Abstract
Background
Postpartum depression (PPD) affects 18% of postpartum women globally. PPD is a heterogeneous condition with diverse presentations, which may differ in underlying etiology, potential outcomes, and optimal treatment. This study aimed to identify and validate PPD subtypes using comprehensive phenotypic and genetic data from a large, nationwide cohort.
Methods
In this population-based cohort study, we used data from the Norwegian Mother, Father, and Child Cohort Study and Medical Birth Registry of Norway (MoBa), and applied unsupervised clustering (Uniform Manifold Approximation and Projection [UMAP] + Density-Based Spatial Clustering of Applications with Noise [DBSCAN]) to identify PPD subtypes among 7859 women with PPD (training set n = 5239). All pregnant women in Norway were eligible to enroll in MoBa between Jun 1, 1999 and Dec 31, 2008. Among women with their first recorded pregnancies in MoBa, they were considered to have clinically significant PPD if they scored ≥8 on the 6-item version of the Edinburgh Postnatal Depression Scale (EPDS) (equivalent to ≥11 on the full EPDS) administered at six months after birth. Input variables included psychiatric symptom severity, psychiatric history, trauma history, pain, and substance use. We then characterized clusters and tested associations with perinatal-relevant auxiliary variables and polygenic scores (PGS) for eight psychiatric and neurodevelopmental conditions. Findings were validated in a reserved test set (n = 2620).
Findings
Nine clusters were identified in the training set (n = 5239), with seven replicating in the independent test set (n = 2620). Clusters were differentiated by symptom severity, onset timing, trauma history, and pain during pregnancy. In the training set, the cluster with the most severe symptoms was characterized by depression + trauma and comprised 28% [1483/5239] of training sample. This cluster showed marked socioeconomic adversity and elevated genetic risk for psychiatric conditions compared to other clusters (ADHD PGS OR 1.19, 95% CI 1.12–1.28; major depression PGS OR 1.15, 95% CI 1.07–1.23; PTSD PGS OR 1.09, 95% CI 1.02–1.17). In the test set, this cluster also comprised 28% [729/2620] and showed consistent associations with ADHD (OR 1.23, 95% CI 1.12–1.35), major depression (OR 1.09, 95% CI 0.99–1.20), and PTSD (OR 1.17, 95% CI 1.06–1.29). Conversely, the mild PPD cluster (10% [498/5239] of training sample) showed a protective profile (ADHD PGS OR 0.81, 95% CI 0.73–0.90; major depression PGS OR 0.85, 95% CI 0.77–0.94). In the test set, it represented 10% (263/2620), with consistent protective associations for ADHD (OR 0.84, 95% CI 0.72–0.96) and major depression (OR 0.77, 95% CI 0.67–0.89). Two early-onset clusters demonstrated distinct profiles despite similar symptom severity. The cluster characterized as early-onset PPD + pain (6% of training set [310/5239] and 7% of test set [178/2620]) had the highest prevalence of somatic conditions, including migraines (19% [58/310] training; 15% [26/178] test), nausea (47% [144/310] training; 53% [94/178] test), prenatal hospitalization (41% [128/310] training; 35% [62/178] test), and birth by caesarean section (10% [31/310] training; 8% [14/178] test), whereas the cluster characterized as early-onset PPD + anger (6% [307/5239] of training sample; 6% [158/2620] of test sample) showed many fewer physical health burdens. The clusters in the training and testing set had moderate to high concordance (ARI = 0.76 [95% CI 0.74, 0.77], FMI = 0.79 [95% CI 0.77, 0.81]).
Interpretation
This study identifies clinically relevant PPD subtypes with distinct genetic and obstetric profiles, highlighting the importance of trauma-informed care, pain management, and holistic obstetric approaches in PPD prevention and treatment. Limitations of this study include reliance on self-reported data, lack of diversity in the genetically homogeneous Norwegian sample, and genetic analyses subject to power constraints, which may limit generalizability and reduce precision of results. Despite these limitations, identification of reproducible subtypes with different risk factors and outcomes provides a framework for developing targeted interventions and advancing precision psychiatry in maternal mental health.
Postpartum depression (PPD) affects 18% of postpartum women globally. PPD is a heterogeneous condition with diverse presentations, which may differ in underlying etiology, potential outcomes, and optimal treatment. This study aimed to identify and validate PPD subtypes using comprehensive phenotypic and genetic data from a large, nationwide cohort.
Methods
In this population-based cohort study, we used data from the Norwegian Mother, Father, and Child Cohort Study and Medical Birth Registry of Norway (MoBa), and applied unsupervised clustering (Uniform Manifold Approximation and Projection [UMAP] + Density-Based Spatial Clustering of Applications with Noise [DBSCAN]) to identify PPD subtypes among 7859 women with PPD (training set n = 5239). All pregnant women in Norway were eligible to enroll in MoBa between Jun 1, 1999 and Dec 31, 2008. Among women with their first recorded pregnancies in MoBa, they were considered to have clinically significant PPD if they scored ≥8 on the 6-item version of the Edinburgh Postnatal Depression Scale (EPDS) (equivalent to ≥11 on the full EPDS) administered at six months after birth. Input variables included psychiatric symptom severity, psychiatric history, trauma history, pain, and substance use. We then characterized clusters and tested associations with perinatal-relevant auxiliary variables and polygenic scores (PGS) for eight psychiatric and neurodevelopmental conditions. Findings were validated in a reserved test set (n = 2620).
Findings
Nine clusters were identified in the training set (n = 5239), with seven replicating in the independent test set (n = 2620). Clusters were differentiated by symptom severity, onset timing, trauma history, and pain during pregnancy. In the training set, the cluster with the most severe symptoms was characterized by depression + trauma and comprised 28% [1483/5239] of training sample. This cluster showed marked socioeconomic adversity and elevated genetic risk for psychiatric conditions compared to other clusters (ADHD PGS OR 1.19, 95% CI 1.12–1.28; major depression PGS OR 1.15, 95% CI 1.07–1.23; PTSD PGS OR 1.09, 95% CI 1.02–1.17). In the test set, this cluster also comprised 28% [729/2620] and showed consistent associations with ADHD (OR 1.23, 95% CI 1.12–1.35), major depression (OR 1.09, 95% CI 0.99–1.20), and PTSD (OR 1.17, 95% CI 1.06–1.29). Conversely, the mild PPD cluster (10% [498/5239] of training sample) showed a protective profile (ADHD PGS OR 0.81, 95% CI 0.73–0.90; major depression PGS OR 0.85, 95% CI 0.77–0.94). In the test set, it represented 10% (263/2620), with consistent protective associations for ADHD (OR 0.84, 95% CI 0.72–0.96) and major depression (OR 0.77, 95% CI 0.67–0.89). Two early-onset clusters demonstrated distinct profiles despite similar symptom severity. The cluster characterized as early-onset PPD + pain (6% of training set [310/5239] and 7% of test set [178/2620]) had the highest prevalence of somatic conditions, including migraines (19% [58/310] training; 15% [26/178] test), nausea (47% [144/310] training; 53% [94/178] test), prenatal hospitalization (41% [128/310] training; 35% [62/178] test), and birth by caesarean section (10% [31/310] training; 8% [14/178] test), whereas the cluster characterized as early-onset PPD + anger (6% [307/5239] of training sample; 6% [158/2620] of test sample) showed many fewer physical health burdens. The clusters in the training and testing set had moderate to high concordance (ARI = 0.76 [95% CI 0.74, 0.77], FMI = 0.79 [95% CI 0.77, 0.81]).
Interpretation
This study identifies clinically relevant PPD subtypes with distinct genetic and obstetric profiles, highlighting the importance of trauma-informed care, pain management, and holistic obstetric approaches in PPD prevention and treatment. Limitations of this study include reliance on self-reported data, lack of diversity in the genetically homogeneous Norwegian sample, and genetic analyses subject to power constraints, which may limit generalizability and reduce precision of results. Despite these limitations, identification of reproducible subtypes with different risk factors and outcomes provides a framework for developing targeted interventions and advancing precision psychiatry in maternal mental health.
| Original language | English |
|---|---|
| Article number | 103540 |
| Number of pages | 18 |
| Journal | eClinicalMedicine |
| Volume | 89 |
| Early online date | 8 Oct 2025 |
| DOIs | |
| Publication status | Published - 1 Nov 2025 |
Bibliographical note
Publisher Copyright:© 2025 The Authors.
