This prospective study is a prespecified secondary analysis of the multicenter randomized controlled trial (RCT) conducted as a part of the Innovation Fund Project Mind:Pregnancy (01NVF17034) in the state of Baden-Württemberg, Germany. Participating centers included the university hospitals of Heidelberg and Tübingen, as well as more than 200 gynecological practices, as previously described and published [19]. Initial screening for eligibility took place routinely from February 2019 to October 2020 with 5299 pregnant women and was conducted using the EPDS. Women with subclinical signs of depression scoring more than 9 were invited to participate in the study. The patients underwent a diagnostic assessment and psychological evaluation by a trained psychologist. A short version of the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition-Clinician Version (SCID-5-CV), was used to exclude acute needs for psychological treatment, including acute psychotic episodes, acute bipolar affective symptoms, acute substance abuse, borderline personality disorder, and acute traumatic events without a relation to the current pregnancy.

Additional inclusion criteria encompassed being 18 years of age or older, adequate proficiency in German, a singleton pregnancy, no known neonatal malformations or anomalies, less than 29 weeks of gestation at screening, having health insurance coverage through one of the participating statutory health insurance providers, and residing within the state of Baden-Württemberg. The exclusion criteria have been described in detail in the previously published study protocol [19].

Study participants were randomly assigned at a ratio of 1:1 to the intervention group (IG) or the control group (CG). The IG was given access to the digital mindfulness–based intervention (eMBI) with 8 weekly sessions during weeks 29 to 36 of pregnancy, lasting 45 minutes, whereas the CG received treatment as usual (TAU) without any restrictions or regulations within the study period. TAU referred to routine obstetric and mental health care available outside the study. Participants in the CG did not receive any additional study-specific psychological intervention, mindfulness training, or therapist contact through the study. The use of external pharmacological or nonpharmacological treatment was not restricted and was documented where available.

Both groups completed the questionnaires digitally via an application and had access to an online pregnancy guidebook. For further information about the intervention, see the study protocol [19].

This prespecified secondary analysis was restricted to participants from the parent RCT who had elevated depressive symptoms at screening (EPDS>9) and an ICD-10–diagnosed affective or stress-related disorder at diagnostic assessment. The analyzed sample included 147 participants, of whom 73 were allocated to the IG and 74 to the TAU-CG. Participants without an ICD-10 affective or stress-related disorder were part of the parent trial but were not included in the present analysis.

Diagnoses included depressive disorders (ICD-10: F32-F33), anxiety disorders (F40-F41), somatoform or dissociative disorders (F44-F45, F48.1, and F68.0), and stress-related disorders or acute stress reactions (F43 and F62). The present analysis compared outcomes between the IG and CG within this diagnostically defined subgroup.

Assessments were conducted digitally at predefined time points from pregnancy to postpartum follow-up. Baseline information, diagnostic status, and sociodemographic characteristics were assessed before randomization. Depressive symptoms were assessed using the EPDS, generalized anxiety using the State-Trait Anxiety Inventory-State (STAI-S) and State-Trait Anxiety Inventory-Trait (STAI-T), and pregnancy-related anxiety using the Pregnancy-Related Anxiety Questionnaire-Revised (PRAQ-R). The main antenatal follow-up assessments were conducted during late pregnancy, including gestational week 34 (T4) and gestational week 36 (T5). Postpartum follow-up assessments were conducted at 4‐6 weeks post partum (T6) and 6 months post partum (T7). The primary outcome was EPDS symptom severity at T6.

The extent of missing data was summarized descriptively by group, time point, and outcome. Missingness patterns were examined to identify whether missing values increased over time and whether they differed between the study groups.

The study protocol was approved by the ethics committee of the Medical Faculty of the Universities of Heidelberg (S-744/2018) and Tübingen (952/2018BO2), Germany, and was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. All personal data were collected and processed subject to confidentiality and the European General Data Protection Regulation (EU-GDPR) [33]. All participants provided signed written consent and received €100 (1€=US $1.14 as of July 27, 2026) each as financial compensation. The study participants had the right to request information about the personal data collected from them. Participation in the study was voluntary after all study content and objectives had been declared and written consent had been given; participation could be withdrawn at any time without any disadvantages for the patient. Patient data and clinical data were recorded pseudonymously and stored centrally in an electronic case report form (eCRF). This study follows the CONSORT (Consolidated Standards of Reporting Trials; Checklist 1 [34]) statement and the SPIRIT (Standard Protocol Items: Recommendations for Intervention Trials [35]) guidelines. The study was registered with the German Clinical Trials Registry (DRKS00017210).

A total of 147 participants were included in the analysis, with 74 women in the CG and 73 in the IG (Figure 1). Baseline demographic and socioeconomic characteristics were broadly comparable between the groups. For example, maternal age, parity, educational level, partnership status, and baseline EPDS scores showed no clinically relevant between-group imbalances (Table S1 in Multimedia Appendix 1).

The most frequently documented primary diagnoses were recurrent depressive disorders (ICD-10: F32-F33), anxiety disorders (F40-F41), and stress-related disorders (F43). The distribution of these diagnostic groups was comparable between the IG and CG (Table 1).

Secondary end points included generalized anxiety symptoms assessed using the STAI-T (trait component) and STAI-S (state component), as well as pregnancy-specific anxiety symptoms assessed using the Pregnancy-Related Anxiety Questionnaire (PRAQ).

In this prespecified secondary analysis of pregnant women with elevated depressive symptoms and ICD-10–diagnosed affective or stress-related disorders, allocation to the eMBI was associated with greater reductions in depressive symptoms than TAU. The most consistent effects were observed during late pregnancy and at 4-6 weeks post partum, with the latter representing the time point with the highest incidence of postpartum depression [36]. Response and remission outcomes supported the clinical relevance of these findings, although substantial postpartum missingness and attenuation of effects at 6 months post partum warrant cautious interpretation.

Regarding anxiety symptoms assessed with the STAI-T and STAI-S, the IG demonstrated significantly lower symptom severity at T4 and T6 compared to the CG. This highlights the clinical significance of the intervention, as the early postpartum phase constitutes a critical window for both the prevention and treatment of maternal mental health disorders. The anxiety findings also suggest that the intervention may have had a more immediate effect on situational anxiety than on more stable dispositional anxiety. STAI-S reflects temporary feelings of tension, nervousness, and worry, whereas STAI-T captures a more stable tendency toward anxiety. The more consistent pattern for state anxiety is plausible, as mindfulness-based exercises may primarily support momentary emotion regulation and stress reduction. Changes in trait anxiety may require longer intervention exposure, repeated practice, or additional therapeutic components.

Although the between-group differences at T7 were no longer statistically significant, this time point reflects a considerably longer interval after the intervention, during which treatment effects may naturally attenuate as postpartum environmental and psychosocial factors increasingly dominate symptom trajectories. Thus, the attenuation of effects at 6 months post partum suggests that an 8-week antenatal intervention may not be sufficient as a stand-alone approach for women with preexisting affective or stress-related disorders. The transition into the postpartum period is accompanied by changing sleep patterns, caregiving demands, and evolving psychosocial stressors, which may reduce the durability of effects achieved during pregnancy. Future digital interventions may, therefore, benefit from a maintenance phase, postpartum booster sessions, or optional stepped-care components that provide additional support after childbirth.

The importance of our study lies in the demonstrated benefits of a digital intervention especially for predisposed women at risk for adverse mental health outcomes, which has not been assessed before in the context of digital mindfulness interventions during pregnancy. It is feasible to assume that a coexisting mental disorder diagnosis contributed to the effectiveness of the digital intervention in the long-term improvement of maternal mental health. The importance of a mindfulness-based intervention for the reduction of depressive and anxiety-related symptoms is highlighted by our results, as significant improvements were observed beyond the use of the digital application [13,37].

The present findings are consistent with prior work demonstrating similar efficacy for such electronic versions of mindfulness training [38-40]. It is known that a digital approach for health interventions may be effective for patients as depression treatment [41], older patients with anxiety or depression symptoms [42], pregnant women experiencing intimate partner victimization [22], and patients with comorbid depression and substance abuse [43]. The necessity of an adaptation to the individual needs of the patient and of personal interaction with a health care professional or psychologist has been pointed out [42]. Our results are in line with the findings of Zemestani and Fazeli Nikoo [21] in that a mindfulness intervention improves both mindfulness and symptoms of mental disorders. However, our study extends existing research by including a significantly larger sample size and incorporating a digital intervention.

Overall, our findings provide clear evidence of the benefits of the intervention compared to routine care and are consistent with existing international treatment recommendations. The Canadian Network for Mood and Anxiety Treatments (CANMAT) guidelines for the management of perinatal depression explicitly recommend the use of psychosocial and psychological interventions, including mindfulness-based approaches, as first-line treatment options for mild to moderate cases, emphasizing their safety and acceptability in pregnancy and postpartum populations [44]. Zuranolone has recently become available as an oral pharmacological option specifically indicated for postpartum depression. This development expands postpartum treatment options but does not address the same clinical context as an antenatal eMBI. Because the present study did not compare eMBI with pharmacological treatment, our findings should not be interpreted as evidence of superiority over medication. Instead, they suggest that digital mindfulness–based support may serve as a scalable nonpharmacological adjunct within stepped perinatal mental health care [20].

Strengths of the study have already been published and include the prospective, longitudinal design with a follow-up period of up to 5 months after childbirth, the multidimensional approach, the broad range of confounders taken into account, and the large number of screened and included patients contributing to the high power of the RCT [42]. Associated diagnoses were made based on DSM (Diagnostic and Statistical Manual of Mental Disorders) and ICD criteria, and a broad range of confounders was taken into account. The large number of screened and included patients contributes to the high power of our RCT while comparable other research in many cases lacks an active CG [45]. Importantly, both pharmacological and nonpharmacological treatments were systematically assessed, thereby minimizing the risk of bias due to unmeasured concomitant therapies.

Nevertheless, some limitations must be considered. Although pharmacological treatments were recorded, residual confounding by socioeconomic or environmental variables cannot be excluded. Moreover, due to the relatively low prevalence of certain mental health diagnoses, the impact of less common disorders on treatment efficacy could not be systematically evaluated. Furthermore, adjustment for baseline EPDS scores cannot fully account for clinical heterogeneity, including differences in disorder chronicity, recurrence, comorbidity, or prior treatment history. Future research should address these factors, include larger diagnostic subgroups, and explore the long-term sustainability of intervention effects beyond the early postpartum period. At T7, effects were no longer statistically significant, which must be interpreted in light of the substantial temporal distance from the intervention period, during which declining adherence, competing postpartum demands, and nonspecific recovery processes may have diluted detectable group differences.

A further limitation is the substantial proportion of missing outcome data during the postpartum assessments, particularly in the IG. At T6, EPDS values were missing for 47.9% (35/73) of participants in the IG and 40.5% (30/74) in the CG. Although sensitivity analyses using CIR and J2R assumptions supported the robustness of the short-term effects, the higher dropout rate in the IG requires cautious interpretation. Because no formal reasons-for-dropout analysis was available, we cannot determine whether the missingness reflected general postpartum burden, reduced motivation to complete app-based assessments after the intervention had ended, perceived lack of need following symptom improvement, or friction caused by the absence of direct therapist contact. Furthermore, the observed-case responder and remission analyses may be susceptible to selection bias because participants with missing postpartum EPDS data were excluded from the logistic models; consequently, the estimated ORs may overestimate clinical benefit if missingness was related to symptom severity, treatment response, or postpartum burden.

Future studies should systematically collect reasons for dropout and app engagement metrics to clarify whether digital interventions become burdensome during the early caregiving phase.

In pregnant women with ICD-10 affective or stress-related disorders, an eMBI was associated with meaningful reductions in depressive symptoms, with the most reliable improvements observed during late pregnancy and at 4-6 weeks post partum—the clinical period of highest vulnerability to perinatal depression. Anxiety symptoms also improved but with less consistency, and effects at 6 months post partum were attenuated.

Taken together, our study provides robust evidence that digital mindfulness–based programs may serve as an effective, guideline-concordant adjunct to existing therapeutic strategies, with particular benefits for women at elevated risk due to preexisting psychiatric diagnoses. The improvements were most pronounced during late pregnancy and the early postpartum period, when women are particularly vulnerable to developing postpartum depression. These findings provide strong support for integrating scalable, digital mindfulness–based programs into perinatal mental health care in line with Canadian and European guideline recommendations.