future studies should incorporate objective biomarkers (e.g. cortisol levels), or clinician-rated structured diagnostic interviews to enhance confidence in outcome measures. Finally, inadequate reporting of randomisation processes and allocation concealment was noted in 41.0% of the studies. Comprehensive disclosure of these methodological details is crucial for establishing allocation validity and supporting the reliable application in clinical practice. ^{Reference Schulz and Grimes60}

For clinical application, psychological treatments demonstrate comparable effects to pharmacological treatments for perinatal depression. ^{Reference Brown, Wilson, Ayre, Robertson, South and Molyneaux61,Reference McDonagh, Matthews, Phillipi, Romm, Peterson and Thakurta62} For instance, CBT showed a treatment effect comparable to that of selective serotonin reuptake inhibitors (s.m.d. −0.30, 95% CI −0.55 to −0.05), ^{Reference Brown, Wilson, Ayre, Robertson, South and Molyneaux61} though direct comparisons remain scarce. Furthermore, when translating the results into clinically important changes on scale scores, we observed that the 95% CrIs for most treatments, including CBT, were wide and crossed the clinical significance boundary. This indicates uncertainty about their generalisable benefits across populations. Therefore, we advocate that medical decisions be guided by careful consideration of individual contextual factors, such as symptom severity. Given medication safety concerns during pregnancy and breastfeeding, ^{Reference Battle, Salisbury, Schofield and Ortiz-Hernandez.13} including risks of neonatal adaptation syndrome ^{Reference Grigoriadis, VonderPorten, Mamisashvili, Eady, Tomlinson and Dennis63} and persistent pulmonary hypertension, ^{Reference Grigoriadis, VonderPorten, Mamisashvili, Tomlinson, Dennis and Koren64} psychological treatments constitute important first-line options for mild-to-moderate symptoms. ⁶⁵ Combined psychotherapy and pharmacotherapy should be reserved for severe depression, treatment-resistant cases or when rapid symptom reduction is critical for maternal or infant safety. ^{Reference Dunlop66}

The present study identified substantial variation in treatment effects across studies, consistent with previous perinatal meta-analyses that reported high heterogeneity (I ² up to 97.6%). ^{Reference Cuijpers, Franco, Ciharova, Miguel, Segre and Quero14,Reference Sockol, Epperson and Barber15,Reference Sockol18,Reference Leng, Yin and Ng19} The association between higher risk of bias and larger depression effects reinforces caution when interpreting findings for treatments with limited or methodologically weaker evidence bases. The significant income-level effect for anxiety is particularly concerning given that 74.3% of trials were conducted in high-income settings, raising concerns about generalisability to low- and middle-income settings. Younger maternal age predicted better treatment response for both conditions, potentially reflecting a lower comorbidity burden, higher engagement or generational differences in mental health literacy. ^{Reference Saccone, Gragnano, Ilardi, Marrone, Strina and Venturella67,Reference Farrer, Leach, Griffiths, Christensen and Jorm68} Beyond participant-level moderators, control condition differences also contributed to effect size variation. Consistent with the nocebo hypothesis, ^{Reference Chavarria, Vian, Pereira, Data-Franco, Fernandes and Berk69,Reference Furukawa, Noma, Caldwell, Honyashiki, Shinohara and Imai70} comparisons against the waitlist control yielded more favourable efficacy results than those against TAU in this study. This finding suggests prior NMAs merging waitlist and TAU may have overestimated efficacy for interventions primarily compared against waitlist control, highlighting the critical need to specify and standardise control conditions in future research.